WPS4869 P olicy R eseaRch W oRking P aPeR 4869 Firms' Productive Performance and the Investment Climate in Developing Economies An Application to MENA Manufacturing Tidiane Kinda Patrick Plane Marie-Ange Véganzonès-Varoudakis The World Bank Middle East and North Africa Region March 2009 Policy ReseaRch WoRking PaPeR 4869 Abstract Drawing on the World Bank Investment Climate (depending on the industry) for the quality of various Assessment surveys, this paper investigates the infrastructure, the experience and education level of relationship between firm-level technical efficiency and the labor force, the cost of and access to financing, as the investment climate for 22 developing economies and well as different dimensions of the government-business eight manufacturing industries. The authors first propose relation. The analysis reveals that some industries, more three measures of firms' productive performance: labor exposed to international competition, are more sensitive productivity, total factor productivity, and technical to investment climate deficiencies. For some industries, efficiency. They show that, on average, enterprises in this is also true for small and medium domestic the Middle East and North Africa have performed enterprises that do not have the possibility to influence poorly compared with other countries in the sample. their investment climate or choose their location. These The exception is Morocco, whose various measures of findings bear clear policy implications by showing that firm-level productivity rank close to the ones of the increasing firms' size and improving the investment most productive economies. The analysis also reveals climate (in particular of small and medium firms and that the competitiveness of countries in the region has industries more exposed to international competition) been handicapped by high unit labor cost, compared could constitute a powerful means of industrial with main competitors like China and India. The development and competitiveness, in the Middle East empirical results show then? that the investment climate and North Africa region in particular. matters for firms' productive performance. This is true This paper--a product of the Middle East and North Africa Region--is part of a larger effort in the region to use enterprise surveys to identify constraints on productivity and growth in developing countries. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. For information, contact jdethier@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Firms' Productive Performance and the Investment Climate in Developing Economies: An Application to MENA Manufacturing Tidiane KINDA Patrick PLANE Marie-Ange VÉGANZONÈS-VAROUDAKIS CNRS-CERDI, Université d'Auvergne 65, Boulevard François Mitterrand 63000 Clermont-Ferrand, FRANCE Keywords: Manufacturing firms, Productivity, Investment Climate, Developing Countries, MENA. JEL classification: D24, O14, O57. Corresponding author: Marie Ange Véganzonès (veganzones@aol.com) Firms' Productive Performance and the Investment Climate in Developing Economies: An Application to MENA Manufacturing 1- Introduction Recent developments of the economic literature have put the investment climate at the center of economic performance. It is now well documented that the investment climate can significantly affect investment, productivity, and growth 1. A growing literature suggests, in particular, that successful market-based economies need good governance institutions 2. Governance exerts a strong influence on the investment climate. On the empirical side, several studies have related economic performances to different measures of governance 3. The role of security of property rights is one of the best documented and supported by the data 4. Some authors have also tested the role of corruption 5 and, to a lesser extend, regulation 6 and bureaucratic quality 7. More recently, the literature has evaluated firm performance and its determinants using enterprise surveys data 8. This approach, still quite new, aims at strengthening the institutional literature by providing microeconomic foundation. Investment climate is defined by the World Bank as the policy, institutional and regulatory environment in which firms operate (see World Bank, 2005). Key factors affecting the investment climate are corruption, taxation, regulatory framework, quality of bureaucracy, legal environment, availability and quality of infrastructures, availability and cost of finance, factor markets (labor and capital), technological and innovation support. A good investment climate reduces the cost of doing business and leads to higher and more certain returns on investment. The forward- looking nature of investment underlines the importance of a stable and secure environment. A poor investment climate is also seen as constituting barriers to entry, exit and competitions. The World Bank (2004) reports as well that a better investment climate improves bureaucratic performances and predictability, and contributes to the effective delivery of public goods that are necessary for productive business. In MENA, various studies point out the deficiencies of the investment climate. This is the case of the World Bank (2004) for governance, as well as of country studies based on enterprises surveys, in particular the World Bank Investment Climate Assessments (ICA) of Egypt (2005 and 2006), Morocco (2001 and 2005), and Algeria (2002). Doing Business (World Bank, 2009) also places MENA low on business climate indicators 9. These deficiencies have been reported as contributing to the poor economic accomplishment of the region 10. Although MENA countries are, in average, defined as middle income-economies, growth and investment performances in particular have been disappointing for more than three decades 11. Attractiveness to FDI has also been weak, as well as competitiveness and exports of manufacturing 12. MENA competitiveness has constantly been affected by poor exchange rate policies and insufficient economic reforms. But other factors, such as the investment climate, can also explain the low productivity and the high production costs at the firm level. The World Bank Investment Climate (ICA) surveys collect data on inputs and outputs, as well as on various aspects of the investment climate at the firm level. ICA surveys produce both subjective evaluations of obstacles, as well as other more objective information with direct link to cost and productivity on the themes of infrastructure, human capital, technology, governance, and financial constraints. These standardized surveys of large samples of firms from different sectors permit comparative measures of firms' productive performance. They also provide information to estimate the contribution of investment climate to these 2 performances. In a context of increasing pressure of globalization, ICA surveys can be seen as an instrument for identifying key obstacles to firms' productivity and competitiveness. They can be used as a support to policy reforms for an increased economic growth. Drawing on the World Bank firm-surveys, this paper analyzes the relationship between investment climate and firm-level productivity for a large number of developing countries (22 among which 5 MENA economies) and eight manufacturing industries. We first propose different measures of firms' productive performances, such as Labor Productivity (LP), Total Factor Productivity (TFP), and Technical Efficiency (TE) using a production frontier approach. These indicators are compared to each others as well as across countries, in order to position MENA manufacturing amongst a wide range of firms from other regions. We reveal that enterprises in MENA perform in average poorly, compared to other countries of the sample. The exception is Morocco, whose various measures of firms' productive performance always rank close to the one of the most productive firms in the sample. An originality of our approach has also been to generate a few composite indicators of investment climate using Principal Component Analysis (PCA), which summarizes well the key dimensions of the investment climate. This has allowed as well tackling the problem of multicolinearity when explaining firm productive performances with a wide range of correlated IC variables. We define four dimensions of the investment climate: the Quality of Infrastructure (Infra), the Business-Government Relations (Gov), the Human Capacity (H), and the Financing Constraints (Fin). We use, as well, city or region-sector averages to reduce the endogeneity problem underlying the investment climate (IC) variables. The analysis finally shows that investment climate matters for firms' productive performances. This has been done by estimating an efficiency function explaining firm-level productivity for each of our 8 manufacturing industries. The paper is organized as follows. The second section introduces different concepts of firm- level productivity and discusses the advantages and limits of the different measures. Section three presents the investment climate (ICA) surveys and summarizes their main limitations. The fourth section presents and compares across countries our different estimations of firms' productive performances by industry. The fifth section introduces the investment climate indicators used in the empirical analysis, and calculates our four broad IC indicators. The sixth section highlights MENA investment climate deficiencies. In the seventh section, we examine whether the various dimensions of the investment climate constraints firms productive performances. The last section concludes 2- Measures of Firm-Level Productivity: Methodological Aspects Our first challenge has been to measure firms' productive performance in a relevant way. We propose different approaches and measures. We first consider a non parametric model of productivity, which consists in calculating productive performances without estimating a production function. Non parametric measure of productivity constitutes a simple and already meaningful way of assessing for example Productivity of Labor (LP) and Total Factor Productivity (TFP). Another way has been to calculate firms' productive performance from a parametric production frontier. This more sophisticated methodology allows to identify the most efficient firms of the sample and to compare MENA firms' performances to them. 3 2.1- Non Parametric Measures of Productivity and Unit Labor Cost Productivity can easily be calculated as the ratio of an output to a specific factor of production (defined as Productivity of Labor when the factor is labor), or to all relevant factors of production (called Total Factor Productivity, TFP). In this paper, we only refer to productivity levels because of the limited time dimension for the production factors (two to three years at the best) and no time dimension for the IC variables. Our analysis thus focuses on comparisons of firm-level productivity among enterprises, industries and countries 13. In the empirical analysis, we first discuss Labor Productivity (LP), which gives a first idea of the productive performance of the firms. Productivity of Labor has the advantage not to be affected by the error in measuring the capital stock. However, the technology is partially described and calculation of productivity suffers from the omission of this variable. Productivity of Labor can be complemented by calculations of Unit Labor Cost, defined as the ratio of firm average wage to firm's labor productivity. This indicator allows comparisons of the cost of labor across countries in competition in the world markets. Firms' productivity of labor can also be biased by the choice of the exchange rate when converting production into US$. This is less the case of TFP because the same rate applies to the output (Y) at the numerator, and to the intermediate consumption (ICons) and capital stock (K) in the denominator. Non parametric Total Factor Productivity (TFP) constitutes another simple (and also more complete) way to evaluate firms' productive performances. Under the hypothesis of constant returns to scale, (i.e., perfect competition for goods but also for factors that are remunerated at their marginal productivity) weights of Intermediate Consumptions (ICons) and of Labor (Wages, W) are estimated as the ratio of the cost of these factors to the Total Cost of Production including profit (Y). The contribution of Capital (K) is then calculated as the complement to one. The advantage of this approach, based on the Solow residuals, is that it does not require the inputs to be exogenous or the inputs elasticity to be constant. The disadvantage is that two hypotheses have to hold: (a) constant returns to scale; and (b) competitive input markets. Another limitation can be seen in the fact that productivity, being calculated as the residual of the production function, is considered as a random variable, which makes it difficult to justify that some exogenous factors can explain productive differences. The equation is as follows: Yi TFPi (1) L1i IC i2 i K i(11i 2 i ) i Wi IC 1i , 2i i (2) Yi Yi 2.2- Parametric Production Functions and Production Frontiers In the parametric approach, TFP is calculated as the residual of an estimated production function, thus relaxing the hypotheses of constant returns to scale (but not automatically of productivity as a random variable). Various hypotheses can be made regarding the technology of production. The Cobb Douglas and the Translogarithmic production functions are the most commonly used. Although both present good mathematic properties, the elasticities of the production to the inputs are easy to read and to interpret with the Cobb Douglass technology. 