72719 POVERTY THE WORLD BANK REDUCTION AND ECONOMIC MANAGEMENT NETWORK (PREM) Economic Premise JUN 2012 SEPTEMBER 010 • Numbe 89 • Number 1 Quality of Government and Living Standards Francesco Grigoli and Eduardo Ley It is generally acknowledged that a government’s output is difficult to define and its value is hard to measure. The practical solution adopted by national accounts systems is to equate output value to input costs, but well-documented inefficiencies in government activities make this approximation questionable. One solution is to purge from gross domestic product (GDP) the fraction of government inputs that is wasted. This note illustrates such a correction, computing corrected per capita GDP on the basis of two studies that estimate efficiency scores for several dimensions of government activities. Results show that the correction could be significant and reorder the rankings of living standards. Despite its acknowledged shortcomings, GDP per capita is practice frontier. When this is not the case, there is an over- still the most commonly used summary indicator of living statement of national production. This in turn could result in standards. Much of the policy advice provided by interna- misleading conclusions, particularly in cross-country com- tional organizations is based on macroeconomic magnitudes parisons, given that the size, scope, and performance of pub- as shares of GDP, and framed on cross-country comparisons lic sectors vary so widely. of per capita GDP. However, what GDP actually measures Moreover, in national accounts systems, this attributed may differ very significantly across countries for several rea- nonmarket value added is further allocated to the household sons. This note focuses on a particular source for this hetero- sector as “actual consumption.� As Deaton and Heston geneity: the quality of public spending. Broadly speaking, (2008) put it: the “quality of public spending� refers to the government’s [...] there are many countries around the effectiveness in transforming resources into socially valuable world where government-provided health outputs. and education is inefficient, sometimes in- Everywhere around the world, nonmarket government volving mass absenteeism by teachers and accounts for a big share of GDP, and yet it is poorly mea- health workers [...] so that such ‘actual’ sured—namely, the value to users is assumed to equal the pro- consumption is anything but actual. To ducer’s cost. Such a framework is deficient because it does not count the salaries of AWOL government allow for changes in the amount of output produced per unit employees as ‘actual’ benefits to consumers of input, that is, changes in productivity (for a recent review adds statistical insult to original injury. of this issue, see Atkinson [2005]) It also assumes that these This “statistical insult� logically follows from the United inputs are fully used. To put it another way, standard national Nations System of National Accounts (SNA) framework once accounting assumes that government activities are on the best “waste� is classified as income—since national income must be 1 POVERTY REDUCTION AND ECONOMIC MANAGEMENT (PREM) NETWORK    www.worldbank.org/economicpremise either consumed or saved. Absent teachers and health care inputs rather than with the actual inputs, which include workers are all too common in many low-income countries waste. (Chaudhury and Hammer 2004; Kremer and others 2005; Indicators of inefficiency exist for the overall public sec- Chaudhury and others 2006; World Bank 2004). Beyond tor and for specific activities such as education, health care, straight absenteeism, which is an extreme case, there are usu- transportation, and other sectors. However, they are far from ally significant cross-country differences in the quality of pub- being uncontroversial. Sources of disagreement include: lic sector services. omission of inputs and/or outputs, temporal lags needed to Moving beyond this state of affairs, there are two alterna- observe variations in the output indicators, choice of mea- tive approaches. One is to try to find indicators for both out- sures of outputs, and mixing outputs with outcomes. put quantities and prices for direct measurement of some Empirical efficiency measurement methods first con- public outputs, as recommended in SNA 93 (but yet to be struct a reference technology based on observed input-output broadly implemented). The other is to correct the input costs combinations, using econometric or linear programming to account for productive inefficiency, namely, to purge from methods. Next, they assess the distance of actual input-out- GDP the fraction of these inputs that are wasted. This note put combinations