Report No. 29825-PE Peru Opportunities for All Peru Poverty Assessment December 2005 Poverty Reduction and Economic Management Sector Unit Latin America and the Caribbean Region Document of the World Bank TABLE OF CONTENTS OPORTUNIDADESPARA TODOS: PRINCIPALESMENSAJESY RECOMENDACIONESEN MATERIA DEPOLITICAS ....................................................................................................................................... i NATURALEZA, DISTRTSUCIONY EVOLUCIONDELA POBREZA CRECIMIENTOY POBREZA: PERSPECTIVAGENERAL .......................................................................................................... iv ii ................................ ................. OPORTUNIDADESECONOMICAS PARA LOSPOBRESDE LAS AREAS URBANAS ............................................................. v OPORTUNIDADESECONOMICAS PARA LOS POBRES DELAS AREASRURALES ............................................................ ix ACCESOLOSSERVICIOSPUBLICOSY A LASINSTITUCIONESPUBLICAS.................................................................. A xi EXPOSICION AL RlESGOY MOVILIDAD SOCIAL ....................................... .................................................. xi11 ... OPPORTUNITIESFORALL: MAINMESSAGESAND POLICY RECOMMENDATIONS ......................... 1 THENATURE, DISTRIBUTIONAND EVOLUTIONOFPOVERTY..... ..................................................... 1 GROWTH AND POVERTY: AN OVERVIEW ............................ ................................................... 3 ECONOMIC OPPORTUNITIESFORTHE URBANPOOR .............. ................................................... 4 ECONOMIC OPPORTUNITIESFORTHE RURALPOOR .................................................................................................... 7 ACCESS PUBLICSERVICESAND INSTITUTIONS....................................... TO .......................................... EXPOSURERISKAND SOCIAL MOBILITY TO ............................................... ........................................................ 11 INTRODUCTION .................................................................................................................................................... 15 1. POVERTYANDINEQUALITY INPERU,1997-2004 ............................................................................... 17 POVERTY INEQUALITY AND UPDATE ...................................................................................................................... 17 Poverty and Inequality in Peru in 2004............................................................................................................ 17 POVERTY AND INEQUALITY TRENDS PERU, 1997-2004 IN ................................. ............................................... 19 Methodological changes in the ENAHO and implications for poverty and inequality comparisons................ 19 Evolution of Poverty and Inequality Trends, I997-2004 .................................................................................. 2I CONCLUSIONS ......................................................................................................................................................... 27 2. ECONOMICGROWTHAND ITSIMPACT ONPOVERTYANDINEQUALITY ............................... 29 ECONOMIC GROWTH TRENDS AND PATTERNS .............................................................................................. ECONOMIC GROWTH AND POVERTY ............................................................................................................. TheBasic Facts....................................................................................................................................... Distribution of economicgrowth. ............................................................................................................ 33 A CLOSERLOOK THENATURE AT OF ............................................................................................. ..35 Thespeed ofgrowth ................... ............................................................................................................ 35 Thecomposition of growth......... ....................................................................................................... 35 THE (WEAK) LINKS BETWEENGROWTH, EMPLOYMENT AND INVESTMENT ............................................................... 38 Making do with existing resources.................................................................................................................... 40 Animal Spirits Ajler All .............................................................................................................. 41 Policy implications....... ............................................................................................. LOOKINGAHEAD ..................... ........................................................................................................ 44 How Much Poverty Reducti mic GrowthAlone Buy?............................................................ Therolefor redistribution ..... ............................................................................................................ 47 CONCLUSIONS ..................... ........................................................................................ 3. NATURE,DISTRIBUTIONAND EVOLUTIONOFPOVERTY ............................................................. 51 POVERTYPROFILE.................................................................................................................................................. 52 Living conditions and characteristics of thepoor ............................................................................................. 52 Correlates of poverty.................................................................... ................................................ 55 Therole of household characteristics versus the role of geograp ................................................ 57 MICRO-DETERMINANTS OF POVERTY DYNAMICS .............................. ................................................ 59 Flows In and Out ofpoverty ........................................................ ................................................ 59 Determinantsof Changes in the Distribution of Income.............. ................................................ 61 ALTERNATIVE MEASURES OF POVER TY .................................................................................................................. 63 Unsatisfied Basic Needs................................................................ ................................................. 64 Caloric Deficit ............................................................................... ................................................. 65 POVERTY MEASURESTARGETINGINSTRUMENTSFORSOCIALPROGRAMS AND ..................................................... 66 Poverty Maps ........................................................................................... ................................................... 66 ................................................... 67 CONCLUSIONS................................................................................................ Proxy-Means Indicators........................................................................... ................................................... 68 4. ECONOMICOPPORTUNITIESFORTHEURBANPOOR .................................................................... 69 LABOR MARKETS URBANPOVERTY......................................................................................................... AND Employment,Labor Incomeand Poverty in Urban ............................................ 72 Recent Labor Market Trends....................................... Determinants of Participation in Formal and Informal Activities .................................................................... .................................................... 74 CONSTRAINTS EMPLOYMENT TO CREATIONINURBAN FTHEMANUFACTURING SECTOR 77 Barriers to Employment Creation and Business De ......................................................... 77 Labor Legislation and Employment.................................................................................................................. 81 Policy Recommendations............................................. .................................................. 85 INF~RMALECONOMIC OPPORTUNITIES ANDURBANPOVERTY .............................................................................. 87 Productivity of Se~-Employmentand Small Businesses.................................................................................... 88 Thecost and causesof info~ality.................................................................................................................... 94 Policy Recommendations.................................................................................................................................. 96 CONCLUSIONS ....................... ..................................... .................................................................................. 97 5. ECONOMICOPPORTUNITIESFORTHE RURALPOOR .................................................................... 99 RURALPOVERTYINPERU: A HETEROGENEOUS .................................................................................... REALITY 100 INCOME INRURAL SOURCESAND POVERTY AREAS.............................................................................................. 101 102 Agricultural and Non-Agricultural Income: Substitutesor Complements?.................................................... Income Distribution in Rural Areas: Patterns and Trends............................................................................. 103 Is Non-AgriculturalEmploymenta Door out of Poverty?............................................................................... 104 THE DETERMINANTS ACTIVITIES........................................................................................................................................................... OFPARTICIPATIONINANDPROFlTABILITY OFAGRICULTURAL AND NON-AGRICULTURAL 106 Income-Generating Strategies: Participation in Agricultural and Non-AgriculturalActivities ..................... 106 Returns to Agricultural and Non-Agricultural Activities: Therole of Access toAssets and Markets............. 108 THEIMPAC OF PUBLIC INTERVENTIONS ON POVER TY ......................................................................................... 115 PROMOTING INCLUSIVE RURALGROWTH ............................................................................................................. 119 CONCLUSIONS....................................................................................................................................................... 122 6. VULNERABILITYAND EXCLUSION ..................................................................................................... 123 SHOCKS AND COPINGSTRATEGIES ........................................................................................................ TheNature and Incidence of Shocks......... .............................................................................. Theimpact ofshocks ....................................................................................................................................... 125 Coping with Shocks: Strategiesand Egectiveness ............................................................................... 125 Policy Implications................................................ .............................................................................. 130 SOCIALMOBnITI ............................................................ .............................................................................. 135 EconomicMobility .......................................................................................................................................... 135 Education Mobility.......................................................................................................................................... 137 Whatdo these Results Imply? Supply-SideversusDemand-SideInterventions.............................................. 138 ACCESSTO PUBLICTRANSFERS, SERVICESAND INST~TIONS ............................................................................... 139 Access to public institutions ............................................................................................................................ Access to public services................................................................................................................................. 140 144 Policy implications.......................................................................................................................................... 145 CONCLUSIONS ....................................................................................................................................................... 148 BIBLIOGRAPHY .......................................... ......................................................................................................... 149 ANNEX 1:METHODOLOGICALISSUESREGARDINGPOVERTYMEASUREMENTINPERU, 1997- 2003.......................................................................................................................................................................... 157 157 A comparison between the ENAHO and the ENNIV ....... Changes in the ENAHO 1997-2003................................................................................................................ ................................. ANNEX 2: GROWTHINCIDENCECURVES-ROBUSTNESS CHECKS ................................................... 162 ANNEX 3: INTERNATIONAL COMPARISONS .............................................................................................. 165 ANNEX 4: PROGRESSTOWARDSTHE MILLENNIUM DEVELOPMENTGOALS ............................... 167 ANNEX 5: THE MININGSECTORAND POVERTYREDUCTION ............................................................. 168 STATISTICALANNEX ........................................................................................................................................ 171 CHAPTER1............................................................................................................................................................ 171 CHAPTER2............. ......................................................................................................... 173 CHAPTER4.................................................................................................................................. ............178 ACKNOWLEDGEMENTS The Peru Poverty Assessment was prepared by a team led by Carolina Sanchez-Paramo and including Marina Bassi, Javier Escobal, Norman Loayza, Leonard0 Lucchetti, Rossana Polastri and Walter Sosa. Valuable research assistance was provided by Magdalena Bendini, Pablo Lavado, Marco Ortiz and Guillermo Vuletin. Many thanks are due to: Ariel Fiszbein, Peter Lanjouw, Humberto Lopez, Edmundo Murrugarra and Pedro Olinto as peer reviewers. Lorena Alcazar, Lisa Bhansali, Leah Belsky, Juan Chacaltana, Daniel Cotlear, Jishnu Das, Franz Drees-Gross, Adolfo Figueroa, Vicente Fretes-Cibils, Javier Herrera, Aart Kray, Teresa Lamas, Jose Roberto Lopez-Calix, Gilbert0 Moncada, Eduardo Moron, John Newman, Jaime Saavedra, Renzo Rossini, CorneliaTesliuc and Gustavo Yamada for valuable inputs and comments. Farid Matuk, Nancy Hidalgo and the rest of the Household Survey Group in the ZnstitutoNacional de Estadisticas e Znforrnacidn (INEI) for granting the team access to the data used for the analysis and providing constant technical assistanceand support. Fernando Zavala and his team at the Ministry of Finance, members of the Peru Country Team and staff at the World Bank office inLima for support inWashington and Peru. Anne Pillay, Michael Geller and Christopher Humphrey for invaluable help in the preparation of the final document. OPORTUNIDADESPARA TODOS PRINCIPALES MENSAJES Y RECOMENDACIONES ENMATERIA DEPOL~TICAS 1. ElPer6enfrenta altosniveles depobreza y desigualdad. En2004, unpocomAs de lamitad de la poblaci6n del Perh era pobre y cerca del 20 por ciento vivia en la pobreza extrema. Aunque es dificil hacer comparaciones regionales debido a la utilizaci6n de diferentes lineas de pobreza en cada pais, 10s niveles de pobreza en el Perd son inferiores a 10s de Ecuador y Colombia per0 estAn por encima de 10s de Argentina y Brasil si consideramos una linea de pobreza de 2 d6lares por dia. Los niveles de pobreza son significativamente mAs altos en las Areas rurales, mientras las iireas urbanas -y en particular el kea metropolitanade Lima- presentanel mayor grado de desigualdad. L a desigualdad, medida por medio del coeficiente de Gini, era de 0,43, un nivel inferior a1 promedio latinoamericano, que es de 0,52, per0 todavia elevado segdn esthndares internacionales. 2. En 10s ultimos aiios 10s indices de pobreza han tardado en reaccionar frente a1impresionante crecimiento econ6mico del pais. Luego de mostrar una disminuci6n durante la dCcada de 1990, la pobreza aurnent6 tras la crisis econ6mica de 1998, mientras la pobreza extrema se mantuvo estable. La recuperaci6n econbmica iniciada en 2001 ha tenido un fuerte impacto positivo en la reducci6n de la pobreza extrema, per0 el avanceen materiade tasas de pobreza se ha limitado a las hreas rurales. 3. Uno de 10s prop6sitos fundamentales de este informe es explicar por quC el crecimiento no se ha traducido en una reducci6n mhs riipida del nivel de pobreza. Elprimer motivo es simplemente que la reducci6n de la pobreza toma tiempo, sobre todo teniendo en cuenta el patr6n de volatilidad econdmica que se ha presentado en el Penj durante las ultimas dCcadas, hecho que hace que 10s empresarios Sean renuentes a invertir en actividades generadoras de empleo. AdemAs, tras 10s aiios de alta inversi6n de la dCcada de 1990, a 10s que sigui6 la crisis de 1998, muchas empresas han tenido capacidad instalada ociosa que apenas ahora, despuks de cinco aiios de crecimiento, se estA resolviendo. Por lo tanto, el buen manejo macroecon6mico de 10s dltimos aiios est6 comenzando a dar frutos desde el punto de vista de la reducci6n de la pobreza y la generaci6n de empleo, y es probable que estos efectos se aceleren si el gobierno mantiene politicas macroecondmicas s6lidas. 4. Este informe tambiCn seiiala 10s diversos obsthculos que impiden que el crecimiento econ6mico lleve a una mayor reducci6n de la pobreza en el Per& y que deben enfrentarse reformando las politicas gubernamentales. El crecimiento se ha concentrado en las industrias dedicadas a la extracci6n de recursos naturales, que tienen un alto coeficiente de capital y generan poco empleo, y en 10s sectores agricola de las Areas rurales e informal de las urbanas, caracterizados por una baja productividad y, por lo tanto, un bajo crecimiento de 10s salarios. El fortalecimiento del vinculo entre el crecimiento y la existencia de un mayor numero de empleos mejor remunerados requiere intervenciones con politicas microecon6micas que eleven 10s incentivos para la generaci6n de empleo en el sector formal y diversifiquen la actividad econ6mica hacia keas con mayor intensidad de mano de obra, como son las exportaciones no tradicionales. Aunque esa diversificacidn ha comenzado -asi lo demuestra el rApido crecimiento de las exportaciones no tradicionales en 10s ultimos aiios- una reforma de las politicas puede contribuir a ese proceso. Estas reformas deben estar acompaiiadas de medidas para mejorar el capital humano y el acceso a 10s servicios publicos por 10s pobres, lo que haria que el acceso a las oportunidades econ6micas fuera mAs equitativo. i NATURALEZA, DISTRIBUCION Y EVOLUCION DELA POBREZA 5. En2004, unpoco r d s de la mitad de la poblaci6n del Pedvivia en lapobreza' mientras cerca de una quinta parte vivia en la pobreza extrema (vCase la tabla 1). Los niveles de pobreza son significativamente mAs altos y mAsprofundos en las keas rurales, en comparaci6n con las urbanas. Enlas Areas rurales, la pobreza alcanza el 72 por ciento y la pobreza extrema el 40 por ciento, mientras en las Areas urbanas esas tasas son del 40 por ciento y el 8 por ciento respectivamente. En la sierra y en la selva 10s niveles de pobreza tambiCn son mAs altos y mAs profundos en comparacidn con la costa.2 La mayoria de las variaciones en las tasas de pobreza entre las regiones se puede atribuir a las variaciones en las caracteristicas de 10s hogares y en el acceso a 10s servicios bAsicos y a la infraestructura de vias de comunicacibn, 6 s que a las diferencias geogrhficas como la altitud y la temperatura. En otras palabras, 10s hogares comparables tienen probabilidades similares de ser pobres, independientemente de las caracteristicas geogrAficas de su regi6n de residencia. L a desigualdad, medida por medio del coeficiente de Gini, es del 0,43 en el pais, per0 es inferior en las Areas rurales, en particular en la parte rural de la costa, y en Lima presentael nivel m i s alto. Tabla1.Indicadoresde lapobreza en 2004 Pobreza . Pobreza extrema Coeficiente de Tasa Brecha Severidad Tasa Brecha Severidad Gini Nacional 51,6 18,O 19,2 2,1 0,43 Area de residencia Urbana 40,3 12,4 7,9 0,7 0,39 Rural 72,s 28,3 40,3 4,8 0,32 Regi6n geografica Costa urbana 37,l 10,6 6 2 0s 0,34 Lima Metropolitana 36,6 10,4 3,4 02 0,40 Costa rural 533 16,4 14,6 0 s 0,32 Sierra 67,7 27,2 36,s 4,5 0,39 Selva 59,s 19,7 26,4 2 2 0,36 Fuente: ciilculosde 10s autores apaTtir de datosde laEncuestaNacionalde Hogares(ENAHO) de 2004, realizadapor el InstitutoNacionalde Estadisticae Informhtica(INEI).La muestraanualcubre el period0de enero adiciembrede 2004. 6. L a pobreza aument6 de manera significativa como resultado de la crisis econ6mica de 1998 y despuCs se ha mantenido estable alrededor del nuevo nivel a1 que lleg6 luego de la crisis, mientras la pobreza extrema no se vi0 afectada por la crisis y ha disminuido desde 2001. Est0 indica que aunque la pobreza ha tardado en reaccionar a1 crecimiento econ6mic0, tal reacci6n ha comenzado y ha tenido el mayor impact0 en 10s m i s pobres de 10s pobres. 7. Sin embargo, la evoluci6n de la tasa nacional de pobreza oculta diferencias importantes entre las keas urbanas y las rurales y de una regi6n a otra. Mientras la tasa de pobreza aument6 entre 1997 y ' Las tasas de pobreza de este informe se basan en el gasto en lugar del ingreso, porque el gasto suele permanecer mis estable durante el transcurso de un aiio y ademhs porque en las encuestas de hogares se tiende a reportar el gasto con m8s exactitud que el ingreso. Por lo general, a1 Per6 se le divide en tres regiones geogrhficas: la "sierra" (las montaiias), la "selva" y la "costa". 11 2000 en las Areas urbanas y en las rurales, entre 2001 y 2003 se mantuvo estable en las primeras per0 disminuy6 en las segundas (vCase la tabla 2). Delmismo modo, despuCs de 10s aumentos generalizados en todas las regiones debido a la crisis, la pobreza sigui6 aumentando ligeramente en Lima, mientras en la selva disminuy6 y en el resto del pais se mantuvo estable. Tabla 2. Porcentaje de la poblaci6nque vive por debajo de la linea de pobreza, 2001-04 Zona geogrhfica 2001 2002 2003 2004 Urbana 42,O 41,O 393 40,3 Rural 77,l 77,7 75,7 72,5 Total nacional 54,3 53,s 52,2 51,6 8. Encomparacion con lo anterior, la evolution de la pobreza extrema fue similar entre las Areas urbanas y las rurales, y entre una regi6n y otra durante ese periodo. L a pobreza extrema se mantuvo estable entre 1997 y 2000 y mejor6 a unritmo constanteentre 2001 y 2004 (vCase la tabla 3). Las mejoras en la pobreza extrema se concentraron en las Areas rurales y en 10s departamentos mAs pobres, entre ellos Ayacucho, Apun'mac, Cusco y Cajamarca. Durante este periodo tambiCn se presentaron mejoras significativas en la brecha de.pobreza y en la severidad, que se redujeron de 20,9 a 183 y de 10,7 a 8,9 respectivamente entre 2001y 2004. Tabla 3. Porcentaje de la poblaci6npor debajo de la linea de pobrezaextrema, 2001-04 Zona geogrhfica 2001 2002 2003 2004 Urbana 10,2 9,4 8 3 799 Rural 49,8 51,7 45,9 40,3 Total national 24,l 24,2 21,9 19,2 Fuente: chlculos de 10s autores a partir de datos de las ENAHO 2001-04 (INEI). Lascifrasde 2001y 2002correspondenalcuartotrimestre del aiio. Las cifras de 2003 correspondena mayo-diciembre. Las cifras de 2004correspondena enero-diciembre. 9. La pobreza, medida por el indice de necesidades bhsicas insatisfechas (NBI), ha disminuido a unritmoconstantedurante 10s filtimos diez afiosa medida que se expandia el acceso a la infraestructura y 10s servicios bisicos. L a parte de la poblacidn que tenia por lo menos una NBIha bajado de 57 por ciento en 1993 a 40 por ciento en 2003. El avance ha sido importante tanto en las Areas urbanas como en las rurales, ya que durante este periodo el indice de NBIha bajado de 42 por ciento a 25 por ciento en las Areas urbanasy de 90 por ciento a 71por ciento en las rurales. Sin embargo, la mayoria de la disminuci6n se dio durante la dCcada de 1990, antes de que 10s niveles de inversi6n pfiblica se recortaran significativamente, hecho que fren6 el ritmo de avance en el acceso a 10s servicios bisicos. 10. La divergencia entre la evolucidn de las medidas monetarias de la pobreza y las no monetarias, que son mAs estructurales, durante un periodo de significativa perturbacibnecon6mica no es algo que se presenteexclusivamente en el Peh. El Ecuador tuvo una experiencia similar durante la crisis de 1998-99 e inmediatamente despds de la dolarizaci6n de 2000, cuando la pobreza monetaria aument6 de manera considerable como resultado de la desfavorable situaci6n econbmica, per0 el indice de NBI continuo descendiendo, siguiendo su tendencia de largoplazo. ... 111 CRECIMIENTOY POBREZA: PERSPECTIVAGENERAL 11. L a relaci6n entre crecimiento econ6mico y pobreza es m i s dCbil en el Pertique en otros paises del mundo. Es decir, el Perti necesita crecer a un ritmo mas acelerado que otros paises para reducir la pobreza o incluso para evitar que aumente. En 10s 6ltimos aiios esta dCbil relaci6n entre crecimiento y pobreza ha sido una consecuencia directa de la naturaleza del crecimiento econ6mico en el P e ~que no , ha tenido una base suficientemente amplia debido a cuatro factores. En primer lugar, el crecimiento per capita fue relativamente lento entre 1997 y 2004. Es decir, aunque la econom'a crecio, no lo hizo a1 mismo ritmo en que aument6 la poblaci6n. Durante ese periodo el ingreso y el consumo per cipita se redujeronen -1,6 por ciento y -14 por cientorespectivamente, seg6n10s datos de la ENAH0.3 12. En segundo lugar, el crecimiento estaba sesgado hacia sectores con un alto coeficiente de capital y una baja demanda de mano de obra -om0 la mineria-, o hacia sectores con baja productividad laboral y niveles salariales bajos -om0 la agricultura-, aunque recientemente 10s niveles de empleo tambikn han estado creciendo en otros sectores, como 10s textiles y las exportaciones no tradicionales. Los sectores mineria y agricultura crecieron a tasas anuales de 7,6 por ciento y 3 3 por ciento entre 1997 y 2004. Durante ese periodo el empleo en el sector minero creci6 3,6 por ciento, mientras el empleo en el sector agricola creci6 4,6 por ciento. Del mismo modo, 10s salarios crecieron a una tasa anual de 12,4 por ciento en el sector minero, mientras en el sector ap'cola disminuyeron-1,s por ciento. Sin embargo, no es probable que el crecimiento de 10s salarios en el sector minero haya beneficiado a 10s pobres, dada la composicidn de la fuerza de trabajo empleada en el sector, que tiende a ser relativamente calificada. 13. Entercer lugar, histbricamente el crecimiento ha sido muy volfitil. Desde 1965 hastael periodo 2001-05, el Per6 nunca habia disfrutado de d s de cuatro aiios consecutivos con tasas de crecimiento superiores a13 3 por ciento y, en cambio, habia tenido trece aiios con tasas de crecimiento inferiores a12 por ciento y siete aiios con tasas de crecimiento negativas. A su vez, la volatilidad se hatraducido en altos niveles de incertidumbre entre 10s empleadores y otros agentes econbmicos, factor que ha debilitado 10s incentivos para invertir y contratar a nuevos trabajadores. Este legado de crecimiento volfitil se esta superando de manera gradual mediante un manejo macroecon6mico prudente y politicas de disciplina fiscal del gobiemo. 14. En cuarto lugar, la inversi6n y la creaci6n de empleo en el Per6 han sido bajas porque existe una significativa capacidad instalada ociosa como resultado del deterioro econ6mico de 1998-99. L a utilizaci6n de la capacidad instalada ha aumentado desde 2000, per0 todavia una proporci6n considerable permanece subutilizada. Aproximadamente el 30 por ciento de 10s empleadores declar6 que estaba utilizando el 60 por ciento o menos de la capacidad instalada de su empresa en 2003, lo que representa una reducci6n respecto del nivelde 2000, que era del 37 por ciento; mientras un38 por ciento declar6 que estaba utilizando miis del 80 por ciento de su capacidad, lo que representaunaumentorespecto a1nivel de 2000, que era del 19 por ciento. Los tiltimos datos disponibles indican que el promedio de utilizacibn de la capacidad instalada era de 74 por ciento en octubre de 2005. Del mismo modo, 10s empleadores se mostraban cautelosos en lo que se refiere a la contratacidn de nuevos empleados. En 2003, el 15 por ciento de 10s empleadores declar6 que estaba dispuesto a contratar a nuevos empleados y mas del 20 por ciento que estaba dispuesto a despedir a empleados actuales. En2000, esas cifras eran del 10por ciento y el 30 por cientorespectivamente. Est0contrasta con 10s datos de las cuentas nacionales, que mueshan que el product0 bruto interno (PBI) per &pita creci6 2,5 por ciento durante el mismo periodo. Sin embargo, la pobreza se mide utilizando la ENAHO. iv 15. Como resultado de 10s ultimos tres factores mencionados -la volatilidad econbmica, el crecimiento en 10s sectores con un alto coeficiente de capital y la capacidad instalada ociosa- las tasas de generaci6n de empleo aunque positivas han sido insuficientes, y de inversi6n se hanmantenido bajas, y debido a eso el crecimiento no ha beneficiado a 10s pobres en las Areas urbanas. Aunque el empleo en empresas con d s de diez trabajadores se ha recuperado desde 2000, todavia permanece en niveles inferiores a 10s existentes antes de la crisis. Es d s : la tasa de ocupaci6n se ha mantenido constante, e incluso se ha reducido ligeramente desde 2000, a la par que la composici6n del empleo se ha inclinado hacia una mayor infonnalidad. El n6mero de horas trabajadas tambiCn ha aumentado ligeramente, lo que indica que las necesidadesde mano de obra se pueden haber atendido por medio del us0 m6s intensivo de 10s actuales trabajadores y no mediante nuevas contrataciones. Del mismo modo, aunque 10s niveles de inversi6n han aumentado en dnninos reales desde 2003, las tasas de inversidn han disminuido a un ritmo constante como porcentaje del PBIdesde 1998. 16. El an6lisis precedente indica que las politicas macroeconomicas orientadas a mantener y fortalecer el crecimiento econ6mico y a reducir la incertidumbre pueden garantizar que el crecimiento econ6mico no solo sea sostenible sino que contribuya a la reducci6n de la pobreza. L a mayoria de esas politicas macroecon6micas ya se aplican y deben continuar. Sin embargo, para fortalecer a6n m6s el vinculo entre crecimiento econ6mico y reducci6n de la pobreza se deben complementar con reformas microecon6micas que ofrezcan mayores incentivos para invertir y contratar a nuevos trabajadores tanto en las heas urbanas como en las rurales, y para elevar el nivel del capital humano en la fuerza de trabajo y hacer que las oportunidades econdmicas Seanmis igualitarias, aspectos que se comentan a continuaci6n. OPORTUNIDADES ECONOMICAS PARA LOSPOBRES DELAS AREAS URBANAS 17. Las mejoras recientes en 10s niveles promedio de empleo y salarios en las Areas urbanas no se han traducido en menores tasas de pobreza urbana. Ello se debe a que tales mejoras se han concentrado entre 10s trabajadores formales y d s educados, empleados en compaiiias m6s grandes, y no se han extendido a1sector informal, en el que est6 empleada la mayoria de 10s pobres. L a ocupaci6n (la cantidad total de trabajadores) aument6 en un promedio anual de 4 por ciento entre 2000 y 2004. Sin embargo, como la tasa de participaci6n en la fuerza de trabajo tambiCn aument6, la tasa de ocupaci6n se mantuvo constante o incluso disminuy6 ligeramente durante ese periodo. Entre 1997 y 2004, 10s salarios de 10s respectivamente (en unindice de salarios con una base de loo), mientras 10s de 10s obreros pasaron de 86 gerentes y empleados de oficina del sector formal aumentaron de 117 a 158 y de 94 a 104 a 89. 18. La existencia de dCbiles conexiones entre las empresas grandes y las pequeiias, que ha impedido que el crecimiento de las primeras se filtre hacia las segundas, ha agravado la ausencia de avances en el empleo y en 10s salarios de 10s pobres. En2004, el 62 por ciento de las empresas grandes le vendia como mhimo el 20 por ciento de su producci6na las microempresas y pequeiias empresasy el 46 por ciento le compraba como m6ximo el 20 por ciento de 10s insumos a las microempresas y pequeiias empresas, mientras s610 el 1,2 por ciento de las empresas grandes le vendia m6s del 80 por ciento de su producci6n a las microempresas y pequeiias empresas y el 4,1 por ciento le compraba m6s del 80 por ciento de 10s insumos a las microempresas y pequeiias empresas. 19. Las futuras disminuciones en la pobreza urbana depender6n de la capacidad de la econom'a urbana para generar empleos m6s productivos y mejor remunerados, sobre todo en 10s sectores que emplean a 10s pobres. Elresto de esta secci6n analiza las limitaciones actuales para la creaci6n de empleo urbano por las empresas grandes, medianas y pequeiias, y recomienda politicas para enfrentarlas. Presta especial atenci6n a las rigideces del mercado laboral y a1papel de la legislaci6n laboral. Enla medida en que la mayoria de 10s pobres de las heas urbanas est6 empleada en empresas pequeiias e informales, V tambi6n estudia 10s incentivos para que las empresas informales se formalicen y 10s factores detenninantes de la productividaden las actividades informales, y hace recomendacionesa1respecto. La eliminaci6nde laslimitacionesa lageneraci6n de empleo urbano 20. El25 por ciento de las empresas manufactureras que participaronen el Estudio sobre el clirna de inversidn en el Perzi (Banco Mundial 2003) declar6 que, si no enfrentara limitaciones, le gustaria aumentar el ndmero de trabajadores que emplea de manera permanente, mientras menos del 5 por ciento dijo que lo disminuiriay el 70 por ciento sostuvo que lo mantendria en el nivel actual. Los motivos de las diferencias entre la cantidad actual y la deseada de empleados contratados y despedidos varian un tanto segdn el tamaiio de la empresa. Sin embargo, queda claro que 10s costos laborales no salariales y 10s costos de 10s despidos establecidos por la ley, y en menor medida la incertidumbre sobre la demanda futura de 10s productos de la empresa, constituyen 10s motivos fundamentales de esa diferencia. 21. Enteoria, las condiciones del empleo permanenteen el Perd son buenas y existe unalto grado de protecci6n en comparaci6n con el de otros paises de la regidn y del mundo. Por ejemplo, en 1999 se calculaba que el costo promedio de un despido equivalia a 13,8 salarios mensuales promedio, una cifra inferior a la de 1987, que era de 15,0, p r o bastante m6s alta que el promedio regional de 5 5 o que el promedio de 1,5 existente en 10s paises industrializados de habla inglesa. Del mismo modo, la ley protege las relaciones laborales y, en menor medida, el acceso a la seguridad social. Sin embargo, es relativamente f6cil eludir estas disposiciones contratando a 10s nuevos trabajadores con contratos temporales en lugarde permanenteso funcionando en el sector informal. Por ejemplo, el empleo temporal y por horasrepresentael 20 por ciento de todo el empleo asalariadoen el sector privado y el 50 por ciento de todo el empleo por contrato enLima Metropolitana. 22. Como resultado de lo anterior, 10s mercados laborales en el P e d son flexibles defacto, aunque esa flexibilidad conlleva el costo de una proteccibn laboral y un acceso a la seguridad social bajos y distribuidos de manera poco uniforme. Solo el 18 por ciento del empleo urbano y el 52 por ciento del empleo asalariado est6 contratado en t6nninos que cumplen plenamente con la legislaci6n laboral peruana, mientras en Amkrica Latina y el Caribe esos porcentajes son del 40 por ciento y el 60 por ciento respectivamente. Del mismo modo, el incumplimiento de las normas sobre salario m'nimo equivale a un poco menos de una cuarta parte de la poblaci6n pertinente del Pert?, mientras en Am6rica Latina y el Caribe ese porcentaje es del 10por ciento. Esta dicotom'a entre 10s altos niveles de protecci6nde iure y 10s bajos niveles de protecci6n defacto es comdn a otros paises de la regi6n en 10s que 10s esfuerzos por hacerque la ley se cumpla son d6biles y la incidenciade la informalidad es elevada. 23. Elimpact0de la legislaci6nlaboral se extiende mAs all6 delempleoformal permanentey afecta 10s niveles totales de empleo y la composici6n del empleo, como l o ha documentado de manera exhaustiva la bibliografia sobre 10s mercados laborales del Pert?y de otros paises. Esos patrones se deben en gran medida a las diferencias de 10s costos salariales y no salariales entre el empleo permanente y el temporal (e informal). Las siguientes son dos medidas que reducirian esas diferencias sin aumentar el costo relativo del empleo formal: 0 Reducir 10s costos de 10s despidos con el propdsito de aumentar 10s incentivos para contratar. Los costos de 10s despidos se podrian reducir de varias formas. El enfoque m6s drastic0 requiere una reducci6n de las indemnizaciones por despido. Esta medida se podria aplicar solamente a 10s nuevos contratos, o bien se podria considerar la posibilidad de pagarles compensaci6n a 10s trabajadores contratados de conformidad con el r6gimen anterior. Entre 10s enfoques menos dr6sticos podria estar vi un aumento de 10s periodos de prueba para 10s nuevos trabajadores y un us0 m6s flexible de 10s "motivos econ6micos" como causade 10s despidos. 0 Reducir 10s costos no salariales. El Peni estd entre 10s paises de la regi6n que tienen la legislaci6n d s generosa en lo que respecta a vacaciones pagadas, junto con Brasil y Panami. Esto no s610 es muy costoso para 10s empleadores (especialmente teniendo en cuenta la baja productividad laboral), sino que en realidad s610 un porcentaje pequeiio de 10s trabajadores disfruta del period0 de vacaciones, hecho que indica que a 10s actuales niveles de ingreso 10s trabajadores estin dispuestos a vender tiempo libre a cambio de un ingreso adicional. Un enfoque mis flexible podria ser que las vacaciones estuvieran en funci6n de la experiencia del trabajador o del tiempo que ha ejercido su cargo y, por lo tanto, de su productividad laboral. Proporcionar incentivos para la formalizacidn 24. El alto nivel de empleo informal del Pen3 es costoso para 10s hogares, las empresas y el gobierno. L a baja productividad entre las empresas informales se traduce en ingresos mis bajos para quienes estin empleados en el sector: el ingreso laboral promedio por hora en el sector informal es inferior en un 50 por ciento a1del sector formal, inclusive cuando se comparan trabajadores similares en empleos similares. Los trabajadores del sector informal tambiCn carecen de acceso a la protecci6n social vinculada a1 empleo: por ejemplo, a prestaciones de salud y a las pensiones o a las indemnizaciones por despido. Aunque algunos de esos trabajadores pueden haber renunciado voluntariamente a esas prestaciones a cambio de salarios d s elevados o de mayor flexibilidad, m6s del 50 por ciento de 10s pobres que trabajan en el sector informal lo hacen de manera involuntaria, seg6n 10s estudios. Las empresas informales tambiCn enfrentan diversos obsticulos costosos, entre ellos el acceso limitado a1 crkdito, a las asociaciones empresariales y a 10s programas del gobierno que promueven las actividades econ6micas y las exportaciones. Desde el punto de vista del gobierno, el incumplimiento en el pago del impuesto a las ventas y de 10s impuestos laborales entre las empresas informales tiene un impacto negativo en 10s ingresos tributarios. 25. Los altos niveles de informalidad y 10s bajos niveles de generaci6n de empleo permanente se pueden atribuir a 10s lentos y costosos procedimientos de registro de las empresas, a 10s complejos procedimientos para que estas presenten sus declaraciones tributarias y a las inflexibles nonnas laborales (que resultan particularmente onerosas para las pequeiias empresas). Ya comentamos las politicas para reducir 10s costos del empleo (permanente). Entre las otras intervenciones dirigidas a proporcionar incentivos para la formalizacibn de las pequeiias empresas y aumentar su acceso a1crkdito podria estar la racionalizacibnde: 0 Los procedimientos de registro de Ias pequefias empresas. Una reducci6n de 10s trirnites de tal manera que 10s costos de 10s procedimientos de registro bajen a un nivel equiparable a1 de sus cornpetidores d s cercanos de dentro y fuera de la regi6n haria d s ficil que las empresas cumplieran con ellos. L a legislaci6n recientemente aprobada para implementar un sistema especial simplificado de registro para las microempresas y las pequeiias empresas, y para permitirles que hagan aportes 6 s reducidos a 10s sistemas de pensiones y de salud, constituye un paso en esa direcci6n. Desafortunadamente, la puesta en marcha de estas y otras reformas contempladas en la nueva ley no ha tenido hasta ahora 10s efectos deseados, pues solo 3.500 empresas informales se han registrado fomlmente. Serfa importante comprender 10s motivos del lirnitado impacto de esas medidas. vii Los mecanismos para presentar declaraciones tributarias. Ya existe un regimen especial y simplificado para que las microempresas y las pequeiias empresas presenten sus declaraciones, per0 estas y otras empresas se podrian beneficiar con una mayor simplificaci6n. Por ejemplo, se podria considerar la posibilidad de establecer la presentacidn basada en caractensticas fhcilmente observables de las empresas y de acuerdo con tablas tributarias predeterminadas. Tales sistemas facilitan la presentacih de declaracionesa las empresas que no se apoyanen sistemasde contabilidad totalmente formales, que con frecuencia son costosos, e incluso a empresas que interactuan con un gran nlimero de socios informales. Elevar la productividadde la pequeiia empresa informal 26. Aunque la promoci6n de la formalidad en el sector privado debe ser una prioridad, el hecho de que el sector informal es muy grande en el Perit tambiCn constituye una realidad. Aproximadamente la mitad de 10s pobres urbanos que trabajan lo hacen por cuenta propia, todos ellos en el sector informal, y un30por ciento mhstrabaja paramicroempresasopequeiias empresas, muchasde las cuales tambien son informales. Del mismo modo, el 40 por ciento de todos 10s empresarios informales (empleados por cuenta propia o no) son pobres, mientras entre 10s empresarios formales esa proporci6nes del 15 por ciento. Por consiguiente, identificar 10s factores detenninantes de la productividad de las actividades informales e implementar politicas orientadas a elevarla es fundamental para ayudar a 10s pobres urbanos a superar la pobreza. 27. Existe una variaci6n significativa en la productividadde las pequeiias empresas informales, en las que la productividad se mide en funcih del valor agregado por trabajador. Esta variaci6n se puede atribuir a las diferencias en las caracteristicas de 10s empresarios, 10s trabajadoresy las empresas. Debido a esto, 10s menores niveles de productividad entre 10s empresarios pobres, y por consiguiente 10s mhs bajos salarios que reciben sus trabajadores, se pueden atribuir a 10s mhs bajos niveles educativos tanto de 10s empleadores como de 10s empleados, a 10s menores niveles de integraci6n a 10s mercados y a1menor acceso a la infraestructura bisica. 28. Las diferencias entre 10s empresarios pobres y 10s no pobres en lo que respecta a sus pricticas comerciales y las caracteristicas de sus empresas no son el product0 de factores independientes sino interrelacionados. Por ejemplo, el us0de prhcticas de mercado y el acceso a1capital y a la infraestructura guardan correlaci6n con la ubicaci6n de la empresa. Es mis probable que las empresas que operan en locales comerciales utilicen algdn tip0 de contabilidad y que empleen una proporci6n mayor de trabajadores remunerados que las que funcionan en las calles o en las casas de sus propietarios. AdemAs, el acceso a maquinarias y a otras herramientas es mayor entre las empresas que funcionan en locales comerciales o no comerciales que entre aquellas que operan en las calles, mientras el us0de unvehiculo es mucho m6s frecuente entre estas liltimas, en parte debido a que lo utilizan como sustituto de un local propiamente dicho. Por bltimo, el manejo de unnegocio desde unlocal comercial guarda correlaci6n con unmayoracceso a10s servicios de telefonia y aguapotable. 29. Entonces, se podrian alcanzar niveles d s altos de productividad entre 10s trabajadores informalespor cuenta propiay las pequeiiasempresas informales con estas medidas: Mejorar el nivel de capacitaci6n tanto de 10s empresarios como de 10s trabajadores asalariados. El aumento general del nivel de capacitacibn de la fuerza laboral se puede lograr invirtiendo en la educaci6n formal (que se comenta rnhs adelante, en el phrrafo 40) y mejorando la pertinencia y la cobertura del sistema de capacitaci6n. No es necesario que el sector publico sea el proveedor de la capacitaci6n. En lugar de eso el Estado puede ofrecer incentivos para que las empresas contraten la ... Vlll capacitaci6n que desean con proveedores privados debidamente acreditados. El programa ProJoven, que proporciona capacitaci6n a trabajadores jovenes, se podria extender para cubrir otros grupos demogrAficos. Aumentar el acceso a 10s locales comerciales y el us0 de prhcticas de mercado. A 10s pequeiios empresarios que operan en las calles se les podrian ofrecer espacios comerciales en mercados u otros emplazamientos a cambio de un alquiler. Ese alquiler podria aumentar con el paso del tiempo para facilitar la inversi6n en las primeras etapas y tambiCn para reflejar 10s posibles aumentos futuros de productividad. El mayor acceso a estos espacios serviria como plataforma para el suministro econdmico de infraestructura bAsica y servicios comerciales como pricticas de gerencia y de contabilidad, acceso simplificado a1 crCdito y servicios legales, l o que a su vez se traduciria en una d s alta productividad. TambiCn ayudaria a descongestionar las calles y las Areas en las que funcionan esos negocios mejorandoel trhfico y disminuyendo 10s peligros. OPORTUNIDADES ECONOMICAS PARA LOSPOBRES DELAS AREAS RURALES 30. Los hogares de las Areas rurales obtienen la mayoria de sus ingresos a partir de actividades agricolas, per0 existen diferencias importantes entre 10s hogares pobres y 10s no pobres en lo que se refiere a sus estrategias para generar ingresos. Los hogares pobres tienden a depender exclusivamente de la agricultura, mientras 10s no pobres tienden a participar tambiCn en actividades no agricolas. Es d s : existe una mayor probabilidad de que 10s hogares pobres dependan de una sola fuente de ingresos, mientras 10s no pobres tienen una mayor capacidad para diversificar el riesgo relacionado con la generacibn de ingresos a1no depender exclusivamente de una fuente determinada. Alrededor de la mitad de 10s hogares rurales obtiene la totalidad de sus ingresos del trabajo por cuenta propia en el sector agricola, mientras el resto combina la agricultura con otros tipos de trabajo. Las tasas de pobreza son significativamente mAs altas entre aquellos empleados en el sector ap'cola (80 por ciento) que entre 10s que trabajan en sectores diferentes a1agricola (60 por ciento). 31. L a mayor parte de las variaciones en 10s ingresos de 10s hogaresde las keas rurales se debe a la variaci6n de 10s ingresos no agricolas procedentes del trabajo asalariado. Es d s : la participacibn del ingreso agricola disminuye a medida que aumenta el ingreso total. Aunque estos hechos simplificados parecen indicar que el empleo no ap'cola proporciona una via para salir de la pobreza, en realidad la mayoria de 10s hogares de las Areas rurales tiende a obtener ingresos tanto del sector agricola como de 10s sectores no agricolas: es decir que dependen de estrategias generadoras de ingreso en lugar de depender de sectores o actividades determinados. L a capacidad de un hogar de implementar una estrategia generadora de ingreso rentable determina su estatusen ttrminos de pobreza. 32. L a participaci6n en estas estrategias generadoras de ingreso estA en funci6n de las caracten'sticas y atributos de 10s hogares. Los mejores atributos de 10s hogares (como un nivel de educacibn superior) y el acceso a la infraestructura y a 10s servicios pfiblicos les permiten usar estrategias que incluyan actividades no agricolas, mientras la propiedad de activos agricolas y la falta de liquidez hacenque sea d s car0 para 10s hogares abandonar las estrategiasque incluyen actividades ap'colas. 33. En10s hogares, tanto laproductividadagricola como el ingreso laboralguardan una correlaci6n positiva con el capital humano (como un nivel de educaci6n superior) y el acceso a1 crkdito, a 10s servicios bhsicos, las telecomunicaciones y la infraestructura de vias de comunicaci6n. En el Ambit0 regional, el rendimiento de las actividades agricolas y asalariadas depende de la profundidad y el dinamismo de 10s mercadosregionales y de 10s niveles totales de productividad. L adensidad de poblacibn y el acceso a la infraestructura son mayores en la costa que en otras Areas, a pesar de las importantes mejoras alcanzadas durante la dCcada de 1990 en la sierra y en la selva. Ambos factores podrian contribuir a crear mercadosmAs integrados y dinhmicos en las Areas rurales, y a conectar mejor a las Areas ix urbanas y rurales. Por consiguiente, para superar las diferencias regionales, se deben realizar inversiones orientadas a mejorar el retorno a 10s activos, servicios y mercados en esas Areas que estAnrezagadas. 34. A1evaluar las diferentes alternativas de politicas para el sector rural es preciso considerar tres aspectos importantes. En primer lugar, la naturaleza de la pobreza rural es heterogknea y varia significativamente de una regi6n a otra, de modo que Ias intervenciones y 10s proyectos tienen que tener en cuenta las particularidades locales para garantizar una eficacia mixima. En segundo lugar, debido a que la tierra es escasa en relaci6n con la poblacibn que debe mantener y a1hecho de que la productividad agricola es baja, muchas de las personas que estAn empleadas en la actualidad en actividades agn'colas tendrian que mejorar de manera espectacular su productividad o abandonar la agricultura para salir de la pobreza. Est0 implica que la estrategia de desarrollo rural para el Pen3 debe ser multisectorial y tener presente la interacci6n entre las actividades agn'colas y las no agricolas. En tercer lugar, es importante advertir que en las zonas rurales ya existen varios programas que apoyan intervenciones en las Areas que identificamos antes y que proporcionan una estructura por medio de la cual el gobierno puede esforzarse por alcanzar la meta de tener un crecimiento rural inclusivo. Sin embargo, estos programas adolecen de diversos problemas que deben ser resueltos si se quiere que las futuras intervenciones Sean eficaces. 35 * Teniendo en cuenta las anteriores consideraciones, y si se quiere que 10s pobres de las Areas rurales se beneficiende las oportunidades econ6micas que genera el crecimiento econ6mico general, tres Areas fundamentales requieren la acci6n del gobierno: 0 Integrar a las Areas rurales a 10s mercados nacionales para aumentar las oportunidades econ6micas. Las acciones d s obvias para facilitar 10s contactos entre agentes y el transporte de mercancias entre las Areas rurales y urbanas son simplemente el mejoramientode la red de carreteras, y en particular de 10s sistemas secundarios y terciarios, para pennitir a 10s productores llevar sus productos a 10s mercados de una manera rApiday econ6mica, y la inversi6n en telecomunicaciones en las heas rurales, para pennitir que 10s residentes rurales tengan un acceso oportuno a la informaci6n pertinente sobre 10s mercados. El sector pfiblico tambiCn puede tomar medidas para facilitar la transmisi6n de conocimientos y tecnologia de las Areas urbanas a las rurales y para desarrollar relaciones econ6micas estables que garanticen una demanda constante de productos agn'colas y no agn'colas para el procesamiento industrial ylo la exportaci6n y crear incentivos para la produccibn a granel. e Mejorar el acceso a1crhdito entre 10s productores rurales. El crkdito rural estA restringido por las dificultadesde muchos productores, especialmenteaquellos que tienen fundos pequeiios, paracumplir 10s requisitos administrativos y las garantias que exigen las instituciones financieras. Por ello, la mayoria del crCdito existente es informal o lo suministran pequeiias cooperativas de ahorro y crCdito. Esas cooperativas, asi como otras instituciones con prop6sitos similares, como 10s grupos de crCdito para mujeres, deben fortalecerse. Las disposiciones sobre crCdito deben modificarse para pennitir el us0 como garantia de 10s activos familiares, como la maquinaria o el ganado, y tomar precauciones para no aumentar el riesgo del crkdito y el incumplimiento en 10s pagos, complementando el mayor acceso con un mejor seguimiento. A1 mismo tiempo deben continuar 10s esfuerzos por aumentar la titulaci6n de tierras. TambiCn es necesario tener en cuenta la alta preponderancia de la propiedad comunitaria de la tierra entre las poblaciones indigenas y el impact0 negativo que eso puede tener en la capacidadde las personasque viven en esas comunidades para acceder a1crCdito. 0 Mejorar 10s niveles del capital humano en las Areas rurales. La mejora de 10s niveles y estindares educativos en las Areas rurales se puede lograr con una serie de intervenciones que incluye: (i) la expansi6n de la educaci6n bilingiie mediante del suministro de materiales de enseiianzay aprendizaje adecuados y el reclutamiento y capacitacibn de maestros que hablen el quechua; (ii) la expansi6n de la educaci6n secundariapor medio de la educaci6n formal o la educaci6n a distancia; (iii) la creaci6n de incentivos para asistir a la escuela mediante programas de transferencia condicionada de dinero o X mejoras en 10s programas de alimentaci6n y nutricion que se ofrecen en las escuelas (en el pArrafo 38 se proporcionan mAs detalles sobre el tema de la educacih). L a asistencia tCcnica tambiCn se puede mejorar. El proyecto Innovaci6n y Competitividad para el Agro Peruano (INCAGRO), el Proyecto de Reduccidn y Alivio a la Pobreza (PRA) y el Fondo de Cooperaci6n para el Desarrollo Social (FONCODES) ofrecen algunos servicios de extensi6n piiblica, per0 una gran cantidad de pequeiios cultivadores y de residentes pobres de Areas rurales todavia est6 excluida debido a su alto costo. Por eso son necesarios mayores esfuerzos que apoyen el suministro de asistencia t6cnica basada en la demanda y acompaiiada por asistencia sobre comercializaci6n y gerencia. ACCESOA LOSSERVICIOSPUBLICOSY A LAS INSTITUCIONESP-LICAS 36. El acceso a 10s servicios piiblicos es un asunto relevante en diferentes campos, esencial para ayudar a 10s pobres a desarrollar capital humano y tambiCn para proteger a 10s vulnerables, tanto en las Areas rurales como en las urbanas. TambiCn es un Ambito directamente susceptible a la selecci6n de politicas, y por consiguiente es un blanco probable de mejoras que contribuyan a fortalecer 10s vinculos entre el crecimiento econ6mico y la reducci6n de la pobreza. El acceso a 10s servicios piiblicos como la asistencia a la salud, la educacibn y la protecci6n social es bajo entre 10s pobres, entre 10s grupos indigenas y en las Areas rurales. Los pobres tambiCn tienen menos probabilidades que 10s no pobres de entrar en contact0 con diferentes instituciones piiblicas, que van desde las oficinas de 10s gobiernos central y local hasta 10s bancos estatales y el sistemajudicial. 37. Este informe no ofrece un anAlisis profundo de 10s sectores de educaci6n, salud y protecci6n social (estos aspectos se comentan en Peru: Accountability in the Social Sectors, 2005), sino que toma nota de las principales cuestiones en esos campos, y especificamente de su relaci6n con el mejoramiento del capital humanode 10s trabajadores que buscan superar la pobreza. 38. Elsector educaci6npresentadiversasdebilidades criticas quereducensu impact0 en 10s pobres, tanto en las Areas urbanas como en las rurales. Muchos pobres consideran que no vale la pena asumir el costo de oportunidad que representa la educacibn, loque disminuye la demanda de esta. Envista de que la calidad de la educaci6n es baja y las oportunidades disponibles a1egresar de las escuelas son limitadas, muchas familias prefieren hacer que sus hijos trabajen y recibir un ingreso adicional, incluso si es pequeiio, en lugar de que asistan a la escuela. Esto sucede sobre todo en las Areas rurales, que no tienen suficientes buenos maestros pues estos prefieren trabajar en las Areas urbanas. Ademis, el ausentismo de 10s maestros es rnuy elevado en las Areas rurales. L a asistencia a las escuelas indigenas es especialmente baja, en buena parte porque no existen suficientes planes de estudios bilingiies o biculturales para la poblaci6n indigena. Otro problema, que tambiCn es m& pronunciado en las Areas rurales, es la oferta limitada de cupos de preescolar y secundaria, que tienen bajas tasas de matricula. 39. Elevar la calidad y la cobertura de la educaci6n requiere politicas de incentivos a la demanda y a la oferta, como: e Promover una mayor demanda de educacibn. Es posible inducir aumentos de la demanda de educaci6n reduciendo de hecho sus costos (tanto de 10s directos como de 10s de oportunidad) por medio de programas de transferencia condicionada de dinero (PTC) o becas, y mediante la implementaci6n de horarios de estudios flexibles que permitan a 10s niiios y a 10s j6venes participar en otras actividades durante el dia. El Perii lanz6 recientemente un PTC llamado "Juntos" (el cuadro 5.3 del informe principal contiene detalles) y podria aprender de experiencias similares en la regibn, como "Bolsa Familia" en Brasil, "Oportunidades" en MCxico y "Bono de Desarrollo Humano" en Ecuador. xi Mejorar la asignacidn y la calidad de 10s maestros. En las ireas rurales se han puesto en marcha planes pilotos de incentivos para mejorar la asistenciade 10s maestros.Tales planes se debenexpandir a escala nacional y complementar con el suministro de capacitaci6n y materiales para 10s maestros, sobretodo en las ireas de educacidn bilingue y multigrado. Tambitn sera importante garantizar que el proceso de descentralizacibn no limite la capacidad de las autoridades para administrar 10s recursos del sector de unamanera eficaz y eficiente. Mejorar la oferta y la calidad de la educacidn bilingue. Para mejorar la asistencia de 10s estudiantes indigenas y especialmente de las nifias a las escuelas es necesario elevar el ntimero de maestros capacitados en educaci6n bilingue y multigrado y diseiiar y distribuir en esas escuelas 10s materiales didicticos correspondientes. Hacia el futuro, 10s esfuerzos por eliminar las barreras culturales a1 acceso deben aprovechar la mayor responsabilidad del sector frente a las autoridades locales y 10s usuarios que ha generadoel proceso de descentralizaci6n. Aumentar la oferta de educacidn preescolar y secundaria. Es factible lograr mejoras en la oferta de educaci6n preescolar por medio de modalidades de educaci6n no formal, como 10s centros de educaci6n infantil manejadospor mujeres, que reciben capacitaci6n y apoyo financier0 del gobierno a cambio del suministro de servicios bisicos de educaci6n. Se pueden lograr mejoras en la oferta de educaci6n secundaria poniendo en prictica modalidades alternativas y m6s flexibles de educacih, como laeducaci6n a distancia. 40. El sector salud tambitn enfrenta obsthculos relacionados con la demanda y la oferta que le impiden tener un impacto mayor en el mejoramiento de las vidas de 10s pobres. El Seguro Integral de Salud (SIS), que elimina las cuotas que pagan 10s usuarios y reembolsa por el sistema de honorarios por servicios todos 10s costos variables en que hayan incurrido 10s proveedorespiiblicos durante el suministro de un paquete bisico de prestaciones (principalmente 10s medicamentos esenciales), ha sido un paso importante en el mejoramiento del acceso de 10s pobres a la asistencia a la salud bisica, per0 el costo sigue siendo un problema para muchos de ellos. Adeds, el sistema de asistencia a la salud no cubre de manera suficiente a la poblaci6nindigena, en parte porque no todas las clinicas son sensiblesa 10s asuntos culturales relacionados con la prestaci6n de servicios de salud a este sector poblacional. Las cuestiones administrativas tambikn son unproblema: en el Pen5 distintos proveedores suministran servicios de salud, segiin el Ministerio de Salud y EsSalud. L a existencia de mtiltiples proveedores con diferentes mandatos podria ser la causa de ineficienciasen la asignaci6n de recwsos y en el us0de la capacidad existente. 41. Como en el cas0 de la educacibn, elevar la calidad y la cobertura de 10s servicios de salud requiere politicas de incentivos a la demanda y a la oferta, como 10s siguientes: Elevar la demanda de servicios de salud reduciendo 10s costospara 10s pobres. Aunque el SIS ha representadouna innovaci6n importante, es precis0 realizar mayores esfuerzos para reducir 10s costos directos y de oportunidad de la asistencia a la salud para 10s pobres. Hacer que 10s servicios de salud Sean d s accesibles para 10s pobres, y sobre todo para aquellos que son d s vulnerables, como las madres, 10s bebts y 10s ancianos, debe ser una prioridad. El SIS tambikn debe reducir la asignacih excesiva de recursos a la atenci6n terciaria y concentrarse en 10s niveles primario y secundario. El gobierno debe considerar la posibilidad de expandir 10s servicios subsidiados y de establecer un programa de transferencia condicionada de dinero relacionado con la asistencia a la salud. Reducir las barreras culturales en la asistencia a la salud. Tener en cuenta las expectativas culturales y las creencias de 10s indigenas en todo el sistema de salud puede eliminar o a1 menos mitigar el impacto de las barreras culturales. L a adopcibn, en 1994, del modelo de 10s Centros de Administracih Compartida, CLAS, que se basa en la participacibn de las comunidades locales en la xii planificacibn y la administraci6n de 10s centros primarios de asistencia a la salud, ha constituido un avance importante en esa direcci6n y debe expandirse. e Aumentar la eficiencia de 10s proveedores p6blicos de servicios de salud y la coordinacih entre ellos. Para mejorar la eficiencia del sistema de salud, el Ministerio de Salud (MINSA) ha firmado una sene de acuerdos de gestibn con las autoridades regionales de salud. Estos acuerdos vinculan 10s recursos con el desempeiio y 10s resultados. Hacia el futuro, entre 10s principales desafios relacionados con esos acuerdos de gesti6n esthn el seguimiento y la publicaci6n de 10s resultados de su ejecuci6n. Ademhs, para potenciar a1 mAximo el us0 de la capacidad actual, el MINSA ha tratado de lograr una mejor coordinaci6n con EsSalud. Ello ha resultado dificil desde el punto de vista politico, per0 10s esfuerzos deben continuar. Eso sera particularmente importante en un Bmbito cada vez mhs descentralizado en el que el riesgo de fragmentaci6n del sistema puede aumentar de manera significativa. EXPOSICION AL RIESGOY MOVILIDAD SOCIAL 42. L a baja productividad, 10s bajos niveles de ingreso y las limitadas oportunidades econ6micas no son las unicas barreras que 10s pobres deben superar. Su restringida capacidad para protegerse contra el riesgo por medio de la diversificaci6n de 10s ingresos y para ahorrar hace que 10s pobres Sean mhs vulnerables a 10s shocks econ6micos. Del mismo modo, 10s bajos niveles de movilidad social, medidos como la correlaci6n entre la experiencia de 10s padres y 10s logos de 10s hijos, tienden a perpetuar las desigualdades en lo que se refiere a ingresos y capital. Riesgoy vulnerabilidad 43. Aproximadamente el 20 por ciento de 10s hogares inform6 que sufri6 un shock en 2003. Tanto 10s hogares pobres como 10s no pobres estuvieron sujetos a shocks y tenian probabilidades de perder ingresos y activos como resultado de estos. Los shocks econ6micos fueron mhs frecuentes en las Areas urbanas, mientras 10s desastres naturales lo fueron en las rurales. A d e d s , dentro de las Areas urbanas y rurales era mhs probable que 10s hogares pobres experimentaran desastres naturales y accidentes, mientras era mhs probable que 10s nopobres sufrieran shocks econ6micos. 44. A1 intentar sobrellevar 10s shocks, 10s hogares pobres tendian a emplear estrategias basadas en aumentar la oferta de mano de obra o reducir el consumo, mientras era d s probable que 10s hogares no pobres dependieran de estrategias basadas en activos, como reducir 10s ahorros, o estrategias orientadas a 10s mercados, como solicitar prCstamos o hacer efectiva una p6liza de seguros. 45. Las estrategias que emplearon 10s pobres fueron menos eficaces que aquellas que usaron 10s no pobres para ayudar a sus hogares a superar el impacto de 10s shocks. Las estrategias a las que recurrieron 10s pobres tienen una eficacia reducida, ya que existe unlimite a1n6mero de horas que pueden trabajar las personas y es dificil bajar el consumo por debajo del nivel de subsistencia. Los hogares que se encuentran d s cerca de esos limites cuando se presenta un shock tendrrin d s dificultades para superar sus efectos. Envista de que esos hogares tienden a ser 10s d s necesitados, esto genera uncirculo vicioso de pobreza y vulnerabilidad. Por consiguiente, las intervenciones orientadas a mejorar la capacidad de 10s hogares pobres para ahorrar y tener acceso a 10s mercados financieros, asi como su acceso a medidas eficaces de proteccih social, pueden contribuir bastante a romper ese circulo vicioso. 46. Para dotar a 10s pobres de 10s mecanismos para ayudarse a s i mismos cuando se presentan shocks se requiere intervenciones destinadas a ampliar su base de activos, aumentar su acceso a 10s Xlll ... servicios e instrumentos financieros y facilitar el us0 de seguros de larga enfermedad o invalidez y de seguros de riesgos catastrbficos. Adem&, las medidas de proteccih social del sector publico para 10s pobres deben mejorar su focalizaci6ny su capacidad parareaccionar con rapidez a las crisis. Ayudar a 10s pobres a ampliar su base de activos. Aparte de las intervenciones destinadas a aumentar la productividad y 10s ingresos de 10s pobres, que ya han sido discutidas, el gobierno debe tomar medidas para mejorar el acceso a la vivienda y a la tierra y la seguridad de estos, que son con frecuencia 10s activos mAs valiosos de 10s pobres. Aumentar el acceso a viviendas adecuadas en las ireas urbanas y promover la titulaci6n de la tierra y la vivienda tanto en las Areas urbanas como en las rurales les permitiria a 10s hogares pobres usar esos activos como garantias para la obtencibn de crCditos en cas0 de necesidad. La titulaci6n tambiCn contribuiria en mucho a la activacih de 10s mercados de vivienda y terrenos, que actualmente tienen un volumen muy escaso de operaciones, especialmente en las Areas rurales, y por ende elevaria el valor de esos activos para el momento en que la liquidez sea necesaria. Otra opci6n es mejorar las transferencias ptiblicas a 10s pobres por medio de un programa de transferencia condicionada de dinero, un paso que el P e d ya estfi considerando (vCase el cuadro 5.3 del infonne principal). Esos programas logran el doble objetivo de proporcionar un alivio de corto plazo a la pobreza y promover las inversiones en capital humano a mediano plazo. Aumentar el acceso a 10s servicios financieros. Se podria salvar la brecha que existe entre 10s pobres y el sistema financiero mediante la expansi6n de 10s servicios de cajero autodtico a las Areas pobres y el suministro de programas de introduccih a1 tema financiero para 10s hogares pobres. TambiCn se podria lograr un mayor contact0 entre 10s hogares pobres y el sistemabancario mediante la canalizacibn a travCs de 10s bancos de 10s pagos que realizan 10s programas sociales, como se hace por ejemplo en Ecuador con el Bono de Desarrollo Humano. TambiCn se pueden crear instrumentos financieros concebidos para 10s pobres, como las cuentas de ahorro que pagan intereses d s bajos per0 no exigen un saldo m'nimo, o instrumentos de base comunitaria como 10s planes rotatorios de ahorro y crkdito. 