4 In the case of a parametric production function, production is derived from the optimization problem of the firms, which in perfect competition maximize current and expected profits by equating production prices to their marginal costs. This hypothesis does not permit any waste of resources or organizational weaknesses. The production frontier approach, however, allows for non optimal behaviors of the firms. Enterprises can be positioned in regard to the most efficient firms that define an empirical production frontier. Firm-level Technical Efficiency (TE) can then be defined as the firms' productivity gap (or efficiency gap) to the "best practice", the empirical practice of the firms which are located on the production frontier. The deterministic parametric production frontier approach can be implemented in a rather simple way, under the restrictive assumption that the production does not suffer from the classical disturbances. The higher positive residual of the regression is used as a correction term, to defining the most efficient observation. The other observations are positioned comparatively to this most efficient observation. Correction is applied to the intercept of the regression for all the observations, except this one. The residual of estimation ( ui ) is a random variable, uncorrelated and independent of the right-hand side variables. ui can be transformed as an indicator of efficiency of value 1 (or 100% when expressed in percent) when ui = 0. For the firms of the sample for which the residual is not zero, ui measures the potential performance gain that these enterprises can achieve. The deterministic parametric production frontier is specified as follows: y i f ( xi , ) u i , u i 0 (3) With - Y: Production - X: Production factors -: Parameters of the equation - ui : Technical Efficiency (TE) - i: Firm index In the stochastic model, the likelihood estimation method is typically applied to estimate a "composite" error term which is split into two uncorrelated elements. The first term (v), which is a random variable, represents the external shocks to the firm. These shocks, independent and identically distributed, follow a normal distribution, with zero average and ² standard deviation. The second term represents the Technical Efficiency (-u). We will suppose that u follows a truncated normal distribution. In this specification, firms' productive performances are not assimilated to a random variable and can then be explained by exogenous factors. The interest of this approach can also be seen in the fact that TEs have a relative form and can be compared across countries and regions. Although there is a wide range of choices as regard the statistical distribution of the efficiency term (u), the ranking of firms according to the efficiency term is generally not sensible to the choice of the specific distribution (Coelli, Prasada Rao and Battese, 1998). Equation is as follows: y i f ( xi , ) u i vi (4) With - Y: Production - X: Production factors 5 -: Parameters of the equation -v: External shocks -u: Technical Efficiency (TE) - i: Firm index 2.3- Explaining Technical Efficiency A complementary approach, when having calculated Technical Efficiency (TE), is to explain the reasons for firms' diverse performances. Firms' inefficiency can be explained by "exogenous" factors which affect either the technology of production, or the firm's ability to transform inputs into outputs. In the literature, these factors have been estimated in two different ways. A simple method consists in estimating the stochastic production frontier, and in regressing the residual of the estimation (the Technical Efficiency, TE) on a vector of explanatory factors (z). This method is called the "Two Steps" procedure. Different estimation procedures can be used. The simplest way is to run an OLS regression. Another possibility is to apply a Tobit model, in order to address the question of the distribution of the efficiency. The "Two Steps" procedure presents, however, some shortcoming in separating the Technical Efficiency (TE) from the production frontier. When some production frontier inputs (x) are explained by factors affecting efficiency, there is an issue of simultaneity14. Because the Technical Efficiency term (TE) is correlated with the production frontier inputs (x), the likelihood estimation of the stochastic production frontier is biased, due to the omission of important explanatory variables. In fact, a relatively new branch of the literature proposes to estimate the production frontier and the factors explaining inefficiency at the same time. This is the "One Step" procedure. In this case, the parameters of the equation (here and ) are simultaneously estimated by the likelihood estimation method. The stochastic version of the model can be written as follows: y i f ( xi , z i , , ) u i vi (5) With - Y: Production - X: Production factors - Z: Factors explaining Technical Efficiency -v: External shocks -u: Technical Efficiency - / : Parameters of the equation - i: Firm index 3- The ICA Firm Surveys: Data Limitations The World Bank Investment Climate (ICA) surveys collect data on inputs and outputs, as well as on a large variety of quantitative and qualitative (perception-based) indicators of the investment climate. In building the database, we have tried to incorporate as much information as possible. We have integrated in our sample 23 countries which participate in the five main regions of the developing world: Sub-Saharan Africa (AFR), East Asia (EAS), South Asia (SAS), Latin America and the Caribbean (LAC), Middle East and North Africa (MENA, see list of countries in Annex 1) 15. In this sample, MENA is represented by 5 countries: Algeria (2002), Saudi Arabia (2005), Lebanon (2006), Morocco (2000, 2004) and 6 Egypt (2004, 2006) 16. Syria (2003) and Oman (2003), which were initially part of the sample, had to be removed because of a very low rate of answer to the questionnaire. By broadening the initial sample to a large number of countries from different regions, we have intended to compare MENA performances to the ones of emerging economies which appears as major competitors in the world market: China and India in particular. To estimate firm-level productivity, we initially considered a population of almost 20,000 firms, coming from 13 manufacturing industries. This initial sample had to be reduced due to various limitations. Calculation of productive performances requires information on at least 5 variables: (1) production, (2) intermediate consumption, (3) labor, (4) wages, and (5) capital stock. For several enterprises, part of this information appeared difficult to get. For others, answers showed flagrant inconsistencies. Enterprises were eliminated in particular when the calculation of productive performances revealed to be questionable or not in line with the income per capita in the country17. Some industries as well had to be merged, due to insufficient observations. In fine, 12 414 enterprises (3073 for the MENA region) regrouped in eight industries were retained when estimating the production frontiers (see Annex 2) 18. As for inputs and output, investment climate variables are subject to measuring errors. In the surveys, some firms did not report the full range of investment climate measures. Other firms reported numbers that were not credible. This is also due to the fact that most of investment climate factors are qualitative variables of perception, thus allowing answers to vary depending on the firms, the regions or the countries. Our choice has been to keep as many firms as possible, providing sufficient information on a wide range of investment climate variables. Once outliers and incomplete observations were removed, 5002 observations were left, among which 1483 for the MENA region, what represent 34% of MENA initial population and 30% of the total number of enterprises with IC variables (see Annex 2). 19. Another question relates to the endogeneity of the IC variables, due to the qualitative nature of investment climate factors. This is particularly true for perception variables (such as obstacles to operation) for which firms are asked to position their answer on a given scale 20. The perception of the scale might be different across firms, industries, regions and countries. Besides, when answering the questions on their investment climate, firms may be influenced by the perception they have of their own productivity and may attribute their inefficiencies to external factors. High-performing firms, as well, may be proactive in reducing their investment climate constraints, for example by working with the authorities to limit inspections or secure more reliable power supply. They also can choose a location with better infrastructure and production conditions, what relates to the endogeneity of implantation. In the empirical part, we assume these endogeneneities and use appropriate estimation techniques to evaluate the impact of the investment climate on the firms' productive performances. We measure in particular investment climate variables as city or region-sector averages of firm-level observations 21. This also helps to mitigate the effects of missing observations for some firms. Actually, if we take each investment climate indicator at the firm level, we end up with a smaller sample of observations in which all indicators are available. Furthermore, to address the issue of endogeneity of firms' implantation, we restrict the sample to the enterprises that are less likely to choose their location. We define a category of domestically owned firms employing less than 150 workers by excluding from the sample the foreign, as well as large domestically owned firms, Exchange rate constitutes another source of uncertainty which may lead to over or under evaluate firm's productive performances. This rate is used to convert production and production factors into US dollars. Several exchange rates can be chosen to calculate and 7 compare firm-level productivity across countries. In this study, we considered the current market rate in US dollars which has the interest to be the rate that firms use for their economic calculations 22. 4- Estimating Firm-Level Productivity: MENA Performance Gap In this section, we present our three measures of firm-level productivity: Productivity of Labor (LP), Total Factor Productivity (TFP) and Technical Efficiency (TE). The data have been pooled across the 22 countries of our sample 23. Firm-level productive performances are calculated for each of the eight industries. Differences and similarities across countries have been analyzed. A pattern of generally low productive performances is observed in the MENA region, with however some countries showing better results. 4.1- Firm-level Labor Productivity and the Unit Labor Cost Firm-level Productivity of Labor (LP) is estimated as the ratio of the firms' Value Added (Y) to the Number of Permanent Workers. The Value Added is calculated as the difference between Total Sales (S i, j) and Total Purchase of Raw Material -- excluding fuel (IC i, j) 24. We make the hypothesis that firms are price takers, thus purchasing raw material at world price, what looks like an acceptable assumption for the manufacturing industry which is exposed to international competition. In this case, prices in dollar of production and intermediary consumptions are comparable across countries. Equation is as follows: LP i, j = Y i, j /L i, j (6) With - Yi, j: Value Added. - L i, j : Number of Permanent Workers - i / j: Enterprise and country index respectively. Tables 2 and 3 display (by country and by industry) the averages Labor Productivity (LP) and Unit Labor Cost. Unit Labor Cost has been computed by dividing the "Total Wages" (W) by the "Value Added" (Y). For each country, average productivity (Unit Labor Cost) is expressed in percent of the level of the country with the most performing firms (or the country with the lowest Unit Labor Cost). Calculations in level are given in Tables 3.1 and 3.2 in Annex 3 25. The analysis reveals a relatively stable ranking of countries. South African and Brazilian firms perform -- in average and in most industries -- the best. This result is consistent with the relatively high incomes in the two countries (2710 and 2780 dollars per capita respectively, see World Bank, 2005). Morocco (2004)'s firms also participate in the best performances of the sample, especially in Metal & Machinery Products, Chemical & Pharmaceutical Products, Leather and Agro-Processing. As far as other MENA countries are concerned, the ranking remains also rather stable. Egyptian and Lebanese's firms are systematically among the least performing in all industries (although Morocco and Egypt have the same GDP per capita, at around 1300 US dollars in 2003). In Algeria, firm-level Productivity of Labor (LP) ranks an intermediate position, close to India in Agro-Processing and Chemical & Pharmaceutical Products, but behind in Textile and Metal & Machinery Products (firms' performances are always lower than in China). Moroccan's firms thus remain the most performing ones in MENA, with levels of Productivity of Labor (LP) far ahead from the two Asiatic giants, and close to the most productive firms/countries of the sample26. 8 This relative efficiency of some MENA countries, however, is not sufficient to understand the capacity of these countries to promote industrial and export activities. Remuneration of labor is an important factor which should be in line with productivity. By combining information on Productivity of Labor (LP) and the cost of labor, the relative Unit Labor Cost gives an idea of the competitiveness. Table 3 presents some information on the subject. It is worth noticing that the Unit Labor Cost in MENA is one of the highest of our sample of countries. This is particularly true in Algeria and Egypt ­ countries where firm-level Productivity of Labor (LP) is among the lowest ­ but also in Morocco and to some extend Lebanon. In MENA, the Unit Labor Cost is of the same magnitude than in the most performing countries of the sample, sometimes even higher (see the case of Brazil), and by far much superior than in the majority of Asian economies (India, China, Sri Lanka, Bangladesh and Thailand). In China and India, salaries (around 100 US dollars per month for unskilled workers) are far lower than in Morocco (more than the double). In the labor intensive sectors of Textile and Garments, cost of labor is 2 to 2 and a half time higher in Egypt and Morocco than in India. This situation should be seriously addressed, if MENA wants to compete in the world market. If not, MENA will continue to suffer from the faster technological innovation in Asia, where wages remain low. Table 1. Firm-Level Labor Productivity (Country average, in % of the country with the most productive firms) Agro Metal & Chemic Wood Non Metal Country* Textile Leather Garment Processing Machinery & Pharm & & Plastic Products Products Furniture Materials South Africa (2003) 52 100 100 94 97 87 100 Brazil (2003) 100 100 50 50 66 100 38 Morocco (2004) 54 80 54 79 100 91 66 Morocco (2000) 56 94 55 85 48 63 57 Saudi Arabia (2005) 77 92 100 Ecuador (2003) 58 91 80 48 50 54 42 66 El Salvador (2003) 71 59 55 35 28 51 46 China (2002) 52 69 45 31 Thailand (2004) 62 62 45 40 31 43 Guatemala (2003) 43 64 31 26 36 33 48 India (2002) 35 66 53 21 22 17 Honduras (2003) 56 50 29 23 39 21 26 India (2000) 39 48 28 24 Pakistan (2002) 40 35 49 22 17 Tanzania (2003) 35 20 Philippines (2003) 32 32 14 Algeria (2002) 27 21 19 19 31 Bangladesh (2002) 18 53 16 9 11 Nicaragua 2003 13 38 26 17 13 17 16 21 Sri Lanka (2004) 13 27 9 17 28 Zambia (2002) 16 13 24 18 Ethiopia (2002) 11 20 20 10 10 Egypt (2006) 14 15 14 12 16 11 10 13 Egypt (2004) 15 20 14 9 11 11 11 11 Lebanon (2006) 11 17 8 7 Note : * Ranking of countries goes from the ones with the most productive firms to the ones with the least productive firms. Source. Authors' calculations 9 Table 2. Firm-Level Unit Labor Cost (Country average, % of the country with the highest unit cost) Agro Metal & Chemic Wood Non Metal Country* Textile Leather Garment Processing Machinery & Pharm & & Plastic Products Products Furniture Materials El Salvador (2003) 52 100 100 85 100 63 87 Nicaragua (2003) 100 72 80 87 88 100 92 79 Guatemala (2003) 64 83 100 79 87 89 74 Algeria (2002) 73 89 89 96 100 Philippines 2003) 66 92 83 South Africa (2003) 86 97 74 80 88 69 64 Morocco 2004) 81 79 91 75 75 76 60 Honduras (2003) 36 78 88 76 63 96 86 Egypt (2004) 51 66 77 77 55 86 100 57 Egypt (2006) 60 86 76 71 46 80 92 51 Saudi Arabia (2005) 89 59 55 Lebanon (2006) 55 53 61 92 Morocco (2000) 62 62 84 60 58 66 62 Zambia (2002) 46 75 48 88 Brazil (2003) 48 54 72 68 56 49 65 Sri Lanka (2004) 86 64 71 39 32 Bangladesh (2002) 49 34 60 69 55 Ethiopia (2002) 71 25 45 56 55 Ecuador (2003) 48 59 52 50 42 32 62 53 Thailand (2004) 42 56 49 35 52 34 China (2002) 39 41 54 38 Pakistan (2002) 31 41 33 47 51 India (2000) 36 38 37 46 India (2002) 32 27 35 42 35 44 Tanzania (2003) 33 31 Note : * Ranking of countries goes from the ones with the most expensive labor to the ones with the least expensive one. Source. Authors' calculations 4.2-Firm-Level Total Factor Productivity In this section, firm-level Total Factor Productivity (TFP) is calculated from a non parametric production function. Production factors include Labor (L) and Capital (K). Same hypotheses and definitions as before apply to input and output variables. Equation is as follows: PTF i, j = Log(Y i, j) ­ Log (K i, j) ­ Log (L i, j) (7) With - Y i, j: Value Added - L i, j: Number of Permanent Workers - K i, j: Gross Value of Property, Plant and Equipment 10 - : Ratio of Total Wages (W) to Total Production Cost (Y). - = 1- - i / j: Enterprise and country index, respectively Table 3. Firm-Level Total Factor Productivity (Country average, in % of country with the most productive firms) Agro Metal & Chemic Wood Non Metal & Country* Textile Leather Garment Processing Machinery & Pharm & Plastic Products Products Furniture Materials South Africa(2003) 88 100 100 91 82 100 100 Brazil (2003) 100 100 87 100 100 100 91 Morocco (2000) 80 81 79 79 70 90 71 Thailand (2004) 70 90 75 73 78 82 Morocco (2004) 73 64 77 77 70 79 80 Saudi Arabia(2005) 70 68 81 Ecuador (2003) 69 74 76 73 75 72 78 64 El Salvador (2003) 76 70 66 64 61 69 76 Philippines (2003) 64 77 65 Algeria (2002) 65 44 59 66 76 Honduras (2003) 61 72 55 57 84 50 54 Guatemala (2003) 65 67 54 62 56 54 73 India (2000) 67 63 58 58 China (2002) 59 58 56 45 Zambia (2002) 58 52 55 52 Pakistan (2002) 55 58 56 54 48 India (2002) 59 61 49 54 51 50 Tanzania (2003) 55 53 Sri Lanka (2004) 41 51 61 51 56 Bangladesh (2002) 51 46 57 50 44 Nicaragua (2003) 49 51 45 47 42 50 44 52 Ethiopia (2002) 51 34 46 49 36 Lebanon (2006) 35 39 40 37 Egypt (2004) 41 36 35 39 34 33 36 43 Egypt (2006) 37 30 33 41 34 34 31 38 Note: * Ranking of countries goes from the ones with the most productive firms to the ones with the least productive firms. Source: Authors' calculations Table 3 presents the firm-level relative TFP by industry, under the reasonable assumption that a sector-based technology leads to a more homogeneous production function. As for Productivity of Labor, results are presented in percent of the average TFP of the most performing country (detailed calculations are given in Annex 4). Conclusions are quite similar than for Productivity of Labor. A first conclusion concerns the ranking of the most performing countries. As previously, South Africa and Brazil present, in most industries, the most productive firms. These countries are again followed by Morocco, which firms' productive performances are quite good in most industries. When compared to Brazil, Moroccan firms show a TFP gap of 10 to 30 percent depending on the sector, what is less than the revenue gap between the two countries (47 %, and 38.5% in PPP respectively). As far as other MENA countries are concerned, ranking is also quite similar than for Productivity of Labor (LP). As previously, Egypt and Lebanon rank at the bottom of the sample (with a limited number of enterprises for the latter country), while 11 Algeria stays in an intermediate position. TFP calculations thus confirm the productivity gap assessed through Productivity of Labor 27. 