Regional convergence in Italy, 1891–2001: testing human and social capital

Abstract

The article aims to present and discuss estimates of levels of human and social capital in Italy’s regions over the long term, i.e., roughly from the second half of the nineteenth century up to the present day. The results are linked to newly available evidence for regional value added in order to begin to form an explanatory hypothesis of long-term regional inequality in Italy: convergence in value added per capita is tested in light of the neoclassical exogenous growth approach, which incorporates human capital and social capital as conditioning variables into a long-term production function. In contrast with conventional wisdom (e.g. Putnam 1993), we find that social capital was not a significant predictor of economic growth in post-Unification Italy: It grew in importance only in the last decades. Conversely, human capital was more important in the first half of the twentieth century. Results suggest that there was not one single conditioning variable over the long run, thus supporting the view that, in different periods, conditioning variables can be determined by technological regimes.

Appendix: regional estimates of human capital, social capital and fertility: sources and methods

1.1 Human capital

The weighting scheme of the human capital composite indicator has been illustrated in Sect. 3. It was devised with the aim of reducing arbitrariness when choosing between the different components of the index, while allowing for the shift in importance from primary to advanced education which took place over the long run. As can be seen from Tables 8, 9, for Liberal Italy, the index of human capital is composed of literacy, with a weight decreasing from 70% in 1871 to 38% in 1911, and primary and secondary enrolment rate, with a weight growing from 30% in 1871 to 58% in 1911; the shares assigned to tertiary and higher education still are negligible. These latter began to grow significantly in the interwar period, touching one-fifth by 1951; in this year, primary and secondary enrolment rate has the lion share of the indicator, almost 70%, whereas to literacy is left a mere 13%. As around the mid of the twentieth century, the overwhelming percentage of the Italian population had become literate, these changes follow the assumption—which seems only reasonable—that differences in a minority of illiterate people no longer affected economic growth in any significant way. During the post–Second World War economic boom, compulsory education extended throughout the country and regional differences became less and less important (see again Tables 8, 9). Furthermore, as the industrialization spread and technological regimes changed, the characteristics of human capital as a factor of production arguably changed too: i.e., what mattered rather more was tertiary (non-compulsory) education attended from age 14 to 19 and later higher education, from age 19 to 24 and above. This change is captured by the human capital indicator: from 1971 to 2001, about one-third of the indicator is made up of tertiary education enrolment rate, and the share of higher education enrolment rate has grown from 10 to 22%. By 2001, when combined, these two indices total almost 60% of the human capital indicator; primary and secondary enrolment rate is around 40%, and literacy has fallen to a negligible share.

Table 8 A composite and dynamic index of human capital for Italy’s regions: components and weights (part 1)

Table 9 A composite and dynamic index of human capital for Italy’s regions: components and weights (part 2)

A serious problem arises when dealing with regional figures for university enrolment: as evident in Tables 8, 9, these tend to seriously underestimate small regions (Aosta Valley, Trentino-Alto Adige, Lucania), which did not host a university or higher education institute for most of the twentieth century. University students from small regions temporarily emigrated to the cities of the larger regions, often returning to their homes after gaining their degrees—this interregional mobility increased markedly during recent decades. However, mobility between the three macro-regions (North-West, Centre/North-East, South and islands) was much lower and, even when the flows were remarkable, namely in the mobility from the South to the North, the return rates were considerably lower, so much so that, by and large, emigrating students could be considered as effectively acquired by the host regions. For these reasons, I calculate and use a new index of university attendance, which is based on macro-regional scores and on the regional enrolment ratios of tertiary education, according to the formula: newUr = Tr/Tm*Um, where U is university attendance, T is tertiary education, r is the region and m the macro-region. In other words, I assume that each region follows the university attendance of its macro-region, proportional to its rate of tertiary education enrolment (due to lack of space, the new index is omitted, but it can be easily derived from the figures in Tables 8, 9).

It is worth stressing that we always refer to the gross enrolment ratio, i.e., the total number of students enrolled (including repeaters and other students above/below the respective age brackets) as a percentage of the official population for a given level of education. Of course, the net enrolment ratio (the students enrolled of the official age group for a given level of education, as a percentage of the corresponding population) and even more attendance rates (the number of people attending a given level of education, as a percentage of the corresponding population), which in turn can be gross or net, would be preferable; but the available sources are not suitable for any long-term and consistent quantification of these variables (cfr. A’Hearn et al. 2011, pp. 180–181),³⁰ as with most of the countries around the world. As far as we can tell, however, North–South differentials in attendance, and more in general differences between enrolment and attendance, were in the liberal age higher than later (A’Hearn et al. pp. 187–193): i.e., they decreased over the course of the twentieth century.³¹ As a consequence, a possible revised human capital indicator, which would allow for regional changes in attendance over the long run, would probably show even higher convergence than the human capital indicator here presented.

The problem of the quality of education is maybe different, regional inequality probably persisted, but the solution (or lack of it) here proposed is analogous. Out of necessity: today PISA (Programme for International Student Assessment) data, which measure the knowledge and skills of 15-year-old students for many countries including Italy, show some North–South differences, but they are available only from 2000 onwards (Nardi 2001).