from the best-practice frontier. These dis- illustrates this type of correction. tances, properly scaled, are called efficiency measures or scores. An input-based efficiency measure shows the extent to which Measuring Living Standards the amount of inputs can be reduced without reducing the GDP’s shortcomings in measuring economic welfare are well level of output. Thus, an efficiency score, say, of 0.8 means known. For example, it is often pointed out that GDP does that using best practices observed elsewhere, 80 percent of not capture differences in leisure or in longevity; it does not the inputs would suffice to produce the same output. reflect differences in inequality or in poverty; and it does not Here, corrections to GDP are based on the efficiency take into account the effect of economic activity on the envi- scores estimated in two papers: Afonso, Schuknecht, and Tan- ronment. This has led to alternative attempts to enlarge the zi (2010) for several indicators referred to a set of 24 coun- concept of GDP, one of the earliest being the Measure of Eco- tries, and Evans and others (2000), which focuses on health, nomic Welfare developed by Nordhaus and Tobin (1972). for 191 countries based on WHO data. Ideally, corrections The recent Report by the Commission on the Measurement of should be based on input-based technical-efficiency studies Economic Performance and Social Progress prepared for the that deal exclusively with inputs and outputs and do not French government by Stiglitz, Sen and Fitoussi (2010) pres- bring outcomes into the analysis. The reason is that public ents an insightful up-to-date summary of the issues. Some of sector outputs interact with other factors to produce out- the report’s main recommendations include: (i) using net in- comes, and here cross-country heterogeneity can play an im- come- or consumption-based measures along with wealth portant role driving cross-country differences in outcomes. rather than gross production-based aggregates; (ii) broaden- Unfortunately, investigation did not discover any technical- ing measures to nonmarket activities; and (iii) considering a efficiency studies covering a broad sample of countries re- dashboard of indicators for the quality of life, environment, stricted to input-output analysis. In particular, these two and sustainability. studies deal with a mix of outputs and outcomes and there- In addition, in the context of the public sector, many gov- fore the results reported here should be seen as illustrative. ernment activities are intermediate inputs for production ac- Furthermore, it should be underscored that the level of waste tivities rather than genuine final outputs. Government ser- identified for each particular country varies significantly vices used effectively as inputs by firms are called across studies, which implies that any associated measures of “instrumental expenditures� by Nordhaus and Tobin (1972). GDP adjusting for this waste will also differ. These instrumental expenditures should be deducted from Corrected GDP the aggregate measure of net income. Nonetheless, this note restricts itself to the SNA framework, where GDP is taken as The first step in the proposed correction of GDP numbers is a measure of production, not welfare. The issue of netting out removing government waste, as reported by recent studies instrumental expenditures from output is also ignored. that look at the efficiency of the government sector. For illus- The simple point emphasized here is that once it is rec- trative purposes, the efficiency scores estimated in Afonso, ognized that the effectiveness of the government’s “produc- Schuknecht, and Tanzi (2010) are used, which estimate pub- tion function� varies significantly across countries, the sim- lic sector efficiency indicators for different categories. This ple convention of equating output value to input cost must note focuses on the indicators corresponding to the function- be revisited. Thus, if it is acknowledged that the same output al categories of health and education. The corresponding (res- could be achieved with less inputs, it is more appropriate to caled) efficiency scores are used to compute a corrected esti- credit GDP or gross national income (GNI) with the required mate of the contribution of public health and education 2 POVERTY REDUCTION AND ECONOMIC MANAGEMENT (PREM) NETWORK    www.worldbank.org/economicpremise services to GDP. Next, the estimated waste from GDP is Figure 1. GDP Loss Due to Health and Education Waste versus per purged using the average (1998–2002) functional shares re- Capita GDP ported in Afonso, Schuknecht, and Tanzi (2010, table 1). 