Education and Health in G7 Countries: Achieving Better Outcomes with Less Spending
  • 1 0000000404811396https://isni.org/isni/0000000404811396International Monetary Fund

Enhancing the efficiency of education and health spending is a key policy challenge in G7 countries. The paper assesses this efficiency and seeks to establish a link between differences in efficiency across countries and policy and institutional factors. The findings suggest that reforms aimed at increasing efficiency need to take into account the nature and causes of inefficiencies. Inefficiencies in G7 countries mostly reflect lack of cost effectiveness in acquiring real resources, such as teachers and pharmaceuticals. We also find that high wage spending is associated with lower efficiency. In addition, lowering student-teacher ratios is associated with reduced efficiency in the education sector, while immunizations and doctors' consultations coincide with higher efficiency in the health sector. Greater autonomy for schools seems to raise efficiency in secondary education.

Abstract

Enhancing the efficiency of education and health spending is a key policy challenge in G7 countries. The paper assesses this efficiency and seeks to establish a link between differences in efficiency across countries and policy and institutional factors. The findings suggest that reforms aimed at increasing efficiency need to take into account the nature and causes of inefficiencies. Inefficiencies in G7 countries mostly reflect lack of cost effectiveness in acquiring real resources, such as teachers and pharmaceuticals. We also find that high wage spending is associated with lower efficiency. In addition, lowering student-teacher ratios is associated with reduced efficiency in the education sector, while immunizations and doctors' consultations coincide with higher efficiency in the health sector. Greater autonomy for schools seems to raise efficiency in secondary education.

I. Introduction and Main Conclusions

A key policy challenge in G7 countries is to improve the performance of education and health systems while containing their cost. Education and health outcomes are critically important for social welfare and economic growth and thus, spending in these areas constitutes a large share of public spending. But there is concern about the efficiency of such spending. In education, there are questions about the ability of school systems to maximize the potential of students and respond effectively to changes in the demand for education outcomes. In health, there is concern about the rapid rise of the cost of health care and the impact on competitiveness, as well as trade-offs between the efficiency and equity of health systems.

This paper attempts to assess the efficiency of education and health spending in G7 countries. It asks whether countries could achieve better education and health outcomes at current levels of spending or, conversely, whether countries could have the same outcomes at lower levels of spending. We seek to establish a link between observed differences in efficiency across G7 countries and discuss the role of policy and institutional factors in explaining efficiency differences. The paper also discusses efficiency-enhancing reforms for the education and health sectors. The analysis involves addressing complicated issues of what drives outcomes in education and health and should be regarded as exploratory. Follow-up work on data and other parts of the analysis are needed for more definitive answers.

The paper is organized as follows. Section II provides the background and relevant literature for the analysis. Section III describes trends in education and health spending and outcomes in G7 countries. It also explores the issue of how efficiency in the education and health sectors in the G7 can be measured, and how observed differences between countries may be related to policy choices and institutions. Section IV makes some suggestions for reforms in the education and health sectors in G7 countries. The technical aspects of the analysis are elaborated in Appendix I.

Based on the quantitative analysis of education and health spending and outcomes for the G7, this paper finds that:

  • Public spending on education and health systems varies greatly in G7 countries, and so do education and health outcomes. Spending is particularly high in relation to outcomes in education and health in France, Germany, the U.K., and the U.S.; that is, the question of how to increase the efficiency of spending on education and health is most relevant for these countries. On the other hand, Canada’s education spending is relatively efficient, as is health spending in Italy and Japan.

  • Part of the differences in spending efficiency can be attributed to exogenous factors such as GDP, demographics, and differences in lifestyle.

  • Policies and institutions are also associated with differences in efficiency. In particular, countries that spend a relatively large share of their education and health budgets on wages and salaries tend to be less efficient. Also, lower student-teacher ratios are associated with reduced efficiency in the education sector, while immunizations and doctors’ consultations (but not the number of doctors per se) are positively correlated with efficiency in the health sector. Greater autonomy for schools seems to raise efficiency in secondary education.

  • Effective education and health reform should aim at enhancing efficiency. This should take into account the stage at which the inefficiencies arise. Further, reforms should seek to balance devolution (of responsibility and resources) and enhanced market competition with regulation to ensure accountability.

  • Cross-country studies, such as this one, can provide important insights into policy challenges that countries face. However, further work on data and sectoral issues is needed to deepen the findings of this paper.

II. Education and Health Spending, Outcomes, and Economic Growth: Background and Literature Review

A large volume of research has emphasized the importance of education outcomes to human development, economic growth, and productivity.1 However, the findings also note the mixed evidence for the relationship between education spending and student performance in developed nations. Recent G8 statements have recognized the need to improve all aspects of the quality of education and the promotion of high standards in education of mathematics, science, technology, and foreign languages.2

On health issues, the G7 countries have focused on complex issues of high and rapidly rising cost and concerns about equity. Cost-enhancing technological advances and, to a lesser extent, aging populations and increased demand for health services as populations become wealthier, are pushing up health care prices faster than general price levels. This has prompted governments to introduce measures to reduce the cost of health care (Cutler, 2002; Newhouse, 1992). But efforts in G7 countries to constrain health spending by rationing or increasing competition have run into concerns about disadvantaged groups’ access to health care.

A key issue in the debate about education and health spending is how (and how strongly) it translates into sectoral outcomes and economic growth. Underlying the policy discussions on education and health is a concern that public and private spending is high in relation to outcomes (education attainment, health status, and economic growth). Prior research on this has yielded the following insights (see Table 5):

  • Education attainment is a key driver of economic growth, with attainment at higher levels of education gaining importance as economies become more developed. Health status is also found to have a positive impact on growth.

  • However, the evidence for a positive relationship between education spending and attainment is mixed, with a number of studies finding no evidence of a statistically significant relationship. Similarly, there is limited evidence for a positive link between health spending and health status. But public spending has been found to benefit the poor by enhancing their access to health services.

  • Selected policies, institutions, and environmental factors, on the other hand, have a clear impact on education attainment. In particular, family background and teacher quality have been identified as important explanatory variables for student achievement. School autonomy and emphasis on assessing student performance are likely to raise education attainment, while teacher unions’ influence may have a negative impact. The effect of other factors on education attainment is ambiguous; these include the impact of class size, decentralization, and the relative importance of public and private education.

  • Key factors for explaining health status include lifestyle (e.g., the consumption of alcohol and tobacco, and diet), income level, occupational and socioeconomic status, urbanization, and medical technology. In addition, education attainment is an important explanatory factor for health outcomes. As in education, the impact of the mix of private and public spending on outcomes is not clear.

  • Improved achievement at lower levels of education promotes both equity and economic growth. However, improved achievement at higher levels of education may benefit the well-off most. In health, there is a strong tradeoff between policies aimed at enhancing cost effectiveness of spending and improving the equity of outcomes.

III. Spending and Outcomes in Education and Health: Empirical Analysis

This section discusses cross-country empirical evidence on the relationship between spending and outcomes in the education and health sector. It starts with a description of recent trends in education and health spending and outcomes. Then, differences in the relationship between spending and outcomes are assessed for G7 countries. Finally, an attempt is made to identify factors that may affect these differences.

A. Trends in Education and Health Spending and Outcomes

Spending on education and health varies considerably within the group of G7 countries.3 Total education spending (including funded from private sources) has been considerably higher in the U.S. than in other G7 countries, particularly at the tertiary level (Figure 1 and Table 6 in Appendix II). Average G7 spending on primary and secondary education is near average levels in the OECD but G7 countries spend more on average on tertiary education.4 Public health spending is highest in Germany and the U.S., and lowest in Italy, Japan, and the U.K. (Figure 2 and Table 6).5 Average public health spending in OECD countries is below spending levels in G7 countries.

