Human capital and productivity

Human capital and productivity Angel de la Fuente* Instituto de Análisis Económico (CSIC) January 2011 Abstract This paper surveys the empirical literature on human capital and productivity and summarizes the results of my own work on the subject. On balance, the available evidence suggests that investment in education has a positive, significant and sizable effect on productivity growth. According to my estimates, moreover, the social returns to investment in human capital are higher than those on physical capital in most EU countries and in many regions of Spain. Keywords: human capital, productivity, growth, measurement error JEL Classification: O40, I20, O30, C19 _____________________ * This paper has been prepared for a special issue of Nordic Economic Policy Review on productivity and competitiveness. It draws heavily on joint work with R. Doménech and other coathors that has been partially financed by the European Commission, the OECD, the research department of BBVA and the Spanish Ministry of Science and Innovation (through grant no. ECO2008-04837/ECON and its predecessors). 1 1. Introduction One of the most distinctive features of the "new" theories of economic growth has been the broadening of the relevant concept of capital. While traditional neoclassical models focused almost exclusively on the accumulation of physical capital (equipment and structures), more recent contributions have attributed increasing importance to the accumulation of human capital and productive knowledge and to the interaction between these two factors. The empirical evidence, however, has not always been consistent with the new theoretical models. In the case of human capital, in particular, a number of studies have produced discouraging results. Educational variables are often not significant or even enter with the "wrong" sign in growth regressions, particularly when these are estimated using differenced specifications or panel techniques. The accumulation of negative results in the literature during the second half of the nineties generated a growing skepticism about the role of schooling in the growth process and even led some authors (see in particular Pritchett, 2001) to seriously consider the reasons why educational investment may fail to contribute to productivity growth. Many researchers in the area, however, held on to more optimistic views. They (we) argue that the negative results found in certain studies can be explained by technical problems that have a lot to do with the difficulty of measuring human capital correctly. This article provides a quick review of several strands of a literature that provides evidence in support of this hypothesis and a more detailed summary of my own work on the subject. The paper is organized as follows. Section 2 sketches the theoretical framework that has guided most studies of the contribution of education to economic growth, reviews the main empirical specifications used in the literature and briefly discusses some of its key results. Section 3 highlights some of the shortcomings of the cross-country schooling data sets most commonly used in the early empirical literature, discusses their implications for attempts to estimate the contribution of education to productivity growth and introduces a convenient indicator of data quality that can be used to quantify the information content of alternative schooling series and to estimate the size of the bias caused by measurement error. Section 4 summarizes the main findings of a series of papers I have written mostly in collaboration with Rafael Doménech. In them, we construct new attainment series for 21 OECD countries and for the regions of Spain, develop measures of the information content of these and other schooling series and estimate a variety of growth specifications for both samples. Using these results we have also constructed a set of metaestimates of the coefficient of human capital in an aggregate Cobb-Douglas production function that correct for the downward bias generated by measurement error. With this correction, we find that the contribution of investment in human capital to productivity growth is positive, quite sizable and implies rather respectable social returns that, for most territories in our two samples, compare quite favorably with those on physical capital. 2 2. Human capital and economic growth: an overview of the literature Theoretical models of human capital and growth are built around the hypothesis that the knowledge and skills embodied in humans directly raise productivity and increase an economy's ability to develop and to adopt new technologies. In order to explore its implications and open the way for its empirical testing, this basic hypothesis is generally formalized in one of two (not mutually exclusive) ways. The simplest one involves introducing the stock of human capital (which will be denoted by H throughout this paper) as an additional input in an otherwise standard aggregate production function linking national or regional output to the stocks of productive inputs (generally employment and physical capital) and to an index of technical efficiency or total factor productivity (TFP). The second possibility is to include H in the model as a determinant of the rate of technological progress (that is, the rate of growth of TFP). This involves specifying a technical progress function that may include as additional arguments some indicator of investment in R&D and a measure of the “technological gap”, that is, of the distance between each country’s productive