Banking development, socioeconomic structure and income inequality

G Model

ARTICLE IN PRESS

JEBO-4124; No. of Pages 24

Journal of Economic Behavior & Organization xxx (2017) xxx–xxx

Contents lists available at ScienceDirect

Journal of Economic Behavior & Organization journal homepage: www.elsevier.com/locate/jebo

Banking development, socioeconomic structure and income inequality Alexandra D’Onofrio a , Raoul Minetti b,∗,1 , Pierluigi Murro c a b c

Casmef (Luiss University), Italy Michigan State University, United States Lumsa University, Italy

a r t i c l e

i n f o

Article history: Received 9 January 2017 Received in revised form 1 August 2017 Accepted 9 August 2017 Available online xxx JEL: G21 G38 O15

a b s t r a c t Using rich, unique data from the Italian local credit markets (provinces), this paper investigates the impact of local banking development on income inequality and the role of the socioeconomic structure in this link. Exploiting the Italian historical banking regulation to instrument for the local presence of bank branches, we find that local banking development mitigates income inequality. However, the finance-inequality nexus manifests itself only in relatively advanced areas. When we study the structural channels of influence, we uncover evidence that banking development can reduce inequality by affecting geographical mobility and urbanization, while it has modest effects through the development of material infrastructures and human capital. © 2017 Elsevier B.V. All rights reserved.

Keywords: Income inequality Financial development Socioeconomic structure

1. Introduction Growing inequality is one of the biggest economic challenges of our time and its causes are increasingly debated among economists and policy makers (The Economist, 2012). Financial sector development has been shown to be highly effective in promoting economic growth (Levine, 2005), thus suggesting the importance of a deeper analysis of how it can be used to alter income distribution and foster pro-poor economic growth.2 The relationship between financial development and income distribution is also important for understanding the process of economic development. Income distribution can influence savings decisions, the allocation of resources, incentives to innovate, and public policies (Aghion et al., 1999; Bagchi and Svejnar, 2015). The mechanisms whereby financial development can influence income inequality still remain the subject of an intense debate, however.

∗ Corresponding author at: Department of Economics, 486 W. Circle Drive, 110 Marshall-Adams Hall, Michigan State University, East Lansing, MI 48824-1038, United States. E-mail address: [email protected] (R. Minetti). 1 For their comments and suggestions, we especially wish to thank participants at The International Conference on Financial Development and Economic Stability (FDES-2016), jointly organized by Durham University Business School, IPAG Business School, and International Society for the Advancement of Financial Economics (ISAFE), in partnership with the Journal of Economic Behavior & Organization. We also thank several other seminar and conference participants for helpful comments and conversations. All remaining errors are ours. 2 Levine (2005) offers a comprehensive survey of theory and empirics of financial development and growth. http://dx.doi.org/10.1016/j.jebo.2017.08.006 0167-2681/© 2017 Elsevier B.V. All rights reserved.

Please cite this article in press as: D’Onofrio, A., et al., Banking development, socioeconomic structure and income inequality. J. Econ. Behav. Organ. (2017), http://dx.doi.org/10.1016/j.jebo.2017.08.006

G Model JEBO-4124; No. of Pages 24

2

ARTICLE IN PRESS A. D’Onofrio et al. / Journal of Economic Behavior & Organization xxx (2017) xxx–xxx

In this paper, we investigate the impact of local banking development on income inequality exploiting the unique empirical setting offered by the Italian banking regulation and using rich, hand-collected data from the Italian local credit markets (the 103 Italian provinces). We also take a step towards understanding whether local banking development affects income distribution by influencing salient aspects of the socioeconomic structure, such as urban structure, geographical mobility, development of material and immaterial infrastructures, and formation of human capital. Although these structural factors are viewed as fundamental drivers of income inequality, evidence on their role in the finance-inequality nexus remains scant. Italy is ideally suited for the purposes of our investigation. The Italian financial system is strongly bank-based, hence banks can play a key role in income distribution.3 In addition, the banking system is segmented across provinces with provinces exhibiting, in turn, highly heterogeneous degree of banking development in the early 2000s (the period of our investigation). Most importantly, this heterogeneous banking development can be considered as fairly exogenous as it reflects the historical regulation of the Italian banking system which imposed severe restrictions on lending and branching in the provinces from 1936 till the late 1990s.4 And, when we consider the structural channels of influence, these are often reputed to have an important role in income inequality in Italy (Ascher and Krupp, 2010; Farina and Franzini, 2015). Exploiting the heterogeneous tightness of the 1936 banking regulation to tackle the possible endogeneity of banking development, and after controlling for a battery of province-specific controls, we find that local banking development has a significant negative effect on the income Gini coefficient. A 10 percent increase of the bank branch density induces a 1.9 percent reduction in the Gini. And this finding is confirmed when using alternative measures of income inequality. However, the detected effect is not robust through the sample, suggesting the existence of non-linearities in the banking development-inequality nexus. When we split our sample according to broad geographical areas, the negative effect of banking development on income inequality is indeed significant only in the northern sub-sample. One possible explanation is that the relationship between banking development and inequality manifests itself only for a sufficiently high level of economic development, given that the North of Italy is more industrialized and rich. That is, as financial development is costly to implement, catch-up effects could start manifesting only after income crosses a certain threshold value.5 In the second part of the paper, we explore the channels whereby banking development may affect income inequality. Our focus is on the role of the socioeconomic structure, including urban structure, geographical mobility (inter-province migratory flows), development of material and immaterial infrastructures, and human capital formation. The importance of these structural factors for income inequality has been long acknowledged in the literature (see, e.g., Calderon and Serven, 2004; World Bank, 2003; Ferreira, 1995; Lopez, 2003; Behrens and Robert-Nicoud, 2014a) and yet very little is known about their role in the finance-inequality nexus. Our findings suggest that banking development has a sizeable effect on geographical mobility and urban structure, while it has only modest consequences on the formation of human capital and material infrastructures. Further, when we add the indicators of human capital and material infrastructures to the regressions, the effect of banking development on inequality remains essentially unaltered. By contrast, when we control for the indicators of geographical mobility and urban structure, these indicators tend to absorb the effect of banking development on inequality. Overall, the results suggest that urban structure and geographical mobility may constitute two important channels through which banking development shapes the income distribution. Interestingly, this appears to be true especially for the North of Italy, suggesting that the detected non-linearity in the banking development-inequality nexus might at least partly reflect the workings of an urban structure and geographical mobility channel. The remainder of the paper unfolds as follows. Section 2 reviews the literature and discusses theoretical predictions. Section 3 provides a general outlook on the history of local banking development in Italy and on the dynamics of income distribution in Italian provinces. Data and methodology are described in Sections 4 and 5, while Sections 6 and 7 present the empirical results. Section 8 concludes.

2. Theoretical predictions and prior literature Financial development can promote economic growth by mitigating financing constraints. The issue of which segments of the population benefit more from the resulting growth acceleration has not been conclusively addressed. For example, financial development could benefit the poor by expanding labor demand and creating opportunities for small and mediumsized enterprises. However, it could also generate higher profit margins for wealthy, established entrepreneurs. In this section, we relate our analysis to the previous theoretical and empirical literature on the finance-inequality nexus.

3 The capitalization of the Italian stock market is relatively low compared to other advanced economies. For example, according to World Bank data, in 2014 the ratio between the stock market capitalization and the GDP in Italy was 27%, while it was 151% in the United States. 4 Between 1936 and 1990, in Italy the number of bank branches grew by 137% versus 1228% in the United States. 5 Our findings are confirmed by employing a rolling regressions technique to show graphically the evolution of the coefficient on financial development when per capita GDP increases (see, e.g., Rousseau and Wachtel, 2002, for an analogous approach). The coefficient becomes negative and significant after a certain level of the median per capita GDP and after the financial sector has achieved a reasonably high level of development. The rolling regression graphs are available from the authors upon request.

Please cite this article in press as: D’Onofrio, A., et al., Banking development, socioeconomic structure and income inequality. J. Econ. Behav. Organ. (2017), http://dx.doi.org/10.1016/j.jebo.2017.08.006

G Model JEBO-4124; No. of Pages 24

ARTICLE IN PRESS A. D’Onofrio et al. / Journal of Economic Behavior & Organization xxx (2017) xxx–xxx