Figure 1.
Figure 1.

Total Education Spending per Student by Level of Education, 2003

(PPP dollars)

Citation: IMF Working Papers 2007, 263; 10.5089/9781451868265.001.A001

Source: OECD Education at a Glance 2006, www.oecd.org/edu/eag2006.1/ Excludes Mexico and Turkey (because of outlying data). Countries are ranked by level of total secondary education spending.2/ Data for primary and secondary education are averaged for Canada.
Figure 2.
Figure 2.

Total Health Spending per Capita by Source, 1998–2001

(Period average in PPP dollars)

Citation: IMF Working Papers 2007, 263; 10.5089/9781451868265.001.A001

Source: OECD Health Data 2006, www.ecosante.fr.1/ Excludes Mexico and Turkey (because of outlying data). Countries are ranked by level of public health spending.

G7 countries have markedly increased overall (public and private) spending on education and health in real terms over the last decade. Between 1995 and 2003, real overall spending per student from public sources increased at an average rate of 2.2 percent per year in primary and secondary education and 1.8 percent in tertiary education (Table 1). Spending trends differed considerably between G7 countries, with Germany posting relatively small spending increases and Canada and Japan increasing their spending at rates well above the average for the G7. Average health expenditure from public and private sources increased by 3.8 percent per year in real per capita terms in the G7 over the period 1995–2003. Increases in health spending were significantly larger for the U.K. and U.S. than for Germany.

Table 1.

Trends in Health and Education Spending, 1995–2003

article image
Sources: OECD Education at a Glance 2006, www.oecd.org/edu/eag2006; OECD Health Data 2006, www.ecosante.fr; and IMF staff calculations.

The average annual real percent increase over 1995–2003 in per student public and private spending in primary and secondary (including post-secondary non-tertiary) and tertiary education.

The average annual increase in real per capital health public and private spending over 1995–2003.

Excludes Mexico and Turkey (because of outlying data).

The rapid increase in the cost of health care has put pressure on public health budgets in G7 countries. Average public spending on health increased from 6.4 percent of GDP in 1995 to 7.1 percent of GDP in 2003. This increase was highest for France, Italy, and the U.K. (see Table 1). Public spending for education declined slightly in the G7 countries, from an average of 4.8 percent of GDP in 1995 to 4.6 percent of GDP in 2003. This mostly reflected a substantial decline in public education spending in Canada.6

Education and health outcomes generally improved in G7 countries in recent years (Table 2).7 Except for Japan, all G7 countries posted increases in the average test scores for mathematics proficiency under OECD’s Programme for International Student Assessment (PISA) between the 2000 and 2003 rounds.8 Improvements in education participation can be gauged from trends in graduation rates of successive cohorts (i.e., if later cohorts have higher graduation rates, education participation has increased over time). The current cohort of 25–34 year olds is generally better educated than the preceding cohort which is presently aged 35–44 years, as indicated by the larger share that attained at least upper secondary and tertiary education in the younger of the two cohorts. France, Italy, and the U.K. posted the highest gains, while there was no substantial progress in the U.S. and even a loss of education attainment in Germany. In health, indicators of mortality (standardized death rates and infant mortality) significantly improved in all G7 countries between 1995 and 2003, with the most progress made by Germany, Italy, and Japan.

Table 2.

Trends in Health and Education Outcomes

article image
Sources: OECD Education at a Glance 2006, www.oecd.org/edu/eag2006; OECD Health Data 2006, www.ecosante.fr; and IMF staff calculations.

Percent difference in scores on comparable components of the PISA mathematics test between the 2000 and 2003 rounds.

Difference in the percent of the population that has attained upper secondary education between the age group 35–44 and 25–34 in 2004.

Percent difference in the percent of the population that has attained tertiary education between the age group 35–44 and 25–34 in 2004.

Annual percent change during 1995–2002.

Annual percent change during 1995–2003.

Excludes Mexico and Turkey (because of outlying data).

Trends in education and health outcomes are not closely related to spending trends in G7 countries. A comparison of Tables 1 and 2 shows that the U.K. achieved above-average improvements in education outcomes with relatively modest increases in real spending, while the U.S. achieved less with more additional resources. Similarly, trends in real health expenditure and health status show that Germany has been able to achieve more gains with a smaller increase in real resources than Canada, the U.K., and the U.S.9 These issues will be investigated in more depth in the following section.

B. The Relative Efficiency of Education and Health Spending

The relative efficiency of education and health spending is assessed by comparing expenditure levels and associated outcomes in G7 countries and other OECD countries. This is done using the Data Envelopment Analysis (DEA) technique, which was developed for estimating best-practice frontiers and relative efficiency in business applications. In this case, DEA is used to assess the relationship between spending (inputs) and outcomes (production) across countries. The general principle is that countries which achieve the same or better outcomes with lower levels of spending than other countries in the sample are the most efficient and determine the best-practice frontier.10 The relative spending efficiency of other countries can be measured by how far away they are from the best-practice frontier. This method is intuitively appealing and has the advantage of being flexible and parsimonious in its assumptions. Appendix I provides a detailed discussion of DEA.

Education and health spending in the G7 countries are generally not among the most efficient in the OECD (Figures 412). Japan is on the best-practice frontier for health-adjusted life expectancy (HALE), standardized death rates, and infant mortality. This means that among the OECD countries with health spending at or below Japan’s level, no country scored better on these health outcomes. Table 3 shows how G7 countries are placed in the ranking of efficiency scores in the sample of OECD countries. G7 countries generally rank in the bottom two quartiles of the rank distribution (particularly France, Germany, the U.K., and the U.S.). Yet, Canada ranks relatively high in education, and Italy and Japan rank relatively high in health.

Table 3.

Efficiency of Education and Health Spending in G7 Countries Relative to the OECD

(Distribution by quartiles of the ranking of efficiency scores for OECD countries) 1/

article image
Sources: OECD Education at a Glance 2006, www.oecd.org/edu/eag2006; OECD PISA, http://pisaweb.acer.edu.au/oecd_2003/oecd_pisa_data_s1.html; OECD Health Data 2006, www.ecosante.fr; and IMF staff calculations.

Canada’s efficiency scores in education ranked, on average, at the 15th percentile of the overall ranking of efficiency scores in the sample of OECD countries. This places Canada in the top (1–25) quartile of the OECD ranking distribution. The rankings are based on the point estimates of the input-oriented bias-corrected efficiency scores (see Appendix I).

Based on overall (public and private) spending by level of education and associated outcome indicators, including the mean and within-country variation in PISA scores for mathematics and graduation rates for upper-secondary and tertiary education. The OECD countries are ranked by efficiency score for each outcome indicator. Finally, the rankings for all education outcome indicators are averaged to get the average country ranking for education spending.

Based on average rankings of spending efficiency (see previous footnote) based on public health spending and associated outcome indicators, including infant, child, and maternal mortality rates, and healthy life expectancy and standardized death rates.

These initial findings for spending efficiency can be further disentangled by looking at the different stages of the process that links spending and outcomes. In particular, it is useful to separate the impact of spending on intermediate outputs (the first stage of the process from spending to outcomes) and the impact of these intermediate outputs on outcomes (the second stage of the process).11 The link between spending and intermediate outputs reflects the cost of procuring and producing the intermediate outputs, that is, issues of cost effectiveness. Countries may also find that their education and health outcomes are low in relation to the available intermediate outputs—in this paper this is referred to as system inefficiency. See Appendix I for more details on these concepts.