0 Facilitar el acceso de 10s pobres a uningreso m'nimo y a 10s seguros de riesgoscatastrbficos. El acceso a un ingreso m'nimo se puede suministrar en la modalidad de 10s programas temporales de asistencia social condicionada a1 trabajo, un ejemplo de 10s cuales es el programa "A Trabajar" del Peru, o en la forma de pensiones no contributivas en el cas0 de personas mayores o discapacitadas, una opci6n cuya sostenibilidad fiscal se tendria que estudiar cuidadosamente antes de su implementacih. Los hogares pobres pueden tener acceso a seguros de riesgos catastr6ficos a travCs del gobierno o de proveedores privados. Aunque el suministro de seguros de desastre por el sector privado es bastantecomun en 10s paises desarrollados y entre 10s hogares acomodados, la liquidaci6n irregular de 10s siniestros, la carencia de titulos de propiedad de las viviendas y las tierras y la vivienda de calidad inferior hacen que sea dificil asegurar a 10s pobres. Sin embargo, existen experiencias exitosas en este aspect0 en Areas urbanas, como la de Manizales en Colombia, que puedenofrecer lecciones utiles. 0 Aumentar el acceso de 10s pobres urbanos y rurales a programas eficaces de protecci6n social. La implementacidnde un sistema de pensiones no contributivas para 10s ancianos necesitadospodria ayudar a prevenir el riesgo de pobreza en la tercera edad, sujeta a la limitaci6n de sostenibilidad fiscal ya mencionada. Del mismo modo, 10s programas que entiendan 10s factores determinantes del riesgo de 10sj6venes (las caracteristicas individuales, las experiencias de la familia, y 10s efectos que tengan 10s pares y el vecindario) y que hagan Cnfasis en la prevencih (por ejemplo, que minimicen el riesgo para el ingreso futuro por medio del suministro de incentivos para terminar la educacih secundaria), pueden contribuir a reducir la vulnerabilidad y el riesgo entre 10sj6venes. Ademis, 10s programas de busqueda de empleo y colocaci6n laboral y 10s servicios de guarderia infantil para las madres pobres xiv puedenaumentar la participacibn en el mercado laboral entre 10s hogarespobres, especialmente 10s de las Areas urbanas. En el sector rural, intervenciones como la introduccibn de nuevas semillas y variedades de pastura y la oferta de capacitacidn ap'cola bisica pueden contribuir a mejorar la seguridad alimentaria y 10s niveles nutricionales en Cpocas de crisis. Movilidad social 47. La movilidad social, medida como la relacidnentre las caracteristicas de 10s padres y 10s hijos y representada por la educacibn y la movilidad ocupacional, es baja y persistente en el Peh. Los recientes aumentos en la movilidad han sido el resultado de avances generales en 10s logros educativos y de cambios en la estructura productiva de la econom'a, en lugar de ser el resultado de una mayor igualdad en las oportunidades educativas y econbmicas, y han estado concentrados en la mitad de la distribucidn (del ingreso). 48. Est0 tiene implicaciones importantes para la adopcidn de politicas en un pais en el que 10s niveles de desigualdad todavia son elevados. El hecho de que la mayoria de 10s avances hayan sido impulsados por aumentos generales en el acceso muestra que las medidas que eleven la oferta, como la construccidn de escuelas y el aumento del n6mero de maestros, han sido eficaces para atraer a rrhs nifios hacia laescuelay mantenerlos en ella. Por otro lado, las evidencias de unescaso avance en relacidncon la "democratizacibn de la educacibn" indican que hay espacio para tipos alternativos de intervenciones que intenten transformar directamente la relacidn entre antecedentes socioeconbmicos y culturales por una parte y logros educativos por la otra; es decir, intervenciones que estimulen la demanda, como las becas basadas en el ingreso, las transferencias condicionadas de dinero y las intervenciones que enfrenten diferencias culturales, como la educacibn bilingue (que ya se comentb). Dado que la exclusidn social sigue siendo un problema para ciertos grupos, enfrentar el problemade la movilidad social, por medio de la promocidnde la movilidad educativa, se convierte en una prioridad. xv xvi OPPORTUNITIESFORALL: MAINMESSAGESAND POLICYRECOMMENDATIONS 1. Perufaces high levels of poverty and inequality. In2004, just over half of Peru's population was poor and about 20 percent were extremely poor. Although regional comparisons are difficult due to the use of different poverty lines across countries, Peru's poverty levels are below those of Ecuador and Colombia, but above those of Argentina and Brazil basedon a US$2/day poverty line. Poverty levels are significantly higher in rural areas, while urban areas-most notably metropolitan Lima-are the most unequal. Inequality, measuredby the Gini coefficient, stood at 0.43-below the Latin American average of 0.52, but still highby internationalstandards. 2. Poverty has been slow to respond to the country's impressive economic growth in recent years. After improvements during the 1990s, poverty increased in the wake of the 1998 economic crisis, while extreme poverty remained stable. The economic recovery since 2001 has had a strong positive impact in reducing extreme poverty, but progress on poverty rates has been limited to rural areas. 3. A main focus of this report is to explain why growth has not translated into more rapid poverty reduction. The first reason i s simply that poverty reduction takes time, particularly considering the pattern of economic volatility in Peru over the past several decades, which makes businesspeople reluctant to invest injob-creating endeavors. As well, in the wake of the high investment years in the 1990s followed by the 1998 crisis, many businesses have excess production capacity that i s only now, after five years of growth, being worked out. Thus, the good macroeconomic managementof recent years i s beginning to show fruit in terms of poverty reduction andjob creation, and this will likely accelerate if the government keeps inplace sound macroeconomic policies. 4. This report also points to a number of obstacles that inhibit economic growth from leading to greater poverty reductioninPeru, and which should be addressedby government policy reforms. Growth hasbeen focused on natural resource extractionindustries, which are highly capital-intensive and generate few jobs, and inthe rural agricultural and urban informal employment sectors, which are characterized by low productivity and, therefore, low wage growth. Strengthening the linkage between growth and more, better-paid jobs requires micro-level policy interventions to increase incentives to formal-sector employment and diversify economic activity into more labor-intensive areas, such as non-traditional exports. While this diversification has begun-as evidenced by the rapid growth of non-traditional exports in recent years-further policy reforms can help this process along. These reforms should be accompanied by measures to boost human capital and access to public services by the poor, thereby increasing equity ineconomic opportunity. THENATURE,DISTRIBUTIONAND EVOLUTIONOFPOVERTY 5. In 2004, just over half the Peruvian population was in poverty: while about one-fifth was in extreme poverty (see Table 1). Poverty is significantly higher and deeper in rural compared to urban areas. Rural poverty stands at 72 percent and extreme rural poverty i s 40 percent, while inurban areas the rates are 40 percent and 8 percent, respectively. Poverty is also higher and deeper in the Sierra and the Selva compared to the Costa.' Most of the regional variation in poverty rates can be attributed to 4. This report bases poverty rateson consumption as opposedto income, becauseconsumptiontends to stay more stable over the course of a year, and also because consumption tends to be reported more accurately than incomeonhouseholdsurveys. 5. Peru is commonly divided into three geographic regions: "Sierra" (mountains), "Selva" (jungle), and "Costa" (coast). variation inhouseholdcharacteristics and in access to basic services and road infrastructure, rather than to geographical differences, such as altitude and temperature. In other words, observationally equivalent households have similar probabilities of being poor irrespective of the geographic characteristics of their region of residence. Inequality as measuredby the Gini coefficient i s .43 nationally, but i s lower inrural areas, especially the ruralCosta, and highest inLima. Table 1:Poverty indicators in2004 Poverty Extreme poverty Gini Headcount Gap Severity Headcount Gap Severity National 51.6 18.0 8.4 19.2 5.3 2.1 0.43 Area of residence Urban 40.3 12.4 5.3 7.9 1.8 0.7 0.39 Rural 72.5 28.3 14.1 40.3 11.7 4.8 0.32 Geographic region UrbanCosta 37.1 10.6 4.5 6.2 1.4 0.5 0.34 Metropolitan Lima 36.6 10.4 4.1 3.4 0.6 0.2 0.40 Rural Costa 53.5 16.4 7.0 14.6 3.1 0.5 0.32 Sierra 67.7 27.2 13.9 36.5 10.9 4.5 0.39 Selva 59.5 19.7 8.8 26.4 6.3 2.2 0.36 Source: Authors' calculationsusingdatafromENAHO2004 (INE1)-Annual samplecoveringthe periodJanuaryto December2004. 6. Poverty increased significantly as a consequence of the 1998 economic crisis and has remained stable around its new post-crisis level afterwards, while extreme poverty was unaffected by the crisis and has improved since 2001. These findings suggest that while poverty has been slow to react to economic growth, this reaction has begun and it has had the greatest impact on the poorest of the poor. 7. The evolution of the national poverty rate, however, hides important differences across urban and rural areas and across regions. While poverty rose between 1997 and 2000 in both urban and rural areas, itremained stableinthe former butdeclinedinthe latterbetween2001and2003 (seeTable 2). Similarly, after generalized increasesacross regions due to the crisis, poverty continued to increase slightly inLima, while it improved inthe Selva andremainedstable inthe rest of the country. Geographic Zone 2001 2002 2003 2004 Urban 42.0 41.0 39.5 40.3 Rural 77.1 77.7 75.7 72.5 NationalTotal 54.3 53.8 52.2 51.6 Datafor 2001and2002correspondto the last quarter. Datafor 2003 correspondto May-December. Datafor 2004correspondto January-December. Table 3. Percent of PopulationBelow Extreme Poverty Line, 2001-04 Geographic Zone 2001 2002 2003 2004 Urban 10.2 9.4 8.9 7.9 Rural 49.8 51.7 45.9 40.3 NationalTotal 24.1 24.2 21.9 19.2 Source: Author's calculationusingdatafrom ENAHO2001-2004(INEI). 2 Datafor 2001and2002 correspondto the last quarter. Datafor 2003 correspondto May-December. Datafor 2004 correspondto January-December. 8. By contrast, the evolution of extreme poverty was similar across urban and rural areas and across regions during this period. Extreme poverty remained stable in 1997-2000, and improved steadily in 2001-2004 (see Table 3). Improvements in extreme poverty were concentrated in rural areas and among the poorest departments, including Ayacucho, Apurimac, Cuzco y Cajamarca. Significant improvements can also be seen inthe poverty gap and severity duringthis period, which declined from 20.9 to 18.5 and from 10.7 to 8.9, respectively, during2001-2004. 9. Poverty as measured by the Unsatisfied Basic Needs index (UBN) has declined steadily during the last 10 years as access to basic services and infrastructure expanded. The fraction of the population with at least one UBNhas fallen from 57 in 1993 to 40 percent in 2003. Progresshas been significant in both urban and rural areas, as the UBNhas declined from 42 to 25 and from 90 to 71percent inurban and rural areas respectively during this period. Most of this decline, however, took place during the 1990s, before public investment levels were cut significantly, leading to a slow-down in progress on access to basic services. 10. The divergence between the evolution of monetary and more structural, non-monetary measures of poverty duringa period of significant economic turmoil i s not exclusive to Peru. Ecuador had a similar experience duringthe 199819 crisis and immediately after the 2000 dollarization, when monetary poverty increased significantly as a consequence of the economic downturn but the UBN index continued to decline along its long-term trend. GROWTH POVERTY: ANOVERVIEW AND 11. The relationshipbetween economic growth and poverty i s weaker in Peru than inother countries inthe world. That is, Peruneedsto grow faster than other countries to lower poverty or evento prevent it from increasing. In recent years this weak relationship between growth and poverty has been a direct consequenceof the nature of economic growth in Peru, which has not been sufficiently broad-based, due to four factors. First, growth inper capita terms was relatively slow during 1997-2004. That is, although the economy grew, it could not keep up with population growth. Income and consumption per capita declined by -1.6 and -14 percent respectively during this period according to data from the Encuesta Nucional deHogares (ENAH0).6 12. Second, growth was biased towards sectors with highcapital intensity and low demand for labor, such as mining, or towards sectors with low labor productivity and wage levels, such as agriculture, although more recently employment levels are now growing in other sectors as well, such as textiles and non-traditional exports. The mining and agricultural sectors grew at annual rates of 7.6 and 3.5 percent during 1997-2004. During this period employment in the mining sector grew by 3.6 percent, while employment inthe agricultural sector increasedby 4.6 percent. Similarly wages grew at an annual rate of 12.4 percent inthe miningsector, compared to a decline of 1.8 inthe agricultural sector. Wage growth in the miningsector, however, i s unlikely to have benefited the poor, given the composition of the labor pool employed inthe sector, which tends to be relatively skilled. 13. Third, growth has historically been highly volatile. Untilthe 2001-2005 period, since 1965 Peru had never enjoyed more than four consecutive years with growth rates above 3.5 percent, while it has had 6. This contrast with data from national accounts that shows that GDP per capita grew by 2.5 percent during the same period. However, poverty is measured using the ENAHO. 3 13 years of growth rates below 2 percent and seven years with negative growth rates. Volatility inturn has translated into high levels of uncertainty among employers and other economic agents, undermining incentives to invest and hire new workers. This legacy of volatile growth i s gradually being overcome through the government's prudent macroeconomic management and disciplined fiscal policies. 14. Fourth, investment and employment creation in Peru have been low because there i s a significant amount of excess capacity inthe productive system resulting from the 1998-99 economic downturn. The use of installed capacity has increased since 2000, but there i s still a significant share that i s currently underutilized. Approximately 30 percent of employers declared to be using 60 percent or less of their business' installed capacity in2003, down from 37 percent in 2000, while 38 percent declared to be using more than 80 percent, up from 19 in 2000. The latest available data indicates that average usage of installed capacity stood at 74 percent in October 2005. Similarly employers appeared wary about hiring new workers. In 2003, 15 percent of employers declared to be willing to hire new workers and over 20 percent declared to be willing to fire existing ones, compared to 10and 30 percent respectively in2000. 15. As a result of the last three factors listed above-conomic volatility, growth incapital-intensive sectors, and excess capacity-employment generation although positive has been insufficient and investment rates have remained low, and as a consequence growth has failed to benefit the poor in urban areas. Although employment in firms with more than 10 employees has recovered since 2000, it still remains below pre-crisis levels. Moreover the overall employment rate has remained constant, or even .declined slightly since 2000, at the same time that the composition of employment has shifted towards higher informality. The number of hours worked has also increased slightly, suggesting that labor needs may have been covered through more intensive use of already employed workers rather than through new hires. Similarly, although investment levels have increased in real terms since 2003, investment rates have been declining steadily as a percentageof GDP since 1998. 16. The analysis above indicates that macro-level policies to maintain and strengthen economic growth, and to reduce uncertainty can ensure that economic growth i s not only sustainable but contributes to poverty reduction. These macro policies are largely in place and should be continued. However, to further strengthen the link between growth and poverty reduction, they must also be complemented by micro-level reforms that will offer greater incentives to invest and hire new workers in both urban and rural areas, and also to raise the level of human capital in the workforce and equalize economic opportunities. These issues are discussed below. ECONOMIC OPPORTUNITIESFORTHEURBANPOOR 17. Recent improvements in average urban employment and wage levels have failed to translate into lower urban poverty rates because they have been concentrated among formal, more educated workers, employed in larger firms, and have not extended to the informal sector where most of the poor are employed. Overall employment (total number of workers) increased by an annual average of 4 percent between 2000 and 2004. However, because the labor force participation rate also increased, the overall employment rate remained constant or even declined slightly duringthis period. Wages of managers and white collar workers in the formal sector increased (on a wage index with 100 as the base) from 117 to 158 and from 94 to 104, respectively, between 1997 and 2004, compared with a change from 86 to 89 among blue collar workers. 18. The lack of employment and wage gains for the poor has been aggravated by weak linkages between large and small firms which have prevented growth among the former from trickling down to the latter. In2004,62 percent of all large firms sold at most 20 percent of their productionto micro and small firms and 46 percent bought at most 20 percent of their inputs from these same firms, compared to 1.2 4 and 4.1 percent that sold or bought more than 80 percent of their production and inputs, respectively, to andfrommicro and smallfirms. 19. Future declines in urban poverty will depend on the capacity of the urban economy to generate more productive, well-paid jobs, particularly in those sectors that employ the poor. The rest of this section analyzes and offers policy recommendations to tackle existing constraints to urban employment creation by large, medium and small firms, paying special attention to labor market rigidities and the role of labor legislation. Moreover, because most of the urban poor are employed in small, informal businesses, it also examines and makes recommendations on incentives for informal firmsto formalize themselves and the determinants of productivity of informal activities. Removing Constraints to UrbanEmployment Creation 20. Twenty-five percent of all manufacturing firms interviewed in the Peru Investment Climate Survey (World Bank, 2003) declared that, if faced with no constraints, they would like to increase the number of permanent workers they employed, compared with less than 5 percent that would decrease it, and 70 percent that would maintain it at its current level. The reasons given for the difference between actual and desired hiringand firing vary somewhat by firm size. It is clear, however, that legislated non- wage labor costs, firing costs and, to a lesser extent, uncertainty about future demand for the firm's products constitute the key reasonsfor this difference. 21. On paper, permanent employment conditions in Peru are good and protection i s high, compared to other countries inthe region and the world. For instance, average firing costs in 1999 were estimated to be equal to 13.8 averagemonthly wages, down from 15.0 in 1987, but still significantly higherthan the regional average of 5.5 or the English-speaking industrialized countries average of 1.5. Similarly the legislation protects labor relationships and, to a lesser extent, access to social security. These regulations, however, are relatively easy to avoid by using temporary rather than permanent contracts to hire new workers, or by operating in the informal sector. For instance temporary and hourly employment represents 20 percent of all private sector salaried employment and 50 percent of all contract-based employment inMetropolitanLima. 22. As a result, labor markets inPeru are defacto flexible, althoughthis flexibility comes at the cost of low andunevenly distributed employment protection and access to social security. Only 18 percent of all urban employment and 52 percent of all salaried employment is hired infull compliance with Peruvian labor regulation, compared to 40 and 60 percent respectively in LAC. Similarly non-compliance with minimumwage equalsjust under one quarter of the relevant population inPeru, while it is 10percent in Latin America and the Caribbean (LAC). This dichotomy between high levels of protection dejure and low levels of protection defacto is common to other countries in the region where enforcement i s weak and the incidence of informality is high. 23. The impact of labor legislation extends beyond formal permanent employment to affect overall employment levels and composition, as has been extensively documented in the literature on Peruvian and other labor markets. To a large extent these patterns result from differences in wage and non-wage costs between permanent employment and temporary (and informal) employment. Two measures that would reduce these differences without increasing the relative cost of formal employment are: e Reduce firing costs to increase incentives to hire. Firing costs could be reduced in a number of ways. The most drastic approach would call for a reduction in severance payments. This measure could be applied only to new contracts or, alternatively, compensation for workers hired under the previous regime could be considered. Softer approachescould include an increase intrial periods for new workers and a more flexible use of "economic reasons" as a cause for firing. 5 0 Reduce non-wage costs. Peru is among the countries in the region with the most generous legislation in terms of paid vacations, together with Brazil and Panama. This i s not only very costly to employers (especially considering low labor productivity), but only a small percentage of workers actually enjoys the vacation period, suggesting that at current income levels workers are willing to sell leisure time for additional income. A more flexible approach could be to make vacation a function of the worker's experience or tenure and, hence, of labor productivity. ProvidingIncentives for Formalization 24. Peru's high level of informal employment i s costly for households, fums and the government. Low productivity among informal businesses translates into lower earnings for those employed in the sector-average hourly labor income in the informal sector is 50 percent below that of the formal sector, even when similar workers in similar jobs are compared. Informal workers also lack access to employment-linked social protection, such as health and pension benefits or severance payments in the event of job loss. Although some of these workers may have voluntarily foregone such benefits in exchange for higher wages or more flexibility, more than 50 percent of the poor working in the informal sector do so involuntarily, according to surveys. Informal firms also face a number of costly obstacles, including limited access to credit, business associations, and government programs promoting economic activities and exports. From the government's perspective, non-compliance with labor and sales taxes among informal firms has a negative impact on fiscal revenues. 25. High levels of informality and low levels of formal permanent employment creation can be attributed to slow and costly business registration procedures, complex tax filing procedures for businesses, and inflexible labor regulation (particularly onerous on small firms). Policies directed at lowering (permanent) employment costs were discussed above. Other interventions aimed at providing incentives for formalization of small business and increasing their access to credit could include streamlining of: 0 Registration procedures for small firms. A reduction of red tape to bring the cost of registration procedures in line with those of close competitors in and outside the region will make it easier for firms to comply with these requirements. Recently approved legislation to implement a special, simplified registration regime for micro and small f i m s and to allow these firms to pay lower contributions to the pension and health systems constitutes a step inthis direction. Unfortunately the actual implementationof this and other changes contemplated in the new law has so far not yield the desired effect, with only 3,500 existing informal fm having formally registered. It would be important to understandthe reasonsfor the limitedimpact of these measures. 0 Tax Tding mechanisms. A special, simplified filing regime for micro and small firms already exists, but these andother firms could benefit fromfurther simplification. Forinstance, filing on the basis of readily observable business characteristics and according to predetermined tax tables could be considered. These systems make it easier to file taxes for firms that do not rely on fully formal, and often costly, accounting and even firms that interact with a large number of informal partners. IncreasingProductivity of Small InformalBusinesses 26. While promoting formality in the private sector should be a priority, it i s also a reality that the informal sector is very large in Peru. Approximately half of the working urban poor are self-employed, all of them informally, and an additional 30 percent work for micro or small fums, many of which are also informal. Similarly 40 percent of all informal entrepreneurs (self-employed or otherwise) are poor, comparedto 15 percent of formal entrepreneurs. Thus, identifying the determinants of the productivity of 6 informal activities and implementing policies aimed at increasing this productivity are key in order to help the urbanpoor step out of poverty. 27. There is significant variation inthe productivity of small, informal businesses, where productivity i s measured as value added per worker. This variation can be attributed to differences in entrepreneur, worker and firm characteristics. As a result, lower levels of productivity among poor entrepreneurs and, consequently, lower wages among their employees can be explained by lower levels of education of both employers and workers, lower levels of market integration and lower accessto basic infrastructure. 28. Differences between poor and non-poor entrepreneurs in terms of business practices and characteristics are not the product of independent factors but rather are inter-related. For instance the use of market-oriented practices, and access to capital and infrastructure are correlated with business location. Businessesthat operate out of a commercial locale are more likely to use some form of accounting and to employ a larger share of paid workers than businesses that operate in the street or out of the entrepreneur's home. In addition access to machinery and other tools i s higher among businesses in commercial or non-commercial locals than among those run in the street, while the use of a vehicle i s much higher among the latter-partly due to its use as a substitute for a proper local. Finally, running a business froma commercial locali s correlated with higher access to phone and water services. 29. Higher levels of productivity of informal self-employment and small businesses could then be achieved by: 0 Increase the level of skills of both entrepreneurs and salaried workers. General increases in the skill level of the labor force can be achieved by investing informal education (discussed inparagraph 40 below) andor by improving the relevance and coverage of the training system. Training does not need to be provided by the public sector, but rather incentives can be put in place by the government for firms to contract the desired training with private, and properly accredited, providers. The existing ProJoven program, which provides training for young workers, could be extended to cover other demographic groups. a Increase access to commercial locales and the use of market-oriented practices. Commercial spaces for small businesses in markets or other locations could be offered to those operating in the street in exchange for a rental (leasing) fee. This fee can be made to increase over time to both facilitate early investment and reflect potential future gains in productivity. Increased access to such spaces would serve as a platform for the economical provision of basic infrastructure and business services such as management and accounting practices, simplified access to credit, and legal services, which inturn translate intohigher productivity. It would also contribute to the decongestion of those streets and areas where these businesses would otherwise operate, easing traffic and decreasing hazards. ECONOMICOPPORTUNITIESFORTHE RURALPOOR 30. Rural households obtain most of their income from agricultural activities, but important differences exist between poor and non-poor households in terms of their income-generating strategies. Poor householdstend to rely on agriculture exclusively, while non-poor households tend to also engage in non-agricultural activities. Moreover, poor households are more likely to rely on a single source of income, while non-poor households are better able to diversify income risk by not relying exclusively on one particular source. Roughly half of allruralhouseholds obtain all income from self-employment inthe agricultural sector, while the rest combine agriculture with other types of work. Poverty rates are significantly higher among those employed in the agricultural sector (80 percent) than among those employed inthe non-agricultural sector (60percent). 7 31. Most of the variation in rural household income i s explained by variation in non-agricultural income from salaried employment. Furthermore, the share of agricultural incomedeclines as total income increases. Although these stylized facts seem to suggest that non-agricultural employment provides an exit out of poverty, in reality most rural households tend to obtain income from both the agricultural and non-agricultural sectors4.e. they rely on income-generating strategies rather than on particular sectors or activities. A household's ability to implement a profitable income-generating strategy determines its poverty status. 32. Participation in these income-generating strategies i s a function of household characteristics and endowments. Better household endowments (such as higher education) and access to infrastructure and public services allow householdsto use strategiesthat include non-agricultural activities, while ownership of agricultural assets and lack of liquidity make it more costly for households to abandon strategies that include agricultural activities. 33. At the household level, both agricultural productivity and labor income are positively correlated with human capital (such as higher education), access to credit and to basic services, telecommunications and road infrastructure. At the regional level the returns to agricultural and salaried activities depend on the deepness and dynamism of regional markets and on overall levels of productivity. Population density and access to infrastructure are higher in the Costa than in other areas, despite important improvements during the 1990s in the Sierra and the Selva. Both factors could potentially contribute to create more integrated and dynamic markets in rural areas, as well as to better connect rural and urban areas. Therefore, investments aimed at improving access to assets, services and markets in those areas that are lagging behind mustbe undertaken inorder to overcome regional differences. 34. Three important issues must be kept in mind when assessing policy options for the rural sector. First, the nature of rural poverty is heterogeneous and varies significantly across regions, so that interventions and projects need to take into account local specificities to ensure maximumeffectiveness. Second, because land i s scarce relative to the population it has to support and agricultural productivity is low, many of those currently employed in agricultural activities will have to dramatically improve their productivity or abandon agriculture in order to escape from poverty. This implies that a rural development strategy for Peru must be multi-sectoral and consider the interaction between agricultural and non-agricultural activities. Third, it i s important to note that there already exist several programs in rural areas that support interventions in the areas identified above and that provide a structure through which the government can work towards the goal of rural inclusive growth. These programs, however, suffer from several problems that needto be addressedif further interventions are to be effective. 35. With the above considerations in mind, three key areas require government action if the rural poor are to benefit from the economic opportunities generatedby overall economic growth: 0 Integrate rural areas into national markets to increase economic opportunities. The most obvious actions to facilitate contacts between agents and the transport of merchandise between rural and urban areas are simply to improve the road network, particularly secondary and tertiary road systems, to allow producers to get their productsto market quickly and inexpensively, and to invest in telecommunications in rural areas, to allow rural dwellers to have timely access to relevant market information. The public sector can also take actions to facilitate knowledge and technology transmission from urban to rural areas, and to develop stable economic relationships that ensure a constant demand of agricultural and non- agricultural products for industrial processing andlor exports, and createincentives for producinginbulk. e Improve access to credit among rural producers. Rural credit i s restricted by the difficulties of many producers, particularly those in small farms, to comply with the administrative and guarantee requirements of financial institutions. As a result most existing credit i s informal, or provided by 8 small loans and savings cooperatives. These cooperatives need to be strengthened and so do other institutions with similar goals, such as women's credit groups. Credit regulations needto be modified to allow for the use of family assets, such as machinery and livestock, as collateral, while taking care of not increasing credit risk and defaults by complementing increased access with better monitoring. At the same time efforts to increaseland titling should continue. Provisions should also be made to account for the highprevalence of communal property of land among the indigenous population, and the negative impact that this may have of the capacity of the individuals in these communities to access credit. e Increase humancapital levels in rural areas. Improving educational levels and standards in rural areas can be achieved through a series of interventions including: (i) expanding bilingual education through the provision of adequate teaching and learning materials and the recruitment and training of Quechua-speaking teachers; (ii) expanding secondary education, either through formal schooling or through distance learning; and (iii) creating incentives for school attendance through conditional cash-transfer programs or improvements in the feeding and nutritionprograms offered in schools (a more detailed discussion on the issue of education i s provided in paragraph 40 below). Technical assistance can also be improved. Some public extension services are provided by INCAGRO, PRA and FONCODES, but a large number of small farmers and the rural poor are still excluded due to their high cost. Further efforts to support the provision of demand-driven technical assistance, accompaniedby marketingand managerial assistance are then necessary. ACCESSTOPUBLICSERVICESAND INSTITUTIONS 36. Access to public services i s a cross-cutting issue, essential to help the poor build human capital and also to protect the vulnerable, both inrural and urban areas. It is also an area directly susceptible to policy choices, and i s hence a likely target for improvements to help improve links between economic growth and poverty reduction. Access to public services such as health care, education, and social protection i s low among the poor, among indigenous groups, and in rural areas. The poor are also less likely thanthe non-poor to come incontact with various public institutions,rangingfromcentral andlocal government offices to public banksto the judiciary system. 37. This report does not offer an in-depth analysis of the education, health and social protection sectors (such discussion can be found in Peru: Accountability in the Social Sectors, 2005), but instead notes some of the main issues in these areas, specifically as they relate to the problem of improving the humancapitalof workers seekingto pullthemselvesout ofpoverty. 38. The education sector exhibits a number of critical weaknesses that reduce its impact on the poor, inboth urban and rural areas. On a broad level, many poor people find the opportunity cost represented by education to be not worth it, reducing the demand for education. Because education quality i s low, andopportunities available after finishingschool to be limited, many families prefer to havetheir children work and earn extra income, even if only a small amount, rather than attend school. This i s particularly true in rural areas, which are underserved by good teachers who prefer to work in urban areas. As well, teacher absenteeism i s very high inrural areas. Indigenous school attendance is especially low, in good part because adequate bilingualhicultural curricula for the indigenous population are not sufficiently available. Another problem, also more pronounced inrural areas, i s the limited supply of pre-school and secondary school, which have low enrolment rates. 39. Increasing the quality and coverage of education will require demand- and supply-side policies suchas: 9 Promote increased demand for education. Increases in the demand for education can then be induced by effectively lowering its costs (both direct and opportunity costs) through conditional cash- transfer (CCT) programs or scholarships, and through the implementation of flexible schooling schedules that allow children and youngsters to engage in other activities during the day. Peru has recently launched a CCT, Juntos (see Box 5.3 in the main report for details), and could learn from similar experiences in the region such as Bolsa Familia in Brazil, O p o ~ u ~ i d ~ineMexico, and s Bono de Desarrollo Humano inEcuador. Improve the allocation and quality of teachers. Incentives schemes aimed at increasing teacher attendancehave been implemented inpilot form inrural areas. These schemes should be expanded to the national level and complemented with the provision of teacher training and materials, particularly in the areas of bilingual and multilevel education. It will also be important to ensure that the decentralization process does not hinder the capacity of the authorities to manage the sector's human resourceseffectively and efficiently. Improve the supply and quality of bilingual education. To improve school attendance by indigenous students, especially indigenous girls, increase the number of teachers trained inbilingual and multi-level education, and develop and distribute the relevant school materials to these schools. Looking ahead, efforts towards the elimination of cultural barriers to access should take advantage of the increased accountability of the sector towards local authorities and users brought about by the decentralization process. Increase the supply of pre-school and secondary education. Improvements in the supply of pre- school education can be achieved through non-formal schooling modalities such as women-operated child education centers, which receive training and financial support from the government in exchange for the provision of basic education services. Improvements in the supply of secondary education can be obtained through alternative, more flexible schooling modalities, such as distance learning. 40. The health sector also faces both demand- and supply-side obstacles to having a greater impact improving the lives of the poor. The Seguro Integral de Salud (SIS), which eliminates user fees by reimbursingpublic providers on a fee-for-service basis for all variable costs incurredduringthe provision of a basic benefit package (mainly essential drugs and medical supplies), has been an important step in improving access of the poor to basic health care, but cost i s still a problem for many poor people. The indigenous population i s also underservedby the health care system, inpart because not all healthclinics are sensitive to cultural issues related to health care for the indigenous. Administrative issues are also a problem: health services are provided by different suppliers in Peru, including the Ministry of Health (MSP) and the ESSALUD. The existence of multiple providers with different mandates can potentially cause inefficienciesinthe allocation of resourcesand inthe use of existing capacity. 41. As is the case with education, increasing the quality and coverage of health services requires demand- and supply-side interventions such as: 0 Increase the demand for health services by lowering costs for the poor. While SIS has been an important innovation, further efforts are neededto reduce direct and opportunity costs for health care for the poor. Making health services more accessible to the poor and particularly to those who are more vulnerable among them, such as mothers, infants and the elderly, should be a priority. SIS should also reduce excessive resource allocations to tertiary care, and focus on the primary and secondary levels. The government should consider expanding subsidized services andlor institutinga conditional cashtransfer programrelated to health care. 0 Reduce cultural barriers in health care. Better accommodating the cultural expectations and beliefs of indigenous people within the health system can eliminate or at least mitigate the impact of cultural barriers. The adoption of the CLAS model in 1994, based on the participation of local 10 communities in the planning and management of primary health care centers, has constituted an important move inthis direction, and should be expanded. Increase the efficiency of and coordination among public health providers. Inorder to increase efficiency in the health system, the MINSA has signed a series of management agreements with regional health authorities. These agreements link resourcesto performance and outcomes. Looking ahead, the main challenges regarding the management agreements include the monitoring and publication of performance results. Inaddition, inorder to maximize the use of existing capacity the MSP has sought better coordination with ESSALUD. This has provedpolitically difficult, but efforts should continue. This will be particularly important in an increasingly decentralized environment where the riskof fragmentation inthe system may rise significantly. EXPOSURETORISKAND SOCIALMOBILITY 42. Low productivity, low income levels and limited economic opportunities are not the only barriers the poor must surpass. Restricted capacity to hedge against risk through income diversification and to save makes the poor more vulnerable to shocks. Similarly low levels of social mobility, measured as the correlation between parental background and children's outcomes, tend to perpetuateexisting inequalities interms of incomeandendowments. Riskand vulnerability 43. Approximately 20 percent of all households reported that they suffered a shock in 2003. Both poor and non-poor households were subject to shocks and were likely to lose income and assets as a consequence of these shocks. Economic shocks were more prevalent in urban areas, while natural disasters were more frequent in rural areas. In addition, within urban and rural areas, poor households were more likely to experience natural disasters and accidents, while non-poor households were more likely to suffer economic shocks. 44. In coping with shocks, poor households tended to use behavioral strategies, such as increasing labor supply or cutting down consumption, while non-poor households were more likely to rely on assets- based strategies, such as reducing savings, or market-based strategies, such as requesting a loan or cashing an insurance policy. 45. The strategies implementedby the poor were less effective than those of the non-poor inhelping households overcome the impact of shocks. There are limits to the effectiveness of the behavioral strategiesused by the poor since individuals can only work so many hours and since it i s difficult to bring consumption under the subsistence level. Households that are closer to these limits at the time of a shock will find it harder to overcome its effects. Becausethese households tend to be the neediest, this creates a vicious circle of poverty and vulnerability. Consequently, interventions aimed at increasing poor households' capacity to save and to access financial markets, as well as their access to effective safety nets can go a long way inbreaking this vicious circle. 46. Enabling the poor help themselves in the face of shocks will require interventions aimed at broadening their assets base, increasing their access to financial services and instruments, and facilitating the use of income or catastropheinsurance. Further, public-sector safety nets for the poor should improve targeting and their ability to react quickly to crises. Help the poor broaden their asset base. Apart from interventions directed at increasing the productivity and earnings of the poor (discussed above), the government should take measures to improve access to and security of housing and land, often the most valuable asset held by the poor. Increasing access to adequate housing in urban areas and promoting housing and land titling in both urban and rural areas would allow poor households to use them as collateral for credit if necessary. 11 Titling would also go a long way inactivatingwhat are currently very thin housingand landmarkets, especially in rural areas, and thus increasing the value of these assets when liquidity is needed. A further option is to improve public transfers to the poor through a conditional-cash transfer program, a step Peru is current considering (see Box 5.3 in the main report). These programs serve the double objective of providing short-term poverty alleviation and promoting medium-term human capital investments. Increase access to financial services. Bridgingthe gap that exists between the poor and the banking system could be done by expanding ATM services to poor areas, and by providing financial literacy programs for poor households. Increased contact between poor households and the banking system could also be achieved by channeling social programpayments through banks, as i s done for example inEcuador inthe case of the Bono de Desarrollo Humano. Special financial instruments catering to the poor could also be created, for example savings accounts that pay lower returns but do not require a minimumbalance, or community-based instruments such as rotating saving and credit schemes. Facilitate access to income and catastrophic insurance for the poor. Income insurance can be provided in the form of temporary workfare programs, of which Peru's A Trabajar is an example, or as non-contributory pensions in the case of older or disabled individuals-an option whose fiscal sustainability would have to be carefully examined prior to its implementation. Involving private insurance companies may be possible. Poor householdscan access catastrophic insurance through the government or through private providers. Although provision of disaster insurance by the private sector is fairly common in developed countries and among well-off households, irregular settlements, lack of housing and landtitles and sub-optimal housing makes the poor hardto insured. There exist, however, successful experiences in this regard in urban areas, such as that of Manizales inColombia, that can offer useful lessons. Increase access to effective safety net programs for the urban and rural poor. The implementation of a non-contributory minimum pension system for the needy elderly could help prevent the risk of poverty in old age, subject to the fiscal sustainability constraint mentioned above. Similarly programs that understand the determinants of youth risk (individual characteristics, family background, peer and neighborhood effects) and emphasize prevention (e.g., minimizing future income risk by providing incentives for secondary education completion) can help reduce vulnerability and risk among youth. Further, job search and placement programs and day care services for poor mothers can increase labor market participation among poor households, especially inurban areas. Inthe ruralsector, interventions suchas introducingnew seeds andpasturingvarieties and offering basic agricultural training can help improve food security and nutritional levels in times of crisis. Socialmobility 47. Social mobility, measured as the relationship between parental and children's characteristics and proxied by education and occupational mobility i s low and persistent in Peru. Recent increases in mobility have been the result of across-the-board gains in educational attainment and changes in the productive structure of the economy, rather than the result of higher equality of educational and economic opportunities, and have been concentrated inthe middle of the (income) distribution. 48. Inacountry where inequality levelsare still high, this has importantpolicy implications. The fact that most progress has been driven by general increases in access shows that supply measures, such as school construction and a higher number of teachers, have been effective in getting and retaining more children in schools. On the other hand, evidence that little progress has been made regarding the "democratization of education" indicates that there i s room for alternative types of interventions that directly attempt to transformthe relationshipbetween socio-economic and cultural backgroundon the one hand, and education achievement on the other-that is, demand interventions, such as income-based scholarships, conditional cash-transfers, and interventions that address cultural differences, such as 12 bilingual education (discussed above). Given that social exclusion continues to be a problem for certain groups, tackling the issue of social mobility, through the promotion of education mobility, becomes a priority. 13 14 INTRODUCTION Peru faces high levels of poverty and inequality, and poverty has been slow to respond to the country's impressive economic growth in recent years. The main focus of this report i s to explain why growth has not translated into more rapid poverty reduction. This report also points to a number of obstacles that inhibit economic growth from leading to greater poverty reduction in Peru, and which should be addressed by government policy reforms. The report i s structured as follows. Chapter 1 presents a poverty and inequality update and discusses poverty and inequality trends for 1997-2004. Chapter 2 analyzes the nature of economic growth and its impact on poverty and inequality. Chapter 3 discusses the nature, distribution and evolution of poverty in Peru during 1997-2003. Chapters 4 and 5 focus on the productive sectors in urban and rural areas respectively, and discuss policies aimed at increasing economic opportunities for the poor. Lastly, Chapter 6 examines the issues of vulnerability, measured as exposure to shocks, and exclusion, measured as (the lack of) social mobility and access to basic services, infrastructure and public institutions. 15 16 1. POVERTYAND INEQUALITY INPERU,1997-2004 1.1 The first step towards a comprehensive analysis of poverty and inequality and their determinants, both in a static and a dynamic.context, i s to get the poverty and inequality facts straight. How many people are poor or extremely poor in Peru? Are inequality levels highcompared to other countries in the region and the world? How have poverty and inequality changed over time? 1.2 Inthis chapter we present poverty and inequality figures to answer these questions. The first section examines the latest available poverty and inequality numbers for Peru using data from the Encuesta Nacional de Hogares 200415 (ENAHO), while the second section discusses poverty and inequality trends for 1997-2004. Because the ENAHO has undergone significant methodological improvements over the years covered by this report, the second section also presents a brief discussion on methodological issues regarding the construction of time-consistent poverty and inequality figures using these data. 1.3 The main findingof the chapter can be summarized as follows: Poverty and inequality are high in Peru. In 2004 51.6 percent of the population was poor and 19.2 percent was extremely poor. Although international comparisons are difficult due to the use of different poverty lines across countries, Peru's poverty levels are below those of Ecuador and Colombia, but above those of Argentina and Brazil based on a US$2/day poverty line. Poverty and extreme poverty are higher in rural than in urban areas. They are also higher in the Sierra and the Selva than in the Costa. Inequality, measured by the Gini coefficient, stood at 0.43. This is lower than the regional averagebut highfor international standards. Poverty increased significantly as a consequenceof the 199819economic crisis, and has been slow to respond to the country's impressive economic growth in recent years. Between 2001 and 2004 the national poverty rate declined from 54 to 51.6. Incontrast extreme poverty and inequality remained constant during 1997-2000 and have improved significantly since. This improvement, however, results from declines inrural poverty only. POVERTY AND INEQUALITY UPDATE 1.4 For the purpose of this report we measure poverty and inequality on the basis of consumption, rather than income. This responds to a number of reasons, the most important of which being that consumption fluctuates less that income duringthe year and that people tend to report consumption more accurately than income. 1.5 We construct poverty measures using data from the Encuesta Nacional de Hogares (ENAHO), administered by the Instituto Nacionalde Estadisticas e Informacibn (INEI). Although the ENAHOi s not the only nationally representative household survey that contains information on income and consumption, it i s the survey that provides the most up-to-date information-2003, compared to 2000, the last year for which the Encuesta de Nacional Niveles de Vida (ENNIV), administered by Cuhto, is available. Poverty and Inequality in Peruin2004 1.6 Over half the population of Peru i s poor and about a fifth i s extremely poor, according to data from the ENAHO 2004/57. These figures, however, hide important differences across urban and rural areas, regions and departments. Poverty and extreme poverty are significantly lower in urban than in 7. Thesefigures measurepoverty andextremepovertyduringJanuary-December2004, usingdata collectedduring ENAHO 200314 (May 2003-April2004) andENAHO200415 (May2004-April2005). 17 rural areas, and so are their depth, measured by the poverty gap, and severity. Across regions poverty is lowest in Metropolitan Lima and highest in the Sierra, and the same can be said about extreme poverty, and the depth and severity of poverty (Table 1.1). Similarly, poverty and extreme poverty rates vary significantly across departments. Poverty and extreme poverty are highest in Huancavelica, at 88 and 74 percent respectively, and they are lowest inMadre de Dios, at 20 and 5 percent respectively (Figure 1.1). Table 1.1: Poverty indicatorsin2004 Poverty Extreme poverty Gini Headcount Gap Severity Headcount Gap Severity National 51.6 18.0 8.4 19.2 5.3 2.1 0.43 Area of residence Urban 40.3 12.4 5.3 7.9 1.8 0.7 0.39 Rural 72.5 28.3 14.1 40.3 11.7 4.8 0.32 Naturalregion UrbanCosta 37.1 10.6 4.5 6.2 1.4 0.5 0.34 Metropolitan Lima 36.6 10.4 4.1 3.4 0.6 0.2 0.40 Rural Costa 53.5 16.4 7.0 14.6 3.1 0.5 0.32 Sierra 67.7 27.2 13.9 36.5 10.9 4.5 0.39 Selva 59.5 19.7 8.8 26.4 6.3 2.2 0.36 Source: Authors' calculations usingdata from ENAHO2004 (INE1)-Annual sample coveringthe periodJanuary to December 2004. 7 The level of inequality in Peru i s lower than the regional average, but it is high by international standards (De Ferranti et alia, 2004). Moreover inequality, as poverty, varies significantly across urban and rural areas and across regions. Urban areas appear to be slightly more unequal than rural ones, mostly due to high inequality levels in Metropolitan Lima. Across regions, the Sierra and Metropolitan Lima exhibit the highest inequality levels (Table 1.1). Figure 1.1: Poverty and extreme poverty ratesvary across departments 1Pouerty B]Extremep e t t y Source: Authors' calculationsusingdata from ENAHO2004 (INE1)-Annual sample covering the periodJanuary 2004-December2004. 18 POVERTYANI) INEQUALITY TRENDS INPERU, 1997-2004 1.8 In this section we examine changes in poverty and inequality over the 1997-2004 period. In constructing poverty and inequality trends we must account for the fact that the ENAHOhas undergone a series of methodological improvements duringthe last few years that make it impossible to consistently compare raw average poverty figures over time. We briefly describe below the most significant methodological changes introduced in the survey since 1997 and discuss a series of empirical strategies that can be implemented inorder to ensure comparability of information across time and space. We then present and compare poverty and inequality trends constructed usingthese different strategies. Methodological changesinthe ENAHO and implications for poverty andinequalitycomparisons 1.9 The ENAHO has undergone a series of methodological improvements during the last few years. These changes include (i) the adoption of a new sampling framework to account for newly-developed areas, (ii) a revision of the way the poverty line is calculated and updated over time, and (iii) a modification inthe periodicity of the survey. As a result of these changes it i s impossible to compare raw poverty and inequality figures between 1997-2000 and 2001-04. 1.10 Although these changes have undoubtedly improve the quality and relevance of the survey they pose important empirical challenges for those interested in analyzing time trends before and after the changes were implemented. We pay attention here to two of these changes (points (i) and, particularly, (iii)), moreadetaileddiscussiononthemotivationforandthespecificsoftheseandotherchanges while i s presentedinAnnex 1. Change in the samplingframework 1.11 In 2001 the INEIreplaced the existing sampling framework of the ENAHO based on the 1993 Population Census with a new one based on the 1999 Pre-Census. Under the new framework a series of newly developed areas, mostly urban, were included in the ENAHO sample. Although the inclusion of these new areas contributed to improve the representativeness of the ENAHO, it meant that the raw poverty figures obtained from the survey were not comparable over time. The INEI then developed mechanismto re-weight the 2001-2003 samples in order to make them comparable to the 1997-2000 (i.e. the corrected sample weights eliminated the new areas and increased the relative weights of the old areas). These comparable figures, however, are somewhat sub-optimal because they do not take into account the standards of living of households in the newly developed areas-i.e. they are not nationally representative. Changes in thesurveyperiodicity 1.12 In 2003 the INEItransformed the ENAHO into a continuous household survey. This change impliedthat duringthe collection of ENAHO 200314 and ENAHO 200415 a small number of households were interviewed each month during May 2003 (2004)-April 2004 (2005) up to a total of 20,000 interviews in the year. In previous years a similar number of households were interviewed but only during the fourth quarter. As a result information on household income and consumption, on which poverty calculations are based, i s available on a monthly basis for May 2003 (2004)-April2004 (2005) in the ENAHO 200314 (200415) but only for October-December in the ENAHO 1997-2002. The question then arises as to what information to use in order to compare poverty and inequality between 2002 and 2003. 1.13 Several authors have documented that income and consumption levels vary significantly across quarters within the same years using household survey data for China and selected Eastern European 19 countries (World Bank, 2003c; Gibson et alia, 2001). If this is the case poverty and inequality comparisons across time must take seasonality into account. This can be done indifferent ways. We can compare poverty figures for a specific season or period of time onZy. In the case of the ENAHO this implies constructingpoverty and inequality trends basedon data for the fourth quarter only (Le. 1997.N- 2004.N) since data is only available for this quarter in 1997-2002. This i s the approach followed by Casas and Yamada (2005) andothers. 1.14 Alternatively we can try to control for seasonality by adjusting income and consumption levels across seasons before comparing them over time. A simple visual examination of the ENAHO 200314 and ENAHO 200415 data suggests that average income and consumption levels vary across quarters (Figure 1.2), and a statistical test rejects the null hypothesis of no seasonality for at least the national and urban extreme quarterly poverty rates (Casas and Yamada, 2005). This lack of conclusiveness has to be interpreted with caution, however, because the quarterly sample size i s small, and thus standard errors associated with quarterly estimates are large, and because the test is performed using only eight data points. As a result a more formal and conclusive examination of the existence of a seasonal pattern for Perurequires a longer time series. Figure 1.2: Poverty and extreme poverty vary by quarter, although not significantly May-June JulySepl. Ckt.-Dec, Jan,-March Apnl-June JulySept. 0Ct:Dec 2003 2003 2003 2004 2004 2004 2004 Source: Authors' calculationsusing data from ENAHO2003 and2004 (INEI). FromMay 2003 to December 2004. Proposed strategies 1.15 Given the changes discussed above and the comparability problems they pose we propose a series of alternative empirical strategies for the construction and comparison of poverty and inequality data for 1997-2004. A brief description of these strategies follows and the results that each one generates are presentedinthe next section. 1.16 We start by using the corrected sample weights constructed by the INEIto compare information between 1997-2000 and 2001-03. Poverty and inequality trends constructed in this way are consistent over time but, because they exclude those living in newly developed areas, have the serious disadvantage of not being nationally representative. For this reason we present these figures in order to give a fust approximation to the evolution of poverty and inequality during this period, but rely on the "uncorrected" 2001-2004 numbers for the rest of the analysis. 20 1.17 Using the "uncorrected" numbers we then present two sets of comparisons. We first examine figures basedon data from the fourth quarter only. Becauseinformation i s collectedduringthe same time period each year these figures are not affected by potential seasonal variation inincome and consumption. The scope of this comparison, however, i s limited because the small sample size of the ENAHO 2003.N and the ENAHO 2004.N only allows us to construct poverty and inequality figures at the national level and for urban and rural areas separately. In order to be able to examine trends at a more disaggregated level we needto use all data from the ENAHO200314 and the ENAHO200415. We do this inour second set of comparisons. 1.18 Finally we calculate 12-month moving averages for various poverty measures for the period May 2003-December 2004. The exercise then generates 9 data points that closely trace the evolution of poverty and inequality over a period of 20 months while correcting for potential seasonality since they all contain observations for all 12months. Figure 1.3: Poverty trends usingcomparabledata from ENAHO 1997-2003 -I 1097 m a lsDD ZMX) m, XXIZ +005 /-Poverty headcount - national Gini national - 1 I Note: Extremepoverty figures for 2000 are not entirely comparableto those of 1997-1999due to the impactof no-responsesinthe lower part of the incomedistribuhon. Source: Authors' calculationsusingdata from ENAHO 1997-2003 (INEI). Evolution of Poverty andInequality Trends, 1997-2004 1.19 In this section we examine the evolution of poverty and inequality in Peru during 1997-20048. We pay attention to national trends, trends in urban and rural areas and trends in different regions. We also examine changes in poverty and extreme poverty at the departmental level. Indoing so we use the four different empirical strategiesdiscussed above: (i) Trends basedon "corrected" sampling weights for ENAHO2001-03 (Figure 1.3), (ii) Trends based on data from the fourth quarter only (Columns 1to 7 in Tables 1.2a and 1.2b). This i s the methodology applied by Casa and Yamada (2005) and others. (iii) Trends based on data from the fourth quarter for 1997-2002 and on annual data for 2003 and 2004 (Columns 1 to 6 and column 8 in Tables 1.2a and 1.2b). This i s the methodology appliedby INEI. 8. A detailed povertyprofile is presentedand discussed inChapter 3 21 (iv) Trends based on 12-month moving averages using data from May 2003 to December 2004 (Figure 1.4). 1.20 Our mainresultsregarding strategies(i) (iii) summarized below andinTable 1.3: to are National trends. Results on national poverty and extreme poverty trends are fairly robust to the choice of an empirical strategy for the purpose of the comparison. All three data series suggest that poverty increased significantly between 1997 and 2000 as a consequence of the economic crisis, and remained stable afterwards. In contrast extreme poverty was unaffected by the crisis, and has declined significantly since 2001. When the full annual 2003 and 2004 samples are used for the comparisons a more positive picture appears regarding poverty changes between 2002 and 2004 with both poverty and extreme poverty coming down slightly. It is important to notice that 2001-2003 poverty and extreme poverty levels differ significant under strategy (i), on the one hand, and strategies (ii) (iii) the other. This is due to the introduction and on in the 2001-2003 samples of new developed areas. As we mentioned above these are mainly marginal urban areas so that households residing here are poorer than the average urban household and, as a result, their inclusion resulted in an overall increase in the poverty rate between 2000 and 2001. This increase, however, cannot be interpreted as a "real" increase inpoverty since the pool of householdsbeing compared in2000 and 2001 i s different. Urban and rural trends. During 1997-2000 poverty increased and extreme poverty remained stable in both urban and rural areas. During2001-2004 poverty rates stabilized and extreme poverty rates declined but only in rural areas. As was the case with national trends poverty and extreme poverty seem to decline the most when the full annual samples for 2003 and 2004 are used for the comparison, althoughthese changes are small. Regional trends. Again increases in poverty took place across the board during 1997-2000 while extreme poverty rates remained constant. Changes inpoverty and extreme poverty during 2001-2004 can only be calculated under strategy (iii) above. Using these data all regions except Metropolitan Lima saw some improvement in poverty rates. These changes were most significant in the Sierra. Duringthis same period extreme poverty increased inMetropolitan Limaand declined in the Sierra andthe Selva. Departmentaltrends. Changes inpoverty and extreme poverty at the department level can only be calculated for 2001-2004 since the ENAHO is only representative at the regional level before then. During this period poverty and extreme poverty declined in most departments. Particularly significant were the declines in Ayacucho, Apurimac, Cuzco y Cajamarca, which are among the poorest departments in the country (Figure 1.4). Although these changes are not statistically significant due to the small size of departmental samples, they suggest that recent economic growth has benefited the poorest among the population-something that we c o n f m inthe next chapter. 