4.3- Firm-Level Technical Efficiency Firm-level Technical Efficiency (TE) is based on the likelihood estimation procedure. As seen in section 2.2., this method allows splitting the error term into two independents factors: the error term (v), which follows a normal distribution, and the Technical Efficiency (u), which obeys a truncated normal distribution. The technology of production explains the Value Added (Y) by the Capital (K) and the Labor (L). Same hypotheses and definitions as before apply to input and output variables. Equation is as follows: Log(Y i, j) = Log (K i, j) + Log (L i, j) + dum i, u i, j v i, j (8) With: - Y i, j: Value Added - L i, j: Number of Permanent Workers - K i, j: Gross Value of Property, Plant and Equipment - dum j: Country-dummy variables - , : parameters of the equation - vi, j: Error term - u i, j: Technical Efficiency (TE). - i / j: Enterprise and country index respectively. Production frontiers have been estimated by industry. As mentioned before, this leads to more homogeneous production frontiers and makes it easier to attribute the residual to differences in efficiency. Differences in coefficients of capital and labor have justified this choice; against an alternative assumption consisting in estimating the same production frontier for all sectors, with specific sector-based dummies (see Table 4). Table 4: Estimations of the Stochastic Production Frontiers Dependant Variable: Value Added Independent Textile Garment Leather Agro Metal & Chemic Non Metal Wood Variables Processing Machinery & Pharm & Plastic & Products Products Materials Furniture Log (labor) 0.659 0.811 0.826 0.695 0.877 0.673 0.886 0.941 (30.53)*** (42.69)*** (20.20)*** (31.22)*** (33.21)*** (22.21)*** (22.35)*** (29.18)*** Log (capital) 0.354 0.260 0.277 0.404 0.289 0.444 0.281 0.228 (24.87)*** (20.96)*** (11.00)*** (28.62)*** (18.52)*** (22.89)*** (13.54)*** (12.79)*** Intercept 2.007 1.350 1.419 1.863 1.716 2.065 1.419 1.644 (18.94)*** (9.22)*** (9.81)*** (13.99)*** (15.61)*** (15.39)*** (9.73)*** (11.51)*** ²u 0.33 0.22 0.80 0.73 1.12 0.39 1.30 0.79 ² 0.99 0.92 1.40 1.47 1.76 1.13 1.86 1.19 ²u/ ² 0.33*** 0.24*** 0.57*** 0.50*** 0.64*** 0.35*** 0.70*** 0.66*** (6.17) (3.00) (6.33) (8.17) (12.80) (5.00) (10.00) (13.20) Observations 2011 2800 634 2190 1622 1274 907 1033 Note: * Significance level 10 %; ** 5 %; *** 1 %. Z statistics are into brackets. Regressions include country-dummy variables. Source: Authors' calculations 12 Table 4 presents the estimation results of the production frontiers. In most industries, the sum of the coefficients relative to labor and capital inputs is close to one. It is a little bit higher for some sectors than can be suspected to face investment indivisibilities. In comparison with other sectors, Textile is probably the most exposed to the competition and the production technology does not reject this hypothesis. For all industries, the coefficients are strongly statistically significant at the 99% level of confidence. Table 5. Firm-Level Technical Efficiency (Country average, in % of country with the most productive firms) Agro Metal& Chemic Wood Non Metal Country* Textile Leather Garment Processing Machinery & Pharm & & Plastic Products Products Furniture Materials South-Africa 2003 85 100 100 100 89 100 100 Brazil 2003 100 100 87 80 98 100 62 Morocco 2004 58 70 81 70 100 72 92 Saudi-Arabia 2005 72 76 81 Morocco 2000 67 76 80 71 68 83 70 Thailand 2004 64 93 67 65 47 66 Ecuador 2003 57 86 61 61 63 60 57 63 El Salvador 2003 40 62 65 58 55 63 66 Guatemala 2003 51 77 45 57 45 48 67 Honduras 2003 58 66 42 48 60 37 48 India 2000 47 66 45 34 India 2002 42 56 66 41 46 32 Pakistan 2002 43 49 61 40 31 China 2002 46 45 51 35 Philippines 2003 36 53 39 Algeria 2002 33 35 39 38 54 Nicaragua 2003 22 55 41 34 38 30 31 49 Tanzania 2003 43 32 Zambia 2002 29 30 41 21 Sri Lanka 2004 17 37 26 33 39 Bangladesh 2002 24 41 32 28 19 Ethiopia 2002 20 30 36 22 23 Egypt 2004 21 30 21 17 22 17 19 32 Egypt 2006 17 15 22 22 25 14 19 24 Lebanon 2006 21 23 16 13 Note : * Ranking of countries goes from the ones with the most productive firms to the ones with the least productive firms. Source. Authors' calculations Table 4 also specifies the percentage of the residual explained by the Technical Efficiency (TE). It can be seen that, in all industries, the efficiency term accounts for a significant part of the total residuals and is statistically significant at 99%. This result justifies the production frontier approach, against the production function approach. In this model, TE explains from 24% of the error term in Garment to 70% in Non Metallic & Plastic Materials. TEs are distributed in an interval of 0 to 1 (1 is the value of the sector's most efficient firms; see Annex 5). In Table 5, TEs are in percent of the average TE of the most performing country. In average, our results for Technical Efficiency (TE) are close to the ones obtained for the non parametric TFP under the hypotheses of constant returns to scale. The ranking of countries, in particular, remains unchanged. As previously, Brazilian and South African's firms show the best performances in all industries, along with Moroccan's firms. Only in Garment and Leather, Moroccan's firms are surpassed by Thailand and Ecuador respectively. Ranking of MENA countries, as well, is unchanged. 13 4.5- Firm-Level Productivity Measures: A High Correlation Annex 6 displays, by industry, the Spearman coefficients of correlation of our three measures of firm-level productivity. All coefficients are highly significant and show a high degree of correlation between the different measures. This is the case in all industries, but more specifically in Wood & Furniture, Non Metallic & Plastic Materials, and Metal & Machinery Products (after Agro-Processing, Chemicals & Pharmaceutical Products, Leather, and Textile). This result justifies our choice of Technical Efficiency (TE) as a measure of firm- level productivity to be explained by the countries/ industries investment climate (see section 6). It will also make our findings more general, because they can be extrapolated to the different indicators of firm-level performance. 5- Assessing the Investment Climate of the Manufacturing Industries Another step in our analysis has been to differentiate and categorize the different dimensions of the investment climate. The World Bank Investment Climate (ICA) surveys provide information on a large number of investment climate (IC) variables -- in addition to general information on firms' status, productivity, sales and supplies. These IC variables are classified into 6 broad categories: (a) Infrastructures and Services, (b) Finance, (c) Business- Government Relations, (d) Conflict Resolution/Legal Environment, (e) Crime, (f) Capacity, Innovation, Learning, (g) Labor Relations. In the surveys, there are multiple indicators that cover a similar theme. Within the same theme, the correlation between indicators can be high. One solution consists to limit the number of indicators. This can however lead to a biased estimation, due to the omission of important explanatory variables. Also, it is not sure that the IC variables retained are good proxy of investment climate. A solution to overcome these problems consists in generating a few composite indicators. Because we intend to determine which investment climate variables are more detrimental to firm performances, we tried to take into consideration an as large as possible set of IC variables which are not typically used in the literature. Since these variables are likely to be correlated, we applied Principal Component Analysis (PCA) to produce a limited number of composite indicators 28. Based on the ICA surveys, we defined the investment climate by four broad categories: "Quality of Infrastructure" (Infra), "Business-Government Relations" (Gov), "Human Capacity" (H), and "Financing Constraints" (Fin). As seen in section 3, our choice of indicators has been restricted by important data limitations. This is also why we have not been able to cover all aspects initially developed in the surveys. Indicators have been selected on the bases of being available for the countries of our sample, as well as capturing the different key dimensions of the investment climate. Besides, we have tried to complete as much as possible the qualitative (perception-based) IC indicators by quantitative information, in order to get a better picture of the investment climate in each industry/country. The Quality of Infrastructure indicator (Infra) has been defined by six variables: Obstacle for the operation of the enterprise29 caused by deficiencies in (a) Telecommunications, (b) Electricity, and (c) Transport; (d) Does the Firm Own or Share a Generator, (e) if yes, which Percentage of Electricity Comes from that Source; Does the Enterprise have access to (f) E- mail or (g) Internet in its Interaction with Clients and Suppliers. Infrastructure deficiencies constitute an important constraint to private sector development in developing countries (see 14 World Bank, 1994). In the literature, deficiency in infrastructure is seen as a burden for enterprises operations and investment. Infrastructures are considered, as well, as a complementary factor to other production inputs. In particular, infrastructure stimulates private productivity by raising profitability of investment 30. Furthermore, infrastructure increases firms' productive performances by generating externalities across firms, industries and regions 31. The "Business-Government Relations" indicator (Gov) includes three to six variables (depending on the industries): Obstacle for the operation of the enterprise caused by (a) Tax Rate, (b) Tax Administration, (c) Customs and Trade Regulations, (d) Labor Regulation, (e) Business Licensing and Operating Permits, and (f) Corruption. This indicator illustrates the capacity of the government to provide an investment-friendly environment and reliable conditions to the private sector. Corruption is seen as having an adverse effect on firms' productive performances. This fact is well documented and often described as one of the major constraints facing enterprises in the developing world (see the World Bank, 2005). Corruption increases costs, as well as uncertainties about the timing and effects of the application of government regulations (see Tanzi and Davooli, 1997). Taxation and regulations have also a first order implication on firms' costs and productivity. Although government regulations and taxation are reasonable and warranted in order to protect the general public and to generate revenues to finance the delivery of public services and infrastructures, over-regulation and over-taxation deter productive performances by raising business start-up and firms' operating costs. The "Human Capacity" indicator (H) is represented by three to four variables: Obstacle for the operation of the enterprise caused by deficient (a) Skill and Education of Available Workers; (b) Education level 32 and (c) Years of Experience of the Top Manager; (d) Training of the Firm's Employees. Human capital constitutes an essential factor of firms' productive performances. Human capital stimulates capital formation by raising firms' profitability. Human capital is also at the origin of positive externalities 33. Because skilled workers are better in dealing with changes, a skilled work force is essential for firms to adopt new and more productive technologies (see Acemoglu and Shimer, 1999). Besides, new technologies generally require significant organizational changes, which are better handled by a skilled workforce (see Bresnahan, Brynjolfsson and Hitt, 2002). Human capital gives also the opportunity to the enterprises to expand or enter new markets. The "Financing Constraints" indicator (Fin) consists of three variables: Obstacle for the operation of the enterprise caused by: (a) Cost, and (b) Access to Financing; (c) Access to an Overdraft Facility or a Line of Credit. Access to (and cost of) financing represent major determinant(s) of firms' productive performances. Access to financing allows firms to finance more investment projects, what leads to an increased productivity through higher capitalistic intensity and technical progress embodied in the new equipments. Besides, financial development has a positive effect on productivity as a result of better selection of investment projects and higher technological specialization through diversification of risk. A developed financial system creates more profitable investment opportunities by mobilizing and allocating resources to the projects that will generate the most surplus (see Levine, 1997, for a synthesis). All four aggregated indicators have been generated at the branch level, thus defining in each country the specific investment climate of each industry. This has implied to produce 32 aggregated indicators (four indicators multiplied by eight industries) by applying Principal Component Analysis (PCA) to the initial indicators 34. The analysis usually treats investment 15 climate as exogenous determinant of firms' performance. As seen in section 3.3, however, this is not always the case. In order to address this issue, we have measured IC variables as city or region-sector averages of firm-level observations. This has helped, as well, to increase the number of observations by integrating in the sample firms for which information is insufficient. This has been done for "Infrastructure" and "Business-Government Relations". For "Human Capacity" and "Financing Constraints", however, the initial indicators having been interpreted as specific to each firm, information has been kept at the firm level (except for the variable "Skill and Education of Available Workers") . The initial IC indicators are presented in Annex 7, along with some information on firms' characteristics. The figures highlight well, in average, MENA deficient investment climate, as well as the specificities of the industrial sector in the region. 