8 POR Table 1 shows the percentage of GDP losses due to pub- lic waste in education and health for the selected countries. MAL Overall, the size of the correction is quite remarkable; the SAZ GDP loss (in percentage of GDP) 6 EST average loss amounts to 4.1 percentage points of GDP, while LIT CZH averages for education and health are 1.5 and 2.6. Given an LAT HUN average spending of 4.6 percent of GDP on education and POL SLV IRE SLK 4.0 percent of GDP on health, this means that 32.6 and 65.0 BRA GRE 4 MEX percent of the inputs are wasted in the respective sectors. TUR THA Note that the best-practice frontier that is used as reference BUL CYP to compute the efficiency scores is constructed on the basis ROM CHL 2 MAU of this particular set of 24 countries. Increasing the refer- KOR ence group to a larger set of countries can only make these efficiency scores worse, as the reference technology becomes richer. 0 SGP 7.5 8.0 8.5 9.0 9.5 10.0 GDP per capita (log) Table 1. GDP Losses Associated with Wasted Public Resources (percentage of GDP, averages 1998–2002) Source: Authors’ calculations based on efficiency scores in Afonso, Schuknecht, and Tanzi (2010). Country Education Health Sum Brazil 2.2 2.0 4.2 Figure 1 plots the GDP losses against the corresponding Bulgaria 0.1 2.6 2.7 per capita GDPs. For this set of countries, there is no strong Chile 1.2 1.0 2.2 discernible pattern; the points scatter rather uniformly over Cyprus 2.2 1.1 3.3 the plot area. Perhaps it could be argued that the range of cor- Czech Republic 0.6 4.8 5.4 rection sizes increases with the level of income—the lower en- velope of the scatter slopes negatively, while the upper enve- Estonia 2.8 3.0 5.9 lope slopes positively. Greece 0.5 3.6 4.1 Another matter of interest is whether the per capita Hungary 1.3 3.9 5.2 GDP ranking is altered at all due to the correction (that is, Ireland 1.0 3.5 4.5 whether any country changes relative position). This reorder- Korea, Rep. of 0.5 1.0 1.5 ing happens on 9 occasions in the 24 countries. In the scatter Latvia 2.8 2.1 4.9 plot (figure 1), the candidates are pairs of countries where one Lithuania 2.5 3.1 5.6 is almost vertically on top of another, but slightly to the right, Malta 1.7 4.7 6.4 and where the vertical (correction) distance is substantial. For Mauritius 1.2 0.7 1.9 example, the Republic of Korea overtakes Cyprus; Cyprus, in turn, almost catches up with Greece, Brazil overtakes Lithua- Mexico 2.4 1.2 3.7 nia, and Poland overtakes Estonia. Poland 1.8 2.8 4.6 The World Health Organization (WHO) study by Evans Portugal 3.1 4.8 7.8 and others (2000) covers health in both advanced and devel- Romania 0.0 2.5 2.5 oping economies. The average GDP loss is 0.9 percentage Singapore 0.0 0.0 0.0 points (the median is 0.8 percent of GDP). This is lower than Slovak Republic 0.8 3.8 4.6 the estimate in table 1 for health, reflecting the lower level of Slovenia 0.0 4.6 4.6 health spending in the wider country data set used in the South Africa 3.7 2.5 6.2 WHO study. The losses are uniformly distributed over the per capita GDP range. Baldacci and others (2008) find that in Thailand 2.3 1.0 3.3 countries suffering from poor governance, the positive effects Turkey 1.2 2.6 3.9 of increased spending on education is reduced, and those re- Average 1.5 2.6 4.1 sulting from higher health spending can be completely negat- Source: Authors’ calculations based on efficiency scores in Afonso, Schuknecht, ed. Rajkumar and Swaroop (2008) also show that in a context and Tanzi (2010). 3 POVERTY REDUCTION AND ECONOMIC MANAGEMENT (PREM) NETWORK    www.worldbank.org/economicpremise Figure 2. GDP Loss Due to Health Waste versus per Capita GDP of low quality of governance, increased expenditures in health NAM and education are not reflected in improved social outcomes. ZMB Given the high correlation between income and governance, poorer countries tend to have more ineffective governments. 3 At the same time, they tend to spend less on health. The com- LSO FSM bined effect is a broadly uniform distribution of waste, as fig- GDP loss (in percentage) ZAF ure 2 shows. SWZ 2 MDA BWA Because inefficiency scores are sector specific, a “virtual EST MOZ MLI GNQ BLR MDV HUN experiment� can be performed by asking what would be the MNG WSM GHA JOR SVKCZE DNK MWI GNB KGZ AGO TKM DJI BOL LVA HRV ATG NZL USA implications if these inefficiencies applied, on average, UGA KAZ MKD DEU ETHTCD MDG GUY SVN BHS FIN throughout all public sector activities. What would be the ex- 1 BFA KHM BIHCOG COL URY BRB CAN ERI BEN BGR CUB LBN ISL NOR tent of the “missing� GDP? Figure 3 shows the distribution of MRT BTN VUT VCT LCA LUX TJK VNM SEN PHLSYR GMB CMR SLV TTO MEX KOR BHR CHE the correction versus per capita GDP and technical-efficiency ALB GTM GRC ARE JPN COD NPL GIN PAK EGY CYP