1.21 The main conclusions presented inTables 1.2a, 1.2b and 1.3 do not vary when 12-month moving averages are considered in an attempt to control for potential seasonality in consumption and poverty figures, as explained above (Figure 1.5). 22 Figure 1.4: Poverty andextreme poverty declined inmost departments between 2001and2003 Poverty headcount IvI&IMI iinasl apn- Mqq. A- caw- u r n -IF. NyLm 6- dvrn UL&"aC La-- urn iDI-0 wnek? M q - Q Y P W PUa ?No SUlUUlh i- T"Te6a iuyd 0 0 100 x ) O 300 PO0 u10 BOO 700 800 B O lo00 Percentageof the popuratlon Extreme Dovertv,headcount -- - m C I X _-cI i) m i L.LZ..-r 3- -1 .". _w.o Ylrl**D.= uac- PSC4 P J* =m E l U..C .r_. 7- .->a 3 0 .c3 23 0 3 c C PO3 50 3 63 C 73 0 83 c Percentage of the population I.2W4 02001- Source Authors' calculations using data from EUAHO 2001 IV and 2003/4 -Annual sample covenng the penod May 2003-ApnI 2004 (ISEI) 1.22 Inaddition recent improvements inextreme poverty and inequality have been accompanied by a reduction in the depth of poverty. The poverty gap, which measures the fraction of average per capita income that would have to be transferred to all poor individuals in order to bringthem to the poverty line, declined from 20.9 to 18.3 percent between 2001 and 2004 (not shown). This improvement, however, was entirely driven by a decline in rural poverty depth from 35.6 to 27.2 percent, while urban poverty depth remained increased slightly from 13.0 to 13.6 percent. 23 3 h m*d & P m ? c 3 -+'" n h tt '" +3? 3 *p? N, % h 3 h 09T + I? z 70 - 0 9 "r- I?: c, W b, 2 1.23 Finally consumption inequality increasedduring 1997-2000 and has declined since. This decline, however, has taken place only after 2002 as the improvements inextreme poverty inboth urban and rural areas have consolidated (Table 1.4). Table 1.3: Summary of changes inpoverty and extremepoverty under different strategies 1997-2000 2001-2004 2001-2004 1997-2000 2001-2004 2001-2004 (QIV) (QIV+annual) (QIV) (QIV+annual) Poverty headcount Extreme poverty headcount National + Area of residence Urban Rural ++ Natural region UrbanCosta Metropolitan Lima ++ Rural Costa + Sierra +I= Selva + Source: Authors' calculationusingdatafrom ENAHO 1997-2004(INEI). Figure 1.5: When controllingfor seasonality using12-monthmovingaverages, poverty remainsstable and extremepoverty declines 0.6 tn 0 0.5 VI e t 0.4 .-VI C > 0.3 0 /-Poverty --*Extreme poverty1 Source: Authors' calculationusingdatafromENAHO 1997-2004(INEI). 26 National MetropolitanLima Other urbanareas Ruralareas 1997 0.46 0.42 0.38 0.36 1998 0.47 0.44 0.37 0.36 1999 0.47 0.44 0.37 0.34 2000 A 0.41 0.37 0.35 0.3 1 2001 0.45 0.39 0.35 0.35 2002 0.47 0.42 0.37 0.34 2003 B(my-dic) 0.44 0.41 0.36 0.3 1 2004 BOm-dic) 0.43 0.40 0.35 0.32 CONCLUSIONS 1.24 In this chapter we have argued that poverty in Peru increased significantly during 1997-2000. Economic growth during 2001-04 halted this negative trend but did not generate significant improvements inthe overall poverty rate. It did however contribute to a significant reduction inextreme poverty, particularly in rural areas. The rest of this report i s devoted to understanding the relationship between economic development and poverty, bothfrom a macro and a microeconomicperspective. 27 28 2. ECONOMICGROWTHAND ITSIMPACTONPOVERTYAND INEQUALITY9 2.1 Economic growth i s a necessary condition for sustainable poverty reduction, but not a sufficient one. Poverty reduction is hard to attain in the absence of economic growth, since it would have to be achieved through substantial redistribution of income from the rich to the poor. However, economic growth does not automatically translate into poverty reduction unless i s accompanied by employment creation and better economic opportunities for the poor. 2.2 Inthis chapter we examine the relationship between economic growth and poverty reduction in Peru during 1997-2004, and analyze the factors that make this relationship weak or strong. The chapter i s structured as follows. The first section briefly describes the basic growth trends. The second section investigates the relationship between aggregate economic growth and poverty, as well as the distribution of economic growth across income levels. The third section takes a closer look at the nature of economic growth, paying attentionto its sectoral composition. The fourth section analyzes the relationshipbetween growth, employment creation and investment. And, finally, the fifth section discusses the potential for poverty reduction inthe future under various growth scenarios. 2.3 The mainfindings of the chapter can be summarized as follows: Economic growth in 2001-2004 in Peru has benefited those at the bottom of the income distribution more, especially in rural areas where it has translated into a decline in extreme poverty and in inequality. Growth, however, has not been sufficiently broad-based and has failed to bringthe overall poverty headcount down, particularly inurban areas. Economic growth i s positively correlated with poverty reduction, but the relationshipbetween growth and poverty is weaker in Peru than inthe average country. When compared to the rest of the world Peru needs to grow faster than the average country to lower poverty or even to prevent it from increasing. The weak relationship between (recent) growth and poverty i s due to two factors. First growth inper capita terms was slow and investment rates declined during 1997-2004 as a percentage of GDP. Second growth and investment were biased towards sectors with low capacity to generate employment, such as mining, or to generate income, such as agriculture. Inaddition to these trends, historically volatile growth rates have resulted into high levels of uncertainty and low levels of entrepreneurial confidence on the sustainability of economic growth. As a result recent economic growth has been slow to translate into higher employment and wage levels, particularly in those sectors that employ the poor and in urban areas. Investment and employment creation in urban areas have been low because there is a significant amount of excess capacity in the productive system and, especially, because employers expectations about future growth are uncertain. Consequently interventions aimed at preserving macroeconomic stability, strengthening growth and improving entrepreneurial expectations can go a long way in ensuring sustainable, poverty-reducing growth. Economic growth alone, however, can have only a limited impact on poverty and needs to be combined with economic and social interventions that specifically target the poor and help them take advantage of the economic opportunities generatedby growth. ECONOMICGROWTHTRENDSAND PATTERNS 2.4 The last 15 years have been witness to significant variation in the growth perfoknce of Peru. The economic progress of the early 1990s, when GDP and GDP per capita increased at an annual rate of 9. This chapter is basedon backgroundwork preparedby the report team and on existing work presentedinWorld Bank 2003 (PER), 2004 (Mining) and2004 (ICA). 29 5.8 and 3.9 percent respectively, was washed away by the 1998 economic crisis and by the economic stagnation that followed-GDP grew at 1.0 percent and GDP per capita actually declined by 0.9 percent between 1998 and 2000. Since 2001, however, the country's macroeconomic performance has been robust and vulnerabilities have been reduced. The aftermath of the economic crisis came to an end in 2001, and economic growth resumed at a strong 4.9 percent in2002, followed by 4.0 and 4.8 percent in 2003 and 2004 (Table 2.1). The most recent estimates place economic growth in 2005 at around 5.8 percent. 1991-1997 1998-2000 2001 2002 2003 2004 Annualgrowthrates (%) GDP 5.8 1.o 0.2 4.9 4.0 4.8 GDPper capita 3.9 -0.9 -1.4 3.3 2.2 2.5 Growth since 2001 has been accompanied by fiscal responsibility and declining fiscal deficits, indicating that this may be the beginning of a positive and sustainable trend. The issue of the sustainability of economic growth i s an important one, given Peru's past of volatile growth and successive economic ups and downs. Since 1965 Peru has never had more than 4 consecutive years with growth rates above 3.5 percent, while it has had 13 years of growth rates below 2.0 percent and 7 years with negative rates (Figure 2.1). Figure 2.1: Economic growthduring the last three decadeshas been insufficientto raise livingstandards over time I Source: Authors' calculationsusingdatafromthe Banco Centralde Reserva 2.6 We will argue in this chapter than Peru's traditional pattern of growth is not conducive to long- term per capita income gains (real income per capita today i s equivalent to that the early 1980s and early 1970s) and, as a result, poverty reduction. We will also discusspolicy measuresaimed at ensuring strong, sustainable and inclusive growth inthe future. But first we briefly review recent poverty and inequality trends. 30 ECONOMIC GROWTHAND POVERTY 2.7 The evidence discussed so far suggests that recent years have been witness to significant declines inextreme poverty and inequality-a development that should not be underestimated since it implies that living standards for those at the bottom of the distribution have improved. However the fact that these changesreflect mainly the experience of rural areas, together with the lack of improvement interms of the overall incidence of poverty, also indicates that there are still important challenges ahead. 2.8 The rest of this chapter in devoted to better understanding the relationship between economic growth and poverty inPeru in order to draw useful lessonsfor the future. We start by analyzingthe basic facts regarding this relationship and by examining the distribution of economic growth across different income groups during 1997-2003. We then take a closer look at the nature and composition of growth, and to the impact that growth has had on employment creation and investment. TheBasicFacts 2.9 Poverty appears to be responsive to GDP and GDP per capita growth over the medium-term in Peru. Using data from the Banco Central de Reserva (BCR), ENNIV and ENAHO, we analyze the relationship between economic growth, measured as changes in GDP and GDP per capita, and poverty reduction during 1985-2003. Periods of negative or weak economic growth (1985-91 and 1997-2000) are associated with significant increases in poverty, while periods of strong economic growth (1991-94 and 1994-97) are associatedwith important decreases inpoverty (Figure 2.2). Figure 2.2: Poverty is responsiveto GDP and GDPper capita growth -41 . ... " . . .. . . " . . -......... - . . .. . -. . Source: Authors' calculations usingdata from BCR, INEIand Cu6nto (2004). 2.10 However, when compared to other countries in the world, Peru needs to growth faster than the average country to achieve a given reduction in poverty. Following Kraay (2003), Loayza and Polastri (2004) compare the experience of Peru in 1985-97, 1997-99, 1999-2002 and 1997-2002 in terms of changes ingrowth and poverty to that of other countries. Economic growth and poverty reductionappear to be correlated for the average country in the world, represented by the solid line in Figure 2.3. 31 Moreover in this country poverty rates are stable in the absence of economic growth. In contrast Peru appears to need positive growth in order to prevent poverty from increasing, as illustratedby the fact that all data points corresponding to Peru are located above and to the right of the solid (regression) line-i.e. for any given change inpoverty, Peru needs to growth faster than the average country. Figure 2.3: Peru needsto grow faster than the average country in the world to reduce poverty 4 1 Average AnnualGrowthof GDP per capita I Note: For the sake of clarity all datapointshavebeenexcluded from the graphwith the exception of those correspondingto Peru. The solid line representsthe predicted relationshipbetween (logarithmic) changesinGDPper capita andchanges inpoverty (AGDP per capita = -0.0113 -0.8209 Apoverty). * Source: Kraay (2003)andauthors' calculations usingdata from INEIand National Account Statistics (BCR). 2.11 This difference between Peru and the international norm can be given two alternative explanations. On the one hand it i s possible for poverty to be more responsive to economic growth inthe long-term in Peru than in the average country, at the same time that the relationshipbetween growth and poverty i s very unstable inthe short-term inPeru(i.e. the regression line that fits the relationshipbetween growth and poverty inPeru i s steeper than that of the average country). On the other hand, it i s possible that poverty is as responsive to growth in Peru as in other countries, at the same time that there are country-specific factors that make it necessary for Peru to grow faster than the average country to maintain a stable poverty rate (i.e. the regression line that fits the relationship between growth and poverty inPeru has the same slope as that of the averagecountry, but it has a positive intercept). 2.12 In order to discriminate between these two alternative explanations, Loayza and Polastri (2004) examine the relationship between growth and poverty across different regions and departments in Peru. They find that neither i s poverty very sensitive to growth, nor has the relationship between both changed over time. Given this, they conclude that the second explanationprovided above appears to be the most sensibleto describe the case of Peru. 2.13 What are then the country-specific factors that caused these differences between Peru and the average country? Although a full examination of this question i s beyond the scope of this chapter, we volunteer the idea that high levels of income inequality may be such a factor based on recent researchon the issue of pro-poor growth that has shown that high initial levels of inequality can potentially hamper the impact of growth on poverty (Ravallion, 2004; Ravallion and Chen, 2003; Easterly, 2002). 32 2.14 Given that inequality i s high in Peru and that the country needs to grow faster than the average country to bring poverty down, the question then arises as to whether recent economic growth has failed to translate into lower poverty rates because it has been insufficient or because it has benefitedthe non- poor relatively more than the poor. To turnto the issueof the distribution of growth next. Distributionof economic growth 2.15 We analyze how economic growth has been distributed across income (or expenditure) groups during 1997-2000 using information on per capita expenditures. We consider two different periods, 1997-2000 and 2001-03, and compute changes in average per capita expenditure for each percentile of the per capita expenditure distribution and for each period. This calculation allows us to identify those groups (or percentiles) whose per capita expenditure grew more or less than the national average. We then plot these changesfor all percentiles inthe form of an expenditure growth incidence curvelo. 2.16 Changes in per capita expenditures varied across percentiles and across periods of time". Economic growth was not distributed uniformly across individuals and householdsduring 1997-2000 and 2000-03. In both periods, per capita expenditure growth was negative among the top percentiles and positive among the bottom percentiles. Moreover, because economic growth was negative during 1997- 2000 and positive during 2001-03, a larger share of the distribution experienced positive increasesinper capita expenditure in2001-03 than in 1997-2000 (Figure 2.4)12. Figure 2.4: Per capita expenditures grew amongthe oor and declined among the rich during 1997-2003 ISource: Authors' calculationsusingdatafrom ENAHO 1997-2003 (INEI). 2.17 These patterns are consistent with the changes in extreme poverty and poverty described above. Positive per capita expenditure growth among the bottom percentiles has contributed to the observed improvements in extreme poverty. However, because per capita expenditure has been relatively weaker inthe middle of the distribution, it has failed to translate into lower poverty rates. Inparticular the share 10. For a detailed methodological description on the construction of growth incidence curves, see Ravallion and Chen (2003). 11. Methodological changes in the ENAHO make it impossible to calculate growth incidence curves for 1997- 2003 without correcting the data (see Chapter 1 and Annex 1for a description of such changes). Some of the results discussed here are not robust to such corrections. We have decided to present the uncorrected 1997- 2000 and 200143 calculations here since they are the most representative of changes at the national level, together with adetaileddiscussionon the effect of corrections onthese resultsinAnnex 2. 12. Casas andYamada (2005) obtain very similar patternsfor 2001-04. 33 of total consumption attributed to the poorest 20 and 50 percent of the population grew from 3.8 to 4.7 and from 18.6 to 20.2 during2001-04, respectively (Casas and Yamada, 2005). 2.18 Changes inper capita expenditures also vary across urban and rural areas. During 1997-2000 all groups inurban areas experienced negative expenditure growth, compared to only those inthe top half of the distribution in rural areas. Similarly, expenditure growth in urban areas was almost zero in2001-03, while most groups experienced positive growth inrural areas (Figure 2.5). Figure 2.5: The impact of the economiccrisiswas strongerand the recoveryweaker inurbanthan in ruralareas. Urban Rural 199p2ooo =-r ... .......... ... , II 2 r m - m .............. . --1 . - ~ 1)- Source: Authors' calculations usingdata fromENAHO 1997- 103 (INEI). 2.19 Again these patterns are consistent with observed changes inrural and urban extreme poverty and poverty rates. Inrural areas positive per capita growth among the bottom percentiles in 1997-2000 and generalized growth across the whole distribution in2001-03, have translated into declines inextreme and non-extreme poverty rates. Inurban areas the situation i s somewhat more sober since decreases in per capita expenditure during 1997-2000 have not been compensated by increases in 2001-03, causing poverty ratesto stall or even increase, as has beenthe case inLima.13 2.20 In sum recent economic growth in Peru has benefited those at the bottom of the income distribution. As a result it has translated into a decline in extreme poverty and in inequality, but only in 13. Ibid. 34 rural areas. Growth, however, has not beensufficiently broad-based and as a result has failed to bringthe overallpoverty headcount down, particularly inurban areas. 2.21 Inthe next section we take a closer look at the nature of economic growth inPeruduringthis period in an attempt to understand why recent growth has not been more effective in terms of overall poverty reduction and, especially, interms of urbanpoverty reduction. A CLOSERLOOK THENATURE OFGROWTH AT 2.22 Inthis section we examine the nature of economic growth in Peru in 1997-2004 from two different perspectives: its speed and its sectoral composition, while inthe next we analyze the relationship between economic growth, investment and employment levels. The speedofgrowth 2.23 Economic growth was slow during 1997-2004. Gross Domestic Product (GDP) and GDP per capita growth were positive and significant between 2001 and 2004. However, due to the severe economic downturn of 1998-99, the balance for the 1997-2004 period i s 2.5 percent increase inGDP per capita. 2.24 Moreover, measures of income and consumption obtained from household surveys suggest that the recovery may have been weaker than implied by National Account data. Consumption per capita declined by 11.4 percent between 1997 and 2001, and by an additional 2.5 percent between 2001 and 2004, while income per capita fell by 1.7 percent in the first period and grew by 0.1 in the second one (Table 2.5). Given that it i s measured usinghousehold-level data, it is then no surprise that poverty did not decline in2001-03. Table 2.5: Income growth measuredusinghouseholddata hasbeenslower than GDPgrowth measuredusingthe NationalAccounts GDPper capita Consumptionper capita Income per capita (base 1994) (ENAHO) (ENAHO) 1997-IVI2001-IV -3.2% -11.4% -1.7% 2001-IVI2004-IV 1 5.7% -2.5% 0.1% Source:Herrera (2004) andFrancke (2005). 2.25 Although a full exploration of the causes of these discrepancies is beyond the scope of this report, several explanations could be considered, including (i) underreporting among the richest 1percent of the population in the ENAHO sample, which would bias consumption and income figures downwards; and (ii) inadequate imputation of consumption of durable goods and housing in the ENAHO, although this would only explain differences inconsumption levels (Francke, 2005) The compositionof growth 2.26 For economic growth to have a positive impact on poverty, it needs to generate employment and income for those who need it the most. Identifying what sectors or activities have acted as growth engines during the recent economic recovery, and examining to what extent growth in these sectors has the potential to generate a demand for the kind of labor the poor can provide i s then a first step towards understanding why the link between economic growth and poverty reduction has been weak in Peru in 2001-03. 35 2.27 Recent economic growth has been driven by external demand and by growth in the mining and agricultural sectors, and to a lesser extent by domestic demand in recent years. Exports are the fastest growing component of aggregate demand, having increased from 11.7 percent of GDP in 1997 to 17.3 percent in 2004 (Figure 2.6). At the same time, mining and agriculture have been the engines of growth from a sectoral point of view, with annual growth rates close to 7.9 and 3.8 percent respectively in 1997- 2004, and consequently have increased their share of GDP from 4.7 to 6.7 percent in the case of mining and from 8.3 to 8.9 percent in the case of agriculture (Figure 2.7). Growth in exports and growth in the miningand agricultural sectors are intimately connected since mineral and traditional and non-traditional agricultural products representmore than 60 percent of all exports (Table 2.6). Figure 2.6: Recent economic growth has beendriven by exports... urce: Authors' calculations usingdata fromCuinto (2004) andBCR. 36 Figure 2.7: ...and by growthinthe mining and the agricultural sectors Source: Authors' calculationsusingdatafrom Cufinto (2004), 1970 1980 1990 1999 2003 2004 Percentageof total exports Traditional exports 97 78 70 68 71 71 Mining 45 46 45 49 52 55 Petroleumand derivatives 1 20 8 4 7 5 Agriculture (traditional only) 15 6 5 5 2 3 Fishing 29 5 10 10 9 9 Non-traditional exports 3 22 30 32 29 28 Manufacturing 2 16 18 21 20 19 Other 1 6 12 11 9 9 2.28 The mining sector has a lower potential for employment creation than the agricultural sector for any given growth rate becauseit employs a small share of the labor force and has a low outputelasticity of employment. The mining sector employs 0.7 of the overall labor force, compared to 33.6 in the agricultural sector, and has an output-elasticity of employment of 0.4, compared to 0.8 inthe agricultural sector. This implies that a 1 percent increased in output will generate a 0.4 percent increase in employment in the mining sector and a 0.8 percent increase in the agricultural sector. Given the growth rates experienced in each sector since 1997, this means that employment in the mining sector has grown by 3.6 percent, or 0.02 percent of the labor force, while employment in the agricultural sector has grown by 4.6 percent, or 1.5 percent of the labor force (Table 2.7). 37 Table 2.7: Inrecent years growthhasconcentratedinsectors with different potential for employment generation Annual Annual Annual output Labor Output labor Output wage growth share elasticity of growth Meanwage elasticity of growth (%I (%I labor (%IO) wage (%I 1997-2004 1997-2004 1997-2004 MininglOil 7.62 0.66 0.43 3.58 10.61 0.65 12.4 Agriculture 3.47 33.63 0.84 4.60 1.28 0.46 -1.79 Electricityand Water 4.13 0.27 0.37 -3.79 8.23 0.74 7.11 Fishing 4.05 0.64 ' 0.59 5.33 4.44 0.81 11.61 Services 2.3 1 31.39 0.72 1.31 4.17 0.50 0.17 Manufacturing 2.15 9.25 0.66 1.35 4.44 0.53 7.43 Commerce 1.77 20.15 0.74 -0.05 2.66 0.51 -2.07 Construction I -1.09 4.00 0.64 -2.36 3.62 0.61 -0.48 Source: Authors' calculationsusingdatafromthe INEIand the BancoCentralde Reservafor 1997-2003. 2.29 In contrast wages are much more responsive to output growth in the mining than in the agricultural sector, althoughthis i s not likely to benefit the poor. Wages have growth at an annual rate of 12.4 percent in the mining sector, compared to -1.79 in the agricultural sector (Table 2.7). Rapid wage growth inthe miningsector, however, i s unlikely to have benefitedthe poor given the composition of the labor poolemployed inthe sector, which tends to be relatively skilled. 2.30 Given that mining and agriculture tend to be concentrated in rural areas, rapid growth in these two sectors, combined with sluggish growth in other sectors with higher concentration in urban areas, such as services or commerce, can help explain why rural areas appear to have done relatively better than urbanones interms of poverty reduction. 2.31 The potential for further improvements in rural poverty based on growth in the mining and agricultural sectors alone i s limited, however, in the absence of other policy interventions. We discuss this issue for the case of the miningsector in more detail inAnnex 4 and for the case of the agricultural sector inChapter 5. THE (WEAK) LINKSBETWEENGROWTH,EMPLOYMENTAND INVESTMENT 2.32 Employment and (labor) income generation provide the main links betweeneconomic growth and poverty reduction. We therefore turn our attention next to the relationship between growth, employment creation and investment, and the reasons why the connection between themhas been weak. 2.33 Even as economic growth has spreadfrom miningand agricultural into other sectors, employment generation and investment rates have remained low. According to data collected by the Ministry of Labor, employment in firms with more than 10 employees has recovered since 2000, although still remains below pre-crisis levels. This optimistic picture, however, blurs if we instead use employment information from the ENAHO. The overall employment rate has remained constant, or even declined slightly since 2000, at the same time that the composition of employment has shifted towards higher informality (Figure 2.8). The number of hours worked also increased slightly between 1997 and 2003, suggesting that labor needs may have been covered through more intensive use of already employed workers rather than through new hires-a point we return to in the next section. Similarly, although investment levels have increased in real terms since 2003, investment rates have been declining steadily as a percentageof GDP since 1998 (Figure 2.9). 38 Figure 2.8: Economicgrowth hasfailed to translate into higher employment ... -7 _I Employmernindex Employmentrate IilfOllll&ty Hcurswed (MOL) (ENAHO) Source: Authors' calcuationsusing data from the Ministry of Labor and the ENAHO 1997- 2003 (INEI). Figure2.9: ...or higherdomesticinvestment (Investment as a percentage of GDP) I --- 990 1991 1992 1993 1994 1995 1996 ?997 1998 1999 2000 2001 2002 2003 2004 ISource: Authors' calculationsusingdatafromCuiinto(2004). 2.34 In sumthe evidencediscussedabove suggests that recent economic growth was low over 1997- 2003, andthat, once growth started to accelerate in 2000-2001, it was biased andnon-inclusive in that it concentratedinsectors with low capacityto generateemployment andlabor income and failed overall to stimulate new (formal) hiringandinvestment. 2.35 These conclusionscanthenbe connectedwith our earlier discussiononthe distribution of growth andits impact onpoverty. Onthe onehand, faster growth inthe agricultural sector and, to a lesser extent, inthe miningsectorm y explain why ruralareas seemto havedonebetterthantherestof thecountry in terms of poverty andextremepoverty reduction since, as we will discuss inChapter 5, the ruralpoor rely heavily on agricultural activities. On the other hand, low growth and employment creation in other sectors, suchas servicesandcommerce,may explain why urbanpoverty rateshavefailed to decline since, 39 as we will discuss in Chapter 4, the urban poor, who rely extensively on their labor to make a living, tend to concentrate on these sectors. Finally the acceleration in employment creation that has taken place in recent months, particularly outside Lima, has not yet been enough to generate significant declines in poverty. 2.36 Employers make decisions about hiring and firing workers and about investment on the basis of their existing resources and their expectations about the future. We examine both factors here. Making do with existing resources 2.37 Although the use of installed capacity has increased since 2000, there is still a significant share that is currently underutilized. Approximately 30 percent of employers declared to be using60 percent or less of their business' installed capacity in 2003, down from 37 percent in 2000,14 while 38 percent declared to be using more than 80 percent, up from 19 in 2000 (Figure 2.10). The latest available data from the BCR indicates that average usage of installed capacity stood at 74 percent in October 2005, leaving still ample marginfor production increasesusingexisting capacity. Figure 2.10: Capacity useis increasing,but there is still a large share ofunusedinstalled capacity 194) 2mo m1 xm xv19 0Upto20% fa Between20% and40% =Between 40% and 60% 0Beween60%and80% faBetween80%and 1W% Note: Dataobtained from a survey to the managers of the 120largest firms in Peru as of 2002. The survey was administeredin2003, betweenJuly 15 andAugust 15. Source: Authors' calculationsusingdatafromCuhto (2004). 2.38 As long as there is enough idle capacity inthe system, employers do not have strong incentives to invest, unless this capacity has become obsolete or their expectations about the future suggest that more capacity will be needed. Hence the existing slack in terms of unused capacity may partly explain why investment rates havenot been very responsive to economic growth up to now. 2.39 Infact the evidence suggests that there exists a positive relationshipbetween economic growth and increasing capacity use on the one hand, and employers' willingness to invest on the other. Sixty percent of employers said they would be willing to invest in Peru in 2003, compared to 50 percent in 2000, while a still large 40 percent said they would be unwilling, compared to 50 percent in200015(Table 2.8). 14. Data obtainedfrom a survey conductedby CuBnto, a think-tank, among the managers of the largest 120firms inPeruas of2002. 15.. %id. 40 1999 2000 2001 2002 2003 al Willing to invest 63.9 48.1 55.0 42.5 58.2 Not willing to invest 29.5 47.1 38.7 56.0 38.2 No answer 6.6 4.8 6.3 1.5 3.6 2.40 Employers also appear reluctant to hire new workers. In 2003 only 15 percent of employers declared to be willing to hire new workers, compared to over 20 percent who declared to be willing to fire existing ones (Figure 2.11). Although these figures represent an improvement over those of 2000, when 10and 30 percent of employers said they would like to hire and fire workers respectively, they make it clear that economic growth has not been sufficient to stimulate employment creation in the absence of positive expectations about future economic opportunities. We then turn to the issue of expectations.I6 Figure 2.11: Willingness to hire new workers hasincreased, butis still low 1989 XEO awl Hm I IMFiringWSame =Hiring] Source: Authors' calculationsusingdatafrom Cuhto (2004). Animal SpiritsAfter All 2.41 Business decisions such as hiring new workers, installing new machinery or re-organizing production, to name a few, are not easily reversible since in the future it will be costly to fire these workers, dispose of the new capital or revert to the old methods if they new ones do not yield the desired results. For this reason employers will undertake such changes only if they believe that they will be profitable not only today buttomorrow. 2.42 Employers' perceptionabout the constraints faced by their own businesshas improved since 2000 as economic growth has become more robust. At that point, 40 percent of all employers identified lack of demandas their main problem, compared to 27 percent in2003.17 This change is likely to be the result of improved economic prospects associated with positive economic growth, and is in line with the improvements inemployers' willing to invest and hire described above. 16. Ibid 17. Ibid. 41 2.43 However, employers not only take into account recent growth, but also worry about its sustainability. Historically volatile growth rates, uncertainty in economic policy, lack of predictability of the legal and regulatory framework and the non-enforcement of laws and contracts were among the highest rated problems by Peruvian firms in 2002 (World Bank, 2003). Close to 79 percent of firms said that laws were fairly, highly or completely unpredictable, which i s very high even by international standards. This unpredictability, combined with implementation issues, such as corruption in public procurement contracts and the difficulty of resolving a dispute through the court system, generate a level of uncertainty that highly disrupts the environment for doing businesses in Peru. Over 70 percent of the firms in the sample identified uncertainty of economic policy as a ``major'' or "very severe" obstacle to operations and growth. 2.44 Uncertainty concerning the business environment inthe country has translated into a deterioration of employers' confidence in the sustainability of growth. Entrepreneurial confidence has declined steadily since 2001, after a short recovery following the 1998 economic crisis'* (Figure 2.12). The early termination of the Fujimori administration in 2000, amidst mounting evidence of flagrant public corruption and mismanagement, left the country inpolitical turmoil and economic pessimism. The sober administration of Paniagua's transitional government brought some relief, and for a few months there was optimism with the advent of the Toledo administration. Despite solid macroeconomic fundamentals, disappointment, however, soon ensued as the new government failed to provide the political stability the country needed as well as to eliminate uncertainty regarding existing legislation. In particular, the possibility of a reversal in the labor-market reforms of the 1990s looms large in investors' predictions about the future of the Peruvianeconomy. Figure 2.12: Employers' confidence hasdeclined since 2001 _ _ I Source: Apoyo, Opinihy Mercado, Survey of Entrepreneurs. 2.45 High uncertainty and low confidence can then affect growth by reducing incentives to invest in new machinery and external training, which in turn affect the adoption of technology and the improvement of productivity and hinder long-term growth and profitability. Regression analysis indicates that if uncertainty was lowered to "moderate" for all firms, 36 percent of firms could be expectedto make 18. Dataobtainedfrom a survey conductedby Apoyo, athink-tank, among employersinPeru. 42 investments which would help them to increase profits by at least 1.5 percent and another 35 percent of firms could increaseprofits by over 3 percent (World Bank, 2003b. 2.46 Insuminvestment andemployment creation have beenlow becausethere is a significant amount of excess capacity in the productive system and, especially, because employers expectations about future growth are uncertain. Consequently policy interventions that result inhigher expectations, such as those aimed at maintaining macroeconomic stability and improving the investment climate, can go a long way inensuring that economic growth is not only sustainablebutcontributes to poverty reduction. We outline some potential interventions aimed at these objectives below. Policy implications 2.47 As we pointed out at the beginning of this chapter, recent economic growth has been accompanied by fiscal discipline and stability. The Government of Peru should continue to work inthis direction in order to ensure the sustainability of growth. 2.48 Maintaining fiscal discipline, at the same time that fiscal space is created for poverty reduction will require, among others, fiscal reforms that create space for pro-poor spending and infrastructure investment and that improve the quality of spending, as well as interventions to improve Peru's investment climate and entrepreneurial expectations. The former could include, among other measures (i) the reformof the civil service, (ii) harmonizationand simplification of the tax system, (iii)reduction the a in public debt, (iv) an increase in fiscal transparency and accountability, especially in the context on ongoing decentralization, and (v) better targeting of social expenditure and reduced overlap among social programs. These interventions could be combined with others specifically designed to identify and implement effective investment projects, such as (i) the selection of projects on the basis of impact and sustainability, and (ii) the implementation of mechanisms that allow for the use of public resources to leverage privatemoneys (e.g. public-private partnerships and minimum-subsidy schemes). 2.49 Reducing uncertainty associated with economic policy, the legal and regulatory system and the judiciary systemwill require, among others": A clear articulation of and a more open consultation about the GOP's medium-term agenda. The GOP can reduce uncertainty about economic policy by more clearly articulating inmedium-term agenda and by institutingstronger modes of consultations with the private sector, particularly with micro, small and medium enterprises (SMEs), which are not well represented under current arrangements. Such a process would allow for the identificationof winners and losers from potential reforms prior to their implementation, as well as for the consideration of compensation mechanisms, hence facilitating consensus building. The Macroeconomic Framework prepared and updated annually by the MEF and recent efforts to introduce multi-annual budgeting constitute steps in the rightdirection. The implementationof simpler and more transparent administrative procedures. Peru should continue simplifying administrative procedures at the central government level and extend this effort to municipalities. Also, the process of public procurement should be revised at the central, regional andmunicipallevels to address the highdegreeof corruption inthe awardof public goods andservice contracts. The implementationof a clear and coherent legal framework in accordance with the medium- term agenda economic agenda. The GOP should ensure that existing legislation i s correctly aligned 19. The recent Investment Climate Assessment prepared recent by the World Bank (World Bank, 2004-ICA) analyzes Peru's investmentclimate and identifies priority areas for government action. We discuss the report's key policy recommendationsregarding uncertainty. 43 with its medium-term agenda. This required that previous legislation i s modified andlor repealed when necessary to reduce confusion over which section of the legislation has precedence and to facilitate implementation. It also requires that the ambiguity regarding the status of certain pieces of regulation (i.e. possible reversal inthe labor-market reforms of the 1990s) be resolved. An improvement in the functioning and image of the judiciary. Efforts should be intensified to improve court processes by reducing time, ensuring transparency in proceedings and, most of all, improving enforcement. Alternative dispute resolution mechanisms for the private sector should be revamped, so that arbitration results are truly binding and enforceable. Communications between the judiciary and the private sector should be improved, allowing for feedback on what i s needed and how it can be provided. 2.50 Preserving fiscal stability and reducing uncertainty will become even more important as the upcoming general elections approach, the pressure to trump fiscal discipline for populist spending mounts, and speculation about the next government and its policy agendaincreases. LOOKING AHEAD 2.51 The discussion in this chapter has been centered around the relationship between economic growth and poverty reduction, as well as around the factors that make this relationshipweaker or stronger. We want to conclude this discussion by presenting the results from a very simple simulation exercise that evaluates the potential impact that future growth, alone or accompanied by redistributive taxation, can have on poverty under different scenarios. 2.52 In a companion study to this report, Sosa-Escudero and Lucchetti (2004) model the micro- determinants of (labor) income using household data from ENAHO 1997-2003 and use the results to simulate the potential impact of various policy interventions on income, poverty and inequality. We focus here on the results from the simulation pertaining to different growth and taxation scenarios, while we postpone a detailed discussion of both the methodology they use and the results they obtained from their estimations to Chapters 3 and6. 2.53 All simulations are performed under a long-term horizon. Inparticular, the authors calculate changes in poverty and inequality rates between the present and the year 201520. Because results are projected more than ten years into the future, it i s important to remember that they should be interpreted as statistical explorations in a strong ceteris paribus context, and not as accurate depictions of the actual impact that different intervention may have on poverty and inequality. 2.54 Finally, because the authors use labor earnings rather than total income in their calculations, the poverty and inequality figures presented inthe paper do not coincide with those already discussed inthis chapter. To avoid confusion we have then standardized current levels of poverty and inequality to equal 100. How MuchPoverty ReductionCan Economic Growth Alone Buy? 2.55 We start by examining the impact of uniform growth in per-capita income at different rates, ranging from 1 to 10 percent. The exercise then assumes that everybody's income grows at the same annual rate and evaluates changes inpoverty and inequality betweenthe present and 2015. 20. The choice of 2015 as a reference responds to the fact that it is also the year when the Millennium DevelopmentGoals are expectedto be achieved 44 2.56 Income per capita growth has a moderate impact on poverty and a somewhat larger impact on extreme poverty under realistic assumptions about the speed of growth.'l Poverty would decline by 10 points if everybody's income grew at a constant rate of 1percent per year. This figure would increase to 30 and 45 points under the 3 and 5 percent per capita income growth scenarios respectively. Similarly extreme poverty would decline by 13points with annual per capita income growth of 1percent, and by 35 and 50 points with annual per capita income growth of 3 and 5 percent respectively (Table 2.9). Table 2.9: (Uniform) Growthhasa moderateimpacton poverty Present g=l% g=3% g=5% g=8% g=10% National Poverty headcount 100.0 90.8 72.7 55.8 36.3 27.2 Extreme poverty headcount 100.0 87.2 66.7 50.9 33.9 24.2 Gap 100.0 87.8 66.6 49.6 31.2 22.9 Severity 100.0 86.5 64.0 46.8 28.9 20.9 Urban Poverty headcount 100.0 86.7 62.2 42.1 20.9 13.5 Extreme poverty headcount 100.0 79.9 49.5 32.3 18.8 13.5 Gap 100.0 83.2 55.9 36.6 19.7 14.2 Severity 100.0 81.8 54.2 36.2 21.8 16.9 Rural Poverty headcount 100.0 95.6 85.0 71.9 54.4 43.1 Extreme poverty headcount 100.0 91.4 76.7 61.6 42.7 30.5 Gap 100.0 91.7 75.3 60.1 40.5 29.9 Severity 100.0 89.6 70.3 53.6 33.5 23.5 Inequality Gini - 100.0 100.0 100.0 100.0 100.0 100.0 ~~ Source: Sosa-EscuderoandLucchet 2.57 Urban areas would benefit relatively more fromgrowth inthe long-term than rural areas. Poverty would fall by 15 points in urban areas, compared to 5 points in rural areas under the assumption of 1- percent annual per capita income growth; the differences between both grow larger as the growth rate increases (Table 2.9). 2.58 Finally inequality i s not affected since standard inequality measures, such as the Gini coefficient, are invariant to scale shifts that more the distribution to the right by a constant multiplicative factor (Table 2.9). 2.59 We consider next the impact that growth inparticular sectors can have on poverty and inequality. For this purpose, we evaluate changes inpoverty and inequality under the assumption that wages in each sector grow at 5 percent per year, allowing wages to increase in one sector at a time?* Because the 21. The Ministry of Finance in Peru predicts annual GDP growth rates between 4 and 5 percent for 2004-07 (Marco Macroeconomico, MEF2003). Assuming that the population grows at 1-1.5 percent a year (following official estimations), the MEFpredictions would yield a GDP growth rateof 3.5-4 percent. 22. The Ministry of Finance in Peru provides predictions for output growth in different sectors during 2004-07 (Marco Macroeconomico, MEF, 2003): Agriculture: 2004-1.8, 20054.6, 2006-4.8, 2007-5.0; Manufacturing: 2004-4.7, 20054.8, 2006-4.8, 2007-5.1; Construction: 20044.1, 2005-5.7, 2006- 45 impact of each sector's growth i s evaluated independently, the changes in poverty generated by this exercise cannot be directly compared with those presented above and are meant exclusively as a tool to help our understanding of the role of the composition of growth. 2.60 The largest changes in poverty are associated with growth in the commerce and the agricultural sectors. Five percent annual growth in the income of those employed inthe commerce sector leads to an 11point decline inpoverty, while 5 percent growth in agricultural income leads to a 10 point decline. The impact of the commerce sector is larger in urban areas, and that of the agricultural sector i s more noticeable inrural areas due to differential patterns of employment across the urban and rural poor (Table 2.10). 2.61 In addition the largest changes in extreme poverty and inequality are associated with income growth in the agricultural sector. Extreme poverty and inequality would fall by 22 and 35 points respectively if income in the agricultural sector grew at 5 percent per year. This i s explained by the fact that most of the extreme poor live in rural areas and depend on agricultural activities for survival (Table 2.10). Table 2.10: The largestchanges inpoverty are associated with incomegrowth inthe commerceand agricultural sectors Present Agric! Ind. Manuf. Constr. Comm. Utilities Finances Gov. Levelin2015under 5% growth Naciona1 Poverty headcount 100.0 89.7 97.6 97.8 97.2 88.9 96.0 98.5 91.7 Extreme poverty headcount 100.0 78.0 98.1 97.8 96.8 89.9 97.0 99.3 93.2 Gap 100.0 80.1 97.6 97.5 97.0 89.6 96.6 98.7 93.1 Severity 100.0 73.8 97.7 97.7 97.3 90.8 97.3 99.1 94.2 Urban Poverty headcount 100.0 95.5 96.0 96.2 95.6 81.9 93.3 97.3 86.9 Extreme poverty headcount 100.0 86.9 96.2 95.5 94.1 78.6 92.9 98.3 86.4 Gap 100.0 91.1 95.9 95.2 94.7 80.6 93.3 97.4 87.6 Severity 100.0 88.7 96.0 95.2 94.7 80.9 93.9 98.0 88.4 Rural Poverty headcount 100.0 83.0 99.5 99.7 99.1 97.1 99.2 99.8 97.2 Extremepoverty headcount 100.0 72.9 99.2 99.1 98.4 96.5 99.4 99.9 97.2 Gap 100.0 71.2 99.0 99.3 98.8 97.0 99.4 99.9 97.6 Severity 100.0 64.2 98.8 99.3 99.0 97.1 99.5 99.9 97.9 Ineaualitv Gini - 100.0 94.1 100.3 102.0 100.3 102.8 102.2 103.4 102.8 Note: Includesmining. A Source: Sosa-Escuderoand Lucchetti (2004), 2.62 Insumthe largestimprovements inpoverty andincomedistributionsareassociatedwithgrowth inlabor-intensive sectors, and in sectors that employed a large fraction of the poor population, ineither urban or rural areas. However, the overall impact on poverty and inequality of growth alone i s limited. We explore then the impact of redistributive taxation, first on its own and then combined with economic growth. 5.9, 20074.0; Commerce: 2004-3.2, 20054.3, 2006-4.0, 20074.5; Services: 2004-3.5, 2005-4.0, 2006-4.0, 20074.5. Assuming that wages growth at the speed of output, a 5-percent growth rate falls on the upper-boundof the MEFpredictions. 46 The role for redistribution 2.63 To simulate the effect of redistributive taxation we combine a proportional tax with and egalitarian distribution of tax revenue. That is, we assume that everybody pays a fixed fraction (10,20 or 30 percent) of their income as taxes and receives a transfer from the government equal to WN, where R i s the total amount of money collectedthroughthe tax and Ni s the size of the population. 2.64 Redistributive taxation has a very small impact on poverty, even when high rates are considered, and a somewhat more important impact on extreme poverty and inequality. A 30 percent tax rate i s associated with a 1 point decline in poverty and 40 and 25 point decline in extreme poverty and inequality-equivalent to the changes associatedwith per capita income growth of 3 percent (Table 2.11). 2.65 These changes are the result of significant redistribution between urban areas, where poverty increases slightly as a result of taxation, and rural areas, where poverty and particularly extreme poverty decline as a result of taxation (Table 2.11). 2.66 The relative ineffectiveness of this pure and naive redistributive policy i s due to the fact that poverty rates are highand mean income low inPeru, hence any given tax extracts little in absolute terms from the richto be given to the poor. Present t=10% t=20% t=30% Level in2015 Naciona1 Poverty headcount 100.0 99.9 99.6 99.0 Extremepoverty headcount 100.0 89.3 76.2 57.6 Gap 100.0 88.1 76.2 64.4 Severity 100.0 77.4 58.0 41.8 Urban Poverty headcount 100.0 100.5 101.6 102.6 Extremepoverty headcount 100.0 85.3 67.9 47.3 Gap 100.0 91.9 83.9 75.9 Severity 100.0 83.5 68.6 55.5 Rural Poverty headcount 100.0 99.1 97.2 94.9 Extremepoverty headcount 100.0 91.6 81.1 63.6 Gap 100.0 85.O 70.0 55.2 Severity 100.0 73.5 51.2 33.0 Inequality Gini - 100.0 92.2 83.5 74.6 Source: Sosa-Escuderoand Lucchetti (2004). 2.67 Not surprisingly then combining redistributive taxation and growth produces larger changes in poverty and extreme poverty than taxation alone, since as time passes taxes are applied on a relatively richer population on average (Table 2.12). The same 30 percent tax considered above yields a 45 point decline in poverty and a 90 point decline in extreme poverty when paired with 5 percent per capita income growth. 47 ,Table2.12: Redistributive taxation, combined with growth, can have an impact on poverty andextreme poverty Present t=10% t=20% k30% Levelin2015 Nacional Poverty headcount 100.0 68.3 62.4 54.5 Extreme poverty headcount 100.0 52.1 34.0 9.3 Gap 100.0 51.2 36.6 23.3 Severity 100.0 40.4 22.8 11.0 Urban Poverty headcount 100.0 58.9 54.8 49.4 Extreme poverty headcount 100.0 35.3 23.5 11.0 Gap 100.0 45.6 35.6 26.2 Severity 100.0 38.4 25.5 15.6 Rural Poverty headcount 100.0 79.3 71.3 60.5 Extreme poverty headcount 100.0 61.9 40.2 8.3 Gap 100.0 55.8 37.5 20.9 Severity 100.0 41.7 21.1 8.0 Inequality-Gini 100.0 92.3 83.8 74.9 2.69 The same cannot be said, however, about the combination of 5 percent growth with low taxation (10 percent). This can be explained by the existence of a large mass of individuals around the poverty line. When this i s the case redistributive policies move part of this mass to the left, or below the poverty line, hence increasing the headcount rate. This implies then that for redistributive taxation to have a higher impact without resorting to high average taxation rates, it would have to become more progressive-i.e. to apply higher rates to higher income levels andlor to exempt low income levels from paying taxes. CONCLUSIONS 2.70 We have analyzed in this chapter the relationship between economic growth and poverty reduction in Peru during 1997-2003.We have shown that, although growth was beenpro-poor, especially inrural areas, it was beeninsufficientto bringpoverty down becauseitfailed to generateemployment and investment. 2.71 We have also explored the potential impact that economic growth can have on poverty in the future under several scenarios, to conclude that under realistic assumptions growth alone or combined with (too simple) redistributive taxation can only have a limited impact. 2.72 What can be done then to enhance the effect of economic growth? The case of Chile i s illustrative inthis regard (Box 2.1). Chile, like Peru, experienced several years of strong growth with no poverty reduction, until President Aylwin introduced a package of more progressive interventions, economic and social, to accompany existing growth policies. These interventions had the objective of helping the poor benefit from the opportunities generated by growth. Peru i s already experimenting with 48 redistributive mechanisms based on contributions through the Mining Canon and others that channel resources from the mining industry to the poorest regions and local governments and which are expected to have a positive impact onpoverty. These are welcome steps, but more needs to be done. Box 2.1: Economic growth and poverty reduction: The case of Chile The deep financial crisis that hit Chile in 1982-83 led the economy to a recession that resulted in a 16 percent decline of GDP and 30 percent unemployment. Many banks and businesses failed. Almost half of the country's population fell below the poverty line, and as of 1983, 30 percent qualified as extreme poor. After the implementation of considerable changes to Pinochet's orthodox neoliberal model in 1985, including measures to encourage exports and make available debt relief to businesses that had been hardly hit by the recession, the economy recuperated and began a path of strong, steady economic growth, with GDP growth averaging 7 percent per year between 1985 and 1990. The immediate improvements in macroeconomic indicators, including a surge in employment generationand real wages, did not translate into a reduction of the country's high poverty levels right away, and as of 1987,5 millionpeople(over 45 percent of the population) were still poor. Ittook a number of years of substantial growth and macroeconomic stability for poverty to start to recede, and finally, towards the late 1980s, poverty levels dropped6 percent, to 39 percent. Aware of the very high level of poverty that still affected the country, the government of Patricio Aylwin, first electedpresidentafter the returnto democracy in 1990, left the central economic policies of the military regime unchanged while introducing a progressive twist. An active pro-growth agenda was combined with aggressive social policies focusedon equity, poverty reduction, and humancapital development. Substantialincreases in social investments,rangingfrom targetedemploymentprogramsto reforms ineducationand health, were financedby a tax reform program approved in 1990 that increased tax collection by 3 percent of GDP in subsequent years. The combination of a sound and stable macro framework with the government's active role on the social front was continued by the democraticgovernmentsthat followed Aylwin's presidency. The results were outstanding: poverty reduction accelerated, and by 1998, the level of population living in poverty had declined to 20 percent, less than half of that in 1987. The "growth with equity'' approach proved extremely effective, benefiting vulnerable groups across the board, including the elderly, poorly educated, and male and female household heads in both rural and urban areas. Poverty continued to fall after 1998. Despite being hit by a recessiontriggered by the international financial turmoil in the late 1990s, Chile was able to sustaincountercyclical social policies thanks to public savings accumulated in the good years, which made it possible to expand expenditures in social programs without endangeringeconomic stability, budgetbalance, or increasingcountry risk. The crucial role that Chile's sound macroeconomic framework and strong growth since the mid-1980s played in poverty reduction is widely accepted. However, the extent to which the country's social agenda was effective incomplementing growth to achievepoverty alleviation deserves further attention. A study conductedby Meller (1999) analyzed the factors driving Chile's outstanding poverty reduction. Using Cowan and De Gregorio (1996)'s "growth efficiency for poverty reduction" concept, Meller compares the poverty-to-GDP elasticity for four time periods between 1987 and 1996. H e found that this elasticity is 50 percent lower during the 87-90 period of military rule than in the three periods with democratic governments (90-92, 92-94, and 94-96). In other words, whereas a 1% GDP increase between 1987 and 1990 reduced the percentage level of poor by 0.75%, the same increase inGDP caused a 1.15% poverty reduction during the 1990s. Meller concludes that high growth rates are necessary but not sufficient for poverty reduction, and that the social policies implemented by democratic governmentsconstituted an important element for poverty reduction. Meller estimates that, due to social programs implemented during the 1990s, the total number of poor dropped by an additional 675,000 (of a total of 1,677,000 poor), which suggeststhat while 60 percentof Chile's poverty reductionbetween 1990and 1996canbe attributed to economic growth, socialpolicies explain the remaining 40 percent. 2.73 Chapters 4, 5 and 6 inthis report will identify barriers that prevent the poor from benefiting from growth and propose policies to help them overcome these barriers. Chapters 4 and 5 will identify economic opportunities for the urban and rural poor respectively and discuss policies that contribute to create these opportunities Chapter 6 will pay attention to the challenges the poor faced, particularly in terms of risk and vulnerability and terns of exclusion for public services and institutions, and provide policy recommendations to help the poor surpass some of these challenges. 49 50 3. NATURE,DISTRIBUTIONAND EVOLUTIONOFP0VERTY2j 3.1 The first step in the analysis of the relationship between poverty and public policy i s understanding how poverty is distributed, both geographically and across households of different characteristics, and how (ifat all) this has changed over time. Inthe chapter we examine the determinants of poverty inboth a static and a dynamic framework, as well as the correlationbetween monetary poverty and alternative welfare measures. We also explore the implications that our discussion on the nature of poverty have when thinkingabout targeting social programs. 3.2 The chapter is structured as follows. The first section presents an updated poverty profile for 2003 and discusses the determinants and correlates of poverty, as well as the role of geographic as a determinant of persistent differences in poverty across regions. The second section presents a more detailed discussion on the micro-determinants of poverty dynamics. It identifies the factors behind changes in the.distribution of income and, hence, in poverty and inequality, and examines some of the forces driving exit from and entry into poverty among rural and urban households. The third section examines the relationship between monetary poverty and other measures of welfare, such as the Unsatisfied Basic Needs index and the caloric deficit. Finally the fourth section discusses the appropriateness of various targeting tools for social programs given these results. 3.3 The mainfindings of the chapter can be summarized as follows: There are significant and persistent differences between poor and non-poor households. The poor tend to live in larger households, be less educated, and have lower access to basic services than their non-poor counterparts. They are also more likely to be unemployedor informally employed in urban areas, and more likely to work inthe agricultural sector inrural areas. Finally households headedby an individual who speaks an indigenous language are more likely to be poor than other households, even after controlling for other observable differences. Inaddition there are significant differences in poverty across areas and regions. Poverty is higher deeper in rural than in urban areas and in the Sierra and the Selva than in the Costa. These differences can be explained almost entirely by household characteristics, access to basic services and roadinfrastructure rather than by geography. Poverty i s a dynamic and multidimensional phenomenon and, as a consequence, traditional static poverty profiles based on single poverty measures only present a limited picture of the actual reality of poor households. Flows in and out of poverty are large. In any given year approximately 25 percent of all households make a transition between poverty and non-poverty. These transitions are not random but rather dependon both household characteristics andendowments and onthe occurrence of shocks. Inaddition changes inhousehold characteristics and endowments, and intheir returns, drive most of the observed variation in the distribution of income over time and across space and, consequently, most of the variation in poverty trends and poverty differences across geographic regions. Recent increases in poverty have been the result of changes in the returns to education and changes in the compositionof employment. Monetarypoverty, on which the above results are based, i s only one of the possible ways inwhich we can measure welfare. The Unsatisfied Basic Needs index and the caloric deficit are two other measures currently used inPeru. Although all three measures of poverty are positively correlated and expectedto evolve in similar ways over the long-term, they do not produce the same poverty profiles inthe short-term. 23. This chapter is basedon background work preparedby the report team, as well as on existing work by Escobal and Torero (2003) and Herrera andRoubaud(2002). 51 a The fact that who is poor andwho is not changesover time, together with the notion that poverty can be defined in various different ways has important implications for the design and selection of targetingtools for social programs, an issuecurrently under discussion inPeru. POVERTY PROFILE Livingconditions andcharacteristicsof the poor 3.4 Although the poor everywhere live in marginal circumstances with respect to housing conditions and access to employment and basic services, living conditions vary widely across the country. We briefly discuss some of the differences and commonalities that exist between the poor and the non-poor, and across regions below. Our focus is on demographic characteristics, access to services and employment, since these fit best with the scope of the report. 3.5 Income and expenditure levels: Poor households have significantly lower levels of income and expenditure than non-poor ones, and these differences are larger in urban areas, particularly in Metropolitan Lima, than inrural ones (Table 3.1). Table 3.1: The demographic characteristics of the poor vary across urbanand ruralareas Urban (excludingLima) Metropolitan Lima Rural rota1 NonPoor Poor Total NonPoor Poor Total NonPoor Poor Per capita Total Income 396 530 166*** 818 1037 237 *** 172 304 100 *** Per capita Expenditures 340 450 149*** 577 719 199 *** 168 285 104 *** Size of Household 4.4 3.8 5.4*** 4.4 4.0 5.6 *** 4.4 3.3 5.0 *** Dependency rate A 29.0 27.7 31.2*** 28.2 27.9 29.1 39.0 37.6 39.7 *** Age of head Lessthan 25 years old 3.5 3.7 3.0 2.2 2.3 1.8 4.0 4.6 3.7 25 to 55 years old 65.3 62.1 70.9*** 64.7 60.3 76.3 *** 61.7 54.5 65.6 *** More than 55 years old 31.2 34.1 26.1*** 33.1 37.3 21.9 *** 34.3 40.9 30.7 *** Gender (% Female Head) 23.4 24.6 21.5 24.4 26.4 19.2 *** 16.0 19.2 14.2 *** Maritalstatus ofhead Cohabiting 23.3 17.7 33.0 *** 20.2 14.5 35.4 *** 28.8 23.6 31.6 *** Married 47.6 49.9 43.7 *** 49.5 52.1 42.6 *** 47.3 43.1 49.6 *** WidowlDivorced 22.5 24.3 19.4 *** 23.8 25.O 20.7 ** 19.4 25.1 16.4 *** Single 6.6 8.1 4.0 *** 6.5 8.4 1.3 *** 4.5 8.2 2.5 *** Educationof head NoEducation 6.2 4.6 9.0 *** 2.3 2.0 3.2 * 16.5 12.8 18.5 *** Primary 32.9 26.7 44.0 *** 22.3 18.3 33.4 *** 57.4 50.5 61.1 *** HighSchool 37.6 37.3 38.2 * 45.1 42.6 52.2 *** 22.0 28.0 18.8 *** Mire*** thanHighSchool *** *** ** 23.3 31.5 8.7 30.3 37.1 11.3 4.1 8.7 1.6 *** Source: Authors' calculationssuingdatafrom ENAHO2003 (INEI) 3.6 Household characteristics: There exists a clear difference between poor and non-poor householdsinterms of their size and composition. Poor households are significantly larger than non-poor 52 ones and, as a consequence, tend to have higher dependency ratios (i.e. number of dependentsper income earner). Households with certain types of people are also more likely to be poor or non-poor irrespective of household size. For instance, households headed by an elderly person are less likely to be poor. Female-households, however, do not appear to be more likely to be poor than male-headed households (Table 3.1). 3.7 Ethnicity: Defining who i s and who is not indigenous i s a complicatedundertaking inPeru, since there does not seem to be a unique criterion that distinguishes the indigenous and non-indigenous populations. For the purpose of this report, we will classify people as indigenous if they speak one of more indigenous languages. However, because this definition may not be entirely. satisfactory, we also present a brief discussion on the issue of ethnicity identification inBox 3.1. 3.8 Households headed by an indigenous person exhibited significantly higher poverty rates than other households, even after controlling for other observable characteristics. Differences between indigenous and non-indigenous households are more marked inurbanthan inrural areas (Box 3.1). 3.9 Education: The education level of the household head i s strongly related to the household poverty status. The average poor household head has primary education compared to secondary education among non-poor household heads (Table 3.1). 3.10 Housing conditions: The poor are less likely to own a title on their house or to own a house at all, especially in urban areasx. They are also more likely to suffer from overcrowding and to reside in inadequate housing (Table 3.2). Urban (excludingLima) MetropolitanLima Rural Total Non-Poor Poor Total Non-Poor Poor Total Non-Poor Poor Ownership Owner 68.3 68.7 67.6 *** 70.1 70.7 68.5 85.9 81.2 88.4 *** Owner (no title) 5.8 4.4 8.2 *** 3.6 1.8 8.5 *** 1.1 1.5 0.9 *** Rent 8.3 9.5 6.1 *** 10.6 11.5 8.0 ** 1.7 3.4 0.9 *** Other 17.7 17.4 18.1 *** 15.7 16.0 15.1 11.3 13.9 9.9 *** Housing conditions Number ofpersons perroom 1.6 1.2 2.2 *** 1.5 1.2 2.3 *** 2.2 1.4 2.6 *** Inadequatehousing 7.9 4.4 14.0 *** 7.4 3.9 16.7 *** 14.7 14.3 14.9 Overcrowdedhousing 7.8 3.1 16.0 *** 6.0 2.2 16.1 *** 18.5 6.1 25.2 *** Access to Public Services Water 70.6 77.5 58.7 *** 82.9 88.2 68.7 *** 32.8 40.2 28.7 *** Electricity 91.2 95.8 83.2 *** 98.0 98.5 96.9 ** 31.8 45.9 24.1 *** Sanitary services 61.1 72.3 41.6 *** 81.4 87.0 66.3 *** 5.7 10.8 2.8 *** 3.11 Access to services: Access to water, electricity and sanitation is lower among poor households than among non-poor ones irrespectiveof their area of residence, althoughdifferences inaccess tend to be more acute inrural areas (Table 3.2). 24. For a detaileddiscussion on the impact of land titling on urban poverty, see Cantuarias and Delgado (2004), Cuhto (2001), Field(2002), FieldandTorero (2004),andWorld Bank (1998). 