6- Investment Climate: How Do MENA Economies Perform? Chart 1 in Annex 8 confirms what we know of MENA investment climate. When MENA is compared to the non MENA countries of the sample, the region always ranks below. This is true for all four dimensions of the investment climate. MENA investment climate is in average of poorer quality than in East Asia (EAP), Africa (AF), Latin America (LA) and South Asia (SA) -- except for the quality of infrastructures which appear as less a constraint than in this last region (see Chart 2 in Annex 8). These findings, which are in line with the literature (see World Bank, 2005), can clearly be related to the disappointing firm productive performances assessed previously. A more detailed analysis reveals, however, differences across countries and indicators. It is again Morocco who seems to suffer the least from IC limitations, except from financing constraints. Quality of infrastructures, business-government relations and human capacity inadequacies do not appear very much higher than in South Africa, the country where firms' productive performances are in average the highest (see Chart 3 in Annex 8). On the opposite, firms in Lebanon appear to face strong inadequacies in infrastructures and business- government relations. Egypt and Saudi Arabia are in an intermediate position, with however relatively high deficiencies in the business-government relation, in particular in Egypt. These results are also in line with our findings on firms' productive performances (see section 3). As far as the different dimensions of the IC are concerned, a disaggregated approach shows which specific aspects are of more concern in the region. Limitations in all three components of the financing constraint demonstrate MENA deficit and cost of funding. This is also true for most dimensions of human capacity and of government-business relations (in particular the tax rate and administration, the labor regulations, and the licenses and operating permits, see Table 1 in Annex 7). MENA deficient financial system, as well as firms difficulties (SMEs in particular) in finding affordable credit, are important aspects often emphasized in the literature, at the same time as the limitations of various dimensions of the business environment and the lack of training and expertise of the labor force 35. As far as the quality of infrastructures is concerned, our results are more mitigated than usually highlighted in the literature. If MENA firms seems, in average, to face more constraints in electricity delivery (more enterprises rely on a generator), as well as in internet connection, quality of telecommunications and transports does not appear as very strong obstacles to operation (see Table 1 in Annex 7). Although this finding looks somehow in contradiction with the conclusions of several studies, differences may be due to our small number of MENA countries, as well as to the presence of Morocco whose quality of infrastructures is not perceived as a strong limitation 36. These results are confirmed at the country level, with 16 Morocco experiencing more deficiencies in the various dimensions of the financing environment, Egypt and Saudi Arabia in different aspects of the government-business relation and Lebanon in all components of the quality of infrastructure and government-business relation (see Charts 4 to 9 in Annex 8). Finally, MENA enterprises are characterized by a smaller size and a lower export orientation than in the rest of the sample (see Table 1 in Annex 7). Morocco, however, show a high export rate, in particular in the Textile, Leather and Garment industries, as well as Lebanon, in Wood and Furniture. Morocco presents as well an above average foreign participation in the capital of the firms (see Chart 10 in Annex 8). 7- Is the Investment Climate Explaining Firm-Level Productivity? In this section, we estimate several variants of a model of Technical Efficiency (TE) which explains the production frontier and the factors contributing to the efficiency at the same time (following the one step procedure, see section 2). Investment climate is first defined by a few indicators of infrastructures, human capacity, government-business relation and financing constraints. To overcome multicolinearity, we introduce then our four IC composite indicators: Infra, H, Fin and Gov. After having controlled for endogeneity of IC variables, we finally address the question of endogeneity of implantation, by restricting our sample to the domestic firms of less than 150 workers. We show that our results are unambiguous and robust to the different specifications and samples of firms. 7.1- Common Model with Individual Indicators of Investment Climate Our empirical model considers a same representation for all industries. This model is estimated at the branch level, thus allowing the coefficients to vary across branches. We explain firms' production frontiers and Technical Efficiencies (TEs) at the same time by regressing the logarithm of the production factors (capital and labor), as well as various plants characteristics and investment climate variables, on the logarithm of the firms' value added. At this first stage of investigation, we have used initial IC variables before aggregation. The model is as follows: ln(y i,j) = c i + 1 ln(l i,j) + 2 ln(ki,j) + Sizei,j + Foreigni,j + Exporti,j + 1 RegElecti,j + 2 RegWebi,j + 1 Credi,j + 2 AccessFi,j + 1 EduMi,j + 2 ExpMi,j + 3 Trainingi,j + 1 × RegLreguli,j + 2 × RegCorrupi,j + c + vi,j: (9) With: y i,j Value Added 37 l i,j: Number of Permanent Workers ki,j: Gross Value of Property, Plant and Equipment Sizei,j: Size of the firm Foreigni,j: Foreign capital (% of firm's capital) Exporti,j: Export (% of firm's sales) RegElecti,j: Electricity delivery (obstacle for the enterprise, regional average) RegWebi,j: Utilization of Internet (regional average) Credi,j: Overdraft facility or credit line 17 AccessFi,j: Access to financing (obstacle for the enterprise, regional average) EduMi,j: Level of education of the top manager (number of years) ExpMi,j: Experience of the top manager (number of years) Trainingi,j: Training of workers RegLregi,j: Labor regulation (obstacle for the enterprise, regional average) RegCorrupi,j: Corruption (obstacle for the enterprise, regional average) c i: Country-Dummy variables c: Intercept vi,j: Error terms i / j: Enterprise and country index respectively The choice of the IC variables has been based on their availability for as many firms/ industries/ countries as possible, as well as on capturing the different key dimensions of the investment climate. Our variables cover properly the four axes of the investment climate defined in the previous section. To address the problem linked to the endogeneity of the IC variables when estimating the TE frontier models, we have considered the city region-sector averages of Electricity delivery (RegElect), Access to Internet (RegWeb), Labor regulation (RegLreg), and Corruption (RegCorrup). The number of explanatory variables, however, has been limited by the multicolinearity between several IC variables when estimating the TE frontier models. Other individual variables have consisted in: the percentage of sales exported by the firms (Export), the percentage of foreign ownership of firms' capital (Foreigni,j), as well as the firm size (Sizej). Export is a factor of productivity by confronting firms to international competition. Foreign ownership, as well, increases productivity if foreign investors bring new technologies and management techniques. As for the size, we intend to test the hypotheses of scales economies and increasing returns to scale in big enterprises 38. It is worth noting that the expected sign for these variables is negative, due to the fact that the one step procedure explains firm-level inefficiency. The same precautions must be taken when interpreting the sign of the coefficients of the other variables. Country-dummy variables have also been introduced when estimating the production frontiers. There are good reasons to think that production may vary across countries for motives other than production factors. The country dummies can pick up the effect of countries specific factors, such as endowment in natural resources, national-level institutions, macro or political instability, trade policy, etc... Country-dummy variables are intentionally not included in the second part of the equation, when explaining Technical Efficiencies (TEs), since they could reduce the impact of some IC variables. Equation (9) has been estimated on unbalanced panels, going from 380 observations (in Leather) to 1601 observations (in Garment) depending on the industry. A Cobb-Douglass production function has been chosen to estimate the production frontiers. We have also maintained our previous assumption as regard the specification of the technology, as well as of the Technical Efficiency (TE). Although the sample size modifies when incorporating the regressors explaining the firm distance to the frontier, the coefficients of the technology are marginally (but downward) affected. These modifications display the potential impact of the interactions and the limitation that we would face when estimating the Technical Efficiency (TE) determinants through the two stage method, as previously discussed 39. Sector-based estimates are presented in Table 6. . 18 Table 6. Estimation Results: Common Model with Individual IC Variables (Dependant Variable: Value Added) Agro Metal& Chemic Wood Non Metal Independent Textile Leather Garment Industry Machinery & Pharm & & Plastic Variables Products Products Furniture Materials ln(l ) 0.657 0.789 0.735 0.560 0.871 0.540 0.883 0.860 (16.14)*** (28.82)*** (7.12)*** (13.32)*** (21.75)*** (11.09)*** (18.78)*** (10.18)*** ln(k) 0.321 0.255 0.242 0.395 0.268 0.444 0.235 0.249 (14.61)*** (14.93)*** (7.18)*** (24.64)*** (13.21)*** (20.01)*** (11.28)*** (8.81)*** Intercept 0.720 1.597 1.993 3.780 1.654 2.985 0.157 1.251 (1.55) (4.21)*** (2.25)** (5.79)*** (4.88)*** (6.08)*** (0.55) (2.22)** Size 0.018 -0.105 -0.092 -0.195 0.600 -0.193 -0.316 0.014 (0.11) (0.21) (0.48) (2.57)** (0.96) (1.92)* (1.29) (0.07) Foreign -0.242 -0.384 -0.011 -0.005 -0.397 -0.005 -0.000 -0.007 (0.53) (0.43) (1.30) (3.36)*** (1.16) (1.88)* (0.01) (1.07) Export -0.006 -0.183 -0.007 -0.001 -0.107 -0.005 -0.019 -0.009 (1.06) (1.43) (2.87)*** (1.06) (0.97) (1.64) (1.22) (1.32) RegElect 0.077 0.323 0.228 0.042 1.006 0.053 -0.025 0.068 (0.54) (0.60) (1.94)* (0.83) (1.92)* (0.86) (0.16) (0.60) RegWeb -2.641 2.138 0.329 -0.426 0.768 -0.757 -1.542 -0.847 (2.43)** (1.26) (0.94) (2.07)** (0.50) (3.39)*** (1.77)* (1.57) Cred -1.011 -2.421 -0.403 -0.144 -1.842 -0.085 -0.304 -0.554 (2.08)** (2.42)** (2.74)*** (2.38)** (2.07)** (1.02) (1.25) (2.26)** AccessF 0.006 0.118 0.059 0.044 -0.022 0.068 0.126 -0.051 (0.11) (0.65) (1.41) (2.34)** (0.11) (2.43)** (1.74)* (1.22) Training -0.135 0.234 -0.142 -0.217 0.428 -0.123 -0.400 -0.103 (0.43) (0.33) (0.93) (3.23)*** (0.56) (1.22) (1.34) (0.59) EduM -0.148 -0.282 -0.076 -0.064 -0.673 -0.073 -0.096 -0.158 (2.02)** (1.53) (2.08)** (3.03)*** (2.61)*** (1.96)* (1.46) (2.84)*** ExpM -0.037 0.045 -0.000 -0.003 0.014 -0.002 -0.006 -0.000 (2.26)** (1.50) (0.05) (0.90) (0.48) (0.38) (0.56) (0.04) RegLregul 0.024 -0.827 -0.069 0.007 0.362 0.020 -0.112 -0.006 (0.13) (1.52) (0.50) (0.10) (0.70) (0.20) (0.53) (0.05) RegCorrup 0.081 0.074 0.168 -0.054 -0.272 -0.008 0.073 0.124 (0.51) (0.17) (1.53) (0.96) (0.59) (0.11) (0.52) (1.40) Constant 1.460 -2.422 1.493 3.388 -2.612 2.358 1.279 1.568 (2.87)*** (1.25) (2.00)** (5.45)*** (1.34) (4.94)*** (1.91)* (2.66)*** Observations 942 380 1601 1494 838 695 774 480 sigma_u 0.75 1.69 0.77 0.90 1.46 0.75 1.10 0.64 sigma_v 0.86 0.81 0.54 0.43 0.76 0.46 0.57 0.67 Wald chi2 1351.45 2787.67 241.01 1306.40 2484.52 1060.30 1321.23 300.67 Prob > chi2 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Notes: The one step procedure explains firm-level inefficiency. Variables Size, Foreign and Export are expected with a negative coefficient. All regressions contain country-dummy variables when estimating the production function. * Significant at 10%; ** significant at 5%; *** significant at 1%. Absolute value of z statistics are in parentheses. Source. Authors' estimations. A first set of conclusions concerns the production frontiers. Our regressions confirm the choice to estimate a production frontier by industry. Elasticites of capital and labor reveal to be different from one industry to another. Impact of capital is strong in Chemicals & Pharmaceutical Products, Agro-Processing and, to a lower extend, Textile. On the opposite, elasticity of labor is higher in Metal & Machinery, Non Metal & Plastic Materials, Wood & Furniture, Leather and Garment. These industries look like being more intensive in labor, although two of them: Metal & Machinery and Non Metal & Plastic Materials, are usually considered as applying more capitalistic technologies in developed economies. This result is confirmed by the computation of the ratio of the two elasticities (capital/ labor). All coefficients are highly significant (at 1% level), what stresses the robustness of our results 19 Another result shows that we are close to the constant returns to scales, legitimating the hypothesis underlying the non parametric TFP measures (see section 2.1). Our estimations also highlight that some differences in production frontiers can be explained by countries specific conditions. This hypothesis is supported by the data, as country-dummies are well significant at this stage of estimations. More interesting, our estimations verify that differences in the investment climate participate in firms' Technical Efficiencies (TE) discrepancies. This is true for all aspects of the investment climate, except for the "Government-Business" relations. Our results confirm that a good quality of infrastructure (proxied by the quality of the electric network and the availability of internet access), a satisfactory access to financing, as well as the availability of expertise at the firm level (such as education and experience of the manager, and training of the employees) are important factors for enterprises productive performances. This outcome, which is consistent with the theory, makes a real contribution to the empirical literature by validating, for a large sample of industrial firms in developing countries, the role of a substantial set of IC variables on firms' productive performances. This finding appears, however, quite different from one industry to another. First, as expected, it looks like that estimation has suffered from the colinearity of several IC variables. In fact, although each broad category of IC variables (except Government-Business Relation) ends up being significant in almost all industries, it is very rare to find two significant IC variables in the same category40. Impact of IC variables can also vary. Access to credit seems more detrimental in Leather, Metal & Machinery Products and Textile (the estimated coefficient of this variable in higher than in the other industries). Access to the internet looks more critical in Textile and Wood & Furniture. As for Human Capacity, the education of the top manager should be more a high priority in Metal & Machinery Products, Textile and Non Metal & Plastic Materials. Interestingly, Textile and Metal & Machinery Products look more sensitive to IC deficiencies. Beside, firms' performances depend on more dimensions of the IC in these two sectors. This finding may be explained by the fact that these industries are more exposed to international competition and need a supportive investment climate to be able to compete efficiently. As for Business-Government Relations, neither labor regulations (RegLreg), nor corruption (RegCorrup) emerge as an obstacle to firms productive performance, although this outcome has to be considered with caution because of the probably high correlation between explanatory variables. Difficulties have also occurred in validating the impact of other individual variables. Firms' size (Size) and foreign ownership of capital (Foreign) justify scales economies and externalities linked to participation of foreign capital in just two sectors (Agro-Processing, and Chemical & Pharmaceutical Products). Export orientation (Export) appears as a determinant of productivity in only one industry: Garment. This result meets, however, with what we know about this sector, where external competitive markets are a stimulating source for a high productivity level. Identically, regressions results are poor in two sectors: Leather and Wood & Furniture 41. All these difficulties, when individual factors are considered, explain why we have then focused our analysis on a few IC composite indicators. These indicators are tested econometrically in the next section. 