technology and the best practice frontier. In what follows, I will refer to the first of these links between human capital and productivity as level effects (because the stock of human capital has a direct impact on the level of output) and to the second one as rate effects (because H affects the growth rate of output through TFP). Box 1 develops a simple model of growth with human capital that formalizes the preceding discussion and incorporates both effects. Box 1: A descriptive model of human capital and growth ____________________________________________________________ This box develops a simple model of growth and human capital that has two components: an aggregate production function and a technical progress function. The production function will be assumed to be of the Cobb-Douglas type: (B.1) Yit = Ait Kit α kHit α hLit α l where Yit denotes the aggregate output of country i at time t, Lit is the level of employment, Kit the stock of physical capital, Hit the average stock of human capital per worker, generally measured by school attainment, and Ait an index of technical efficiency or total factor productivity (TFP) which summarizes the current state of the technology and, possibly, omitted factors such as geographical location, climate, institutions and endowments of natural resources. The coefficients α i (with i = k, h, l) measure the elasticity of output with respect to the stocks of the different factors. An increase of 1% in the stock of human capital per worker, for instance, would increase output by α h%, holding constant the stocks of the other factors and the level of technical efficiency. Under the standard assumption that (B.1) displays constant returns to scale in physical capital and labor while holding average attainment constant, (i.e. that α k + α l = 1), we can define a per capita production function that will relate average labor productivity to average schooling and to the stock of capital per worker. Letting Q = Y/L denote output per worker and Z = K/L the stock of capital per worker and dividing both sides of (B.1) by total employment, L, we have: (B.2) Q = AZ α kH α h ____________________________________________________________ 3 Box 1 -- continued ____________________________________________________________ The technical progress function describes the determinants of the growth rate of total factor productivity. I will assume that country i's TFP level can be written in the form: (B.3)Ait = BtXit where Bt denotes the world "technological frontier" (i.e. the maximum attainable level of efficiency in production given the current state of scientific and technological knowledge) and Xit = Ait/Bt is (an inverse indicator of) the "technological gap" between country i and the world frontier. It will be assumed that Bt grows at a constant and exogenous rate, g, and that the growth rate of Xit is given by (B.4) Δ xit = γ io - λ xit + γ Hit where xit is the log of Xit and γ io a country fixed effect that helps control for omitted variables such as R&D investment. Notice that this specification incorporates a technological diffusion or catch-up effect. If λ > 0, countries that are closer to the technological frontier will experience lower rates of TFP growth. As a result, relative TFP levels will tend to stabilize over time and their steady-state values will be partly determined by the level of schooling. ____________________________________________________________ Some recent theoretical models suggest that the accumulation of human capital may give rise to important externalities that would justify corrective public interventions. The problem arises because some of the benefits of a more educated labor force will typically "leak out" and generate output gains that cannot be appropriated in the form of higher earnings by those who undertake the relevant investment, thereby driving a wedge between the private and social returns to education. Lucas (1988), for example, suggests that the average stock of human capital at the economy-wide level increases productivity at the firm level holding the firm's own stock of human capital constant. It is also commonly assumed that the rate effects of human capital on technical progress include a large externality component because it is difficult to appropriate privately the full economic value of new ideas. Azariadis and Drazen (1990), and implicitly Lucas (1988) as well, stress that younger cohorts are likely to benefit from the knowledge and skills accumulated by their elders, thus generating potentially important intergenerational externalities that operate both at home and in school. The literature also suggests that human capital can generate more diffuse "civic" externalities, as an increase in the educational level of the population may help reduce crime rates or contribute to the development of more effective institutions. From theory to data: alternative approaches to empirical analysis Empirical studies of the effects of human capital on productivity (or more broadly, of the determinants of economic growth) have followed one of two alternative approaches. The first one involves the specification and estimation of an ad-hoc equation relating growth in total or per capita output to a set of variables that are thought to be relevant on the basis of informal theoretical considerations. The second approach is based on the estimation of a structural relation between the level of output or its growth rate and the relevant explanatory variables 4 that is derived from an explicit theoretical model built around an aggregate production