3

2.1. Theoretical predictions on finance and inequality There is a broad theoretical literature on the finance-inequality nexus (see, e.g., Seven and Coskun, 2016, for a detailed overview). Besides Kuznets’ pioneer study of inequality (Kuznets, 1955), some classical economists stressed that income inequality can channel resources towards agents with high saving propensity, boosting capital accumulation and growth (Lewis, 1954; Kaldor, 1955). Later, in the 1990s a strand of studies started to model explicitly the relationship between finance and inequality. Several studies in this strand suggest that financial development can abate income inequality and poverty by mitigating informational asymmetries and credit enforcement costs, which may be especially binding on poor households and entrepreneurs with scarce internal funds and pledgeable collateral (Banerjee and Newman, 1993; Galor and Zeira, 1993; Aghion and Bolton, 1997; Galor and Moav, 2004). The theoretical models emphasize different channels whereby financial development can reduce inequality.6 If the poor lack access to credit, they are prevented from investing in education and, hence, from seeking more remunerative jobs. Financial development may allow low-income individuals to invest in education, thus mitigating inequality (Galor and Zeira, 1993; Galor and Moav, 2004; Aghion and Bolton, 1997).7 A second channel focuses on the ability of the poor to become entrepreneurs. By ameliorating credit constraints, financial development may decrease collateral requirements and borrowing costs, fostering entrepreneurship and firm formation (Banerjee and Newman, 1993). In Matsuyama (2000) highly productive projects entail a sizeable fixed cost. Because of credit constrains, only the wealthy can afford such costs, while poor households become lenders. In Buera (2009) financial frictions inhibit entrepreneurship so that wealthy agents can found firms, while poor ones are stuck with salaried jobs. Financial development may also alter the distribution of income through an increased labor demand by firms rather than through an increased access to credit by the poor (Beck et al., 2010). The increased labor demand may benefit low-income workers. In contrast with the above models, some theories stress that the poor primarily rely on informal (e.g., family) connections to obtain funds, so that the development of the formal financial sector especially helps the rich (see, e.g., Claessens and Perotti, 2007, for a discussion). In this strand, some theoretical models predict that, while financial development broadens the opportunities of the poor, the relationship might be not linear but depend on the level of economic development. Greenwood and Jovanovic (1990) show how the interaction of financial and economic development generates an inverted U-shaped relationship between income inequality and financial development that reminds Kuznets’ hypothesis that growth may increase income inequality at early stages of development and reduce it at later stages. In particular, in Greenwood and Jovanovic (1990) financial intermediaries produce information about projects but joining them entails a fixed cost. At early stages of development, only the rich can pay this fixed cost, so that financial development exacerbates inequality. As the economy develops, the financial system becomes accessible to the poor. In a similar vein, Greenwood and Smith (1997) and Townsend and Ueda (2006) underscore that important non-linearities can occur in the financial development-inequality nexus, because the development of sophisticated financial institutions may entail sizeable fixed costs.8 2.2. Empirical findings on finance and inequality A growing empirical literature tests the impact of financial development on income distribution (see, e.g., DemirgucKunt and Levine, 2009, for a review).9 Some cross-country studies assess the relation between financial development and national Gini coefficients. Li et al. (1998) estimate a negative relationship between financial development (proxied by the M2 to GDP ratio) and the Gini coefficient in a sample of 49 countries over the 1947–1994 period. Clarke et al. (2006) employ a sample of 83 developed and developing countries over the 1960–1995 period and find that, in the long run, financial development is associated with lower inequality. Using data for the 1960–2005 period for 72 countries, Beck et al. (2007a) uncover a negative relationship between financial development and the Gini, which holds when controlling for a wide array of country-specific factors, and when using panel instrumental variable methods to control for endogeneity. They also find that financial development exerts a disproportionately positive impact on the relatively poor, i.e. the bottom quintile of the income distribution (see also Deininger and Squire, 1998; Dollar and Kraay, 2002; Ravallion, 2001; Kappel, 2010, for the impact of financial development on poverty reduction, and Claessens and Feijen, 2007, for its role in reducing undernourishment).10

6

For a review and discussion of early studies, see, e.g., Aghion and Bolton (1992). Mookherjee and Ray (2003) show that when human capital accumulation generates externalities across professions, and financial markets are underdeveloped, persistent inequality is inevitable in any steady state. 8 Furthermore, some models imply that if financial development reduces income inequality and the rich save more than the poor, this could slow down physical capital accumulation and growth and increase poverty (Bourguignon, 2001). However, in Galor and Moav (2004), at later stages of development, human capital gains importance relative to physical capital and better financial development allows poor agents to invest in education, promoting growth. 9 Although we focus on the role of local banking development in income inequality, it is worth mentioning a strand of studies that investigate the possible differential impact of banks and stock markets. Kpodar and Singh (2011) find that when institutions are weak, bank-based financial systems better mitigate poverty but, when institutions become more sophisticated, market-based financial systems become more effective in poverty reduction. Demirguc-Kunt and Levine (2008) find evidence that financial institution development exerts a stronger impact on income distribution and poverty than financial market development. 10 There are also studies that find that financial development does not reduce inequality, however. Claessens and Perotti (2007) show that in countries with historically high inequality, distortions in the institutional environment generate unequal access to finance, exacerbating inequality. Charlton and 7

Please cite this article in press as: D’Onofrio, A., et al., Banking development, socioeconomic structure and income inequality. J. Econ. Behav. Organ. (2017), http://dx.doi.org/10.1016/j.jebo.2017.08.006

G Model

ARTICLE IN PRESS

JEBO-4124; No. of Pages 24

4

A. D’Onofrio et al. / Journal of Economic Behavior & Organization xxx (2017) xxx–xxx

In cross-country studies, it can be difficult to properly address endogeneity issues, for example due to omitted institutional and regulatory factors that could influence both banking development and inequality. Recently, some studies have then performed country-level analyses to mitigate endogeneity. The focus on the local level in a single country allows to exploit within-country variation, reducing the risk of omitted variable bias. Gine and Townsend (2004) find that in Thailand the access to financial services has a negative impact on income inequality through an increase in labor demand. One natural experiment sometimes used to mitigate endogeneity problems is the change in banking regulation and banking sector policies within a country. Burgess and Pande (2005) study the effects of a state-led bank branch expansion program enacted in Indian states between 1977 and 1990. Using this natural experiment, they find that between 1977 and 1990 the state-led branch expansion into rural unbanked locations reduced rural poverty. Beck et al. (2010) find that the bank deregulation of the U.S. states tightened the income distribution by boosting incomes in the lower tail. Our paper contributes to this small group of studies by exploiting the unique empirical setting offered by the exogenous heterogeneity in bank access to credit within Italy due to the 1936 Italian banking regulation and to the historical segmentation of the local (provincial) Italian credit markets. 2.3. The links between finance and inequality A clear understanding of the socioeconomic mechanisms linking finance and inequality is still missing. As noted above, most studies emphasize human capital formation, entrepreneurship, and labor demand as the channels whereby financial development can affect inequality (see Beck et al., 2010, for a test of these mechanisms). There is little evidence on the effects that financial development can have on income inequality through its impact on the socioeconomic structure, such as urbanization and geographical mobility, material and immaterial infrastructures, and formation of human capital. Yet, a broad literature argues that these structural factors can play a key role in reducing income inequality and poverty (see, e.g., Calderon and Serven, 2004; Ferreira, 1995; Lopez, 2003; World Bank, 2003). Urbanization and migration also play an important role in the analyses of Kuznets (1955) and Williamson (1965). In Kuznets (1955), migration from low-income areas to higher-income ones can lead to an inverted-U curve (inequality first increases and then declines as the economy develops). Williamson (1965) also puts forward the idea that migratory flows can affect inequality. And a recent strand of studies show theoretically and empirically the key role that the urban structure has played in the increase in inequality in industrialized countries (see, e.g., Baum-Snow and Pavan, 2013; Behrens and Robert-Nicoud, 2014a,b. In Italy these socioeconomic structural factors are reputed to have a strong influence on income inequality, which makes Italy a good empirical laboratory to isolate their possible role in the finance-inequality nexus. Exploiting this observation, in this paper we take a step towards examining whether these structural channels matter in the relationship between local financial development and inequality. 3. Institutional background The Italian financial system is dominated by the banking sector, since the capitalization of the stock market is still relatively low. The strong importance of the banking sector makes the Italian financial system close to that of other countries of continental Europe and of Japan. In 2000, in Italy the ratio of bank credit over the GDP equalled 70.33 percent, a figure similar to that of France (81.29 percent), Belgium (77.34) and Finland (51.38).11 Most importantly, the high dependence of Italian firms on banks is analogous to that observed in other countries of continental Europe (see, e.g., De Bonis et al., 2012, for a detailed overview of the Italian banking system). Thus, although the lessons from Italy are not necessarily transferable to other countries, the analysis can also provide useful insights for the banking development-inequality nexus in other countries. The Italian banking system is segmented across local credit markets (provinces). As a geographical and administrative unit, a province can be compared with a U.S. county. Provinces represent the appropriate measure of local banking markets in Italy also according to the definition used by public authorities in their regulatory activity. For example, the Italian Antitrust Authority considers the province as the relevant market for banking activities. Moreover, at the time of deregulation of entry in the banking market in the 1990s, the Bank of Italy defined the local market as the provincial one.12 In Italy, a strong local presence of bank branches is crucial for access to credit and for financial inclusion because it is particularly difficult for households and firms to borrow in a market other than the local one (Petersen and Rajan, 2002; Guiso et al., 2004, 2013). There is indeed evidence that in Italy, due to informational disadvantages, banks entering new provincial markets suffer from higher loan default rates (Bofondi and Gobbi, 2006). The estimated effects of banking development on income distribution may suffer from endogeneity problems. That is, although we control for a battery of relevant characteristics of the provinces, unobserved factors could drive both banking

Stiglitz (2008) and Law and Tan (2009) find that in developing countries stock market liquidity does not benefit the poor. Jauch and Watzka (2015) uncover evidence that financial development increases income inequality in a sample of 138 countries. 11 On average, over the 2000–2010 period in Italy the ratio of bank credit over the GDP was 72.36 percent, similar to that of France (82.02 percent), Belgium (85.23) and Finland (84.35). 12 The deregulation process entailed various legislative acts and implementation procedures over the 1990s. Guiso et al. (2003) treat 1998–2000 as the period in which the deregulation process could be considered to be concluded.

Please cite this article in press as: D’Onofrio, A., et al., Banking development, socioeconomic structure and income inequality. J. Econ. Behav. Organ. (2017), http://dx.doi.org/10.1016/j.jebo.2017.08.006

G Model

ARTICLE IN PRESS

JEBO-4124; No. of Pages 24

A. D’Onofrio et al. / Journal of Economic Behavior & Organization xxx (2017) xxx–xxx

5

Table 1 Summary statistics.

Piemonte Valle D’aosta Lombardia Trentino-Alto Adige Veneto Friuli-Venezia Giulia Liguria Emilia-Romagna North Toscana Umbria Marche Lazio Center Abruzzo Molise Campania Puglia Basilicata Calabria Sicilia Sardegna South Italy

Gini coefficient

Theil coefficient

Branches per 1000 inhab GDP per capita

Unemployment rate

Mean

Std. Dev.

Mean

Std. Dev.

Mean

Std. Dev.

Mean

Std. Dev.

Mean

Std. Dev.