For the most part, the relatively low average spending efficiency in G7 countries seems to reflect issues of cost effectiveness, at least in tertiary education and health. Table 4 shows that the majority of G7 countries have better system efficiency in education and health than other OECD countries on average (i.e., the ratio shown for the G7 average is smaller than 1).12 The results are especially striking for the U.K., which shows very high system efficiency in both tertiary education and health, although spending efficiency in the U.K. is well behind the OECD average. The average results for the G7 imply that lack of cost effectiveness is a major driver of spending inefficiencies in the G7 in tertiary education and health. This may be addressed by increasing competition between suppliers of goods and services to education and health institutions and tackling overspending on specific spending items (e.g., due to overstaffing). In secondary education, the situation is reversed, and G7 countries (except Germany) score worse on system efficiency than on overall spending efficiency. This suggests that for secondary education, options for reducing public spending may be found by assessing whether the mix of intermediate outputs is optimal for achieving improved outcomes, or whether incentives for an effective use of intermediate outputs are adequate.

Table 4.

Spending and System Efficiency in Education and Health 1/

article image
Sources: OECD Education at a Glance 2006, www.oecd.org/edu/eag2006; OECD PISA, http://pisaweb.acer.edu.au/oecd_2003/oecd_pisa_data_s1.html; OECD Health Data 2006, www.ecosante.fr; and IMF staff calculations.

Ratio of efficiency rankings of G7 countries and the average ranking in the sample of OECD countries. A ratio of 1 implies that the country ranks at the average of the OECD sample; a lower ratio signifies an above average efficiency and a ratio in excess of 1 implies that the country is less efficient than OECD average. See Appendix I for a more detailed explanation of the underlying calculations.

Based on efficiency rankings from Table 3.

Based on efficiency rankings using as inputs the average of various intermediate education outputs (see Table 9) and as production various outcome indicators (see Table 3 and Table 7).

Based on efficiency rankings using as inputs the average of various intermediate health outputs (see Table 10) and as production various outcome indicators (see Table 3 and Table 8).

The efficiency findings are reasonably robust and in line with results from similar assessments of spending efficiency. The rankings of G7 countries do not change very much if overall health spending is used in the efficiency analysis instead of public spending (see Table 4). The findings also do not seem substantially affected by outliers or choice of outcome indicator.13 Furthermore, the findings of this paper are broadly consistent with findings from other research on the efficiency in education and health, although some results differ due to variations in methodology, choice of measures for spending and outcomes, and the selection of sample countries.14

C. Achieving Better Outcomes with Lower Spending

Strengthening the link between spending and outcomes would allow countries to achieve the same outcomes at lower levels of spending or achieve better outcomes at the same level of spending. In order to understand how countries can do this, it is important to determine whether the DEA results are policy-related or reflect factors outside of the direct influence of policy makers. This is done below through a statistical analysis of the efficiency results.15

Cross-country differences in efficiency results are associated with expenditure composition and institutional arrangements in education and health. In particular, wage spending is negatively associated with efficiency in education, while health efficiency tends to be lower in countries with more health workers (Tables 11 and 12).16 This finding indicates that reduced compensation spending may increase efficiency. However, it could also reflect the possibility that countries with poor past education and health outcomes are spending more on staffing. The mix of intermediate outputs is also important: lowering student-teacher ratios in education is associated with lower efficiency, while immunizations seem to be a cost-effective way of boosting health status, and thereby is associated with higher efficiency.17 Greater autonomy for schools and emphasis on academic ability in student selection for schools seems to raise efficiency in secondary education, while decentralization and large out-of-pocket payments are associated with lower efficiency in the health sector. In tertiary education, efficiency is negatively associated with decentralization and private spending. Finally, exogenous factors, such as GDP,18 urbanization,19 socioeconomic background, and lifestyle also affect efficiency.

For G7 countries looking to enhance the efficiency of their education and health spending, the findings could have several implications:

  • The composition of spending in G7 countries tends to be tilted toward compensation, which seems to hamper efficiency in both education and health.

For education, the evidence for a negative associated between overspending on compensation and efficiency is quite strong (see Table 11). The negative association between efficiency in achieving PISA test scores and student-teacher ratios suggest that overstaffing may play a role. In addition, the literature suggests that salary levels may also contribute to inefficiencies.20 In particular, it has been found that although the quality of teachers is an important driver of educational attainment, higher compensation does not in itself increase teacher quality (Blau, 1996, and Rivkin, Hanushek, and Kain, 2005). Among the G7 countries, Japan, Germany, and Italy have high wage bills in education. In Italy, a relatively low student-teacher ratio seems a key driver for the relatively high costs of compensation. In Germany, on the other hand, classroom sizes are relatively large and salary levels may be more of a factor in inefficiencies.21

For health, the data strongly suggest that overstaffing is associated with efficiency problems. Health employment density as well as general practitioner density are negatively correlated with efficiency for all health outcome indicators. At the same time, numbers of doctors’ consultations and hospital beds have a positive association with the efficiency of spending for infant, child, and maternal mortality. It seems, therefore, that how (and how often) health workers interact with patients matters for improving health outcomes in a cost-effective manner.22 Several of the G7 countries (Canada, France, Germany, and the U.S.) have a substantially higher number of health practitioners than the OECD on average, pointing to a potential source of inefficiency in these countries. The U.S. not only has a relatively high number of health practitioners, but also markedly fewer doctors’ consultations per capita than other G7 and OECD countries.

  • Policies at the school level is positively associated with the efficiency of education spending, while the relation with decentralization is ambiguous. The devolution of responsibility to the school level is positively correlated with efficiency scores: in countries where more school principals have hiring authority, the efficiency scores of PISA mathematics test scores and the distribution of scores are higher (see also Woessmann, 2000). In Japan, Germany, and Italy—where efficiency scores for education are relatively low—principal autonomy in hiring teachers is substantially lower than in the OECD on average. Moreover, relying on student records for admission to secondary school appears to be a cost-effective practice which boosts efficiency scores of upper-secondary graduation rates. However, decentralization of spending is not associated with higher or lower efficiency of spending in secondary education, while it is negatively correlated with efficiency in tertiary education. This echoes findings from the literature that the link between performance and decentralization in education is ambiguous (see Table 5). This may reflect the need to balance school autonomy and decentralization with accountability for results (e.g., Woessmann, 2006), which may be a complicated task in the context of fiscal decentralization.

  • The increasing cost of health care has prompted countries to institute reforms, including by increasing out-of-pocket payments for users of health services. It is therefore unsurprising that out-of-pocket payments are associated with lower efficiency—countries where efficiency problems are large have raised out-of-pocket payments to reduce inefficiencies and the fiscal burden posed by the health sector (see Cutler, 2002). But it also suggests that such measures have only achieved limited success and those countries remain relatively inefficient. Aside from the effect of out-of-pocket payments, the share of private spending does not have an impact on efficiency in health. This may reflect adverse selection issues related to private health insurance and incentives for insured persons to overconsume health services.23 The negative correlation between decentralization and health spending efficiency we find for the OECD does not have a clear implication for the G7. It reflects the relatively high efficiency of EU new member states, where a relatively small share of health spending is allocated at the subnational level.

IV. Concluding Remarks

The analysis of this paper breaks new ground by assessing the relationship of spending efficiency in education and health with policies and institutions. The question of how better outcomes can be achieved at a lower cost is central to the concept of efficiency. This is also a critical question for fiscal policy, and a better understanding of what drives efficiency allows a more rational approach to balancing needs in education and health with limits imposed by fiscal affordability and sustainability. In particular, the analysis of this paper suggests directions for education and health policies that could raise efficiency, including:

  • Education and health reforms need to pay close attention to issues of staffing and wages. Countries where the education and health sectors are overstaffed and compensation payments account for a relatively high share of spending may pay a price in the form of reduced efficiency. This seems to be a relevant issue for all G7 countries, perhaps with the exception of the U.K. In this context, policy making should take into account that a larger education and health workforce will not automatically lead to an increase in education and health service delivery. This suggests that policy makers should carefully monitor indicators of capacity utilization (e.g., average class size and hospital bed occupation rates).