ITA MCO scores. Technical efficiency is positively correlated with per GEO IDN DOM VEN SAU ESP FRA 0 ZWE IRQ MAR JAM OMNMLT SGP capita GDP. As before, the correction is roughly uniformly 5 6 7 8 9 10 11 distributed across the range of per capita GDP. The effects of GDP per capita (log) lower efficiency scores and lower spending broadly compen- Source: Authors’ calculations based on efficiency scores in Evans and others sate for each other. Thus, poorer countries with more ineffec- (2000). tive government also spend a smaller share of GDP in public services, so any correction of the sort discussed here is going to be small. The scatter of technical efficiency versus total Figure 3. Technical Efficiency Scores, per Capita GDP, and Total Loss waste displays an upper envelope: 5 6 7 8 9 10 the estimated waste is bounded by 1.0 the efficiency score. Finally, the focus turns here to .8 the country rankings of living technical ef�ciency standards, the GNI per capita .6 (scores) computed using the World Bank’s .4 Atlas methodology.1 As noted, this is the measure that the World .2 Bank uses for classifying countries in income groups, as well as to set 10 lending eligibilities. What is the ef- 9 fect on the ranking of the correc- 8 tions discussed here? Consider the GDP per capita correction based on the health ef- 7 (log) ficiency scores of Evans and others 6 (2000) applied to the value added 5 of public administration and de- fense for the 2009 GNI. The result is reordered country ranking 15 where 51 countries out of 93 change their relative positions. total waste 10 Since the value-added variable is (percent of GDP) available only for nondeveloped 5 countries, the same correction is performed on the wage bill—to 0 cover a larger set of countries. The .2 .4 .6 .8 1.0 0 5 10 15 portion of reordered countries is still higher than 50 percent; 59 of Source: Authors’ calculations based on efficiency scores in Evans and others (2000). 116 countries are repositioned. In both corrections, about 70 per- 4 POVERTY REDUCTION AND ECONOMIC MANAGEMENT (PREM) NETWORK    www.worldbank.org/economicpremise cent of the reordering happens in the lower half of the origi- Executive Boards, or their management. A more detailed ex- nal ranking and the average shift is approximately equal to position of the views expressed here can be found in Grigoli two positions. and Ley (2012). Concluding Remarks About the Authors The current practice of estimating the value of the govern- Francesco Grigoli is an Economist in the International Monetary ment’s nonmarket output by its input costs is not only unsat- Fund’s Fiscal Affairs Department, where he works on public isfactory, but also misleading in cross-country comparisons of expenditure and fiscal policy–related issues. Formerly, he was living standards. Since differences in the quality of the public a consultant of the World Bank’s Public Sector Group in the sector have an impact on the population’s effective consump- Poverty Reduction and Economic Management (PREM) Net- tion and welfare, they must be accounted for in any compari- work, and he was a Visiting Scholar at Columbia University.  He sons of living standards. This note included illustrative cor- holds a PhD in development economics from the University of rections of the input costs to account for productive Insubria, Italy, and a MA in international economics from the inefficiency, thus purging from GDP the fraction of these in- University of Sussex, UK. puts that is wasted. Eduardo Ley is Lead Economist in the Economic Policy and The results suggest that the magnitude of the correction Debt Department of the PREM Network of the World Bank, could be significant. When correcting for inefficiencies in where he works on fiscal policy issues. Formerly he held posi- the health and education sectors, the average loss for a set of tions at the IMF, Fundación de Estudios de Economía Aplicada 24 European Union member states and emerging econo- in Madrid, Resources for the Future in Washington, DC, and the mies amounts to 4.1 percentage points of GDP. Sector-spe- Universidad Carlos III de Madrid. He holds a PhD in econom- cific averages for education and health are 1.5 and 2.6 per- ics and a MA in statistics from the University of Michigan. centage points of GDP, implying that 32.6 and 65.0 percent Note of the inputs are wasted in the respective sectors. These cor- rections are reflected in the GDP per capita ranking, which 1. The Atlas conversion factor for any year is the average of a gets reshuffled in 9 cases out of 24. In a hypothetical scenario country’s exchange rate (or alternative conversion factor) for where the inefficiency of the health sector is assumed to be that year and its exchange rates for the two preceding years, representative of the public sector as a whole, the rank reor- adjusted for the difference between the rate of inflation in dering would affect about 50 percent of the 93 countries in the country and that in a number of developed countries. the sample, with 70 percent of the reordering occurring in For more details, see http://data.worldbank.org/indicator/ NY.GNP.PCAP.CD. the lower half of the original ranking. These results, howev- er, should be interpreted with caution, as the purpose of this References note is to call attention to the issue, not provide fine-tuned Abraham, Katharine G., and Christopher Mackie. 