0.346 0.357 0.359 0.374 0.351 0.342 0.364 0.359 0.355 0.355 0.342 0.348 0.360 0.354 0.350 0.357 0.364 0.362 0.352 0.362 0.373 0.348 0.361 0.357

0.025 0.024 0.028 0.032 0.030 0.026 0.029 0.025 0.028 0.027 0.026 0.025 0.042 0.031 0.031 0.034 0.036 0.040 0.037 0.040 0.040 0.034 0.038 0.033

0.263 0.264 0.286 0.294 0.271 0.250 0.273 0.277 0.274 0.267 0.244 0.256 0.262 0.261 0.250 0.253 0.262 0.263 0.246 0.256 0.271 0.239 0.258 0.266

0.033 0.040 0.048 0.046 0.042 0.037 0.040 0.035 0.042 0.036 0.031 0.032 0.065 0.044 0.043 0.046 0.050 0.054 0.049 0.054 0.056 0.046 0.052 0.046

0.666 0.779 0.679 0.948 0.740 0.737 0.593 0.801 0.721 0.645 0.607 0.739 0.479 0.620 0.508 0.424 0.285 0.327 0.415 0.252 0.357 0.439 0.361 0.575

0.118 0.016 0.079 0.103 0.089 0.102 0.043 0.093 0.120 0.078 0.059 0.069 0.115 0.122 0.057 0.049 0.039 0.035 0.014 0.023 0.031 0.060 0.088 0.195

23,331 25,891 26,011 27,493 25,627 24,456 22,065 26,393 25,145 23,309 20,229 22,082 21,185 22,276 18,159 16,438 14,112 14,044 15,809 13,735 13,857 16,503 14,925 20,988

2385 1541 3267 2311 2033 2034 1950 2504 2918 2581 1330 2185 4283 3118 991 1408 1164 1324 1318 1422 1383 2133 2067 5309

5.316 5.073 4.102 3.105 4.312 4.534 6.361 3.928 4.523 5.380 5.955 5.202 9.047 6.274 7.923 9.332 13.018 13.633 12.345 14.351 16.170 12.411 13.201 7.913

1.559 3.031 1.128 0.866 1.068 0.991 2.072 1.454 1.607 1.620 1.115 1.469 1.688 2.209 1.524 1.077 3.010 2.362 1.849 3.559 4.913 2.497 4.167 4.865

development and income inequality (or the socioeconomic structure, in the second part of the analysis). And reverse causality could also affect the estimates, as the income distribution in a province might influence the local presence of bank branches. To tackle the possible endogeneity of banking development, we then construct instruments capturing exogenous restrictions on the entry of banks into local credit markets. To this end, we exploit information on the regulation of local banking markets introduced in Italy in the late 1930s. In response to the 1930–1931 banking crisis, in 1936 the Italian Government approved a Banking Law with the objective of enhancing bank stability through severe restrictions on bank entry. The Banking Law imposed strict limits on the ability of different types of banking institutions to open new branches. Guiso et al. (2003, 2004) show that this banking law deeply affected local credit markets in the decades that followed. Entry into the provincial banking markets was liberalized only during the 1990s, following also the introduction of European directives about the coordination of banking regulations across the European Union. Two elements make our instruments suitable. First, the 1930s regulation deeply affected the structure of local banking markets for several decades, and indeed at the end of the 1990s this structure closely resembled that of the 1930s (Guiso et al., 2004; Minetti et al., 2015). Second, as shown by Guiso et al. (2003, 2004), the constrictiveness of the regulation in the provinces stemmed from “historical accident” and in particular it reflected the interaction between waves of bank creation prior to 1936 and the history of Italian unification. Therefore, the tightness of the banking regulation was not correlated with economic features of the provinces. 4. Data and measurement This section describes our indicators and data for income inequality and financial development as well as the set of conditioning information. We collected data from four main sources. First, we hand-collected and elaborated data from the municipality-level database on tax revenue compiled by the Department of Finance of the Italian Ministry of Economy and Finance. We also obtained data from the Statistical Bulletin of the Bank of Italy; the province-level database of the Italian National Statistics Office (Istat); and the Geoweb Starter, a database containing local, provincial and regional statistical information produced by the Istituto Guglielmo Tagliacarne (a major Italian research institute). Since measures of income distribution by province are not available, we computed them starting from the income data. We obtained from the Department of Finance the spreadsheets on the distribution of taxable income for each of the 8056 Italian municipalities over the 2001–2011 period. For each municipality and each year, we have the frequency and the average income of 28 to 30 income classes. Employing this information, we aggregated the data assigning each municipality to its province and then computed the indicators used in the inequality literature.13 First, we derived the Gini coefficient of income distribution from the Lorenz curve. Larger values of the Gini imply greater income inequality. The Gini coefficient is equal to 0 if everyone has the same income, and it is equal to 100 if a single individual receives the income of the entire

13

Acciari and Mocetti (2013) also use tax records to study the geography of income inequality in Italy.

Please cite this article in press as: D’Onofrio, A., et al., Banking development, socioeconomic structure and income inequality. J. Econ. Behav. Organ. (2017), http://dx.doi.org/10.1016/j.jebo.2017.08.006

G Model

ARTICLE IN PRESS

JEBO-4124; No. of Pages 24

A. D’Onofrio et al. / Journal of Economic Behavior & Organization xxx (2017) xxx–xxx

6 Table 2 Correlation matrix.

Gini index Theil index Branches per 1000 inhab Gini index 1.000 0.949 Theil index Branches per 1000 -0.134 inhab 0.000 -0.147 GDP per capita 0.000 Unemployment 0.116 0.000 Agriculture (share) 0.051 0.084 -0.150 Manufacturing (share) 0.000 -0.093 Construction (share) 0.002 Trade openness -0.095 0.001

GDP per capita

Unemployment Agriculture (share)

Manufacturing Construction (share) (share)

1.000 0.064

1.000

0.030 0.097 0.001 -0.110 0.000 -0.153 0.000 0.055

0.839 0.000 -0.829 0.000 -0.493 0.000 0.558

1.000 -0.845 0.000 -0.694 0.000 0.572

1.000 0.546 0.000 -0.637

1.000 -0.497

1.000

0.062 -0.191

0.000 -0.245

0.000 -0.275

0.000 0.234

0.000 0.261

-0.240

1.000

0.000 0.084 0.005

0.000 0.524 0.000

0.000 0.646 0.000

0.000 -0.590 0.000

0.000 -0.612 0.000

0.000 0.638 0.000

-0.270 0.000

Trade openness

1.000

Note: The table reports the pairwise correlation coefficients for the variables included in the regression analysis. p-values are reported in italics.

province. As an alternative measure of income distribution, we computed the Theil index. Similar to the Gini, the Theil is also increasing in the degree of income inequality: if all individuals have the same income, the index equals 0, while it is equal to ln(n) if one individual receives all of the province’s income, where n is the number of individuals. As a robustness check, we also computed other income inequality metrics, such as the income ratio between the median and various percentiles, to study the effect of banking development not only on inequality in the population as a whole, but also on the shape of the income distribution. Results, available upon request, are qualitatively similar. The literature on the relationship between local banking development and economic growth has put forward several indicators to proxy for the development of financial intermediaries. We concentrate on bank branch density by province (number of branches normalized by the population) as a measure of the level of development of the local credit markets.14 This is a standard measure in the empirical banking literature (see, e.g., Degryse and Ongena, 2005). This measure is reputed to be a key indicator of financial inclusion and financial access, which are fundamental aspects that influence the nexus between banking development and inequality (Beck et al., 2007b). In particular, the rationale behind this variable is that it is a suitable provider-based metric of the demographic penetration of banking services in the provincial credit markets (the key credit markets for Italian households and firms) and, hence, of the accessibility of banking services. Further, bank branch density shows a large inter-provincial dispersion and captures the dimension of banking development that is largely affected by the 1936 banking regulation (see Benfratello et al., 2008, and Section 3). Finally, the structure of the Italian banking system reduces the utility to consider alternative measures of banking development such as the number of ATM per capita in the province. In Italy, ATM machines are typically located in the premises of bank branches and the correlation between the branch density and the ATM density is very high, equalling 0.91 in the 2001–2011 period. Nonetheless, as a robustness check, we also experiment with the ATM density as an indicator of local banking development (the results remain virtually unchanged).15 In line with the literature, we also use the Herfindahl-Hirschman Index (HHI) of branch concentration in the province, to study the effect of local banking competition on income inequality (see Berger et al., 2004). Data on local branch density and concentration were obtained from the Statistical Bulletin of the Bank of Italy. We use information from Istat for per capita GDP, unemployment rate, the distribution of workers among sectors, the percentage of population in small municipalities, the inter-province migratory flows and the percentage of the population with at least a secondary school degree. The index of infrastructure endowment, as well as the indices of street, railway,

14 Being located in the same province is important for banks’ ability to collect “soft” information through personal interactions between local loan officers and firms (see, e.g., Petersen and Rajan, 1994; Presbitero and Rebelotti, 2014). Furthermore, Bonaccorsi Di Patti and Gobbi (2001) show that provinces with a high number of bank branches relative to their population have a larger credit supply. 15 We also experimented with using the ratio of branches over area. Relative to the ratio of branches over population, this indicator turned out to have a lower correlation with the instruments capturing the tightness of the 1936 banking regulation. This can plausibly reflect the specific prescriptions of the 1936 banking regulation and how these prescriptions influenced the local expansion of banks’ branch networks. Indeed, in assessing the effects of this regulation, Guiso et al. (2003) focus on the ratio of branches over population. As already noted by Beck et al. (2007b), while the geographical and the demographic banking penetration exhibit some positive correlation, they tend to capture different aspects. Our measure has also advantages over a measure of credit over GDP in the province. First, it is available on a homogeneous basis for long periods while there are breaks in the series for total loans by province, due to the reclassification of “Istituti di Credito Speciale” (see also Benfratello et al., 2008). Second, as also stressed by Beck et al. (2007b), credit over GDP reflects one dimension of financial development (financial system depth) but it overlooks financial inclusion and access which are better captured by banks’ demographic penetration in local credit markets.