53 Box3.1: Measuring the size and living standardsof the indigenouspopulation. Measuring the size of the indigenous population. Peru has a large and diverse indigenous population. The 2001 EncuestaNacional de Hogares(ENAHO) includesthe following questionsregarding ethnicity: What is your native language? What languagedo you use most frequently? What racelethnic group do youbelong to? What native language didldo your grandparentsiparents use? Because these questionsrespond to different concepts of ethnicity, ranging from language-based to culture-based,they produce different estimates of the size of the indigenous population-which vary from 47.7 percent of all households accordingto the broadest definition to 25.4 percent according to the narrowest, comparedto the widely accepted figure of 30.0 percentfrom the 1994PopulationCensus. For the purposeof this report, we will use definition(1) below. Table B3.1.1. The size of the indigenouspopulation varies with the definition of indigenous By area All Lima Other urban Rural % of all households Non-Spanishnative language(1) 33.7 3.1 11.4 19.2 Self-identification as indigenous(2) 42.6 5.0 16.3 21.3 (1) or (2) 45.2 5.4 17.3 22.5 Most frequently uses non-Spanishlanguage(3) 25.4 2.0 7.1 16.3 Headof householdgrandparentslparents'native 47.7 6.7 17.8 23.2 languagewas non-Spanish(4) Moreover, the term `indigenous' hides significant heterogeneity-although households headed by Quechua speakers represent a majority (75 percent of all indigenous households), and additional 15 percent of households are headedby Aymara speakers(12 percent) and Amazon indigenous(3 percent)respectively. Measuring living standards among the indigenous population. Indigenous households exhibit higher rates of poverty and extreme poverty than non-indigenoushouseholds, although important differences exist betweenurban and rural areas. In particular, while poverty rates are lower in urban than in rural areas for all households, the relative differences betweenindigenous and non-indigenoushouseholds are more markedinthe former, especially inLima, than inthe latter. Table B3.1.2: The incidenceof poverty is higher among indigenoushouseholds All Indigenous Non-indigenous Poverty headcount 46.8 63.8 42.0 Extremepoverty headcount 20.1 35.3 16.6 Poverty gap (FGT1) 17.4 26.2 13.2 Severity (FGT2) 8.7 14.1 6.2 Table B3.1.3: Poverty rates are lower and differences between indigenousand non-indigenoushouseholds higher inurban areas. All Indigenous Non-indigenous National II46.8 1I 63.8 39.6 Lima 37.2 Other urban 52.3 38.0 Rural 1 72.2 78.6 65.8 Source: AuthorscalculationsusingdatafromENAHO(2001) Although the differences between indigenous and non-indigenous households result partly from differences in the 54 endowmentsof both groups, indigenoushouseholdare still 11percentmore likely to be poor thanotherwisesimilar non- indigenous households once these are taken into account. Once we disaggregateby area of residence, only in Lima do endowments seem to fully explain the difference in poverty rates between the two groups, which, given that relative differences where largest here, suggest that differences in endowments between indigenous and non-indigenous households are more markedinthe capital than elsewhere. Table B3.1.4: Indigenous ethnicity is positively correlated with poverty, even after taking into account differences inendowments. National 0.113*` Lima Other urban Rural (6.17) Source: Trivelli (2004). Numberscorrespondto the coefficientof an indigenousethnicity indicatorvariableinalogit modelfor the determinants of poverty(poor= 1,non-poor=0). The modelincludes informationon areaof residence,householdsize andcomposition,demographic characteristics of the householdheadandother members,andemployment characteristics of the householdhead. **Differentfrom0 at the 1percent significancelevel. Interestingly, despite the differences in monetary poverty, only 22 percent of all indigenous households responded positively to the question"Do you consider your householdto be poor?', comparedto 23 percent of all non-indigenous households, thus suggestingthat monetarypoverty may not captureindigenoushouseholds' welfare appropriately. Source: Trivelli (2004). 3.12 Employment: Employment is the main source of income for the large majority of households, and thus one of the main determinants of poverty. Although there are no significant differences in household head employment rates between poor and non-poor households, the percentage of household members that i s employed i s larger inthe latter. 3.13 The type and sector of employment also have an impact onpoverty. The informal sector provides employment to a higher share of the poor than the non-poor, especially inLima. Self-employment and blue-collar employment are positively correlated with poverty and so i s employment in agricultural activities. Correlates of poverty 3.14 Inthis section we useconditional probit models, inwhich householdpoverty statusis modeled as a function of "exogenous" variables, to examine more formally the relationship between poverty and some of the variables discussed above. This approach helps us identify variable that are directly correlated with poverty, once the effect of other variables is controlled for. We estimate these models separately by area and region to account for variation in the living conditions of poor and non-poor households. 3.15 We briefly discuss here the results from these estimations, identifying common correlates across urban and rural areas, while the full-model estimates are provided inthe Statistical Annex. 55 3.16 Common urban factors: Urban poverty across regions appear to be associated with (i) larger household sizes and a larger number of dependents, (ii) low levels of education of the household head, (iii) accesstoservices,(iv)unemploymentofthehouseholdheadandorlowlaborforceparticipation low levels among other household members, and (v) employment in the informal sector of the household head. 3.17 Common rural factors: Rural poverty across regions appears to be associated with (i) larger household sizes and a larger number of dependents, (ii) low levels of education of the household head, (iii) accesstoservices, and(iv) employmentinagricultureofthehouseholdhead,althoughthisis low only significant inthe Selva and the Sierra. Table 3.3: Poverty Profile Urban (excluding Lima) Metropolitan Lima Rural rota1NonPoor Poor TotalNonPoor Poor TotalNon Poor Poor Employment Status of head Employed 34.2 94.0 94.4 93.1 93.3 92.7 99.4 99.3 99.5 Unemployed 5.6 5.9 5.6 6.8 6.6 7.2 0.5 0.6 0.4 Not inthe Labor Force 13.5 15.9 9.2 *** 20.2 23.2 12.0 *** 4.1 6.6 2.8 *** Labor Markets IncomeearnerslAdults(lo+ years old) 67.6 72.4 59.2 *** 64.1 67.9 53.8 *** 61.6 72.0 55.9 *** Informal Sector A Informally employedlAdults (lo+ years 016 44.0 41.8 47.9 *** 33.7 31.7 39.0 *** 70.0 67.1 71.5 *** Informal employmentofhead 71.5 63.8 84.1 *** 60.7 53.7 77.9 *** 93.3 86.3 96.9 *** Employment Category of head Employer 10.0 12.2 6.5 *** 9.49 11.0 5.9 *** 11.0 16.3 8.3 *** White collar employee 20.0 27.4 7.9 *** 27.8 34.6 11.12*** 3.7 8.2 1.3 *** Bluecollar employee 22.0 17.9 28.7 *** 23.3 18.8 34.6 *** 12.4 15.6 10.7*** Self-employed 43.6 37.8 53.2 *** 35.6 31.9 44.7 *** 71.4 57.5 78.6 *** Other 4.4 4.7 3.8 3.8 3.8 3.7 1.6 2.4 1.1 *** Economic Sector of head Primary Activities 26.6 19.5 38.3 *** 2.2 1.9 3.0 ** 85.3 75.4 90.5 *** Public Sector 5.7 7.0 3.6 *** 4.8 5.5 2.9 ** 1.7 2.9 1.0 *** Construction 5.9 4.2 8.6 *** 10.8 8.9 15.3 *** 1.8 2.0 1.6 Manufactures 10.6 10.6 10.6 17.5 17.0 18.7 2.3 2.8 2.0 *** Services I Utilities 51.3 58.8 39.0 *** 64.7 66.6 60.1 9.0 17.0 4.9 *** Notes:***(**)("Difference betweennon-poor andpoori s significantly different fromzero at the 1(5) (10) percentlevel. 3.18 The evidence presented in this section then shows that there exist significant and persistent differences between poor and non-poor households across areas and regions. In particular there i s a heavy concentration of poor households in the most geographically adverse regions, the rural Sierra and the Selva (see Chapter 1). Interestingly, as we discussed below, these differences across regions can be almost entirely explained by differences across household living ineachregion. 56 The role of householdcharacteristics versus the role of geography 3.19 Inexplainingregionaldifferences inpovertyratesonecouldconsider twodifferent views (Table 3.4). The first view postulates that differences arise from the spatial concentration of individuals with poor characteristics and endowments. Under this view identical individuals should have the same probability of being poor irrespective of where they live. The second view assigns a more causal role to geography, so that poor households living in a well-endowed area are more likely to live poverty than those who live inother areas. 3.20 Whether we believe in the first or the second view has important implications for policy making. Inthe world described by the first view investments inindividualcharacteristics, suchas education, are all that i s needed to improve the living conditions of the poor. In contrast under the second view the returns to these characteristics may be a function of the local environment and, consequently, such interventions may fail to produce the desired results. Table 3.4: Regionaldifferencesinper capita expenditurescan be mostlyexplainedby differences inhouseholdaccessto private assetsandinfrastructure Sierra-Costa Selva-Sierra Total (log) difference -0.217 -0.167 Explained -0.241 -0.244 Residual 0.024 0.077 Difference explained by: Householdcharacteristics -0.185** -0.258*' Household size 0.031** -0.064** Years of schooling (household head - HH) -0.061** -0.065** Years of schooling (other members) -0.069** -0.102** Potentiallabor experience (HH) -0.013** -0.024** Gender (HH) 0*000 -0.001 Number of migrants -0.009* -0.005* Savings 0.002** 0.000 Value of durable goods -0.003 0.004 Infrastructure -0.024* -0.064* Schools intown (per capita) 0.024 0.023 Medical centers intown (per capita) 0.010 0.009 UnsatisfiedBasic Needs Index (UBN) -0.058** -0.095** Geography -0.163 0.031 Altitude -0.036 -0.004 Temperature -0.235** 0.173'* Temperature squared 0.117 -0.121 Igneous rocks 0.015* -0.004* Sedimentary rocks -0.004 -0.009 Soil depth -0.022 -0.005 Location 0.050 0.039 Urbanization 0.055 0.038 Distance to provincial capital 0.005 0.001 Geography * location 0.081* 0.007* Urbanization altitude * 0.081* 0.007* Source: EscobalandTorero (2003). 3.21 Escobal andTorero (2003) use Census data and LSMS data to explore whether geography has an effect on living standards once observable household and individual Characteristicshave been controlled 57 for. In order to do this they first model consumption as a function o f household and individual characteristics, private assets, access to public assets and geographic factors, and then use this model to explain differences inconsumption levels across the three regions: Costa, Sierra and Selva. 3.22 They conclude that most o f the difference in per capita expenditures across regions can be accounted for by differences in private assets and infrastructure, measured using a "basic needs" index. Inother words observationally equivalent householdshave similar probabilities ofbeing poor irrespective ofthe geographic characteristics, such as altitude or temperature, o f their regionof residence(Table 3.4). 3.23 These results, however, do not indicate that geography does not matter, as the authors correctly point out, but rather that its influence on consumption comes through a spatially uneven distributioninthe provision of public infrastructure. 3.24 A direct conclusion fromthis observation is that public investments ininfrastructure inareas with adverse geography have significant redistributionaland equalizing potential. Inthis context recent trends inpublic investment are worrisome. Bothinabsolute andrelative terms the amount of resources devoted to public invest have declined steadily between 1999 and 2002, recovering only slightly afterwards (see Box 3.2). Although part of this decline can be explained by the privatization of large state-owned telecom and electricity enterprises duringthe 1990s, progress towards higher access to public services and road infrastructure has recently stalled even though important needs persist. Box 3.2: Public investment inbasic servicesandroad infrastructure The role of infrastructure as a key determinant of long-term growth and sustainablepoverty and inequality reductions has been emphasizedby numerousauthors, most recently Easterly and Serven (2003) and Calderon and Serven (2004). In this box we briefly review recent trends in public investment expenditures, access to basic service and to road infrastructureinPeru, andproposeafew policy options aimed at increasingand sustainingpublic investmentlevels. Public investments levels declined significant, both in absolute and relative terms, between 1999 and 2002, and have only recoveredslightly since. The level of public investmentfell from5,657 millions of Nuevos Soles, or 4.9 percentof GDP, in 1999 to 3,575 millions of Nuevos Soles, or 2.8 of GDP, in 2002. Although the level of investment had recoveredto 4,083 millions of Nuevos Soles by the end of 2004 and further increases are projected by the Ministry of Finance, the estimated2007 level is still below the 1999one (Table B.3.2.1). As a consequenceof this deterioration in public investment levels, progresstowards increasedaccess to basic services and road infrastructure stalled after 2000 both inurban and rural areas compared to previous periods. The fraction of households with access to tap water increased from 59.9 to 60.9 percent between 2000 and 2003, compared with an increase from 54.8 to 59.9 duringthe three years prior. The same can be said about access to electricity and sanitation (Table B3.2.1). Similarly progressinextending road infrastructure was mainly due to local investmentson rural roads, rather than to investmentsfunded by the central government (not shown). Access to tap water indwelling Access to electricity Toilet indwelling Total Urban Rural Total Urban Rural Total Urban Rural 1997 54.8 73.3 17.7 69.2 91.8 23.8 45.6 66.3 4.1 2000 59.9 73.5 33.2 72.8 95.2 28.9 47.1 69.5 3.3 2003 60.9 32.8 72.4 31.8 47.5 5.7 3.25 A slowdown in public investment in basic services and road infrastructure can then severely hinder future output and employment growth and, as a result, future poverty reduction. We will discuss in more detail the role of access to basic services and road infrastructure as determinants of economic opportunities for the urban and rural poor inChapters 4 and 5 respectively. 58 MICRO-DETERMINANTS OFPOVERTYDYNAMICS 3.26 We have so far discussedpoverty and its determinants in a static manner, as if takinga snapshot. However, poverty i s a dynamic phenomenon. Household endowments and the returns to these endowments change over time and as a result, so does poverty. Similarly persistent differences in living standards across areas and regions can be mitigated or exacerbatedby policy interventions. 3.27 We analyze flows in and out of poverty and study the micro-determinants of poverty dynamics from two different angles: changes inthe distribution of income over time, and the role of geography in explaining spatial variation inincome. Flows Inand Out of Poverty 3.28 Inthinkingabout poverty a distinction has often beenmade between permanent and transitory poverty, where the former i s perceived as being the result of structural household and individual characteristics while the secondi s thought of as being the result of a transitory (or reversible) shock. 3.29 Whether most poverty is permanent or transitorily has important policy implications. For instance, one may want to consider social assistance policies in the first case, while interventions that promote access to credit and insurance may be more effective in the second (for a more detailed discussion onthis, see Chapter 5). 3.30 A potential way to assess the prevalence of each "type" of poverty is to analyze flows in and out of poverty. Herrera and Roubaud (2002) do this inthe case of Peru during 1997-99 usingthe panel data component of ENAHO. This panel contains information on 1,720 households for all three years, and for larger numbers when pairs of years are considered (Table 3.5). 1997 1998 1999 Total numberof households 4,022 4,044 2,218 1997-98panel 2,709 2,709 1998-99panel 1,872 1,872 1997-99panel 1,720 1,720 1,720 3.31 Year-to-year flows in and out of poverty are large and stable over time. About 20-25 percent of all households change poverty status in-between years. Inaddition almost 40 percent of poor households exit poverty in any given year, while about 15 percent of non-poor households fall into poverty. Herrera and Roubaud (2002) check for the robustness of these figures to ensure that they are not driven by false transitions due to measurements error in income, and find that at least 85 percent of all transitions between states are driven by changes inincome larger than 30 percent but within the boundaries of what i s considered "plausible shocks" in a context of macro instability and absence of social protection networks (Table 3.6). 3.32 Transitions over three years also exhibit significant mobility, although certain within-state inertia exists. Approximately 35 percent of all households changes states at least once duringthose three years, while the rest remain poor (13 percent) or non-poor (52 percent). However households do not make transitions at random. Rather the probability that a particular household changes states seems to be contingent on hist0ry-i.e. a household that did not make a transition between 1997 and 1998 was less likely to do so between 1998 and 1999 (Figure 3.1). 59 1997 1998 Poor Non-Poor Total Poor 1 62.4 37.6 100.0 (29.2) Non-Poor 13.6 86.4 100.0 (70.8) Total 27.9 72.1 100.0 1998 1999 Poor 63.0 37.0 100.0 (27.9) Non-Poor 19.4 80.6 100.0 (72.1) Total 31.6 68.4 100.0 1 Figure 3.1: Three-year transitionsare still highbut are history-dependent 1997 1998 1999 Total 1999 Source: HerreraandRoubaud(2002). 3.33 These seem to suggest that households with certain characteristics are more likely to remain either poor or non-poor, while others are more likely to change states. In order to investigate this, the authors first classify households in three categories according to nature of transitions during 1997-99: (i) permanently poor, (ii) poor, and (iii) never transitorily poor (poor during 1or 2 years out of three). They then study the determinants of persistent poverty, when compared to no poverty, and the determinants of exit from and entry intopoverty. 3.34 Household and household head characteristics are important determinants of persistent poverty, while shocks, both demographic and economic, and labor market attachment, measured as the fraction of formally employed household members, are important determinants of transitions. In addition neighbor characteristics seems to have some impact on both types of comparisons (Table 3.7) 60 Table 3.7: Household characteristics are important determinants of persistent poverty while shocks are important determinants of transitory poverty ._ Persistent Poverty Exitfrom Poverty Entry into Poverty versus versus versus Never Poor Persistent Poverty Never Poor Head of household Age L t Gender (Male) Primary or less 1' t Secondary Self-employed UnemployedlNLF Employment inagriculture t7-tt f Employment inmanufacturing Household Size L % members0-9 years old % members 10-15years old % members60 and above L Number of LFparticipants L Informality rate T No landlhousetitle Access to tap water 1' Access to electricity 1 Access to toilet L Averageeducationlevel L L Neighborhood Average income Average educationlevel Informality rate Shocks Loss of householdhead 1 Loss of employmentof householdhead t Loss of employment by other members L t Increaseinformallv emvloved members 1' Note: tindicatesthat significantlyincrease leprobabilityofthe x first state relative to the second, while1indicates that x significantly decreasesthe probabilityof the first staterelative to the second. Source: Herrera and Roubaud(2002). 3.35 Insumflows inand out of poverty appear to be relatively large and stable over time. Moreover whether a particular household changes poverty status or not over a certain period of time i s a function of both household characteristics and endowments, and of shocks. We explore the role of household characteristics and endowments are determinants of poverty dynamics, bothacross time and space, betow andpostpone the discussion on shocks and their impact to Chapter 6. Determinantsof Changesinthe Distribution of Income 3.36 Income levels and, consequently, poverty and inequality levels can change as a result of changes in household characteristics and endowments, of changes in the returns to these characteristics and endowments, or of changes inboth. For instance, other things being equal income may rise and poverty 61 decline as average education levels increase or as the wages of female workers grow closing the gender wage gap. 3.37 Inacompanion study to this report, Sosa-EscuderoandLuccheti (2004) usehousehold-level data to analyze the micro-detenninants of changes in poverty and inequality in Peru during 1997-2002. In particular they explore the role of changes in (i) workers' human capital (measured by education and experienced, proxied by age and age squared), (ii)workers' demographic characteristics, (iii) job characteristics (measured by sector of activity, and type of employment, formal or informal), and (iv) geographic indicators (region and area of residence). 3.38 Their main findings are summarized below (Table 3.8) and a detailed explanation of their methodologicalapproachis provided inBox 3.3: Table3.8: Changesininequality andpoverty(per capitaincome) Inequality Poverty Extreme Poverty Gini FGT(0) FGT(1) FGT(2) FGT(0) FGT(1) FGT(2) Observed change1997-2002A 1.9 6.3 3.9 2.7 3.7 2.0 1.4 Returns 0.0 -9.9 -7.2 -5.3 -8.9 -4.7 -2.9 Education 0.4 0.1 0.2 0.2 0.3 0.2 0.2 Gender -0.1 -1.2 -1.0 -0.8 -1.5 -0.7 -0.4 Experience -0.3 -8.8 -6.4 -4.7 -7.7 -4.2 -2.7 Endowments -0.1 4.1 2.8 2.2 3.5 2.2 1.9 Education 0.1 -0.3 -0.2 -0.1 -0.3 -0.1 0.0 Industry of employment 0.0 0.6 0.6 0.6 0.7 0.7 0.7 Informality 0.1 0.4 0.2 0.2 0.3 0.2 0.2 Hoursof work -1.0 2.0 1.1 0.6 1.2 0.4 0.2 Regionlaxeaof residence 0.7 1.4 1.1 0.9 1.6 1.o 0.8 Unobservables 1 -0.6 0.5 0.2 0.1 0.3 0.1 0.1 Note: Observedchanges inpoverty and inequality are different from those presented inTables 1.3a 1.3b and 1.4 because the authors calculatepoverty andinequality on the basis on (labor) income rather thanconsumption. Source:Sosa-Escudero and Lucchetti (2004). Changes in the returns to education, particularly increases in the relative returns to higher levels of education, contributed to increase both poverty and inequality. Among household heads (typically, middle-aged male workers) real wages declined significantly for all educational categories except tertiary, while among non-household heads they increased slightly for all levels except incomplete primary. Both effects have contributed to an increase in wage dispersion across education levels. Given the relative importance of household-head earnings as a fraction of total household income, it i s plausible for the first effect to be driving the observed changes. In contrast, changes in the educational structure of the work force contributed to poverty reduction, while having a mild impact on inequality. As the average level of education increases, so does the averagewage paidto workers since wages are positively correlated with education levels. An increase in the wage differential between male and female workers, combined with a small increase in the percentageof male workers in the labor force resulted in higher inequality and lower poverty. Changes in the returns to experience had a poverty reducing, equalizing effect. This i s due to a decrease in wage differences between young (less experienced) and old (more experienced) workers, which could be the result of a more rapid depreciation of older workers' skills (due, for instance, to technological change) or of the observed decline injob tenure (see Chapter 3 for a discussion on the issueofjob tenure). 62 Regarding job characteristics, changes in the distribution of workers across industries and between the formal and informal sectors increased poverty, but had a negligible effect on inequality. In additionchanges inthe number of hours worked resulted inhigher poverty and inequality. 0 Different growth patterns across regions and areas of residence resulted in both higher poverty and inequality. 0 Changes in unobserved characteristics, such as school quality, labor market connections and unmeasuredskills, have had an equalizing effect, but have also increased poverty Box3.3: Explainingchanges inpoverty and inequality usingmicro-simulations Microeconomic simulations of counterfactual distributions are helpful to characterize past distributional changes and to simulatethe distributional impact of changes ineconomic factors and public policies. The characterization of past distributional changes relies on a decompositionexercisethat analyzes the driving forces behind observed changes inthe distribution of income. For this purposethe distribution of (labor) income at time t is modeledas a function of observable and unobservable individual characteristics, so that past changes in income can be attributed to (marginal) observed changes in such characteristics. Other sources of income, such as non-labor incomeor transfers, are not consideredinthe analysis. This model can then be used to simulate counterfactual scenarios that evaluate the potential impact that changes in these characteristicscouldhaveonthe distribution of income. For instance, inorder to assess what the impact of an increase inthe wages perceivedby workers with tertiary educationwould be, a counterfactualdistribution is generated using time-t values for all individual characteristicsbut the returns to tertiary education, which is set at the desired level. In other words, the counterfactual distribution is the ones that would have been observedhad the wages of tertiary workers increasedwhile all other individual characteristicsremainedconstant. Sosa-Escudero and Lucchetti (2004) apply this methodology to analyze changes inthe distribution of income in Peruduring 1997-2002,and to simulate the potential impact of changes in various individual characteristics and policy variables on this distribution. For this purpose they estimate earnings equations using information on (i) workers' human capital (education and experience, proxied by age and age squared), (ii) workers' demographic characteristics, (iii)job characteristics (sector of activity and informality indicators), and (iv) geographical location (region and area or residence). These equations are estimated separately for heads and non-heads of households, and for urbanand rural areas, usingOLS and quantile regressionmethods. The results from the decomposition exercise are discussed inChapter 2 of this report, while the results from the simulationexerciseare presentedinChapter 1. Source: Sosa-Escuderoand Lucchetti (2004). 3.39 In sum recent changes in poverty and inequality in Peru result from several forces acting in opposite directions. In general changes in the returns to individual characteristics have contributed to poverty reduction, while changes in the individual and job Characteristics have pushed in the opposite direction. ALTERNATIVE MEASURES OFPOVERTY 3.40 Poverty i s a multidimensional phenomenon and, as a result, monetary measures of poverty, to which we have devoted all of our attention so far, can only provide us with a limited picture of the daily reality of the poor. Inthis section we explore two alternative definitions of poverty currently used in Peru, the index of Unsatisfied Basic Needs (UBN) and the caloric deficit (CD), and their relationship to monetary poverty. 63 Unsatisfied Basic Needs 3.41 The Unsatisfied Basic Needs (UBN) index measures the fraction of the population subject to certain structural deficiencies. The index i s a weighted summary of five-types of household-level variables and indicators: (i)characteristics of the dwelling, (ii) overcrowding, (iii) of access to degree basic sanitation services, (iv) the presenceof children of schooling age currently not attending school, and (v) the householddependency rate. 3.42 The UBNindex i s thought of as providing a measure of structural depravation or poverty and, as such, i s not as sensitive to the businesscycle as monetary poverty. Infact changes inthe UBNindex tend to reflect secular trends in economic growth and social investments. Both measures, however, are positively correlated. That is, areas with highlevels of monetary poverty also tend to have highvalues of the UBNindex (Figure3.2). I* Figure3.2: Monetarypoverty and the UBNindex are positively correlated at the departmentlevel 100% Huancavelica 80%. _ _ UBN6O0/o 40% 20% 20% 40% 60% 80% 100% Headcountratio I Source: Authors' calculations usingdatafrom ENAHO2002 and 2003 (INEI). 3.43 Poverty measuredby the UBNindex has declined steadily during the last 10years. The fraction of the population with at least one UBNhas fallen from 56.8 in 1993 to 40.3 percent in 2003. Progress has been significant inboth urban and rural areas. The UBNdeclines from 42.4 to 25.3 and from 90.1 to 71.3 percent inurban and rural areas respectively duringthis period. 3.44 Most of this decline, however, took place duringthe 1990s, before public investment levels were cut significantly and progress in access to basic services stalled (Box 3.2). Further advancement in the reductionof the population share with U B N s will hence depend, among other factors, on the recovery of public investment levels (Table 3.9). 64 Table 3.9: Unsatisfied Basic Needs Index (UBN) Presenceof At least one Inadequate Inadequate school-age High UBN dwelling Overcrowding sanitary children not dependency services currently rate enroiled 1993 Census National 56.8 13.8 25.0 36.5 10.6 13.6 Urban 42.4 10.3 18.4 19.8 7.O 8.7 Rural 90.1 22.0 40.4 75.3 18.8 25.0 ENAHO 2001.IV National 41.9 11.6 18.8 23.4 2.8 2.6 Urban 28.9 9.7 13.6 12.3 1.3 1.9 Rural 69.8 15.6 30.1 47.4 5.9 4.1 ENAHO 2002.IV National 39.9 12.2 17.5 21.7 3.1 2.3 Urban 26.5 9.4 12.0 9.7 1.5 2.0 Rural 70.0 18.4 30.0 49.0 6.8 3.0 ENAHO2003.W National 40.3 11.5 16 23.7 3.3 1.5 Urban 25.3 6.5 7.6 10.2 1.5 1.2 Rural 71.3 16.9 28.1 51.2 6.3 1.9 Source: Cubto (2004) andauthors' calculationsusing data fromENAHO 2003.N (INEI) Caloric Deficit 3.45 The caloric deficit measures the fraction of the population living inhouseholds where per capita food intakes are insufficient interms of their caloric content; where individual minimumcaloric levels are established according to demographic characteristics and area of residence. 2001 2002 2003 2004 National 33.3 35.8 34.2 35.0 Urban 26.2 29.4 31.4 31.9 Rural 46.6 47.7 39.2 40.6 3.46 The caloric deficit i s therefore thought of as providing an alternative measure for extreme poverty. The extreme poverty line equals the monetary value of a basic food basket that provides the minimumcaloric intake. However, becausemoney is fungible, the fact that household income is equal to or above the extreme poverty line does not guarantee that household food consumption is sufficient to satisfy minimumcaloric needs, while the caloric deficit provides a direct measure of actual nutritional deficiencies. 3.47 The nutritional status of the population, measured as the caloric deficit, has deteriorated over the past three years, particularly inurban areas. The percentage of the urban population with inadequate food 65 intakes has increased a whooping 22 percent between 2001 and 2004. This contrasts with the observed decline inextreme poverty, bothinurbanareas and across the country duringthis period (Table 3.10). POVERTY MEASURESTARGETINGAND INSTRUMENTSFORSOCIAL PROGRAMS 3.48 Social programs and transfers can be targeted using several tools and criteria. Targeting tools or mechanisms include poverty maps, which target benefits at a geographic level, proxy-means indicators, which target benefits at the household or individual level, and self-targeting mechanisms, such as the one employed in A Trabujar in Peru, which also target benefits at the household or individual level. In addition geographic areas, households or individuals can be selectedas potential beneficiaries on the basis of a series of welfare criteria, ranging from income-based criteria, such as monetary poverty, to criteria basedon structural deficiencies, suchUnsatisfiedBasic Needs (UBN)index. 3.49 The choice of a particular tool or criterion should then depend on the nature and main objectives of the program(s) being targeted. Programs directed to individuals or households, such as cash-transfer programs or programs that support the elderly, are better targeted using household-based tools, while programs direct to other units, such as school feeding programs or infrastructure investments, are better targeted using geographic criteria. Similarly programs aimed at alleviating income constraints or at minimizing the impact of cyclical shocks are better targeted using income-based criteria, while those aimed at improving living conditions over the medium-term are better targeted using criteria based on structural needs. 3.50 What does this imply for Peru?We briefly comment here on existing targeting tools, as well as on the ongoing discussion about the possible adoption of a proxy-means index as a new tool, and provide a few recommendations both. Poverty Maps 3.5 1 Since the late 1990s Peru has used data from the 1993 Population Census to target certain social programs at the district level. Peru also has a Poverty Map that combines data from the 1993 Population Censusand from household surveys to produce district-level poverty estimates. 3.52 Although the map has been updated periodically by using the latest available household survey (Le. the last update uses the 1999 ENAHO), it suffers from a certain degree of obsolesce due to two factors. First, by relying on the 1993 PopulationCensus sampling framework it fails to account for newly developed settlements, particularly in urban areas. Because most of these settlements are poor, this has caused the distribution of social programs to be slightly biased against urban areas, as we pointed out above. Second, economic circumstances in Peru have changed significantly since the map was first constructed. Although the use of recent household data accounts for some of these changes, the effectiveness of this strategy is limited since the accuracy of the map deteriorates rapidly as the dates on which the Census and the household survey were collected grow apart. 3.53 To illustrate this point we compare the department-level rankings produce by the latest poverty map (1994 Census and 1999 ENAHO) and by the 2003 ENAHO. The correlation between both rankings i s high, but there are important disagreements between them, as measured by changes in the relative position a each department. Inparticular six departments go up or down in the rankingby 5 or more positions-a significant amount given that the total number of departments is 25 (Figure 3.3). Although this i s clearly a very rough exercise that ignores that benefits are targeted at the district and not at the department level, and that significant variation inpoverty rates exists within departments, it suggests that there is a certain disconnect between the country portrayed by the map and the actual allocation of poor householdsacross areas and regions. 66 Figure 3.3: Changesinrankingproducedby poverty map ISource: Authors' calculationsusingdatafrom 1999PovertyMapand ENAHO2003 (INEI). Proxy-MeansIndicators 3.54 The Government of Peru is currently considering the development of a proxy-means indicator to target certain social programs. Although Peru has usedthis kindof mechanismto target certainprograms, such as ProJoven, inthe past, it has not done so at a large scale. 3.55 The main advantage associatedwith the use of these indices it that they can help reduce leakages inthe presenceof significant within-district heterogeneity, or when programs target a specific population group rather than a specific income group. 3.56 Their construction, however, requires the collection of a significant amount of information for every household that has the potential of becoming a programbeneficiary. This process can be expensive, although various strategiesto minimize costs, such as self-reporting through "convocatonas populares" in rural areas, can be considered. 3.57 Moreover, as was the case with poverty maps, they are subject to obsolescence as household characteristics and circumstances change over time (i.e. a recent evaluation of the proxy-means index used inEcuador indicates that the information usedto construct the index should be updated every 4 or 5 years). This i s particularly true if the index gives significant weight to household income or expenditure levels, as opposed to, say, the characteristics of the dwelling, since the former are more sensitive to cyclical fluctuations. Also because households are more likely to change economic status than entire geographical areas, the impact of business cycles i s more acute for proxy-means indices than for poverty maps. 3.58 In thinking about developing such a tool the GOP should then consider a series of factors including:25 25. The Government of Peru is currently implementing a series of surveys representative at the provincial and district levels with the objective of gatheringrelevant datafor targeting purposes. 67 Potential savings from better, more effective targeting of certain programs. This could be estimated through simulation exercises using information on existing within-district heterogeneity and current geographical targeting rules. The cost of collecting the required information. This process could be made more economical by linkingitto the Census, orby usingself-reporting (especially indisperserural areas). The tool's sustainability over time. The extent to which the information contained in the index depreciates over time will be a function of both its methodological design and the rate of change of overall economic and social conditions. Responsivenessto the latter can be enhancedby allowing the tool to remain "alive" in-between waves of information collection. This can be done using "ventanillas" where households that do not currently qualify as beneficiaries can request to be administered the index's questionnaire and where irregularities regarding current beneficiaries can also be reported by third parties. In both cases, reclassification f a particular household should be subject to verification of the information. CONCLUSIONS 3.59 Inthis chapter we have shown that there exist significant andpersistent differences between poor and non-poor households in terms of their demographic characteristics, access to basic services and infrastructure, and employment status. 3.60 We have also argued, however, that poverty is a dynamic and multidimensional phenomenon and that, as a consequence, traditional static poverty profiles based on single poverty measures only present a limitedpicture of the actual reality of poor households. 3.61 Finally we have explored the implications of our discussion on the nature of poverty for the design and selection of targeting tools for social programs, an issuecurrently under discussion inPeru. 68 4. ECONOMICOPPORTUNITIESFORTHEURBANPOOR% 4.1 Employment constitutes the main and frequently the only source of income for most families living in urban areas, so more often than not the lack of it leads to poverty. Labor income accounts for more than 75 percent of total income among urban households. As a result, the focus of this chapter will be on labor markets and the capacity of the urban economy to generate employment and income and, therefore, reduce poverty. 4.2 The chapter i s structured as follows. The first section discusses the relationship between labor markets and poverty in urban areas paying attention to the relative importance and the determinants of participation in formal and informal activities and to recent labor market trendsz7. The second section uses data from the urban manufacturing sector to study existing constraints to employment creation, with a particular focus on the role of labor legislation. The third section examines the nature and productivity of informaleconomic opportunities available to the poor, and analyzes the cost of informality. Finally the fourth section concludes. 4.3 The main findingsof the chapter can be summarized as follows: The average urban household obtains most of its income in the form of labor income, but important differences exist between poor and non-poor households in terms of their income generating strategies. Poor households rely relative more on informal activities and are less able to minimize income risk by diversifying across different economic activities than non-poor ones. Higher participation inthe informal sector can be explained by lower education levels among poor household members. Recent improvements in average employment and wage levels have failed to translate into lower urban poverty rates because they have been concentrated among formal, more educated workers, employed in larger firms, and have not extended to the informal sector where most of the poor are employed. Future declines in urban poverty will therefore depend on the capacity of the urban economy to generate more productive, well-paid jobs, particularly in those sectors that employ the poor. Firms identify labor legislation rigidities and uncertainty about future sales as the main constraints to formal employment creation. Peru's labor legislation i s very protective de jure, but offers low and unequal coverage de facto due to the extensive use of temporary contracts and the high incidence of informality. As aresult the impact of labor legislation extends beyondformal permanent employment to affect overall employment levels and its composition. Most of the urban poor are employed in small, informal businesses. Low levels of productivity among poor entrepreneurs and, consequently, lower wages among their employees can be explained by lower levels of education of both employers and workers, Lower levels of market integration and lower access to basic infrastructure. High levels of informality are associated with inequality in access to social protection and with productivity and fiscal costs. Slow and costly business registration procedures, inflexible labor regulation, high levels of uncertainty among employers, and low and costly access to credit are the mainreasonsbehindhighinformality rates. 26. This chapter is based on background work prepared by the report team, as well as on existing work by Saavedraand Torero (2003) and Jaramillo (2004a). 27. For this purpose informality'will be defined on the basis on firm size (10 or fewer employees) andlor on the basis of the individual's employment status (self-employed), rather than on legalistic terms (keeping of formal accounting, contributions to Social Security, etc.). This decisionresponds to data availability restrictions. 69 Q As a result policy interventions aimed at promoting formal employment creation, increasing the productivity of informal activities and creating incentives for the formalization of informal businesses can go a long way inreducingurbanpoverty. LABOR MARKETSURBANPOVERTY AND 4.4 Urban households obtain most of their income through the use of their labor and other productive assets. The productivity of these assets vanes across sectors and activities and so do their returns. As a result the sector and type of employment individuals and households have access to bears an impact on income and poverty. In addition changes in economic and labor market conditions over the business cycle can also have an impact on the nature of economic opportunities available to urban households, bothpoor and non-poor. 4.5 In this section we discuss the relationship between type (e.g. salaried or self-employment) and sector (e.g. formal or informal) of employment and poverty, and examine the determinants of participation in different sectors. We also review recent labor market trends and discuss how they have affected urban poverty, if at all. Employment, Labor Income and Poverty inUrbanAreas 4.6 There exists significant variation in the incidence of poverty across groups of people with different labor market status. Poverty rates are highest among the informal salaried and self-employed (35 percent) and among the unemployed (33 percent). And they are lowest among the formal self- employed (5 percent), employers (17 percent) and formal salaried workers (18 and 14 percent for private andpublic workers respectively). Individuals who are out of the labor force exhibit poverty ratesthat are in between those of the previous groups (28 percent), probably due to the presence of pensioners among them (Table 4.1). Table 4.1: Unemployment, self-employment and informal employment are positively correlated with poverty Extreme poor Poor Percentageof individualsingroup Employer 1.6 17.3 Formal Salariedprivate 1.2 17.8 Salariedpublic 1.4 13.9 Informal Salaried private 4.8 34.8 Self-employed Formal 0.9 5.0 Informal 5.1 34.8 Unemployed 3.3 33.3 NLF 4.1 27.9 4.7 Moreover most of the variation in poverty rates can be explained by differences between sectors rather than by differences between types of employment. On average poverty rates are higher among workers inthe informal sector than among their formal counterparts, irrespective of whether they are self- employed or salariedemployees. 70 4.8 Welfare differences across formally and informally employed individuals translate into welfare differences across households, as the capacity of their members to access formal or informal employment determines the nature of income sourceshouseholds depend on. 4.9 Poor households rely relatively more on informal activities than non-poor households. Households in the bottom quintile of the income distribution (the poorest households) obtain 95 percent of household labor income from the informal sector, compared to 40 percent for households in the top quintile (the richest households). In addition informal self-employment i s particularly important among poor households, while members of non-poor ones are more likely to be formal salaried workers, both in ;he privateand public sectors (Figures 4.la and 4.lb). Figure 4.la: Poor householdsobtainmost of their incomefrom informalemployment ... Note: All workers infirms with 10or fewer employeesand all self-employed individuals (with the exception of professionals)are consideredinformal. Source: Authors' calculations usingdata from ENAHO2003.N (INEI). Figure4.lb: ...especially from self-employment. 1 2 3 * 5 Incomequlntlles /BFormal Public Formal Pwate0Informal. Non-saiariedI#lnf-l- - - Salaried1 Note: All workers infirms with 10or fewer employeesand all self-employed individuals (with the exception of professionals)are consideredinformal. Source:Authors' calculations usingdatafrom ENAHO 2003.IV (INEI). 71 4.10 Informal activities are not exclusively limited to poor households, however, nor are formal ones limited to non-poor ones. Twenty-five and 35 percent of household income comes fromformal activities for those in the second and third quintiles respectively. Similarly, informal self-employment commands about 35 percent of household income among those in the fourth and fifth quintiles (Figu4e 3.la and 4.lb). 4.11 Finally non-poor households seem to be better able than poor ones to minimize income risk by diversifying across different economic activities. Poor households rely on fewer labor income sources and thus are more exposed to sector-specific shocks. For instance informal self-employment represents 80 and 50percent of total labor incomefor householdsinthe first and secondquintiles. Incontrast, richer households have access to a larger number of income sources, each of which account for at most 30 percent of total labor income (Figure 4.lb). 4.12 The capacity of individuals and, as a result, households to access formal and informal employment i s a function of both individual characteristics and labor market conditions. In addition the latter can change over the business cycle and have an impact on the employment opportunities available to urban households. We analyze both the micro-determinants of access to employment in different sectors and recent labor markets trends below. Determinantsof ParticipationinFormaland Informal Activities 4.13 Inordertoexaminethedeterminants ofparticipationwedistinguishbetweenformalandinformal activities inurban areas, where the informal sector includes workers in firms with 10or fewer employees and unregistered, non-professional self-employed individuals. We also concentrate on private salaried and self-employment and exclude from our analysis employers, public sector employees, workers inthe domestic service sector and all those employed in family businesses. Finally we differentiate between salaried and self-employed individuals in the informal sector in response to the evidence provided in the literaturethat the reasonsfor being informal vary significantly across bothgroups (Maloney, 2003). 4.14 We use regression analysis to examine the impact of individual and job characteristics on the likelihood of (different types of) employment in the formal and informal sectors. We estimate two different models for this purpose: a probit model that distinguishes between formal and informal workers, and an ordered logit model that distinguishes between formal workers, informal salaried workers and those self-employed inthe informal sector. Our results can be summarized as follows (Table 4.2): Demographic characteristics: Male household heads are more likely to be formally employed than other male members of the household. Incontrast, both female heads and other female household members have a higher probability of being informal than their male counterparts. Older workers appear more likely to participate in informal activities according to the probit model, although this result i s not significant. The role of age as a determinant of participation in formal and informal activities appears to be more nuanced, however, once we distinguish between informal salaried and self-employment in the ordered logit model. Inparticular younger workers are more likely to be salaried employees inthe informal sector, while older workers are more likely to be self- employed, suggesting that the role of formal and informal employment may vary with age (Le. informal employment can constitute an entry port into the labor market for young salaried workers, and a profitable employment alternative for older self-employed individuals). Education: The probability of being informally employed decreases with education, and it does so more rapidly for higher levels of education. In the ordered logit model the negative correlation between education and informality is stronger among salaried workers than among the self-employed, which i s again consistent with the idea of informal self-employment representing a viable option for older (and skilled) workers lookingfor higher flexibility. 72 Table 4.2: Participationinformal and informalactivitiesis a function of worker andjob characteristics Probit Ordered Iogit Informal Informalsalaried Informal non-salaried Formal Marginal effects (dyidx) Householdhead -0.041'' -0.045** -0.010=* 0.059" (0.006) (0.010) (0.003) (0.013) Female 0.085** 0.021** 0.004** -0.026'* (0.006) (0.010) (0.002) (0.012) FemaleHHhead 0.056** 0.042** 0.003'* -0.049** (0.010) (0.013) (0.001) (0.014) Age 0.006 -0.014** -0.003** 0.018** (0.001) (0.002) (0.000) (0.002) Age squared 0.000 0.000** 0.000** -0.000** (0.000) (0.000) (0.000) (0.000) Education Incomplete primary Baseline category Completeprimary -0.012 -0.001 -0.002 0.001 (0.010) (0.008) (0.003) (0.010) Incomplete secondary -0.052** -0.001 -0.002 0.001 (0,010) (0.008) (0.004) (0.010) Complete secondary -0.117** -0.032** -0.008'* 0.040** (0.009) (0.007) (0.002) (0.010) Tertiary -0.177" -0.077** -0.025"' 0.097** (0.010) (0.010) (0.005) (0.017) Industry of employment Agriculture Baseline category Construction 0.182** 0.256** -0.071** -0.191" (0.005) (0.038) (0.022) (0.017) Manufacturing 0.134** -0.035 -0.011 0.040 (0.006) (0.020) (0.010) (0.030) Services 0.407** 0.044** 0.019" -0.072** (0,010) (0.019) (0.009) (0.028) Utilities -0.170*' -0.179" -0.324** 0.493** (0.049) (0.011) (0.060) (0.071) Occupation Managerflrofessionals Baseline category White collar 0.460" 0.861" -0.193** -0.693** (0.009) (0.024) (0.011) (0.034) Bluecollar 0.734** 0.702** 0.154** -0.888*' (0.012) (0.031) (0.008) (0.024) Other 0.290** 0.815** -0.490'* -0.343** (0.003) (0.012) (0.005) (0.012) Regiondummies Yes Yes Yes Yes Year dummies Yes Yes Yes Yes Note: "(') Significantly diffe it from 0at the 10(5) percentlevel. All workers infirms with 10or fewer employees and all unregisteredself-employed individuals, with the exceptions of professionals,are consideredinformal. Source: Author's calculations using datafromENAHO 1998-2003(INEI). Industry of employment: Workers in the construction, manufacturing and service sectors have a higher probability of being employed informally than those in primary activities-the baseline category-or in the utilities sector. Higher incidence of informality in the construction and service sectors i s partly the result of seasonal cyclicality, while higher incidence of informality in manufacturing is a reflection of the large number of micro and small firms inthe sector. The positive effect of employment inconstructionand services on the probability of being informal is driven by its impact on the probability of informal salaried employment, rather than informal self-employment. Occupation: The probability of being informally employed decreases as we move from blue to white-collar activities according to the probit model-a distinction that can serve as a proxy for skill 73 requirements, hence confirming the existence of a positive relationshipbetween formality and human capital. Once we distinguish between informal salaried and self-employment, however, interesting differences arise. In particular, employment in white collar activities increases the probability of being a salaried worker in the informal sector, while employment in blue collar activities increases the probability of being self-employed inthe informal sector. 4.15 Insummale primary earners, older and more educated workers are on average more likely to be formally employed than other individuals. We findevidence, however, that these relationships may vary as individuals progress along their working careers-:.e. some older and more educated workers may contemplate informal self-employment as an attractive alternative to formal employment. An issue we will return to later on inthis chapter. 4.16 Differences across demographic and, particularly, education groups in terms of access to formal and informal employment, combined with the fact that the poor tend to be less educated than the non- poor, explain why the poor are more likely to be informally employed. In addition this relative concentration of the poor in the informal sector, combined with recent labor market trends, explains, as we discuss next, why urban poverty has not been very responsive to economic and, more recently, employment growth in urban areas. Recent Labor Market Trends 4.17 The macroeconomic developments of 1997-2004 had a direct impact on urban labor markets. Urban employment in formal firms with 10 or more employees fell as a consequence of the 1998 economic crisis and started to recover only in 2002. Wages paid by these fums also declined after 1998, but bounced up faster than employment surpassing pre-crisis levels by 2003. The same general employment and wage trends can be detected in Lima, where increases in employment have been accompanied by increasesin unemployment as labor market participation rates have gone up inresponse to better economic prospects (Table 4.3). Table 4.3: Urbanemployment and wages declinedafter the 1998crisis and recoveredas economic growthresumed. / 1997 1998 1999 2000 2001 2002 2003 2004 Urbanareas Employmentindex A 114.7 111.7 105.8 103.1 101.9 104.3 106.1 109.4 Wage index 98.0 98.1 95.2 96.5 96.8 NA 101.6 MetropolitanLima Employment index A 113.6 111.1 104.9 101.2 101.8 103.8 105.8 108.2 Unemployment rate 8.6 6.9 9.4 7.8 8.8 9.7 10.3 Wage index 85.8 84.3 83.0 82.8 83.5 86.2 88.5 88.7E Managers 117.3 129.0 131.4 136.4 139.3 143.0 153.9 157.7E White collar employees 93.6 98.1 101.1 101.6 102.8 102.4 105.4 104.4E Bluecollar employees 85.8 84.3 83.0 82.8 83.5 86.2 88.5 88.6E 4.18 Improvements in average wage and employment levels, however, have not benefited all workers and sectors equally. The bulk of labor income adjustment fell primarily on low-skilled workers (blue collar workers or "obreros"), compared to managers or white collar workers (Table 4.3). In addition, as we pointed out in Chapter 1, the composition of urban employment has changed as its level has recovered. First, improvements in employment infirms with 10or more employees have been driven by improvements in large f r m s (Le. f m s with 50 or more employees), especially in Lima (Figure 4.2). Second, if we consider all urban employment, rather than just employment in fm with 10 or more 74 .employees, and pay attention to the employment rate2', rather than the level of employment, we observe that this rate has remained constant during 1997-2003, at around 70 percent29. Finally the informality rate has increased significantly inrecent years, from 75 percent in 1997 to 80 percent in2000 to 83 percent in 2003. Figure 4.2: Increasesinemployment have beenconcentrated inmedium andlarge firms... a' 1 2001 El2002 2003 12004 1 Note: Employment figures correspondto firms with 10or moreemployees. Source: Ministeriode Trabajo (Informe EstadisticoMensual, 2001-04). 4.19 One important limitation of the data produced by the Ministry of Labor, on which the numbers above are based, i s that it only covers formal employment. Casas and Yamada (2005) use the ENAHO 2001-2004 to document overall labor market trends, both formal and informal, and reach similar conclusions. Overall employment levels have increased but the quality of employment, proxied by the incidence of underemployment, has deteriorated and hourly earning for the average workers have remained constanti s real terms. 4.20 Furthermore weak linkages between large and small firms have prevented positive growth among the former from trickling down to the latter. In2003 about 46 of all large firms interviewed by CuBnto, a think-tank3', declared that they buy at most 20 percent of their inputsfrom small suppliers and 60 percent declare that they sell at most 20 percent of their output to small clients (Table 4.4). 28. Percentageof (potential) labor force participants who are employed. 29. SeeFigure 2.8 inChapter 1. 30. Encuesta de Opini6n (Cuinto, 2004). 75 Table 4.4: The linkagesbetweenlargeandsmall and mediumfirms are weak Sales Purchases 2001 2002 2003 2001 2002 2003 Percentageof large firms that sellslbuys from medium and small firms: upto 20% 45.9 44.0 62.0 42.3 36.6 46.5 Between20% and 40% 24.3 26.1 18.6 12.6 21.6 12.9 Between40% and 60% 12.6 15.7 4.7 14.4 12.7 2.9 Between60% and 80% 9.0 3.7 1.2 13.5 14.2 5.9 Between 80% and 100% 2.7 3.7 1.2 9.9 6.7 4.1 Noanswer 5.4 6.8 12.3 7.2 8.2 27.7 4.21 Given these developments and the fact that the poor are more likely to be employed as blue collar workers and in small andor informal f m s than the non-poor, it i s no surprise that recent improvements in wages and employment levels in urban areas has done little to bringurban poverty down (Table 4.5). Sixty-five percent of the working poor is employed inblue collar occupations, compared to 45 percent of the non-poor, and hence were exposed to the decline in wages that took place during 1997-2000. Similarly only 17 percent of the poor work in firms with 50 or more employees, compared to 30 percent of the non-poor, so that positive increases in employment levels among these f m s could only have limitedimpact onthis group. Table 4.5: The poor are morelikely to work inblue collar occupationand ininformal andlor smallfirmsthanthe non-poor Poor Non-poor Percentageof workers ingroup Occupation Managers 3.7 15.7 White collar employees 20.4 37.2 Blue collaremployees 66.5 43.6 1 Informality rate 83.9 67.3 Firmsize 0-9 76.2 60.2 10-49 6.7 8.2 50+ 17.1 31.5 4.22 Future declines in urban poverty will therefore depend on the capacity of the urban economy to generate more productive, well-paid jobs, particularly in those sectors that employ the poor. The . acceleration inemployment and wage growth observed in2004 and the first months of 2005 constitutes a step in the right direction, but further progress will be needed. We examine this issue inthe second part of this chapter. We use fm-level data from a recent Znvestment CZimateSurvey conducted by the World Bank in Peru to analyze existing constraints to urban employment creation by large, medium and small firms, paying special attention to labor market rigidities and the role of labor legislation. In addition, because most of the urban poor are employed in small, informal businesses we also examine the determinants of productivity of informal activities, as well as the barriers to formality these and other firms face. 76 CONSTRAINTS TO EMPLOYMENT CREATION IN URBAN AREAS: AN ANALYSIS OF THE MANUFACTURING SECTOR 4.23 Inthis section we discussthe existence andnatureof constraints to employment creation inurban areas using information on manufacturing employment. Although manufacturing represents only 12 percent of total urban employment in the private sector, there are important insights to be gained regarding the overall functioning of the urban economy. Neighboring manufacturing and service sector firms operate invery similar economic environments, are subject to the same macroeconomic shocks and face the same set of labor and businessrules and regulations. Therefore, to the extent that the decision to grow or to be formal or informal depends on the economic environment rather than on the sector the firm operates in, observations on the behavior of manufacturing firms will be informative about the potential behavior of service firms. Barriers to Employment CreationandBusinessDevelopment 4.24 We use firm-level data collected in Peru by the Investment Climate Project (Development Economics Research Group, The World Bank). The sample contains information on 534 manufacturing firms located in urban areas in Ancash, Arequipa, Ica, La Libertad, Lima and Callao, Puno, Piura and Ucayali. For the purpose of the analysis, we divide firms in three groups according to their size: small, with 1to 10employees and (normally) not subject to labor inspection; medium, with 11to 99 employees; and large, with more than 100 employees. Large firms are more likely to have more than one establishment, to count with public or foreign participation and to engage inexport activities than medium and small firms (Table 4.6). Table4.6: Firmcharacteristicsvary withfirmsize All Sillt3ll Medium Large (0 to 10) (llto 99) (loo+) Numberof firms 534 199 256 79 Average size 85.3 5.6 34.0 452.3 Age of firm (years) 17.6 13.3 18.2 26.1 Number of establishments 1.4 1.5 1.1 1.8 Foreign participation (% of total capital) 5.6 1.5 4.6 18.9 Exporter (% of all firms) 44.0 21.6 50.0 84.8 Share of production inexports (%) 22.0 9.1 24.1 47.7 Total sales in2001(Nuevos Soles 1,OOO)A 20,457.3 6,014.6 3,365.8 89,150.2 Total costs in2001(Nuevos Soles 1,OOO)A 15311.9 4,375.6 3,178.2 66,250.9 Foodand beverages 6.7 2.0 5.0 24.0 Textiles 43.6 47.7 44.9 29.1 Wood and paper 11.9 13.5 12.8 5.0 Chemical 20.0 14.0 22.2 27.8 Metal 11.0 13.5 10.1 7.5 Other 6.5 9.0 4.9 6.3 Note: A Dataon sales andcosts is only availal :for 116firms (36 small firms, 58 medium fiims and 22 large firms). Source: Authors' calculations using data from the Investment Climate Assessment (World Bank, 2003b) 77 4.25 All interviewed firms were asked whether, if faced with no constraints, they would increase, decrease or maintain the actual number of permanent workers employed at the frm, and, when choosing to increase or decrease it, by how much. Using this information we construct estimates of firms' unconstrainedpreferencesregarding net employment creation. 4.26 Twenty-five percent of all firms declared they would like to increase the number of permanent workers they employed, compared with less than 5 percent that would decrease it, and 70 percent that would maintain it at its current level. Since firms were also asked to report the magnitude of the desired change, we can calculated a hypothetical `unconstrained' net employment creation rate, which stands at 4.5 percent (Figure 4.3). Figure4.3: Twenty-five percent of all firms would like to hire more workers in unconstrained Source: Authors' calculationsusingdata from the InvestmentClimateAssessment (World Bank, 2003b) 4.27 Unfortunately we can not benchmark this figure by comparing it to the actual rates of employment creation and destruction among firms inthe survey becausethis information i s available only for a very small number of these firms and, when available, it appears to be inconsistent with other employment data in the survey. We can, however, compare it to changes in urban employment levels in the manufacturing sector in the last few years, as recorded by the Ministry of Labor. The manufacturing employment index increased from 88.0 in 2000 to 92.0 in 2004, a 4.5 percent increase. That is, the desired net employment creation reported by f r m s in the sample in 2001 i s equivalent, in percentage terms, to that observedinurban areas over a span of 4 years. 4.28 The share of firms reporting desired changes i s largest among medium firms, while reported desired changes are largest among small firms. Medium-size firms more frequently report they would like to increase or decrease the number of permanent workers they employ than other firms. However, conditional on a desire to increaseor decrease employment, it i s small fums that report the largest "ideal" changes. Regarding hiring (firing) small firms report they would increase (decrease) the size of their workforce by 64 (47) percent, compared to 40 (24) and 28 (12) percent among medium and large firms respectively. 4.29 Differences in desired changes across firms are consistent with the idea that small firms are relatively more constrained when thinking about employment changes, and particularly employment growth, because most of them operate informally and significant increases in size would force them to become formal. 78 4.30 The reasons given for the difference between actual and desired hiring and firing behavior vary somewhat with fm size, and do not appear to be symmetric with respect to the desired course of action. This is not surprising given that small firms are less likely to be unionized and to comply with certain labor regulations than larger firms, as long as they remain small. For instance, firing costs appear to be a greater constraint when thinkingof contractingthan when thinkingof expanding, probably due to the fact that in the first case they would have to be incurred immediately while in the second they would only materialized if the newly employed workers were to be fired in the future. Similarly firing costs also appear to be a stronger constraint for mediumand large firms who are more likely to comply with existing regulations than for small ones. This asymmetry could indicate that a reduction in legislated firing costs could initially leadto a reduction inpermanent employment andor a substitutionof permanent employees for temporary ones. Uncertainty regardingfuture sales also plays an important role in shaping hiringand firing decisions, which is consistent with our discussion in Chapter 2. Finally unions and bureaucratic procedures with the Ministry of Labor do not seemto impose to heavy a burden on either hiringor firing (Table 4.7). All Small Medium Large (0 to 10) (llto 99) (loo+) Not increasing Firing costs 50.0 39.3 50.0 68.4 Non-wagelaborcosts 73.8 69.6 76.9 68.4 MOLprocedures 6.1 15.1 3.8 0.0 Unions 1.5 0.0 1.2 5.2 Sale expectations 29.2 39.3 26.9 21.0 Not decreasing Firing costs 70.0 NA 76.9 NA Non-wagelabor costs 55.0 NA 69.2 NA MOLprocedures 5.0 NA 0.0 NA Unions 5.0 NA 0.0 NA Sale expectations 15.0 NA 15.3 NA 4.31 Although the discussion in this chapter i s focused on the impact of labor legislation on employment creation, it is important to keep inmindthat there are numerous other factors that also affect employment and, more generally, business decisions by firms. Infact insome cases these factors may be considered more importantor bindingconstraints than labor legislation. 4.32 Firms interviewed in the Investment Climate Survey were asked to identify constraints to business expansion. The main constraints mentioned were uncertainty regarding economic performance and policy, informal and unfair competition and the limited access to and highcost of credit as the most important constraints (Table 4.8). We have already discussed the issue of uncertainty (Chapter 2). We discuss informality in the next section and summarize the results on access to and the cost of credit presentedinthe Peru's Investment Climate Assessment (World Bank, 2003b) inBox 4.1. 