function and, possibly, a technical progress function of the type described in Box 1. This basic framework for the "structural" analysis of the determinants of growth can give rise to a large number of empirical specifications. Some of the most common examples are discussed in Box 2. The production function can be estimated directly with the relevant variables expressed Box 2: Some common empirical specifications ____________________________________________________________________________________ For estimation purposes it is generally convenient to work with the production function written in logarithms or in growth rates. Using lower case letters to denote logarithms, and the combination of lower case letters and the symbol " Δ " to denote growth rates, the production function given by equation (B.1) in Box 1 yields the following two specifications: (B.5) yit = ait + α kkit + α hhit + α llit + ε it (B.6) Δ yit = Δ ait + α k Δ kit + α h Δ hit + α l Δ lit + Δε it where ε it and Δε it are stochastic disturbances. One difficulty that arises at this point is that both of these equations contain terms that are not directly observable (in particular the level of TFP, ait, or its growth rate, Δ ait). To proceed with the estimation, it is necessary to make further assumptions about the behavior of these terms. Different assumptions will generate different econometric specifications. The simplest possibility is to assume that the rate of technical progress is constant over time and across countries, i.e. that Δ ait = g for all i and t. In this case, g can be estimated as the regression constant in equation (B.6) and ait is replaced in equation (B.5) by aio + gt , where aio and g give rise to country-specific constants and a common trend respectively. An alternative and more sophisticated approach is to specify Δ ait in equation (B.6) as a function of other variables. One possible specification is the one given by the technical progress function described by equations (B.3) and (B.4) in Box 1. When data on factor stocks or their growth rates are not available (or are not considered reliable), a generalized Solow model can be used to approximate these variables in terms of observed investment rates. In such a model, long-term equilibrium values of factor ratios are simple functions of investment rates, and the behavior of these ratios away from such an equilibrium can be approximated as a function of investment rates and initial income per worker. If we are willing to assume that most countries are reasonably close to their long-run equilibria, equation (B.5) can be replaced by an equation relating output per worker to investment rates in physical and human capital. Otherwise, the relevant equation will involve the growth rate of output and it will include initial output per worker as an additional regressor in order to pick up transitional dynamics along the adjustment to the long-run equilibrium. Two rather standard specifications of the resulting steady state and convergence equations (which do not allow for rate effects) would be (B.7) qit = aio + gt + α k 1α kα h ln skit δ +g+nit + α h 1α kα h ln shit δ +g+nit and (B.8) Δ qit = g + β (aio + gt) + β ⎝ ⎜ ⎛⎠ ⎟ ⎞ α k 1α kα h ln skit δ +g+nit + α h 1α kα h ln shit δ +g+nit - β qit where q is the log of output per worker, sk and sh stand for investment in physical and human capital measured as a fraction of GDP, n for the rate of growth of employment or the labor force and δ for the rate of depreciation (which is assumed to be the same for both types of capital). The parameter β measures the speed of convergence towards the long-run equilibrium or steady state and can be shown to be a function of the degree of returns to scale in both types of capital considered jointly and of the length of the period over which we are taking observations. ____________________________________________________________________________________ 5 in levels or in growth rates when reliable data are available for the stocks of all the relevant production inputs. Alternatively, its parameters can be recovered from other specifications (convergence and steady state equations) that are designed for estimation when only data on investment flows (rather than factor stocks) are available. These specifications can be derived from a production function by replacing factor stocks or their growth rates by convenient approximations in terms of investment rates using the procedure developed by Mankiw, Romer and Weil (1992) within the framework of a generalized Solow model with several types of capital. Empirical evidence: a bird’s eye view A large number of empirical studies have analyzed the relationship between human capital and economic growth using the different specifications I have outlined above.1 Early attempts in this direction, by and large, produced positive results that tended to confirm economists’ traditionally optimistic views regarding the macroeconomic payoff to investment in education. Landau (1983), Baumol et al (1989), Barro (1991) and Mankiw, Romer and Weil (1992), among many others, find that a variety of educational indicators have the expected positive effect on output growth. During the second half of the nineties, however, a new round of empirical papers produced rather disappointing results on the effects of schooling on aggregate productivity. Unlike most previous studies, most of these papers used pooled quinquennial data and relied on either panel techniques or the use of differenced specifications to control for unobserved country heterogeneity. In this setting, educational