Please cite this article in press as: D’Onofrio, A., et al., Banking development, socioeconomic structure and income inequality. J. Econ. Behav. Organ. (2017), http://dx.doi.org/10.1016/j.jebo.2017.08.006

G Model

ARTICLE IN PRESS

JEBO-4124; No. of Pages 24

A. D’Onofrio et al. / Journal of Economic Behavior & Organization xxx (2017) xxx–xxx

7

Fig. 1. Gini coefficient: 2010. Note: Our calculations on Italian Department of Finance and Istat data. The map shows the level of the Gini coefficient, in 2010 in the 103 Italian provinces, classified in quintiles.

IT and broadband, education and cultural structures are instead obtained from the Geoweb Starter database of the Istituto Guglielmo Tagliacarne. Table 1 contains summary statistics at the regional level (a region comprises different provinces).16 The table reveals that the average income inequality, measured by the Gini index or the Theil index, is similar among the three Italian macro-areas (North, Center and South). On average, the regions located in the South of Italy exhibit a lower per capita GDP (14,925 euro, that is lower than the national average of 20,987 euro) and a higher unemployment rate (13.20 versus a national average of 7.91). Table 2 reports the correlation matrix. The Gini and the Theil index are highly positively correlated with each other. Per capita GDP shows a low correlation with the inequality indicators, suggesting that inequality is not necessarily driven by economic development. Fig. 1 displays a map of the 103 Italian provinces organized by the value of the Gini coefficient. The North features the highest number of provinces with the lowest value of the Gini coefficient. If we consider the average trend of inequality per macro areas (Fig. 2), we can notice that, although slightly declining in all areas, on average southern provinces exhibit higher level of inequality than northern ones.17 Fig. 3 plots the average (2003–2010) of branch density and the Gini index of inequality at the regional level, slightly less disaggregated than the provincial one. Income inequality tends to be lower when the level of banking development is higher (we interpolate the data using a weighted least squares method that provides a generally smooth curve). Such preliminary considerations call for a deeper analysis of the interactions between local banking development and income inequality. 5. Empirical methodology We assess the relationship between local banking development and income distribution using the following regression set-up Yp = a1 + b1 Bp + b2 Cp + εp

(1)

16 Table A.1 of the Appendix reports more detailed definitions of the variables. For matter of space, we relegate summary statistics at the provincial level to Table A.2 of the Appendix. 17 In 2009, Istat data on income distribution show that most resident households in Italy (about 58 percent) had a net income lower than the average annual amount (29,766 euro).

Please cite this article in press as: D’Onofrio, A., et al., Banking development, socioeconomic structure and income inequality. J. Econ. Behav. Organ. (2017), http://dx.doi.org/10.1016/j.jebo.2017.08.006

G Model JEBO-4124; No. of Pages 24

ARTICLE IN PRESS A. D’Onofrio et al. / Journal of Economic Behavior & Organization xxx (2017) xxx–xxx

8

Fig. 2. Gini coefficient evolution: 2003–2011. Note: Our calculations on Istat data.

Fig. 3. Local banking development and income inequality. Note: Our calculations on Bank of Italy and Istat data for Italian regions.

with p = 1, . . ., 103. Yp is a measure of income inequality (e.g., the logarithm of the Gini index or of the Theil index) in province p, Cp is a vector of province level control variables, and εp is the error term.18 The variable of interest is Bp , a measure of local financial development (e.g., the log of bank branch density) in province p. The coefficient b1 captures the impact of banking development on income inequality. In running the cross-sectional regressions, we experiment both with the year 2001 and with the average over the years 2001–2011. In addition to the cross-sectional specification, we also perform regressions using panel data for the 2001–2011 period.19 As noted, considering the local entities (provinces) of a single country enables us to reduce the risk of omitted variable bias and to implicitly control for differences in formal institutions. Nevertheless, there remains the possibility that banking development and inequality are jointly determined and that unobserved factors are correlated with both. To address these possible endogeneity issues, we use an instrumental variable (IV) approach. Let Ip be a vector of instruments that are correlated with local banking development but affect income inequality only through the banking channel. The effect of these instruments on Bp is captured by b4 in the local banking development equation: Bp = b3 Cp + b4 Ip + up

(2)

where Cp refers to the control variables in (1), Ip is the vector of instruments and up is the residual.

18 19

In robustness tests, we also use the Gini in levels, instead of its logarithm. In the panel specification, by considering 103 provinces and eleven years of data (2001–2011), we end up with 1133 province-year observations.

Please cite this article in press as: D’Onofrio, A., et al., Banking development, socioeconomic structure and income inequality. J. Econ. Behav. Organ. (2017), http://dx.doi.org/10.1016/j.jebo.2017.08.006

G Model

ARTICLE IN PRESS

JEBO-4124; No. of Pages 24

A. D’Onofrio et al. / Journal of Economic Behavior & Organization xxx (2017) xxx–xxx

9

Table 3 Banking development and income inequality: baseline estimations. Variables

(1)

(2)

(3)

(7)

(8)

Average 2001–2011

Panel 2001–2011

Year 2001

OLS Gini (log)

2SLS Gini (log)

OLS Gini (log)

2SLS Gini (log)

ArellanoBond Gini (log)

OLS Theil (log)

2SLS Theil (log)

Arellano-Bond Theil (log)

−0.049**

−0.174**

−0.051**

−0.170***

−0.150***

−0.112**

−0.411***

−0.350***

(0.021) 0.068*

(0.069) 0.131**

(0.022) 0.198***

(0.061) 0.253***

(0.055) 0.244***

(0.043) 0.238**

(0.155) 0.389***

(0.126) 0.593***

(0.040) 0.005

(0.056) −0.021

(0.061) −0.000

(0.065) −0.043

(0.058) −0.013

(0.096) 0.006

(0.130) −0.057

(0.134) −0.031

(0.015) 0.039 (0.098) −0.196**

(0.018) 0.179 (0.136) −0.222***

(0.023) −0.205 (0.137) −0.232***

(0.030) −0.119 (0.142) −0.222***

(0.010) −0.053 (0.133) −0.195***

(0.031) −0.063 (0.204) −0.193

(0.038) 0.273 (0.307) −0.255*

(0.021) −0.245 (0.297) −0.151

(0.079) −0.091

(0.071) 0.227

(0.074) −0.168

(0.074) −0.211

(0.074) −0.259

(0.156) −0.432

(0.147) 0.330

(0.158) −0.672

(0.262) −0.011*

(0.371) −0.010

(0.299) −0.011

(0.297) −0.013

(0.250) −0.009

(0.555) −0.029**

(0.811) −0.027

(0.560) −0.023

(0.006) 0.006 (0.010) 0.058*** (0.021)

(0.008) −0.003 (0.011) 0.022 (0.027)

(0.007) 0.002 (0.010) 0.062*** (0.023)

(0.008) 0.003 (0.010) 0.046** (0.023)

(0.006) −0.003 (0.010) 0.025 (0.023)

(0.014) −0.007 (0.020) 0.099** (0.043)

(0.018) −0.029 (0.023) 0.013 (0.059)

(0.014) −0.030 (0.022) 0.009 (0.053)

Year 2001

Branch density (log) Per capita GDP (log) Unemployment (log) Agriculture (share) Manufacturing (share) Construction (share) Trade openness (log) Center South Instruments Saving banks 36 N. branches 36 Year fixed effects Observations R-squared F instruments Overid p value

N 103 0.685

0.129*** (0.041) 0.005** (0.002) N 103 0.577 10.22 0.436

N 103 0.361

(4)

0.140*** (0.041) 0.004*** (0.002) N 103 0.186 12.34 0.913

(5)

Y 1133

0.999

(6)

N 103 0.435

Panel 2001–2011

0.129*** (0.041) 0.005** (0.002) N 103 0.189 10.22 0.339

Y 1133

0.998

Note: The table reports regression coefficients and associated standard errors (in parentheses). The time-span of the regressions, the dependent variables and the estimation method are reported at the top of each column. The set of instruments includes the number of bank branches in the province in 1936 (per 100,000 inhabitants) and the number of savings banks in the province in 1936 (per 100,000 inhabitants). ***Significant at 1%, **significant at 5%, *significant at 10%. The table reports the value of the R2 and the value of the F-statistics for a test of the weakness of the instruments. The table also reports the p-values of a Sargan test, as a test of overidentifying restrictions.

We first estimate the empirical model in (1) by OLS. Then, we estimate the model (1)–(2) using a two-stage least square (2SLS) estimator. Finally, in the panel specification for the 2001–2011 period, we use the Arellano-Bond model, to account for the dynamic dimension of the panel. For the IV approach, we need an appropriate set of instruments. Following Guiso et al. (2004) and Herrera and Minetti (2007), we exploit the 1936 banking law which subjected the Italian banking system to strict regulation of entry until the 1990s.20 The 1936 banking law imposed strict limits on the ability of different types of banking institutions to open new branches. In particular, each banking institution was assigned a specific geographical area of competence based on its presence in 1936. Banks’ ability to grow and lend was restricted to that area. A further directive issued in 1938 specified that national banks (banche di interesse nazionale) could open branches only in the main cities; local commercial banks and cooperative banks could open branches within the boundaries of the province where they operated in 1936; savings banks could operate within the boundaries of a region (which includes various provinces). Since in 1936 the prevalence of the different categories of banking institutions varied across Italian provinces, the tightness of the banking regulation in the following decades was highly heterogeneous across provinces. Bank entry in local credit markets was fully liberalized only towards the end of the 1990s. Guiso et al. (2004) demonstrate that these banking laws deeply affected creation and location of new bank branches in the decades that followed 1936. Thus, we expect that the regulation shaped the local banking structure during the decades in which it was in place and that this impact persisted for several years after the removal

20

For other studies exploiting the 1936 banking regulation with difference objectives see, e.g., Benfratello et al. (2008) and Minetti (2011).