  • Spending efficiency can be enhanced by gearing the composition of spending toward cost-effective intermediate outputs. For instance, immunizations yield large benefits in terms of health status at a relatively low cost, and countries should ensure that public spending is not a constraint to achieving better immunization rates. Reducing student-teacher ratios, on the other hand, tends to be a relatively inefficient way of improving education outcomes.

  • Reforms aimed at increasing efficiency need to take into account the stage of the spending process at which inefficiencies arise. In particular, it is important to distinguish lack of cost effectiveness (mostly a problem in health and tertiary education) from system inefficiency (a frequent problem in secondary education). Issues of cost effectiveness may be tackled by increasing competition between suppliers of goods and services to education and health institutions, as well as addressing any bias in the composition of spending toward wages and salaries (e.g., by moderating salaries and tackling overstaffing). System efficiency can be improved by reforms that promote competition between education and health institutions in a framework of appropriate accountability for the quality of service delivery.

  • Effective education and health reforms need to balance increased devolution of responsibility and resources with enhanced market competition and regulatory frameworks that ensure accountability for results. Shifting responsibility for education and health to the private sector and subnational governments is unlikely to achieve efficiency gains alone, but may be part of a balanced reform strategy. In education, such a reform strategy could include enhanced autonomy for schools to hire teachers and selecting students on merit. In health, progress may be made by providing incentives for cost savings. With cost-increasing technological progress and aging driving up the cost of health care, reducing inefficiencies in the health sector will be key for achieving sustainable fiscal policies and overcoming obstacles to international competition in G7 countries. However, it is critical for reforms aimed at increasing efficiency to address concerns about accountability that may arise from an increased role for competition. In addition, concerns about equity need to be taken into account. In particular, disadvantaged groups are often dependent on public spending for access to basic education and health services. Reforms should be designed to balance equity concerns (and the related risk of reversal of reforms in the future) with the need to increase spending efficiency.

Appendix I. Data, Data Envelopment Analysis, and Second-Stage Analysis

In this paper, the DEA technique is used to assess the relationship between spending and outcomes. This method allows us to identify countries where the ratio of outcomes over spending is relatively high in comparison with other countries where it is comparatively low. A second-stage statistical analysis is used to assess whether differences between countries can be accounted for by factors out of the control of policy makers, or whether changes in expenditure and sectoral policies may make a difference. This appendix explains key technical aspects of the data used in the paper, the DEA methodology and the second-stage analysis.

A. Some Data Issues

The paper uses PPP deflated spending that eliminates unit cost differences across countries in the education and health sector. The idea is to use a measure of spending that is comparable across countries in the sense that, assuming equal cost efficiency, a similar package of education and health services could be bought at the same level of spending.24 Expressed as a ratio of GDP, the spending measure will be biased against countries with lower levels of GDP—at a given level of spending as a share of GDP, richer countries will be able to procure more education and health services than poorer countries. In PPP terms this bias will likely be closer to zero than with the alternatives, although some bias against richer countries may remain. In any event, such bias should be limited within the relatively homogeneous group of OECD countries.

The DEA exercise benchmarks the relationship between spending and outcomes in G7 countries against the OECD. The sample of G7 countries is too small for a meaningful comparison of spending levels and outcomes, so the sample was expanded to the OECD (excluding the outliers Mexico and Turkey). The DEA exercise links spending and outcomes as follows:25

  • Total spending per student in secondary education is related to average PISA mathematics test scores, the distribution of those test scores (measured as the ratio of the score at the 25th percentile in the country to the score at the 75th percentile), and secondary education graduation rates.26

  • Total spending per student in tertiary education is related to tertiary education graduation rates.

  • Public spending per capita on health is related to infant, child, and maternal mortality rates, as well as HALE and population standardized death rates.27

The choice of these outcome indicators is driven by data availability and reflects the variables that have been used in other studies of education attainment and health status (e.g., WHO, 2000, and OECD, 2005 and 2006b). These are also the outcome indicators of choice in many of the studies that have assessed the link between education and health outcomes and economic growth. A limitation of these variables is that they may not capture the full range of outcomes pursued with education and health spending.

B. Data Envelopment Analysis

The framework for analyzing the relationship between spending and outcomes derives from the empirical analysis of production efficiency. This framework provides a method for assessing the relative efficiency with which production units convert input items into product items (i.e., technical efficiency). It is based on a straightforward concept of dominance (see Figure 3): as unit A achieves the same number or more product items than unit E with fewer input items, unit A is more efficient than (i.e., dominates) unit E. By the same token, unit E is also less efficient than units B, C, and D. The difference between the input items used by unit A and E can be used as a measure of the inefficiency of unit E relative to unit A. (Alternatively, this could be measured by the difference in production items.)

Figure 3.
Figure 3.

Efficiency and the Best-Practice Frontier

Citation: IMF Working Papers 2007, 263; 10.5089/9781451868265.001.A001

The most efficient units in a sample provide the parameters for an initial estimate of the best-practice frontier. One of the most common ways for determining the best-practice (or production possibility) frontier is DEA (a more detailed discussion of DEA can be found in Zhu, 2003). The best-practice frontier is illustrated in Figure 3 by the solid line that connects the best-practice units A, B, C, and F. Because these are the most efficient units in the sample, they are assigned an efficiency score of 1. The efficiency scores of the less efficient units (D and E) depend on their distance to the best-practice frontier (the lower the efficiency score, the less efficient the unit). Several measures of the distance to the frontier can be used. Here we adopt the Farrell input-oriented efficiency score and we only assess cases with one input and one production item. In this simple case, it is straightforward to calculate the efficiency score for unit E as the ratio of the number of inputs needed at a minimum (i.e., at the best-practice frontier) to achieve its level of production (this is 0.2, see Figure 3) and the number of inputs actually used by unit E (1.0). In other words, the efficiency score of unit E is 0.2. This can be interpreted as an indicator of the cost savings that could be achieved from efficiency enhancement. That is, if unit E was fully efficient, it would only need 20 percent of the inputs it is using to achieve the same production level.

Simple DEA estimation produces upward-biased estimates of the efficiency scores which need to be corrected. In this paper, spending is taken as the input, measures of education achievement and health status are used to measure production, and OECD countries are the production units. However, estimating the best-practice frontier for the OECD countries from the observations of education and health spending and outcomes is subject to bias, for which a correction needs to be made. This bias stems from the fact that since we only observe a sub-sample of the possible outcomes representing all feasible combinations of spending and outcomes, we do not know the exact position of the best-practice frontier. Suppose, for example, that the inputs in Figure 3 represent education spending and the product item PISA test scores. Also, suppose that education spending or PISA test scores were initially not observed for country B. Then the best-practice frontier would be drawn through the countries A, C, and F, following the dashed line between the observations for countries A and C. However, suppose the observation for country B becomes available. Then, the best-practice frontier would shift outward to the line that connects countries A, B, C, and F. It is straightforward to see that, as a general principle, as more information becomes available about the feasible production combinations, the best-practice frontier may shift outward but cannot move inward. This one-sided error means that estimating the best-practice frontier with a finite sample is subject to a bias. Since efficiency scores are measured in relation to the frontier, the estimation of scores is subject to the same bias in finite samples (i.e., efficiency is overestimated unless a correction is made for the bias).

Corrections are made for the estimation bias in the best-practice frontier and efficiency scores through bootstrapping. This paper uses a method proposed by Simar and Wilson (2000), and is based on the assumption that the frontier that envelops all possible production combinations of input and product items is smooth. A key issue is how quickly the estimated efficiency scores converge to their unbiased true values if the sample of observations is expanded.28 In the case of one input and one product item the convergence rate is fast enough to yield acceptable estimates of efficiency scores and build confidence intervals. Table 3 in the main text presents rankings for the G7 countries relative to the OECD based on the point estimates of the input-oriented bias-corrected efficiency scores.