79 Small Difference S Medium Difference Large (0to 10) and M (llto 99) MandL (loo+) Percentageoffirmsingroup Labor Workforce skill level 7.0 ** 16.0 * 10.1 Infrastructure Telecommunications 5.0 3.5 3.7 Electricity 10.5 11.3 10.1 Transport 8.0 7.4 6.3 Access to land 15.5 12.5 ** 1.2 Financialresources Access to (national)credit 50.7 ** 38.6 ** 22.7 Access to (international)credit 48.2 ** 34.1 29.1 Cost of credit 66.1 ** 58.8 I. 31.6 Economicenvironment Uncertaintyabout economicpolicy 69.3 74.2 ** 56.9 Macroeconomic instability 67.8 ** 59.3 ** 44.3 Corruption 59.7 58.2 49.3 Crime 58.2 ** 50.0 ** 36.7 Informalityandunfair competition 69.8 71.0 tli 58.2 Smuggling 67.8 62.1 ** 43.0 Source: Authors' calculations usingdatafrom the Investment ClimateAssessment (WorldBank, 2003b) Box4.1: Limitedaccessto and highcost of credit appearsto bea concernfor Peruvianbusinesses Limited access to and high cost of credit constitute one of the most important constraintscurrently facing Peruvian firms accordingto arecentWorld Bank report on the investmentclimate inPeru (World Bank, 2003b). Inthis Box we briefly summarizedthe report's results and recommendationsregardingthese problems. InPeru, financial markets have low penetration, with only 45 percent of firms having loans, in contrast with 85 percent in Malaysia. In addition, the cost of finance is very high, particularly for small and medium enterprises (SMEs), which pay high interest rates and must post high levels of collateral. Peru has improved dramatically over the last several years on the quantity and quality of information available to financial intermediaries. However, due to very high inefficiencies injudicial enforcement of contracts, the financial system continues to depend on high levels of collateral, particularly for new clients. This has the effect of stifling dynamismand entry of new firms and ensuringthat those who do manageto enter are less likely to grow quickly. A recent report by the IDB estimates that improvements in contract enforcement by the judiciary could lead to increases in sales or 25 percent or more and increases in the order of 9 percent for investment (Herrero and Henderson, 2003). Because asset registries also function poorly andjudicial proceedings for asset recuperationare complex and lengthy, SMEs typically cannot use their machinery (often their largest assets) as collateral for loans and thereforefind it difficult to invest, expand or upgradetheir technological capabilities. To improve access to credit, a reform of the moveable asset registries should be carried out to form a coherentintegrated system that allows the identification and prioritization of claims, and facilitates their rapid, low cost registration and search. Such a registry system should facilitate the reclamation of assets by all parties and not only banks. An efficient, well functioning asset registry combined with streamlineddebt collection proceedingscan help revolutionize credit markets in Peru and allow the creation of new instruments that can reach smaller firms. This in turn should free up a large fixed asset class (machinery and equipment), currently not eligible for collateralization for most borrowers, particularly SMEs. Source: World Bank (2003b). 80 Labor Legislationand Employment 4.33 We examine next the nature of labor legislation and the extent to which it impacts employment levels and employment creation. In analyzing the nature of existing legislation we distinguishbetween what the legislationi s intended to do (de jure effectiveness) and what the legislation actually achieves (de facto effectiveness) while in commenting on its impact we rely on the extensive literature already available. Finally we draw some policy recommendations based on the results presented on this section regarding employment creation and the role of labor legislation as a deterrent to further employment growth. The Nature of Labor Legislation: De Jure versus De Facto Protection 4.34 In order to examine the nature of labor legislation, or any legislation for that matter, it is important to distinguishbetween the levelof protection that the legislation intents to provide and the level of protection that it actually provides once the extent to which the legislation i s enforced is taken into account. This distinction is especially important when discussing potential changes to the legislation since well-intentioned principles may result into unintentional but inequitable outcomes once the law i s implemented. 4.35 Peru's legislation provides highlevels of employment and worker protection on paper-it offers good employment conditions and access to social protection and is protective of employment and of labor relationships. Botero et alia (2003) compare labor legislations across 85 countries focusing on three different dimensions: (i) employment regulation (available contractual forms, employment conditions and firing costs), (ii) relationships regulation (collective bargaining, participation of workers in labor management, conflict resolution mechanisms), and (iii) security regulation (access to pensions, social health and unemployment insurance). The authors measure a country's relative position using indices that vary from 0 (low protection) to 1(highprotection for each of the three elements within dimensions (inparenthesesabove) andfrom0to 3 for eachdimension3'. 4.36 Permanent employment conditions are good and protection i s high, compared to other countries inthe regionand the world (Table 4.9). For instance, Heckmanand Pages (2002) estimate averagefiring costs in 1999 to be equal to 13.8 average monthly wages, down from 15 in 1987, but still significantly higher than the regional average of 5.5 or the English-speaking industrialized countries average of 1.5. Similarly the legislationprotects labor relationships and, to alesser extent, access to social security (Table 4.10). Employment conditions Severancepay Average firing costs 0 (low protection) to 1or 3 (high protection) Monthly wages 0-3 0-1 Peru 1.67 0.70 13.8 Latin America 0.75 0.50 5.5 Industrialized countries 0.40A 0.12A 1.5 All countries 0.63 0.35 31. For adetailedmethodological discussionon the construction of these indices, see Boteroet alia (2003). 81 Table 4.10: Labor legislationprotects labor relationships andaccess to Social Security Labor relationships Access to Social Security 0 (low protection) to 3 (high protection) Peru 2.29 1.24 Latin America 1.45 1.75 Industrialized countries 0.53A 2.26 A All countries 1.24 1.67 4.37 These regulations, however, are relatively easy to avoid by using temporary rather than permanent contracts to hire new workers, or by operating in the informal sector. Temporary and hourly employment represents 20 percent of all private sector salaried employment and 50 percent of all contract-based employment in Metropolitan Lima (Jaramillo, 2004). This reflects that fact that restrictions on temporary or hourly hirings are significantly lower in Peru, with an index value of 0.28 (where 0 i s easy and 1i s difficult), compared to those in Latin American or the industrialized countries averages, with index values of 0.55 and 0.5 respectively. Inaddition we have already mentioned that the informality rate inurban areas is close to 70 percent. Income quintiles 1 2 3 4 5 Percentageof individuals ingroup Private sector (formal) contract 0.5 0.6 3.0 5.9 10.1 Private sector Social Security 0.5 2.5 5.4 10.5 16.1 Private sector Firmwith 20+ employees +++ 0.7 1.5 4.5 7.7 11.4 Table 4.12: Access to employment-linked socialprotection varies with type of contract Pensions 33.5 99.7 84.8 6.0 Bonuses 58.9 90.9 88.1 29.9 Individualunemploymentaccounts(CTS) 29.1 63.2 52.4 0.4 Table 4.13: Coverage of health andpension systems and of permanentsemployment varies by income level Income quintiles ~ 1 2 3 4 5 Percentageof individuals ingroup Healthinsurance I 100.0 10.0 18.3 18.9 20.6 32.2 Pensions 100.0 9.9 18.2 16.9 18.1 36.9 Permanentcontract I/100.0 8.2 16.6 19.2 19.2 36.8 Source:Jaramillo (2004a). 82 4.38 As a result labor markets inPeru are de facto flexible, although this flexibility comes at the cost of low and unevenly distributed employment protection and access to social security.32 Heckman and Pages (2002) conclude that only 17.9 percent of all urban employment and 51.9 percent of all salaried employment is hired in full compliance with Peruvian labor regulation, compared to 39.5 and 60.0 respectively in LAC. The authors also estimate that non-compliance with minimumwage equals 23.5 percent of the relevant population inPeru, while it i s 10.0inLAC. In addition to low levels, coverage i s unevenly distributed with higher access among non-vulnerable groups. MacIsaac and Rama (2001) analyze differences inaccess to severance pay inthe event of job loss by income quintile and by type of contract and show that it is significantly higher among richer workers, workers with access to social security and workers employed in firms with 20 or more employees (Table 4.11). Jaramillo (2004a explores differences inaccess to social protection by type of contract and income levels and finds similar patterns (Table 4.12 and 4.13).33 The author also shows that only a quarter of those entitled to received severance pay compensation actually do so due to the high costs of enforcement, in terms of both time and money, associatedwith bureaucratic and legalprocedures. 4.39 In sum, labor legislation in Peru offers high levels of protection de jure but low levels of protection de facto due to the highincidence of informal employment and the extended use of temporary contracts in the formal sector. This, however, does not mean that labor legislation does not have an impact of employment levels and, particularly on employment composition. We briefly discuss existing literatureon these topics next. The impact of labor legislation on employment 4.40 Several authors have analyzed the impact of various aspects of Peruvianlabor legislation on labor demand, employment tenure andjob tenure, the composition of total employment, and labor productivity makinguse of legislative changes inthe early 1990s as an identificationstrategy. Their main conclusions can be summarized as follows: Labor demand. Saavedra and Torero (2004) examine the impact of firing costs, measured as expected severance payments, on labor demand and show that it i s negative and significant, both at the fmand sector level. They also find that the magnitude of the coefficient decreased in the post- reform period, suggesting that labor market deregulation in the early 1990s effectively brought about a reduction in the cost of employment. Inaddition, they find evidence that labor market flexibility facilitated formal firms' adjustment to desired employment levels during the different stages of the businesscycle. Employment demand elasticity appears to have increasedduringthe latter part of the 1990s, though evidence i s not very robust. Also, it is not possible to establish the extent to which this phenomenon can be attributed to the reduction of firing costs or the larger utilization of temporary contracts (Saavedra and Maruyama, 1999). Price and output elasticities increased as well, likely due to the fact that the reforms made it easier to adjust to the desired employment levels. Employment turnover and job tenure. Saavedra and Torero (2004) also examine the extent to which increased labor market flexibility prompted an increase in turnover and a reduction of job duration, in particular in the formal sector. Using data for 1985-1997, they find that mean tenure startedto decline in 1992, coinciding with the beginning of labor market legislationchanges. Though the reduction inmean tenure may be related to a certain extent to the recovery initiated in 1993, the observed decline cannot be explained solely by cyclical movements of the economy. They also find that the fall in tenure was larger (and statistically significant) for formal workers than for informal 32. This dichotomy betweendejure and de facto protection levels is not exclusive to Peru and can be observedin many other countries in and outside the region were enforcement is weak and informality rates are high (Saavedra, 2004; World Bank, 2004e). 33. Jaramillo's figures are similar qualitative but different inmagnitudefrom those inTesliuc (2005), who shows that 1.6 percentof those inbottom quintile and25.1 of those intop quintilecontribute to pensionsystem. 83 ones, and that the differences in mean tenure between the formal and informal sectors fell significantly during the 1990s. In addition Saavedra and Torero (2004) shows that worker rotation increased considerably since 1993 and argues that this change cannot be explained by demographic growth or changes in sector distribution alone, as simulations of the evolution of employment mean duration over time controlling for demographic and sectoral characteristics yield no significant changes. He also shows that employee rotation was considerably larger for salaried than independent workers, who are not directly affected by legislation, and that, after the reforms, employment duration diminished at a faster rate for formal salaried workers than for any other group. These results indicate that the observed increase in employment rotation and decline in mean tenure can be attributed to changes inlabor legislation, and particularly to employers increasedability to use temporary contracts and to lower firing costs. a Composition of employment. Labor market reforms in the early 1990s, combined with the economic expansion that begun in 1993, translated into higher employment levels, both inthe formal and informal sectors. Positivenet employment creation duringthis period, however, hides significant changes in the composition of the labor force (Saavedra and Torero, 2004). Layoffs were biased towards older workers while hiringwas biasedtowards younger ones, since the relative costs of firing tenured, more expensive workers had decreased significantly after the reforms and younger workers could more easily adapt to new technologies. Inaddition, SaavedraandMaruyama(1999) alsoreport a significant shiftto temporary employment contracts after the reforms were introduced. Temporary employment increased from 19 to 4.4 percent of all contracts between 1986 and 1997, and it grew by 400 percent among industrial medium and large enterprises. Although these increases in temporary employment may seem puzzling in the context of increased hiringand firing flexibility and lower permanent employment costs, employers' decision to use these contracts widely may have been responded to various reasons, including (i) labor cost minimization, (ii) prevention of unionization, and particularly (iii) possibility that the the new legislation was revoked given the distinct lack of consensus building that characterized the environment inwhich labor reforms were formulated. Saavedra and Chong (1999) also report an increase in the number of informal workers during the 1990s, in spite of lower permanent employment costs and greater flexibility. The authors identify a number of factors that contribute to explain this rise, such as improvements in tax collection, demographic shifts and technological changes that affected the productivity of different types of workers and sectors. In addition, higher informality may have been employers' and workers' response to increases in non-wage costs in 1987 and in 1990 associated with changes in caps and minimums in several contributions and with the increase in the National Housing Fund (FONAVI) and inpension contributions. 0 Productivity. Changesinthe composition of the labor force appear to be correlated with a decline in labor productivity, even in the face of economic growth. The Investment Climate Report for Peru (World Bank, 2003b) presents evidence showing that total productivity is lower among enterprises with a higher percentageof unregistered workers. This result comes as no surprise, since total factor productivity grows as processes become more efficient and working practices improve overtime. Workers without a contract are typically the least experienced, the least skilled and least likely to receive significant training from the firm, which limits their potential contribution to total factor productivity growth. In addition Saavedra and Torero (2004) finds a significant positive relation between salaries and productivity in the years after the reform. Wages increased only in those sectors that experienced increases in productivity, and vice versa. The author concludes that the slow growth of labor productivity that characterized Peru during the 1990s was a major factor behind the slow rate of growth of both for formal and independent workers' wages. 84 Policy Recommendations 4.41 We have argued inthe two previous sections that employment legislation affects not only those employed in the formal sectors, but rather all firms and workers. Legislation intended, in principle, to protect workers can, in practice, result in low and inequitable levels of protection across workers employed in different sectors or under different types of contract. Inparticular a small group of permanent workers (insiders) enjoys high levels of protection at the expense of their non-permanent counterparts (outsiders). In addition labor legislation has an impact on the overall level of employment and on its composition. These results have important implications for policy making. 4.42 To a large extent these patterns result from differences in wage and non-wage costs between permanent and temporary employment. According to estimations by Peru's Ministry of Labor the hourly labor cost of permanent workers is 1.4-1.5 times that of temporary workers and 5-6 times that of workers with no contract (Ministry of Labor, 2004a). These differences reflect both higher wages and higher levels of protection (i.e. non-wage costs) among permanent workers than among their non-permanent counterparts. 4.43 We present below some policy options that can contribute to diminish the cost gap between permanent and other contracts. Intheory this gap can be closed by either increasing the cost of temporary employment, by decreasing the cost of permanent employment or both. However, inpractice increasing the cost of temporary employment without simultaneously decreasing that of permanent employment would only result inhigher relative costs of formal employment and higher levels of informality. 4.44 Moreover, although they do contribute to increase de facto flexibility, reforms that introduce flexibility at the margin but leave the main rules of the game unchanged do not constitute sustainable long-term solutions to the problem of excessive rigidity of permanent employment legislation. The experience of Spain after the introduction of temporary contracts in the early 1980s provides a good example inthis regard34.This implies that comprehensive rather than partialreforms mustbe undertaken. 4.45 Finally the implementation of either option will generate winners and losers and thus will face political economy constraints. This needs to be taken into account and compensatory measures put in place when appropriate andlor necessary. 4.46 Increasing the cost of temporary employment could require, among other factors: 0 The limitation of their use to temporary activities. This could be done by allowing the use of these contracts only under specific circumstances that justify doing so because of their temporary nature- for instance, replacing a sick or absent worker, or handling seasonal increases in the f m ' s activity. A reduction in the number of contractual modalities available could help make this process more transparent. 0 The limitation of contract duration. Limits could also be imposed onthe maximumduration of any specific contract and on the maximum number of consecutive contracts that can be offered to the same worker. It could also be stipulated that if this limit is surpassed the worker becomes automatically a permanent employee. Indoing this, however, caution must be exercised in order not mitigate incentives for employers to dismiss otherwise productive workers immediately before these limits arereached. 34. A comprehensive discussionof the Spanish experience can be found in Dolado, Garcia-Serrano and Jimeno (2002); Hernanz, Jimeno and Kugler (2003), and Garcia-serrano(1998) amongothers. 85 4.47 Inevaluatingthese proposals we must also take into account that certain groups of workers may be hurt by temporary contracts becoming more expensive. These are likely to be workers that we would consider vulnerable, such as women, young workers and the unskilled. To minimize the impact that these measures can have on these groups, it would be important to complement them with compensatory measures. These could range from targeted training programs to the introduction of special contracts (Le. apprenticeship, etc) aimed at easingtheir labor market insertion. 4.48 Decreasing the cost of permanent employment couldrequire, among other factors: Reducing firing costs. High firing costs discourage permanent hiring and lead to unequal access to social protection. Firing costs could be reduced in a number of ways. The most drastic approach would call for an actual reduction in the severance payments. Because this measure is likely to encounter significant opposition, it could be applied only to new contracts or, alternatively, compensation for workers hired under the previous regime could be considered. Softer approaches could include an increase of trial periods for new workers and a more flexible use of "economic reasons" as a cause for firing. Inadditiontothenegativeimpactoffiringcostsonemployment, complicated andcostlyenforcement procedures leadto low effective coverage. This could be mitigatedby allowing for private arbitration under certain circumstances-for instance in cases when firing occurs because of "economic reasons". Reforming personal unemployment accounts (CTSs). Personal unemployment accounts provide protection to the same group of workers that are covered by firing cost regulation. As a result the potential decrease inworker protectionbrought about by a reduction infiring costs couldbe mitigated by reforming individual unemployment accounts (Compensacion por Tiempo de Servicios, CTS). This reform should aimto restore the spirit under which CTSs were created. Currently workers have almost unrestricted access to the money in these accounts independent of their employment status so that their protection potential is highly undermined. The idea behind these accounts i s that they should help workers smooth income shocks associated with spells of unemployment. Hence a minimumbalance should be required before workers can have access to the money, and this access should be restricted to a portion of the remaining funds. This minimumbalance should in turn be a function of the average duration of employment and unemployment spells. Jaramillo (2004a) calculates that given the dynamics of Peruvian urban labor markets, this balance should be equal to four times the average monthly salary. In addition employer and employee contributions could vary depending on whether the balance inthe CTS is below or above this minimum. Reconsidering profit-sharing rules. Although in theory profit sharing-rules can be understood as incentives for higher productivity, their current implementation inPeru makes it difficult for them to function as such. The fraction of profits to which workers are entitled varies by sector, but this variation is not related to sectoral differences in productivity and rather i s regulated by law. This creates strong incentives for firms to resort to accounting tricks inorder not to declare profits. Infact only 2.9 percent of all workers covered by the system actually receive their share, so that the purpose of such rules is entirely deceived. To transform them into effective incentives, profit sharing rules should be negotiated between employers and workers (maybe within nationally fixed minimumand maximum levels) and linked to observable outcomes. Moreover agreements should be made public. Finally additional efforts should be madeto increase accounting transparency by eliminatingpotential loopholes incorporate taxation legislations and by promoting randomindependent audits. Reducing non-wage costs. The main components of non-wage costs in Peru are bonuses and stipulated (paid) vacations. While it could be argued that bonuses constitute delayed wages and as such are paid for the employees inthe form of lower monthly salaries, Peru i s among the countries in the region with the most generous legislation in terms of paid vacations, together with Brazil and Panama. Vacations have undoubtedly a positive effect on workers' welfare, but they can be costly if labor productivity i s low as i s the case in Peru. Moreover survey data indicates that only a small 86 percentage of salaried workers actually enjoys the stipulated vacation period suggesting that workers may be willing to sell leisure time for additional income. A more flexible approach could be adopted by allowing stipulated vacation to vary with worker's experience and, hence, with labor productivity since generally more experience workers tend to be more productive. 4.49 The Labor Commission of the Peruvian Congress has been working on a draft for a new labor code for the last couple of years and this project i s now under discussion. Some of the recommendations proposed above are already contemplated in this draft, and some others are not (see Box 4.2 for a discussion on the draft proposal). Moreover, even if this new proposal i s approved, it will be difficult to immediately implement all reforms since doing so will be costly and likely to encounter opposition from certain groups. Prioritizationcould then be established on the basis of a set of criteria that account for the reforms' costs and benefits. On the cost side, these criteria should include both the economic costs of setting up the necessary compensation mechanisms for those posed to lose from the reforms and the political economy costs associatedwith consensusbuildingand bargaining with all interested parties. On the benefit side, the criteria should include the reforms' potential impact on the cost gap between permanent andtemporary empl~yment~~as well as the incentives for formalization they may generate. Box4.2: The current proposalfor the Ley General del Trabajo The Labor Commission within the Peruvian Congress has been working on a draft for a new labor code for the last two years. In this box we briefly revise the proposals included in this draft regarding permanent and temporary contracts, and commenton their potential impact given the discussionpresentedinthis chapter. Permanent employment. The current draft contemplates an increase in firing costs for permanent workers, by increasing both the minimum and maximum compensations. It also extends the number of causes subject to severance pay to include those related to the firm's operations-i.e. economic, technical or structural reasons, and closing of business. Temporary employment. The draft proposesmeasuresto restrict the useof temporary contractsandareductionin the different types on contractsavailable. What could be the consequences of these changes?Simultaneously increasingthe cost of firing permanent workers and hiring temporary ones could lead to higher informality. It could increasejob duration, but also unemployment duration. Moreover it could lead to higher inequality inlabor market outcomes by granting more protection to those who already have a job while making it harder to find formal employment for those who do not. Finally, by increasing adjustment costs it could create disincentives for firms to innovate and adopt new technologies, thus reducing their productivity and long-run growth potential (Jaramillo, 2004a). INFORMALECONOMIC OPPORTUNITIES AND URBANPOVERTY 4.50 The performance of informal entrepreneurs and small businessesis intimately linked to poverty in urban areas. Approximately half of the working urban poor are self-employed, all of them informally, and an additional 30 percent work for micro or small firms. Similarly 40 percent of all informal entrepreneurs(self-employed or otherwise) are poor, compared to 15percent of formal entrepreneurs. 4.5 1 Identifying the determinants of the productivity of informal activities and implementing policies aimed at increasing this productivity are key in order to help the urban poor step out of poverty. This, 35. Several authors have tried to quantify the relative importance of the various components of wage and non- wage costs. For a comprehensive discussion on this issue, the reader can consult Jaramillo (2004a) and Ministeriode Trabajo (2004a). 87 however, must be done with the understanding that informality i s costly for households, businesses and the government. In this section we tackle both issues and propose a series of policy interventions aimed at, on the one hand, increasing earnings among those in the informal sector and, on the other, reducing informality. Productivityof Self-Employment and Small Businesses 4.52 We study the determinants of productivity of informal self-employment and small businesses, as well as the relationship between value-added, wages and employment in the case of the later using data from the "trabajador independiente" module of the 2003 ENAHO. This module asks self-employed individuals and small entrepreneursa battery of questions about the organization and productivity of their business. 4.53 Using this information we discuss the differences between poor and non-poor entrepreneurs in terms of their reasons to be informal and the average productivity of their business. We measure productivity as value-added or value-added per worker, which we construct using information on sales, self-consumption and costs of production inputs provided in the survey. This measure is, of course, imperfect36and, as a consequence, the results discussed here should be interpreted as being suggestive of (qualitative) association among variables rather thanof causal (quantitative) relationships. 4.54 Involuntary informality i s more prevalent among the poor than the non-poor but, among those who chose to be informal voluntarily, the reasonsto do so are similar across bothgroups. Fifty percent of the poor declare to be informal because they could not find employment inthe formal sector, compared to 35 percent of the non-poor. Conditional on being voluntarily informal, however, the differences between both groups are very small. Approximately half of those who say they chose to be informal did so because their earnings were higher than they would have been in the formal sector, and an additional 30 percent claimed to prefer the informal sector because of its higher flexibility (Table 4.14). Tabie 4.14 Involuntary informalityis moreprevalent amongthe poor than the non-poor, butthe reasonsto beinformal are similar across bothgroups Poor Non-poor Involuntary (Couldnot find formalemployment) 47.9** 35.8 Voluntary 52.1*' 64.2 Obtains higherearnings 47.8" 50.3 Hasmoreflexibility 28.2 28.7 Runsfamilybusiness 7.6` 6.3 Other 16.3* 14.2 4.55 Average levels of productivity vary across sectors and with market size, while within sectors and markets businesses managed by the poor are less profitable than those managed by the non-poor. Monthly value-added i s highest in the manufacturing sector and lowest in the commerce sector. In addition, businesses operating in larger markets3' are more productive than those servicing smaller ones, although the differences are insignificant for areas with population under 10,000. Differences also exist within sectors and areas between poor and non-poor entrepreneurs. Inparticular businesses run by the 36. Our measure of profits or value added does not account, for instance, for the depreciation of capital, such as machinery or vehicles, used in production, or for the rental costs of land or housing when the business is operated from the entrepreneur'shome. 37. We assume that all businesses service their area of residenceso that the size of their potential market can be proxiedby the size of the survey strata inwhich the householdresides. 88 poor produce less value added than those run by the non-poor in all sectors but manufacturing and irrespective of market size (Table 4.15). Table 4.15: Businessmanagedby poor entrepreneurs are less profitablethan those managedby non-poor ones 1 All Poor Non-poor All businesses 235.7 149.1'* 281.9 Economic sector Manufacturing 420.5 348.9** 412.2 Commerce 142.5 81.2** 173.2 Services 258.6 127.5** 328.0 Market size Morethan 100,001 households 381.6" 201.7"' 463.0 Between20,001 and 100,000households 219.7" 140.7** 258.4 Between 10,001and20,000 households 132.7$ 107.0'* 147.9 Between4,001 and 10,OOO households 125.9 131.0 122.4 Between401 and4,000 households 154.8 135.7" 171.8 4.56 These differences have important implications in terms of welfare, both for the entrepreneurs themselves and for their employees. Business value-added can be thought of as labor income from self- employment in the informal sector, so that an increase in value-added directly translates into an increase inhousehold income. Similarly, becausemore profitable businesstend to employ more workers andpay higher salaries, an increase in value-added also leads to an improvement in welfare among informal salaried employees. In particular an increase in value-added of 926 Nuevos Soles, equivalent to one paid workers after controlling for worker and business characteristic^.^^ standarddeviation, is correlated with a 3 percent increase infirmsize and a 2 percent increaseinwages of 4.57 The question then arises as to what explains value-added levels and, consequently, the differences in value-added that we observe between poor and non-poor entrepreneurs. Are they a function of the demographic characteristics of entrepreneurs? Of the way in which they operate their business? Of the quantity and quality of the inputs they use? After examining existing differences and commonalities between both groups and using regression analysis to identify the determinants of value-added, we will conclude below that the answer is a combination of all these factors. Dgferences betweenpoor and non-poor entrepreneurs 4.58 The most important difference between poor and non-poor entrepreneurs in terms of their personal characteristics arises from their respective level of education. Forty percent of all poor entrepreneurshas at most completed primary education and 11percent has completed tertiary education, 38. These figures are obtained as follows. We regress both the number of workers in the firm and (log) hourly wage of paid workers on value added, entrepreneurcharacteristics(demographic characteristicsand education level) and business characteristics (sector of activity, location, access to markets, and inputs). The wage regressionalso includes information on worker demographics and educationlevel. The coefficient onprofits in the employment equation is 0.0001, significant at the 1 percent level. Multiplied.by 926 Nuevos Soles, this yields a change firm size of 0.09, or 3 percent of the average firm size for business with 1or more employees (2.6). The coefficient on profits inthe wage equation is 0.00003,significant at the 1percentlevel. Multiplied by 926 Nuevos Soles, this yields a change inhourly wages of 2.7 percent. 89 compared to 20 and 30 percent of all non-poor ones respectively (Table 4.16). Besides educational differences both groups appearto be fairly similar interms of their demographic composition. Table 4.16: Non-poor entrepreneursare moreeducated... Poor Non-poor Percentageof individualsingroup Headof household 42.6** 39.6 Female 53.3'* 55.3 Femaleheadofhousehold 12.2 12.0 Age 37.0** 38.1 Incompleteprimary 22.1" 10.9 Completeprimary 17.0** 11.1 Incompletesecondary 22.2" 12.5 Completesecondary 27.1" 31.8 Tertiary 11.4** 29.5 4.59 Regardingbusiness operation, non-poor entrepreneursappear to be better integrated into markets than poor ones in the sense that they more often have access to (commercial) outlets and follow more market-oriented practices. Non-poor entrepreneurs are more likely to operate out of a outlet, be it in a public or private market or at home, thanpoor ones. The difference inaccess i s the result of lower levels of ownership among poor entrepreneurs, combined with their limited capacity to pay rental costs. Similarly non-poor entrepreneursare more likely to keep some form of business accounts, either formally or informally, and to employ more paid workers than poor ones. Interestingly these differences do not result from differences inthe fraction of poor andnon-poor entrepreneursthat are self-employed nor from differences infirm size among those who runsmall businesses (Table 4.17). Table 4.17: ...haveaccessto commercial outletsand follow market-orientedpracticesmore often,... I Poor Non-poor Percentageof businessesingroup Access to commercialoutlet No outlet 64.3** 54.8 Privatelpublicoutlet 8.8" 14.7 Owner 23.2 23.8 Other (e.g. rental) 76.8 76.2 At home 26.7** 30.3 Owner 78.5 80.2 ' Other (e.g. rental) 21.5 19.8 Businesspractices No (formallinformal) 82.7** 72.3 accounting Sue of business Self-employed(%) 69.4 70.7 1or moreemployees 30.6 29.3 Total number of workers 1.4 1.4 Share of unpaid(family) 86.2" 77.0 workers 4.60 Non-poor entrepreneurs also seem to use more and better inputs and to have more access to infrastructure than poor ones. A larger fraction of non-poor entrepreneurs declares to employ and own machinery andlor furniture, but these differences disappear once we account for the fact that non-poor 90 entrepreneurs are more likely to operate out of a outlet than poor ones. No significant differences exist regarding the use of vehicles or smaller tools. Inaddition firms runby non-poor entrepreneursemploy a smaller share of female workers and a larger share of more educated workers (Table 4.18). Finally non- poor entrepreneurshave higher access to water, electricity and phone services thanpoor ones, irrespective of whether they conduct their business inthe street, ina formal outlet or at home (Table 4.19). Table 4.18: ...usemore c ita1and skilled labor, ... Poor Non-poor Percentageof businessesingroup Machinery 18.8" 26.2 Operates frompubliclprivate outlet 38.7 40.0 Owner of asset 88.0* 90.3 Owner of asset (inoutlet) 87.3, 86.9 Furniture 25.7*' 33.4 Operates from publiclprivate outlet 53.4** 61.3 Owner of asset 91.5 91.2 Owner of asset (inoutlet) 90.0" 85.0 Vehicle 17.7 16.2 Operates from publiclprivate outlet 4.8 3.6 Owner of asset 54.2" 62.7 Owner of asset (inoutlet) 91.6 91.3 Tools 56.0 56.8 Operatesfrompubliclprivate outlet 69.3* 64.0 Owner of asset 90.9** 92.4 Owner of asset (inoutlet) 94.7 92.5 Percentageof workers infirm(excluding employer) Characteristicsof labor force infirm Shareof female workers 54.3** 49.1 Share of workers with secondary or higher education 68.8" 78.4 Share of workers with tertiary education 9.5** 20.8 Average numberof years of tenure infirm 2.7 2.8 Note: "'"Significantly different from average for non-poor at 5 (10) percent level. Poor Non-poor Percentageof businesses ingroup Water 11.5** 15.7 Operatesfromprivatelpublic outlet 35.5 37.3 Operatesfromhome 31.O* 33.7 Electricity 22.3** 31.8 Operatesfrom privatelpublic outlet 54.2** 62.0 Operatesfromhome 66.3" 75.2 Telephone 0.4'; 2.8 Operatesfromprivatelpublic outlet 1.6*' 5.0 Operatesfromhome 1.2 6.8** 4.61 Interestingly these differences in business practices and characteristics are not the product of independent factors but rather are inter-related. Inparticular the use of market-oriented practices, and access to capital and infrastructure are correlated with business location. Businessesthat operate out of a commercial outlet are more likely to use some form of accounting and to employ a larger share of paid workers than businesses that operate inthe street or out of the entrepreneur's home. Inaddition access to 91 machinery and other tools i s higher among businesses in commercial or non-commercial outlets than among those runinthe street, while the use of a vehicle is much higher among the latter-partly due to its use as a substitute for a proper outlet. Finally runninga business from a commercial outlet i s correlated with higher access to a phone and water (Table 4.20). No outlet Operatesfrom publidprivate outlet Operatesfrom home Rental Owner (FormaYinformal) accounting 21.3 46.3 49.2 32.7 Share of paid workers 6.6 19.9 18.1 6.9 Accessto capital Machinery 10.7 43.6 43.7 44.6 Vehicle 26.1 4.2 4.4 3.5 Tools 49.1 63.2 61.5 66.7 Access to infrastructure Water 42.9 42.9 34.9 Electricity 70.1 66.7 75.7 Telephone 11.9 10.2 8.7 4.62 Although they make it clear that poor and non-poor entrepreneurs conduct their businesses in different ways and under different conditions, the simple tabulations presented here do not allow us to identify what the effect that each of the variables considered i s on value-added, either independently or once other factors are taken into account. As a result they can not be used to disentangle the relative importance of these variables in terms of explaining differences in productivity between poor and non- poor entrepreneurseither. We useregression analysis below to tackle these issues. Determinantsof Value-Addedper Worker 4.63 Using an OLS model we analyze the determinants of value-added per worker (i.e. value-added per worker). We do this in two steps. We first consider the effect of different sets of variables (demographic characteristics and education level of the entrepreneur, market size and business practice indicators, and business characteristics) in a series of partial models, and then estimate a full model that includes all variables. Our results can be summarized as follows (Table 4.21): Demographic characteristics and education: Businesses run by household heads, male, older or more educated entrepreneurs have higher level of value-added per worker. The same i s true about businessesrun by individuals who are voluntarily informal (Model 1). The relationship between the demographic characteristics and education level of the entrepreneur and the firm' value-addedremain when all determinants are consideredjointly in the full model, although the effect of lower levels of education becomesinsignificant (Model 4). Marketsize and businesspractices: The larger the market the businessserves inthe higher the level of value-added per worker (Model 2). These differences remain significant once other factors are taken into account (Model 4). Similarly keeping some form of business accounting and employing a larger share of paid workers (not shown)39are both positively correlated with profit levels in the partial and full model. 39. The effect of employment of paid workers on value-added per worker is estimated using a restricted sample of firms with two or more workers and equals 168.70 (with a standard error of 22.59). 92 Table 4.21: Value added per worker is a function of entrepreneur and businesscharacteristics, and of market size and businesspractices Model 1 Model2 Model3 Model 4 Demographic characteristics Dependentvariable: Value addedper worker (Nuevos Soles) Householdhead 43.03*' 34.88** (17.38) (17.32) Female -136.38** -127.99** (16.30) (17.13) FemaleHHhead -21.31 -10.57 (23.09) (23.01) Age 8.34" 7.31" (2.03) (2.03) Age squared -o.oa*' -0.07** (0.02) (0.02) Education Incomplete primary Baselinecategory Complete primary 1.51 -7.15 (18.44) (18.33) Incomplete secondary 42.30** 21.49 (18.36) (18.35) Complete secondary 53.94'* 26.46 (16.86) (16.95) Tertiary 190.21** 145.19** (17.54) (17.98) Voluntarily informal 54.04** 46.87** (10.50) (10.54) Access to markets More than 100,001 households Baselinecategory Between20,001 and 100,000households -18.71 -8.84 (20.08) (19.69) Between 10,001 and20,000 households -108.71** -92.63** (25.11) (24.53) Between4,001 and 10,000 households -98.62*' -68.65" (24.31) (23.76) Between401 and4,000 households -82.03** -38.64 (24.78) (24.37) (Formdinformal) Accounting 127.65** 110.19" (11.33) (11.84) Businesscharacteristics Self-employed 55.71** 70.19** (11.60) (11.61) No outlet Baselinecategory Operates from privatelpublic outlet (all) 109.49'* 58.11** (15.89) (15.64) Operatesfrom home -8.67' -4.56 (12.98) (13.02) Uses machinery 51.85** 28.22** (12.69) (12.43) Uses vehicle 58.71** -3.90 (15.63) (15.99) Usestools -30.97** -30.63** (11,09) (10.93) Sector of activity Manufacturing Baselinecategory Commerce -131.36** -86.89** (16.34) (16.81) Services -67.46** -39.09** (16.21) (16.15) Regiondummies Yes Yes Yes Yes Number of observations 9.041 9.041 9.041 9.041 Nore: ""'Significantly differentfromzero at the 5 (10) percent le Source: Authors' calculations using data fromENAHO2003 (INEI). 93 Business characteristics: One-person businesses have higher profit levels per worker than those with two or more workers in both types of models. In addition value-added per worker are significantly higher among business operating from formal outlets, and among businesses that make use of machinery or vehicles, even after controlling for location. All effects, with the exception of vehicle use, are robust to the inclusion of other variables in the model. The level of education of the firm's work force, measured as the share of workers with secondary or higher education, i s also positively correlated with value added per worker (not shown).40 Finally average profit levels in the commerce and services sectors are lower than those in the manufacturing sector in both models (Models 3 and 4). 4.64 These results suggest that differences in productivity levels between businessesrun by poor and non-poor entrepreneurscan be attributed to differences inthe level of education of the entrepreneur, inthe size of the market the firm has access to, in the extent to which business practices implemented by the firmare market-oriented, and inother business characteristics. Consequently policies that can help poor entrepreneurs overcome one or more of these barriers can go a long go in easing urban poverty. We discuss some of these polices at the end of this section. The cost and causes of informality 4.65 Important as it i s to support poor informal entrepreneurs, it has to be done with the understanding that informality i s costly for households, firms and the government. Low productivity among informal businessestranslates into lower earnings for those employed inthe sector-average hourly labor income inthe informal sector is 50percent below that of the formal sector, even after controlling for worker and job characteristic^.^' Informal workers also lack access to employment-linked social protection, such as health and pension benefits or severance payments inthe event ofjob loss. And although some of these workers may have voluntary foregone such benefits inexchangefor higher wages or more flexibility, we noted above that more than 50 percent of the informal poor are so involuntarily. Similarly non- compliance with labor and sales taxes among informal firms has a negative impact on fiscal revenues. Finally high informality rates are associated with important costs for both formal and informal firms, as discussedinBox4.3. Box 4.3: Informality iscostlyfor bothformal and informal firms. According to the results of a recent Investment Climate Survey conducted in Peru by the World Bank, formal employers rated informality the second most severe constraint to business growth and development. Subsequent analysis of the survey's data showedthat infact firms with larger numbers of informal competitors were less productive than those facing lower levels of informal competition. These productivity losses were caused by formal businesses operating at a disadvantage when competing with informal ones, which manage to reduce their production costs by not complying with existing regulation. For instance, a firm could lower its costs by about 40 percent by not paying non-wage benefits to its workers and avoiding sales taxes. Despite the relative advantage of reduced production costs, informal firms also faced important costs that hinder their potential for growth. Often they had limited access to credit from formal financial or other institutions, which negatively impacted their capacity to invest in innovation through new machinery and equipment. They also lacked the means to protect their property rights, business transactions and contracts, as well as the incentives to invest in training. All these limitations jeopardized the capacity of informal firms to grow. As aresult, firms with alarge share ofunregisteredworkers were found to be less productive than their counterparts, even after controlling for other observable differences. I Source: World Bank (2003b). 40. The effect of the education level of the work force on value-addedper worker is estimated using a restricted sample of firms with two or more workers andequals 34.86 (withastandarderror of 20.19). 41. This figure corresponds to the estimated coefficient in an earnings regression of a dummy variable taking a value of 1 for those employed in the informalsector. The modelalso includes worker andjob characteristics, andcontrolsfor type of employment (e.g. salaried, self-employed). 94 4.66 Long and costly business registration procedures, or red tape, and rigid labor regulation are reported by employers to be the main reasons for informality (World Bank, 2003b). Although the creation of the Unified Business Registry in 1990 dramatically reduced red tape, registration remains a slow process that can take up to 100days to complete, compared with 43 days in Colombia and 27 days in Chile. In addition the monetary costs of registration are relatively high compared to that of other countries with which Peru competes with in international markets (e.g. \the cost of registration equals approximately 36 percent of per capita income inPeru, but only 27 and 10percent inColombia and Chile respective~y).~' Table 4.22: The cost of startinga businessinPeruis significantlyhigher than inother countriesinand outside the region Peru Ecuador Colombia Chile UnitedStates Number of procedures 10 14 14 9 5 Time (indays) 98 92 43 27 5 Cost (% of income per capita) 36.4 47.4 27.4 10.0 0.6 Minimumcapital (% of incomeper capita) 0.0 10.4 0.0 0.0 0.0 4.67 Of the procedures needed to start a business, obtaining a business license from the local municipality i s by far the costliest process both in terms of time and money, taking on average up to 34 days and costing US$439 (World Bank, 2003b). Furthermore this process appears to be particularly onerous for small firms, which can spendupto 43 days getting their operating license. 4.68 Labor regulations, although intended to protect workers, can also constitute a significant obstacle to employment creation and, more generally, business growth. As discussed above, Peru has heavy regulations regardingfiring costs and non-wage compensation payments despite the reforms implemented inthe early 1990s. Theseregulations can increaselaborcosts by upto 70percent43,and althoughitcould be argued that these costs can be absorbedby formal workers inthe form of lower wages, a recent study by the Inter-American Development Bank (2001) shows that a 10-percent increase in mandatory non- wage costs rises the costs of labor between 3 and 7.5 percent. 4.69 The increasesinthe cost of labor implied by compliance with current regulationcan be relatively heavier for small firms because of their lower average level of profits and labor productivity. Profits and labor productivity, measured as value-added and value-added per worker, are significantly lower in small firms, with fewer than 10 employees, than in medium and large ones.44 Moreover, although wages paid by these fmare below those paid by mediumand large ones to reflect differences inlabor productivity, their wage bill tends to account for a larger percentage of total cost becauseof lower capital-labor ratios. As a result proportional increases in labor costs as the ones generated by various labor regulations can have a higher impact on the overall productivity and make incentives for informality stronger among small fm. 4.70 It is then important to consider policy interventions aimed at reducing red tape and labor costs and thus at attracting informal business into the formal sector. Becausemost of the poor are employedin 42. Other authors have measuredthe cost of registration in terms of time and money. De Soto (1986) estimated that in the 1980s the registration process took up to 420 days and cost US$1,200. More recently Jaramillo (2004b) estimates these figures to be 70 days and US$117 based on data from the Encuestade la Micro y PequefiaEmpresa(2003). 43. This figure is calculatedas follows: Christmasand National Holiday bonuses (16.6 percent), paid vacation and holidays(13.3 percent), tenurebonus(9.7 percent), health and accident insurance(16 percent), training fund (1 percent), andother bonuses and contributions (13 percent). 44. According to data fromPeruInvestmentClimate Survey (World Bank, 2003b) 95 small firms, which exhibit higher rates of informality, interventions targeted towards these fm may yield higher pay-offs interms of poverty reduction and extended (though maybe limited) protection to the poor. We discuss some suchinterventions below. Policy Recommendations 4.71 Inthis section we have focused on informal economic opportunities. We have analyzed their productivity and, more generally, discussed the costs and causes of informality. We have concluded that, given the relative importance of labor income from informal sources as a fraction of household income among poor households, interventions aimed at increasing the productivity of informal activities can be an effective way of improving the welfare of these households. We have also argued, however, that informality i s costly for households, firms and the governments, so that these interventions should be complemented with others aimed at providing incentives for formality, particularly for small firms. We briefly discusshere both types of interventions. 4.72 Increasing the productivity of informal self-employment and small businesses will require, among other factors: 0 An increase inthe level of skills of both entrepreneurs and salaried workers. General increases in the skill level of the labor force can be achieved by investing in formal education (discussed in Chapter 5) andor by improving the relevance and coverage of the training system. There i s significant evidence for Peru and for the region that training, although costly, pays off in the form of higher labor productivity. The existing system, however, fails to provide relevant training for most workers. Training does not need to be provided by the public sector, but rather incentives can be put in place by the government for firms to contract the desired training with private, and properly accredited, providers. These incentives can rely on making training expenses deductible or on (partially) matching training resources invested by fm,and can be targeted to informal fm by conditioning them on easy-to-monitor indicators such as firm size or worker education levels. The ProJoven program, which provides training for young workers, replicates some of these principles andcould beextendedto cover other demographic groups. 0 An increase in access to commercial outlets and in use of market-oriented practices. Commercial spaces for small businesses in markets or other locations could be offered to those operating in the street in exchange for a rental (leasing) fee. This fee can be made to increase over time to both facilitate early investment and reflect potential future gains in productivity. Increased access to such spaces has important advantagesthat can make them cost effective. On the one hand they contribute to the decongestion of those streets and areas where these businesses would otherwise operate, easing traffic and decreasing hazards. On the other hand these spaces can be used as a platform for the economical provision of basic infrastructure and business services, such as management and accounting practices, simplified access to credit, and legal services, which in turn translate intohigher value-added. Given that these initiatives should be self-sustainable over time, the role of the public sector, and particularly of local authorities could be that of a catalyzer, helping attract potential invertors, and a coordinator rather thanthat of direct provider. 4.73 Improving incentives for firms, particularly small fums, to become formal will require, among other factors: 0 Further simplification of registration procedures. A reduction of red tape to bring the cost of registrationprocedures in line with those of close competitors in and outside the region will make it easier for fmto comply with these requirements. Recently approved legislation that considered the implementation of a special, simplified registration regime for micro and small f m s constitutes a 96 step in this direction. This legislation proposes several measures aimed at reducing the time needed for and the monetary cost of obtaining operating licenses from municipalities. Unfortunately the actual implementation of this and other changes contemplated in the new law has so far not taken place. The drafting and approval of a `reglamento' that makes the law operational could accelerateits implementation. A word of caution is necessary, however, regarding the potential impact of a simplification in registration procedures in the absence of other reforms. If the overall benefits from being informal are perceived to outweigh those of being formal it is unlikely that decreasing a one-time cost (Le. the cost of registration) will attract a large number of informal firms into the formal sector. Therefore simplifyingregistrationprocedures to make themcheaper should be accompanied by othersmeasures that make formality more attractive, such as the ones discussed above regarding labor legislation and the ones presentedbelow. The simplificationof tax filing mechanismsfor smallfirms. A special, simplified filing regime for micro and small firms already exists, but these firms could benefit from further simplification. For instance, filing on the basis of readily observable business characteristics and according to pre- determinedtax tables could be considered. These systems make it easier to file taxes for fmthat do not rely on fully formal, and often costly, accounting and even firms that interact with a large number of informalpartners. The implementation of a special labor regime for small firms. The new legislation mentioned above also included measures to allow micro and small firms to pay lower contributions to the pension and health systems and to reduce the duration of paid vacation. Again accelerating the effective implementationof this legislation could help increating incentives for formalization among these firms. 4.74 Inany case and regardless of the type of incentives chosen, it is important to point out that they must be appropriately packaged to minimize monitoring costs and gradually phased out so as to provide incentives for initialtake-up and for future firmgrowth. CONCLUSIONS 4.75 In this chapter we have argued that the evolution of urban poverty is closely linked to the functioning of urban labor markets since labor income constitutes the main, and frequently the only source of income for urban households. 4.76 The poor tend to be informally self-employed or to be employed in small (informal) business, so that policies aimed at increasing the productivity of informal activities can help reduce urban poverty. However, because informality is associated with lower levels of productivity, these policies should be combined with others that make it attractive for informal businesses to become formal. 4.77 More broadly future sustained improvements in urban poverty will require further employment creation and income generation in those sector that tend to employ the poor. More flexible labor legislationcan contribute to this goal. 97 98 5. ECONOMIC OPPORTUNITIES FOR THE RURAL~ 0 0 ~ ~ 5 5.1 Poverty is more widespread and significantly deeper in rural than in urban areas. Seventy two percent of the ruralpopulationi s poor and 40 percent is extremely poor, compared with 40 ands 8 percent respectively in urban areas. The poverty gap in rural areas, at 28 percent, is more than double that of urbanareas. 5.2 In Peru, as elsewhere, the income of poor households has traditionally been tied to the agricultural sector. Recently, however, the non-agricultural sector has become a more prominent source of income and employment in rural areas, both for poor and non-poor households. This chapter is devoted to the analysis of rural poverty in Peru with a focus on income-earning opportunities for households inruralareas. The chapter explores the distribution of households across both sectors, as well as its implications, and analyzes the potential impact of various public interventions on rural income and poverty. 5.3 The rest of the chapter i s structured as follows. The fEst section briefly explores differences across regions interms of the incidence of ruralpoverty and its responsiveness to economic growth. The second section examines the contribution of agricultural and non-agricultural income sources to total household income and the relationshipbetweenemployment ineach sector andpoverty. The thirdsection studies the determinants of participation in and the returns to agricultural and non-agricultural activities. The fourth section builds on these results and analyzes the potential impact of public interventions on rural income and poverty, by simulating their effect on the participation in and the returns to different economic activities. Finally, the fifth section outlines, based on the results of the chapter, the basic necessary elements for an effective development strategy aimed at promoting inclusive growth in rural areas. 5.4 The main findingsof the chapter can be summarized as follows: There exist significant differences across geographic regions in terms of the nature of rural poverty and its responsivenessto growth. These differences respond to variation in household characteristics and endowments acrossregions, andto the extent to which rural areas are integrated with urbanareas and innationalmarkets. The average rural household obtains most of its income from agricultural activities, but important differences exist between poor and non-poor households in terms of their income-generating strategies. Poor households tend to rely on agriculture, while non-poor households tend to engage in non-agricultural activities. Moreover, poor households are more likely to rely on a single source of income, while non-poor households are able to better diversify income risk by not relyingexclusively on one particular source. Participation in and the returns to these income-generating strategies are a function of household characteristics and endowments, access to markets and of policy levers. Both agricultural productivity and labor income are positively correlated with human capital, access to credit and public investment inbasic services, telecommunications and roadinfrastructure. The impact of public investments is higher for the non-poor than for the poor as a result of the better quality of their endowments and their higher degree of market integration. This difference, however, becomes smaller when two or more public interventions are implemented simultaneously inthe same area due to the existence of complementarities across interventions. A strategy aiming to promote inclusive rural growth must then consider policies directed towards increasing the endowments of the rural poor and improving access to markets, basic services and 45. This chapter is based on background work prepared by the report team and on existing work by Escobal and Torero (2003) andEscobal(2003). 99 infrastructure. This strategy should be responsive to regional heterogeneity, be comprehensive in order to encompass both the agricultural and no-agricultural sectors, and be spatially integrated to account for complementarities across interventions RURALPOVERTY INPERU: A HETEROGENEOUSREALITY 5.5 Rural poverty is higher and deeper than urban poverty all over Peru, but there exist significant differences across geographic regions and departments in terms of poverty rates. Seventy two percent of the rural population i s poor and 40 percent i s extremely poor, compared with 40 ands 8 percent respectively in urban areas. These numbers, however, mask a substantial amount of variation across regions, with rural poverty rates being lowest in the Costa and highest in the Sierra, and even across departments, with rural poverty ratesbeing lowest inMadre de Dios and highest inHuanuco (Figure 5.1). Figure 5.1: There existssignificantvariation inruralpoverty ratesacrossdepartments. Rural Poverty rates in Peru (Confidence intervals of regional rates- 2002) 95% 90% 85% 80% 75% 70% 65% 60% 55% 50% 45% 40% 35% 30% 25% 20% 11 Source: Authors' calculations usingENAHO 2002.W (INEI). 5.6 Similarly there are important differences across regions inthe extent to which poverty respondsto economic growth. We use (regional) estimates of the output-elasticity of poverty to examine these differences. According to these estimates, a one-percent increase in output would lead to a 1.03-1.38 percent decline in rural poverty at the national level, making rural poverty slightly less responsive to economic growth than urban poverty (Table 5.1). At the regional level, rural poverty appears to be most responsive to growth inthe Costa (output-elasticity between -0.941 and -1.323), and least responsive in the Sierra (output-elasticity between -0.559 and-0.873). 5.7 These regional differences reflect variation in the depth of poverty since the further away from the poverty line the poor are, the more growth will be required to reduce poverty. The poverty and extreme poverty gaps are larger inthe Sierra and Selva thaninthe Costa. 5.8 They also reflect differences across regions in household characteristics, endowments and the degreeof market development. The presenceof indigenous populations i s higher and education levels are lower in the Sierra than in other regions. Access to public services, such as schools and roads, i s also more inadequate in the Sierra, both for poor and non-poor households, than in the Costa and the Selva. 100 Finally urban and rural areas are more integratedeconomically inthe Costa than inthe rest of the country. These differences are robust to controlling for other observable regional characteristics (Tables A.7, A.8 and A.9 inthe Statistical Annex). All rural RuralCosta RuralSierra RuralSelva All urban Output-elasticityof poverty 1997 -1.283 -1.092 -0.737 -0.937 -1.367 (0.074) (0.099) (0.055) (0.085) (0.052) 1998 -1.385 -1.046 -0.699 -1.296 -1.358 (0.074) (0.091) (0.051) (0.107) (0.056) 1999 -1.221 -1.161 -0.647 -0.995 -1.358 (0.102) (0.121) (0,081) (0.113) (0.061) 2000 -1.366 -1.323 -0.873 -0.915 -1.393 (0.107) (0.150) (0.084) (0.130) (0.080) 2001 -1.035 -1.176 -0.559 -0.720 -1.364 (0.062) (0.141) (0.040) (0.074) (0.046) 2002 -1.087 -0.941 -0.680 -0.755 -1.274 (0.049) (0.080) (0.037) (0.059) (0.043) Note: Standarderrors inparentheses. Source: Duclos, EstebanandRay (2003). 5.9 In sum, the reality of rural poverty across Peru is heterogeneous and does not easily fit a single model-something we will come back to when we discuss the basic necessarily elements for a successful rural development strategy. INCOME POVERTYINRURALAREAS SOURCESAND 5.10 Rural households obtain more than three quarters of their income through the use of their labor and other productive assets, such as land or livestock, inagricultural and non-agricultural activities. As a result the sector and type of employment households have access to bears an impact on householdincome and poverty. On average households that derive most of their income from agricultural activities, be it salaried or non-salaried, have lower income levels and higher poverty rates that those who have access to non-agriculturalactivities (Table 5.2). Table 5.2: Poverty ratesare higher amonghouseholdswho dependon agricultural activitiesthan those who do not Extreme poverty Poverty Agricultural Salaried 48.7 80.7 Self-employed 53.3 79.9 Non-agricultural Salaried 24.8 52.8 Self-employed 41.6 71.9 5.11 In this section we analyze the relationshipbetween the main sector and type of employment and income. We examine the relative importance as sources of income of salaried and self-employment in agriculturaland non-agriculturalactivities and the extent to which they are substitutes or complements. 101 Income DistributioninRuralAreas: Patternsand Trends 5.12 Agricultural activities are the main source of income for the average rural household, but their relative importance varies with income (Figure 5.2). Agricultural activities, especially agricultural self- employment, account for 40 percent of all household income, compared to 30 percent for non-agricultural activities and an additional 30 percent in the form of capital income and private transfers (e.g. remittances). Agricultural activities, however, become less important as household income rises. They account for almost 60 percent of total income for households in the first quintile of the rural income distribution (i.e. the poorest 20 percent of the population) but only for 30 percent for household in the fifth quintile (i.e. the richest 20 percent of the rural population). This decline in the importance of agricultural income i s accompanied with an increase inthe importance of non-agriculturalincome, which accounts for 9 and 45 percent of household per capita income among the poorest and richest households respectively. Figure 5.2: The relativeimportance of different income sources varies with household income... I 1 2 3 4 5 1 Note: Quintilesdefined usingper capita householdincome. Source: Authors' calculations usingdata fromENAHO 2002.W (INEI). 