7.2- Common Model with Composite Indicators of Investment Climate In this specification, the IC individual variables have been replaced by our four composite indicators: Quality of Infrastructure (Infra), Business-Government Relations (Gov), Human Capacity (H), and Financing Constraints (Fin). This model allows introducing much more IC 20 variables than previously 42. Like in the first empirical model, we have considered a same representation for all industries. The model is still estimated at the branch level and explains the logarithm of the firms' value added and Technical Efficiency (TE) in one step. Other control variables are unchanged. The model is as follows: ln(y i,j) = c i + 1 ln(l i,j) + 2 ln(ki,j) + Sizei,j + Foreigni,j + Exporti,j + 1 RegInfrai,j + 2 ,RegGovi j + 3 Hi,j + 4 Fini,j + c + vi,j: (11) Estimation results reinforce our previous findings (see Table 7). Production frontiers are robust to the introduction of different IC variables, with little changes in returns to scales or in the elasticities of production factors across industries. Countries specific conditions are also validated by the data. One of the most interesting outcomes, nevertheless, concerns the investment climate which four dimensions are now significant with the expected sign. As we actually explain firm-level inefficiency, a positive coefficient is expected for three out of our four indicators. This is the case of RegInfra, RegGov and Fin, which are interpreted as obstacle for the operation of the firms. On the opposite, H being constituted of variables which are supposed to improve Technical Efficiency, a negative coefficient is expected for this variable (see section 5 for the definition of the axes of the composite indicators). Beside, our model validates the impact of a much more substantial number of IC variables incorporated in the aggregated indicators. This result has to be stressed because it is the first time (to our knowledge) that the empirical literature brings evidence of the role of such a significant set of IC variables for such a large and diversified sample of industrial firms. It is also of first important for MENA, knowing the deficiencies of different dimensions of the investment climate which improvement could constitute a powerful mean of boosting firms' efficiency and of catching up with more efficient and competitive countries. Improving the financial environment in Morocco, the government-business relation in Egypt, Saudi Arabia and Lebanon, and the quality of infrastructure in Lebanon in particular would go in this direction. Findings by industry bring, as well, quite interesting comments. Human Capacity (H), Infrastructure (Infra), and Financing Constraints (Fin) appear to be the most robust investment climate factors for firm-level productivity. All three broad indicators explain quite well productivity discrepancies in most industries while Business-Government Relations (Gov) constitutes a less constant dimension. Our empirical analysis also reveals that some industries: Textile (for H, Infra and Fin), Metal & Machinery Products (for H and Gov) and Wood & Furniture (for H and Fin ) appear more sensitive and vulnerable than others in front of a deficit of their investment climate (the estimated coefficients of the IC variables are higher for these industries). This comment may be extended to Non Metal & Plastic Materials and Garment for, respectively, Human Capacity (H) and Government-Business Relation (Gov). These findings confirm in a different way some conclusions of the previous model. As mentioned before, this result may be due to the fact that most of these industries face international competition. This fragility justifies that a particular attention be paid when taking decisions that may affect these sectors' investment climate. This also means that the pay off of an improvement of the investment climate would be more substantial in these industries, which could play a leading role in the industrial capacity and export development of the countries. This conclusion is all the more important for the MENA economies, knowing the high specialization of some of them (Morocco and Egypt in particular) in the Textile and Garment industries (see Table 2 Annex 7). Improving the investment climate in these two 21 sectors would greatly help to resist to the strong international competitions and reinforce the export orientation of the two countries. Table 7. Estimation Results: Common Model with Aggregated IC Variables (Dependant Variable: Value Added) Agro Metal & Chemic Wood Non Metal Independent Textile Leather Garment Industry Machinery &Pharm & & Plastic Variables Products Products Furniture Materials ln(l ) 0.637 0.778 0.879 0.551 0.885 0.578 0.836 0.923 (16.01)*** (27.90)*** (15.19)*** (12.54)*** (25.26)*** (11.84)*** (17.87)*** (15.50)*** ln(k) 0.337 0.252 0.196 0.397 0.258 0.447 0.248 0.254 (15.06)*** (16.57)*** (7.40)*** (24.54)*** (13.11)*** (20.05)*** (11.91)*** (9.31)*** Intercept 1.081 2.149 1.326 4.302 1.883 2.868 1.738 1.223 (2.01)** (5.93)*** (4.62)*** (5.77)*** (5.90)*** (4.26)*** (4.54)*** (2.78)*** Size -0.809 -0.333 -0.037 -0.212 -0.159 -0.198 -0.490 0.273 (1.54) (1.77)* (0.33) (2.75)*** (0.22) (1.99)** (2.22)** (1.10) Foreign -0.426 -0.006 -0.014 -0.005 -0.541 -0.006 0.004 -0.019 (0.90) (0.76) (0.50) (3.48)*** (1.05) (1.72)* (0.54) (1.28) Export -0.016 -0.020 -0.078 -0.001 -0.114 -0.008 -0.017 -0.186 (0.81) (1.95)* (1.81)* (1.14) (1.04) (1.49) (1.53) (1.08) RegInfra 0.762 -0.079 -0.057 0.014 0.833 0.204 0.262 0.318 (2.52)** (0.66) (0.95) (0.27) (1.83)* (2.35)** (1.71)* (2.32)** H -0.716 -0.138 -0.116 -0.253 -1.174 -0.147 -0.488 -0.768 (1.76)* (0.79) (1.08) (5.03)*** (1.52) (1.71)* (2.33)** (2.24)** ,RegGov -0.259 -0.072 0.185 -0.047 0.706 -0.068 -0.060 0.136 (1.21) (0.72) (2.48)** (1.48) (1.70)* (1.39) (0.54) (0.86) Fin 0.778 0.219 0.035 0.124 0.257 0.148 0.330 -0.208 (2.40)** (1.68)* (0.50) (3.86)*** (0.54) (2.67)*** (2.36)** (1.26) Constant -0.961 0.162 0.506 3.243 -6.121 1.508 0.703 -0.522 (0.95) (0.19) (1.84)* (4.82)*** (2.83)*** (2.32)** (1.04) (0.71) Obs 929 433 1555 1481 826 741 750 461 sigma_u 1.31 1.11 0.25 0.91 1.98 0.70 1.10 0.56 sigma_v 0.86 0.60 0.73 0.37 0.65 0.56 0.53 0.75 Wald chi2 1579.56 2375.90 925.66 1343.79 3117.04 1010.55 1490.81 893.91 Prob > chi2 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Notes: The one step procedure explains firm-level inefficiency. The expected sign of the IC aggregated variables is positive for RegInfra, RegGov and Fin, and negative for H (see definition of variables in section 5). Variables Size, Foreign and Export are also expected with a negative coefficient. All regressions contain country-dummy variables when estimating the production function. * Significant at 10%; ** significant at 5%; *** significant at 1%. Absolute value of z statistics are in parentheses. Source. Authors' estimations By using our IC aggregate indicators, however, we don't always better explain productivity. This is somehow the case of Metal & Machinery Products and Agro-Processing, but essentially of Garment for which a very few aspects of the investment climate seem to help firms to perform better 43. No improvement is seen, either, in Leather, which is again poorly explained by the model. This fact is, however, largely compensated by the tremendous gain of information through the large set of IC variables now explaining firm-level productive performances, as well as by the validation of another variable of interest: the Government- Business Relation (Gov) 44. We will also show in the next section that Garment is better explained by the data, when dealing with small and medium domestic firms. Another addition of the model consists in validating the role of more plants characteristics in explaining firm-level Technical Efficiency (TE). This is true for the variable Size, which justify scales economies in four industries instead of two previously: Wood & Furniture and Leather in addition to Agro-Processing and Chemicals & Pharmaceutical Products. This constitutes an interesting result that would justify a policy of concentration of small enterprises, which importance in developing countries is well documented. Concentration 22 could be seen as a powerful means of boosting efficiency and competitiveness of the industrial sector, thus contributing to industrial development and economic growth. Besides, export orientation (Export) explains externalities linked to export activities in Leather, in addition to Garment (with a stronger coefficient for Garment), what confirms the exposure to international competition of these two industries. Increase the export capacity of some industries appears, though, as another mean to stimulate firm's efficiency and to promote a diversified economic growth, where industry is subject to play a major role. 7.3- Technical Efficiency and Endogeneity of Firms' Implantation Another test of robustness has consisted in addressing the question of a possible endogeneity in firms' location. City or region-sector averages IC indicators would not be exogenous regressors if, for example, more efficient firms tend to establish in locations where the investment climate is better. In order to evaluate this bias, we have re-run our previous model on a set of firms which are less likely to choose their location. This had led to eliminate foreign firms and large domestically owned firms. Following Dollar et al. (2005), we define our new sample as the domestically owned firms employing less than 150 workers 45. Results of this new set of estimations still confirm our previous findings (see Table 8). A first conclusion concerns the investment climate, which impact on firms' performances is still validated by the data. This is true for all four dimensions of the investment climate. This result confirms that small and medium domestic firms are sensitive as well to changes in the different dimensions of the investment climate. A detailed analysis also reveals that the influence of the investment climate can be different for this category of firms. This is the case in Textile, Garment and Non Metal & Plastic Materials, where impact of IC variables is stronger than for the whole sample (see section 6.2). In Textile, this is true for all three significant dimensions of the investment climate (Infra, H and Fin). In Garment, Financing Constraints (Fin) and Infrastructure (Infra) appear now as constraints for small firms' productive performances, in addition to a stronger impact of Business-Government Relations (Gov). Besides, small firms in Non Metal & Plastic Material are more sensitive to limitations in Infrastructure (Infra) and Human Capacity (H). This outcome is likely to show that, in the three industries, big and foreign firms can resist more to a degradation of the investment climate. This finding also tends to confirm that big enterprises have the possibility to influence positively their investment climate, or to establish in locations where the investment climate is more favorable. This outcome can be considered as of first importance, knowing the potential of job creation of small and medium enterprises. Actually, it is well documented that small businesses generally deal with poor investment climate. They have, for example, a more difficult and more expensive access to the financial system. They have not the power, as well, to lobby policy makers to get better regulations. They also attract less qualified people who prefer higher salaries in bigger enterprises. They have less the capacity to compensate deficient infrastructure, buying a generator or paying for expensive internet connections (in addition to the fact that they don't choose their location (see World Bank, 2005). This makes of this category of firms a great potential for an improved performances of the industrial sector. This is particularly true for our MENA economies, which are characterized by a relatively small size of firms (see Table 1 Annex 7). 23 It is also interesting to note that, when focusing on small and medium domestic firms, we find that more IC variables explain firm-level performances. This is due to the fact that big enterprises are less sensitive to IC limitations and bias downward the estimated coefficients when dealing with the whole sample. Restricting the sample to small and medium firms better highlights the impact of IC and firms characteristics on firm-level performances and competitiveness, thus drawing more substantial policy implications. This is well illustrated by the case of Garment, for which very few IC variables were previously significant. Table 8. Estimation Results: Common Model with Aggregated IC Variables and Sample Differentiation (domestic firms with less than 150 employees) (Dependant Variable: Value Added) Agro Metal & Chemic Wood Non Metal Independent Textile Leather Garment Industry Machinery &Pharm & & Plastic Variables Products Products Furniture Materials ln(l ) 0.547 0.882 0.975 0.460 0.834 0.549 0.779 0.981 (9.01)*** (23.30)*** (15.3)*** (5.92)*** (15.85)*** (6.74)*** (11.39)*** (14.07)*** ln(k) 0.319 0.252 0.177 0.384 0.251 0.390 0.223 0.252 (12.38)*** (16.25)*** (6.06)*** (18.61)*** (10.87)*** (13.89)*** (10.01)*** (8.74)*** Intercept 2.153 1.732 -0.309 2.105 1.903 2.426 2.238 1.024 (4.18)*** (5.00)*** (0.93) (2.33)** (4.86)*** (3.11)*** (3.16)*** (2.42)** Size -2.897 0.045 0.186 -0.357 -2.331 -0.345 -0.412 0.678 (1.91)* (0.27) (0.84) (2.88)*** (1.76)* (2.44)** (2.51)** (1.37) Export -0.417 -0.010 -0.003 -0.005 -0.475 -0.016 -0.013 -0.316 (0.98) (1.85)* (0.81) (1.79)* (0.99) (1.49) (1.62) (0.95) RegInfra 1.170 -0.127 0.763 0.007 0.869 0.161 0.157 0.472 (2.13)** (1.09) (2.97)*** (0.09) (2.01)** (1.83)* (1.60) (1.72)* H -1.352 -0.133 -0.276 -0.201 -1.103 -0.108 -0.263 -1.444 (2.04)** (0.86) (0.77) (2.82)*** (1.35) (1.23) (2.50)** (2.09)** RegGov -0.171 -0.105 1.552 -0.045 0.424 -0.067 -0.063 0.154 (0.51) (1.17) (2.88)*** (1.05) (0.94) (1.48) (0.81) (0.54) , Fin 1.170 0.222 0.665 0.093 0.496 0.146 0.178 -0.520 (2.05)** (1.97)** (3.33)*** (1.96)** (0.99) (2.60)*** (2.42)** (1.57) Constant -0.254 -0.348 -3.389 1.894 -1.468 1.509 1.389 -2.307 (0.15) (0.44) (2.40)** (2.58)*** (0.73) (3.33)*** (3.36)*** (1.40) Observations 730 359 1093 1123 639 607 650 395 sigma_u 1.42 1.02 0.28 0.73 1.41 0.43 0.80 0.91 sigma_v 0.90 0.45 0.73 0.77 0.71 0.80 0.51 0.69 Wald chi2 663.77 1615.42 763.09 787.50 1175.83 479.31 576.56 796.86 Prob > chi2 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Notes: The one step procedure explains firm-level inefficiency. The expected sign of the IC aggregated variables is positive for RegInfra, RegGov and Fin, and negative for H (see definition of variables in section 5). Variables Size, Foreign and Export are also expected with a negative coefficient. All regressions contain country-dummy variables when estimating the production function. * Significant at 10%; ** significant at 5%; *** significant at 1%. Absolute value of z statistics are in parentheses. Source. Authors' estimations In addition, our estimations confirm once more that it is the small and medium firms more exposed to international competition that suffer the most of the deficiencies of their IC. This is particularly true for Textile and Garment, but also to some extent for Non Metal & Plastic Material. This finding still verifies that an improvement of the IC would particularly benefit to this category of firms, which competitiveness and export capacity could be significantly boosted. Finally, another result tends to confirm the importance of the size as a factor of productivity and efficiency. Actually, small and medium domestic firms appear to gain more from concentration than big and foreign ones (what looks like a reasonable outcome). This is the case in Textile and Metal & Machinery Products where the variable Size is now significant, as 24 well as in Agro-Processing and Chemicals & Pharmaceutical Products where the coefficient of this variable shows a significant increase compared to previous estimations. This constitutes an interesting result that would again justify a policy of concentration of small enterprises 46. 