As an alternative to DEA, efficiency scores could be estimated using stochastic estimation methods (e.g., WHO, 2000, and Greene, 2004). The main advantage of DEA is that it is parsimonious in its assumptions for deriving the best-practice frontier—the key assumptions are that all units have access to the same set of production technologies, free disposal of inputs and outputs, and convexity of the set of production possibilities. Drawbacks of DEA include sensitivity to measurement error and to sample selection bias. These drawbacks are less pronounced in stochastic estimation methods, but these may be more sensitive to misspecification as the efficiency estimates are obtained from a model that describes the production process. This model needs to capture variations in efficiency between countries as well as other sources of heterogeneity that may explain differences in the relationship between spending and outcomes across countries (e.g., Greene, 2004). When assessing the impact of spending on outcomes across countries, the dilemma faced with stochastic frontier estimation is that a large number of observations is needed for a reasonable estimation of the various model parameters and efficiency, but that as the number of countries in the sample is expanded, they also become more heterogeneous (especially when low-income countries are mixed with OECD countries) which leads to further challenges in the specification of the model. Specification of the model is also complicated by a lack of understanding about what drives differences across countries in their relationship between spending and outcomes. Taking this into account, DEA seems most suited for the undertaking of the analysis of this paper.

The analysis of the link between spending and outcomes in education and health is complicated by the fact that this relationship is indirect. Spending has no direct impact on outcomes. Rather, spending translates into intermediate inputs (e.g., school buildings and service delivery contracts with physicians) which are combined to produce intermediate outputs (e.g., school instruction hours and patient-doctor consultations). These intermediate outputs in turn are used to promote better outcomes.

The DEA results can be disaggregated to determine at which stage of the spending process inefficiencies arise. In particular, the paper attempts to disaggregate what happens in the stage from spending to intermediate outputs (cost effectiveness) and from intermediate outputs to outcomes (system efficiency). This is done by comparing overall spending efficiency (the overall measure of efficiency from spending to outcomes as discussed above) and system efficiency (see Table 4) as follows. First, an index of intermediate outputs in health and education (by level) is created. For secondary education, the intermediate output indicator is the index of countries’ average ranks for the levels of student-teacher ratios, number of computers per student, and average hours of compulsory instruction time in secondary education. For tertiary education, the intermediary output indicator is the countries’ rank for the level of the student-teacher ratio. For health, the intermediate output indicator is the index of the countries’ average ranks for number of hospital beds, physicians per capita, health workers per capita, immunizations, and doctors’ consultations. Countries are indexed such that the lower the rank (the higher the number) of a country’s index, the more intermediate outputs the country uses. Second, efficiency scores are calculated, using the intermediate output index as an input and associated outcomes (mean and within-country variation in PISA mathematics scores and graduation rates for secondary education; graduation rates for tertiary education; and infant, child, and maternal mortality rates, as well as HALE and standardized death rates for health). Third, the resulting system efficiency rankings are averaged, expressed as a ratio of the average OECD ranking, and compared with similar ratios for spending efficiency.

C. Second-Stage Analysis with Efficiency Scores

The second-stage analysis attempts to explain observed differences in the relationship between spending and outcomes. This will allow us to answer the question of how countries can strengthen the link between spending and outcomes. The second stage uses correlations and regressions of efficiency scores and associated factors to explain observed differences in the relationship between spending and outcomes.

Given the close relationship of spending and outcomes with income levels, correlations of efficiency scores and associated factors are conditional on GDP. GDP per capita has a very strong negative impact on efficiency scores (see footnote 22). Many of the factors that are associated with efficiency are also closely related to income level. In order to avoid attribution of factors whose effects on the variation in efficiency cannot be separated from the effect of GDP, only GDP per capita and factors unrelated to GDP per capita are considered in the second-stage analysis of this paper.

Regression analysis provides insight into the relative importance of factors that are associated with efficiency. The standard approach for the second stage is to regress the DEA efficiency scores on a set of explanatory variables (e.g., see Afonso and St. Aubyn, 2006). Since the Farrell input-oriented efficiency scores are limited to between 0 and 1, sometimes a truncated regression framework is used. However, this yields biased coefficient estimates due to serial correlation between the observations (the source of this is the same as the bias in the estimation of efficiency scores discussed above—see Simar and Wilson, 2007). Instead, Simar and Wilson (2007) propose another set of bootstrap procedures. However, these still require a larger sample size than we have available (less than 28 observations, given that some countries for which we have efficiency scores will drop out of the regressions because of nonavailability of data for the independent variables).

To circumvent the problem of small sample size, the regression model is reformulated in differences. Instead of estimating the DEA efficiency scores for each country, the model proposed here estimates differences in efficiency scores between countries. The model only considers differences in efficiency scores between countries where one country has a higher level of outcomes and lower spending than the other country (i.e., one country dominates the other under DEA). This can be illustrated in Figure 3. Country E is dominated by countries A, B, C, and D, but not by country F. Achieving better results would mean for country E that it moves in a left-and-upward direction, into the area outlined by the dotted vertical and horizontal lines. For country E to achieve such an improvement, it would be important to know what drives the differences in its efficiency score from the scores of countries A–D.29 The difference with country F, on the other hand, is not very relevant for country E in its quest for better outcomes with lower spending, because moving toward country F would imply an increase in spending. Writing the model in differences in efficiency scores instead of levels of the scores boosts the sample size considerably.30 For the example illustrated in Figure 3, country E is represented by four observations in the new data set (the differences in efficiency scores with countries A, B, C, and D). Separate regressions are estimated for the various education and health spending and outcome indicator combinations.

The regression coefficients are estimated using a bootstrap method to address serial correlation. The difference regression is susceptible to the same type of bias and related serial correlation as the level regressions, although the bias may now be positive as well as negative depending on which country in the pair has the larger bias. The method used here is equivalent to algorithm 1 of Simar and Wilson (2007), which they developed to address the effects of serial correlation on coefficient estimates in second-stage regressions of efficiency score levels. It makes use of Monte Carlo draws of error terms in the regression for the bootstrap. Their algorithm was adjusted to the regression model in differences by making a straightforward change to ensure that the Monte Carlo draws of the error terms were taken from a distribution with the correct truncation for our regression model in differences.31

Cluster correction further helps reduce the correlation problem, but the model still suffers from incorrect standard errors due to serial correlation. The cluster correction relaxes the independence assumption and allows dependence within clusters (all differences with the dominated country for each dominated country) but not across clusters. Specifying cluster correction affects the estimated standard errors and variance-covariance matrix of the estimators, but not the estimated coefficients. However, because of the difference specification the sample-wide serial correlation remains and limits inference due to inaccurate standard errors.

The regression results are presented in Tables 13 and 14. They are very similar to the findings from the correlation analysis in Tables 11 and 12 (although it is striking to see spending on in-patient care emerge with a negative sign across the board in the regressions of health efficiency while in the pairwise correlations it is not highly associated with efficiency). However, because of the limitations of the regression due to inaccurate standard errors, the analysis of the paper focuses on the second-stage correlations.

Appendix II. Tables and Figures

Table 5.

Links Between Economic Growth and Spending and Outcomes in Education and Health

article image
article image
article image
article image
article image
Table 6.

Education and Health Spending

(PPP dollars)

article image
Sources: OECD Education at a Glance 2006, www.oecd.org/edu/eag2006; and OECD Health Data 2006, www.ecosante.fr.

Data are for 2003.

Data are the average for 1998–2001.

Only average spending is available for primary and secondary education.