Please cite this article in press as: D’Onofrio, A., et al., Banking development, socioeconomic structure and income inequality. J. Econ. Behav. Organ. (2017), http://dx.doi.org/10.1016/j.jebo.2017.08.006

G Model

ARTICLE IN PRESS

JEBO-4124; No. of Pages 24

A. D’Onofrio et al. / Journal of Economic Behavior & Organization xxx (2017) xxx–xxx

10

Table 4 Banking development and income inequality: non-linearity. Variables

(1)

(2)

(3)

Year 2001 2SLS

Branch density (log) Per capita GDP (log) Unemployment (log) Agriculture (share) Manufacturing (share) Construction (share) Trade openness (log) Instruments Saving banks 36 N. branches 36 Year fixed effects Observations R-squared F instruments Overid p value

(4)

(5)

(6)

Average 2001–2011 2SLS

(7)

(8)

(9)

Panel 2001–2011 Arellano-Bond

North Gini (log)

Center Gini (log)

South Gini (log)

North Gini (log)

Center Gini (log)

South Gini (log)

North Gini (log)

Center Gini (log)

South Gini (log)

−0.227**

−0.137**

−0.096

−0.295***

−0.070

0.056

−0.247**

−0.090

0.061

(0.110) 0.295***

(0.069) 0.066

(0.137) −0.061

(0.104) 0.431***

(0.081) 0.280***

(0.134) −0.178

(0.099) 0.410***

(0.072) 0.233***

(0.131) −0.156

(0.080) −0.027

(0.045) −0.006

(0.085) 0.019

(0.075) −0.088***

(0.085) 0.030

(0.187) −0.015

(0.073) −0.024*

(0.046) 0.005

(0.150) −0.000

(0.020) 0.583 (0.498) −0.172

(0.028) 0.073 (0.172) −0.073

(0.033) −0.026 (0.122) −0.119

(0.031) 0.386 (0.469) −0.193*

(0.051) 0.027 (0.227) −0.142*

(0.044) −0.460** (0.212) −0.399*

(0.012) 0.311 (0.450) −0.182*

(0.024) 0.102 (0.186) −0.174*

(0.019) −0.397** (0.191) −0.362*

(0.112) 0.156

(0.070) −1.540*

(0.151) 0.752*

(0.113) −0.182

(0.078) −1.506*

(0.232) 0.570

(0.105) −0.001

(0.095) −1.698***

(0.190) 0.272

(0.545) −0.045*

(0.837) −0.040**

(0.447) −0.001

(0.423) −0.026

(0.899) −0.036**

(0.552) 0.002

(0.376) −0.005

(0.571) −0.030*

(0.404) 0.002

(0.026)

(0.017)

(0.005)

(0.029)

(0.017)

(0.008)

(0.021)

(0.015)

(0.007)

0.068 (0.063) 0.004** (0.002) N 46 0.046 5.998 0.760

0.120** (0.056) 0.013** (0.005) N 21 0.591 4.513 0.229

0.224 (0.228) 0.007 (0.006) N 36 0.541 0.985 0.947

0.089 (0.070) 0.003* (0.002) N 46 0.210 6.491 0.744

0.100* (0.059) 0.009** (0.003) N 21 0.816 4.241 0.511

0.196 (0.216) 0.007 (0.006) N 36 0.366 1.236 0.128

Y 506

Y 231

Y 396

0.991

0.671

0.592

Note: The table reports regression coefficients and associated standard errors (in parentheses). The time-span of the regressions, the dependent variables and the estimation method are reported at the top of each column. The set of instruments includes the number of bank branches in the province in 1936 (per 100,000 inhabitants) and the number of savings banks in the province in 1936 (per 100,000 inhabitants). ***Significant at 1%, **Significant at 5%, *Significant at 10%. The table reports the value of the R2 and the value of the F-statistics for a test of the weakness of the instruments. The table also reports the p-values of a Sargan test, as a test of overidentifying restrictions.

of the regulation. Hence, we expect the local tightness of the 1936 banking regulation to be correlated with the current local banking development. On the other hand, as shown by Guiso et al. (2003, 2004), the distribution of types of banks across provinces in 1936, and hence the constrictiveness of regulation in a province, stemmed from “historical accident”. For instance, the strong presence of savings banks in provinces of the North East and the Center stemmed from the fact that this institution originated in Austria and started to operate first in the provinces dominated by the Austrian Empire (Lombardia and the North East) and in close-by states (especially Tuscany and the Papal States). In addition, the regulation was not designed looking at the needs of the single provinces. The differences in the restrictions imposed on the various types of banks were related to differences in the connections of the different types of banks with the Fascist regime. Therefore, the regulation is unlikely to have any direct impact on income inequality nowadays. We choose as instruments two indicators that Guiso et al. (2004) employ to characterize the local structure of the banking system in 1936: the number of bank branches in the province (per 100,000 inhabitants) and the number of savings banks in the province (per 100,000 inhabitants). Based on the discussion above, provinces with more bank branches in 1936 and with relatively more savings banks should have suffered less from the regulatory freeze. After estimating the baseline empirical model on the full sample, we split the data in three sub-samples corresponding to the three broad geographical areas (North, Center, South) in which Italy is usually subdivided. Our goal is to study the presence of non-linearities in the finance-inequality nexus. Finally, we explore dimensions of the socioeconomic structure of the Italian provinces through which banking development may have affected income inequality. In particular, we focus on the role of urban structure, migration, education and material infrastructures. 6. Banking development and income inequality In this section, we investigate the relationship between local banking development and income inequality. Table 3 displays the baseline results (heteroskedasticity-robust standard errors, clustered at the province level, are reported in parentheses). Please cite this article in press as: D’Onofrio, A., et al., Banking development, socioeconomic structure and income inequality. J. Econ. Behav. Organ. (2017), http://dx.doi.org/10.1016/j.jebo.2017.08.006

Branch density (log) HHI of branches

(1)

(2)

(3)

(4)

(5)

Controlling for bank concentration

Using ATM density

Year 2001

Year 2001

Panel 2001–2011 ArellanoBond Gini (log)

−0.249**

−0.179***

(0.108) −0.635 (0.925)

(0.062) −0.123 (1.054)

Unemployment (log) Agriculture (share) Manufacturing (share) Construction (share) Trade openness (log) Instruments Saving banks 36 N. branches 36 Region fixed effects Year fixed effects Observations R−squared

(10)

Gini index on level

Year 2001

Year 2001

−0.148***

−0.074**

−0.062***

−0.054***

(0.050) 0.002 (0.184)

(0.029)

(0.021)

(0.020)

−0.163** (0.077)

−0.168** (0.073)

2SLS Gini (log)

Average 2001–2011 2SLS Gini (log)

(11)

2SLS Gini

−0.103*** (0.037)

Branch density 1995 (log) Per capita GDP (log)

(9)

Panel 2001–2011 ArellanoBond Gini

ATM density (log)

Panel 2001–2011 ArellanoBond Gini (log)

(8)

Branches in 1995

Average 2001–2011 2SLS Gini

2SLS Gini (log)

Average 2001–2011 2SLS Gini (log)

(7)

−0.123**

−0.045**

0.147**

0.262***

0.243***

0.145**

0.262***

0.207***

(0.054) 0.089*

(0.019) 0.064**

0.055**

0.094***

0.089***

(0.067) −0.052

(0.074) −0.047

(0.055) −0.013

(0.072) −0.022

(0.082) −0.049

(0.054) −0.008

(0.050) −0.002

(0.025) −0.010

(0.024) −0.009

(0.024) −0.015

(0.021) −0.004

(0.043) 0.413 (0.408) −0.233**

(0.042) −0.073 (0.366) −0.223***

(0.010) −0.054 (0.132) −0.196***

(0.021) 0.063 (0.152) −0.230***

(0.034) −0.235 (0.180) −0.241***

(0.009) −0.167 (0.155) −0.239***

(0.018) 0.056 (0.176) −0.236***

(0.012) −0.078 (0.081) −0.104***

(0.008) 0.076 (0.059) −0.095***

(0.011) −0.041 (0.051) −0.081***

(0.004) −0.014 (0.048) −0.072***

(0.093) 0.549

(0.086) −0.261

(0.073) −0.258

(0.077) 0.335

(0.079) 0.037

(0.067) −0.178

(0.087) 0.194

(0.030) −0.024

(0.031) 0.089

(0.027) −0.080

(0.027) −0.097

(0.681) −0.018

(0.692) −0.015

(0.270) −0.009

(0.406) −0.014*

(0.381) −0.015

(0.266) −0.008

(0.399) −0.006

(0.142) −0.002

(0.158) −0.005

(0.106) −0.005

(0.089) −0.003

(0.017)

(0.018)

(0.006)

(0.008)

(0.009)

(0.007)

(0.009)

(0.004)

(0.003)

(0.003)

(0.002)

0.068 (0.063) 0.004** (0.002) Y N 103 0.234

0.120** (0.056) 0.013** (0.005) Y N 103 0.229

0.100** (0.040) 0.006*** (0.002) Y N 103 0.558

0.101** (0.045) 0.006*** (0.002) Y N 103 0.066

0.178*** (0.056) 0.007** (0.003) Y N 103 0.449

0.167*** (0.056) 0.007** (0.003) Y N 103 0.109

0.122*** (0.042) 0.005** (0.002) Y N 103 0.573

0.134*** (0.041) 0.004*** (0.002) Y N 103 0.184

Y Y 1133

Y Y 1133

Y Y 1133

Note: The table reports regression coefficients and associated standard errors (in parentheses). The time-span of the regressions, the dependent variables and the estimation method are reported at the top of each column. The set of instruments includes the number of bank branches in the province in 1936 (per 100,000 inhabitants) and the number of savings banks in the province in 1936 (per 100,000 inhabitants). ***Significant at 1%, **significant at 5%, *significant at 10%. The table reports the value of the R2 and the value of the F-statistics for a test of the weakness of the instruments. The table also reports the p-values of a Sargan test, as a test of overidentifying restrictions.