5.13 Inaddition richer households seemmore able to diversify income risk by not relying excessively on one particular source (Figure 5.2). Households in the top quintile obtain the largest share of their income, 30 percent, from salaried employment inthe non-agricultural sector, followed by 25 percent from non-salaried employment inthe agricultural sector. This contrasts with households inthe bottomquintile who obtain over 50 percent of their income from self-employment inagricultural activities alone. 5.14 These patterns, however, are not static and the relative importance of different income sources varies over time (Table 5.3). The share of agricultural income declined significantly between 1997 and 2002 as a result of a decrease inproducer prices. This decline was almost entirely compensatedfor by an increase in the share of private transfers, suggesting that households relied on migration and (national andlor international) remittances to cope with (transitory) income shocks generatedby agricultural price changes-a hypothesis that i s confirmed by our analysis of shocks and coping strategies inChapter 6. 102 Table 5.3: ...and over time 1998 2002 %of totalper capitaincome-Allruralhouseholds Agricultural Salaried 9.4 9.8 Self-employed 35.9 31.9 Non-agricultural Salaried 18.7 18.1 Self-employed 17.5 12.2 Capital 1.6 1.6 (Private) Transfers 16.9 26.3 5.15 In sum, the poor depend more heavily than the rich on agricultural activities and, although they are able to substitute one income source for another inthe face of shocks, they appear to be less able to diversify the overall composition of income. Given that poverty rates associated with agricultural employment are higher than those of non-agricultural employment, the question then arises as to whether bothtypes of activities are substitutes for or complements to each other and, more broadly, whether non- agriculturalactivities provide a door out of poverty. We turnto these two issuesnext. Agricultural and Non-Agricultural Income: Substitutesor Complements? 5.16 Agricultural and non-agricultural activities appear to be complements at the department level- the finest level of disaggregation we can use given the data to proxy for common economic area or market-and substitutes at the household level. The positive correlation between the sizes of different economic sectors within departments i s largely due to market size effects since economically powerful areas tend to be prosperousacross sectors (Table 5.4). Table 5.4: Agricultural and non-agricultural activitiesare complementsat the department level. .. Agricultural-Salaried (A - S) Note: Shadedcellsindicatedsignificant correlationsat the 10percentlevel. Source: Authors' calculations usingdata fromENAHO 2002 (INEI). 5.17 In contrast the correlation between the share of agricultural and non-agricultural sources in household income i s negative and significant, irrespective of income and for both salaried and non- salaried activities (Table 5.5). Generally speaking, and inthe absence of significant improvement inlabor productivity in either sector, substitutability across sectors is likely to result from the existence of an upper bound for overall household labor supply. There are, however, important differences between poor and non-poor households. The former substitute between agricultural non-salaried and salaried employment among poor households, suggesting that access to land (or the lack thereof) may be an important factor, while the latter tend to substitute all agricultural activities with salaried non-agricultural ones, suggestingthat higher returns to education inthe non-agricultural sector may play a role. 5.18 Given these results the question then arises at to whether the non-agricultural sector i s the stepping-stonefor poor rural households out of poverty. We explore this issue next. 103 Table 5.5:...and substitutes at the household level. A - S A - S E N A - S NA SE - Correlation coefficient All rural households Agricultural-Salaried (A - S) 1.oo Agricultural-Self-employed (A - SE) Non-agricultural-Salaried (NA - S) Non-agricultural-Self-employed (NA - SE) 1.oo Agricultural-Salaried (A - S) 1.oo Agricultural-Self-employed (A - SE) Non-agricultural-Salaried (NA - S) Non-agricultural-Self-employed (NA - SE) 1.oo Agricultural-Salaried (A - S) 1.oo Rural households in5" quintile Agricultural-Self-employed (A - SE) Non-agricultural-Salaried (NA - S) Non-agricultural-Self-employed (NA - SE) Note: Dataare correlations betweenincome shares 10oercent level. Source: kuthors' calculationsusingdatafromENAHO 2002 (INEI). I s Non-Agricultural Employmenta Door out of Poverty? 5.19 Most of the variation in household income i s due to variation in non-agricultural income, particularly income from salaried employment (Table 5.6). We calculate the percentageof total income inequality that can be explained by the variation in each of its components, and find that more than 50 percent i s due to inequality in non-agricultural salaried income (36 percent) and non-agricultural non- salaried income (15 percent)-a contribution that surpasses these activities' share intotal income (30 and 13 percent respectively). In other words, access to non-agricultural activities explains most of the differences in income levels between poor andrichhouseholds. Table 5.6: Most of the variation inper capita household income is generated by variation innon-agricultural income Meanincome Variation Correlationwith Explained share of total coefficient total income income variation Agricultural Salaried 141.61 2.50 0.27 8.8 Self-employed 164.60 2.91 0.45 20.1 Non-agricultural Salaried 311.09 2.02 0.61 36.0 Self-employed 137.15 2.96 0.40 15.1 Capital 24.59 11.72 0.39 10.6 5.20 Moreover, the share of agricultural income intotal incomedeclines as the latter rises (Figure 5.3). As we pointedout above, agricultural incomerepresents about 60percent of householdincomeamong the poorest households, but only 46 and 30 percent among households in the 3 and 5 quintiles of the rural income distribution respectively. 104 Figure5.3: The shareof agricultural incomeinhouseholdincomedecreases with the latter, butboth continue to be correlated asincomeincreases AgricultureIncomebyTotalIncomeLevel w ~ ~ - 7 3,000 .. ' / -2,500 00 0 0 - 2,000 0 0 _ - 0 4 1,500 I 0% ' /- 0 0 1,000 2,000. 3,000 4,000 5,000 6,000 7,000 8,000 Totalincome I Source: Authors' calculationsusingdata fromENAHO2002 (INEI). 5.21 These stylized facts seemto imply that non-agricultural activities do indeedprovide a way out of poverty. However the economic reality of rural areas i s more complex than the evidence presented so far suggests. First, it i s important to notice that, despite the decline in its share of total income, agricultural income continues to rise with household income suggesting that household characteristics, and not only sector characteristics, play a role inthe determination of the profitability of a particular activity. Second, 30 percent of rich households' income still comes from agricultural activities suggesting that there may exist barriers to total substitutability away from agricultural activities and that, in the presence of such barriers, sectoral diversification may be more important than total substitution as a means to avoid poverty. 5.22 Inreality, with the exception of those that rely exclusively on agricultural self-employment, most households tend to obtain income from more than one source, thus relying on income-generating strategies. It i s the success of these strategies, rather than that of a particular activity, that will ultimately determine whether a householdi s poor or not. 5.23 There i s not, however, a linear relationship between household income and income diversification. Some households are "pushed" to diversify their activities beyond the agricultural sector to cope with external shocks to their own farming, while others many be "pulled" into the non- agricultural sectors becauseit often pays more than farmingand rewards certain assets, such as education, better. 5.24 A household's ability to implement a certain strategy will then depend on the participation of its members on the activities that form this strategy, while the success of the strategy ingenerating sufficient income will depend on the returns that their labor and other assets obtain in those activities. In the next section we study the issues of participationand profitability. 105 THE DETERMINANTS PARTICIPATION IN AND PROFITABILITY OF AGRICULTURAL OF AND NON- AGRICULTURAL ACTIVITIES 5.25 The three most frequently used income-generating strategies, pursued by over 75 percent of all households, combine employment in one or two activities. Namely, 50 percent of all households obtain all income from self-employment in the agricultural sector, 15 percent combines this activity with non- agricultural self-employment, and an additional 11 percent combines it with salaried agricultural employment. Table 5.7: Some income-generating strategies are more likely to provide a way out of poverty than others (A-SE) + (A-S) 48.6 78.3 11.2 (A-SE) 50.0 75.5 49.9 (A-SE) -I-(NA-SE) 41.0 71.3 15.8 (A-S) 28.7 59.9 3.6 (A-SE) + (NA-S) 26.6 56.7 4.9 (NA-SE) 19.6 41.7 3.2 (NA-S) 7.2 16.7 3.0 Other 24.5 54.3 8.5 5.26 The implementation of these and other commonly used strategies yields different income levels, although some patterns arise. Strategiesthat include self-employment inthe agricultural sector tend to be associated with lower income levels and higher poverty, whereas strategies that include non-agricultural activities, particularly salaried employment, tend to be associated with higher income levels and lower poverty. We nextexamine what determines the use of one strategy or another. Income-GeneratingStrategies: ParticipationinAgricultural and Non-Agricultural Activities 5.27 As we mentioned above, the pattern of income diversification between agricultural and non- agricultural activities i s the result of differences inhousehold characteristics, assets and endowments. In addition when markets do not operate in a competitive or efficient manner, personal and institutional constraints can play an important role indetermining participation ina particular sector. 5.28 In order to explore this issue, we estimate the effect that various household characteristics and access to (prodgct) markets (proxied by market size) have on the probability of choosing a particular sector of employment or income-generating strategy46. The results from this exercise are presented on Tables A.10and A.1347inthe Statistical Annex and the mainconclusions are summarized below. 5.29 The maindeterminants of participationina specific employmentsector are: 46. In particular we use an ordered probit model to estimate the impact of household and household head characteristicson the probability of choosing a specific sector of employment or a specific income-generating strategy. The model includes (i) gender of householdhead, (ii) humancapital of householdhead (measuredby years of education and experience, proxied by age and age squared), (iii)household characteristics (composition, crowdedness and recipiency of remittances), (iv) household access to infrastructure (water, sewerageand electricity), and (v) density of local market (proxied by size of population). 47. Household-level summary statistics by main sector of employment and by income-generatingstrategy are presented inTablesA. 11and A.12 inthe StatisticalAnnex. 106 0 Gender of household head: Female-headedhouseholds are more likely (14 percent) to allocate their time to the non-agriculturalsector, especially to non-salaried activities. 0 Education of household head: Households with more educated heads have a higher probability of participating in non-agricultural activities, particularly in the salaried sector. In particular the probability of participation in non-agricultural salaried activities increases by 1.4 percent with each year of education. 0 Access to basic services: Members of households with access to electricity are less likely to be self- employed in the agricultural sector, and more likely (6 percent) to participate in non-agricultural activities. 0 Market size: Households that reside in larger towns have a higher probability (13 percent) on engaging innon-salaried activities, both inthe agricultural and non-agriculturalsectors. 5.30 As a result the mains factors underlyingthe choice of a specific income-generating strategy4' are: Gender of household head: Female-headed households are more likely to choose strategies that include participation in the non-agricultural sector; that is, strategies that yield higher income levels and thus lower poverty. This suggests that there i s no discrimination inrural labor markets, a result supported by Valdivia and Robles (1997) and Escobal (2001b) who fmd no gender bias in income diversification. Education of household head: Households with more educated heads tend to use strategies that are associated with higher levels of income by combining salaried and non-salaried non-agricultural activities, and self-employment in agriculture. The fact that more educated individuals and households tend to diversify rather than fully substitute across economic activities could be suggestive of the existence of barriers to full exit from the agricultural sector, such as thin land rental andsale markets-an impression that i s confirmed by Escobal(2001b) as discussedbelow. Access to basic services: Households with access to electricity have a lower probability of choosing strategiesthat include self-employment inthe agricultural sector-typically low income strategies. Market size: Households living in larger populations tend to use mixed strategies that combine activities inboth the agricultural and the non-agricultural sectors. This seems to suggest that higher population density and larger markets make it easy for households to diversify across sectors and, therefore, to minimize the impact of shocks generatedby sectors-specific factors. 5.31 Escobal(2001b) confirms the importance of locationandendowments as determinants of income- generating strategy choices in his examination of the factors underlying sectoral shares in household income shares. He finds that ownership of fixed agricultural assets, such as land and cattle increase the share of agricultural income, particularly from self-employment, and that access to credit increases the share of incomegeneratedthrough non-salaried activities, bothagricultural and non-agricultural. 5.32 Insum, better householdendowments (such as highereducation) andaccess tobasic servicesand markets allow households to use strategies that include non-agricultural activities, while ownership of agricultural assets and lack of liquidity make it more costly for households to abandon strategies that include agriculturalactivities. 5.33 Itis importantto note, however, that there is significant variation within strategies. For instance, two farmers may be devoting the same amount of labor and effort to their plot with very different results. Similarly, two salaried employees in the food-processing sectors may be paiddifferent wages according to their productivity. Finally, two artisans may be able command different prices for their handicrafts 48. We haveonly examinedthose strategies that are most frequently usedaccordingto the data. These strategies are describedinTable 5.7. 107 depending on the markets they have access to. We then turn our attention to the returns to agricultural and non-agricultural activities and their determinants. Returns to Agricultural and Non-Agricultural Activities: The role of Access to Assets and Markets 5.34 Forty percent of household income and 50 percent of labor income i s generatedinthe agricultural sector. Similarly 30 percent of household income and 40 percent of labor income is associated with salaried activities contracted through labor markets, and these figures go up to 40 and 50 percent respectively ifwe take into account non-agriculturalself-employment. 5.35 Moreover, the relationship between agricultural and salaried income and total income varies with the latter. As we pointed out above, poor households tend to depend more heavily on agricultural activities, particularly self-employment, than better-off ones. Similarly, the share of salaried incomerises rapidly as income increases, and then declines steadily as we move towards the top of the income distribution (Figure 5.4). This inverted-U pattern results from differences in the incomediversifying strategies employed by poor and rich households-the former tend to diversify within the agricultural sector, so that the initial increase in the share of salaried income responds to increases in agricultural salaried income (Figure 5.5); as income increases, however, households engage first in non-agricultural salaried activities and then innon-agricultural non-salaried activities, giving rise to the overall decline in the share of salaried income among richer households. Figure 5.4: The relative importance of salaried activities varies with income... Wage IncomebyTotalIncome Level -_ 173,0MI I 10 0 1.W 2 , m 9c.x 4 , m 5 . m 6.W 7 . m 8,ODo Tow 1-ma I-% wage imms - wage Imm/ Source: Authors' calculationsusingdata from ENAHO2002.N (TNEI). 108 Figure 5.5: ...as householdssubstitute away from agricultural activities I 1 I I I I I I 2 3 4 3 Soirrce: Authors' calculations using data from EXA H 0 2002.W (NEI). 5.36 We find a strong correlation between agricultural productivity and hourly labor income on the one handandtotal income on the other (Tables 5.8 and 5.9). Agricultural productivity amongfarmers in the richest quintiles i s double that of farmers in the poorest quintile, althoughproductivity levels are low for all income groups. Similarly hourly labor income increases by a factor of five as we move up the incomedistribution, particularly inthe non-agricultural sector. Household income quintiles 1 2 3 4 5 NuevosSoles per hectare(Median value) All ruralareas 609.2 823.0 1,020.0 1,113.8 1,600.0 Costa 1,450.0 1,870.4 3,137.5 4,200.0 5,685.0 Sierra 509.3 642.3 646.6 799.6 749.1 Selva 775.0 776.1 866.2 3,320.5 9,252.0 Household income quintiles 1 2 3 4 5 Hourly labor income (1998 Soles-Metsopolitan Lima) All labor income 0.71 1.19 1.63 2.23 3.82 Agricultural Salaried 1.10 1.54 2.15 2.39 3.21 Self-employed 0.67 1.07 1.37 1.87 2.99 Non-agricultural Salaried 0.94 1.98 2.34 3.27 5.84 Self-employed 0.90 1.26 1.78 2.24 3.36 109 5.37 For these reasons, we concentrate our attention on the determinants of agricultural productivity and the returns to salaried employment. We expect both to be a function of household characteristics, as well as of the extent to which households have access to markets where these characteristics are rewarded. We fist analyze the role of household characteristics (assets and services) and then discuss the role of markets. 5.38 The main determinants of agricultural productivity, measuredas the value of output per hectare, are: Land distribution: Agricultural productivity in Peru i s low due to, among other factors, the high level of fragmentation inlanddistribution. The average farmer has access to small quantities of land that are often divided into several plots. According to Peru's last Agricultural Census (1994), 24 percent of all farms owed less than 1hectare of land, 55 percent owed less than 5 hectaresand only 5 percent owed 30 or more hectares. The reasons for the highlevel of fragmentation in land holdings vary by region. Inthe Costa they can be traced to the 1969 Agrarian Reform, which imposed limits on rural landholdings (150 hectares for irrigatedcoastal land) and converted most large landholdings into cooperatives, the majority of which subsequently broke up into small plots. In contrast, in the Sierra farmers maintain fragmented land holdings as a way of managing risks by cultivating lands in different ecological zones. This practice, together with a complicated mechanismof land division for bequests and land grouping through inter-community marriages, explain an important part of the fragmentation process inthis region. The small size of the average plot inturnmakes it difficult andoften expensive for individual farmers to access credit and invest inproductivity-enhancingassets, such as machinery, or infrastructure, such as irrigation. The level of mechanization of small farms i s therefore low, while only 30 percent of all agricultural land i s irrigated. Household characteristics: Productivity increases with the level of education of the household head. This relationship is particularly strong for small plots (less than 1hectare) and intermediate levels of education (primary and some secondary education). The average productivity of a small plot increases from 750 to 947 soles per hectare, a 25 percent increase, as the household head completes primary education, and to 1,886, and additional 100percent increase, as she acquires some secondary education. In contrast, the average productivity of a large plot grows from 4,045 to 5,356 soles per hectare, a 32 percent increase, and to 5,449, an additional 2 percent increase, under the same circumstances (Table 5.10). This positive relationship between agricultural productivity and human capital is not exclusive to the case of Peru-Foster and Rosenweig (1996) and World Bank (2004e) document similar effects for the cases of IndiaandEcuador respectively. Access to assetslendowments:Average productivity is positively correlated with the possessionof a land title and with access to electricity (Table 5.10). Plots tendered by farmers with a land title are 30 percent more productive than the rest. A land title may encourage farmers to make productive investments on their land. It may also lift liquidity constraints that precluded such investments inthe past by allowing farmers to use their land as collateral. Similarly plots in farms with access to electricity are 7 percent more productive than those in farms without it. Access to electricity may have a direct impact inproductivity by, for instance, makingit possible for water to be pumped from a well for irrigation purposes. It may also serve as an indirect measure (or proxy) for access to other productivity-enhancinginfrastructure or machinery. Access to services: Access to services, such as technical assistance and credit, is also positively correlated with agricultural productivity. Plots cultivated by farmers with access to technical assistanceare 15 percent more productive than other plots. Technical assistance can be though of as additional human capital andlor technology invested on the plot. Similarly access to credit may allow farmers to make productive investments in the presence of liquidity constraints. Evidence of the positive relationship between technical assistance and credit and productivity has also been provided for the case of Ecuador (World Bank, 2004e). 110 Table 5.10: Agricultural productivity increaseswith education, possessionof land title, and access to electricity, technical assistance andcredit AI1 rural Lessthan 1 Between 1and More than5 hectare 5 hectares hectares Prod %of Prod % o f Prod % o f Prod %of pop pop pop pop Totai 2,046 1,078 2,346 5,606 Education (head) Incompleteprimary of less 1,773 15.6 750 15.7 2,767 15.1 4,045 14.0 Completeprimary 1,888 60.0 947 62.6 2,136 63.3 5,356 64.7 Incomplete secondary 2,280 18.7 1,886 19.3 2,096 15.9 5,449 14.9 Complete secondaryor more 2,948 5.6 1,014 2.6 3,667 5.6 11,433 6.2 Access to technical assistance Yes 2,333 19.5 No 2,034 80.5 Access to electricity Yes 2,125 56.2 1,288 54.5 2,377 60.2 4,837 74.1 No 1,991 43.8 1.011 45.5 2.299 39.8 7,889 25.8 Access to credit inlast 12 months Yes 3,118 5.9 No 1,969 94.1 Access to landtitle Yes 2,625 30.0 No 2,020 70.0 ,Number of observations 1092 100.0 346 31.6 514 47.0 232 21.2 Source:Authors' calculationsusingdata .omLSMS 2000 (Cuhto). 5.39 Although the figures presented here are mere average differences and could be driven by factors other than the variables of interest, they illustrate the point ina simple manner. Escobal(2001b) confirms the robustnessof these relationships ina regression context, where other factors are controlled for. 5.40 The main determinants of the returns to agricultural and non-agricultural salaried em~loyment are.49 0 Sector of employment: Hourly wages are higher in the non-agricultural sector than in the agricultural one, and within the former they are highest in the high-skilled manufacturing and the business sectors and lowest in the low-skilled manufacturing and transport sectors even after controlling for other observable worker characteristics (Sosa-Escuderoand Luccheti, 2004). Salaried workers in the non-agricultural sector earned between 18 (low-skilled manufacturing) and 36 (business) percent more than their counterparts in the agricultural sector. These differences may reflect, among other factors, variation inlabor productivity across sectors. 0 Householdcharacteristics: Hourly wages increasewith human capital. More educated workers are paidhigher wages (Figure 5.5). Individuals with complete primary and secondary education received wages that are 15 and 30 percent higher respectively than those of workers who have not completed primary (Sosa-Escudero and Luccheti, 2004). In addition the education wage premium is higher in the non-agricultural sector, particularly for workers who have completed secondary education (Figure 5.5). Experience, proxiedby age and age squared, is also positively correlated with wages. 49. Estimates are from an earnings equation using (log) hourly wages as the dependent variable and a series of controls including: (i) worker characteristics (gender, education, age and age squared), (ii)information on sector and type of employment, (iii)regional controls. The model was estimated separatelyfor urbanandrural areas, andfor householdheadsandother members. ill Female workers receive lower wages than their male counterparts, and the gender gap i s larger for household heads than for other workers. This result contrasts with the fact that gender didnot play a significant role inexplaining participationindifferent economic sectors. Figure 5.5: The returnsto salaried employment increasewith education - 6 6 -2 0 2 4 -5 0 5 AgflCUihVB per caolta Incomeper hD~rlog scale) NcrAgilwitilrepercapto lncsmeper hour(lopscale) I Source: Authors' calculationsusingdata fromENAHO2002 (INEI). 5.41 The patterns described so far generally hold across the country, but there exists substantial variation in agricultural productivity levels and in the importance of and returns to salaried employment across regions. Average agricultural productivity i s higher in the Costa than inthe Sierra and the Selva (Figure 5.6). The average plot inthe Costa yields 3,699 soles per hectare, compared to 1,834 and 1,469 in the Sierra and the Selva. The same can be said about the prevalence of and the returns to salaried employment. Forty-five percent of all hours worked are devoted to salaried activities in the Costa, compared with 20-25 in the Sierra and the Selva (Table 5.11). Similarly the returns to salaried employment, measured in terms of hourly wages, are also higher in the Costa than in the other regions, bothinthe agriculturaland non-agriculturalsectors (Figure 5.7). Figure 5.6: Agricultural productivity is higher, ... Distribution of Land Productivity by Region .4- f . .3- . . . x .e Y 8 ~ Ruralcosta .2- / n ,` ,-J RuralSierra RuralSelva .l- 0- 4 6 8 10 12 Logoutputper hectare(soles) Source: Authors' calculations using data from LSMS 2000 (Cubto). 112 Table 5.11: ...salaried employment is more prevalent, ... Costa Sierra Selva All rural % of weekly hours worked Agricultural Salaried 28.5 7.6 13.0 11.9 Self-employed 39.1 66.7 58.6 60.9 Non-agricultural Salaried 13.1 12.1 12.4 12.3 Self-employed 19.2 13.6 16.0 14.9 Source: Authors' calculationsusingdatafrom ENAHO2002.W (INEI). Figure 5.7: ...andhourly wages are higher inthe Costa than inthe Sierra andthe Selva I ISource: Authors' calculationsusingdata from ENAHO 2002.N (INEI). 5.42 This variation i s the result.of regional differences in both endowments and returns to these endowments. To mention a few examples: regarding endowments, the average size of a plot i s larger and the share of irrigated land is higher, the average worker i s more educated, and access to infrastructure is more prevalent in the Costa than in the Sierra and the Selva; regarding the returns to these endowments, the returns to education, both in agricultural and salaried activities, are higher in the Costa than in the other regions (Tables 5.12 and 5.13) Table 5.12: The returns to education inthe agricultural sector.... All rural costa Sierra Selva Prod %of Prod % o f Prod % o f Prod % o f pop POP pop pop Education(head) Incomplete primary of less 1,773 15.6 3028 14.0 1399 15.8 1517 16.0 Complete primary 1,888 60.0 2761 60.2 2067 59.6 1377 61.2 Incomplete secondary 2,280 18.7 5600 18.6 1576 19.0 1585 17.9 Complete secondary or more 2,948 5.6 5295 7.2 3003 5.6 1397 4.8 Note: Agricultural productivity measuredas value of outputperhectare. Source: Authors' calculationsusingdatafromENAHO2002.N (INEI). 113 Table 5.13: ...andin salariedactivitiesare higher inthe Costathan inthe Sierra and the Selva All rural costa Sierra Selva Median %of Median %of Median %of Median hourly hourly hourly hourly % of income Pop income Pop income Pop income POP Education(head) Incomplete primary of less 1.29 49.7 1.87 49.9 1.09 53.1 1.29 46.2 Completeprimary 1.61 23.8 2.30 18.7 1.24 25.3 1.61 24.1 Incomplete secondary 1.78 12.2 2.15 12.8 1.46 9.6 1.77 14.6 Completesecondaryor more 2.43 14.2 2.82 18.5 2.17 12.0 2.43 15.0 5.43 Given their correlation with household income, regional differences in agricultural productivity and the returns to salaried activities in turn translate into regional income differences. In2002 average per capita income in the Sierra, at 146 soles per month, was 60 percent that of the Costa, and this figure was 72 percent for the Selva (Table 5.14). Costa versus Sierra 69.7 30.3 100.0 Costaversus Selva 13.1 89.9 100.0 Sierraversus Selva 127.1 -27.1 100.0 Averageper capita income (2002)Soles per month costa 238.7 Sierra 146.4 Selva 173.1 5.44 The question then arises as to how much of much of the observedregionalincome differences can be attributedto variation inendowments versus variation inthe returns to these endowments. We explore this by performingpairwise comparisons, and find that differences between the Sierra and the other two regions are mainly explained by differences in the returns to endowments, while differences between the Costa and the Selva are mostly due to differences inendowments (Table 5.14). 5.45 While regional differences in endowments are a function of several factors, including household preferences, geography and public interventions, regional differences in returns are in part the result of the thickness and dynamism of regional markets and of productivity. Population density and access to infrastructure are higher inthe Costa than in other areas, despite important improvements inthis area the duringthe 1990sinthe Sierra and the Selva, and bothfactors could potentially contribute to createmore integrated and dynamic markets inruralareas, as well as to better connect rural and urban areas. 5.46 Inorder to overcome regional differences, important investments aimed at improving access to assets, services and markets inthose areas that are laggingbehind must be undertaken (see Box 5.1 for a discussion on the impact of road infrastructure on rural poverty). The nature of these investments will dependon the characteristics of the area and, more importantly, on its potentialfor future productivity and economic growth conditional on such investments. Investments inportable assets, such as education and health, that allow individuals to migrate and benefit from economic opportunities elsewhere are preferable in areas with low growth potential, while investments in fixed, productive assets, such as basic services 114 and infrastructure, can help reduce poverty in areas with high growth potential. Inthe next section we evaluate the impact that a series of such investments in portable and fixed assets can have on household income and poverty. Box 5.1: The benefitsof RuralRoads: Broadeningincomeopportunitiesfor the poor Rural roads are of great importance inPeru, as they play a crucial role in the integration of the country's irregular topography and diverse ecology and climate. Peru's geography makes road construction, rehabilitation, and maintenance a particularly challenging endeavor, and though the importance of good quality rural roads in supporting access to healthcare, education and employment opportunities is not underestimated in Peru, the high constructioncosts imposedby the irregular terrain, combined with sparse and scatteredpopulations that benefit from rural roads, make investmentsin this type of infrastructure less appealingto politicians than other projects yielding higherpolitical returns. The World Bank financed two Rural Road Rehabilitation and Maintenance projects (in 1995 and 2001), which, among other things, provided backing for road rehabilitation, as well as routine and periodic maintenanceof rural roads. Escobal and Ponce (2002) analyzed the project's impact on a number of important well-being indicators, including the level of income and spendingon per capita consumption. The study comparedhouseholds livinginareas that benefited from the roadrehabilitation program with householdsinareas that were not coveredby the program, controlling for initial conditions and households' characteristics. To estimate the effect of rehabilitation, the authors used propensity score matching, with some variations used to make this methodology compatible with the database used. The study focusedon two types of roads: "caminos vecinales" and "caminos de herradura". The "caminos vecinales" are dirt roads connecting towns and villages via public service or freight trucks; they typically connect to secondary roads and provide rural populations access to urban areas. On the other hand, "caminos de herradura" are paths used to transport goods, generally located in areas with irregular terrain; their quality is very low andthey are less frequently used. Results show that the improvement of rural roads had a very positive impact on the lives of the people in areas benefited by the program. Inparticular, access to new sources of income outside agriculture was one of the most positive outcomes of the rehabilitation program. The study also found that the improvement of rural roads generated an increasein household income, though this increase did not necessarily translate into an expansion in consumptionspending. Instead, it was found that the additional income was usedto increasesavings, in most cases by acquiring additional cattle. Escobal and Ponce argue that this result may suggest that improvements in the quality of roads are not perceived as permanent, either due to little or no maintenance of rehabilitated roads or because, even if regular maintenance were to be a permanent component of the rehabilitation program, households do not consider these actions sustainablein the long term, and choose to save what they consider temporary gains from temporarily better infrastructure. The authors arguethat ifreturns from road improvement investments are to be maximized, regular road maintenance should be guaranteed, both to justify these type of investments and to provide households with a clearer picture to enable informed investmentandconsumption decisions. Inaddition, Escobal and Ponce found that householdsneighboring "caminos vecinales" tended to benefit more from rehabilitation than those neighboring "caminos de herradura", which may be related to the very different initial conditions observed in households close to each type of road. The available data indicates that households with access to improved "caminos vecinales" have higher levels of education, larger extensions of arable land, and more access to public infrastructure than households located on improved "caminos de herradura". Due to data limitations, it was not possible to carry out a comparative study of how the program benefited households near each type of road. Further understanding of the complementarities between initial conditions and access to adequate quality roads would be helpful for the designof public programsinrural areas. I Source:EscobalandPonce(2002) THEIMPACTOFPUBLICINTERVENTIONSONPOVERTY 5.47 We consider a series of different interventions aimed at increasing humancapital levels (access to primary and secondary schoolingor access to sewerage) andlor increasing local physicalcapital (access to a public phone) and access to markets (access to a main road). We examine the impact that such 115 interventions can have on household income and poverty based on simulations by Escobal (2002) and Escobal and Torero (2000). 5.48 The authors simulate the implementation of one or more interventions using data from the ENNIV. When two or more interventions are implemented simultaneously, they allow for the existence of complementarities across interventions-that is, they allow for their joint effect to be larger than the sum of their separate effects. Their use of household-level data is motivated by the wealth of information contained in the ENNIV, but has the problemof forcing them to base their results on a single cross-section which could produce endogenous estimates-i.e. in the case of a positive correlation between access to a public phone and household income levels we cannot distinguish between the following two competing explanations: (i) access to a phone actually increasesincome or (ii) people rich are more likely to live near or have access to a phone. In order to be able to derive policy recommendations from the results of this exercise we then rely on complementary analysis by Escobal and Torero (2002) who show that (alternative indicators of) access to basic services and infrastructure, has a positive causal effect on household income.5o 5.49 The main results associated with the implementationof a sin@ intervention can be summarized as follows. First, all interventions being considered (access to a public phone, access to primary and secondary school, access to sewerage, and access to a main road) have a positive impact on household expenditures, although the averagemagnitude of the impact differs across interventions (Table 5.15). Householdincome quintiles 1 2 3 4 5 Percentageincreaseinhouseholdincome Access to public phone 1.72 3.75 5.45 6.10 12.04 Access to primary and secondary school 3.27 3.45 4.47 5.87 6.97 Access to sewerage 3.41 3.53 4.11 4.07 7.57 Access to mainroad (1hour reduction intravel time) 0.95 1.04 1.30 1.17 1.52 Access to mainroad (2 hour reduction intravel time) 1.90 2.09 2.61 2.36 3.06 5.50 Second, the average magnitude of the impact associated with each intervention increases with expenditure. For example, access to a public phone increases expenditure by 12 and 1.7 percent among households in the fifth and the first quintiles respectively. This could be explained if richer households have access to better economic opportunities where the returns to these interventions are higher on average-for instance, by being more connectedto markets. 5.51 Third, the most effective interventionfor the poorest quintiles is not necessarily so for the richest quintiles. Access to primary and secondary school and access to sewerageproduce the largest percentage increasesinexpenditure among households inthe first quintile (3.27 and 3.41 percent respectively), while access to a public phone produces the largest increase for households in the fifth quintile (12 percent). This could be explained if the returns to these interventions vary by economic activity or sector, and the poor are concentrated inactivities or sectors that are different from those inwhich the richconcentrate- something we already documented above. It could also be explained by the fact that the poor tend to _ _ _ ~ 50. The authors use information from the Population Censuses and the ENNIV to construct the following indicators: number of schools and healthcenters inthe area (inper capita terms), the Unsatisfied Basic Needs indeed, the level of urbanization and the distance to the provincial capital. Their results are presented and discussedindetailedinChapter 2. 116 concentrate in areas with low productivity and growth potential and thus are more likely to benefit from investments inportable assets and lesslikely to benefits from investments inproductive, but fixed assets. 5.52 Fourth, as a result of their differential impact across the expenditure distribution, these interventions tend to benefit the non-poor more than the poor (Table 5.16) and, within the poor, the non- extreme poor more than the extreme poor. However, it is important to notice that the potential welfare impact of a small percentage increase in household expenditure i s significantly larger among poor householdsthan non-poor ones. Table 5.16: ...andthis impact is larger for non-poor thanfor poor households Access to public phone 8.26 3.87 Access to primary and secondary school 6.24 3.75 Access to sewerage 6.04 3.43 Access to mainroad (1hour reduction intravel time) 1.37 1.06 Access to mainroad (2 hour reduction intravel time) 2.76 2.14 5.53 We now consider the impact associated with the simultaneous implementation of two or more interventions. The mainresults can be summarized as follows. First, the conclusions postulated about the implementation of a single intervention still hold when two or more interventionsare implemented at the same time-their impact on expenditure is positive, and varies with the nature of the interventions andthe level of householdexpenditure (Table 5.17). Table 5.17: Most public interventionsappear to be complementary to each other 1:Access to public phone 2: Accessto primary and secondary school 3:Access to sewerage 4: Access to mainroad (1hour reduction intravel time) 5: Access to mainroad (2hour reduction intravel time) Household income quintiles 1 2 3 4 5 Percentageincrease inhouseholdincome 1 + 2 5.06 7.34 10.17 12.33 19.85 1+3 33.44 36.05 37.70 39.18 42.49 1 + 4 4.25 6.33 8.07 8.74 14.82 1 + 5 6.84 8.97 10.75 11.43 17.67 2 + 3 6.79 7.10 8.77 10.18 15.06 2 + 4 4.25 4.53 5.83 7.11 8.59 2 + 5 5.24 5.62 7.20 8.37 10.24 3 + 4 0.95 1.04 1.30 1.17 1.52 3 + 5 1.90 2.09 2.61 2.36 3.06 1 + 2 + 3 37.81 40.75 43.86 47.35 52.42 1 + 2 + 4 8.38 10.83 13.99 16.08 23.93 1 + 2 + 5 11.81 14.43 17.93 19.96 28.16 1 + 3 + 4 37.66 40.47 42.47 43.82 47.35 1 + 3 + 5 42.02 45.04 47.40 48.63 52.38 2 + 3 + 4 7.80 8.22 10.17 11.47 16.81 2 + 3 + 5 8.82 9.34 11.60 12.77 18.59 1 + 2 + 3 + 4 42.17 45.32 48.84 52.27 57.62 1 + 2 + 3 + 5 46.67 50.05 53.99 57.35 62.99 Source: Escobdand Torero (2000). 117 5.54 Second, the impact of simultaneous interventions surpasses in most cases the sum of the impact of individual ones, providing evidence of potential complementarities across different interventions (Table 5.18). 5.55 Third, even though non-poor households continue to benefit more than poor householdsfor any intervention (individual or joint), complementarities between interventions tend to reduce the gap in differential impact between rich and poor households, especially when three or more interventions are considered. For instance, the expenditure increase generated by having access to a public phone among non-poor households i s more than double that among poor ones (8.27 versus 3.87 percent), while combinations of this and any other interventionyield smaller differentials (e.g. access to a public phone and a main road increase expenditure by 13.70 and 9.09 percent for non-poor and poor households respectively). Table 5.18: Complementarities between interventionscontribute to close the gap indifferentialimpact between rich andpoor households 1:Access to public phone 2: Access to primary and secondary school 3: Access to sewerage 4: Access to mainroad(1 hour reduction intravel time) 5: Access to mainroad (2 hour reduction intravel time) Non-poor Poor Percentageincreaseinhouseholdincome 1 + 3 39.05 36.11 1+5 13.70 9.09 1 + 2 15.02 7.76 2+3 12.66 7.31 1 + 4 10.95 6.45 2+5 9.18 5.96 2 + 4 7.70 4.85 3+5 2.76 2.14 3 + 4 1.37 1.06 1+3+5 49.47 45.19 1+2+3 48.58 41.21 1 + 3 + 4 44.58 40.58 1+2+5 22.93 14.95 1 + 2 + 4 18.91 11.30 2+3+5 15.77 9.60 2 + 3 + 4 14.21 8.45 1 + 2 + 3 + 5 58.80 50.63 1 + 2 + 3 + 4 53.61 45.84 5.56 Escobal (2002) takes this analysis one step further and explores the channels underlying the impact of public interventions or sets of interventions on household incomelexpenditure. Inparticular he decomposes changes in household income into changes in labor supply across sectors, measured as participation in the sector, changes in the average income obtained from each sector, and analyzes how both factors are affected by the type of interventions we have considered so far. The mainconclusion of the exercise i s that access to infrastructure increases participation and average income in the non- agriculturalsector, especially innon-salaried activities. 118 5.57 Useful as these simulations are, however, we would like to conclude this section with a word of caution. The results we have presented here account only for the (direct) benefit-side, while a full assessment of the profitability or a particular intervention or set of interventions should ideally also account for its costs. PROMOTING INCLUSIVE RURALGROWTH 5.58 Based on the evidence presented in this chapter we can identify four key areas in need of government action if the rural poor are to benefit from the economic opportunities generatedby overall economic gr~wth.~'The areas are: promotion of rural growth through higher integration of rural areas into national markets, increasesin human capital levels, improved access to credit and financial markets, and investment in infrastructure. We propose a series of policy interventions that could be implemented ineachone ofthese areas below. 5.59 Promoting rural growth and increasing the integration of rural areas in national markets will require, among other actions: The strengthening of economic connectionsbetween urban and rural areas: This could be done by developing adequate road communications to facilitate contacts between agents and the transport of merchandise between rural and urban areas. It could also be done by facilitating knowledge and technology transmission from urban to rural areas, and contributing to develop stable economic relationships that ensure a constant demand of agricultural and non- agricultural products for industrial processing andlor exports, and create incentives for producinginbulk. The development of wholesale and local markets. The lack of accessibleoutlets for production in the form of wholesale markets i s one of the major constraints for increasesinoutput inrural areas. In the past the poor quality of rural infrastructure, especially in the Sierra and the Selva has made it difficult to develop wholesale markets becausethe cost of transportation and handlingoften exceeded the profits from the sale. 5.60 Increasing human capital levels inruralareas will require, among other factors: 0 An increase in the coverage and quality of rural education. Improving educational levels and standards inrural areas can be achieved through a series of interventions including: (i) the expansion of bilingual education through the provision of adequate teaching and learning materials and the recruitment and training of quechua-speaking teachers; (ii) the expansion of secondary education, either through formal schooling or through distance learning; and (iii) the creation of incentives for school attendance through conditional cash-transfers or improvements in the feeding and nutrition programs offered inschools (see Chapter 5 for a more detailed discussion on the issueof education). 0 An increase inaccessto technical assistance. Most of rural Peru has little or no access to technical assistance or extension services. The system of public extension services was essentially dismantled inthe 1990s, and a privatemarket for technical serviceshas replaced it. Although subsidized access to these services is provided by INCAGRO, PRA and FONCODES, a large number of small farmers and rural poor are still excluded due to their high cost. Further efforts to support the provision of demanddriven technical assistance, accompanied by marketing and managerial assistance are then necessary. 5.61 Increasing access to credit among ruralproducers will require, among other actions: 51. For an extensive discussion on rural development in the Sierra region the reader should consult the recent RuralDevelopmentStrategy for the Sierrapreparedby the World Bank (2003). 119 The strengthening of rural financial institutions. Rural credit is restricted by the difficulties of many producers, particularly those in small farms, to comply with the administrative and guarantee requirements of financial institutions5*. As a result most existing credit i s informal, or provided by small loans and savings cooperatives. These cooperatives need to be strengthened and so do other institutionswith similar goals, such as women's credit groups. The modification of regulation on collateral. Credit regulation needs to be modified to allow for the use of other family assets, such as land and livestock, as collateral. At the same time efforts to increasetitling (of housing and, particularly, land) should continue. Provisions should also be made to account for the high prevalence of communal property of land among the indigenous population, and the negative impact that this may have of the capacity of the individuals inthese communities to access credit. 5.62 Improving access to and the quality of infrastructure will require, among other factors, and increaseinpublic resources and partnerships with the private sector as discussedin Chapter 1and inBox 2.4. 5.63 The evidence presented in this chapter has also helped us identify important issues regarding the implementation of policy interventions aimed at rural development. First we have argued that the nature of rural poverty i s heterogeneousand varies significantly acrossregions, so that interventions and projects need to take into account local specificities to ensure maximumeffectiveness. For instance inareas with a hostile andor poor natural environment, it will be important to invest inmobile assets and capabilities, such as education and health, that can follow the individual, while i s areas that are rich inlocal resources it will be important to promote interventions that maximizethe returns to these resources. 5.64 Second we showedthat becauselandi s scarce relative to the population it has to support, many of those currently employed in agricultural activities will have to dramatically improve their productivity or abandon agriculture in order to escape from poverty. This implies that a rural development strategy for Peru must be multi-sectoral and consider the interaction between agricultural and non-agricultural activities. It also implies that this strategy must take into account that migration, to urban areas or to other countries, may constitute an optimal coping strategy in areas where potential productivity and economic growth i s low. 5.65 Third we have provided evidence of the existence of complementarities across different policy interventions, implying that a comprehensive spatial approachto rural development can yield highreturns interms of poverty reduction. This, however, must be done attending to the issues of costs and equity. The simultaneous introduction of multiple services may come with too higha bill for poor householdsto afford, so that in some instances sequencing may be preferable. Similarly when resources are scarce we may decide to spread, rather than concentrate them so that a larger fraction of the population can benefit from them. 5.66 Finally it i s important to notice that there already exist several programs in rural areas that support interventions inthe areas we have identified and provide a structure through which the GoP can work towards to the goal of rural inclusive growth. These programs, however, suffer from several problems related to some of the implementation issues identified above that need to be addressed if further interventions are to be effective. Escobal and Valdivia (2004) critically evaluate the most important of these programs, and we present a summary of their mainfindings inBox 5.2. 52. For instance, the FOGAPI (Fondo de Garantia de Prestamo para la Pequeiia Industria) has not been able to extent its services to agricultural micro and small enterprises. 120 Box 5.2: An overview of ruraldevelopment programsinPeru Escobal and Valdivia (2004) examine the objectives and impact of the most important rural developmentprograms in terms of budget (i.e. a total of 18 programs with an annual budget of US$460millionor 0.9 percent of GDP). We summarizetheir main findings inthis box. The authorsclassify the programsaccordingto their main objective into five categories: (i) market development, (ii) social infrastructure, (iii) humancapital investment, (iv) temporary relief (or social protection), and (v) management 3 f natural resources (Table B5.2.1). At the national level most programs and, consequently, most resources are devotedto humandevelopment investment. The distribution of resources, however, varies by region. For instance in the Sierra, which receives 50 percentof the resources budgeted for these programs, relatively more attention is paid to the management of natural resources, mainly through PRONAMACHS. Programs are targeted in different ways according to their goals and objective population. Programs in categories (ii) are and (iii) targeted using the PovertyMap, while programs incategory (iv) are self-targetedat the individual level. Table B5.2.1 Mainpu c programs for developmentfocused onrural areasby category Program Type of program Annualbudget (US$ million) PETT Developmentof marketsfor production(Uni-markets) 7.6 PRA Developmentof marketsfor production(Mu1ti-markets) 3.3 Corredor Puno-Cusco Developmentof marketsfor production(Mu1ti-markets) 2.9 MejorandoTu Vida Social infrastructure 42.0 ProviasRural Social infrastructure 31.9 HTEL Socialinfrastructure 5.7 SaludBhicaParaTodos Supportonhumancapitalinvestment 55.3 Vas0 de Leche Supportonhumancapitalinvestment 90.2 AlimentacibnInfantil Support onhumancapitalinvestment 11.2 Wawa-Wasi Support onhumancapitalinvestment 2.1 PANFAR Support onhumancapitalinvestment 17.9 PACFO Support onhumancapitalinvestment 20.7 DesayunosEscolares Supportonhumancapitalinvestment 48.3 A Trabajar Rural Temporary relief 26.7 Apoyo Alimentario aComedores Temporary relief 32.5 PAR Temporary relief- Transfer of assets andproductioninputs 0.4 MARENASS Sustainablemanagementofnaturalresources ` 2.1 PRONAMACHCS Sustainablemanagement ofnaturalresources 57.1 Total 458.2 Source:EscobalandValdivia (21x The authors also provide a critical assessment of the overall effectiveness and impact of these programs. They concludethat: The design and implementation of most programs are still focused on isolated actions and goals, and do not respond to a spatial vision of development where complementarities across interventions are possible. An integrated approach, however, can be found in some interventions, especially those financed by international donors. The extent to which programs empower beneficiaries during the decision making process varies across programs. In some cases project goals are demandAriven, while in others demand is "induced" using some validation mechanism (Le. town meetings where the offer coming from some central planning mechanism is transformed into local demand). Moreover, even when empowerment mechanisms are in place, the level of involvement of central authorities remains high leaving little space for direct decision-making regarding the allocation of resourcesby the intendedbeneficiaries. Finally projects are not evaluatedsystematically, making it difficult to measuretheir impact or their viability in different areas andlor project sizes. Baselinesare rarely collected, and neither are various project models tested using pilots before up-scaling. Source:EscobalandValdivia (2004). 121 CONCLUSIONS 5.67 We have argued inthis chapter that living standards and poverty inrural areas are the product of the income-generating strategies implemented by poor and non-poor households. Participationin and the returns to these strategies are a function of household characteristics and endowments, access to markets, basic services and infrastructure, and of policy levers, 5.68 We have also shown that non-poor households benefit more from public investments in basic services and road infrastructure because they have better endowments and enjoy higher degrees of market integration. Differences across household, however, tend to disappear when two or more public interventions are implemented simultaneously due to the existence of complementarities across interventions. 5.69 For these reasons, a strategy aiming to promote inclusive rural growth must consider policies directed towards increasing the endowments of the rural poor and improving access to markets, basic services and infrastructure. This strategy should be responsive to regional heterogeneity, be comprehensive in order to encompass both the agricultural and no-agricultural sectors, and be spatially integrated to account for complementarities across interventions. And, above all, it should tailor the nature of its interventions to the productivity and economic growth potential of the area under consideration, investing inportable assets when this potential i s low and in fixed productive assets when thispotential ishigh. 122 6. VULNERABILITY AND EXCLUSION53 6.1 The discussion presented in Chapters 3 and 4 focused on the role of economic opportunities, or the lack thereof, as a determinant of poverty. Low productivity and low income levels, however, are not the only barriers the poor .must surpass. Limited saving capacity and access to financial markets and safety nets make the poor more vulnerable to shocks. Similarly low levels of social mobility and concentration in poorly endowed areas with deficient access to public services contribute to exclude the poor not only from the benefits of current economic growth but also from future economic and social opportunities. 6.2 This chapter discusses the issues of vulnerability and exclusion. The fnst section examines the incidence and nature of shocks, as well as the coping strategies used by poor and non-poor households and their effectiveness. The next two sections analyze the issue of social exclusion from two different perspectives. The second section looks at the problem through the lens of time by focusing on social mobility and studying it from two different but related angles-occupational mobility and educational mobility. The third section considers exclusionfrompublic services and institutions. 6.3 Our mainfindings can be summarized as follows: Both poor and non-poor households are subject to shocks and are likely to lose both income and assets as a consequence of these shocks. In coping with shocks poor households tend to use behavioral strategies, such as increasing labor supply or cutting down consumption, while non-poor households are more likely to rely on asset-based strategies, such as reducing savings, or market- based strategies, such as requesting a loan or cashing an insurance policy. The strategies implemented by the poor seem to be less effective than those of the non-poor in helping households overcome the impact of shocks. Social mobility, measured as the relationship between parental and children's characteristics and proxied by education and occupational mobility, i s low inPeru. Moreover recent increases in social mobility have been the result of across-the-board gains in educational attainment and changes in the productive structure of the economy, rather than the result of higher equality of educational and economic opportunities, and have been concentrated inthe middle of the (income) distribution. Access to public services, to which the bulkof social spending is devoted, i s low among the poor and inrural areas. There are also important andpersistent differences in access between the indigenous and non-indigenous populations. Inaddition the poor are less likely than the non-poor to come in contact with various public institutions, ranging from central and local government offices to public banks to the judiciary system. These differences between poor and non-poor households, indigenous and non-indigenous households and rural and urban areas pose important challenges for the social sectors, particularly inthe context of the ongoing decentralization process. SHOCKSAND COPINGSTRATEGIES 6.4 Both poor and non-poor households are subject to economic, demographic and other types of shocks. However the poor are generally more vulnerable to these shocks than the non-poor in terms of boththeir impact and their duration. Inthis section we use information from the ENAHO2003 to analyze the incidence and impact of different types of shocks, as well as the coping strategies used by poor and non-poor householdsand the extent to which these strategies are successful. 53. This chapter is based on background work preparedthe report team and on existing work by Benavides and Valdivia (2004), Pasquier-Doumer (2002) andBenavides(2002). 123 The Nature andIncidenceof Shocks 6.5 We consider four different types of shocks: economic, demographic, shocks caused by natural disasters or accidents, and other shocks. Economic shocks capture the loss of employment of the household head or some other household member. Demographic shocks refer to the illness or death of one or more household members, as well as to changes in the composition of the household (e.g. abandonmentby headof household). Natural disastersrefer to all weather-related shocks, while accidents include bothunintended events (e.g. job injury)and crime (e.g. robbery). 6.6 There are no significant differences regarding the overall incidence of shocks across poor and non-poor households (Table 6.1). Approximately 20 percent of households declare having suffered at least one shock in the previous year. Natural disasters and accidents are the most prevalent shocks, affecting 13 percent of households, followed by economic and demographic shocks, affecting 4.7 and 1.6 percent of householdsrespectively. Table 6.1: The nature and incidenceof shocksvaries with incomeand across areas National Lima Other urban Rural Non Non Non Non Total Poor Poor Total Poor Poor Total Poor Poor Total Poor Poor Percentageof households ingroup At least one shock 19.4 20.7 18.4 18.2 16.0 19.0 17.5 19.4 16.4 22.2 22.8 21.2 Economic A 4.7 3.9 5.4 8.1 8.1 8.1 6.0 6.9 5.4 1.1 1.0 1.2 Demographic A 1.6 1.3 1.8 1.3 1.3 1.3 2.1 1.9 2.2 1.2 1.0 1.6 Disasters or accidents A 13.1 15.3 11.5 9.6 7.2 10.5 9.2 10.0 8.8 19.7 20.7 18.0 Other 1 0.5 0.5 0.5 0.2 0.2 0.1 0.8 1.0 0.7 0.5 0.4 0.8 Note:A Economic shocks: employment lossor change intypelsector of employment; demographicshocks: changes in householdcomposition(birth, death, marriage, divorce); disasters or accidents: weather-relatedshocks and/or accidents (e.g. job injuy,crime). Source: Authors' calculations usingENAHO 2003 (INEI). 6.7 Some differences exist, however, across areas and between poor and non-poor households once the nature of shocks i s taken into account. Economic shocks are more prevalent in urban areas, while natural disasters are more frequent in rural areas. This i s due to the fact that urban households rely relatively more on labor markets and rural householdsrely relatively more on agricultural activities for a living. 6.8 Inaddition, withinurbanandruralareas, non-poor householdsaremorelikelyto suffer economic shocks while poor householdsare more likely to experience naturaldisastersandaccidents. This is due to two factors. Firstnon-poor householdstend to concentrate inthe formal sector of the economy and, thus, are susceptible to unemployment spells and the loss of employment benefits that it implies, while poor households generally rely on informal income sources (Table 6.1). Second the poor tend to reside in areas that are more prone to these types of shocks, or, more likely, in areas that lack the appropriate infrastructure to deal with shocks. For instance, slums inurban areas are highly susceptibleto flooding in the event of heavy raindue to the absence of proper sewerageand draining systems. This i s illustratedby the higher incidence of this type of shocks among poor households in Lima's marginal areas or C O ~ O S (Table 6.2). Similarly, the rural poor are vulnerable to losses caused by erosion and landslides because their property i s usually situated in more disadvantageouslocations and divided into small separate plots making it hard to undertake preventive investments such as terracing and the construction of retaining walls. 124 Lima Center - Lima Conos - Female-headed. Poor N o nPoor Poor NonPoor Poor NonPoor Percentageof households ingroup A t least one shock 13.3 18.6 19.8 18.3 25.3 20.4 Economic 8.2 9.4 5.7 8.0 5.0 5.7 Demographic 0.0 1.2 1.7 2.4 5.7 5.4 Disasters or accidents 5.5 8.9 13.5 8.6 15.0 9.6 Other 0.0 0.2 0.0 0.4 0.5 0.3 6.9 More formally the role that sector of employment and (access to) infrastructure play inexplaining the variation in household susceptibility to shocks is confirmed when we estimate the likelihood of suffering a particular shock as a function of household characteristics using regression analysis. Household informality rates are negatively correlated with the probability of suffering an economic shock, while accessto water, electricity and sanitation reduces the probability of suffering a shock caused by a naturaldisaster. 6.10 Finally, female-headed households, both poor and non-poor, appear to be more vulnerable to shocks than male-headed ones, irrespective of their income status. The question arises, however, as to whether it may have been a shock that caused the household to be female-headed, rather than the other way around. This i s particularly plausible in the case of demographic shocks, which could capture the absence or even the death of the previous (male) household head. Since we do not have information as to the situation of the householdat the time of the shock, we cannot rule out this possibility. The impactof shocks 6.11 Shocks can have a negative impact on income, on wealth and assets, or on both, depending on their nature and severity. Both in urban and rural areas income losses are most frequent after economic shocks, while wealth and asset lossesare most common after a natural disaster or an accident (Table 6.3). 6.12 Similar percentages of poor and non-poor households declare to have suffered losses in income and wealth after a shock. This, however, should not be interpreted as an indication that the impact of these shocks i s the same for both types of households since information on the actual amounts lost i s not available and, even if losses were similar in magnitude, their impact on poor households would be relatively more severe given that their income i s closer to subsistence levels and that they have fewer assets. This i s confirmed by the evidence presented inChapter 3 on the determinants of flows inand out of poverty and, inparticular, on the role played by economic and demographic shocks in explaining entry into (transitory) poverty. Copingwith Shocks: Strategies andEffectiveness 6.13 Most households that suffer a shock resulting in the loss of income or wealth try actively to cope with it by usingexisting savings and assets, askingfor a loan or cashing inan insurance policy, increasing labor supply or reducing consumption. The choice of a particular strategy depends on several factors, ranging from the household's saving capacity prior to the shock to its ability to access financial markets or to cut down consumption. These factors inturn are related to both income levels and area of residence so that we observe significant differences between the strategies usedby poor andnon-poor households in urban and rural areas. 