Excluding Mexico and Turkey (because of outlying data).

Table 7.

Outcome Indicators in Education 1/

article image
Sources: OECD Education at a Glance 2006, www.oecd.org/edu/eag2006; and OECD PISA, http://pisaweb.acer.edu.au/oecd_2003/oecd_pisa_data_s1.html.

The PISA test was given in 2003, upper secondary graduation rate data are for 2004, and the tertiary graduation rate is averaged for available years during 1999–2004. The graduation rate is calculated in gross terms as the number of graduates over the total population of the age group representing the typical age at graduation.

Ratio of the test score at the 25th percentile of the score ranking by student and the 75th percentile.

Excluding Mexico and Turkey (because of outlying data).

Table 8.

Outcome Indicators in Health 1/

article image
Source: OECD Health Data 2006, www.ecosante.fr.

Data for healthy life expectancy are the average for 2001–02, standardized death rates are the average for 2000–02, infant mortality are the average for 2003–04, child mortality are for 2004, and maternal mortality are the average for 2002–03.

Excluding Mexico and Turkey (because of outlying data).

Table 9.

Intermediate Output Indicators in Education 1/

article image
Source: OECD Education at a Glance 2006, www.oecd.org/edu/eag2006.

Intermediary output data are from 2004, except secondary school enrollment which is the average for 1999–2003 and the number of computers per student in secondary education which is for 2003.

Average of lower and upper secondary.

For students aged 7–14 years.

Excluding Mexico and Turkey (because of outlying data).

Table 10.

Intermediate Output Indicators in Health 1/

article image
Source: OECD Health Data 2006, www.ecosante.fr.

Intermediary output data are the average of available data for 2001–04, except for hospital beds which is the average for 2000–02, health worker density which is the average for 2000–03, immunization which is the average for 2003–04, and doctors’ consultations which is the average for 2002–03.

Physicians, nurses, and midwives.

Excluding Mexico and Turkey (because of outlying data).

Table 11.

Correlations of Bias-Corrected Efficiency Scores and Associated Factors for Secondary and Tertiary Overall Education Spending 1/

article image
Sources: OECD Education at a Glance 2006, www.oecd.org/edu/eag2006; the World Bank’s World Development Indicators, http://devdata.worldbank.org/wdi2006/contents/index2.htm; OECD PISA, http://pisaweb.acer.edu.au/oecd_2003/oecd_pisa_data_s1.html; UNESCO Institute for Statistics, http://stats.uis.unesco.org; and IMF staff calculations.

Includes only pairwise correlations between efficiency scores and factors that are not significantly correlated with GDP (and GDP itself). Double ++ (– –) indicates that the efficiency score is positively (negatively) correlated with the associated factor at the 5 percent significance level. Single + (–) indicates that the efficiency score is positively (negatively) correlated with the associated factor at the 10 percent significance level. N.a. signifies that the associated factor is not available.

Table 12.

Correlations of Bias-Corrected Efficiency Scores and Associated Factors for Public Health Education 1/

article image
Sources: OECD Health Data 2006, www.ecosante.fr; the World Bank’s World Development Indicators, http://devdata.worldbank.org/wdi2006/contents/index2.htm; and IMF staff calculations.

Includes only pairwise correlations between efficiency scores and factors that are not significantly correlated with GDP(and GDP itself). Double ++ (– –) indicates that the efficiency score is positively (negatively) correlated with the associated factor at the 5 percent significance level. Single + (–) indicates that the efficiency score is positively (negatively) correlated with the associated factor at the 10 percent significance level. N.a. signifies that the associated factor is not available.

Table 13.

Regression Results for Overall Education Spending Efficiency Scores 1/

article image
Sources: OECD Education at a Glance 2006, www.oecd.org/edu/eag2006; the World Bank’s World Development Indicators, http://devdata.worldbank.org/wdi2006/contents/index2.htm; OECD PISA, http://pisaweb.acer.edu.au/oecd_2003/oecd_pisa_data_s1.html; UNESCO Institute for Statistics, http://stats.uis.unesco.org; and IMF staff calculations.

Bootstrapped truncated regressions with lower bound 0 and upper bound 1 suppressing the constant term and with robust standard errors (in parenthesis) and clustering on each dominated country. The model is formulated in differences in the efficiency scores of pairs of countries where one country has a higher level of outcomes and lower spending than the other country (i.e., one country dominates the other in DEA). ** denotes significance at the 5 percent confidence level.

In regression without intercept, R2 measures the proportion of the variability in the dependent variable about the origin explained by the regression.

Table 14.

Regression Results for Public Health Spending Efficiency Scores 1/

article image
Sources: OECD Health Data 2006; www.ecosante.fr; the World Bank’s World Development Indicators, http://devdata.worldbank.org/wdi2006/contents/index2.htm; and IMF staff calculations.

Bootstrapped truncated regressions with lower bound 0 and upper bound 1 suppressing the constant term and with robust standard errors (in parenthesis) and clustering on each dominated country. The model is formulated in differences in the efficiency scores of pairs of countries where one country has a higher level of outcomes and lower spending than the other country (i.e., one country dominates the other in DEA).** denotes significance at the 5 percent confidence level.

In regression without intercept, R2 measures the proportion of the variability in the dependent variable about the origin explained by the regression.

Figure 4.
Figure 4.

Secondary Education Spending and Average PISA Mathematics Scores 1/

Citation: IMF Working Papers 2007, 263; 10.5089/9781451868265.001.A001

Sources: OECD Education at a Glance 2006, www.oecd.org/edu/eag2006; OECD PISA, http://pisaweb.acer.edu.au/oecd_2003/oecd_pisa_data_s1.html; and IMF staff calculations.1/ The line connects countries with the highest observed efficiency and depicts the best-practice frontier unadjusted for estimation bias (see Appendix I).
Figure 5.
Figure 5.

Secondary Education Spending and the Distribution of PISA Mathematics Scores 1/

Citation: IMF Working Papers 2007, 263; 10.5089/9781451868265.001.A001

Sources: OECD Education at a Glance 2006, www.oecd.org/edu/eag2006; OECD PISA, http://pisaweb.acer.edu.au/oecd_2003/oecd_pisa_data_s1.html; and IMF staff calculations.1/ The distribution of PISA test scores is calculated as the ratio of the average scores for the top quartile and for the bottom quartile of test scores. The line connects countries with the highest observed efficiency and depicts the best-practice frontier unadjusted for estimation bias (see Appendix I).
Figure 6.
Figure 6.

Secondary Education Spending and Upper Secondary Graduation Rates 1/

Citation: IMF Working Papers 2007, 263; 10.5089/9781451868265.001.A001

Sources: OECD Education at a Glance 2006, www.oecd.org/edu/eag2006; and IMF staff calculations.1/ The line connects countries with the highest observed efficiency and depicts the best-practice frontier unadjusted for estimation bias (see Appendix I).
Figure 7.
Figure 7.

Tertiary Education Spending and Tertiary Graduation Rates 1/

Citation: IMF Working Papers 2007, 263; 10.5089/9781451868265.001.A001

Sources: OECD Education at a Glance 2006, www.oecd.org/edu/eag2006; and IMF staff calculations.1/ The line connects countries with the highest observed efficiency and depicts the best-practice frontier unadjusted for estimation bias (see Appendix I).
Figure 8.
Figure 8.

Public Health Spending and HALE 1/

Citation: IMF Working Papers 2007, 263; 10.5089/9781451868265.001.A001

Sources: OECD Health Data 2006, www.ecosante.fr; and IMF staff calculations.1/ The line connects countries with the highest observed efficiency and depicts the best-practice frontier unadjusted for estimation bias (see Appendix I).
Figure 9.
Figure 9.