ARTICLE IN PRESS

2SLS Gini (log)

Average 2001–2011 2SLS Gini (log)

(6)

G Model

JEBO-4124; No. of Pages 24

Variables

A. D’Onofrio et al. / Journal of Economic Behavior & Organization xxx (2017) xxx–xxx 11

Please cite this article in press as: D’Onofrio, A., et al., Banking development, socioeconomic structure and income inequality. J. Econ. Behav. Organ. (2017), http://dx.doi.org/10.1016/j.jebo.2017.08.006

Table 5 Banking development and income inequality: robustness checks.

G Model JEBO-4124; No. of Pages 24

12

ARTICLE IN PRESS A. D’Onofrio et al. / Journal of Economic Behavior & Organization xxx (2017) xxx–xxx

The regressions in columns 1–5 study the impact of local banking development on the Gini coefficient, while columns 6–8 present results with the (log) Theil index as the dependent variable.21 In columns 1–2 and 6–7, we use data from the year 2001, while columns 3–4 report the regressions for the 2001–2011 average, and columns 5 and 8 use panel data from 2001 to 2011. Column 1 shows the OLS results. The estimated coefficient of bank branch density equals −0.049 and is significant at the 5% level. This suggests that an increase by 10% of the bank branch density induces a reduction of approximately 0.5% of the income inequality in the province. In column 2, we treat bank branch density as endogenous and instrument for it using the number of bank branches and the number of savings banks in the province in 1936. The estimated coefficient is −0.174 (again significant at the 5% level). Thus, the 2SLS results suggest that an increase by 10% of the bank branch density induces a reduction of approximately 1.75% of the income inequality in the province. To conserve space, we only show the coefficients on the instrumental variables from the first-stage regression (bottom row of column 2). As expected, branch density in 2001 increases with the number of bank branches and savings banks in the province in 1936. Based on the prescriptions of the 1936 Italian banking regulation (see Guiso et al., 2003), provinces with a larger number of bank branches and savings banks should have suffered less from the regulatory freeze. Results are very similar when we consider the 2001–2011 averages and the panel data. In particular, in column 5 we report the results of the Arellano Bond model in which we employ lagged values of the regressors as internal instruments as well as the set of external instruments comprising the indicators of tightness of the 1936 banking regulation. The coefficient on bank branch density is significant at the 5% level and equal to −0.150. Altogether, the findings in Table 3 support the hypothesis that bank branch density tightens the income distribution. Looking at the set of controls, we find that a higher level of per capita GDP is associated with a higher level of income inequality. Moreover, the percentage of workers in the manufacturing sector is negatively correlated with the Gini index. Finally, inequality appears to be more pronounced in southern provinces. The literature suggests that the effect of financial development on income inequality might be non-linear. For example, in Greenwood and Jovanovic (1990), Greenwood and Smith (1997), Deidda (2006) and Townsend and Ueda (2006) the organization of financial intermediaries is costly at early stages of development so that only the rich can profit from more sophisticated financial institutions. However, as the economy develops, the financial structure becomes more extensive and income inequality declines because financial development helps a larger portion of the population. These models thus predict an inverted U-shaped relationship between financial intermediary development and income inequality. In a similar vein, Lee (1996) notes that financial sector expertise is accumulated in a learning-by-doing process where lenders acquire information by investing their funds. Therefore, the financial sector may have to achieve a certain size before its functioning is sufficiently sophisticated. And in Acemoglu and Zilibotti (1997), the financial sector has to reach a certain size before indivisible, high-returns projects can be funded.22 In Table 4, we report regression results for subsamples of provinces according to broad geographical areas. The three macro regions of Italy (North, Center, South) differ significantly in the degree of economic development. In the baseline regressions of Table 3 the dummies for the Italian macro-regions suggest that southern location matters for inequality while the location in the Center has no explanatory power. In the regressions of Table 4, we instead test whether the indicators for banking development (and the other controls) have a differential effect across the macro-regions. Local banking development enters significantly, and with the expected sign, in the regression for the northern provinces. Coefficients on the control variables behave as in the full sample. Across all the empirical models, the coefficient on financial development is instead no longer significant for southern provinces; this is also partially true for the Center, with the exception of the regression using the year 2001. Since the North of Italy is the area with the highest level of per capita GDP, these findings suggest that the effectiveness of banking development in reducing income inequality varies according to the stage of economic development. In Table 5, we perform a number of robustness tests on the baseline regressions of Table 3. As a first robustness check, we add to the regressions the Herfindahl-Hirschman Index (HHI) of bank branches in the province to allow for a possible role of bank concentration in income inequality. The literature suggests that bank concentration might affect the access to finance (see, e.g., Beck and Demirguc-Kunt, 2008; Degryse and Ongena, 2005). The reader might then wonder whether our instruments affect bank concentration and, in turn, bank concentration shapes the income distribution in the provinces (possibly leading to a violation of the exclusion restriction on the instruments). To assuage such possible concerns we then study whether bank concentration has a role in addition to local banking development. The estimates in columns 1–3 reveal that (this proxy for) banking competition has no statistically significant effect on income inequality. Further, the results for the impact of banking development on inequality remain essentially unaltered. In columns 4–6 of Table 5, we verify the robustness of the results to using an alternative measure of local banking development, the number of ATM per capita in the province. The estimates confirm the findings obtained with our preferred measure of banking development. As a third robustness check, we also experiment with a different timing of the regulatory indicators. As noted, the banking deregulation unfolded

21 From now on, we show only results from the estimations with the Gini index as dependent variable. Results are robust to using the Theil index as a measure of inequality. 22 Kim and Lin (2011) explore whether the effect of financial development on income inequality arises only above a certain threshold of financial development. Clarke et al. (2006) and Ang (2010) find no evidence of an inverted U-shaped relationship. For an empirical analysis of non-linearities in the finance-growth nexus, see, instead, Yilmazkuday (2011).

Please cite this article in press as: D’Onofrio, A., et al., Banking development, socioeconomic structure and income inequality. J. Econ. Behav. Organ. (2017), http://dx.doi.org/10.1016/j.jebo.2017.08.006

G Model

ARTICLE IN PRESS

JEBO-4124; No. of Pages 24

A. D’Onofrio et al. / Journal of Economic Behavior & Organization xxx (2017) xxx–xxx

13

progressively during the 1990s.23 In columns 7–8 we then experiment with using the ratio of branches over population for the mid-nineties (the year 1995). The results remain virtually unchanged. In columns 9–11 of Table 5, we perform a further robustness check with an alternative definition of the inequality measure: instead of using the logarithm of the Gini, we use the Gini in levels as the dependent variable. Again, the results remain virtually unchanged. Finally, we experiment with dropping unemployment or per capita GDP from the regressions, to assuage possible concerns of multicollinearity of these variables with the measure of local banking development. Again, the results carry through (see Table A.3 of the Appendix). 7. Socioeconomic structure and the banking-inequality nexus Banking development can reduce income inequality through various channels. The extant empirical literature has especially concentrated on mechanisms such as labor demand, entrepreneurship, and firm formation. By contrast, there is very limited evidence on the effects that financial development can have on income inequality through its impact on the socioeconomic structure. Yet, there is compelling evidence that structural factors such as urban structure and migration, infrastructures, and education can play a key role in income distribution (see, e.g., Calderon and Serven, 2004; World Bank, 2003; Ferreira, 1995; Lopez, 2003; and references therein). In Italy these structural factors are reputed to have a strong influence on income inequality (Ascher and Krupp, 2010; Farina and Franzini, 2015), and this makes Italy a good empirical laboratory to isolate their possible role in the finance-inequality nexus.24 The objective of this section is to take a step in this direction. 7.1. Disentangling the channels The availability of bank credit can influence agents’ ability to purchase houses and land and move across geographical areas, thus affecting migratory flows and the diffusion of urbanization. Banking development can also allow to finance material and immaterial infrastructures. Infrastructures can help smooth out inequalities, for example by promoting better access to investment opportunities for the poor. Further, a more developed local banking sector can improve agents’ ability to accumulate human capital, helping them finance the costs of education and also making available the resources to finance the development and improvement of schools. Clearly, these structural factors do not necessarily work in the direction of mitigating income inequality. For instance, the development of IT infrastructures may widen the income gap between highly skilled workers and low skilled ones, due to the different ability of skilled and unskilled workers to take advantage of advances in information technology (see, e.g., Autor, 2014). And, starting from Kuznets (1955) and Williamson (1965), various studies have stressed that urbanization and migratory flows can have a non-linear impact on income inequality. In Tables 6–8 we explore different structural channels through which banking development can influence inequality. In Table 6a and b, we focus on the role of urban structure and geographical mobility (inter-province migratory flows). In Table 7, we consider the role of material and immaterial infrastructures. In Table 8, we study the role of human capital formation and education. In each table, we perform two kinds of tests. In Panel A of the table, we add the indicators of socioeconomic structure to the baseline regressions and verify to what extent they tend to absorb the estimated effect of banking development on income inequality. In Panel B, we test directly the effect of banking development on the indicators of socioeconomic structure. This approach to disentangling the channels of influence is in line with other studies. For example, Faber and Gaubert (2016) study the impact of tourism on economic growth in Mexico. To verify to what extent the detected positive effect of tourism on growth occurs through the impact of tourism on infrastructures (airports, seaports, roads or railways), they estimate the effect of tourism on growth both before and after additionally conditioning on measures of development of airport, seaport, road and railway infrastructures. Similarly, Claessens and Feijen (2007) try to disentangle to what extent the detected negative effect of financial development on undernourishment occurs through the impact of financial development on agricultural productivity and on the adoption of productivity-enhancing inputs. To this end, they first estimate the impact of financial development on undernourishment, then the impact of financial development on undernourishment conditioning on the adoption of productivity-enhancing techniques and tools, and finally the impact of financial development on the adoption of productivity-enhancing techniques.25 All these studies stress that, since the additional controls capturing possible channels of influence are unlikely to be causally identified and since they may be affected by the key dependent variable itself (inequality in our setting), the additional results obtained by adding the controls to the regressions should be interpreted with caution. Nonetheless, this approach can constitute a step towards identifying the mechanisms whereby banking development can affect inequality.

