Education and Italian regional development

Abstract

In this paper, we study the connection between growth and human capital in a convergence regression for the panel of Italian regions. We include measures of average primary, secondary and tertiary education. We find that increased education seems to contribute to growth only in the South. Decomposing total schooling into its three constituent parts, we find that only primary education in the South seems to be important, while tertiary education seems to have a negative impact on regional growth. Our main results are robust to the inclusion of additional variables in the regression analysis and the use of an IV estimator. Overall, this study suggests that Italian growth benefited from the elimination of illiteracy in the South, mainly in the 60s. It also suggests a possible relationship between the level of development of an economy and returns to different levels of education, with Italian regions still far from being able to capture the positive returns from higher levels of education.

Introduction

Differences in human capital endowments and their rates of investment have long been recognised as an important element in explaining observed GDP gaps, with theoretical models often emphasising the presence of externalities to education. In particular, a number of growth models imply that public returns to education exceed private returns, often assuming that high average levels of human capital throughout the economy increase the productivity of any given worker.¹ But a higher level of education could be associated with many productivity improving factors not captured by private returns.² In contrast, traditional screening models of education generate the exact opposite result, stressing the possibility of negative externalities. These models are usually associated with higher levels of education. In this case, education only confers credentials used in the labour market to select able workers. But we may think of other mechanisms implying the possibility of negative externalities,³ affecting lower levels of education as well as high.⁴

Note that individual-based micro-analyses will be useless as a guide to public policy when there are important externalities because such analyses will measure only private returns to education. Conversely, macro-studies consider the data of direct interest, namely the returns at the level of the economy.⁵ However, differently from the micro-econometric evidence on return to education, empirical macro-studies show puzzling results as often find that education is not strongly associated with per capita income growth.

It has been claimed that the main problem causing the observed lack of empirical support is that most growth regressions, while using large international data sets, incorrectly impose a single coefficient and thus equal returns on schooling among different countries. This problem is likely to arise when the quality of education is influenced by differences in educational institutions. In this case, one explanation of the observed low returns to education found in large international data sets is that national statistics may not be comparable.⁶ Moreover, it may well be that the quantity of education affects its quality: returns to education may be higher in more educated areas as usually predicted by growth models.⁷ In both cases, standard regressions would produce distorted estimates on education due to the presence of parameter heterogeneity and measurement error problems.

A second problem that may arise when we estimate returns to schooling is that in some cases acquisition of educational skills is not obviously linked with productivity. As noted by Schultz (1962), education may represent not only an investment for individuals but can also be considered as a consumption good and, thus, be privately valued for its own sake. But another interesting example is found in Pritchett (1996), who quotes as in 1988, 50% of university students in Saudi Arabia were studying Humanities, Religion and Theology. While this kind of degree probably represents a good credential in the Saudi Arabian job market, still does not represent an obvious acquisition of growth-enhancing skills. A related problem has been emphasised by Griliches (1997). He observes that in many countries, and especially developing countries, the public sector represents the employer of most of the skilled labour force. This fact may create three sources of distortions when we estimate returns to schooling. Firstly, the output of the Public Sector is certainly badly measured in National Accounts and, possibly, underestimated. Secondly, the literature on developing countries shows many examples where the growth of the Public Sector with the corresponding absorption⁸ of skilled labour force has not been governed by efficiency criteria. Finally, the Public Sector is not obviously an innovative sector while, as predicted by many theoretical growth models, especially shumpeterian models, educational capital is growth enhancing only when allocated in innovative activities.⁹

In this paper we investigate if, dealing with the problems described above, a standard macro-analysis of returns to education would produce significant results. To control for the first problem, we focus on a more homogeneous data set rather than the whole international sample and ask if there has been any role for human capital in the Italian regional economic development. We claim that Italian data are most suitable for a macro-study of returns to education: differently from most regional data sets, the Italian regions are quite diverse in their endowments of human capital—among the European countries, Italy has the highest dispersion of regional education attainment¹⁰—and, since the 60s, has experienced vast increases in the average duration of education at all three levels. Secondly, the Italian regions have common institutions so that, in large part, the data represent a controlled experiment in ceteris paribus variation of labour force educational endowments in a developed economy. Further, there is a large literature showing a clear duality in the Italian economy between the developed North-Centre and the less-developed South, suggesting the presence of two convergence clubs. These two clubs are also characterised by the presence of homogeneous educational institutions in both areas, together with substantial differences in human capital endowments. In fact, with respect to the less-developed South, the richer North-Centre is characterised by larger stocks of human capital. Therefore, this is an ideal sample to test the relationship between quantity and returns to education: allowing for parameter heterogeneity in the two clubs, we analyse if returns to education have been different in these two areas of the country considered separately. Conversely, given the quality and the level of disaggregation of data, the problem of the link between acquisition of educational skills and productivity is certainly more difficult to deal with. Nevertheless, following Griliches, we control in our empirical analysis for the presence of a relatively large Public Sector and check if this may possibly affects our estimates on return to schooling.

We have census data on average years of schooling and primary, secondary and tertiary school attainments distinguished for gender and use information on enrolment rates to construct a yearly data set. Thus, we follow the standard development literature that predicts larger externalities for educated women than men and investigate if differences in male and female education have different impacts on the development of Italian regions. Finally, we ask if different levels of education produce different impacts on growth. In fact, due to their emphasis on the role of technology, most of the theoretical growth models expect that higher levels of educational attainments act more powerfully on growth than, say, primary school. This prediction contradicts micro-econometric evidence, where returns to investments in primary education are usually estimated as the largest.¹¹