125 Economic Demographic DisasterslAccidents Other Poor NonPoor Poor NonPoor Poor NonPoor Poor NonPoor Percentageof households ingroup (conditional on having suffereda shock) National Loss of income 89.0 86.1 71.5 65.5 50.2 52.1 64.5 66.1 Loss of wealth I assets 2.5 1.7 4.3 5.2 26.0 25.5 4.9 16.4 Both 7.5 10.5 10.4 15.0 19.1 13.9 18.2 8.8 None 0.9 1.8 13.9 14.3 4.7 8.5 12.4 8.7 UrbanAreas Lossof income 88.1 86.3 74.9 65.6 61.0 52.8 78.4 68.1 Lossof wealth I assets 2.8 1.7 2.6 4.1 23.1 24.5 0.8 22.4 Both 8.4 10.2 11.2 13.9 9.0 11.8 8.7 3.5 None 0.8 1.9 11.3 16.4 6.9 10.9 12.1 6.0 RuralAreas Lossof income 94.6 82.2 66.3 65.2 46.2 50.9 41.1 62.5 Loss of wealth I assets 1.2 1.8 6.8 9.4 27.0 27.2 11.8 5.1 Both 2.4 16.1 9.1 19.3 22.9 17.7 34.0 18.8 None 1.8 0.0 17.8 6.1 3.9 4.3 13.1 13.7 6.14 Poor households are more likely to spends all their income and hence save less than non-poor ones. In contrast, non-poor households make more frequent use of their own savings and of loans to finance and smooth their consumption over time (Table 6.4). Table 6.4: Poor householdssave lessand have more limited accessto financial marketsthan non-poor ones interms of income. Source: Herrera(2002). 6.15 Consequently, poor households tend to increase labor supply and reduce consumption after suffering a shock, while non-poor households are more likely to use existing assets or to resort to financial markets, either asking for a loan or cashing in an insurance policy. Inaddition the poor have a slightly higher probability of receiving assistance, especially in the event of a natural disaster, than the non-poor, although the relative importance of this strategy is very small compared to those based on behavioral responses at the household level (Table 6.5). 6.16 The differences between the strategies implemented by poor and non-poor households are more marked in urban than in rural areas due to the fact that financial markets, to which the non-poor resort more frequently than the poor, are more developed in the former. More developed financial markets could also explain why the use of loans and insurance i s more prevalent inurban areas, both among poor and non-poor households, while changes in labor supply and consumption are more common in rural areas (Table 6.5). 126 Economic Demographic DisasterslAccidents Other Poor NonPoor Poor NonPoor Poor NonPoor Poor NonPoor Percentageofhouseholdsingroup(conditionalonhavingsuffered ashock) National Reducedsavings I sold assets 20.3 27.6 18.9 24.8 14.7 18.1 29.1 14.9 ReceivedloanI insurance 16.7 20.3 20.7 31.2 10.6 17.8 23.7 24.0 Increasedhouseholdlabor supply 44.2 37.0 50.3 39.1 22.9 25.2 34.6 26.1 Receivedassistance 0.6 0.8 0.0 0.0 3.2 1.9 1.8 0.0 Reducedconsumption 20.4 17.1 5.6 9.9 11.9 8.2 14.6 16.2 Other 5.2 7.8 5.O 6.6 5.1 8.0 16.4 12.9 Nothing 9.0 10.4 11.3 14.8 42.0 31.9 10.3 22.6 Urban ReducedsavingsI sold assets 18.5 28.7 19.8 26.5 17.2 17.1 35.8 12.9 ReceivedloanI insurance 27.1 16.6 16.9 23.2 23.5 24.1 16.8 17.3 Increasedhouseholdlabor supply 43.1 37.O 55.3 39.7 24.2 22.8 31.8 21.1 Receivedassistance 0.0 0.3 0.0 0.4 3.0 1.7 3.4 0.0 Reducedconsumption 14.8 19.6 14.1 11.3 8.4 8.4 12.6 18.7 Other 6.0 6.2 4.2 7.9 7.7 9.0 20.6 10.3 Nothing 10.9 9.3 6.7 14.2 28.8 27.7 19.0 25.0 Rural Reducedsavings I sold assets 19.6 17.3 27.7 18.6 13.8 19.5 41.9 9.6 ReceivedloanI insurance 6.0 22.2 26.3 18.0 6.0 9.2 10.6 19.0 Increasedhouseholdlabor supply 46.6 42.7 34.0 38.1 22.4 27.3 29.2 24.3 Receivedassistance 4.1 0.0 0.0 0.0 3.3 1.6 4.8 0.0 Reducedconsumption 21.3 13.1 7.6 8.6 13.2 9.2 10.8 19.3 Other 6.2 4.5 3.3 10.1 4.1 7.7 24.5 6.5 Nothing 16.0 15.6 13.4 12.7 46.6 36.9 26.9 24.8 6.17 Given these differences, the question then arises as to how effective the strategies implemented by poor and non-poor households are in coping with the impact of shocks. The data suggest that, independently of the nature of the shock, non-poor households seem to be more effective at overcoming its consequences than poor ones. A higher percentage of non-poor households declares to have already overcome the shock or expects to overcome it in the next 6 months. In contrast, relatively more poor households report that it will take more than 12 months to go back to pre-shock welfare levels or, more dramatically, that they will never be able to recover from the shock (Table 6.6). 6.18 Differences in effectiveness are particularly marked in the aftermath of a natural disaster, suggestingthat the loss of wealth and assets has a more irreversiblecharacter for the poor. An idea that is consistent with the exiguous savings capacity of the poor. 6.19 Differences in the effectiveness of the coping strategies selected by poor and non-poor households are the result of two factors. First the nature of the strategies implemented by both groups is different. Poor households tend to rely more on strategies that require immediate behavioralchanges (in labor and consumption), whereas non-poor households minimize such changes by usingfinancial markets and assets. Secondthere are limits to the effectiveness of the behavioral strategies used by the poor since individuals can only work so many hours and it is difficult to bring consumption under the subsistence level. Households that are closer to these limits at the time of the shock will find it harder to overcome its 127 effects. Becausethese households tend to be the most needy, this creates a vicious circle of poverty and vulnerability. Lessthan 6 Between 6 and More than 12 Does not Itwas already months 12months months Never know solved Percentageof householdswho sufferedeach shock (conditional onhaving sufferedashock) Poor Economic 17.7 10.3 15.8 9.0 35.0 12.2 Demographic 14.4 13.1 19.0 12.3 30.1 11.1 Disaster 4.2 7.3 16.3 33.7 24.4 14.1 Other NA 24.0 18.2 20.6 26.0 11.2 Non-poor Economic 19.1 16.7 14.4 10.8 23.4 15.7 Demographic 10.1 9.4 23.5 8.4 24.4 23.9 Disaster 7.3 8.6 15.8 18.3 23.5 26.3 Other 11.2 13.3 10.8 10.5 37.0 16.9 6.20 Finally it i s important to notice that, although our analysis of shocks and coping strategies has been very much focused on the impact of shocks on monetary variables (income and assets) and on household-based coping strategies. Shocks can and do have an effect on other outcomes, such as education and health outcomes. The next two sections explore, in turn, the impact of shocks on non- onetary outcomes, paying particular attention to differences between indigenous and non-indigenous people, and the role of social capital inproviding insurance against shocks. Non-monetary Impact of Economic Shocks 6.21 This section summarizes the results of three recent studies on the impact of economic crisis on human capital outcomes (Scahdy, 2004, and Paxson and Schady, 2004) and the differences inthis impact between indigenous andnon-indigenous groups (Benavides and Valdivia, 2004). 6.22 Impact economic crisis on education. Schady (2004) uses data from the PeruLiving Standards Measurement Survey (LSMS) to analyze the effect of the 1988-1992 economic crisis on children's education, measured as school attendance. The data covers urban areas (covered of rural areas was limited in 1991 due to terrorism) and expands through the period before the crisis (1985/86), the crisis (1991), and its aftermath (1997). The LSMS includes detailed information on household characteristics, educational attainment, years of schoolingcompleted, and current employment. 6.23 Schady finds that attendance remained stable across income groups and that the fraction of children who combined school with work declined significantly during the crisis, particularly among older children. The percentageof children aged 12 to 17 employed and attending school in 1991 was 12 percent, comparedto 31percent in 198516and 20 percent in 1997. Inaddition the author finds that lower rates inchildemployment translate intohigher education levels over time. Children exposed to the crisis have an average of between 0.1 and 0.2 more years of schooling, and that average number of years passed for a given age increasedwith the number of years of crisis exposure. 6.24 What explains these patterns? Schady argues that the observed decline inchild and youth labor i s the result of a decrease inthe opportunity cost of schooling for these groups associated with the fall in real wages generatedby the crisis. The evidence presented inthe paper suggests that children substitute 128 between employment and schooling in response to changes in wages: those who did not combine work with school during the crisis years may have been more likely to focus their efforts in school and consequently make adequategrade progress. 6.25 Schady concludes that, though macroeconomic crises have serious consequences for household welfare, Peru's case demonstrates that they need not always negatively affect education. These findings point the need for further research to understand why macroeconomic crisis lead to a deterioration in education outcomes insome countries, but not inothers, as well as their impact on schooling quality. 6.26 Impact of economic crisis on child health. Christina Paxson and Norbert Schady (2004) examine the effect of the Peruvian crisis on child health and infant mortality using data from the 1986, 1991192, 1996 and 2000 Demographic and Health Surveys (DHS). The DHS are nationally representative, sample women aged 15-49, and include questions on the date of birth, current vital status, and the date of death (if deceased) of all children ever born to the respondent. They also contain information on circumstances surrounding the births of children aged 59 months or less and, for children who are living, height and weight, as well as information on a range of household socio-demographic characteristics, including urban status, maternal education, housing characteristics and ownership of durable goods. Finally the authors supplement the DHS with administrative data on health expenditures and consumption data fromthe 1985186and 1991. 6.27 Paxson and Schady show that infant mortality increased sharply around 1990 across the country. This increasebegins with childrenborn inthe secondhalf of 1989 andpeaks for children born inthe first half of 1990, rising approximately 2.5 percentage points (Le., from 50 per 1000births to 75 per 1000 births), or approximately 17,184 infant deaths. The author also estimate the elasticity of infantmortality with respect to per capita GDP to be -0.973 (t=2.92), suggesting that mortality and per capita GDP were conversely related at the time of the crisis. 6.28 Though the available data does not allow for a detailed identification of the causes behind the observed increase in mortality, the authors document a collapse in public expenditures on health during the crisis period, which possibly led to the important declines in health care utilization observed during the years in which the crisis was most profound. Simultaneously households appearedto have protected expenditures on food, possibly at the cost of expenditure in other items important for infanthealth status, such as health care and medication. Paxson and Schady find no evidence that the increase in infant mortality was due to changes in the composition of the women giving birth (Le. their age, level of education and urbanization), outbreaks of infectiousdisease, or terrorism. 6.29 The authors conclude future research on the reliability of different sources of mortality data and on the importance of changes in household income and consumption relative to changes in public expenditures on health and other services would be important for the design of policies to protect child healthduringmacroeconomic crises. 6.30 Differential impact of the crisis on human capital outcomes across indigenous and non- indigenousgroups. Benavides and Valdivia (2004) reproduce Schady and Paxson's analysis separately for indigenous and non-indigenous groups. They find that the impact of the crisis on infant health outcomes was significantly stronger among quechua and aymara-speaking households, suggesting that these groups are more vulnerable to shocks than their counterparts. Social Capitalas a CopingStrategy 6.31 Inour discussion about shocks and coping strategieswe have focused on the role of income and wealth as potential instruments that households can use to overcome the impact of shocks. We turn our 129 attention now to the role that social relationships, or social capital, can play as both a coping strategy and a stepping stone out of poverty. 6.32 The literature on this issue (World Bank, 2001; Granovetter, 2000) has identified two different types of social relationships: close and remote (or strong and weak). The first type refers to relationships built around neighborhood associations, or locally based social programs (e.g. Comedores Populares in Peru). Households that engage in these relationships tend to have similar characteristics and to reside next to each other. As a result their income levels are correlated and they are likely to suffer common shocks, which can undermine their capacity to help each other in the presence of a negative shock. Therefore these relationships are thought of as providing day-to-day support, rather than as providing insurance against non-idiosyncratic shocks. In contrast the second type of relationships refers to associations with a more mixed membership, such as sports clubs or political parties. Higher heterogeneity minimizes the likelihood of correlated shocks, and i s expectedto generatebroader and new opportunities for diversification. 6.33 At least 50percent of all Peruvianhouseholdsdeclare to have membersthat belongto at least one association. Inaddition, what we have called "strong" social capitalrelationships are more prevalent than "weak" ones. Forty percent of all households reports to be associated with a "strong" organization, compared to 26 percent that reports to be associatedwith a "weak" one (Table 6.7). National Urban Rural Total Poor NonPoor Total Poor NonPoor Total Poor NonPoor Percentageofhouseholdsingroup Participates in: At least one association 50.3 60.3 42.6 44.6 49.6 41.8 70.7 74.0 64.6 Cultureandsports 7.6 5.1 9.6 9.2 6.6 10.6 6.5 5.2 8.9 Neighborhood 14.9 18.1 12.3 10.7 10.6 10.8 26.1 25.8 26.7 ProfessionalI Unions 6.0 2.3 8.9 9.9 4.7 13.0 2.1 1.1 4.0 SocialPrograms 27.0 40.7 16.3 18.7 30.5 11.9 45.2 50.9 34.7 Other 10.9 16.3 6.8 6.6 8.6 5.5 22.3 25.5 16.5 Natureof association: Strong 41.9 58.9 28.6 29.4 41.1 22.7 71.3 76.7 61.4 Weak 24.6 23.7 25.3 25.7 19.9 29.1 30.9 31.8 29.4 6.34 There are, however, important differences between urban and rural areas and between poor and non-poor households. The level of association i s generally higher in rural areas, particularly through "strong" organizations. Similarly poor households exhibit higher overall participation rates in social organizations than richones, particularly in "strong" organizations and inurban areas (Table 6.7). 6.35 Insum, there appears to be more social capital, proxied by association with various groups, among poor households than among non-poor ones. These groups may be able to provide support inthe event of a household-specific shock, such as an illness episode, a death or the loss of employment. They cannot, however, provide support or insurance in the face of a common shock, such as a natural disaster or an economic slowdown. Policy Implications 6.36 We have argued in this section that the poor are more vulnerable to shocks and less effective in overcoming them than the non-poor because of their limited capacity to save and to access financial markets and safety nets. As a result interventions aimed at lifting these constraints can go a long way in breaking the vicious circle of high poverty and high vulnerability. These interventions must take into 130 account the nature of the shocks the poor are exposed to, particularly whether they are idiosyncratic or common to a specific area or group, and the implications that this has for risk-pooling opportunities. They must also be designed in ways that support the poor without making them dependent. We propose below a series of such interventions that can be applied inboth urban andruralareas. Increasingthe savings capacity of thepoor 6.37 Saving is difficult for poor households given their daily needs and their scarce resources. In addition their patterns of accumulation are often not optimal given their limited access to financial markets and the often inappropriate nature of the financial instruments available to them. Hence enabling the poor help themselves will require interventions aimed at (i) broadening their assets base, and (ii) increasing access to financial services andinstruments among the poor. 6.38 Helping the poor broaden their asset base will require, among other factors: 0 An increase in household disposable income. Accumulation of any type requires that income exceeds basic household needs. A first step towards increased savings is increased income, either in the form of higher labor incomeor inthe form of transfers. We already presented inChapters 3 and4 a series of interventions directed at increasing the productivity and earnings of the urban and rural poor respectively. On the issue of transfers, it is possible to use existing programs or to develop new ones that may be more suited to the specific goal we have inmind. Existing public transfers are insufficient to provide significant poverty alleviation due to their low level and regressive distrib~tion~~. An alternative option i s to follow the steps of the numerous countries in and outside the region that have had very positive experiences with conditional-cash transfer programs (e.g. Mexico with Oportunidades; Brazil with Bolsa Escola; Argentina with Jefes y Jefas; Ecuador with the Bono de Desarrollo Humano). These programs serve the double objective of providing short-term poverty alleviation and promoting medium-term human capital investments. The GOP i s currently considering the creation of a conditional cash transfer program, Pro-Peru. We discuss the program's design and objectives in Box 6.1. 0 An increase in the marketability of housing and land: Housing and land are the most valuable assets held by the poor, butthey are often of little use inthe event of a shock due to their low levelsof marketability. Increasing access to adequate housing in urban areas and promoting housing and land titling inboth urbanandruralareas would allow poor householdsto usethem as collateralfor credit if necessary. Titling would also go a long way in activating what are currently very thin housing and land markets, especially in rural areas, and thus increasing the value of these assets when liquidity is needed. 54. Herrera (INEI, 2002b) reports that the poverty in2001 would have increasedfrom 54.8 to 55.4 percent in the absence of public cash transfers. Similarly Tesliuc (2005) argues that poverty in 2003 would have increased from 54.7 to 55.2 percent in the absence of public in-kind transfers in the form of nutrition and feeding programs. 131 Box 6.1: Is Peru ready for a Conditional Cash-Transfer Program?A few observationsonJuntos The Government of Peru has recently proposed the creation of a new conditional cash transfer program (CCT) known as Pro-Peru. CCTs have been widely and successfully used in other countries in and outside the region to promote short-term poverty alleviation and long-term human capital accumulation. The case of OportunidadesinMexico and BolsaEscolainBrazil provide two good examples of such programs. The program will pay beneficiary households approximately 100 Soles (or approx. US$30) per month conditional on children attending school, being taken to the health center for periodic check-ups, andor complying with adequate intakes and use of nutritional complements. The program will serve families with children under the age of 14 residing in poor rural communities in 110 selected districts (to be extended to 320 during 2006), where poor communities are identified using information on poverty, unsatisfied basic needs, chronic malnutrition and social violence. Basedon this generalframework we presentinthis box a brief discussionof some important issues the GOP should consider if and when Pro-Peru is implemented. In particular we pay attention to the issues of coverage and targeting, programobjectives andprogramfinancing. Coverage and targeting. Basedon the selectioncriteria described above the program is expectedto cover 2.5 million people nationally. In 2003 6 million Peruvians were extremely poor and an additional 9 million were non-extremelypoor out of a population of approximately27 million. This impliedthat the program, if well targeted, would cover 30 percentof the extremepoor and 15 percentof all poor. Objectives. The mainobjectives of Juntos appear to be short-run poverty alleviation and long-term human capital accumulationthrough increasedschoolingand access to basic health care. The first objective seems relevant given that access to social protection inrather limitedand unevenly distributed, as we have discussed inthis report. The second objective, however, presents some problems. With respect to primary education, quality rather than coverage is the problem, and the program would do little to solve this issue. Moreover there already exist other social programs, such as the DesayunoEscolar, which promote school attendance. Finally, although coverageis an issue in initial andsecondary education, increasingit would require significant supply interventions in the form of schoolconstructionand provision of teachers, which havenot beenbudgetedfor. The same can be said with respectto health, where distance to the nearest health center and inadequacy of services provided, rather than lack of demand, seem to be the main constraints to access (see discussion later on in this chapter). In addition, vaccination rates do not constitute a problem and, as was the case with education, there exist other programsdesignto increasehouseholddemandfor basic healthcare, suchas PACFO. All of this is not to say the program is unnecessary, but rather that its objectivesneedto be clearly defined according to existing sectoral needs and in coordination with existing programs, and that, when these objectives require complementary supply interventions, additional resources mustbe budgeted. Financing. Social spendinginPeru, at 5 percentof GDP, is low comparedto the region's average, and we have already mentioned that social protection for the poor is almost non-existent. In that sense a program like Juntos could fill up an important gap. Unfortunately available fiscal space to finance this initiative i s small. Providing coverage for 100,000 households during the first phase of the program would cost 10 million Nuevos Soles per month, 120millionNuevos Soles or US$40 millionper year. Similarly providing coveragefor 2.5 million people, or 532,000 households, would cost 53 million Nuevos Soles per month, 636 million Nuevos Soles or US$212 million per year. These figures are equivalent to 8 and 44 percent of the 2004 social protection budget, respectively, and to 1 and 5 percent of the overall social sector budget. To this we need to add set-up and administrative costs, as well as the cost of developingand maintaining appropriatetargetingtools. The 2006 budget for Juntos is 300 million Nuevos Soles, of which 30 percent will be transfer to the education and health sectors to ensure adequate service supply for program beneficiaries. Given that the program's objectivesas currently stated overlap with those of other existing programs, it would be important to recognize that implementing Juntos may imply that other programs need to be scaled-down or retargeted to avoid overlap and inefficiencies. Other risks. Finally it is necessary to point out that the development and implementation of a CCT program requires careful planning and extensive discussion regarding its coverage and objectives, as well as the rules by which beneficiaries will be chosen. If the GOP is committed to the implementationof Juntos, it should not allow the proximity of national elections and the mounting pressure for more populist measuresthese will generate to jeopardize this process. To avoid this risk, the GOP should put forward for discussiona proposal that describes Juntos characteristics and implementation strategy clearly and ensures that beneficiaries are selected and the programis operated in atransparentmanner andaccordingto widely accepted criteria. 132 6.39 Increasing access to financial services and instruments among the poor will require, among other factors: An increaseinaccessto the bankingsystem. Bridgingthe gap that exists betweenthe poor and the banking systemcould be done by expanding ATM services to poor areas, and by providing financial literacy programs for poor households. Increased contact between poor households and the banking systems could also be achieved by channeling social programpayments through banks, as i s done for example inEcuador inthe case of the Bono de DesarrolloHumano. 0 The creation of financial instruments that cater the poor. This could be done by offering savings accounts that pay lower returns but do not require a minimumbalance, or by developingcommunity- basedinstruments such as rotating saving and credit schemes. 6.40 Indesigningandimplementingthe interventions discussedsofar we musttake intoaccount that their benefits may vary across groups, and even among poor households. For instance, given their limited savings capacity the extreme poor are more likely to benefit from interventions aimed at increasing their asset base than from interventions aimed at increasing access to the banking and financial systems. In contrast the non-extreme poor are more likely to benefit from access to financial tools that allow them to save profitably and at the same time maintaina certaindegree of liquidity. Zncreasingaccess to insurance and credit among thepoor 6.41 The poor suffer from higher levels of exposure to risk than the non-poor becausethey have very limited or no access to insurance or credit. As a result they tend to save in liquid assets that can be quickly retrieved in the event of a shock but that yield very low returns. Increasing access to insurance and credit among the poor, as well as providing them with profitable savings opportunities can be an effective way of helpingthe poor cope with shocks. 6.42 We focus here on the issue of insurance, since we already discussed policy options aimed at increasing accessto credit among the urban and ruralpoor inChapter 4 and 5. Indoing so we distinguish between interventions whose objective it i s guarantee a minimumlevel of income for poor households (income insurance), and interventions whose objective is to cope with asset losses associatedwith natural disasters (catastrophic insurance). 6.43 Improvingaccess to insurance will require, among other factors: 0 An increase in access to income insurance. Income insurance can be provided in the form of workfare programs, of which Peru's A Trubajur i s an example, or as non-contributory pensions inthe case of older or disabled individuals-an option whose fiscal sustainability would have to be carefully examined prior to its implementation. Involving insurance companies in the provision of this service to the poor may prove cumbersome, since it requires a careful assessment of the specific risks that the poor face and of their willingness to pay, but not impossible. 0 Catastrophic insurance: Poor households can access catastrophic insurance through the government or through private providers. Inthe first case governments can seek insurance in international markets and, inthe event of a shock, channel the funds provided by the insurance policy to affected areas. Although provision of disaster insurance by the private sector i s fairly common in developed countries and among well-off households, irregular settlements, lack of housing and land titles and sub-optimal 133 housing makes the poor hard to insured. There exist, however, successful experiences in this regard in urban areas, such as that of Manizales inColombia?' that can offer interesting lessons. 6.44 While its cost can be minimized by providing catastrophic insurance, the riskof a disaster can be mitigated by implementing preventive measures. Some hazards, such as floods and landslides can be reduced through engineering solutions but others cannot. Prevention, however, can be useful even in these cases. Avoiding particularly vulnerable or risky settlement areas i s important. In urban areas adequate housing and infrastructure, together with slum upgrading programs can help mitigate the impact of natural disasters. Infrastructure also plays and important role in rural areas, as do measures aimed at minimizing erosion, such as agricultural terracing and reforesting. Building more effective safety netsfor thepoor 6.45 Social protection programs can help prevent and mitigate the impact of shocks and reduce risk aversion among the poor. For these programs to be effective, however, they must be well targeted and have the ability to deliver relief shortly after shocks occur. They must also take into account existing informal social protectionmechanisms. 6.46 Increasing accessto effective safety nets among the urban poor will require, among other factors: The implementation of programs that provide protection against income shocks. We have already discussed two kinds of programs that could serve this purpose above: individual unemployment accountsand workfare programs, such asA Trabajar-Urbano. The main challenge it to adapt these programs to the specificity of urban areas in terms of the benefits they offer, their targeting mechanisms and their entry and graduation rules so that they truly act as countercyclical mechanisms. This last point i s particularly important given our discussion in Chapter 3 on flows in and out of poverty and the large volume of householdsthat enter and exit poverty eachyear. 0 The implementationof programs that target vulnerable groups. The elderly and the youth are particularly at risk in urban areas. Coverage of the formal pension system i s extremely low among the elderly, while existing programs targeted at youth groups lack a focus on prevention. The implementation of a non-contributory minimum pension system for the needy elderly could help prevent the risk of poverty in old age, subject to the fiscal constraint mentioned above. The work of Gill, Packard and Yermo (2004) discusses how this can be done in ways that minimize both fiscal costs and potential disincentives to save or work in old age. Similarly programs that understand the determinants of youth risk (individual characteristics, family background, peer and neighborhood effects) and emphasize prevention (e.g. minimizing future income risk by providing incentives for secondary education completion) can helpreduce vulnerability andriskamong youth. 0 The implementation of programs that facilitate access to the labor force and employment diversification. Interventions that help the urbanpoor take greater advantage of the jobs available in cities are very important in urban areas since, as we discussed in Chapter 5, the rural poor obtained most of their income inthe form of wages. In addition to interventions aimed at increasing average education levels, training programs and programs directly targeted at labor market integration, such as 55. The local authorities have signed an agreement with an insurance company that allows any resident to purchase insurancecoverage through the municipal tax collection system. The insurance contract is priced competitively and is designed so that the insurance company has a direct contractual relationship with the taxpayer who wishes to participate in the program. The municipal authority only acts as a premium collector and is not responsible for any claims under the plan, which remains at all times the responsibility of the company(World Bank, 2004~). 134 job search and placementservices, and day care servicesfor poor mothers can increaselabor market participation andattachmentamongpoor households, thus contributing to incomediversification. 6.47 Increasingaccessto effective safety nets amongthe ruralpoor will require, amongother factors: e The implementationof programs that provide protection against income shocks. As inthe case of urban areas, workfare programs, such as A Trabajar-Rural, and conditional cash-transfers programsare two usefultools inthis regard. e The implementation of programs that facilitate agricultural diversification and access to the non-agricultural sector. The ruralpoor tend to rely on agricultural activities for a living andare less likely then non-poor householdsto diversify their sources of income, as we pointed out in Chapter 5. Interventionsthat helpimprove subsistence agriculture will not lift'households out of poverty but can improve food security and nutritional levels in times of crisis. Such a package could include the introduction of new seed and pasturingvarieties andmethods, andthe provision of training to reduce post-harvestlosses and improve food-conservationtechniques. Similarly interventions that allow the poor to participateinthe non-agricultural sectorcanhelp increaseaverage levels of householdincome at the sametime that they reduceincomerisk. We already discussedinChapter 4 that participation in the non-agricultural sector is positively correlated with human capital levels, access to credit, and access to basic services and road infrastructure. Inaddition workfare and public work programs can be usedto boost employment in andto provide connectionsto the non-agricultural sector amongpoor households. SOCIALMOBILITY 6.48 InChapters 4 and 5 we identified a worker's sector and type of employment and level of educationas key determinantsof individual andhouseholdincome. Indoing this, however, we took these factors, particularly the level of education, as given. 6.49 Inthis section we turn our attention to the impact that parental cultural and socioeconomic background, ratherthat individual andhouseholdcharacteristics,has on workers' occupationalchoice and level of education. We investigate the extent to which parentalbackground is correlated with children's outcomes, andwhether this correlation haschangedover time. 6.50 Why is this important? Peru, like the rest of the LatinAmerica, is a fairly unequalcountry, both in terms of income and endowments. A high correlation between parental background and children's outcomes, or low social mobility, will tend to perpetuatethese inequalities (De Ferranti et alia, 2004), while a low correlation will makeit possiblefor individuals from disadvantagedbackgroundsto breakout of poverty. 6.5 1 We then examine changes in social mobility using two different measures: economic mobility, measuredin terms of occupational mobility, and educational mobility. Inorder to do so we draw from existing work by Benavides (2002) inthe case of economic mobility andPasquier-Doumer (2002) in the case of educationalmobility. Economic Mobility 6.52 Benavides (2002) examines the patterns of occupational mobility across generations, where occupational choices are assumed to be a function of (i) the individual's socioeconomic background, 135 ability and effort, (ii) economic growth and the transformation of the productive structure it overall generates, and (iii) trends such as increasesinthe averagelevels of education and urbanization?6 secular 6.53 For the purpose of the analysis the author defines three cohorts and 9 occupational categories. The three cohorts being considered are individuals between 50 and 65 years of age who enteredthe labor market between 1960 and 1975, a time of rapid industrialization, individuals between 35 and 49 years of age who entered the labor market between 1975 and 1990, a period of negative growth, and finally individuals between 25 and 34 years of age who entered the labor market in the early 1990s, when economic growth was strong again. The choice of occupational categories responds to differences in human capital requirements, sector of employment (e.g. services), and type of employment-forma1 or informal (Table 6.8). For instance, the first category includes highly skilled professionals and government administrators, who have on average 15.6 years of education and earned in 2001 a monthly salary of 1,989 Soles. Occupational categories are then grouped inbroader incomestrata-high-medium or A, mediumor B and low-medium or C-for the sake of the analysis. Table 6.8: Workers can be classified into occupational categoriesaccording to their sector and type of employment andtheir education level Years of Monthly education income Income level (Soles) 1 Professionalsand government administration (high) 15.6 1,989 High-medium (A) 2 Professionalsand government administration (low) 15.5 1,277 High-medium (A) - 3a Small entrepreneurs 10.8 1,270 Medium(B) 3b Self-employed 9.4 735 Medium(B) 4 Salariedworkers 12.5 1,032 Medium(B) 5 Techniciansand other skilled workers (high) 12.2 941 Medium (B) 6 Skilled workers (medium) 9.8 748 Low-medium(C) 7 Unskilledworkers 9.2 622 Low-medium (C) 8 Rural workers 7.7 631 Low-medium(C) Source: Benavides (2002). 6.54 A simple examination of average mobility patterns over time shows that, while upwards mobility i s important, horizontal and downward mobility account for most of the observed occupational mobility. Upwards mobility (moving from C to B or from B to A) represents 40 percent of all mobility, while horizontal mobility (moving within A, B or C) and downward mobility (moving from A to B or B to C) represent 35 and 25 percent of all mobility respectively. In addition, mobility patterns do not differ significantly by cohort. 6.55 A more detailed analysis reveals that overall patterns mask important differences in the experiences of various occupational groups in terms of their likelihood to experience upward, horizontal and downward m~bility.~'A broad characterization of mobility patterns across groups is presented below: darker cells (numbered 1and 2) indicate a high number or density of observations, while lighter cells (4 and 5) indicate the opposite. 56. The author uses data from the ENNIV 2000 and2001, and restrictsthe sample to employed males between25 and 65 years oldresidinginurban areas. 57. The author uses linear and multiplicative logarithmic models to explore the issue of differences in mobility acrossoccupationalgroups. 136 6.56 Mobilityis relatively lower (Le. intergenerational correlation is relatively higher) for those inthe groups A and C than for those in group B (Table 6.9)?8 This implies that social mobility can to a large extent be accounted for by movements within categories in group B, and by movements between Byon the one hand, andA andC on the other. That is, economic mobility is the result of an increase inwhat is generally consideredthe low-middle class. 6.57 This pattern could be consistent with steady increases in average education levels and secular changes in the country's productive structure (Le. decline inthe share of agriculture and indfease inthe share of manufacturing and services), andlor with improvements in the equality of economic opportunities. Inorder to distinguish between these two sets of factors the author examines changes in mobility across cohorts. He finds very stable intra-a rcohort patterns, consistent with secular changes education levels and sectoral employment, and that the observed increase inoccupational mobility is due to the first set of factors. This conclusion is supported by the results on education mobility, orratherthe lackthereof, we discuss next. ~ ~~~ Source: Benavides(2002). EducationMobility 5.58 Average levels of education, measured as the averag of years of schooling, have increased steadily inPeru since the beginning of the twentieth c result of the expansionof the education system (Figure 6.1). Primary education has become universal in urban areas, the educational gap between men and women has declined significantly, and differences inaccess to primary education betweenindigenous and nan-indigenous people have decreased. 6.59 Certain groups, however, have failed to benefit from developments. Improvements in the level of education of the rural population have been less ed, especially among women, while inequality inaccess to secondary andtertiary education increased between indigenous and non-indigenous households. 58. These results are confirmedonce the author quantifies these differences by comparingthe relativeprobability ofmovingacrosscategories, conditional onparentaloccupation. 137 Figure6.1: Increase inyears of education I `*1 I 0 I 1905-1912 1913-1922 1923-1932 1933-1942 1943-1952 1953-1962 1963-1972 Birthcohorts /-Lima ---Other urban-Rural/ ISource: Pasquier-Doumer(2002). 6.60 This increase in the average level of education has been accompanied by improvements in educational mobility, measuredas a weakening of the correlation betweenthe education level of children and their parents. In other words, the likelihood that children are more educated than their parents has increased over time. Educational mobility has risen steadily among urban households since 1913-1922, while it has followed a more erratic pattern for rural ones, improving during 1913 and 1932, then stalling upto 1952, andagain increasingafter that. Inbothareas menand women havehadsimilarexperiences. 6.61 Education mobility can be thought of as the result of bothacross-the-boardincreasesineducation (access) and changesinthe relatively probability of educational success for a given set of socio-economic and cultural characteristics (equality of opportunity). That is, the average education level of the population can increase because everybody completes two more years of education than their parents andlor becausethe causalimpact of parental backgroundon their children's education diminishes. 6.62 Pasquier-Doumer (2002)59argues that most of the observed growth inaverage levels of education and education mobility i s due to generalized improvements inaccess to education. As a result, thoughthe share of illiterate children with illiterate parents has decreased over time, the probability that a child acquires no education given that her parents are illiterate has remained unchanged. The one exception to this pattern has been women in rural areas, for whom educational progress has responded to both improvements inoverall access and more equality of opportunities. What do these ResultsImply?Supply-side versusDemand-Side Interventions 6.63 The fact that most progress in education levels has been driven by general increases in access shows that supply measures, such as school construction and a higher number of teachers, have been effective in getting and retaining more children in schools. On the other hand, evidence that little progress has been made regarding what the author calls "democratization of education" indicates that there is room for alternative types of interventions that directly attempt to transform the relationship 59. The author usesdata from the 1974,1990 and 1996Employment Surveys(Ministry of Labor) andthe 1985-86 and2001ENNIV (Cuhnto). 138 between socio-economic and cultural background and education achievement-that is, demand interventions, such as income-based scholarships, conditional cash-transfers, and interventions that address cultural differences, such as bilingual education. We discuss both supply and demand interventions inmore detail inthe next section. 6.64 Ina country where inequality levels are still high, and where social exclusioncontinues to be a problem for certain groups, particularly the indigenous, directly tackling the issue of social mobility becomes a priority. Inthis regard Box 6.2 presents evidence that the gap between indigenous and non- indigenous people, measuredinterms of poverty and social outcomes, did not decreasedduringthe 1990s despite sustained improvements in access to education and health services across all population groups duringthe period. Box6.2: The gapinpoverty andsocialoutcomesbetweenindigenous andnon-indigenousgroups Benavides and Valdivia (2004) evaluate progress in poverty and other social indicators among the indigenous and non- indigenous population during the 1990s using data from the ENNIV (1992194 and 2000). We briefly summarize their conclusions inthis box. The authors argue that both indigenous and non-indigenous groups experiencedsubstantialimprovements inpoverty and other social indicators during the period. Both poverty and extreme poverty declined, primary enrolment increasedand several infant and maternalhealth and mortality indicators exhibited significant improvements(Table B6.2.1). These changes, however, did not translate into a reduction of the ethnic gap in education and health outcomes since improvements were similar in magnitude across both groups. The only exception to this rule was the relatively higher increaseininstitutional birthamong indigenousmothers. Table B.6.2.1: Poverty and socialindicators, 199214-2000 I measuredwith the ENAHO 1997-2003 Source:Benavidesand Valdivia (2004). ACCESS TOPUBLIC TRANSFERS, SERVICES AND INSTITUTIONS 6.65 Inthe previous two sections we have argued that the poor are more vulnerable and less socially mobile than the non-poor due to, among other factors, their lack of appropriate endowments. We explore here the role that public policy can play in lifting some of these barriers, thus helping the poor integrate better, both socially and economically. For this purpose we consider access to public services and institutions. 139 Access to public services 6.66 We present here a brief description of the differences across income levels, regions, and indigenous and non-indigenous groups in terms of access to education, health, infrastructure and public institutions. A detailed exploration of the determinants of access to public services and institutions i s beyond the scope of this report since it has received significant attention in the context of both the Public Expenditure Review prepared by the World Bank in 2003 and, more recently, the background documents prepared for the RECURSO study (World Bank, under preparation).M) Box 6.3 discusses international comparisons between indigenous andnon-indigenous earnings and education. Box 6.3: Educationand earning differentials betweenindigenousand non-indigenous adults: An international comparison This box discusses differences in educational attainment and labor market earnings between indigenous and non- indigenous populations in various countries, including Peru. In all the countries the rate of secondary-school completion among indigenous people in lower than that of non-indigenouspeople, but there is significant variation in the size of the difference between both groups (Table B6.3.1). Peru appears to be in an intermediateposition, with differential in completion rates that are smaller than those inBolivia, Ecuador or Mexico, but larger than those in the US. or Canada. Table B63.1: Seco lary-schoolcompletionratesare lower amongindigenousthan non-indigenousgroups ... Country Groups Secondary-schoolcompletion Ratioindigenous/ (%) non-indigenous Bolivia(2000) Indigenous 16 0.46 Non-indigenous 35 Ecuador (1999) Indigenous 9 0.32 Non-indigenous 28 Guatemala(2000) Indigenous 8 0.28 Non-indigenous 29 Mexico (2000) Indigenous 8 0.33 Non-indigenous 24 Peru(2003) Indigenous 27 0.56 Non-indigenous 48 us(2000) American-Indian 66 0.88 Restof population 75 Canada(2000) Aboriginal 52 0.75 Canadian-&rn white 69 Note: For Mexico "Indigenous" refersto residents inmunicipalities where more than 70percent of the populationspeaks an indigenous language. "Non-indigenous" refers to residentsof municipalities where les than 10percent of the population speaks anindigenous language. Source: World Bank (2004~). Similarly inall countries the returnsto educationare lower for indigenousworkers than for non-indigenousones, even after controlling for other observable characteristics. As before Peru appears to occupy an intermediate position regardingearningsdifferentials betweenindigenousand non-indigenousworkers, with levels that are higher than those of Bolivia and Mexico but lower than those in Ecuador or the US. However the fraction of this earnings differential thancannotbe explainedby differencesinobservablecharacteristicsis highest inPeru 60. See Annex 5 for abrief description of the objectivesand contentsof the RECURSOstudy. 140 Country Groups Returnsto schooling Ratioindigenous/ Unexplainedearnings (%) non-indigenous differential (%) UrbanBolivia(2000) Indigenous 6 0.49 27 Non-indigenous 9 Ecuador (1999) Indigenous 8 0.64 45 Non-indigenous 4 Guatemala(2000) Indigenous 10 0.54 42 Non-indigenous 11 Mexico (2000) Indigenous 9 0.26 42 Non-indigenous 11 Peru(2003) Indigenous 11 0.57 57 Non-indigenous 11 us (2000) American-Indian 7 0.75 36 Rest of population 6 Canada(2000) Aboriginal 7 0.62 9 Canadian-bornwhite 9 6.67 The main findingsregardingaccess to education canbe summarizedas follows (Tables 6.10a and 6.lob): Poor versus non-poor: Approximately 13percentof children inpoor householdsreport not attending school, comparedwith 8 percent innon-poor households. The differencebetweenpoor andnon-poor households i s more marked in rural areas and at higher levels of education. Important differences also exist between poor households, where children in female-headed households exhibit lower enrolmentratesthanthose inmale-headedones for all age groupsandlevels of schooling. Urban versus rural: The percentageof children of schooling age currently not enrolled in school is higher inrural than urban areas (14 percentinrural areas versus 9.7 percent inLima and 8.2 percent inother urbanareas). This difference is moreacutefor lower levelsof schoolingandamongyounger children. Important differences also exist within urban areas, where non-enrollment at the primary level is higher inmarginal neighborhoods("conos" inLima) than inmoreaffluent ones (not shown). Boys versus girls: Enrollment rates are higher among boys, especially for older groups and higher levels of schooling. These differences are more prominent among children in poor households and among those residing in rural areas. In addition, girls in female-headed households are less likely than boysto enroll in school, particularly secondary school, compared to their counterpartsinmale- headedhouseholdsirrespectiveof income. Indigenous versus non-indigenous: Enrollment rates are lower among poor indigenous children than among poor non-indigenousones. The same is true for non-poor indigenous andnon-indigenous children inrural areas, butnot inurban areas. 6.68 Other factors that appear to be correlated with enrolment even after controlling for income and location differences are the education level of the household head, which has a positive effect on enrollment, andthe age of the child, which has a negativeimpact on enrollment61. 61. The coefficients associated with these variables are significant in a probit regression model that identifies the determinants of the probability that a child is enrolled inschool. 141 Table 6.10a: Access to servicesvaries with income andacrossurbanand rural areas National Lima Other urban Rural Non Non Non Non Total Poor Poor Total Poor Poor Total Poor Poor Total Poor Poor Percentageof populationingroup Education-% not enrolled 5-17 years old All 11.0 8.4 12.7 9.7 9.6 9.8 8.2 7.1 9.2 14.2 9.1 15.6 Boys 10.1 8.7 11.0 10.4 11.2 9.4 7.9 7.0 8.7 11.9 8.4 12.9 Girls 11.9 8.1 14.3 8.9 8.1 10.1 8.6 7.2 9.8 16.7 9.8 18.4 5-11years old All 4.4 1.4 6.1 2.0 1.0 3.2 2.7 0.9 4.0 7.1 2.7 8.0 Boys 4.4 1.4 5.9 2.6 1.5 3.8 2.7 0.7 4.3 6.5 2.5 7.4 Girls 4.5 1.4 6.3 1.4 0.6 2.5 2.6 1.2 3.7 7.7 3.0 8.7 12-17 years old All 19.1 15.6 21.7 18.5 18.8 18.2 14.3 13.1 15.5 24.4 16.0 27.1 Boys 17.1 15.9 18.0 19.0 20.4 16.9 13.7 13.3 14.2 19.2 14.4 20.8 Girls 21.2 15.2 25.5 18.0 17.0 19.5 14.9 12.9 16.7 30.4 17.8 34.2 Health-Received treatment Public hospitallhealthcenter 48.6 36.7 65.3 40.2 33.0 60.8 36.8 27.8 53.8 71.5 65.2 75.4 ESSALUDhospitallhealthcenter 19.6 28.2 7.2 29.2 34.1 15.3 24.2 31.8 10.3 4.7 9.6 1.8 Private hospitallhealthcenter 11.0 15.4 4.8 17.1 20.3 8.4 11.7 14.7 6.1 5.0 8.8 2.7 Other (pharmacists, local healer) 22.3 21.5 23.5 15.3 14.9 16.4 29.6 28.1 30.7 19.6 17.6 20.8 Health Didnot receive treatment - 54.7 46.9 62.5 49.7 46.9 56.1 51.3 45.6 59.3 61.7 50.0 66.3 Health-Reason for no treatment Didnot havemoneylinsurance 25.3 16.0 32.8 16.3 11.7 25.5 23.4 16.0 32.1 32.2 23.0 35.0 Haddifficulty accessing center 8.5 5.9 10.6 1.7 1.7 1.7 5.0 5.6 4.3 15.8 13.2 16.6 Preferredother options 52.3 50.9 53.5 47.1 47.7 45.8 49.9 51.5 48.0 57.7 54.4 58.7 Other 9.4 11.3 7.8 8.3 9.1 6.7 11.5 13.4 9.3 7.8 9.7 7.2 Infrastructure Water 60.9 73.3 45.0 82.9 88.2 68.7 70.6 77.5 58.7 32.8 40.2 28.7 Electricity 72.4 85.9 55.1 98.0 98.5 96.9 91.2 95.8 83.2 31.8 45.9 24.1 Sanitary Services 47.5 64.3 25.9 81.4 87.0 66.3 61.1 72.3 41.6 5.7 10.8 2.8 Gas 47.3 64.3 25.6 80.8 83.8 72.9 58.1 70.9 35.9 8.7 18.9 3.2 Telephone 22.9 36.9 5.1 49.5 61.4 17.8 23.7 33.9 6.1 0.4 0.9 0.0 Distanceto Municivalitv 45.0 22.5 73.8 1.4 1.2 2.0 3.6 2.6 5.4 123.7 95.9 138.8 6.69 Insumchildren inrural areas andinpoor householdsare less likely to be enrolledinschool, and these differences are larger for lower levels of education and among girls. These patterns are the result of several supply and demand factors. First there are important differences in income levels and in school infrastructure across areas andneighborhoods. Secondthe opportunity cost of schooling i s larger inurban areas where markets are more dynamic and the productivity of labor is higher. Third households make decisions about the allocation of resources to the schooling of children based on their expectations on the children's ability and opportunity cost, and on the returns to their education investments. This can lead to intra-household inequality in education outcomes when resources (or credit) are limited or when competing investment opportunities are perceived as beingrelatively more profitable. 6.70 The main findings regarding access to health services can be summarized as follows (Tables 6.10a and6.lob): 142 Table 6.10b: Access to basicservices among the indigenous,particularly the poor, is low National Urban Rural Non Indigenous indigenous Indigenous indigenous Indigenous indigenous Poor Non Non Non poor Poor poor Poor poor Poor poor poor Poor poor Percentage of population ingroup Education-% not enrolled 5-17 years old All 13.6 8.0 10.2 8.4 10.0 7.6 7.1 8.5 18.0 9.8 11.7 8.2 Boys 12.8 6.8 8.7 6.5 9.7 6.3 7.6 7.0 16.7 9.1 9.3 5.3 Girls 14.3 9.2 11.7 10.4 10.3 8.9 6.7 10.2 19.4 10.6 14.2 11.0 5-11years old All 7.7 2.4 8.1 2.3 5.7 2.0 5.0 2.0 10.3 4.2 9.5 3.0 Boys 7.6 2.6 8.1 1.3 5.5 2.3 5.4 0.6 10.3 4.4 9.5 3.2 Girls 7.8 2.1 8.0 3.3 5.8 1.7 4.7 3.5 10.2 3.9 9.5 2.9 12-17years old All 21.4 13.9 13.1 13.7 15.6 13.5 9.8 13.9 28.7 16.1 14.8 13.0 Boys 19.6 11.4 9.5 10.9 15.2 10.8 10.6 12.1 24.9 14.1 9.0 7.4 Girls 23.2 16.4 17.1 16.5 16.0 16.0 9.1 15.8 32.9 18.4 22.0 18.1 Health-Received treatment Public hospitallhealthcenter 78.1 50.5 66.1 40.2 63.2 43.5 59.3 37.0 88.4 70.8 77.8 62.0 ESSALUDhospitallhealth center 7.5 23.6 14.1 29.4 13.5 26.9 18.8 31.6 3.4 14.0 6.2 14.7 Private hospitallhealthcenter 4.2 11.0 6.2 16.8 7.9 12.0 7.0 17.5 1.7 7.8 4.8 12.2 Other (pharmacists, localhealer) 12.6 19.3 15.5 18.4 19.0 22.3 16.9 19.0 8.2 10.7 13.1 15.0 Health Didnot receive treatmer - 29.8 23.7 25.2 21.2 27.4 23.9 23.6 21.0 31.1 23.1 27.6 22.2 Health-Reason for no treatmenl Didnot havemoneylinsurance 38.0 24.0 38.7 20.7 39.7 21.2 41.1 20.1 37.4 30.7 36.2 23.9 Haddifficulty accessingcenter 13.4 2.8 5.0 1.8 0.8 0.4 1.2 1.3 17.9 8.5 8.9 4.4 Preferredother options 52.9 38.1 38.1 26.8 41.0 30.2 27.0 23.5 57.1 56.7 49.6 43.7 Other 22.0 45.2 35.2 57.3 29.9 54.7 39.5 60.4 19.2 23.0 30.8 41.8 Infrastructure Water 41.9 62.2 49.7 75.7 59.1 72.6 63.2 82.0 33.8 42.8 31.8 41.0 Electricity 45.8 76.4 55.8 87.5 84.9 95.1 84.2 96.1 27.2 41.5 18.3 40.1 Sanitary Services 13.4 45.7 30.3 67.8 39.0 67.2 50.7 78.3 1.2 5.6 3.4 10.6 Gas 11.0 43.3 26.5 67.9 31.2 59.5 43.6 76.1 1.4 13.2 4.0 23.1 Telephone 1.2 20.9 5.9 37.8 3.4 31.8 10.4 44.5 0.1 0.4 0.0 1.2 Distanceto Municipality NA NA NA NA NA NA NA N A NA NA NA N A Note: The last year for whichdataon e icitv was collectedinthe ENAHOis 2001. Source: Authors' calculationsusingdatafrom ENAHO2001 (INEI) 0 Poor versus non-poor: Althoughthere do not exist important differences in sickness reports across poor and non-poor households (51 percent of the poor declare to have suffered an illness during the survey period, compared to 53 percent of the non-poor), there i s significant variation inthe way these episodes are treated in poor and non-poor households, interms of boththe actions taken and the kind 143 of care received. The poor are less likely to seek treatment that the non-poor42 percent of the poor report to have done nothing after falling sick, compared to 47 percent of the rich-due to lack or money or medical insurance andor difficulty to access a health center, When they do seek treatment the poor make much higher use of public hospitals and health centers and much lower use of ESSALUD services and privatehospitals than the rich. Urban versus rural: Households in rural areas make use of medical facilities less frequently than those in urban areas. When treatment i s sought the use of ESSALUD centers and private providers is higher in urban than in rural areas, although significant differences exist within urban areas with the use of public facilities beinghigher inmarginal neighborhoods than inmore affluent ones. Indigenous versus non-indigenous: Access to medical treatment is lower among the indigenous population and this difference i s more marked for poor households. The reasons for not seekindreceiving treatment vary across indigenous and non-indigenous households with distance and higher preference for alternative methods of treatment being more prevalent among the former. For those who do receive treatment the use of public hospitals and healthcenters i s more common among the indigenous population. 6.71 In sumthe urban and rural poor contact health care providers less often than the non-poor and when they do so they tend to use public health centers and hospitals rather than private ones. These differences respond to three basic factors. First public services tend to be cheaper than private ones and this makesthemrelatively more attractive for the poor. Second, within public services, the poor are more likely to access those that do not depend on affiliation and contributions to the Social Security system since they tend to be employed informally. Third, the public sector i s more likely to serve relatively poorer and more remote areas, so that access to public services is less costly for the poor in these areas, other things beingequal. 6.72 The main findingsregarding access to infrastructure can be summarized as follows (Tables 6.10a and 6.lob): 0 Poor versus non-poor: Poor households have lower access to infrastructure and tend to live further away from the center of their municipality than non-poor ones. These differences are more marked in rural areas. 0 Urban versus rural: Access to infrastructure is significantly lower in rural than in urban areas, particularly regarding sanitation and access to gas and phone services Inaddition, there is substantial variation in access within urban areas, with higher access rates inLima than in other urban areas and with lower rates inmarginal neighborhoods than inrichones. Indigenous versus non-indigenous: Access to infrastructure is lower among indigenous households thannon-indigenous onesirrespective of incomelevels andlocation. 6.73 Insum access to infrastructure is higher among non-poor households than among poor ones, mostly as a result of location and, to a lesser extent, income. Access, however, represents only one dimension of serviceprovision. Often poor households inurban areas, particularly inmarginal ones, have access to services but these are unreliable (e.g. water and electricity are only available a few hours a day) and of poor quality (e.g. water i s unclean). In other words, urban poor households have access to "overwhelmed" services (World Bank, 2004~). Access to public institutions 6.74 The importance of location and road infrastructure deserves further consideration as a determinant of the extent to which poor households have both access and the capacity to interact with public institutions, since these institutions tend to be concentrated in urban, more prosperous areas. We 144 use information on the frequency of contacts with various public institutions to explore this issue (Table 6.11). The poor interact less frequently with institutions representing both the central and local governments, although contacts with local authorities seem to be more equally distributed in rural areas. Use of financial institutions, even more accessible ones as public banks and saving cooperatives, i s less common among poor households than non-poor ones. Finally, poor households interact less frequently with law-enforcing institutions, such as thejudiciary systemand the police, than non-poor ones. National Lima Other urban Rural Non Non Non Non Total Poor Poor Total Poor Poor Total Poor Poor Total Poor Poor Percentageof populationingroup Government Municipality 27.8 29.3 25.9 23.3 27.3 11.9 31.9 32.5 30.9 27.5 26.5 27.9 MinistryofAgriculture 2.5 2.2 2.8 0.1 0.1 0.0 2.9 2.9 2.7 4.3 5.2 3.9 Ministry of Industry 0.3 0.4 0.1 0.3 0.4 0.0 0.4 0.6 0.2 0.1 0.1 0.1 Financial services Public banks 19.8 25.7 12.2 19.1 22.5 9.4 26.9 32.6 17.4 12.2 17.2 9.8 Justice and safety Judiciary services 3.8 5.0 2.2 3.8 4.9 0.9 4.9 5.9 3.2 2.4 3.1 2.1 Police 6.4 8.4 3.7 9.3 10.6 5.5 6.8 8.0 4.9 2.9 4.4 2.2 6.75 This, however, is not to indicate that distance to a particular institution is the only factor that limits access to it. Education levels, language, and socioeconomic background are also important determinants of the willingness and capacity of households to interact with public institutions. We discussthis inmore detailed inBox 6.4 for the case of thejudiciary system. Policy implications 6.76 The evidence presented in this section has shown that important differences in access to education, health and infrastructure between poor and non-poor households and between indigenous and non-indigenous households still exist. Although significant efforts were made during the 1990s to improve access to these services, the challenge remains. A challenge that i s particularly relevant given the role that these factors play as determinants of income, as discussedinChapters 4 and 5, and of social mobility and integration, as discussed above. Inthis sense the decline in public investment observed in recent years i s not a promising signal in terms of further progress towards poverty reduction and social inclusion. 6.77 We concentrate here on the discussion of policy interventions aimed at'increasing the coverage and quality of education and health services since interventions directed at increasing access to basic services and infrastructure through increases in public investment levels have already been presented elsewhereinthis report. 6.78 The main challenge regarding the implementation of these policies will be that of doing so in an increasingly decentralized environment. Peru i s moving fast towards the decentralization of its social sectors: social assistance, education and health. A full examination of the implications of the decentralization process for the social sectors is beyond the scope of this chapter as it will be extensively analyzed in the RECURS0 study. It i s important, however, to keep in mind what the main obstacles aheadmay be and we do so briefly after discussion our policy recommendations. 145 Box6.6: Improvingaccessto the Judiciary amongthe poor inPeru I Strongjustice sector institutions are critical for empowering the poor; however, access to justice continues to be severely limited for a large portion of the Peruvian population. Justice sector deficiencies hit the poor the hardest, thus aggravating inequalities in access to justice for those living in rural areas, women, and indigenous groups. This is especially important since the poor generally face disputes which are associated with their most basic needs (e.g. food, housing, health care, personal security, intra-family violence, or education), yet they are unable to use thejustice systemto resolvetheir disputes. The high cost of the services is a major deterrentfor low- income groups. The combination of court fees, lawyer fees, living and transportation expenses, and occasional bribesoften makesjustice servicesunaffordable. Institutional weaknesses inthis sector harmthe poor since they do not havethe resources to cope with the systemic failures. The low amount of justice services and their unevendistribution affects rural and indigenouspopulations disproportionately. In the poorest regions (in particular the Sierra), access to justice is more difficult due to acute geographical isolation of the communities and low coverage of the judiciary. Although coverage varies widely across judicial districts, rural areas are generally neglected with courts concentrated in the urban centers and relatively few first instancejudges inthe rural areas. Peruhas aratio of aboutonejudge per 16,000 citizens, and the number of public defenders is insufficient to cover the poor, with only 249 legal aid and public defenders to serve the entire country. In addition, cultural and linguistic barriers make indigenous use of formal justice less likely as lawyers and judges rarely speak indigenous languages, and most importantly, judges' decisions do not reflect the cultural reality inwhich indigenouspeoplelive. Certain obstacles make women's access to justice services more difficult than men's. Overall, more men than women use justice services, whether specialized courts (56.1 percent of men vs. 43.9 percent of women) or justice of the peace (58 percentof men vs. 42 percentof women). Institutional weaknesses of the Family Courts are particularly important as they tend to increase gender inequalities. Women are primarily affectedby Family Courts' low response to domestic violence cases and the lack of adequate legal aid for women seeking child support. In 1998,73.3 percent of child support cases presentedby women and 66.7 percent of those presentedby men did not reachjudgment in the family courts. Similarly, 67 percent of domestic violence cases in 1998 did not reach a judgment. Many more did not even reach the courts, as it is quite common for judges to resist qualifying domestic violence as a serious crime and following the prescribedprocedures. The burden of Family Courts' shortcomings falls mostly on poor women. The Justice Services Improvement Project, approved in March 2004, seeks to equate the demand in Peru for services with their availability and accessibility. Through the "Enhanced Access to Justice Component," the Project expects to provide technical assistance to strengthen the institutional capacity of the Judiciary (Poder Judicial), the Ministry of Justice, the Consejo Nacional de la Magistratura (Judicial Council), and the Academia de la Magistratura (Judicial Academy). The Projects also aims to improve sector performance in terms of quality and timeliness as well as enhance access of the country's mosteconomically and socially disadvantagedgroups. I Source: Contributionby the World Bank team working onthe Justice ServicesImprovementProject. Zncreusingthe coverage and quality of education 6.79 Access to good quality education is a function of both supply and demand factors. The presence of a secondary school in town will make it more likely for youngsters to continue studying after completing primary education because the cost of attendance is smaller the shorter the distance to the school. Similarly a higher need to complement household income will make it less likely for children in poor householdsto continue in school since their labor i s now more valuable. 