23 As noted by Guiso et al. (2003, 2004), during the 1990s Italy experienced various banking reforms. The European Banking Directive enacted in 1992 mandated free entry. In 1993, a new banking law was approved which incorporated the European Banking Directive. In the same year, the separation between short and long-term lending was removed and banks were allowed to underwrite security offerings and equity. Again in 1993, savings banks were transformed from mutual organizations into standard corporations. This facilitated mergers and acquisitions. In 1994, the government started to privatize state-owned banks. 24 For a review of the relevance and effects of migratory flows in Italy, see, e.g., Mocetti and Porello (2010) and references therein. 25 In a different context, Manova (2013) studies whether firm entry is a channel whereby financial development spurs export. She does so by verifying to what extent the estimated effect of financial development on export changes after conditioning on firm entry.

Please cite this article in press as: D’Onofrio, A., et al., Banking development, socioeconomic structure and income inequality. J. Econ. Behav. Organ. (2017), http://dx.doi.org/10.1016/j.jebo.2017.08.006

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

(11)

(12)

Panel A Year 2001

Average 2001–2011

Panel 2001–2011

Year 2001

2SLS

2SLS

Arellano-Bond

2SLS

Manufacturing (share) Construction (share) Trade openness (log) Instruments Saving banks 36 N. branches 36 Region fixed effects Year fixed effects Observations R-squared F instruments Overid p value

Year 1991

Gini (log)

Gini (log)

Gini (log)

Gini (log)

Gini (log)

Gini (log)

Gini (log)

Gini (log)

Gini (log)

Gini (log)

Small Municipality

Gross flow

Net flow

−0.174** (0.069)

−0.165** (0.070) −0.054*

−0.167 (0.104)

−0.219*** (0.079)

−0.170*** (0.061)

−0.155** (0.061) −0.077***

−0.093 (0.072)

−0.222*** (0.057)

−0.150*** (0.055)

−0.133** (0.057) −0.068***

−0.068 (0.066)

−0.207*** (0.061)

0.374** (0.175)

−2.049** (0.807)

−1.069 (1.174)

−0.253* (0.149) −0.010

1.599** (0.754) 0.198

4.358*** (1.522) −0.095

(0.029)

(0.023)

(0.023)

0.003

0.027**

(0.015)

(0.012)

0.029** (0.012)

0.131** (0.056) −0.021

0.119** (0.058) −0.023

0.123 (0.087) −0.021

0.007 (0.007) 0.205*** (0.072) −0.012

(0.018) 0.179 (0.136) −0.222***

(0.017) 0.209* (0.126) −0.201***

(0.018) 0.186 (0.128) −0.221***

(0.017) 0.284 (0.223) −0.119

(0.030) −0.119 (0.142) −0.222***

(0.028) −0.089 (0.125) −0.198***

(0.026) −0.054 (0.108) −0.237***

(0.028) −0.146 (0.193) −0.175***

(0.010) −0.053 (0.133) −0.195***

(0.009) −0.013 (0.118) −0.169**

(0.009) 0.021 (0.097) −0.208***

(0.010) −0.051 (0.193) −0.141**

(0.056) 0.553 (0.507) 0.376*

(0.246) −2.307 (2.708) −1.171

(0.384) 11.735** (5.052) 2.716

(0.071) 0.227

(0.069) 0.546

(0.071) 0.271

(0.081) 0.037

(0.074) −0.211

(0.071) 0.138

(0.063) 0.095

(0.064) −0.167

(0.074) −0.259

(0.071) −0.002

(0.059) 0.031

(0.061) −0.209

(0.226) 4.631***

(1.179) −7.753

(1.926) −3.367

(0.371) −0.010

(0.339) −0.012*

(0.338) −0.010

(0.411) −0.039***

(0.297) −0.013

(0.273) −0.016**

(0.242) −0.010

(0.278) −0.043***

(0.250) −0.009

(0.227) −0.012**

(0.199) −0.007

(0.231) −0.031***

(1.198) −0.036

(7.708) 0.088

(11.860) −0.296

(0.008)

(0.007)

(0.008)

(0.014)

(0.008)

(0.007)

(0.007)

(0.011)

(0.006)

(0.005)

(0.006)

(0.009)

(0.028)

(0.133)

(0.289)

0.129*** (0.041) 0.005** (0.002) Y N 103 0.577 10.22 0.436

0.130*** (0.041) 0.005** (0.002) Y N 103 0.607 9.627 0.357

0.087** (0.044) 0.003 (0.002) Y N 103 0.590 3.448 0.504

0.096* (0.052) 0.004* (0.002) Y N 65 0.104 5.200 0.681

0.140*** (0.041) 0.004*** (0.002) Y N 103 0.186 12.34 0.913

0.141*** (0.041) 0.004** (0.002) Y N 103 0.295 11.52 0.660

0.106** (0.042) 0.002 (0.002) Y N 103 0.436 5.430 0.498

0.110* (0.048) 0.003 (0.002) Y N 65 0.328 8.295 0.637

0.123*** (0.042) 0.004** (0.002) Y N 103 0.449 9.006 0.154

0.086 (0.061) 0.010*** (0.003) Y N 103 0.483 5.729 0.314

0.078 (0.062) 0.006** (0.003) Y N 65 0.435 3.571 0.102

0.253*** (0.065) −0.043

0.229*** (0.066) −0.052*

0.159** (0.080) −0.045*

0.022*** (0.006) 0.259*** (0.064) −0.069**

0.244*** (0.058) −0.013

0.227*** (0.059) −0.015*

0.157** (0.070) −0.014

0.022*** (0.006) 0.257*** (0.060) −0.023**

Y Y 1133

Y Y 1133

Y Y 1133

Y Y 715

0.999

0.984

0.899

0.999

Note: The table reports regression coefficients and associated standard errors (in parentheses). The time-span of the regressions, the dependent variables and the estimation method are reported at the top of each column. In Panel A, we add the indicators of urban structure and migration to the baseline regressions. In Panel B, we test the effect of banking development on the indicators of urban structure and migration. The set of instruments includes the number of bank branches in the province in 1936 (per 100,000 inhabitants) and the number of savings banks in the province in 1936 (per 100,000 inhabitants). ***Significant at 1%, **significant at 5%, *significant at 10%. The table reports the value of the R2 and the value of the F-statistics for a test of the weakness of the instruments. The table also reports the p-values of a Sargan test, as a test of overidentifying restrictions.

ARTICLE IN PRESS

Agriculture (share)

Year 1991

Gini (log)

Net flow 1991 (log)

Unemployment (log)

15)

Gini (log)

Small municipality 2001 (share) Gross flow 1991 (log)

Per capita GDP (log)

(14)

Panel B

A. D’Onofrio et al. / Journal of Economic Behavior & Organization xxx (2017) xxx–xxx

Branch density (log)

(13)

G Model

Variables

JEBO-4124; No. of Pages 24

14

Please cite this article in press as: D’Onofrio, A., et al., Banking development, socioeconomic structure and income inequality. J. Econ. Behav. Organ. (2017), http://dx.doi.org/10.1016/j.jebo.2017.08.006

Table 6a Banking development and economic structure: urbanization and migration.

G Model

ARTICLE IN PRESS

JEBO-4124; No. of Pages 24

A. D’Onofrio et al. / Journal of Economic Behavior & Organization xxx (2017) xxx–xxx

15

Table 6b Banking development, urbanization and migration: further tests. Variables

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Panel A Migration: the main provincial city

Branch density (log) Gross flow (log)

Average 2001–2011

Panel 2001–2011

Year 2001

Average 2001–2011

2SLS

Arellano-Bond

2SLS

2SLS

N. branches 36 + Controls Region fixed effects Year fixed effects Observations R−squared F instruments Overid p value

Year 2001

Gini (log)

Gini (log)

Gini (log)

Gini (log)

Gross flow main prov. city

Net flow main prov. city

Gini (log)

Gross flow other municipalities

−0.115*

−0.181***

−0.090

−0.166***

−3.133***

−2.134**

−0.119*

−1.976**

(0.066) 0.021** (0.009)

(0.046)

(0.060) 0.022** (0.008)

(0.043)

(0.969)

(1.013)

(0.063) 0.005 (0.006)

(0.966)

Net flow (log) Instruments Saving banks 36

Migration: other municipalities

0.011*** (0.004)

0.013*** (0.004)

0.107***

0.162***

0.129***

0.141*

0.112***

0.125***

(0.040) 0.003 (0.002) Y Y

(0.046) 0.003* (0.002) Y Y

Y Y

Y Y

(0.041) 0.005** (0.002) Y Y

(0.046) 0.005** (0.002) Y Y

(0.041) 0.005*** (0.002) Y Y

(0.040) 0.005*** (0.002) Y Y

N

N

Y

Y

N

N

N

N

103 0.388 7.901 0.487

73 0.550 12.681 0.429

1133

803

0.793

0.999

103 0.566 10.22 0.346

73 0.587 11.73 0.664

101 0.358 9.638 0.371

101 0.189 9.804 0.309

Variables

Panel B Panel 2001–2011 (Arellano-Bond) Center

North

Branch density (log) Gross flow (log)

Gini (log)

Gini (log)

Gini (log)

Gini (log)

Gini (log)

−0.125 (0.117) 0.027 (0.018)

−0.226*** (0.081)

−0.047 (0.087) 0.027 (0.019)

−0.117 (0.076)

0.104 (0.126) 0.033 (0.032)

0.154** (0.043)

Net flow (log) + Controls Region and Year fixed effects Observations

South

Gini (log)

Y Y 506

0.033*** (0.008) Y Y 451

Y Y 231

0.001 (0.009) Y Y 209

Y Y 396

−0.016*** (0.003) Y Y 55

Note: The table reports regression coefficients and associated standard errors (in parentheses). The time-span of the regressions, the dependent variables and the estimation method are reported at the top of each column. In Panel A, columns (1)–(4), we add the gross migratory flow or the net migratory flow of the main provincial city to the baseline regressions. In Panel A, columns (5)–(6), we test the impact of bank branch density on the gross migratory flow or the net migratory flow of the main provincial city. In Panel A, column (7), we add the gross migratory flow in the other municipalities of the province to the baseline regressions. In Panel A, column (8), we test the impact of bank branch density on the gross migratory flow of the other municipalities of the province. In Panel B, we reestimate the regressions in columns (11)–(12) of Table 6 separately for the North, Center, and South of Italy. The set of instruments includes the number of bank branches in the province in 1936 (per 100,000 inhabitants) and the number of savings banks in the province in 1936 (per 100,000 inhabitants). ***Significant at 1%, **significant at 5%, *significant at 10%. The table reports the value of the R2 and the value of the F-statistics for a test of the weakness of the instruments. The table also reports the p-values of a Sargan test, as a test of overidentifying restrictions.