variables are often found to be insignificant or even enter with the "wrong" sign in growth regressions. (See for instance Benhabib and Spiegel (1994), Islam (1995), Caselli, Esquivel and Lefort (1996) and Pritchett (2001)). While some researchers have been willing to take such counterintuitive results at face value, many others have been rather skeptical (see for instance Barro (1997)). These authors have tended to attribute negative results on schooling and growth to various econometric and specification problems and to poor data quality. Measurement error, in particular, has been widely recognized to be a potentially important problem for two reasons. First, because the series of average years of schooling commonly used in the literature are likely to contain a lot of noise and, second, because years of schooling can be expected to be a very imperfect measure of skills in any event. The first problem, in addition, is likely to be particularly important in a panel setting, where parameter estimates rely heavily on the time-series variation of the data, because measurement error arising from changes in classification and data collection criteria tends to generate a lot of spurious volatility in the schooling series that will make it difficult to identify its contribution to productivity growth. Although it is too early for the issue to have been conclusively settled, my reading of the evidence accumulated over the last decade or so is optimistic. We have good reasons to believe 6 that the negative results found in some of the previous literature can indeed be largely attributed to deficiencies in the human capital data used in earlier studies. Papers that make use of improved data sets on attainment or allow for measurement error strongly suggest that increases in average schooling do indeed have a substantial impact on productivity growth. Results are generally even stronger and sharper when direct measures of skill levels are used to proxy for human capital, suggesting that improvements in the quality of schooling can have an even larger effect on aggregate output than increases in its quantity. The wave of negative results on the growth effects of education that arrived in the second half of the nineties is clearly associated with the introduction of panel data techniques. While early studies relied on cross-section data (working with a single observation per country that described average behavior over a period of several decades), studies in the second group have used several observations per country, taken over shorter periods, and have employed panel techniques or differenced specifications that basically eliminate the cross-section variation in the data before proceeding to the estimation. While theses estimation techniques have the important advantage that they control for unobservable differences across countries, they also have some disadvantages. Perhaps the main one is that they are more sensitive to measurement error in the data as errors tend to be greater in the time-series than in the cross-section dimension because they tend to cancel out when we work with averages over long periods. This suggests, as I have already noted, that a possible explanation of the negative results obtained in panel data studies has to do with the poor quality of the schooling data that have been used until recently in the growth literature. As we will see below, most of the earlier databases on international schooling levels contain large amounts of noise that can be traced back to various inconsistencies of the primary data used to construct them. The existence of this noise induces a downward bias in the estimation of the coefficients that measure the impact of human capital (that is, a tendency to underestimate their values) because it generates spurious variability in the stock of human capital that is not matched by proportional changes in the level of productivity. A number of recent studies provide evidence that is consistent with this hypothesis. Starting with Krueger and Lindhal (K&L 2001), some authors have constructed statistical indicators of the informational content of different attainment series (reliability ratios) that can be used to calculate the likely size of the attenuation bias and conclude that the value of this ratio is sufficiently low to explain the lack of significance of educational indicators in previous studies. Other authors, including Cohen and Soto (2007), de la Fuente and Doménech (D&D, 2001a and b and 2006) and Barro and Lee (2010), have tried to improve the signal-to-noise ratio in the schooling series by exploiting new sources of information and introducing different corrections. They find that the results concerning the impact of education on growth improve considerably when these revised series are used. I will return to these issues in much greater detail in the following two sections. 1 For a more detailed survey of the relevant literature, see section 3 of the Appendix to de la Fuente and Ciccone (2003). 