Public Health Spending and Standardized Death Rates 1/

Citation: IMF Working Papers 2007, 263; 10.5089/9781451868265.001.A001

Sources: OECD Health Data 2006, www.ecosante.fr; and IMF staff calculations.1/ The line connects countries with the highest observed efficiency and depicts the best-practice frontier unadjusted for estimation bias (see Appendix I).
Figure 10.
Figure 10.

Public Health Spending and Infant Mortality 1/

Citation: IMF Working Papers 2007, 263; 10.5089/9781451868265.001.A001

Sources: OECD Health Data 2006, www.ecosante.fr; and IMF staff calculations.1/ The line connects countries with the highest observed efficiency and depicts the best-practice frontier unadjusted for estimation bias (see Appendix I).
Figure 11.
Figure 11.

Public Health Spending and Child Mortality 1/

Citation: IMF Working Papers 2007, 263; 10.5089/9781451868265.001.A001

Sources: OECD Health Data 2006, www.ecosante.fr; and IMF staff calculations.1/ The line connects countries with the highest observed efficiency and depicts the best-practice frontier unadjusted for estimation bias (see Appendix I).
Figure 12.
Figure 12.

Public Health Spending and Maternal Mortality 1/

Citation: IMF Working Papers 2007, 263; 10.5089/9781451868265.001.A001

Sources: OECD Health Data 2006, www.ecosante.fr; and IMF staff calculations.1/ The line connects countries with the highest observed efficiency and depicts the best-practice frontier unadjusted for estimation bias (see Appendix I).
Figure 13.
Figure 13.

Teacher Salary in Secondary Education and GDP

Citation: IMF Working Papers 2007, 263; 10.5089/9781451868265.001.A001

Sources: UNESCO Institute for Statistics, http://stats.uis.unesco.org; and the World Bank’s World Development Indicators, http://devdata.worldbank.org/wdi2006/contents/index2.htm

References

  • Afonso, António, and Miguel St. Aubyn, 2004, “Non-Parametric Approaches to Education and Health Expenditure Efficiency in OECD Countries,” mimeo (Lisbon: Technical University of Lisbon).

    • Search Google Scholar
    • Export Citation
  • Afonso, António, and Miguel St. Aubyn, 2006, “Cross-Country Efficiency of Secondary Education Provision: A Semi-Parametric Analysis with Non-Discretionary Inputs,” Economic Modelling, Vol. 23 (May), pp. 47691.

    • Search Google Scholar
    • Export Citation
  • Baldacci, Emanuele, Benedict Clements, Sanjeev Gupta, and Qiang Cui, 2004, “Social Spending, Human Capital, and Growth in Developing Countries: Implications for Achieving the MDGs,” IMF Working Paper 04/217 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation
  • Barro, Robert J., and Jong-Wha Lee, 1994, “Sources of Economic Growth,” Carnegie-Rochester Conference Series on Public Policy, Vol. 40, pp. 146.

    • Search Google Scholar
    • Export Citation
  • Barro, Robert J., and Xavier Sala-i-Martin, 1995, Economic Growth (New York: McGraw Hill).

  • Bennett, Jan, 2003, “Investment in Population Health in Five OECD Countries,” OECD Health Working Papers, No. 2 (Paris: Organization for Economic Co-operation and Development).

    • Search Google Scholar
    • Export Citation
  • Besley, Timothy, Miguel Gouveia, and Jacques Drèze, 1994, “Alternative Systems of Health Care Provision,” Economic Policy, Vol. 9, No. 19 (October), pp. 199258.

    • Search Google Scholar
    • Export Citation
  • Blankenau, William, 2005, “Public Schooling, College Subsidies and Growth,” Journal of Economic Dynamics and Control, Vol. 29, No. 3 (March), pp. 487507.

    • Search Google Scholar
    • Export Citation
  • Blau, Francine D., 1996, “Symposium on Primary and Secondary Education,” The Journal of Economic Perspectives, Vol. 10, No. 4, pp. 38.

    • Search Google Scholar
    • Export Citation
  • Bloom, David E., David Canning, and Jaypee Sevilla, 2003, “The Effect of Health on Economic Growth: A Production Function Approach,” World Development, Vol. 32, No. 1, pp. 113.

    • Search Google Scholar
    • Export Citation
  • Bloom, David E., and David Canning, 2005, “Health and Economic Growth: Reconciling the Micro and Macro Evidence,” mimeo (Cambridge, MA: Harvard School of Public Health).

    • Search Google Scholar
    • Export Citation
  • Boucekkine, Raouf, David de la Croix, and Omar Licandro, 2002, “Vintage Human Capital, Demographic Trends, and Endogenous Growth,” Journal of Economic Theory, Vol. 104, pp. 34075.

    • Search Google Scholar
    • Export Citation
  • Coulombe, Serge, Jean François Tremblay, and Sylvie Marchand, 2004, “Literacy Scores, Human Capital and Growth Across Fourteen OECD Countries,” International Adult Literary Survey Monograph Series (Ottawa: Statistics Canada).

    • Search Google Scholar
    • Export Citation
  • Cutler, David M., 2002, “Equality, Efficiency, and Market Fundamentals: The Dynamics of International Medical-Care Reform,” Journal of Economic Literature, Vol. 40 (September), pp. 881906.

    • Search Google Scholar
    • Export Citation
  • De la Fuente, Angel, and Rafael Doménech, 2006, “Human Capital in Growth Regressions: How Much Difference Does Data Quality Make? An Update and Further Results”, Journal of the European Economic Association, Vol. 4, pp. 136.

    • Search Google Scholar
    • Export Citation
  • Deaton, Angus, 2003, “Health, Inequality, and Economic Development,” Journal of Economic Literature, Vol. 41, No. 1 (March), pp. 11358.

    • Search Google Scholar
    • Export Citation
  • Doppelhofer, Gernot, Ronald I. Miller, and Xavier Sala-i-Martin, 2004, “Determinants of Long-Term Growth: A Bayesian Averaging of Classical Estimates (BACE) Approach,” The American Economic Review, Vol. 94, No. 4 (September), pp. 81335.

    • Search Google Scholar
    • Export Citation
  • Filmer, Deon, Jeffrey S. Hammer, and Lant Pritchett, 2000, “Weak Links in the Chain: A Diagnosis of Health Policy in Poor Countries,” World Bank Research Observer, Vol. 15 (August), pp. 199224.

    • Search Google Scholar
    • Export Citation
  • Färe, Rolf, Shawna Grosskopf, Björn Lindgren, and Jean-Pierrre Poullier, 1997, “Productivity Growth in Health-Care Delivery,” Medical Care. Vol. 35, No. 4, pp. 35466.

    • Search Google Scholar
    • Export Citation
  • Glewwe, Paul, and Hanan G. Jacoby, 2004, “Economic Growth and the Demand for Education: Is There A Wealth Effect?Journal of Development Economics, Vol. 74, pp. 3351.

    • Search Google Scholar
    • Export Citation
  • Glied, Sherry, 2003, “Health Care Costs: On the Rise Again,” The Journal of Economic Perspectives, Vol. 17, No. 2, pp. 12548.

  • Gradstein, Mark, and Moshe Justman, 2002, “Education, Social Cohesion, and Economic Growth,” The American Economic Review, Vol. 92, No. 4 (September), pp. 1192204.

    • Search Google Scholar
    • Export Citation
  • Greene, William, 2004. “Distinguishing Between Heterogeneity and Inefficiency: Stochastic Frontier Analysis of the World Health Organization’s Panel Data on National Health Care Systems,” Health Economics, Vol. 13, pp. 95980.

    • Search Google Scholar
    • Export Citation
  • Greenwald, Rob, Larry V. Hedges, and Richard D. Laine, 1996, “The Effect of School Resources on Student Achievement,” Review of Educational Research, Vol. 66, No 3, pp. 36196.