8- Conclusions In this paper, we have empirically verified that investment climate (IC) matters for firms' productive performance. This finding is true for several aspects of the investment climate, in particular the quality of various infrastructures, the experience and education of the labor force, the cost and access to financing, as well as different dimensions of the government- business relation. This outcome (which is consistent with the theory) makes a real contribution to the empirical literature by validating, for a large sample of industrial firms in developing countries, the role of a substantial set of IC variables. Policy implications of our findings are comprehensible by showing what determinants of productivity cause producers to be more efficient, and where should reform be targeted to have the greatest impact on productivity. In most industries, it is the dimension of infrastructures, human capacity and/or financing that more often account for firms' productive performance. Building on these three dimensions of the investment climate would have a large pay-off for the efficiency and competitiveness of the manufacturing industry as a whole. This factor should be kept in mind when dealing with the reform agenda of many developing countries, the MENA region in particular, in which poor investment climate hinders economic development and catch-up with more efficient and competitive countries. A more in-depth analysis also reveals interesting differences across industries. Actually, although most industries appear sensitive to different dimensions of the investment climate, firms in Textile and Metal & Machinery Products look like to suffer more of investment climate limitations. This comment may be broadened, to some extent, to Non Metal & Plastic Materials and Garment. This may be due to the fact that these sectors face international competition and need a supportive investment climate to compete efficiently. This fragility justifies that a particular attention be paid when taking decisions that may affect these sectors. This also means that the pay off of an improvement of the investment climate in terms of productive performances and competitiveness would be more substantial in these industries, which could play a leading role for the industrial capacity and export development of the countries.. This result constitutes an important means of appreciation of the positive impact of investment climate improvement since MENA manufacturing suffers from a deficient integration into the world economy, as well as from a high competition in the world market. Another interesting finding can be seen in the fact that impact of investment climate varies for small and medium (under 150 workers) domestic firms. This is the case in Textile, Garment and Non Metal & Plastic Materials, where investment climate constraints emerge stronger than for the whole sample. This result is likely to show that, in these industries, big and foreign firms have the possibility to influence positively their business environment, and/or establish in locations where the investment climate is better. This finding also implies that improvement of the investment climate of small and medium enterprises in these industries would generate substantial productivity gains and largely boost competitiveness of this category of firms. This outcome has to be considered as of first importance, knowing the significance of small enterprises in developing countries, MENA in particular, as well as their substantial potential of job creation. Interestingly, another result tends to confirm the 25 importance of the size as factor of productivity and efficiency. Actually, small domestic firms appear to gain more from concentration than big and foreign ones. This is the case in Textile and Metal & Machinery Products, in addition to Agro-Processing and Chemicals & Pharmaceutical Products. This constitutes an interesting result that would justify a policy of concentration of small enterprises as a powerful means of efficiency and competitiveness of the industrial sector, thus contributing to industrial development and economic growth Actually, like other developing countries, MENA is increasingly concerned about improving competitiveness and productivity as the region face the intensifying pressure of globalization. This is particularly true in MENA Textile, Garment and Leather industries, in which export specialization can be high in some countries. Among the region, the World Bank firm-surveys provide a standard instrument for identifying key obstacles to productivity, and prioritize policy reforms. This instrument can be used to boost competitiveness and diversify MENA economies. This factor should be taken into consideration if MENA wants to face the increasing international competition of countries such as China and India, which have successfully diversify their economy and benefit, in addition, from low labor costs. Targeting reforms on small and medium enterprises, as well as on those industries and investment climate variables which are the most inadequate and which favor the most productivity would constitute an important element of MENA strategy of growth and employment for the future. 26 References Acemoglu, Daron, and Robert Shimer (1999), "Efficient Unemployment Insurance", Journal of Political Economy, 107 (5): 893-28. Acemoglu, Daron, Simon Johnson, and James A. Robinson (2001), "The Colonial Origins of Comparative Development: An Empirical Investigation", American Economic Review, 91(5): 1369­401. Acs Zoltan, and David Audretsch (1993), "Innovation and Technical Change, the New Learning", in. Libecap, G.D. (ed.), Advances in the Study of Entrepreneurship, Innovation and Economic Growth, JAI Press, Stamford, CT. Algeria (2002), Algeria Investment Climate Assessment, Pilot Investment Climate Assessment, International Finance Corporation, The World Bank, Washington, D.C. (June). Argimon I., Jose Manuel Gonzales-Paramo and P. Hernández de Cos (1997), "Evidence of Public Spending Crowding-Out from a Panel of OECD Countries", Applied Economics, 2:1001-1010. Aschauer D.A. (1989), "Is Public Expenditure Productive?", Journal of Monetary Economics, 23: 177-200, May. Aysan, A., M. Nabli and M-A Véganzonès -Varoudakis (2007a), "Governance Institutions and Private Investment: An application to the Middle East and North Africa", The Developing Economies, 45 (3): 339-377. Aysan, A., G. Pang and M-A Véganzonès-Varoudakis (2007b), "Uncertainty, Economic Reforms and Private Investment in the Middle East and North Africa", Applied Economics, 39 (21), December. Barro Robert J. (1990), "Government Spending in a Simple Model of Endogenous Growth", Journal of Political Economy, 98 (5): part II, S103-S125, October. Bastos, F., and J. Nasir (2004), "Productivity and Investment Climate: What Matters Most?", World Bank Policy Research Paper, 3335, June. Blejer, Mario I and Mohsin S. Kahn (1984), "Government Policy and Private Investment in Developing Countries ", IMF Staff Papers, 31. IMF, Washington, D.C.. Bosworth, Barry, and Susan Collins (2003), "The Empirics of Growth: An Update", The Brookings Institutions. Washington D.C., Processed. Bresnahan, Timothy, F. Erik Brynjolfsson, and Lorin M. Hitt (2002), "Information, Technology, Workplace Organization and the Demand of Skilled Labor: Firm- Level Evidence", Quarterly Journal of Economics, 117 (1): 339-76. 27 Calderon, Cesar, and Alberto Chong (2000), "Causality and Feedback between Institutional Measures and Economic Growth", Economic and Politics, 12(1): 69- 81. Coelli T., D. S Prasada Rao and G. E. Battese (1998), An Introduction to Efficiency and Productivity Analysis, Kluwer Academic Publishers. Djankov, Simeon, Rafael La Porta, Florencio Lopez-de-Silanes, and Andrei Sheifer. (2002), "The Regulation of Entry", Quartely Journal of Economics, 117, February, Dollar. David, Mary Hallward-Driemeier and Taye Mengistae (2005), "Investment Climate and Firm Performance in Developing Economies", Economic Development and Cultural Change, 54 (1): 1-21. Easterly, William, and Ross Levine (2003), "Tropics, Germs, and Crops: How Endowments Affect Economic Development", Journal of Monetary Economics, 50 (January): 3-39. Egypt (2005), Egypt: Toward a Better Investment Climate for Growth and Employment Creation, Middle East and North Africa Region, Private Sector, Financial Sector and Infrastructure Group, The World Bank, Washington, D.C. Egypt (2006), Egypt's Investment Climate: A Survey-Based Update, Middle East and North Africa Region, Private Sector, Financial Sector and Infrastructure Group, The World Bank, Washington, D.C. (December). Eifert, Benn, Alan Gelb and Vijaya Ramachandran (2005), "Business Environment and Comparative Advantage in Africa: Evidence from the Investment Climate Data", The World Bank, Washington D.C., Processed. El Badawi, Ibrahim, A. (2002), "Reviving Growth in the Arab World", Working Paper Series, 0206, Arab Planning Institute, Safat, Kuwait. Escribano, Alvaro., and J. Luis Guasch (2005), "Assessing the Impact of the Investment Climate on Productivity Using Firm-Level Data: Methodology and the Case of Guatemala, Honduras and Nicaragua", World Bank Policy Research Working Paper, WPS3621, World Bank, June Evans, Peter, and James Rauch (2000), "Bureaucratic Structure and Bureaucratic Performance in the Less Developed Countries", Journal of Public Economics, 75 (1): 49-71. Frankel, Jeffrey (2002), "Promoting Better National Institutions: The Role of the IMF". Paper presented at the Third Annual IMF Research Conference, Washington, November 7-8. Griliches, Zvi, and Jacques Mairesse (1995), "Production Functions: The Search for Identification », NBER Working Paper, 5067(March). Gupta, Sandeev, Hamid Davooli and Rosa Alonso-Terme (2002), "Does Corruption Affect Income Inequality and Poverty", Economics of Governance 3(1): 23-45. 28 Hall, Robert and Charles Jones (1999), "Why Do Some Countries Produces So Much More Output per Workers that Others?", Quarterly Journal of Economics, 114 (February): 83-116. Haltiwanger, John (2002), "Understanding Economic Growth. The Need for Micro Evidence", New Zealand Economic Paper, 36 (1): 33-58. He, Kathy S., Morck, Randall and Yeung Bernard (2003), "Corporate Stability and Economic Growth", William Davidson Working Paper, 553. Hernando de Soto (2000), The Mystery of Capital: Why Capitalism Triumphs in the West and Fails Everywhere Else, Random House, New York. Holtz-Eakin, Douglas, and Amy Ellen Schwartz (1995), "Spatial Productivity Spillovers from Public Infrastructure: Evidence from State Highways", NBER Working Paper, W5004. (February). Joumard. Isabelle, C. Liedholm, and D. Mead (1992), "The Impact of Laws and Regulations on Micro and Small Enterprises in Niger and Swaziland", OECD Technical Paper Series, OECD, Paris Knack, Stephen, and Philip Keefer (1995), "Institutions and Economic Performance: Cross Country Tests Using Alternative Institutional Measures", Economics and Politics 7(3): 207­27. Levine, Renelt (1997), "Financial Development and Economic Growth: Views and Agenda", Journal of Economic Literature, 35: 688-726. Leinbenstein, H (1966), "Allocative Efficiency vs X-Efficiency", American Economic Review, 56(3): 580-606. Little, Ian (1987), "Small Manufacturing Enterprises in Developing Countries, The World Bank Economic Review, January Little, Ian, Dipak Mazumdar, and John Page (1987), Small Manufacturing Enterprises: a Comparative Analysis of India and other Economies, Oxford University Press. Loaya, N.V., A.M. Oviedo, and Luis Serven (2004), " Regulation and Macroeconomics Performance", World Bank. Washington, D.C.. Lucas Robert Jr. (1988), "On the mechanics of economic development," Journal of Monetary Economics, 22(1): 3-42 (July). Maddala, G. S. (1988), Introduction to Econometrics, McMillan Publishing Cie, New York. Mankiw, Gregory, David Romer and David Weil (1992), "A Contribution to the Empirics of Economic Growth", Quarterly Journal of Economics, 106: 407-37. Manly Bryan F.J. (1994), Multivariate Statistical Method: A Primer, 2nd ed. New York: Chapman&Hall. 29 Mardia, K. V., J. T. Kent, and J. M. Bibby (1997), Multivariate Analysis, 6th ed. New York: Academic Press. Marschak, Jacob, and Williams Andrews (1944), "Random Simultaneous Equation and the Theory of Production", Econometrica, 12 (3/4): 143-205. Mauro, Paolo (1995), "Corruption and Growth", Quarterly Journal of Economics, 110: 681­712. McMillan, John (1998), "Managing Economic Change: Lessons from New Zealand", The World Economy, 21: 827-43. McMillan, John (2004), "A Flexible Economy? Entrepreneurship and Productivity in New Zealand ", Working Paper, Graduate School of Business, Stanford University, Stanford, CA. Mitra, Arup, Aristomène Varoudakis and Marie-Ange Véganzonès (2002), "Productivity and Technical Efficiency in Indian States' Manufacturing: The Role of Infrastructures", Economic Development and Cultural Change, 50 (2): 395-426. Mo, Park Hung (2001), "Corruption and Economic Growth", Journal of Comparative Economics, 29(1): 66-79. Morocco (2001), Moroccan Manufacturing Sector at the Turn of the Century. Result of the Firm Analysis and Competitiveness Survey, Pilot Investment Climate Assessment, International Finance Corporation, The World Bank, Washington, D.C. Morocco (2005), Evaluation du Climat d'Investissement au Maroc, Middle East and North Africa Region, Private Sector, Financial Sector and Infrastructure Group, The World Bank, Washington, D.C. Murphy, Kevin M., Andrei Shleifer, and Robert W. Vishny (1989), "Industrialization and the Big Push", The Journal of Political Economy, 97 (5): 1003-1026 (October). Nabli, Mustapha K (2007), Breaking the Barriers to Higher Economic Growth, World Bank, Washington, DC. Nabli, Mustapha.K., and Marie-Ange Véganzonès-Varoudakis (2004), " How Does Exchange Rate Policy Affect Manufactured Exports in the Middle East and North Africa", Applied Economics , 36 (19): 2209-2220. Nabli, Mustapha.K., and Marie-Ange Véganzonès-Varoudakis (2007), "Reforms Complementarities and Economic Growth in the Middle East and North Africa", Journal of International Development, 19: 17-54. Nagaraj, Rayaprolu, Aristomène Varoudakis and Marie-Ange Véganzonès (2000), "Long- Run Growth Trends and Convergence Across Indian States: The Role of Infrastructures", Journal of International Development, 12 (1): 45-70 30 North, Douglas C. (1990), Institutions, Institutional Change, and Economic Performance, Cambridge, U.K.: Cambridge University Press. Organization for Economic Cooperation and Development (2001), Businesses' Views on Red Tape, Paris, OECD. http:/ www.oecd.org/publication/e-book/4201101E.PDF. Psacharopoulos, George (1988), "Education and Development: A Review", World Bank Research Observer, Oxford University Press, 3(1): 99-116. Rodrik, Dani (1999), "Institutions for High-Quality Growth: What They Are and How to Acquire them", Paper presented at the IMF Conference on Second Generation Reforms, Washington, November 8-9. Rodrik, Dani, Arvind Subramanian, and Francesco Trebbi. 