7 Another interesting development is the use of cross-country data on direct measures of skill which may provide better proxies for the stock of human capital than years of schooling. While such data are still rather scarce, some recent papers suggest that this is likely to be a very fruitful line of research. Hanushek and several coauthors2 construct indicators of labor force quality using mean country scores in a number of international student achievement tests in mathematics, science and reading, while Coulombe et al (2004) use data drawn from IALS, an international study on the skill level of the adult population conducted by the OECD and Statistics Canada. In both cases, the results of growth regressions point to even larger output effects than those obtained using even revised attainment data. While not entirely free of problems, these estimates do suggest that the quality of education is likely to be at least as important as its quantity and that the return to improvements in schooling quality could be extraordinarily high, for not only are their expected benefits large, but the relevant costs will generally be much lower than those of increasing attainment for they do not involve a further sacrifice of student time and output. 3. Cross-country data on schooling: problems and consequences Most governments gather information on a number of educational indicators through population censuses, labor force surveys and specialized studies and surveys. Various international organizations collect these data and compile comparative statistics that provide easily accessible and (supposedly) homogeneous information for a large number of countries. The most comprehensive regular source of international educational statistics is UNESCO's Statistical Yearbook. This publication provides reasonably complete yearly time series on school enrollment rates by level of education for most countries in the world and contains some data on the educational attainment of the adult population, government expenditures on education, teacher/pupil ratios and other variables of interest.3 The UNESCO enrollment series have been used in a large number of empirical studies of the link between education and productivity. In many cases this choice reflects the easy availability and broad coverage of these data rather than their theoretical suitability for the purpose of the study. Enrollment rates can probably be considered an acceptable, although imperfect, proxy for the flow of educational investment but they are not necessarily a good indicator of the existing stock of human capital since average educational attainment (which is often the more interesting variable from a theoretical point of view) responds to investment flows only gradually and with a very considerable lag. In an attempt to remedy these shortcomings, a number of researchers have constructed data sets that attempt to measure directly the educational stock embodied in the population or labor force of large samples of countries during a period of several decades. These data sets have generally been constructed by combining the available data on attainment levels with the 2 See among others Hanushek and Kimko (2000) and Hanushek and Wossman (2008 and 2009). 3 Other useful sources include the UN's Demographic Yearbook, which also reports educational attainment levels by age group and, in recent years, the OECD's annual report on education in its member countries (Education at a Glance), which contains a great deal of information about the inputs and outputs of the educational system. 8 UNESCO enrollment figures to obtain series of average years of schooling and of the composition of the population or labor force by educational level. The best known early attempts in this line are the work of Kyriacou (1991), the first versions of the Barro and Lee data set (1993, 1996 and 2000) and the series constructed by World Bank researchers (Lau, Jamison and Louat (1991), Lau, Bhalla and Louat (1991) and Nehru, Swanson and Dubey (NSD, 1995). In de la Fuente and Doménech (D&D, 2006) we briefly review the methodology used in these studies and compare the different data sets with each other, focusing in particular on the OECD, where the quality of the available information should in principle be better than in developing countries. The analysis of the different series reveals very significant discrepancies among them in terms of the relative positions of many countries and implausible estimates or time profiles for at least some of them. Although the various studies generally coincide when comparisons are made across broad regions (e.g. the OECD vs. LDCs in various geographical areas), the discrepancies are very important when we focus on the group of industrialized economies. Another cause for concern is that existing estimates often display extremely large changes in attainment levels over periods as short as five years (particularly at the secondary and tertiary levels). To a large extent, these problems have their origin in the deficiencies of the underlying primary data. As Behraman and Rosenzweig (1994) have noted, there are good reasons to worry about the accuracy and consistency of UNESCO's data on both attainment levels and enrollment rates. Our analysis of the different schooling data sets confirms this diagnostic and suggests that many of the problems detected in these data can be traced back to shortcomings of the primary statistics, which do not seem to be consistent, across countries or over time, in their treatment of vocational and technical training and other courses of study, and reflect at times the number of people who have started a certain level of education and, at others, those who have completed it. Attenuation bias and a measure of data quality The poor quality of cross-country schooling data is a serious concern because it tends to obscure the relationship between the variables of interest and generates a tendency to underestimate the impact of human capital on productivity. To understand the origin of the attenuation bias caused by measurement error, assume that the level of productivity, Q, is a linear function