In general, the use of a composite indicator tends to increase regional (North–South) differences in human capital in the last decades, as compared to alternative unvarying measures of human capital such as literacy or the total enrolment rate: It does not come as a surprise that, as illustrated in Sect. 3, the use of alternative (and arguably less questionable) measures does not invalidate the overall result of the article, i.e., that human capital became less important in the last decades (For cross-section estimates confirming this result, see also Felice 2008). The same can be said for out-of-reach more accurate measures, such as attendance rates. These are good reasons to believe that the view taken by the present article is by and large correct. However, Tables 8, 9 are intended to provide any interested reader with full information about the individual components of the index and their trends, in order to make the methodology proposed entirely transparent, and amendable.

1.2 Social capital

In this article, new estimates of social capital at the regional level are presented for two benchmark years, 1871 and 1891, whereas for other benchmarks (from 1901 onwards) the available figures by Giorgio Nuzzo are employed. In order to achieve a comprehensive and coherent long-term picture, the new benchmarks have been reconstructed through a methodology explicitly linked to Nuzzo’s. However, when compared with the benchmarks from 1901 onwards, for 1871 and 1891 there were less indicators available and thus, in order to come to consistent figures, the hypothesis had to be introduced that, for each dimension of social capital, the ratio between the observed variables and the unobserved ones in 1891 was the same as in 1901, and in 1871 the same as in 1891.

As mentioned, Nuzzo’s indicator is a simple mean of social participation, political participation and trust. Social participation is measured by an average of different non-profit institutions, i.e., those which, according to the author, effectively generated social capital (significantly, unions were excluded). For the second half of the nineteenth century, we can rely upon friendly societies, which were also used by Nuzzo for 1901: In both cases, the indicator is the number of members of such friendly societies, as a ratio to the total population. As a first step, data for 1904, 1895 and 1873 (the years for which data were available) have been extrapolated backwards in order to create the 1871, 1891 and 1901 benchmarks, via linear interpolation with the continuous compounding yearly rate. As a second step, in 1891, social participation has been estimated from the member of friendly societies in 1891 and by maintaining, for every region, the 1901 ratio of social participation/members of friendly societies, as in the equation:

$$ R_{sp} 1891 = Rmf1891\times\left( {Rsp1901/Rmf1901} \right) $$

(4)

where R is the region, sp is social participation, and mf is the number of members of friendly societies.³² Finally, this procedure has been replicated for 1871, using the 1891 estimate of social participation in place of Nuzzo’s figure for 1901. The number of friendly societies and the total amount of deposits of the banche popolari have been tested too, alone or in combination (to also include the number of members), but they turned out to be weakly correlated with Nuzzo’s figures.

Nuzzo’s indicator of political participation is an average of the densities of political non-profit institutions, of the share of voters out of the total population at different elections and of an informal indicator based on polls taken from 1993 to 2003 concerning political engagement. For 1901, the author relied only on the density of political non-profit institutions. For the second half of the nineteenth century, we lack this information, but we can avail ourselves of the use of statistics on local newspapers, an indicator in line with Putnam’s approach, where the readers of newspapers are used as a proxy of political participation. In our case, we have data about the number of local newspapers published in 1880 (1454), 1891 (1779), 1895 (1901) and 1905 (3120), which, via linear interpolation with the continuous compounding yearly rate, have been used to create 1871, 1891 and 1901 regional benchmarks; as can be seen from the figures in brackets, the numbers are relatively high, which partly comforts about the reliability of the proxy used. The rest of the procedure is analogous to the one outlined for calculating social participation, with the difference that in this case, regional data on local newspapers are linked to Nuzzo’s index of political participation.³³

Nuzzo’s indicator of trust is measured by the inverse of an average of estimates of violent criminality and of court proceedings, as well as of the share of perceived criminality as determined by polls conducted in 1995 and 2003. For the second half of the nineteenth century, we can make use of almost the same data as those used by Nuzzo for 1901. More in particular, trust is approximated through the inverse of an average of criminal and civil court proceedings in 1901–1904, 1891 and (here only criminal court proceedings) 1871.³⁴ Here too, at this point, the procedure is analogous to the one outlined for the other two dimensions, in this case the data being correlated with Nuzzo’s index of trust. For 1871, since only criminal statistics were available, these were in turn correlated with criminal statistics in 1891 and with the index of trust in 1891.

For each component, the regional data of the two-step procedure are shown in Table 10. By construction, the results are in line with Nuzzo’s benchmarks.

Table 10 The components of social capital in 1871, 1891 and 1901 (Italy = 1)

1.3 Fertility and population growth

Figures of fertility and population growth are shown in Table 11. As can be observed also by the naked eye, usually differences in fertility rates do not automatically equal differences in population growth; in fact, in all the panel models (random and fixed effects), fertility is not significant as a predictor of population growth, although it certainly has a positive effect. More specifically, Southern Italy scored fertility rates above the Italian average throughout the period, but a population growth below the average up to the 1970s, with the exception of the years 1938–1951. Although differences in mortality rates were important to some extent, the main reason for this was migration, both interregional and international, which depopulated Southern Italy through most of the period (but it came almost to a halt in the 1930s, only to resume again in the 1950s). From the 1970s, migration from the south played a much diminished role, and thus for the first time, not having reduced their gap in fertility rates, the Southern regions experienced a population growth higher than the Italian average. This difference is apparently more important than the one in social capital, when it comes to explaining the decline in the south in recent decades. It could be, however, that differences in social capital also determined differences in fertility rates, at least to some extent and perhaps until very recently. In any event, the two variables appear to be highly correlated in the last decades: Their correlation grows from 1911 to 1938 and remains significantly high at least until the 1980s.³⁵ Needless to say, this is another topic that deserves thorough consideration in further research.

Table 11 Fertility rates and population growth for Italy’s regions (Italy = 1)