2002. "Institutions Rule: The Primacy of Institutions over Geography and Integration in Economic Development." NBER Working Paper 9305, National Bureau of Economic Research, Cambridge, Massachusetts. Rodrik, Dani, and Arvind Subramanian (2004), "From Hindu Growth to Productivity Surge: The Mystery of the Indian Growth Transition", Harvard University. Cambridge, Mass. Processed. Saleh, Jahangir (2004), "Property Rights Institutions and Investment", The World Bank Policy Research Working Paper Series, 3311, The World Bank, Washington D.C. Sekkat, Kalid, and Marie-Ange Véganzonès-Varoudakis (2007), "Openness, Investment Climate, and FDI in Developing Countries", Review of Development Economics, 11(4): 607-620. Simon, H. (1955), "A Behavioral Model of Rational Choice", Quarterly Journal of Economics, 69(1): 99-118 (February). Tanzi, Vito, and Hamid Davooli (1997), "Corruption, Public Investment, and Growth", IMF Working Papers, 97/139, International Monetary Fund, Washington, D.C. World Bank (2002), World Development Report 2002: Building Institutions for Market, Oxford University Press, New York. World Bank (2004), Better Governance for Development in the Middle East and North Africa: Enhancing Inclusiveness and Accountability, MENA Development Report, The World Bank, Washington, D.C. World Bank (2005), World Development Report 2005: A Better Investment Climate for Everyone, World Bank and Oxford University Press, Washington, D.C. World Bank (2009), "Doing Business", World Bank, Washington, D.C. 31 Annex 1: List of countries MENA* LAC AFR SAS EAP Algeria (2002) Brazil (2003) Ethiopia (2002) Bangladesh (2002) China (2002) Egypt (2004/2006) Ecuador (2003) South Africa (2003) India (2000/2002) Philippines (2003) Morocco (2000/2004) El Salvador (2003) Tanzania (2003) Pakistan (2002) Thailand (2004) Lebanon (2006) Guatemala (2003) Zambia (2002) Sri Lanka (2004) Saudi Arabia(2005) Honduras (2003) Nicaragua (2003) MENA : Middle East and North Africa; LAC: Latin America and the Caribbean; AFR : Sub Sahara Africa; SAS: South Asia; EAS : East Asia. Annex 2a: ICA Surveys: Data Limitations Metal Chemical Non Industries/ Agro- & & Metal & Wood Total (number of firms Textile Garment Leather Processing Machinery Pharmac. Plastic & and %) Products Products Materials Furniture Total Enterprises 2496 3794 821 2815 2163 1728 1159 1317 16293 MENA Enterprises 761 906 257 655 758 364 487 199 4387 (% total) (30%) (24%) (31%) (23%) (35%) (21%) (42%) (15%) (27%) Total Frontier 1998 2796 634 2184 1604 1270 897 1031 12414 (% total enterprises) (80%) (74%) (77%) (78%) (74%) (73%) (77%) (78%) (76%) MENA Frontier 541 711 167 436 538 241 335 120 3073 (% total MENA) (69%) (78%) (65%) (67%) (71%) (66%) (69%) (59%) (70%) (% total frontier) (26%) (25%) (26%) (20%) (34%) (19%) (37%) (11%) (25%) Total with 942 1604 380 1525 841 738 478 778 5002 IC variables (38%) (42%) (46%) (54%) (39%) (43%) (41%) (59%) (45%) MENA with 215 371 91 228 258 95 162 63 1483 IC variables (% total MENA) (28%) (41%) (35%) (35%) (34%) (26%) (33%) (32%) (34%) (23%) (24%) (15%) (31%) (13%) (34%) (8%) (30%) (% total IC) (23%) Sources: Authors' calculations. 32 Annex 2b: ICA Surveys: Data limitations Industry Countries Total Frontier IC variables (number of firms and %) Textile Algeria (2002) 79 27 (34%) Egypt (2004) 141 92 66 (65%) (47%) Morocco (2000) 200 142 (71%) Lebanon (2006) 11 5 5 (45%) (45%) Morocco (2004) 160 148 144 (93%) (90%) Egypt (2006) 170 111 (67%) Leather Algeria (2002) 14 Egypt (2004) 44 29 19 (66%) (43%) Morocco (2000) 68 36 (53%) Lebanon (2006) 15 Morocco (2004) 80 77 72 (96%) (90%) 25 Egypt (2006) 36 (69%) Garments Egypt (2004) 120 87 52 (73%) (43%) Morocco (2000) 316 216 (68%) Lebanon (2006) 27 13 13 (48%) (48%) Morocco (2004) 334 315 314 (94%) (94%) Egypt (2006) 109 83 (76%) Agro Algeria (2002 51 27 Processing (53%) Egypt (2004) 156 115 90 (74%) (58%) Morocco (2000) 83 44 (53%) Lebanon (2006) 49 16 14 (33%) (29%) Morocco (2004) 72 60 58 (83%) (81%) Saudi Arabia (2005) 94 75 66 (80%) (70%) Egypt (2006) 150 107 (71%) 33 Metal & Algeria (2002) 110 47 Machinery (43%) Products Egypt (2004) 168 119 88 (71%) (52%) Morocco (2000 38 27 (71%) Lebanon (2006) 7 Morocco (2004) 19 19 19 (100%) (100%) Saudi Arabia (2005) 185 163 136 (88%) (74%) Egypt (2006) 231 163 (71%) Chemical & Algeria (2002 52 25 Pharm. (48%) Products Egypt (2004 65 52 41 (80%) (63%) Morocco (2000 77 44 (57%) Lebanon (2006) 6 Morocco (2004) 61 56 54 (92%) (89%) Egypt (2006) 103 64 (62%) Wood & Egypt (2004) 58 31 19 Furniture (53%) (33%) Lebanon (2006) 37 11 11 (30%) (30%) Morocco (2004) 3 Saudi Arabia (2005) 51 37 33 (73%) (65%) Egypt (2006) 50 38 (78%) Non Metal Algeria (2002) 85 41 & Plastic (48%) Materials Egypt (2004) 169 126 93 (75%) (55%) Morocco (2000) 77 48 (62%) Lebanon (2006) 7 Morocco (2004) 77 69 69 (90%) (90%) Egypt (2006) 72 51 (71%) Source.:Authors' calculations 34 Annex 3: Labor Productivity (LP) and Labor Cost (1000 dollars) Wage per Wage per Labor Labor Unit Labor Unit Labor Number of Country capita capita Productivity Productivity Cost Cost firms (average) (median) (average) (median) (average) (median) Textile MENA 2.43 2.20 7.84 4.91 0.54 0.42 272 Algeria (2002) 1.64 1.66 4.27 3.58 0.57 0.44 27 Egypt (2004) 0.71 0.65 4.78 1.93 0.45 0.31 92 Egypt (2006) 0.84 0.82 3.47 1.87 0.50 0.38 127 Lebanon (2006) 0.64 0.66 1.83 1.49 0.44 0.33 5 Morocco (2000) 3.46 2.98 11.74 7.42 0.42 0.38 142 Morocco (2004) 3.71 3.31 10.61 7.11 0.60 0.49 148 Non MENA 1.90 1.34 9.99 5.47 0.39 0.27 1256 China (2002) 3.76 1.45 11.35 6.90 0.35 0.24 39 India (2000) 1.50 1.07 10.64 5.15 0.28 0.22 216 India (2002) 1.58 0.94 10.48 4.66 0.42 0.19 195 Leather MENA 2.50 2.10 5.41 3.74 0.64 0.56 106 Egypt (2004) 0.71 0.51 3.49 1.18 0.51 0.49 29 Egypt (2006) 0.74 0.59 1.40 0.87 1.92 0.64 25 Morocco (2000) 2.69 2.51 5.91 5.50 0.61 0.47 36 Morocco (2004) 3.18 2.70 6.13 4.70 0.69 0.59 77 Non MENA 1.70 1.38 6.80 4.03 0.47 0.35 467 Chine (2002) 1.68 1.27 8.04 4.05 0.38 0.31 53 India (2002) 1.19 0.74 7.55 3.89 0.34 0.20 57 Garments MENA 2.24 2.06 5.31 3.43 0.67 0.57 415 Egypt (2004) 0.68 0.59 1.78 1.10 0.79 0.50 87 Egypt (2006) 0.70 0.60 2.27 1.11 0.61 0.50 83 Lebanon (2006) 0.42 0.40 2.20 1.27 0.90 0.35 13 Morocco (2000) 2.53 2.25 5.28 4.28 0.56 0.55 216 Morocco (2004) 2.74 2.54 6.42 4.17 0.63 0.60 315 Non MENA 1.96 1.36 6.60 3.33 0.73 0.42 1903 Chine (2002) 2.86 1.12 12.35 3.50 0.46 0.36 93 India (2000) 1.34 0.84 7.17 3.74 0.30 0.25 186 India (2002) 1.10 0.89 7.38 4.11 0.36 0.23 206 Agro-Processing MENA 3.77 3.04 15.87 10.01 0.46 0.32 293 Algeria (2002) 2.39 1.99 6.71 4.93 0.53 0.35 27 S. Arabia (2005) 6.78 6.21 27.93 18.42 0.37 0.36 75 Egypt (2004) 0.92 0.62 4.87 2.22 0.40 0.31 115 Egypt (2006) 3.11 0.73 9.28 2.91 0.38 0.29 107 Lebanon (2006) 0.46 0.45 3.01 1.96 0.29 0.25 16 Morocco (2000) 6.16 3.36 24.27 20.23 0.34 0.24 44 Morocco (2004) 6.99 4.87 29.43 18.87 0.72 0.30 60 Non MENA 2.50 1.67 14.91 6.40 0.46 0.27 1751 India (2002) 2.20 0.86 21.10 4.94 0.44 0.17 167 35 Annex 3: Labor Productivity (LP) and Labor Cost (1000 dollars) (end) Wage per Wage per Labor Labor Unit Labor Unit Labor Number of Country capita capita Productivity Productivity Cost Cost firms (average) (median) (average) (median) (average) (median) Metal & Machinery Products MENA 5.98 4.10 19.92 10.93 0.50 0.41 348 Algeria (2002) 2.91 2.51 5.80 3.74 0.80 0.59 47 S. Arabia (2005) 7.01 6.34 26.15 18.32 0.39 0.39 163 Egypt (2004) 5.18 0.79 13.66 2.24 0.55 0.36 119 Egypt (2006) 1.13 0.87 7.23 3.24 0.50 0.30 181 Morocco (2000) 5.02 3.97 18.74 9.49 0.57 0.38 27 Morocco (2004) 9.74 9.58 40.64 19.81 0.44 0.50 19 Non MENA 4.49 3.29 16.44 8.62 0.46 0.36 999 Chine (2002) 2.29 1.38 11.96 6.11 0.30 0.25 150 India (2000) 1.58 1.14 9.53 5.48 0.26 0.24 68 India (2002) 2.17 0.96 16.02 4.45 0.33 0.23 140 Chemical & Pharmaceutical Products MENA 5.24 4.10 21.38 12.96 0.47 0.35 133 Algeria (2002) 2.59 2.32 6.97 5.27 0.66 0.40 25 Egypt (2004) 1.40 0.89 9.26 3.16 0.44 0.36 52 Egypt (2006) 1.75 0.98 8.34 3.04 0.43 0.33 68 Morocco (2000) 6.75 4.51 25.84 17.65 0.34 0.27 44 Morocco (2004) 9.98 7.87 39.06 25.49 0.40 0.31 56 Non MENA 3.21 2.18 18.95 9.24 0.37 0.23 821 India (2000) 5.00 1.18 16.72 6.65 0.23 0.19 208 India (2002) 1.86 0.89 12.76 4.72 0.29 0.18 331 Wood & Furniture MENA 3.40 3.01 8.78 7.65 0.58 0.51 81 S. Arabia (2005) 6.32 5.67 16.86 14.97 0.40 0.36 37 Egypt (2004) 1.07 0.82 2.19 1.71 0.69 0.65 31 Egypt (2006) 0.98 0.87 3.90 1.58 0.95 0.60 39 Lebanon (2006) 0.66 0.64 1.48 1.00 0.81 0.60 13 HORS MENA 2.67 2.23 7.54 4.77 0.58 0.45 914 Non Metal & Plastic Materials MENA 2.16 1.91 8.13 4.77 0.44 0.39 237 Algeria (2002) 2.68 2.50 5.77 4.82 0.69 0.60 41 Egypt (2004) 0.75 0.61 5.18 1.66 0.37 0.34 126 Egypt (2006) 0.82 0.69 9.87 1.97 0.57 0.30 53 Morocco (2000) 3.53 3.14 11.05 8.80 0.53 0.37 48 Morocco (2004) 4.38 3.91 14.83 10.35 0.43 0.36 70 Non MENA 2.95 2.40 11.19 6.92 0.54 0.36 569 Source: Authors' calculations 36 Annex 4: Total Factor Productivity (TPF) Textile Garments Leather Agro-Processing Non MENA Average 3.85 5.08 4.42 4.75 Median 2.88 4.02 3.94 3.51 MENA Average 3.91 5.35 4.13 4.31 Median 2.87 3.97 3.20 3.09 Efficient Country Brazil (2003) Brazil (2003) Brazil (2003) Brazil (2003) Average 6.06 6.82 6.20 7.37 Median 4.71 5.14 5.71 5.61 Algeria (2002) Average 3.65 3.75 Median 3.06 2.46 Saudi A. (2005) Average 5.30 Median 3.94 Egypt (2004) Average 3.09 2.90 3.20 3.61 Median 1.94 2.04 2.06 2.17 Egypt (2006) Average 2.33 2.36 2.00 2.98 Median 1.76 1.93 1.71 2.32 Lebanon (2006) Average 3.40 6.03 2.28 Median 1.66 2.29 2.24 Morocco (2000) Average 4.57 5.75 4.89 6.27 Median 3.77 4.69 4.60 4.44 Morocco (2004) Average 4.49 5.99 4.48 5.20 Median 3.45 4.57 3.63 4.32 China (2002) Average 5.48 4.42 3.34 Median 2.79 3.32 3.31 India (2000) Average 4.38 4.79 Median 3.16 3.70 India (2002) Average 4.07 4.15 3.65 4.16 Median 2.78 2.88 3.47 3.02 37 Annex 4: Total Factor Productivity (TPF) (end) Metal & Chemical & Non Metal Machinery Pharmac. Wood & & Plastic Products Products Furniture Materials Non MENA Average 5.55 4.55 4.95 4.56 Median 4.42 3.65 4.22 3.70 MENA Average 4 .97 4.77 3.78 3.87 Median 3.56 3.66 3.48 3.03 Efficient Country Brazil (2003) Brazil (2003) South Afr (2003) South Afr (2003) Average 7.70 7.86 6.98 5.77 Median 6.45 6.27 6.12 5.08 Algeria (2002) Average 4.79 4.71 5.04 Median 3.81 4.12 3.87 Egypt (2004) Average 3.27 2.93 2.75 3.02 Median 2.18 2.08 2.20 2.20 Egypt (2006) Average 3.69 3.15 3.18 2.44 Median 2.21 2.16 1.92 1.94 Morocco (2000) Average 5.98 7.19 3.95 Median 4.49 5.67 3.60 Morocco (2004) Average 7.70 6.50 4.72 Median 4.54 4.93 4.04 Saudi A. (2005) Average 5.94 5.18 Median 4.37 4.99 Lebanon (2006) Average 2.25 Median 2.26 China (2002) Average 3.86 Median 2.89 India (2000) Average 4.54 4.42 Median 3.77 3.63 India (2002) Average 5.09 3.77 Median 3.27 3.12 38 Annex 5: Technical Efficiency calculated from a Stochastic Frontier Textile Garment Leather Agro-Processing Non MENA Average 0.443 0.621 0.639 0.445 Median 0.436 0.623 0.642 0.440 MENA Average 0.416 0.658 0.576 0.426 Median 0.417 0.659 0.581 0.426 Efficient Country Brazil (2003) South Afr (2003) Brazil (2003) South Afr (2003) Average 0.985 0.988 0.977 0.978 Median 0.985 0.988 0.977 0.978 Algeria (2002) Average 0.327 0.347 Median 0.337 0.346 Saudi Arabia (2005) Average 0.708 Median 0.717 Egypt (2004) Average 0.206 0.204 0.295 0.165 Median 0.184 0.205 0.280 0.152 Egypt (2006) Average 0.164 0.220 0.150 0.212 Median 0.154 0.215 0.124 0.189 Lebanon (2006) Average 0.206 0.227 0.155 Median 0.165 0.216 0.157 Morocco (2000) Average 0.663 0.790 0.741 0.698 Median 0.668 0.792 0.754 0.724 Morocco (2004) Average 0.571 0.802 0.681 0.684 Median 0.585 0.802 0.695 0.697 China (2002) Average 0.455 0.503 0.442 Median 0.450 0.497 0.444 India (2000) Average 0.467 0.653 Median 0.462 0.661 India (2002) Average 0.410 0.650 0.551 0.397 Median 0.402 0.653 0.578 0.373 39 Annex 5: Technical Efficiency calculated from a Stochastic Frontier (end) Metal & Chemical & Non Metal Machinery Pharmac. Wood & & Plastic Products Products Furniture Materials Non MENA Average 0.617 0.428 0.483 0.618 Median 0.623 0.421 0.481 0.633 MENA Average 0.525 0.434 0.455 0.520 Median 0.526 0.430 0.4653 0.516 Efficient Country Morocco (2004) Brazil (2003) South Africa (2003) South Africa (2003) Average 0.973 0.984 0.977 0.971 Median 0.973 0.984 0.977 0.971 Algeria (2002) Average 0.378 0.372 0.525 Median 0.395 0.363 0.563 Saudi Arabia (2005) Average 0.740 0.792 Median 0.758 0.804 Egypt (2004) Average 0.217 0.168 0.189 0.313 Median 0.190 0.149 0.174 0.292 Egypt (2006) Average 0.242 0.139 0.192 0.228 Median 0.217 0.131 0.149 0.214 Lebanon (2006) Average 0.131 Median 0.118 Morocco (2000) Average 0.657 0.820 0.682 Median 0.659 0.816 0.711 Morocco (2004) Average 0.973 0.709 0.890 Median 0.973 0.721 0.890 China (2002) Average 0.337 Median 0.337 India (2000) Average 0.441 0.333 Median 0.452 0.330 India (2002) Average 0.449 0.312 Median 0.439 0.304 40 Annex 6: Sperman Correlation Coefficient of the Three Measures of Firm-Level Productivity Textiles Leather Nobs: 1998 Nobs: 634 TE TFP LP TE TFP LP TE 1 TE 1 TFP 0.7077* 1 TFP 0.7703* 1 LP 0.7615* 0.6012* 1 LP 0.6427* 0.6756* 1 Garments Agro-Processing Nobs: 2796 Nobs: 2184 TE TFP LP TE TFP LP TE 1 TE 1 TFP 0.5571* 1 TFP 0.7047* 1 LP 0.5675* 0.6370* 1 LP 0.7814* 0.5861* 1 Metals & Machinery Products Chemicals & Pharmaceutic Products Nobs: 1604 Nobs: 1270 TE TFP LP TE TFP LP TE 1 TE 1 TFP 0.7483* 1 TFP 0.7349* 1 LP 0.7762* 0.6810* 1 LP 0.7542* 0.6270* 1 Wood & Furniture Non-Metallic & Plastic Materials Nobs: 1031 Nobs: 901 TE TFP LP TE TFP LP TE 1 TE 1 TFP 0.8456* 1 TFP 0.7394* 1 P 0.8885* 0.7532* 1 LP 0.8028* 0.6293* 1 Note : *: significant at 1%,. TE : Technical Efficiency, TFP : Total Factor Productivity, LP : Labor Productivity. Source : Auhtors' calculations 41 Annex 7: ICA Surveys: Table 1 Investment Climate and Plant Characteristics (a) Ho: No NON diff in MENA MENA means Standard Number Standard Number [p- Mean Deviation of firms Mean Deviation of firms values] Size 127.1 266.9 3075 192.4 555.9 9350 0.0 Export (% sales) 16.8 34.1 2987 18.7 35.0 8815 0.0 Foreign ownership (% K) 8.3 25.4 3072 6.2 21.7 9292 0.0 Use of E-mail (% firms) 52.0 50.0 2289 60.5 48.9 8940 0.0 Use of website (% firms) 26.7 44.2 2550 35.6 47.9 8233 0.0 Telecommunication* 4.7 21.2 2493 11.4 31.8 8635 0.0 Electricity* 18.2 38.6 2512 33.2 47.1 8650 0.0 Transport* 7.6 26.5 2332 15.1 35.8 8634 0.0 % firm with generator 44.9 41.8 3040 35.5 48.6 9332 0.0 % elect from generator 15.3 16.6 2999 6 18.7 9110 0.0 Overdraft facility (% firms) 42.6 49.5 3069 56.4 49.6 8519 0.0 Financing Access* 51.5 50.0 2032 34.7 47.6 8492 0.0 Financing Cost* 56.9 49.5 2051 42.0 49.4 8477 0.0 Top Manager Ed. Level 3.9 1.4 2261 4.3 1.5 8083 0.0 Top Manager Exp. (years) 12.5 10.9 2218 8.0 9.0 8260 0.0 % Workers Formal Training 19.8 39.9 3052 39.8 49.0 9248 0.0 Availability Skilled Workers* 30.1 45.9 2505 24.0 42.7 8625 0.0 0.0 0.0 0.0 0.0 Labor Regulation* 26.9 44.3 2505 21.8 41.3 8430 0.0 Tax Rate* 57.0 49.5 2493 41.8 49.3 8628 0.0 Tax Administration* 38.5 48.7 2486 34.8 47.6 8618 0.0 Licence/Operating Permits* 20.8 40.6 2486 15.5 36.2 8408 0.0 Customs/Trade Regulations* 18.4 38.7 2448 24.9 43.2 7844 0.0 Corruption* 40.6 49.1 2489 44.6 49.7 8635 0.0 * Percentage of firms ranking the variable as a major or severe constraint Source: Authors calculations 42 Table 2: Number of firms/ Rank N Textile Leather Garment Agro-Ind Metal Chemical Wood M Algeria (2002) 79 14 51 110 52 2 20% 4% 13% 28% 13% Egypt (2004) 4 141 3 44 3 120 156 3 168 3 65 1 58 1 15% 5% 13% 17% 18% 7% 6% Morocco (2000) 1 200 2 68 2 316 83 38 2 77 3 23% 8% 37% 10% 4% 9% Lebanon (2006) 11 15 27 49 7 6 37 7% 9% 17% 31% 4% 4% 23% Morocco (2004) 2 160 1 80 1 334 72 19 4 61 3 3 20% 10% 41% 9% 2% 8% 0% Egypt (2006) 3 170 4 36 4 109 150 1 231 1 103 2 50 4 18% 4% 12% 16% 25% 11% 5% Saudi Arabia (2005) 94 2 185 2 51 28% 56% 15% 43 Annex 8: Investment Climate and Plant Characteristics (b) Chart 1: Composite IC Indicators (MENA/ non-MENA) 0.70 m_infrast m_humain 0.60 m_business m_finance 0.50 0.40 0.30 0.20 0.10 0.00 -0.10 non MENA MENA Chart 2: Composite IC Indicators (regions) 1.2 m_infrast 1.0 m_humain m_business 0.8 m_finance 0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 SA EAP LAC MENA AF Chart 3: Composite IC Indicators (countries) Infra 2.60 Human Bus-Gov 2.10 Fin 1.60 1.10 0.60 0.10 -0.40 -0.90 -1.40 SouthAfrica Morocco India Egypt Lebanon SaudiArabia 44 Chart 4: Infrastructures (obstales) 3.5 3.0 tel_obs elect_obs transp_obs 2.5 2.0 1.5 1.0 0.5 0.0 SouthAfrica Morocco India Egypt SaudiArabia Lebanon Chart 5 Infrastructures (electricity) 1.2 50.0 generator %el_gen 45.0 1.0 40.0 35.0 0.8 30.0 0.6 25.0 20.0 0.4 15.0 10.0 0.2 5.0 0.0 0.0 SouthAfrica Morocco India Egypt SaudiArabia Lebanon Chart 6: Infrastructures (Internet) 2.0 e-mail internet 1.8 1.6 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0 SouthAfrica Morocco India Egypt SaudiArabia Lebanon 45 Chart 7: Human Capacity 20.0 ed_manag exp_manag 18.0 training skill_obs 16.0 14.0 12.0 10.0 8.0 6.0 4.0 2.0 0.0 SouthAfrica Morocco India Egypt SaudiArabia Lebanon Chart 8: Government-Business Relation 3.5 l_obs tax_rate_obs tax_adm_obs 3.0 cust_obs corr_obs lic_obst 2.5 2.0 1.5 1.0 0.5 0.0 SouthAfrica Morocco India Egypt SaudiArabia Lebanon Chart 9: Financing Constraints 3.5 overdraft fin_acc_obs fin_cost_obs 3.0 2.5 2.0 1.5 1.0 0.5 0.0 SouthAfrica Morocco India Egypt SaudiArabia Lebanon 46 Chart 10: Firms Characteristics 30.0 size foreign X 25.0 20.0 15.0 10.0 5.0 0.0 SouthAfrica Morocco India Egypt SaudiArabia Lebanon Note: All variables "obstacles" are averages of dummies going from 0 (none) to 4 (severe); generator, training, overdraft are averages of dummies 0 or 1 (0 for No and 1 for Yes); e-mail and internet are average of dummies 1 and 2 (1 for Yes and 2 for No); ed_manag is an average of dummies going from 1 to 6; %el_gen is the percentage of electricity coming from a generator; Infra , Gov_Bus, and Fin can be read as obstacles; Human can be read as a capacity; foreign is the percentage of firm's capital own by foreigners; X is the percentage of firm's sales exported; size is a variable calculated from the number of permanent workers. Source: Authors calculations 47 Notes 1 See Bosworth and Collins (2003); Djankov and al. (2002); Dollar and al (2005); Hall and Jones (1999); Haltiwanger (2002); He et al. (2003); Loaya, Ociedo and Serven (2004); OECD (2001); Rodrik, Subramanian (2004); McMillan (1998 and 2004); World Bank (2003, 2004). 