2006. “A waste estimates. Framework for Nonmarket Accounting.� In A New Architecture A natural way forward involves finding indicators for for the US National Accounts, ed. Dale W. Jorgenson, J. Steven both output quantities and prices for direct measurement of Landefeld, and William D. Nordhaus, 161–92. University of some public outputs. This is recommended in SNA 93, but Chicago Press. has yet to be implemented in most countries. Moreover, in re- Afonso, António, Ludger Schuknecht, and Vito Tanzi. 2010. “Pub- lic Sector Efficiency: Evidence for New EU Member States and cent times there has been increased interest in outcomes- Emerging Markets.� Applied Economics 42 (17): 2147–64. based performance monitoring and evaluation of government Atkinson, Anthony B. 2005. Measurement of Government Output activities (see Stiglitz, Sen, and Fitoussi [2010]). As argued and Productivity for the National Accounts. Basingstoke: Palgrave also in Atkinson (2005), it will be important to measure not MacMillan. only public sector outputs but also outcomes, because the lat- Baldacci, Emanuele, Benedict Clements, Sanjeev Gupta, and Qiang ter are what ultimately affect welfare. A step in this direction Cui. 2008. “Social Spending, Human Capital, and Growth in Developing Countries.� World Development 36 (8): 1317–41. is to extend the accounting of the nation’s productive inputs Chaudhury, Nazmul, and Jeffrey S. Hammer. 2004. “Ghost Doc- and outputs, taking into account specific aspects of nonmar- tors: Absenteeism in Rural Bangladeshi Health Facilities.� ket activities, as suggested by Abraham and Mackie (2006) for World Bank Economic Review 18 (3): 423–41. the United States, with the creation of “satellite� accounts in Chaudhury, Nazmul, Michael Kremer, Karthik Muralidharan, and specific areas such as education and health. F. Halsey Rogers. 2006. “Missing in Action: Teacher and Health Worker Absence in Developing Countries.� Journal of Economic Acknowledgment Perspectives 20 (1): 91–116. Deaton, Angus, and Alan Heston. 2008. “Understanding PPPs and The views expressed in this brief are those of the authors and PPP-Based National Accounts.� National Bureau of Economic should not be attributed to the IMF, the World Bank, their Research Paper No. 14499, Cambridge, MA. 5 POVERTY REDUCTION AND ECONOMIC MANAGEMENT (PREM) NETWORK    www.worldbank.org/economicpremise Evans, David B., Ajay Tandon, Christopher L. Murray, and Jeremy Nordhaus, William D., and James Tobin. 1972. “Is Growth Obso- A. Lauer. 2000. “Comparative Efficiency of National Health lete?� Cowles Commission Monograph 319, Yale University. Systems in Producing Health: An Analysis of 191 Countries.� Rajkumar, Andrew S., and Vinaya Swaroop. 2008. “Public Spending GPE Discussion Paper Series No. 29, WHO, Geneva. and Outcomes: Does Governance Matter?� Journal of Develop- Grigoli, Francesco, and Eduardo Ley. 2012. “Quality of Govern- ment Economics 86: 96–111. ment and Living Standards: Adjusting for the Efficiency of Stiglitz, Joseph E., Amartya Sen, and Jean-Paul Fitoussi. 2010. Public Spending.� WP/12/182, IMF, Washington, DC. Report of the Commission on the Measurement of Economic Perfor- Kremer, Michael, Karthick Muralidharan, Nazmul Chaudhury, Jef- mance and Social Progress. http://www.stiglitz-sen-fitoussi.fr/en/ frey Hammer, and F. Halsey Rogers. 2005. “Teacher Absence in index.htm. India: A Snapshot.� Journal of the European Economic Association World Bank. 2004. World Development Report 2004: Making Services 3 (2): 658–67. Work for the Poor. Washington, DC: World Bank. The Economic Premise note series is intended to summarize good practices and key policy findings on topics related to economic policy. They are produced by the Poverty Reduction and Economic Management (PREM) Network Vice-Presidency of the World Bank. The views expressed here are those of the authors and do not necessarily reflect those of the World Bank. The notes are available at: www.worldbank.org/economicpremise. 6 POVERTY REDUCTION AND ECONOMIC MANAGEMENT (PREM) NETWORK    www.worldbank.org/economicpremise