of the stock of human capital, H, given by (1) Q = bH + u where u is a random disturbance. Given this relationship, variations in the stock of human capital, H, will induce changes in Q, and the relative magnitude of the variations in these two variables will allow us to estimate the value of the coefficient b. Now, if H is measured with error, that is, if what we observe is not H itself but a noisy proxy for it, say (2) P = H + ε , 15 Figure 1: Estimated α s vs. SUR reliability ratio -0.25 0 0.25 0.5 0.75 1 -0.25 0 0.25 0.5 0.75 1 catchup fe diff levels pred SUR reliability ratio catch-up differences fixed effects levels ˆ ! s estimate without measurement error In de la Fuente and Doménech (2002 and 2006) we use a procedure of this type to obtain consistent meta-estimates of α s. Working with the three linear specifications estimated above (that is, with all of them except for the catch-up model) and with different assumptions about the nature of measurement error (and in particular about its correlation across data sets and with the remaining explanatory variables in the model), we obtain different estimates of α s which are then adjusted to account for the possible bias generated by the fact that we are working with the average attainment of the entire population rather than that of employed workers. In this manner, we generate a rather broad range of possible values for α s. Under what we consider to be the most plausible assumptions, our results imply values of α s between 0.70 and 0.80. It is worth noting that our smallest lower bound for this parameter is 0.57. This is almost twice as large as Mankiw, Romer and Weil's (1992) estimate of 1/3, which could probably have been considered a consensus value for this parameter in the early 90s and came then to be seen as too optimistic in the light of the negative results in the literature reviewed in an earlier section. Our estimates, by contrast, point to a considerably higher figure and suggest that investment in human capital is an important growth factor whose effects have been underestimated in previous studies as a result of the poor quality of schooling data. 16 4.4. Regional results for Spain Our analysis of Spanish regional data yields qualitatively similar conclusions regarding the contribution of schooling to productivity. In de la Fuente and Doménech (2008), we estimate a catch-up specification using biennial data for the Spanish regions covering the period 1965-95. The specification is identical to the one estimated above for the OECD sample except in that physical capital is now disaggregated into two components, one of which is the stock of productive infrastructures (transport and water supply networks and urban structures). As a proxy for the stock of human capital, we use our own census-based attainment series and an alternative estimate of average years of schooling constructed using Mas et al's (MPUSS, 2002) series on the breakdown of the working-age population by attainment level which is, in turn, based on Labor Force Survey data. Table 3: Growth estimates with alternative schooling series and specifications ______________________________________________ [1] [2] [3] [4] S data from: MPUSS D&D MPUSS D&D α s -0.013 0.835 -0.013 0.835 (0.11) (2.04) (0.11) (4.13) adj. R2 0.749 0.753 0.757 0.763 regional effects all all signif. signif. ______________________________________________ Notes: - All equations include period dummies. - White's heteroscedasticity-consistent t ratios in parentheses below each coefficient. - The employment ratio has been dropped from the equation due to its lack of significance. The estimates of the human capital parameter obtained with both schooling series are reported in Table 3. All equations contain period dummies. Equations [1] and [2] contain a full set of regional dummies, and equations [3] and [4] retain only those regional fixed effects that were significant in the first iteration. Inspection of the table reveals two interesting results regarding the coefficient of human capital ( α s). First, this parameter goes from being non-significant when the MPUSS (2002) data are used to having a large and significant value with our attainment series. This result is consistent with our estimates of the information content of the two series, as the relevant reliability ratio is 0.900 for our data and only 0.035 for MPUSS's attainment series when both are measured in logarithmic differences. Second, our estimate of α s for the Spanish regions (0.835) is higher than those reported above for the OECD data using a similar specification (0.540 with a full set of country dummies and 0.394 when only the significant fixed effects are retained). Again, the explanation seems to lie at least partly in the information content of the different data sets (the relevant reliability ratio for the cross-country attainment series in D&D (2006) was 0.246). In fact, our estimate of α s using Spanish regional data lies well within the range of the meta-estimates obtained by D&D (2006) for OECD countries after correcting for measurement error. 