2 See in particular Frankel (2002) and Rodrik (1999). 3 See for example, Acemoglu, Johnson, and Robinson (2001); Easterly and Levine (2003); Hall and Jones (1999); Knack and Keefer (1995); Rodrik, Subramanian, and Trebbi (2002);. 4 See Easterly and Levine (2003); Knack and Keefer (1995); North (1990); Rodrik, Subramanian and Trebbi (2002); and Saleh (2004). See Acemoglu, Johnson and Robinson (2001); Calderon and Chong (2000) in the context of growth. 5 See Mauro (1995); Gupta, Davooli and Alonso-Terme (2002); Mo (2001); Tanzi and Davooli (1997). 6 See Kerr (2002); Hernando and Soto (2000). 7 See Evans and Rauch (2000). 8 See Bastos and Nasir (2004); Dollar and al. (2005); Eifert and al. (2005); Escribano and Gasch (2005). 9 Some important limitations can be found in the cost and access to financial services (this is the case of Morocco, and to a lower extend of Egypt and Algeria) in the tax system (for example in Egypt, but also Algeria and Morocco), as well as in the regulatory environment (for example in Saudi Arabia,) 10 See see El Badawi (2002); the World Bank (2004a); Aysan et al. (2007a). 11 See Nabli. (2007); Nabli and Véganzonès-Varoudakis (2004); Aysan, et al. (2007a and b). 12 See Sekkat and Véganzonès-Varoudakis, (2007); Nabli and Véganzonès­Varoudakis (2007). 13 Measuring productivity in level, although more restrictive than measuring growth rates (it requires for example specific functional forms of the production function) is less demanding in terms of data quality requirement. It allows, in particular, unbalanced panels with short term dimension, measurement errors, or constant value of IC variables (see Escribano and Guasch, 2005). 14 See Marschak and Andrews, 1944; Griliches and Mairesse, 1995 15 Some countries benefit from two surveys. This is the case of Egypt (2004, 2006), India (2000, 2002) and Morocco (2000, 2004). 16 The year of the survey is into brackets. Lebanon and Saudi Arabia, however, are less represented than the other countries of the region. In the case of Lebanon, the low number of observations makes sometimes results difficult to interpret. For Saudi Arabia, firms' surveys cover only 3 of the 8 branches studied (Agro- Processing, Wood & Furniture, Metal & Machinery). 17 Some inconsistencies have been seen, for example, when the hypotheses of constant returns to scale in TFP calculations led to a negative contribution of capital, or when the residual of estimation of the production frontiers was not in line with the standard deviations of the regressions and influenced too much the estimation of coefficients (we used DFFITS detection method in that case, see Maddala, 1988). For some countries, this has sometimes led to eliminate a whole sector. This has been the case when the number of enterprises left concluded to be smaller than 10 or when more than 70 per cent of the firms had been eliminated from the initial population (we accepted to make an exception for Lebanon). 48 18 For MENA, the loss of information fluctuates from 22% in Garment to 41% in Wood & Furniture (around 30% in Metal & Machinery Products, Non Metal & Plastic Materials and Textile, and 35% in Leather, Agro-Processing and Chemical & Pharmaceutical Products). This loss is of 20% to 25% for the whole sample of countries, what is lower than for MENA. This means that answers in MENA were, in average, less satisfactory than in the other countries of the sample. As for the contribution of MENA to the whole sample, when estimating the production frontiers, it varies from 11% in Wood & Furniture to 37% in Non Metal & Plastic Materials (25% in average for the whole manufacturing industry, see Table 1-a.), what is a bit less than, but consistent with, the contribution of MENA to the initial sample. 19 This percentage is of 45 in the whole sample, what confirms that firms in MENA did not answer the questionnaire as accurately as the rest of the sample. This is the case in all industries, but more particularly in Agro-Processing, Chemicals & Pharmaceutical Products and Wood & Furniture, in which almost 20% less enterprises have given correct IC information. 20 Firms are asked to evaluate their constraints on a scale going from none to very severe. 21 We ensure to get a sufficient number of observations by city and sector. . 22 The choice of an adequate exchange rate depends, among other things, on the exchange rate regime of the country. In presence of a floating exchange rate regime, the volatility of the current exchange rate may affect the perception of the productive performances. This is particularly true for the Labor Productivity (LP). For Total Factor Productivity (TFP), this problem is somewhat attenuated by the fact that the same exchange rate is used to convert intermediate consumptions and capital in the denominator, and production in the numerator. Using current exchange rate introduces, as well, a bias for example when fixed exchange rate policy leads to an overvaluation of the currency or when the floating rate suffers from overshooting. Current exchange rate has the advantage to represent the rate that firms deal with when making their own economic calculations. This is the rate that the producer faces when he competes on external as well as domestic markets. Both, a constant exchange rate or the use of a Purchasing Power Parity (PPP) exchange rate with the US dollar, are surely more problematic for our analysis. PPP conversion rate is useful when comparing purchase power of income per capita. We know that the purchasing power in developing countries tends to be higher than when GDP per capita is converted using nominal exchange rate. But when dealing with production, current rate is more representative of the enterprises' economic reality. The choice of exchange rate does not seem to change radically the perception of the firms' productive performances. The coefficient of correlation of our three measures of firm-level productivity using alternatively current and constant exchange rates is relatively high. 23 Some countries benefit from two surveys, namely Egypt (2004 and 2006), India (2000 and 2002) as well as Morocco (2000 and 2004, see Table 1 in Annex 1). 24 The variable Direct Raw Material Costs is not available from the surveys. 25 Our analysis is based on firm-level productivity average and median. Generally, averages have been found higher than medians (30 % higher in some cases). We are in presence of an unsymmetrical distribution, where a small number of high performing firms increase the average productivity. In this context, the median is more representative of the typology of the firms. The median has also the advantage to be more stable when the size of the sample is changing. The average, however, summarizes well the productive performances of all the firms of the sample. It is these averages that have been used, when calculating the percentages in Table 2 and 3. Medians are, however, also displayed in Table 3, Annex 3. 26 It can be noted that firms in Saudi Arabia seem to perform very well in the sectors covered by the survey (Agro-Processing, Metal & Machinery Products, and Wood & Furniture). This result will be explained in next section. 49 27 Interpretation of results is, however, more difficult for some countries. This is the case of Lebanon, for which the number of observations is too small (5 for Textile and 16 for Agro-Processing) to reach a reliable conclusion. The combination of two surveys for Morocco and Egypt allows more than one hundred observations by branch. Morocco, for example, benefits from 500 enterprises in Garments. In Saudi Arabia, firms present quite good productive performances, although most of the branches suffer also from a relative small number of observations. In Wood & Furniture, firm-level TFP is one of the highest of the sample. This result confirms the conclusion reached for Productivity of Labor.). 28 See Manly (1994); Mardia, Ken and Bibby, (1997); Nagaraj and al. (2000); Mitra and al. (2002); Nabli and Véganzonès-Varoudakis (2007); Aysan and al. (2007a and b). 29 Obstacles' value goes from none (0) to very severe (3). 30 See Aschauer (1989), Argimon et al., (1997), Barro (1990), Blejer and Kahn (1984), Murphy, Shleifer, and Vishny (1989). 31 For spatial externalities, see Holtz-Eakin and Schwartz (1995). 32 Education level goes from primary to post graduate 33 See Lucas (1988), Psacharopoulos (1988), and Mankiw, Romer and Weil (1992). 34 The principal components of the initial variables were extracted for each aggregated indicators. The four composite indicators were then constructed as the weighted sum of two or three principal components, depending of the explanatory power of each component. We chose the most significant principal components whose eighenvalues were higher than one. In this case, we explain around 70 percent of the variance of the underlying individual indicators. The weight attributed to each principal component corresponds to its relative contribution to the variance of the initial indicators (calculated from the cumulative R²). The contribution of each individual indicator to the composite indicator can then be computed as a linear combination of the weights associated with the two or three principal components and of the loadings of the individual indicators on each principal component. For more details on the aggregation method using Principal Component Analysis (PCA) see Nagaraj, Varoudakis, Véganzonès (2000), and Mitra, Varoudakis, Véganzonès (2002). 35 See in particular the World Bank Investment Climate Assessments (ICA) of Egypt (2005 and 2006), Morocco (2001 and 2005), and Algeria (2002). Doing Business 2005-2009 ranks as well MENA particularly low in reforms regarding the labor market, getting credit, enforcing contracts, construction permits, starting a business, closing a business and protection of investors (see the World Bank, 2009). Nabli (2007) also stresses MENA above average licenses, domestic taxation, import duties, regulatory and administrative barriers to firms start up and operations, opaque bidding procedures and official acceptance of uncompetitive practices, unpredictable judicial systems that do not facilitate the restructuring of viable business or the closure of nonviable ones, as well as weaknesses in infrastructure and financial system. With public bank dominating the banking system in many countries and favoring state enterprises, large industrial firms and offshore enterprises, small and medium firms in particular find it difficult to get the startup and operating capital they need. 36 We will also stress a contraction in the answers to the survey which justifies taking into consideration quantitative indicators, to complete the information given by qualitative ones. In the case of electricity, although more firms than in other regions seem to rely on a generator, comparatively less declare electricity as a strong constraint for operating. 37 We will recall that the Value Added is calculated as the difference between "Total Sales" and "Total Purchase of Raw Material -- excluding fuel". 50 38 The new literature on international trade associates firms' size with increasing returns to scale, market imperfections and product heterogeneity linked to technological innovation. The literature on corporate governance, as well, describes the difficulties in inciting and controlling big enterprises, although they are more able to reduce transaction costs and facilitate economic calculations. Small enterprises are described as less capitalistic and more flexible in a volatile environment, in particular in economies characterized by rigidities which encourage the development of the informal economy. 39 For two sectors: Chemicals & Pharmaceutical Products; and Wood & Furniture, coefficients of capital and labor are slightly smaller than in previous estimation (see Table 5). 40 For Infrastructure, the quality of the electrical network (RegElect) appears to increase firms' performances in Garment and Metal & Machinery Products. It is, however, the access to internet (RegWeb) which emerges as a factor of productivity in more industries (Textile, Agro-Processing, Chemical & Pharmaceutical Products and Wood & Furniture). As far as Human Capacity is concerned, level of education of top manager (EduM) is significant in almost all sectors (except Leather and Wood & Furniture), meanwhile number of years of expertise of manager (ExpM) and training of employees (Training) seem to play a role in only one sector each (Textile and Agro-Processing respectively). Same conclusions can be drawn for Financing Constraints, where access to credit line or overdraft facility (Cred) appear to generally stimulate productivity gains (except in Chemical & Pharmaceutical Products and Wood & Furniture), though the qualitative variable of access to financing (AccessF) is significant in only three sectors (Agro-Processing, Chemical & Pharmaceutical Products, and Wood & Furniture). 41 In these industries, a few factors seem to explain efficiency (only access to credit line (Cred) in the case of Leather and, internet access (RegWeb) and access to financing (AccessF) in the case of Wood & Furniture). On the opposite, Agro-Processing, Chemical & Pharmaceutical Products, Garment, and Textile display a broader set of factors explaining firms' productivity gains. 42 : respectively seven and six instead of two for Infrastructures and Business-Government Relations, four instead of three for Human Capacity, three instead of two for Financing Constraints. 43 Loss of information appears essentially for "Human Capacity" and "Infrastructure" for which one of the initial individual indicators was previously significant. 44 Besides, this model explains better Wood & Furniture. 45 Based on the number of observation of the regressions, big foreign enterprises constitute 30% of the sample in Leather, 24% in Agro-Processing, 23% in Metal & Machinery Products, 21% in Textile, 18% in Chemicals & Pharmaceutical Products, 17% in Garment, 14% in Non Metal & Plastic Materials and 13% in Wood & Furniture. 46 The overall explanatory power of the model is not very different for both samples. In Textile, Garment and Agro-Processing, firms' Technical Efficiency (TE) gap seems to be explained by more IC variables or plant characteristics for relatively small domestic enterprises, confirming that these industries are more sensitive to deficiencies in the investment climate. Opposite result is slightly observed in Wood & Furniture and Chemicals & Pharmaceutical Products. As for Leather and Metal & Machinery Products, these sectors are still poorly explained in both samples. 51