6.80 Increasing the coverage and quality of education will require supply- and demand-side policies. Thesecould include, among others: 0 An increase inthe supply of pre-school and secondary education. Enrolment rates inpre-school and secondary education are relatively low in Peru, particularly inrural areas. Improvements in the supply of pre-school education can be achieved through non-fomal schooling modalities such 146 women-operated child education centers, which receive training and financial support from the government inexchange for the provision of basic education services. Improvements inthe supply of secondary education can be obtained through alternative, more flexible schooling modalities, such as distance learning. An improvementinthe allocationand quality of teachers. The distribution of teachers has been traditionally uneven, resulting in better quality teachers favoring urban areas and causing high absenteeismin rural areas. As a consequence the coverage and quality of education i s substantially lower inruralthan in urbanareas. A significant revision of the teacher payroll has been underway for a few years now and i s expected to result in improvements in teacher allocation. In addition incentives schemes aimed at increasing teacher attendance have been implemented in pilot form in rural areas. These schemes include monetary and non-monetary incentives, including teacher training. Lookingahead these efforts should be consolidated (i.e. application of incentive system at the nationallevel) and complemented with the provision of teacher training and materials particularly in the areas of bilingual and multilevel education. It will be important also to ensure that the decentralization process does not hinder the capacity of the authorities to manage the sector's human resourceseffectively and efficiently. An improvement in the supply and quality of bilingual education. Access to education i s particularly low among the indigenous population, especially among indigenous girls. This can be partly remedied by increasing the number of teachers trained inbilingual and multi-level education, and by developing and distributingthe relevant school materials to these kind of schools. Looking ahead effort towards the elimination of cultural barriers to access should take advantage of the increased accountability of the sector towards local authorities and users brought about by the decentralization process. An increaseinthe demandfor education. The demand for education is a function of both its direct cost and the opportunity cost of schooling, as well as of its perceived value, both in the form of returns inthe labor markets and interms of social value. Increasesin the demand for education can then be induced by effectively lowering its costs through conditional cash-transfer programs or scholarships, and through the implementation of flexible schooling schedules that allow children and youngsters to engage in other activities during the day; or by effectively increasing its value through improvement in education quality and cultural relevance (e.g. bilingual education for indigenous people). Zncreasing the coverageand quality of health services 6.81 As was the case with education, an increase in supply is not enough to ensure that the poor, particularly those inrural areas, have access to health services. Economic and cultural barriers contribute to making inequity inaccesspersistent. 6.82 Increasing the coverage and quality of health services will then require supply- and demand-side policies. These could include, among others: a An increase in the supply of basic health services to the poor. Making health services more accessible to the poor and particularly to those who are more vulnerable among them, such as mothers, infants and the elderly should be a priority. The recent creation of the Seguro Integral de Salud (SIS) has been a significant step inthis direction. The SIS eliminates user-fees by reimbursing public providers on a fee-for-service basis for all variable costs incurred during the provision of a basic benefit package (mainly essential drugs and medial supplies). The numbers of mothers and infants receiving attention under the SIS have increased steadily since its creation, and further increases incoverage are expected inthe future. Looking ahead the main challenges for the SIS are the improvement of skilled birth attendance and the control of excessive allocation of resources to tertiary care. 147 An increaseinthe efficiency of andcoordination among public healthproviders. Healthservices are provided by different suppliers in Peru, including the Ministry of Health and the ESSALUD, associatedto the Social Security authority, which cover different population groups. The existence of multiple providers with different mandates can potentially cause inefficiencies in the allocation of resources and inthe use of existing capacity. Inorder to increaseefficiency inthe healthsystem, the Ministry of Health has signed a series of management agreements with regional health authorities. These agreements link resources to performance and outcomes. Looking ahead the main challenges regarding the managementagreementsinclude the monitoring and publication of performance results. In addition, in order to maximize the use of existing capacity the Ministry is seeking better coordination with ESSALUD. This, however, has proved politically difficult. Looking ahead the efforts to promote coordination between both institutions should be renewed. This will be particularly important in an increasing decentralized environment where the risk of fragmentation in the systemmay rise significantly. A reduction of cultural barriers. Access to health services is particularly low among the indigenous population due, among other factors, to cultural barriers. Better accommodating the cultural expectations and beliefs of indigenous people within the health system can go a long way in eliminating or at least mitigating the impact of these barriers. The adoption of the CLAS model in 1994, based on the participation of local communities in the planning and management activities of primary health care centers, has constituted an important move inthis direction. Looking aheadthe GOPshould continue working towards the consolidation of this model. More generally strengthening the responsivenessof healthservices to localneeds, as well as the capacity of communities to monitor healthproviders will be even more important ina decentralized environment. An increase in the demand for health services. The demand for health services is a function of both its cost and its perceived value. Increasesinthe demand for health services can then be induced by effectively lowering its costs through subsidized services, such as that provide through the SIS, or through conditional cash-transfer programs; or by effectively increasing its value through improvements inthe quality and cultural relevance of the service provided. 6.83 Many of the reforms in the education and health sectors described above have been supported through a series of World Bank programmatic and technical operations (the PRSL I-IV). Inaddition, as the decentralization of the social sectors progresses,financial and technical support to further reforms will be provided through the RECURS0 project and through an operation directed to increase accountability for the decentralization of the social sectors. CONCLUSIONS 6.84 We have presented evidence i s this chapter that the poor are more vulnerable and tend to have lower access to basic services than the rich. As a consequencethey are exposed to higher (income) risks and exhibit lower levels of social mobility. 6.85 We have argued that the root of higher vulnerability among the poor can be found in the lack of saving capacity of the poor and intheir limitedaccess to financial markets and safety nets. Policies aimed at removing some of these barriers can then go a long way inreducing vulnerability and exposure to risk among poor householdsinurban and rural areas. 6.86 We have also argued that increasesin social mobility and opportunities will depend on increases in access to basic services, such as education and health, and to public institutions. These in turn will require improvements inthe level and quality of resources, as well as inthe efficiency of and coordination among the different actors in the sectors. All of which will have to be achieved in an increasingly decentralized environment. 148 BIBLIOGRAPHY Alchzar, Lorena. 2004. "Monitoring Social Outcomes and Policies in Peru: the Challenge of Decentralization." Processed. Grade, LimaPeru. Altamirano, Teofilo, James Copestake, Adolfo Figueroa and Katie Wright. 2003. "Poverty Studies in Peru: Towards a More Inclusive Study of Exclusi6n." WeD Working Paper 05, Wellbeing in DevelopingCountries Economic and SocialResearchCouncil ResearchGroup, Bath, UK. Alvarado Perez, Betty and Enrique Chon Yamasato. 2002. "Aproximaci6n a la pobreza de Lima."Processed. World Bank,Washington DC. Arias, Omar and Walter SosaEscudero. 2004. "Subjective and Objective Poverty inBolivia." Processed. The World Bank, Washington, DC. Benavides, Martin. 2002. "Cuando 10s Extremos no se Encuentran: UnAnalisis de la MovilidadSocial y la Igualdad de Oportunidades en el Peru Contemporaneo". Bulletin de 1'Institut Francais d'Etudes Andines, Tome 31, No. 3. Benavides, Martin and Martin Valdivia. 2004. "Metas del Milenio y la Brecha Etnica en Peru''. Processed. Grupode Anhlisis para el Desarrollo (GRADE), LimaPerk Botero, Juan, Simeon Djankov, Rafael La Porta, Florencio Lopez-de-Silanes and Andrei Shliefer. 2003. "The Regulation of Labor". National Bureau of Economic Research Working Paper Series No. 9756. NBER, Cambridge, MA. Calderbn, CCsar and Luis Servtn. 2004. `The Effects of Infrastructure Development on Growth and Income Distribution." Processed. Central Bank of Chile andthe World Bank, Washington, DC. Calvo, Guillermo, ed. 2004. GoodJobs Wanted. Labor Markets in Latin America. The Inter-American Development Bank and the John Hopkins University Press, Washington, DC. Casa, C. and G. Yamada. 2005. "Medicion de Impact0 en el Nivel de Vida de la Poblaci6n del DesempeiioMacroeconomicopara el Period0 2001-2004". Universidaddel Pacifico, Lima,Peru. Chong, Alberto and Jaime Saavedra. 2003. "Explaining Increases in `Legal Informality' During the Ninties inLatin America: The Case of Ped." Processed. De Ferranti, David, Guillermo Perry, William Foster, Daniel Lederman and Alberto Valdes. 2005. Beyond the City. TheRural Contribution to Development. The World Bank,Washington D.C. De Ferranti, David, Guillermo Perry, FranciscoH.G. Ferreira and Michael Walton. 2004. Zneq~alityin Latin America and the Caribbean. Breaking withHistory? The World Bank, Washington D.C. De Ferranti, David, Guillermo Perry, Indermit Gill, J. Luis Guasch, William Maloney, Carolina Shnchez- Phramo and Norbert Schady. 2003. Closing the Gap in Education and Technology. The World Bank, Washington D.C. Diaz, Juvenal. 2004. "Cobertura e incidencia de 10s Programas Sociales Focalizados". Processed. World Bank. 149 Duclos, J., J. Esteban & D. Ray. 2003. "Polarization: Concepts, Measurement, and Estimation". Processed. Available at: http://l32.203.59.36/CIRPEEIcahierscirpee120031fles/CIRPEE03- 0l.pdf Easterly, William, and Luis Servkn, eds. 2003. The Limits of ~tabilization. lnji-astructure, Public Deficits, and Growth in Latin America. Stanford University Press and the World Bank, Palo Alto, CA. Economist IntelligenceUnit,The. 2005. Peru: Country Profile 2005. The Economist. United Kingdom. Escobal, Javier. 2005. "Rural Poverty in Peru: The Role of Agriculture and Non-agriculture Sectors." Processed. Grupode Analisis parael Desarrollo (GRADE), LimaPen?. Escobal, Javier. 2002. "El rol de 10s activos pbblicos en la generaci6n de oportunidades de empleo rural no agropecuario en el Ped." Processed. Grupo de Analisis para el Desarrollo (GRADE), Lima Perb. Escobal, Javier. 2001a. "El beneficio de 10s caminos rurales: ampliando oportunidades de ingreso para 10s pobres." Documento de Trabajo #40, Grupo de Analisis para el Desarrollo (GRADE), Lima Ped. Escobal, Javier. 2001b. "The Determinants of Non-Farm Income Diversification in Rural Peru." World Development29 (3): 497-508. Escobal, Javier and Carmen Ponce. 2002. "El Beneficio de 10s Caminos Rurales: Ampliando Oportunidades para 10s Pobres". Processed. Grupo de Analisis para el Desarrollo, (GRADE) Lima Ped. Escobal, Javier, Jaime Saavedra and Maxim0 Torero. 1998. "Los Activos de Los Pobres en el Perk" Processed. Grupode Analisis para el Desarrollo, (GRADE) Lima Ped. Escobal, Javier, and Maxim0 Torero. 2002. "How to Face an Adverse Geography: The role of Private and Public Assets." Processed. Inter-American Development Bank, Washington, DC. Escobal, Javier, and Maxim0 Torero. 2000. "Measuring the Impact of Asset Complementarities: The Case of Rural Peru." Processed. Grupo de Analisis para el Desarrollo(GRADE), Lima, Perk Escobal, Javier and Martin Valdivia. 2004. "Peru: Hacia una Estrategia de Desarrollo para la Sierra Rural." Processed. Grupo de Analisis parael Desarrollo (GRADE), Lima Pen?. Francke, Pedro. 2005. "iQu6 esta pasando con la pobreza y la distribuci&?'. Processed. Rio Abierto, Lima, Peru. Francke, Pedro. 2004. "iQu6 nos dicen las recientes investigaciones sobreprogramas sociales?Una nota para la discusi6n." Processed. The Peruvian Consortium for Social and Economic Research (CIES), PontificiaUniversidadCat6licadel Ped. Fay, Marianneand Tito Yepes. 2003. "Investing inInfrastructure: What i s Needed from 2000 to 2010?' World Bank Policy ResearchWorking Paper 3102, Washington, DC. 150 Felicio, Mariana and InduJohn-Abraham. 2004. "Peru: Towards a System of Social Accountability." En Breve No.39. Ferreira, Francisco and Michael Walton, eds. 2006. "World Development Report 2006. Equity and Development." Draft. The World Bank, Washington DC. Foster, Andrew D. and Mark R. Rosenzweig. 1996. "Technical Change and Human-Capitalreturns to Investments: Evidence from the Green Revolution". American Economic Review 86(4). Gasparini, Leonardo, Martin Cicowiez Federico GutiCrrez and Mariana Marchionni. 2003. "Simulating Income Distribution Changes in Bolivia: a Micro-Econometric Approach." Processed. World Bank, Washington, DC. Gasparini, Leonardo, Federico Gutierrez and leopoldo Tornarolli. 2005. "Growth, Income and Poverty in Latin America and the Caribbean: Evidence from Household Surveys". Processed. UniversidadNacionalde L a Plata. Buenos Aires, Argentina. Gibson, John, Jikun Huang and Scout Rozelle. "Improving Estimates of Inequality and Poverty from Urban China's Household Income and Expenditure Survey". Working Paper, No. 1028. Department of Agriculture and ResourceEconomics, UC Davis, California. Gill, Indennit, Truman Packard and Juan Yermo. 2004. Keeping the Promise of Old Age income Security in Latin America: A Regional Study of Social Security Reforms. The World Bank. Washington, DC. GonzAlez Mantilla, G., Servh, J.C., Lcjpez, L.,and Burgos, H. 2002, "El Sistema Judicial en el Peri: Un EnfoqueAnalitico a Partirde sus Usosy Usuarios." Processed. The World Bank. Hall, Gillette and Harry Anthony Patrinos. 2004. "Indigenous People, Poverty andHuman Development inLatinAmerica: 1994-2004." Processed. The World Bank, Washington, DC. Heckman, James J. and Carmen PagCs. 2004. Law and Employment. Lessonsfrom Latin America and the Caribbean. The University of ChicagoPress, Chicago, IL. Herrera, Javier. 2003. "La Pobreza en el Per& 2003. Advertencia sobre cambios metodolcjgicos." Processed. Institutde Recherchepour le DCveloppement(IRD)y el INEI. Herrera, Javier, ed. 2002a. Pobreza y Desigualidad en el Area Andina. Bulletin de L'Institut Frangais d'Etudes Andines, Lima, Peru. Herrera, Javier andFrancois Roubaud. 2002. "Dinamica de la PobrezaUrbana en Peru y en Madagascar 1997-99: Unanalisis de Datos de Panel". Bulletin de 1'Institut Francais $Etudes Andines, Tome 31,No. 3. Herrera, Javier. 2004. "Cifras macroeconcjmicas y condiciones de vida de 10s hogares." Processed. InstitutdeRecherchepour leDCveloppement (IRD-DIAL). Hnatkovska, Viktoria, and Norman Loayza. 2003. "Volatility and Growth." Processed. INEI. 2004a. ``Situaci6n del Mercado Laboral en Lima Metropolitana." Znforme de Empleo No. 6. InstitutoNacional de Estadistica e Infodtica, Lima Per& 151 INEI. 2004b. "Indicadores Econ6micos." Infonne deEmpleoNo. 6. InstitutoNacionaldeEstadisticae Infodtica, Lima Perk INEI. 2004c. Perk Principales IndicadoresRegionales 2002-2003. InstitutoNacionalde Estadistica e Infordtica, Lima Peru. INEI. 2003a. Perk CornpendioEstadistico2003. InstitutoNacionaldeEstadisticae Informiitica,Lima Peru. INEI. 2003b. Encuesta Nacional de Hogares2003. Condicionesde Viday Pobreza. InstitutoNacional de Estadisticae Informiitica, Republica de Peru. INEI. 2002a. Condicionesde Vidaen el Perk Evolucicin, 1997-2001. EncuestaNacionalde Hogares (ENAHO). Programa MECOVI-Peh, InstitutoNacionalde Estadistica e Infodtica, Republica de Perti. INEI. 2002b. La Pobreza en el Perri en 2001 Una visicin departamental.. Nacional de Estadistica e Infordtica, Republicade Perti. INEI. 2001. Nuevas Estimaciones de la Pobreza en el Perii, 1997-2000. Nacional de Estadistica e Infodtica, Republicade Perti. INEI. 1998a. Encuesta Nacional de Hogares 1998. IV Trimestre. Condicionesde Vida y Pobreza. Programa de Mejoramiento de las Encuestas de Condiciones de Vida en el Perti. Instituto Nacional de Estadistica e Infodtica, Rep~blicade Per& INEI. 1998b. Peni: Niveles de Viday Pobreza 1998. EncuestaNacionalde Hogares 1998. Colecci6n Estudios e Investigaciones. InstitutoNacional de Estadistica e Infodtica, Repdblica de Ped. INEI. 1997. Perk Medicidn de Niveles de Vida y Pobreza. Encuesta Nacional de Hogares 1998. Colecci6n Estudios e Investigaciones. Instituto Nacional de Estadistica e Informhtica, Repdblica de Per& Jaramillo, Miguel. 2004a "La Regulaci6n del Mercado Laboral en Pedi." Processed. Grupo de Anhlisis parael Desarrollo (GRADE), Lima Perti. Jaramillo, Miguel. 2004b "Transaction Costs in Peru: How much does it cost to start a formal garment firm in Lima?"Paper presented at the Annual Meeting of the International Society for Neo InstitutionalEconomics, Tucson, Arizona. Jaramillo, MiguelandJaime Saavedra. 2003. "Gobernabilidad, Reformas y Desempeiio Econ6micoenel Pex4de 10s 1990s." Loayza, Norman and Rossana Polastri 2004. "Poverty and Growth in Peru." Background Report for Peru's PovertyAssessment. Processed. World Bank, Washington DC. Loayza, Norman, Pablo Fajnzylber and CCsar Calder6n. 2002. "Economic Growth inLatin America and the Caribbean. Stylized Facts, Explanations and Forecasts." Processed. Research supported by the World Bank's Latin America RegionalStudiesProgram. 152 Lopez-Cglix, JosC R. 2004. "Budgetary Protection of Social Spending in World Bank Programmatic Adjustment Loans. Lessons from Peru." Paper for the seminar on "El Futuro de la Protecci6n Presupuestariaen Perii, Lima, Perb. Processed. Lopez-Cdix, JosC R., y Alberto Melo. 2004. A mas disciplina~scaZ,menospobreza. Revisiondel gasto publico en Peru. Inter-American DevelopmentBank, Colombia. Lopez-Chlix, JosC R, and Karina Rozas. 2003. "Preserving Pro-Poor Social Budget Protection in the Context of Decentralization." Processed. World Bank, Washington, DC. MacIsaac, Donna and Martin Ram. 2001. "Mandatory severance Pay: An Assessment of its Coverage andEffect inPeru". Processed. World Bank,Washington, DC. Maloney, William. 2003. "Informality Revisited". Processed. World Bank, Washington, DC. Ministerio de Econom'a y Finanzas. 2004. Marc0 Macroecondmico~ultian~al2005-2007. RepGblica de Perti. Ministerio de Trabajo y Promoci6n del Empleo. 2004a. "Boletin de Econom'a Laboral 28-28." Ministerio de Trabajo y Promoci6ndel Empleo y Programade Estadisticasy EstudiosLaborales, Lima,Perb. Ministerio de Trabajo y Promoci6n del Empleo. 2004b. "Informe Estadistico Mensual." Direcci6n de Promoci6ndel Empleoy Formacih Profesional. Lima, Perti. Ministerio de Trabajo y Promoci6n del Empleo. 2003. "Informe Estadistico Mensual." Direcci6n de Promoci6n delEmpleo y Formaci6nProfesional. Lima, Ped. Ministerio de Trabajo y Promoci6n del Empleo. 2002a. "Informe Estadistico Mensual." Direcci6n de Promoci6ndel Empleo y Formaci6nProfesional. Lima, Ped. Ministerio de Trabajo y Promoci6n del Empleo. 2002b. "Boletin de Economia Laboral 21." Ministerio de Trabajo y Promoci6n del Empleo y Programa de Estadisticas y Estudios Laborales, Lima, Perti. Ministerio de Trabajo y Promocidn del Social. 2001. "Informe Estadistico Mensual." Direccibn de Promoci6n delEmpleo y Formaci6nProfesional. Lima, Ped. Pasquier-Doumer,Laure. 2002. "La Evolucion de la Movilidad escolar Intergeneracionalen el Perii a lo largo del Siglo XX". Bulletinde 1'Institut Francaisd'Etudes Andines, Tome 31, No. 3. Paxson, Christina andNorbert Schady. 2004. "Child Healthand economic Crisis inPeru". Forthcoming inTheWorldBank EconomicReview. Psacharopoulos, George and Harry Anthony Patrinos (eds). 1994. Indigenous People and Poverty in Latin America: An Empirical Analysis. ' Washington, DC: the International Bank for Reconstruction. Saavedra, Jaime and Alberto Chong. 1999. "Strutural Reforms, Institutions and the Informal Sector in Peru". Journal of Development Studies35(4). 153 Saavedra, Jaime and Mhximo Torero. 2004. "Labor Market Reforms and their Impact over Formal Labor Demand and Job Market Turnover: The Case of Peru." InLaw and Employment: Lasons from Latin America and the Caribbean, eds. James J. Heckman and Carmen Pages. Nacional Bureau of Economic Research, Cambridge, MA. Saavedra, Jaime and Eduardo Maruyama. 1999. "Estabilidad Laboral e Indemnizacibn por Despido: Efectos sobre el Funcionamiento del Mercado Laboral Peruano." Processed. Grupo de Anhlisis parael Desarrollo(GRADE), LimaPerti. Schady, Norbert. 2004. "Do Macroeconomic Crises Always Slow Down Human Capital Accumulation?' The WorldBank EconomicReview, 18 (2). Sheahan, John. 2003. The Persistence of Poverty in Peru: Possible answers, their Limits, and their Implications for Latin America. Development Alternatives, 1nc.-Boston Institute for DevelopingEconomies, Boston, MA. Sosa-Escudero, Walter and Leonard0 Lucchetti. 2004. "Exploring the Determinants of Poverty and Income Distribution inPeru." Processed. Centro de Estudios Distributivos, Laborales y Sociales (CEDLAS), UniversidadNacionalde L a Plata, L aPlata, Argentina. Tesliuc, Cornelia M. 2005. "Strengthening the Compact-Social Protection." Processed. The World Bank. Trivelli, Carolina. 2004. "Indigenous Poverty in Peru: An Empirical Analysis." Processed. Instituto de Estudios Peruanos(IEP). Perti. Universidaddel Pacifico. 2003. "Primer Informe: Disefiode Modelos de Prediccibn de Cumplimiento de Objetivosdel Milenio." Processed. Centro de Investigacih de laUniversidaddel Pacifico. Valdivia, Martin and M. Robles. 1997. "Decisiones Laborales en la Economias Rurales del Peru". Processed. Grupo de Anrilisis parael Desarrollo (GRADE),Lima Penj. Webb, Richard and Graciela Fernandez Baca (eds). 2005. Anuario Estadistico. Peru en Numeros 2004. InstitutoCuanto, Lima,Peru. Webb, Richardand Graciela Fernandez Baca (eds). 2004. Anuario Estadistico. Peru en Numeros 2003. InstitutoCuanto, Lima, Peru. World Bank. 2005a. Doing Business in 2005. Removing Obstacles to Growth. The World Bank, the InternationalFinance Corporationand Oxford University Press, New York, NY. World Development Report. 2005b. A Better Investment Climate for Everyone. The World Bank and Oxford University Press, New York, NY. World Bank. 2004a. Country Assistance Strategy Progress Report for the Republic of Peru. World Bank, Washington, DC. World Bank. 2004b. "Project Appraisal Document for a Proposed Technical Assistance Loan in the , amount of US$8.0 million to the Republic of Perufor InstitutionalCapacity for SustainableFiscal Decentralization." Processed. World Bank, Washington, DC. 154 WorldBank. 2004c. The Urban Poor in Latin America. Washington, DC: World Bank. World Bank. 2004d. Proposed Programmatic Social Reform Loan IV in the Amount of US$lOOMillion to theRepublic of Peru. Washington, DC: WorldBank. WorldBank. 2004e. Ecuador Poverty Assessment. Washington, DC. World Bank. 2004f. "Study on the Environmentand Social Dimensions of the Mining Sector in Peru." Processed. The World Bank, Washington, DC. World Bank. 2003a. Restoring Fiscal Discipline for Poverty Reduction in Peru: A Public Expenditure Review. Washington, DC: World Bank. World Bank. 2003b. Peru: Microeconomic Constraintsto Growththe Evidencefrom theManufact~ring Sector. Washington, DC: World Bank. WorldBank. 2003c. Romania-Poverty Assessment. TheWorld Bank, Washington, D.C. World Bank. 2002. Country Assistance Strategyfor the Republic of Peru. World Bank, Washington, DC. 155 ANNEX 1:METHODOLOGICAL ISSUES REGARDINGPOVERTYMEASUREMENTIN PERU, 1997-2003 There exist two different household surveys inPeru, the EncuestaNacional de Niveles de Vida (ENNIV) administered by Cuinto and the Encuesta Nacional de Hogares (ENAHO) administered by the Peruvian Statistical Institute (InstitutoNacional de Estadistica e Infonnacion, INEI). Because both surveys differ interms of their format andbothinstitutions apply different methodologies to measurepoverty, there also exist two different sets of poverty figures (Table A.1.1). Table A.1.1: Poverty headcountmeasuredby the ENNIV and the ENAHO National Urban Rural 1 1991A 57.4 49.9 52.2 36.0 70.8 74.0 53.4 46.4 50.4 34.7 65.6 68.6 1997 1994A 50.7 42.7 48.9 29.7 64.8 66.3 2000 54.1 48.4 47.9 36.9 66.1 70.0 ENAHOfiguresfor 1991and 1994are notentirely comparableto those for 1997and2000. Despite the fact the analysis presented in the previous Peru Poverty Assessment was based on the ENNIV, we decidedto usethe ENAHOfor our poverty calculations due totwo reasons. First, it provides the most up-to-date data (2003, compared with 2000 for the ENNIV). Second, it produces more detailed income and consumption measures and, as a result, more accuratepoverty numbers. Preparing time-consistent, comparable poverty numbers for 1997-2003, however, required that we dealt with a number of methodological improvements to the survey implemented over this period. For this purpose we reliedheavily on the work produced by the INEI,particularly by the HouseholdSurvey team. The purpose of this annex i s then twofold. First we briefly describe the methodological changes in the ENAHO, as well as the strategiesappliedto deal with them. Second, we review the main methodological differences between the ENAHO and the ENNIV, paying special attention to the effect that these differences have on income, consumptionand poverty measurement. A more detailed discussion of both issuescan be found inthe various INEIdocuments cited inthe text. Changesinthe ENAHO1997-200362 During 1997-2003, the ENAHO has undergone significant methodological improvements and revisions regarding the size and composition of the survey sample and the definition and construction of the poverty line. We briefly comment on bothbelow. Changesin thesample size and composition. The size of the survey sample varied during the period (Table A1.2), falling significantly between 1997 and 2000 and increasing afterwards. As a result, the level of representativeness of the survey also changed over time-at the level of the ``d~minio~~"between 1997 and 2000, and at the level of the department from 2001 onwards. 62. This section draws heavily from see INEI (2001), INEI (2002), and Herrera (2003). The interested reader shouldconsultthese documents for a detailed discussionon the issuespresentedhere. 63. There are seven "dominios": Metropolitan Lima, rural Sierra, urban Sierra, rural Costa, urban Costa, rural Selvaand urbanSelva. 157 Table A.1.2: Survey sample size, 1997-2003. Sample size Sample size (unweighted) (weighted) 1997 I 31,280.00 25,178,626.00 1998 35,509.00 27,319,804.00 1999 18,783.00 28,743,428.00 2000 17,177.00 26,732,023.00 2001 75,470.00 27,2 19,122.00 2002 83,807.00 27,483,404.00 2003 1 84,397.00 27,308,177.00 Source: Authors' calculationsusingENAHO 1997-2003 In addition, in 2001 the sampling framework was modified to account for demographic changes as documented in the 1999 Pre-Census. The new sampling framework included new (peripheral) urban areas, especially around Lima, that had developed during the 1990s and thus were not covered by the previous sampling framework basedonthe 1993Population Census (Table A1.3). All New areas Oldareas % of all sampled households National 100.0 18.3 81.7 NorthernCosta 100.0 30.4 69.6 CentralCosta 100.0 24.6 75.4 SouthernCosta 100.0 28.8 71.2 NorthernSierra 100.0 9.9 90.1 CentralSierra 100.0 13.5 86.5 SouthernSierra 100.0 10.5 89.5 Selva 100.0 17.8 82.2 Metropolitan Lima 100.0 19.6 80.4 Changes in thepoverty line. The overall poverty line is a function of the extreme poverty line, which in turns i s a function of three different elements: (i)the definition of the caloric norm (or minimumdaily caloric consumption), (ii) the reference populationfrom which information about the consumption basket that yields the caloric normis obtained, and (iii) construction and valuation of this consumption basket. All three elements have the experiencedchanges during 1997-2001. From 1997 to 2000 the ENAHO used a single caloric norm for all individuals (2,318 calories per person per day), independent of age, gender and area of residence, therefore disregarding differences in nutritional needs across demographic groups and across different types of economic activity. This changed in2001 when a more nuance approachedwas adopted. Usingthe 1993-4 Encuesta Nacional de Propositos Multiples, the INEI estimated caloric requirements by "dominio" for a representative five- member household (two adults and three children), attributing different requirements according to age and gender. Although the average caloric normvaried across "dominios", reflectingdemographic differences, these "dominios" could be grouped intothree categorieswith three different caloric norms (2,232 calories per person per day in metropolitan Lima; 2,194 calories per person per day inthe Costa, the urban Sierra andthe urban Selva; 2,133 calories per person per day inthe rural Sierra andthe ruralSelva). There were also changes in the way the reference population was defined. While in 1997-2000, consumption patterns were identified separately for eachnaturalregion by usingregion-specific reference 158 populations, in 2001 the INEI adopted a single-reference population approach. Under the old methodology, differences in (extreme) poverty lines across regions captured price level disparities and differences infood consumption patterns, but they did not take into account behavioral differences caused by regional variation in income and consumption levels+.g. poor households tend to spend a larger fraction of their income in food that wealthier households. As a consequence, the estimated Engel coefficient* for low-income areas was systematically below that of high-income areas. These problems were solved by adopting a single population of reference for the entire country, together with a regional price deflator that made income and consumption across regions comparable. Finally, boththe composition of the consumption aggregate and the way inwhich the value of the poverty line was updated over time were modified. The set of expenditures to be included in the consumption aggregate was changed to make it compatible with the National Accounts Framework in 2001-for instance, double accounting associated with inter-household transfers was eliminated, the method use to compute consumption of durable goods was improved, indirect taxes not directly associated with the provision of a particular service were excluded from the consumption aggregate. Similarly, the INEI shifted from maintaining the extreme poverty line constant in real terms from year to year and updating the overall poverty line on an annual basis using yearly Engel coefficients, to maintaining the overall poverty line constant using the Consumer Price Index (CPI) for the main cities. This shift eliminated inconsistencies across years inpoverty measurementcaused by the negative correlationbetween total and food expenditure levels. For the purpose of this report we present two series of poverty numbers. The first series incorporates all changes described above for 2001-2003. The advantage of this approach i s that poverty numbers for 2001-2003 take into account the population residing innewly developed areas and are, as a result, more accurate and representative. The disadvantagei s that these numbers are not comparable to the 1997-2000 ones. The second series recalculates the consumption aggregate and the poverty line for 2001-2003 using the sampling framework and the methodology applied in 1997-2000 thus producing comparable poverty numbers for 1997-2003. The table below illustrates the relative importance of sampling changes and changes inthe poverty line inexplaining changesinpoverty rates for 2000-2001 (Table A.1.4). TableA.1.4: Variationinpovertyfiguresdueto samplingandmethodologicalchanges Comparableseries Actual 2000 2001 2001 Total Variation due to changes in Variation due to variation samplingframework changes inpoverty line ~ a t i o n a ~ I 48.4 49.8 54.8 6.4 2.5 2.5 Urban 36.9 35.7 42.0 5.1 4.2 2.I Rural 70.0 75.9 78.4 8.4 -0.8 3.3 A comparisonbetweenthe ENAHOandthe ENNIV65 There are three potential causes for the discrepancies between the poverty numbers produced using the ENAHO and those produced using the ENNIV: (i)sample differences across both surveys, (ii) methodological differences regarding the calculation of poverty lines, and (iii) differences in the computation of household expenditure. We briefly explore all three here using information for ENNN 64. The Engel coefficient captures the relationship between food and total expenditures and it is use to construct the overall poverty line from the extreme poverty line (i.e. income required to afford the basic food consumptionbasket). 65. Thediscussioninthis sectionclosely follows that presentedby J. Herrera inINEI(2001). 159 1997 and ENAHO 1997-most of the conclusions can be extrapolated to ENNN 2000 and ENAHO 2000. Size and composition of sample Although sample size varies significantly across both surveys, sample composition appears to be extremely similar (Table A.1.5). Consequently, this i s not likely to explain differences in poverty incidence between the ENAHOand the ENNIV. Nacional 100.0 100.0 Urban 64.6 64.9 Rural 35.4 35.1 28.5 28.5 17.7 17.8 5.9 5.1 12.8 12.9 `21.9 23.0 5.5 5.7 Rural Selva 7.5 6.8 Totaland extremepoverty lines As we explained above constructing a poverty line involves the following steps: (i) determination of the caloric norm (or minimum caloric intake), (ii) determination of the composition and cost of the basic food-consumption basket (extreme poverty line), and (iii) determination of the cost of the basic consumption basket (total poverty line). There are important differences between the ENNIV and the ENAHOregarding allthree points. Both surveys based their calculation of per capita caloric norms on a representative five-member household, with two adults and three children. However, the ENAHO used a single per capita caloric norm for 1997-2000, and a gender-age-specific per capita caloric norm based on from 2001 onwards, while the ENNIV usedregion-specific per capita caloric norms in 1997 and 2000. This led to significant differences inthe level of minimumper capita caloric intake stipulated ineach survey, with higher intakes contemplated inthe ENNTVthan inthe ENAH066. Inaddition, the ENAHOusesinformation on actual food consumption patterns fromthe 1993-4Encuesta de Propositos Multiples (INEI) to determine the composition of the basic food-consumption basket that provides the minimumper capita caloric intake, whereas the ENNIV uses,a "normative" basket whose compositionreflects the views of nutrition experts rather than realconsumption data. Finally, the ENAHO uses implicit food prices by region, as captured by the survey, to cost the basic food- consumption basket. Incontrast, the ENNNuses various sources to determine food prices: market prices inLima, marketprices inselectedcities for other urbanareas, andprices as captureby the survey for rural areas. 66. For a detailed discussion on the differences in caloric norms between the ENAHO and the ENNIV, see INEI (2001). 160 As aresult, we observe significant differences inthe level of the extremepoverty line. These differences vary across regions, and appear to be largest for the Sierra and the Selva, and smallest for the Costa (Table A.1.6). 1997 2000 ENNIV E N A H O Difference ENNIV ENAHO Difference Soles per capita (monthly) Lima 98.53 117.52 19 101.41 125.91 24 Urban Costa A 86.06 90.12 5 90.55 97.72 8 RuralCosta 76.46 84.88 11 79.33 91.81 16 Urban Sierra 72.20 90.17 25 75.38 97.37 29 RuralSierra 65.96 82.39 25 65.04 90.15 39 Urban Selva 76.88 93.98 22 78.8 102.81 30 RuralSelva 73.35 84.16 15 70.48 91.96 30 Householdexpenditure The expenditure module in ENAHO questionnaire exhibits a higher level of disaggregation than the ENNIV one-the consumption aggregate used by ENAHO contains 58 food items and 113 non-food items, comparedto 31 and55 respectively inthe ENNIV-and this inturn translatesinto a more accurate descriptionof the level andcomposition of expenditureinthe former thanthe latter. Moreover, donations received from institutions or other households are taken into account by the ENAHO, but not by the ENNlV. All inall, this results inhigher estimatedexpenditurelevelswhen we usethe ENAHO. Infact, in 1997average percapitaexpendituremeasuredby theENAHOwas equal to 286.56 soles, significantly abovethe ENNIV's 218.91 soles. Methodological differences across both surveys generate two opposing forces. The higher poverty lines used by ENAHO would tend to produce relatively higher poverty estimates, while higher income level would tend to relatively lower poverty, other things being equal. In order to disentangle the effect of these two forces, Herrera (INEI,2001) performs a series of counterfactualcalculations usingthe ENNIV andENAHOpoverty lines and levels of aggregationof expenditures(Table A.1.7). ENNIV povertyline EHAHOpoverty line ENNIVaggregation ENAHO aggregation ENAHO aggregation ENNIV aggregation ENAHO 1997 ENAHO 1997 ENAHO 1997 ENNIV 1997 National 44.5 36.6 42.7 50.7 Urban 33.1 25.5 29.7 43.0 Rural 65.0 56.6 66.3 64.8 161 ANNEX 2: GROWTHINCIDENCECURVES-ROBUSTNESS CHECKS A short theoretical introduction6' Growth incidence curves (GIC) illustrate the distribution of growth, showing growth rates by quantiles rankedby income, usefulto understandpoverty and inequality trends. Let y denote the income of the p-thquantile at date t. The income for the p-thquantile can be obtained by inverting the cumulative distribution function (CDF) of income of that quantile, denotedby F(y): The growth rate inincome of the p-thquantile canbeexpressedas: Lettingp vary from zero to one, gt (p) tracesout the growth incidence curve. Because F,l (p)= L'(p)h, where L' Xp) i s the Lorenz curve, with slope L' t(p) and mean pt (Gastwirth, 1971), the GIC also can be defined as: Where y, =(p,lp,-,)-1 is the growth rate in,qwhich clearly indicates that if the Lorenz curve does not change, then g,(p)= yr for all p. Inaddition, g,(p)> y, if and only ify ,(p)/p,increasesover time. When the GIC i s positive at all percentile points (g (p) > 0) for all p) there i s fEst-order dominance (FOD) of the distributionat date t over t -1; that is, there i s anunambiguous reductionof poverty between period t -1 and t. Ifthe GIC switches sign, though, it i s not possible to generally infer that a higher-order dominance holds by simply observing the GIC. An upward shift of the GIC implies greater levels of poverty reduction. Finally, a gr (p) decreasing (increasing) function for all p indicates falling (rising) inequality over time. Robustnesschecks As was mentioned in Chapter 2 and as explained in Annex 2, the Encuesta Nacional de Hogares (ENAHO) has undergone significant methodological changes during 1997-2003, the most important of which are (i) the adoption of a new sampling framework in 2001 and, (ii) adoption of a new (spatial) the price deflator in2001. Inorder to minimize the impacts of these changes on our growth incidence calculations, we presented GIC for 1997-2000 and 2001-2003 separately in Chapter 1. In this annex we present some additional robustness checks intended to explore to what extent our results are dependenton some of these changes. Our checks use data from 2001 and 2002 since these are the years for which both sampling frameworks can be used. 67. Basedon Ravallionand Chen (2003). 162 Firstit is importantto examine how meanexpenditure has changedover 1997-2003 usingthe oldandnew samplingframeworks. A comparison of data for 2001 and 2002 (the period in which both frameworks can be made to overlap) reveals a significant difference in mean expenditure levels and small difference inmeanexpenditure trends. This seconddifference beingthe more relevant since GIC depicture changes inexpenditure. 1997 1998 1999 2000 2001 2002 2003 Old samplingframework 1.53 1.54 1.46 1.34 1.34 1.42 New samplingframework 1.24 1.29 1.23 We then construct GIC for 2001-2002 under the old and new framework. The picture we obtained are rather different. Under the old sampling framework, only household in the top 40 percent of the distribution experienced positive (expenditure) growth during 2001-2002. In contrast under the new samplingframework (which is nationally representative), household at the bottomof the distribution also experience positive growth. Given that the difference between both frameworks consists mainly in the inclusion of newly developed, mostly poor areas inthe sample, these results suggest that growth in 2001- 2002 was positive inthese areas. FigureA.2.1: GIC under the oidand newsamplingframeworks. GIC-(Old) Samplingframework basedon 1993PopulationCensus 2001-2002 35.00% 30.00% 25.00% 20.00% 15.00% 10.00% 5.00% 0.00% -5.00% -10.00% -15.00% GIC-(New) Samplingframework basedon2000Pre-PopulationCensus 163 Spell 2001-2002 20.00% 15.00% 10.00% 5.00% 0.00% -5.00% Source: Authors' calculations usingENAHO2001-2002. These results also explain why, if we construct a GIC for 1997-2000 using the old sampling framework we obtain a rather bleak picture interms of expenditure growth. FigureA.2.2: Growth incidencein1997-2000under the old samplingframework . 1997-2002 30.00% 25.00% 20.00% 15.00% 10.00% 5.00% 0.00% -5.00% -10.00% -15.00% -20.00% Source: Authors' calculations usingENAHO 1997-2002. 164 ANNEX 3: INTERNATIONAL COMPARISONS Labor Force Participation Rate LaborForce Participation Rate All Female 1990 .. ~- 2003 1990 ___ 2003 .. Bolivia 72.1 72.8 36.9 38.1 ____I_ Colombia 67.1 70.5 35.9 39.3 Ecuador 61.8 64.7 24.8 29.0 Peru 58.7 62.3 27.5 32.2 Venezuela 64.1 67.0 31.3 35.6 Table A.3.2 Employmentby Sector i Services ~ ~~ All -Male Manufacturing ~ Female All Male gemale All Male Female Bolivia 1990 11 Agriculture 1.2 1.9 0.2 25.1 34.6 11.5 72.8 62.8 87.3 2000 4.9 6.1 3.3 28.2 39.2 14.2 66.8 54.6 82.3 Colombia 1990 1.4 1.9 0.6 30.9 34.7 25.0 67.7 63.4 74.3 2001 22.2 32.5 6.6 18.4 19.2 17.3 59.4 48.3 76.2 Ecuador 1990 6.9 10.3 2.5 26.8 30.0 16.5 66.3 59.7 81.0 2001 7.7 10.2 3.9 24.3 29.6 16.2 67.5 59.8 79.0 Peru 1990 1.2 1.7 0.6 27.3 32.2 19.7 71.5 66.1 79.7 2001 8.8 10.9 6.1 17.9 24.1 10.2 73.3 64.9 83.8 Venezuela 1990 13.4 18.5 2.2 25.3 29.6 15.5 61.2 51.7 82.2 2001 9.6 14.5 1.7 22.1 28.3 12.0 68.2 57.1 86.1 Source: World Deve opment Indicators, The World Bank (2004). Table A.3.3 PovertyLevelinthe Andean Region - 1!1994 1 1995 1996 1997 1998 1999 2000 2001 2002 Change(%) A Bolivia j i 64.7 a 65.4" 63.8a -1.4% Colombia I 4gb 55 14.6% Ecuador 1 40.6' 1i 44.1' 1 8.5% Peru i 42.6= 47.4a 54.ga 54.2" 27.2% Venezuela i 47.0" 44.8a 50.0" 6.4% ~ LAC 165 Table A.3.4 -Extreme Poverty Level inthe Andean Region 1994 1995 ! 1996 2001 i Bolivia I 45.3 42.0 23.43% ~ Colombia 48 14.6% Ecuador 40.6 8.5% Peru 39.5 Venezuela 32.5 31.2 4.68% LAC Note: A Betweenfirst and last data points inTable. Source: SEDLAC: USD2, exceptfor b from Colombia PA 2002 1975- 1980 1985 1990 1995 2000 2002 Bolivia 0.512 0.548 0.58 0.603 0.635 0.67 0.681 Colombia 0.661 0.689 0.706 0.727 0.75 1 0.771 0.773 Ecuador 0.63 0.674 0.696 0.71 0.719 0.735 Peru 0.642 0.672 0.696 0.706 0.733 0.752 Venezuela 0.716 0.73 0.739 0.759 0.768 0.776 0.778 Table A.3.1. GDPand Inflationinthe Andean Region ,440 Source: World DevelopmentIndicators,The World Bank (2004). 166 Targets proposedfor the first Position Progressachieved inPeruinthe 1990s seven MDGs of LAC 1. Reducethe proportion of In1998, thepercentageofLAC The incidenceof extreme poverty inPeruhas peoplelivinginextreme population livingon less thanUS$ 1increasedfrom23.0 percentin 1991to 23.9 percent poverty and suffering from aday PPP was 12and less than in2002, thoughindicators at thenationallevelmask hungerby half between 1990 US$2 PPPa day was 32 (WB important differences acrossregions, and ruraland and 2015. estimates usingPPP 1993 and urbanareas. With current efforts, it is unlikely that householdsurveysdata). the goal to reach 11.5 percentbe achieved. Child malnutrition (under 5) decreasedfrom 10.8 percent in 1992to 7percentin2000. 2. Achieve universal Net enrollment ratio inprimary The net enrollment rate for basic education completion of primary educationinLAC was 97 percentin decreasedfrom 90.6 percent in 1991to 89.5 percent educationby 2015 for boys and 1998 and insecondary education, in2002; at which rateof progress,Perucanmeet the girls and 75 percentcompletion 60 per cent (WDI 2002). 100percentprimary educationcompletion rateby of secondary education. 2015. The proportion of students who reach5h grade to 84.1 percentin2002 from 75.1 percentin 1991. The country could also reachthe 100percent literacy level, which as of 2002, was 96.64 percent. 3. Eliminate gender disparities The ratio of girls to boys attending Though aggregatefigures indicate that this goal was inprimary andsecondary primary and secondary school in achievedby 2002 for primary and tertiary education educationby 2005, andfor all LAC in 1998was 99:lOO (WDI an analysisof desegregateddata indicatesthat levelsby 2015. 2002). gender disparity inrural and extremepoverty areas is actually rising. The girl-boy in secondary school actually decreasedfrom 94.5 percentin 1991to 90.28 percentin2002. Peruvian womenearn23 percentless than men with the same levelof leducationand experience. 4. Reduce infant andchild ]Theunder 5 mortalityrate per 1,000 /Under5 mortality was reducedfrom 92 per 1,000 mortality rates by two-thirds live birthsinthe region was live birthsin 1992to 60 per 1,000 livebkhsin between 1990and 2015. estimatedat 37 for 2000 (WDI 2000, while infant mortality declined from 64 per 2002). 1,000 to 43 per 1,000. Peru is ontrack to reachthis goal. In1995, therewere 188maternal The maternalmortality rate decreasedfrom 265 per deathsper 100,000 live births in 100,OOO live births in 1996to 185per 100,OOO live LAC (estimatesdependon method birthsin2000 (ENDES 2000). The levelof used; UNDP). deliveries that received institutional attentionrose from52.5 to 59.3 percent. The target of 66 per 100,OOO livebirthsby 2015 is unlikely to be achievedwith current efforts. 6. Halt and reverseby 2015 the In2003, the prevalencerateof The averagerateof HIVlAIDSinfectionremained spreadof HIVIAIDS, malaria, HIVIAIDS amongadults aged 15- stable around0.25 percentbetween1996and 2000. andother diseases. 49 inLA was 0.5-0.7 percent, while Perucontinuesto endurehigh incidenceof malaria inthe Caribbeanwas 1.9-3.1 and tuberculosis. percent(WHO 2003). 7. Ensureenvironmental InLAC, in2000, approximately 85 The urbanpopulation withaccessto safe water sustainability - One indicator percentof the population hadaccess dropped from 88 percent to 87 percent, while inmi amongothers is access to an to water (WDI2002). areas increasedfrom 42 percentto 62 percent improved water source. between 1990and2000. Access to sanitation increasedfrom 77 to 79 percent(Human developmentreport). 167 ANNEX 5: THE MININGSECTORAND POVERTYREDUCTION Contemporary mining in Peru has a controversial and ambivalent image among local stakeholders. Mining has developed in poor and sometimes extremely poor rural areas, characterized by economic stagnation, lack of employment opportunities and weak, underdeveloped social capital. However, the presence of profitable mining activities has not always translated into significant improvements in local living standards-Le. high economic growth in departments with large mining investments has not necessarily led to a decline inpoverty (Figure A.5.1). As a result it has often beenarguedthat there is no real connection between the minerals extracted and the wealth generated in a specific area, which, over the years, has strengthenedthe debatable argument that miningdoes not generate wealth nor improve the quality of livingof the affected communities. Figure ASS: Higheconomicgrowthinareaswhere miningis strongdidnot alwaystranslated intolower poverty in2001-2002 I I I Changc inoutput(%) Note: The choice of Moquegua, Tacna, Puno, Arequipa, Cajamarcaand Ancash respondsto the fact that during 1996-2003 approximately 80 percent of the resources distributedby the canon miner0 were directed to these departments Source: Authors' calculationsusingdatafrom ENAHO 2001-2002(INEI) andCuinto (2004). Inthis annex we comment onfour basic aspectsof theinteractionbetweenthe miningsector andthe local community that are likely to have an impact on local living standards and poverty: the capacity of the miningsector to generate employment, the extent to which miningrevenuesrevert to local communities, and the role of miningcompanies and social investors. The redistributiveeffect of the Canon. The creation of the Canon hasresulted into significant increases in the amount of resources transferred to regional and local governments by the Central Government, potentially mitigating some of the problems outlined above. These positive effects have been reinforced by the recent increaseinthe price of metals and minerals in international markets, as well as by changes inthe Canon law that increasethe share ofregionalandlocalgovemment inCanonrevenues. Miningactivities andemployment generation. Most of the formal mininginPerutakes place inremote and economically depressedareas, where employment rates are low and education i s of poor quality and where most of the population is indigenous. These factors explain the high and somewhat disproportionate expectations that local people have regarding large- and medium-size mining 168 investments, which are expected not only to provide jobs, but also to improve access to basic public services. The miningindustry has fulfilled some of these expectations through their community programs, but one cannot expect that miningby itself can solve the issue of local unemployment, since generally, after the development phase, miningrequires only a small number of specialized employees. This is more evident since most potential local workers lack specific training and education to prepare them to become qualified miningworkers. Inaddition to the limited creation of local jobs and the provision of some benefits, mininghas the potential to serve as a catalyst for the creation of local services associated with the mine. This is where the potential for future sustainability resides and where donor and government efforts should focus in terms of channeling the benefits of miningto support capacity buildingand capital investment at the local and regional levels inorder to create businesspartnerships between the mine and the local people. Finally there are some undesirable side effects of mining on local economies that have had limited exposure to a cash economy and that should receive more attention (local inflation of prices, potential increases in criminality and immigration). Some of these problems can be prevented through adequate local employment policies, campaigns to educate the workers' families on the use of cash for improving their livelihoods and the creation of housing saving programs. Distribution of mining revenues. It is clear that mining companies alone cannot and should not be expected to solve the complex issue of the sustainability of the services and goods that they provide to local communities as a matter of compensation. Government intervention and commitment is crucial to any attempt at sustainability, but because resources are always limited, these more remote regions have historically received very little support. However, this trend has started to change due to the derecho de vigencia and the canon minero and to other legal requirements that channel some investment to the municipal, district andprovincial levels. Revenuesfromthe canon represent 27 percent of all transfers to local government. Moreover the volume of resources distributedthrough the canon have increased from 0.1 to 0.11percent of GDPbetween 1996 to 2003. Inadditionrecentchangesinthecriteriausedfor the canondistributionwill linkitdirectly tothe lackof basic necessitiesandthe infrastructure deficit. This changerespondsto the complaints about the fact that, under the previous allocation rule based on population density, some urbanhigh-income municipalities- usually not impacted by mining- received a highshare of the canon. Social investment by mining companies. Not all the investment received by communities comes through local governments and transfers. Internationalmines tend to invest more incommunities than the national entities, and make a greater effort for the sustainable development of the community than in the past. This may be the result of having greater resources, better capacity for community relations and local development, and of being more carefully scrutinized by public interest groups in the international community. National large and medium scale mines may also lack some of the capacity shown by the internationalindustry leaders. In comparison with other sectors, it is estimated that mining companies have a higher rate of social investment: more than half of the mining companies participate in these kinds of investments. The majority have supported rehabilitation of roads (93.3 percent), support to local activities (70 percent), support to local sports (66.7 percent), provision of electricity (63.3 percent), donation of books (60 percent), among others. Yet, ingeneral these benefits tend to be punctual economic compensations based on informal agreements and do not contribute to development frameworks whereby communities are 169 empowered through a process of capacity buildingand understanding of their own priorities. This fairly typical approach on the part of the mining companies does not help in developing a more integral relationship among the stakeholders. Insummoreneedstobedonebeyondthe issueoffinancial resources, tohaveamorebalancedshareof benefits among local stakeholders, lower expectations in communities and address the issue of social sustainability. Recommendations. Before any first contact i s made between a mining company and the local people, the government should lead a formal communication process with affected communities aimed at educating the local residents about the basic economics of mining-the limitedjob intake capacity; the potential for benefiting the community through an agreed process of compensations that would occur within a process of local development with the active participation of the affected population; the potentialfor the creation of local servicesfor the mine; and the distribution of royalties and canon. Channeling of resources to communities through the derecho de vigencia, the canon minero and the recently approved royalties should continue. However, existing concerns that these revenues may end up financing the salaries of the local bureaucraciesinstead of fundingthe development priorities of the local communities should be addressed. To help prevent this, the government should establish fiscal mechanisms to oversee the use of the canon and the new royalties to help ensure that these funds are directed to supporting the community plans for development as established in the law of the canon minero. In this regard it is also worth considering the law of the canon minero be amended to allow resourcesto be usedfor projects other than infrastructure-e.g. the canon and the royalties could be used to finance local capacity buildingfor adequate managementof local business and long-term sustainability of the benefits of mining. Finally, mining operators should, to the extent possible, avoid compensating affected communities with direct cash payments and instead should seek to support the development priorities of the local communities. Formal benefit agreements should have a participatory monitoring mechanism in place to ensure compliance and timely interventions to improve its implementation. From the first stages of miningoperation andthroughout the entire life cycle of the mine, bothshort term physical infrastructure works that will benefit local communities, and long-term local development (LED)projects such as local capacity building should be implemented to promote sustainable development. These types of LED programs are mutually beneficial because they will encourage mines to use and purchase the goods and services of their area of influence for their operations and personnel needs, which will in turn increase economic development and employment opportunities for the local community per se. Inthis regard, the Decreto Supremo 042-2003-EMon "Previous Commitment" has set forth the guiding principles for buying local goods and services and hiring local workers, but in order to make these principles plausible, capacity buildingwhich enables local enterprises to enter into supply contracts with mining companies and supply quality goods as well as training programs to provide local workers the necessaryskills to be employed by the miningcompany must be developed. 170 STATISTICAL ANNEX CHAPTER1 Table SA.1: Statisticsfor simulatedper capita incomeby regions Real 2002 g=l% g=3% g=5% g=8% g=10% Inequality Gini 55.67 55.67 55.67 55.67 55.67 55.67 Poverty (Lima) FGT(0) 48.2 41.5 27.7 17.8 8.45 5.33 FGT(1) 19.3 15.7 10.3 6.67 3.77 2.98 FGT(2) 10.7 8.67 5.78 4.02 2.75 2.37 Exremepoverty (Lima) FGT(0) 14.3 10.9 6.31 4.34 3.29 2.49 FGT(1) 5.31 4.35 3.22 2.67 2.19 1.98 FGT(2) 3.41 2.99 2.49 2.19 1.93 1.83 Poverty (Costa) FGT(0) 57.7 49.5 36.1 24.2 11.7 7.65 FGT(1) 24.1 20.1 13.4 8.62 4.47 3.12 FGT(2) 13.6 11 7.18 4.68 2.66 1.97 Exremepoverty (Costa) FGT(0) 24.5 19.7 11.9 7.41 4.09 2.93 FGT(1) 8.84 7.02 4.5 3.08 1.98 1.58 FGT(2) 4.83 3.91 2.69 1.97 1.38 1.13 Poverty (Sierra) FGT(0) 76.7 72.4 63.3 53.3 39.6 31.1 FGT(1) 45.5 41.5 33.8 26.8 17.9 13.2 FGT(2) 32 28.5 22.2 16.8 10.5 7.34 Exremepoverty (Sierra) FGT(0) 58.8 53.6 44.6 35.6 24.5 17.7 FGT(1) 31.1 27.6 21.4 16 9.6 6.5 FGT(2) 20.4 17.7 13 9.25 5.16 3.37 Poverty (Selva) FGT(0) 76 71.4 60.8 47.5 31.7 23.7 FGT(1) 41.7 37.3 28.7 21.5 13.3 9.36 FGT(2) 27.5 23.9 17.7 12.8 7.52 5.15 Exremepoverty (Selva) FGT(0) 57.4 50 38.2 28.7 17.8 11.8 FGT(1) 26.6 22.8 16.6 11.8 6.69 4.5 1 FGT(2) 16.2 13.7 9.63 6.62 3.65 2.44 171 Table SA.2: Statistics for simulatedper capita income Diffel it redistributive policies by regions Real2002 t= 10% t=20% t=30% Inequality Gini 55.67 51.30 46.5 1 41.51 Poverty (Lima) FGT(0) 48.24 49.33 50.84 52.37 FGT(1) 19.28 18.34 17.44 16.57 FGT(2) 10.69 9.38 8.17 7.07 Exremepoverty (Lima) FGT(0) 14.31 11.57 9.33 6.35 FGT(1) 5.31 3.87 2.59 1.59 FGT(2) 3.41 2.15 1.22 0.59 Poverty (Costa) FGT(0) 57.69 56.83 55.75 54.58 FGT(1) 24.12 21.10 18.11 15.14 FGT(2) 13.57 10.51 7.86 5.61 Exreme poverty (Costa) FGT(0) 24.45 20.10 13.75 7.80 FGT(1) 8.84 5.58 2.94 1.14 FGT(2) 4.83 2.44 0.98 0.26 Poverty (Sierra) FGT(0) 76.65 76.22 75.51 74.66 FGT(1) 45.54 39.42 33.32 27.26 FGT(2) 31.95 24.13 17.43 11.87 Exreme poverty (Sierra) FGT(0) 58.76 54.42 48.73 39.30 FGT(1) 31.11 22.61 14.66 7.57 FGT(2) 20.37 11.69 5.59 1.91 Poverty (Selva) FGT(0) 76.00 75.66 74.63 73.02 FGT(1) 41.72 35.83 29.99 24.23 FGT(2) 27.49 20.54 14.71 9.97 Exreme poverty (Selva) FGT(0) 57.39 51.60 44.99 33.36 FGT(1) 26.58 18.87 11.85 6.06 FGT(2) 16.22 9.09 4.28 1.55 172 I Table SA.3: Poverty, extreme poverty and inequality by department, 2004 Poverty ExtremePoverty Gini Amazonas 60.9 28.9 0.37 Ancash 55.3 23.4 0.38 Apurimac 65.9 30.7 0.36 Arequipa 40.9 10.7 0.37 Ayacucho 64.9 24.9 0.33 Cajamarca 74.2 36.9 0.37 cusco 59.2 25.9 0.40 Huancavelica 84.4 59.9 0.38 Huanuco 77.6 46.9 0.39 Ica 29.2 2.4 0.33 Junin 52.6 18.3 0.34 LaLibertad 48.2 22.5 0.41 Lambayeque 46.7 12.5 0.35 Lima 37.2 4.4 0.42 Loreto 62.7 32.0 0.35 Madre de Dios 20.4 4.5 0.30 Moquegua 37.2 10.5 0.32 Pasco 61.6 27.3 0.34 Piura 60.9 20.8 0.35 Pun0 79.2 49.8 0.38 San Martin 57.1 24.0 0.35 Tacna 26.7 5.2 0.38 Tumbes 21.6 1.1 0.31 Ucayali 55.8 30.2 0.40 173 Table SA. 4: Probabilityof being oor (Probitestimates) Total Lima Urban Rural Marginaleffects Characteristicsof the Head Demographic Age 25-55 years old -0.026 0.059 -0.038 -0.063** (0,032) (0.066) (0.048) (0.032) Age morethan55 years old -0.230*** -0.145*** -0.209*** -0.238*** (0.031) (0.055) (0.042) (0.039) Female 0.044* -0.019 0.030 0.068*** (0.023) (0.048) (0.034) (0.025) MaritalStatus Cohabiting 0.181*** 0.264*** 0.123** 0.129*** (0.033) (0.085) (0.061) (0.032) Married 0.092""" 0.173*** 0.005 0.082** (0.033) (0.065) (0.059) (0.036) Widow I divorced 0.048 0.180* 0.003 0.016 (0.032) (0.096) (0.053) (0.035) Education Primary -0.160*** -0.089 -0.075** -0.147*** (0.021) (0.080) (0.037) (0.021) Secondary -0.272*** -0.195** -0.171*** -0.281*** (0.023) (0.093) (0.038) (0.030) University -0.401*** -0.282*** -0.307*** -0.406*** (0.020) (0.067) (0.031) (0.048) Employment Employer -0.168*** -0.122*** -0.168*** -0.167** (0.029) (0.033) (0.03 1) (0.068) Self-Employed 0.064** 0.021 0.067* 0.031 (0.030) (0.045) (0.035) (0.059) Worker "Obrero" 0.086*** 0.099** 0.076** 0.019 (0.029) (0.046) (0.035) (0.053) Other (Family I domestic worker) -0.034 -0.009 0.007 -0.150 (0.046) (0.071) (0.052) (0.099) InformalSector' 0.178*** 0.133*** 0.128*** 0.246*** (0.027) (0.047) (0.034) (0.055) Industry Public Administration -0.068" -0.050 -0.069 -0.033 (0.037) (0.076) (0.043) (0.062) Construction -0.007 -0.022 0.053 -0.09 (0.030) (0.070) (0.038) (0.066) Industry -0.053** -0.008 -0.064** -0.071 (0.025) (0.071) (0.027) (0.049) ServicesI utilities -0.163*** -0.107 -0.148*** -0.170*** (0.018) (0.074) (0.023) (0.03 1) Householdcharacteristics Sizeof the Household 0.090*** 0.069*** 0.092*** 0.072*** (0.005) (0.008) (0.007) (0.006) 174 % members younger than9 or older than 60 0.352*** 0.357*** 0.341*** 0.251*** (0.026) (0.063) (0.040) (0.028) At leastone migrant inthe household -0.162*** -0.046 -0.137*** -0.196*** (0.025) (0.076) (0.029) (0.034) Income earners over total adults 10years or older -0.349*** -0.409*** -0.322*** -0.204*** (0.026) (0.052) (0.037) (0.031) Employed ininformal sector over total adults 10years or older 0.009 0.144 -0.008 -0.095* (0.045) (0.107) (0.069) (0.052) Infrastructure Water -0.057*** -0.080 -0.028 -0.066*** (0.014) (0.056) (0.022) (0.016) Electricity -0.139*** 0.029 -0.207*** -0.125*** (0.016) (0.063) (0.034) (0.016) Sanitary Services -0.195*** -0.138** -0.199*** -0.102*** (0.017) (0.059) (0.021) (0.036) Ownership Rent -0.019 0.077 -0.053 -0.148** (0.031) (0.053) (0.033) (0.064) Owner of the house(with title) -0.018 0.031 -0.066*** -0.018 (0.017) (0.030) (0.022) (0.022) Owner of the house(without title) 0.107*** 0.192** 0.029 -0.032 (0.035) (0.081) (0.039) (0.060) Regions Urban 0.146*** (0.018) costa -0.092*** (0.020) Sierra 0.120*** 0.124*** 0.247*** (0.020) (0.020) (0.022) Selva -0.041* 0.147*** 0.008 (0.022) (0.025) (0.023) Observations 16,117 1,516 7,203 7,398 PseudoR-2 0.35 Note: Robuststandarderrors inparentheses. significantat 10%; * *: gnificant at 5%;***0.34 0.33 0.27 significant at 1% Source: Authors' calculationsusingdatafrom ENAHO2003 (INEI). 175 Table SA.5: Probitby: gion 2003 2003-2004 costa Sierra Selva Lima Characteristicsof the Head Age 25-55 years old -0.018 -0.083** -0.019 0.059 (0.056) (0.039) (0.057) (0.066) Age morethan 55 years old -0.164*** -0.258*** -0.267*** -0.145*** (0.052) (0.047) (0.062) (0.055) Female 0.095** 0.039 0.036 -0.019 (0.044) (0.027) (0.055) (0.048) MaritalStatus Cohabiting 0.285*** 0.087** 0.087 0.264*** (0.087) (0.037) (0.068) (0.085) Married 0.149* 0.001 0.067 0.173*** (0.080) (0.039) (0.070) (0.065) Widow I divorced 0.065 -0.029 0.068 0.180" (0.083) (0.038) (0.071) (0.096) Education Primary -0.093** -0.153*** -0.164*** -0.089 (0.042) (0.025) (0.050) (0.080) Secondary -0.165*** -0.302*** -0.266*** -0.195** (0.043) (0.032) (0.053) (0.093) University -0.276*** -0.457*** -0.422*** -0.282*** (0.033) (0.036) (0.050) (0.067) Employment Employer -0.157*** -0.192*** -0.064 -0.122*** (0.041) (0.051) (0.073) (0.033) Self-Employed 0.068 0.048 0.099 0.021 (0.048) (0.044) (0.067) (0.045) Worker "Obrero" 0.102** 0.021 0.087 0.099** (0.049) (0.039) (0.063) (0.046) Other (Family I domesticworker) 0.016 -0.121 0.074 -0.009 (0.070) (0.078) (0.103) (0.071) Informal Sector 0.152*** 0.154*** 0.130* 0.133*** (0.040) (0.046) (0.069) (0.047) Industry Public Administration 0.017 -0.109** -0.198** -0.050 (0.069) (0.052) (0.083) (0.076) Construction 0.112** -0.047 -0.048 -0.022 (0.049) (0.049) (0.090) (0.070) Industry -0.079** (0.049) (0.069) (0.008) (0.037) (0.037) (0.059) (0.071) ServicesI utilities -0.085*** -0.190*** -0.260*** -0.107 (0.031) (0.026) (0.035) (0.074) 176 Characteristics of the Household Size of the Household 0.075*** 0.092*** 0.093*** 0.069*** (0.007) (0.007) (0.009) (0.008) % Members younger than 9 or older than 60 0.243*** 0.282*** 0.447*** 0.357*** (0.050) (0.031) (0.057) (0.063) At least one migrant inthe household -0.108** -0.177*** -0.269*** -0.046 (0.045) (0.035) (0.049) (0.076) Income earners over total adults 10years or older -0.325*** -0.213""" -0.319*** -0.409*** (0.046) (0.035) (0.055) (0.052) Employed ininformal sector over total adults 10 years or older -0.01 -0.049 -0.099 0.144 (0.090) (0.057) (0.102) (0.107) Infrastructure Water -0.042 -0.071*** -0.032 -0.08 (0.026) (0.018) (0.030) (0.056) Electricity -0.191*** -0.122*** -0.146"'" 0.029 (0.032) (0.019) (0.035) (0.063) Sanitary Services -0.176*** -0.197*** -0.140*** -0.138** (0.028) (0.026) (0.036) (0.059) Ownership Rent -0.019 -0.063 -0.120"" 0.077 (0.063) (0.041) (0.061) (0.053) Owner of the house (with title) (0.040) (0.029) -0.088** 0.031 (0.028) (0.023) (0.037) (0.030) Owner o f the house (without title) 0.037 0.038 -0.128 0.192"" (0.039) (0.096) (0.089) (0.081) Regions Urban 0.149*** 0.051** 0.339*** (0.028) (0.023) (0.035) Observations 4,325 7,027 3,249 1,516 PseudoR-2 . 0.30 0.33 0.31 0.34 Note: Robuststandarderrors inparentheses.* significantat 10%;** significant 5%; *** significantat at 1% Source: Author's calculationsusingdatafromENAHO2003. 177 oq cq -! N N Q 3 P P 1 TableSA.8 Ruralpovertyprofile includingethnicityvariable - (marginaleffectsof a probit model) (1) (3) (4) Rural (2)Rural Rural Rural coast highlands jungle Gender (1=male) -0.02 -0.00 -0.01 -0.06 (0.02) (0.14) (0.02) (0.06) Age of household's head 0.00087 0.00240 0.00076 0.00086 (0.00067) (0.00483) (0.00059) (0.00198) Maximumyears of educationachievedbya householdmember -0.02 0.01 -0.01 -0.04 (O.OO)*** (0.02) (O.OO)*** (0.01)*** Numberof householdmembers 0.06 0.13 0.05 0.07 (0.01)*** (0.04)**' (0.0l)*** (0.02)*** Ratioof membersyounger than 14years 0.27 0.71 0.21 0.33 (0.05)*** (0.29)" (0.04)"' (0.14)** Ratioof membersolder than 65 years -0.03 0.35 -0.05 -0.07 (0.03) (0.30) (0.03) (0.14) Ethnicgroup(native language) 0.08 -0.08 0.06 -0.00 (0.02)'- (0.15) (0.02)*** (0.06) Numberof rooms -0.03 -0.05 -0.03 -0.03 (0.01)*** (0.03)* (O.Ol)*** (0.02) Earthfloor -0.15 -0.21 -0.11 -0.11 (0.03)"* (0.16) (0.04)"' (0.05)** Precariousceiling 0.04 0.28 0.00 0.05 (0.02)" (0.09)'*' (0.01) (0.11) Drinkablewater insidethe dwelling 0.04 0.02 0.17 (0.06) (0.06) (0.14) Dwellingwith electricity 0.02 0.29 -0.02 0.16 (0.02) (0.08)"' (0.02) (0.04)"* Receivedor sent transfers -0.07 -0.05 -0.05 -0.17 (0.02)*" (0.10) (0.01)"* (0.05)"' Durableassets -0.12 -0.23 -0.09 -0.18 (O.Ol)*" (0.07)*** (0.01)"' (0.04)*** Trasportationassets -0.00 0.01 -0.00 -0.00 (0.00) (0.01) (0.00) (0.01) Livestock 0.03 -0.19 0.02 -0.09 ~ (O.Ol)*** (0.11)' (0.01)" (0.04)** /Numberof observations 4779 252 3647 872 Like1ihood (log): -1811.19 -105.73 -1227.42 -385.00 LR chi2: 16.00 15.00 16.00 16.00 PseudoR2: 0.28 0.36 0.30 0.28 Standarderrors in parenthesis * significantat 10%; ** significantat 5%;*** significantat 1% Source: Own estimates based on ENAHO(HouseholdSurvey)2001 180 Table SA.9 Profileof Rural Dwellers by IncomeGeneratingSources (Probabilityof choosing specific incomesources Waged Waged Non- Non-waged Non-waged Agricultural Agricultural Agricultural Non- Income Income Income agricultural Income dy/dx dy/dx dy/dx dy/dx Age of household head -0.006 -0.015 0.005 -0.013 (0.001) *** (0,001) *** (0.001) *** (0.001) *** Age of household headsquared 0.000 0.000 0.000 0.000 (0.000) '** (0.000) e*, (0.000) f., (0.000) *** Householdhead male 0.002 -0.086 0.137 -0.087 (0.020) (0.017) **.* (0.021) **, (0.021) '** Years of educationof householdhead -0.016 0.014 -0.009 0.002 (0.002) *** (0.001) *** (0.001) .** (0.002) Number of rooms inthe house 0.005 0.011 0.006 0.030 (0.005) (0.004) *** (0.003) ** (0.005) *** YOof members between 14and 65 years 0.152 0.087 -0.049 0.086 (0.026) *** (0.019) I** (0.015) *** (0.028) *** Housewith drinking water -0.054 0.033 0.005 0.032 (0.019) *** (0.013) ** (0.011) (0,019) * Housewith pipeline -0.048 0.091 -0.093 0.132 (0.038) (0.033) I,.. (0.029) *./ (0.043) **I Housewith electricity 0.034 0.062 -0.070 0.061 (0.020) (0.015) .*. (0.013) *** (0.020) '** House sent or received remittances -0.022 0.002 0.007 -0.024 (0.014) (0.011) (0.008) (0.015) Populationintown -0.072 -0.143 0.112 0.118 (0.000) ** (0.000) (0.000) *** (0.000) *** Note 0>0.10: **0>0.05: ***0>0.01 Source: Own'esimations basedon ENAHO IV auarter 181 I I I I I I 7 Io -=?=?zg m 0 7 cd m -b- tNc ?d ' N O 2 u, x N N e N W N m O 2m a?$SZ N