    • Search Google Scholar
    • Export Citation
  • Grogger, Jeffrey, and Derek Neal, 2000, “Further Evidence on the Effects of Catholic Secondary Schooling,” Brookings-Wharton Papers on Urban Affairs, pp. 151201.

    • Search Google Scholar
    • Export Citation
  • Gupta, Sanjeev, and Marijn Verhoeven, 2001, “The Efficiency of Government Expenditure: Experiences from Africa,” Journal of Policy Modeling, Vol. 23 (May), pp. 43367.

    • Search Google Scholar
    • Export Citation
  • Gyimah-Brempong, Kwabena, and Mark Wilson, 2004, “Health Human Capital and Economic Growth in Sub-Saharan African and OECD Countries,” The Quarterly Review of Economics and Finance, Vol. 44, pp. 296320.

    • Search Google Scholar
    • Export Citation
  • Hanushek, Eric A., and Dennis D. Kimko, 2000, “Schooling, Labor-Force Quality, and the Growth of Nations,” The American Economic Review, Vol. 90 No. 5, pp. 1184208.

    • Search Google Scholar
    • Export Citation
  • Hanushek, Eric A., 2002, “Publicly Provided Education,” in Handbook of the Economics of Education, ed. by A.J. Auerback and M. Feldstein (Amsterdam: Elsevier, 4 th ed.).

    • Search Google Scholar
    • Export Citation
  • Hoxby, Caroline Minter, 1996, “How Teachers’ Unions Affect Education Production,” The Quarterly Journal of Economics, Vol. 111, No. 3, pp. 671718.

    • Search Google Scholar
    • Export Citation
  • Jamison, Eliot A., Dean T. Jamison, and Eric A. Hanushek, 2006, “The Effects of Education Quality on Income Growth and Mortality Decline,” mimeo (Stanford, CA: Stanford University).

    • Search Google Scholar
    • Export Citation
  • Kneller, Richard, Michael F. Bleaney, and Gemmell Norman, 1999, “Fiscal Policy and Growth: Evidence from OECD Countries,” Journal of Public Economics, Vol. 74, pp. 17190.

    • Search Google Scholar
    • Export Citation
  • Krueger, Dirk, and Krishna B. Kumar, 2004, “Skill-Specific Rather Than General Education: A Reason for US-Europe Growth Differences?Journal of Economic Growth, Vol. 9, No. 2, pp. 167207.

    • Search Google Scholar
    • Export Citation
  • Lowry, Robert C., 2004, “Markets, Governance, and University Priorities: Evidence on Undergraduate Education and Research,” Economics of Governance, Vol. 5, pp. 2951.

    • Search Google Scholar
    • Export Citation
  • Lucas, Robert E., Jr., 1988, “On the Mechanics of Economic Development,” Journal of Monetary Economics, Vol. 22 (July), pp. 342.

  • McMahon, Walter W., 1998, “Education and Growth in East Asia,” Economics of Education Review, Vol. 17, No. 2, pp. 15972.

  • Milesi-Ferretti, Gian Maria, and Nouriel Roubini, 1998, “Growth Effects of Income and Consumption Taxes,” Journal of Money, Credit and Banking, Vol. 30, No. 4, pp. 72144.

    • Search Google Scholar
    • Export Citation
  • Neal, Derek, 2002, “How Vouchers Could Change the Market for Education,” The Journal of Economic Perspectives, Vol. 16, No. 4, pp. 2544.

    • Search Google Scholar
    • Export Citation
  • Newhouse, Joseph P., 1992, “Medical Care Costs: How Much Welfare Loss?The Journal of Economic Perspectives, Vol. 6, No. 3, pp. 321.

    • Search Google Scholar
    • Export Citation
  • Nixon, John, and Philippe Ulmann, 2006, “The Relationship Between Health Care Expenditure and Health Outcomes. Evidence and Caveats for A Causal Link,” The European Journal of Health Economics, Vol. 7, pp. 718.

    • Search Google Scholar
    • Export Citation
  • Nordhaus, William, D. 2006, “Baumol’s Disease: A Macroeconomic Perspective,” Working Paper 12218, National Bureau of Economic Research (Cambridge, MA: NBER).

    • Search Google Scholar
    • Export Citation
  • OECD, 2005, Health at a Glance (Paris: Organization for Economic Co-operation and Development).

  • OECD, 2006a, Education at a Glance (Paris: Organization for Economic Co-operation and Development).

  • OECD, 2006b, “Working Party No. 1 on Macroeconomic and Structural Policy Analysis. Public Spending Efficiency in Primary and Secondary Education: Performance Indicators,” Economics Department, ECO/CPE/WP1(2006)15 (Paris: Organization for Economic Co-operation and Development).

    • Search Google Scholar
    • Export Citation
  • Or, Zeynep, 2000, “Determinants of Health Outcomes in Industrialized Countries: A Pooled, Cross-Country, Time-Series Analysis,” in Economic Studies No. 30 (Paris: Organization for Economic Cooperation and Development).

    • Search Google Scholar
    • Export Citation
  • Papagni, Erasmo, 2006, “Household Borrowing Constraints, Fertility Dynamics, and Economic Growth,” Journal of Economic Dynamics and Control, Vol. 30, pp. 2754.

    • Search Google Scholar
    • Export Citation
  • Pauly, Mark V., 1995, “When Does Curbing Health Costs Really Help the Economy?Health Affairs, Vol. 14, pp. 6882.

  • Petrakis, P.E., and D. Stamatakis, 2002, “Growth and Educational Levels: A Comparative Analysis,” Economics of Education Review, Vol. 21, pp. 51321.

    • Search Google Scholar
    • Export Citation
  • Poot, Jacques, 2000, “A Synthesis of Empirical Research on the Impact of Government on Long-Run Growth,” Growth and Change, Vol. 31, No. 4, pp. 51646.

    • Search Google Scholar
    • Export Citation
  • Pritchett, Lant, 2001, “Where Has All the Education Gone?The World Bank Economic Review, Vol. 15, No. 3, pp. 36791.

  • Reinhardt, Uwe E., Peter S. Hussey, and Gerard F Anderson, 2004, “U.S. Health Care Spending In An International Context: Why Is U.S. Spending So High, And Can We Afford It?Health Affairs, Vol. 23, No. 3, pp. 1025.

    • Search Google Scholar
    • Export Citation
  • Rivkin, Steven G., Eric A. Hanushek, and John F. Kain, 2005, “Teachers, Schools, and Academic Achievement,” Econometrica, Vol. 73, No. 2, pp. 41758.

    • Search Google Scholar
    • Export Citation
  • Romer, Paul, 1990, “Endogenous Technological Change,” Journal of Political Economy, Vol. 89, No. 5, pp. S71S102.

  • Simar, Léopold, and Paul W. Wilson, 2000, “A General Methodology for Bootstrapping in Non-Parametric Frontier Models,” Journal of Applied Statistics, Vol. 27, pp. 779802.

    • Search Google Scholar
    • Export Citation
  • Simar, Léopold, and Paul W. Wilson, 2007, “Estimation and Inference in Two-Stage, Semi-Parametric Models of Production Processes,” Journal of Econometrics, Vol. 136, pp. 3164.

    • Search Google Scholar
    • Export Citation
  • Wilson, Paul W., 2005, “Efficiency in Education Production among PISA Countries, with Emphasis on Transition Economies,” mimeo (Austin, TX: Department of Economics, University of Texas).

    • Search Google Scholar
    • Export Citation
  • Woessmann, Ludger, 2000, “Schooling Resources, Educational Institutions, and Student Performance: The International Evidence,” Kiel Working Paper No. 983 (Kiel: Institute of World Economics).

    • Search Google Scholar