17 4.5. Some implications The results summarized in the previous sections have some important implications. If average schooling enters the production function with a coefficient within the range of values we have estimated, differences in school attainment are one of the key sources of productivity differentials across both the OECD countries and the regions of Spain and investment in education yields a rather substantial return that, in most territories, compares quite favorably with that available from alternative investment opportunities. Figure 2: Contribution of schooling to relative productivity in 1995 -10% 0% 10% 20% PV Ca Ma Na Ri Ar Va Cnt Cn Ba Mu CL As CM An Ex Ga schooling other factors -20% - Key: An = Andalucia; Ar = Aragón; As = Asturias; Ba = Baleares; Cn = Canarias; Cnt = Cantabria; CL = Castilla y León; CM = Castilla la Mancha; Cat = Cataluña; Va = Valencia; Ex = Extremadura; Ga = Galicia; Ma = Madrid; Mu = Murcia; Na = Navarra; PV = País Vasco; Ri = Rioja. While I don't have the space that I would need to go into detail, I don't want to close this section without at least a brief elaboration on these two statements. Using the estimates given in Table 3 and the underlying data, in D&D (2008) we have calculated the contribution of schooling to the relative productivity of the Spanish regions, defined as log real output per job measured in deviations from the (unweighted) sample average of the same variable. Figure 2 shows the decomposition of each region's relative productivity into a schooling-induced component and a residual that captures the joint impact of all other factors. Using regression weights to average the different regions, we find that the share of schooling in relative productivity was 40% in 1995 -- that is, that for the typical Spanish region schooling accounts for 4/10 of the productivity gap with the sample average.11 A similar calculation for the OECD sample implies a share of schooling in relative productivity of 30%. 11 We define the relative productivity of region i (qreli) as the difference between the region’s log output per employed worker and the average value of the same variable in the sample. The contribution of human capital to relative productivity (csi) is obtained multiplying the coefficient of this factor, α s , by the relative level of schooling (measured in log differences with the geometric sample mean). After constructing these two variables for each region, we estimate a regression of the form csi = a ∗ qreli + ei 18 Our estimates also imply that the social returns to education are quite respectable.12 Combining our results on the productivity effects of human capital with rough estimates of its impact on employment and with data on educational expenditure, we estimate social rates of return ranging from 10.1% to 12.6% in Spain and from 8.3% to 11.5% in the EU15.13 In both samples, these returns compare quite favorably in most cases with those available from alternative investment opportunities. This suggests that in most of these territories a marginal reallocation of investment resources in favor of education would be socially desirable. 5. Conclusion Academic economists have traditionally been rather optimistic about the contribution of education to economic development and have often assigned to the accumulation of human capital a central role in formal models, particularly in the recent literature on endogenous growth. The results of early empirical studies on the determinants of economic growth have been largely consistent with this view. During the second half of the nineties, however, a new round of empirical papers produced rather disappointing results on the subject that sparked a lively controversy in the literature between "skeptics" and "believers" in the salutary effects of schooling on aggregate productivity growth. This paper contains a selective and rather partisan review of some of the relevant literature. After setting the stage, it focuses on a problem (the poor quality of cross country schooling data) that may help explain the discouraging results found in some influential studies, on possible ways to overcome it, and on what happens when this is done. I have argued that, due to various deficiencies of the primary data, the schooling series used in the early empirical literature on growth and human capital contain a considerable amount of noise that generates a very substantial downward bias in estimates of the parameter that measures the contribution of educational attainment to productivity. This conclusion is based on the estimation of a statistical indicator of the information content of the schooling series most commonly used in the literature. It is also reinforced by the finding of a clear tendency for human capital coefficients to rise and become more precise as the information content of the schooling data increases. When this relationship is extrapolated to construct estimates of the value of the schooling coefficient that would be obtained in the absence of measurement error, the exercise suggests that the true value of the elasticity of output with respect to the stock of human capital is almost certainly no lower than 0.60 --that is, around twice as high as the most optimistic estimate of reference in the earlier literature on growth and human capital. where ei is a random disturbance. The coefficient obtained in this manner, a ≅ csi/qreli , measures the fraction of the observed productivity differential that can be attributed to human capital in the sample as a whole. 12 The social rate of return to schooling is defined as the discount rate that equates the present value of the increases in output induced by a marginal increase in average attainment to the present value of the explicit and opportunity costs of schooling. For further details on how this magnitude can be estimated, see de la Fuente (2003). 13 For additional details, see de la Fuente and Doménech (2008) and de la Fuente (2003). 19 References Azariadis, C. and A. Drazen (1990). "Threshold Externalities in Economic Development." Quarterly Journal of Economics, May, pp. 501-26. Barro, R. (1991). "Economic Growth in a Cross Section of Countries." Quarterly Journal of Economics CVI (2), pp. 407-43. Barro, R. (1997). Determinants of economic growth: a cross-country empirical study. Lionel Robbins Lectures, MIT Press, Cambridge, Mass. Barro, R. and J-W. Lee (1993). "International Comparisons of Educational Attainment." Journal of Monetary Economics 32, pp. 363-94. Barro, R. and J-W. Lee (1996). "International Measures of Schooling Years and Schooling Quality." American Economic Review 86(2), Papers and Proceedings, pp. 218-23. Barro, R. and J-W. Lee (2000). "International data on educational attainment, updates and implications." NBER Working Paper no. 7911. (http://www.nber.org/papers/w7911). Barro, R. and J.-W. Lee (2010). “A new data set of educational attainment in the world, 19502010.” NBER Working Paper no. 15902. Cambridge, Mass. Baumol, W., S. A. Batey Blackman and E. Wolf (1989). Productivity and American Leadership: the Long View. MIT Press. Behrman, J. and M. Rosenzweig (1994). "Caveat emptor: cross-country data on education and the labor force." Journal of Development Economics 44, pp. 147-71. Benhabib, J. and M. Spiegel (1994). "The Role of Human Capital in Economic Development: Evidence from Aggregate Cross-Country Data." Journal of Monetary Economics 34, pp. 143-73. Caselli, F., G. Esquivel and F. Lefort (1996) "Reopening the convergence debate: a new look at cross-country growth empirics." Journal of Economic Growth 1(3), pp. 363-89. Cohen, D. and M. Soto (2007). "Growth and human capital: good data, good results." Journal of Economic Growth 12(1), pp. 51–76. Coulombe, S., J. F. Tremblay and S. Marchand (2004). "Literacy scores, human capital and growth across fourteen OECD countries." Statistics Canada. de la Fuente, A. (2003). "Human capital in a global and knowledge-based economy, part II: assessment at the EU country level." European Commission, DG for Employment and Social Affairs, Brussels. http://digital.csic.es/bitstream/10261/1844/1/57603.pdf de la Fuente, A. and A. Ciccone (2003). Human capital and growth in a global and knowledge-based economy. European Commission, DG for Employment and Social Affairs. Office for official publications of the European Communities, Luxembourg. http://ideas.repec.org/p/aub/autbar/576.03.html de la Fuente, A. and R. Doménech (D&D, 2000). "Human capital in growth regressions: how much difference does data quality make?" OECD Economics Department Working Paper no. 262, Paris. de la Fuente, A. and R. Doménech (D&D, 2001a). "Schooling data, technological diffusion and the neoclassical model." American Economic Review 91(2), Papers and Proceedings, pp. 323-27. de la Fuente, A. and R. Doménech (D&D, 2001b). "Educational attainment in the OECD, 196090." CEPR Discussion Paper no. 3390. 20 de la Fuente, A. and R. Doménech (D&D, 2002). "Human capital in growth regressions: how much difference does data quality make? An update and further results." CEPR Discussion Paper no. 3587. (http://ideas.repec.org/e/pde52.html#papers) de la Fuente, A. and R. Doménech (D&D, 2006). "Human capital in growth regressions: how much difference does data quality make?" Journal of the European Economic Association 4(1), pp. 1-36. de la Fuente, A. and R. Doménech (2008). “Human capital, growth and inequality in the Spanish regions.” In U. Stierle, M. Stierle, F. Jennings and A. Kuah, editors, Regional Economic Policy in Europe. Challenges for Theory, Empirics and Normative Interventions, pp. 1544. Edward Elgar. Hanushek, E. and D. Kimko (2000). "Schooling, labor-force quality and the growth of nations." American Economic Review 90(5), pp. 1184-208. Hanushek, E. and L. Woessmann (2008). "The role of cognitive skills in economic development." Journal of Economic Literature 46(3), pp. 607-68. Hanushek, E. and L. Wössman (2009). “Do better schools lead to more growth? Cognitive skills, economic outcomes and causation.” NBER Working Paper no. 14633, Cambridge, Mass. Islam, N. (1995). "Growth empirics: a panel data approach." Quarterly Journal of Economics 110, pp. 1127-70. Krueger, A. and M. Lindahl (2001). "Education for growth: why and for whom?" Journal of Economic Literature XXXIX, pp. 1101-36. Kyriacou, G. (1991). "Level and Growth Effects of Human Capital, A Cross-Country Study of the Convergence Hypothesis." Mimeo, NYU. Landau, D. (1983). "Government Expenditure and Economic Growth: a Cross-Country Study." Southern Economic Journal, January, pp. 783-92. Lau, L, D. Jamison, and F. Louat (1991). "Education and Productivity in Developing Countries: an Aggregate Production Function Approach." Report no. WPS 612, the World Bank. Lau, L., S. Bhalla and F. Louat (1991). "Human and physical capital stock in developing countries: construction of data and trends." Draft mimeo, World Development Report, World Bank. Lucas, R. (1988). "On the Mechanics of Economic Development." Journal of Monetary Economics 22, pp. 3-42. Mankiw, G., D. Romer and D. Weil (1992). "A Contribution to the Empirics of Economic Growth." Quartely Journal of Economics, pp. 407-37. Mas. M., F. Pérez, E. Uriel, L. Serrano y A. Soler (2002). "Capital humano, series 1964-2001," en Capital humano y actividad económica (CD Rom). Fundación Bancaja, Valencia. Nehru, V., E. Swanson and A. Dubey (1995). "A New Database on Human Capital Stocks in Developing and Industrial Countries: Sources, Methodology and Results." Journal of Development Economics, 46, pp. 379-401. Pritchett, L. (2001). "Where has all the education gone?" World Bank Economic Review 15(3), pp. 367-91.