7.2. Urban structure and migration In Tables 6a and b, we focus on the urbanization-migration channel. In Table 6a and b, Panel A, columns 1, 5 and 9 carry over the baseline results of Table 3. The specifications in columns 1–4 are 2SLS regressions using data for the year 2001, while those in columns 5–8 are 2SLS regressions for the 2001–2011 averages. Finally, columns 9–12 report ArellanoBond estimates. In columns 2, 6 and 10, we add to the baseline specification a proxy for urban structure (the diffusion of urbanization), whereas in columns 3–4, 7–8 and 11–12 we insert two indicators of inter-province migratory flows. We capture the urban structure of the province using the percentage of population living in small municipalities (less than 15,000 inhabitants) in 2001. To capture migration, instead, we consider, alternatively, the (log) gross migratory flow of the province Please cite this article in press as: D’Onofrio, A., et al., Banking development, socioeconomic structure and income inequality. J. Econ. Behav. Organ. (2017), http://dx.doi.org/10.1016/j.jebo.2017.08.006

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

(11)

(12)

Panel A Average 2001–2011 2SLS Gini (log)

Gini (log)

Gini (log)

Gini (log)

Gini (log)

Gini (log)

Gini (log)

Gini (log)

Gini (log)

Material Infrastr.

Street index

Railway index

IT Broad. index

−0.188*** (0.065) −0.040*

−0.167** (0.072)

−0.178*** (0.065)

−0.160* (0.082)

−0.170*** (0.055) −0.030

−0.170*** (0.064)

−0.173*** (0.062)

−0.142** (0.069)

−0.151*** (0.054) −0.007

−0.148** (0.059)

−0.150*** (0.055)

−0.119* (0.064)

−0.288 (0.484)

0.182 (0.511)

−0.075 (1.024)

−0.804* (0.420)

(0.018)

Construction (share) Trade openness (log) Instruments Saving banks 36 N. branches 36 Region fixed effects Year fixed effects Observations R-squared F instruments Overid p value

(0.007) −0.004 (0.014)

−0.017* (0.009)

−0.005 (0.014) −0.008 (0.010)

−0.006 (0.009)

0.013

0.032**

0.246*** (0.064) −0.039

0.251*** (0.070) −0.043

0.244*** (0.061) −0.046

(0.016) 0.209*** (0.071) −0.050*

0.033**

0.244*** (0.057) −0.013

0.241*** (0.062) −0.013

0.239*** (0.054) −0.014

(0.016) 0.206*** (0.066) −0.016*

−0.146 (0.315) 0.026

−0.023 (0.410) 0.168

−0.614 (0.696) −0.094

1.071*** (0.385) 0.089

0.127** (0.053) −0.021

0.128** (0.057) −0.019

0.122** (0.051) −0.023

(0.018) 0.115* (0.069) −0.022

(0.016) 0.025 (0.125) −0.295***

(0.019) 0.175 (0.135) −0.222***

(0.018) 0.154 (0.125) −0.251***

(0.017) 0.225* (0.116) −0.217***

(0.028) −0.175 (0.140) −0.295***

(0.030) −0.119 (0.144) −0.226***

(0.032) −0.129 (0.142) −0.239***

(0.027) 0.019 (0.129) −0.221***

(0.010) −0.074 (0.126) −0.211***

(0.010) −0.053 (0.134) −0.199**

(0.010) −0.060 (0.131) −0.208***

(0.009) 0.089 (0.116) −0.189***

(0.141) −3.926*** (0.800) −1.826***

(0.150) 0.401 (1.081) −0.301

(0.361) −1.625 (2.306) −1.678

(0.127) −3.848*** (0.904) −0.298

(0.078) −0.242

(0.072) 0.230

(0.075) 0.073

(0.069) 0.398

(0.072) −0.507*

(0.082) −0.201

(0.074) −0.284

(0.070) 0.168

(0.067) −0.318

(0.081) −0.247

(0.070) −0.305

(0.069) 0.074

(0.460) −11.918***

(0.610) 3.670

(1.236) −9.274

(0.546) −13.928***

(0.302) −0.006

(0.353) −0.010

(0.321) −0.009

(0.296) −0.011

(0.290) −0.008

(0.286) −0.013

(0.306) −0.012

(0.256) −0.015**

(0.232) −0.008

(0.240) −0.009

(0.252) −0.008

(0.210) −0.011*

(3.054) 0.107**

(2.869) 0.062

(5.696) 0.067

(2.690) 0.048

(0.008)

(0.008)

(0.007)

(0.008)

(0.008)

(0.008)

(0.008)

(0.007)

(0.006)

(0.006)

(0.006)

(0.006)

(0.042)

(0.056)

(0.107)

(0.054)

0.131*** (0.045) 0.004** (0.002) Y N 103 0.582 10.12 0.537

0.107*** (0.038) 0.005*** (0.002) Y N 103 0.590 10.22 0.486

0.137*** (0.042) 0.004** (0.002) Y N 103 0.598 10.32 0.499

0.102** (0.046) 0.005** (0.002) Y N 103 0.605 7.309 0.455

0.145*** (0.042) 0.004*** (0.001) Y N 103 0.218 12.81 0.980

0.119*** (0.038) 0.005*** (0.002) Y N 103 0.189 12.77 0.997

0.141*** (0.041) 0.004** (0.002) Y N 103 0.187 10.43 0.952

0.121*** (0.044) 0.004** (0.002) Y N 103 0.321 9.214 0.996

0.129*** (0.041) 0.005** (0.002) Y N 103 0.611 10.22 0.456

0.129*** (0.041) 0.005** (0.002) Y N 103 0.097 10.22 0.001

0.129*** (0.041) 0.005** (0.002) Y N 103 0.146 10.22 0.447

0.129*** (0.041) 0.005** (0.002) Y N 103 0.679 10.22 0.250

Y Y 1133

Y Y 1133

Y Y 1133

Y Y 1133

0.999

0.995

0.991

0.999

Note: The table reports regression coefficients and associated standard errors (in parentheses). The time-span of the regressions, the dependent variables and the estimation method are reported at the top of each column. In Panel A, we add the indicators of material infrastructures to the baseline regressions. In Panel B, we test the effect of banking development on the indicators of material infrastructures. The set of instruments includes the number of bank branches in the province in 1936 (per 100,000 inhabitants) and the number of savings banks in the province in 1936 (per 100,000 inhabitants). ***Significant at 1%, **significant at 5%, *significant at 10%. The table reports the value of the R2 and the value of the F-statistics for a test of the weakness of the instruments. The table also reports the p-values of a Sargan test, as a test of overidentifying restrictions.

ARTICLE IN PRESS

−0.005 (0.013)

IT and broadband index

Manufacturing (share)

Year 2001 2SLS

Gini (log)

Railway index

Agriculture (share)

Panel 2001–2011 Arellano-Bond

Gini (log)

(0.021)

Unemployment (log)

(16)

Gini (log)

Street index

Per capita GDP (log)

(15)

A. D’Onofrio et al. / Journal of Economic Behavior & Organization xxx (2017) xxx–xxx

Material infrastructure

(14)

Panel B

Year 2001 2SLS

Branch density (log)

(13)

G Model

Variables

JEBO-4124; No. of Pages 24

16

Please cite this article in press as: D’Onofrio, A., et al., Banking development, socioeconomic structure and income inequality. J. Econ. Behav. Organ. (2017), http://dx.doi.org/10.1016/j.jebo.2017.08.006

Table 7 Banking development and economic structure: material infrastructure.

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

(11)

(12)

Panel A Average 2001–2011 2SLS Gini (log)

Gini (log)

Gini (log)

Gini (log)

Gini (log)

Gini (log)

Gini (log)

Gini (log)

Gini (log)

Secondary degree

Infrastr. for education

Cult. and recr. struct

−0.174** (0.069)

−0.172*** (0.059) −0.003

−0.173*** (0.067)

−0.166*** (0.060)

−0.170*** (0.061)

−0.168*** (0.055) −0.001

−0.170*** (0.062)

−0.172*** (0.060)

−0.150*** (0.055)

−0.151*** (0.052) −0.000

−0.143** (0.064)

−0.181** (0.078)

0.201 (0.414)

0.067 (0.479)

0.569 (0.859)

Construction (share) Trade openness (log) Instruments Saving banks 36 N. branches 36 Region fixed effects Year fixed effects Observations R-squared F instruments Overid p value

(0.002) 0.010

(0.019)

(0.019)

0.048* (0.026)

−0.013

0.004

0.253*** (0.065) −0.043

0.261*** (0.072) −0.042

0.247*** (0.072) −0.045

(0.011) 0.251*** (0.070) −0.043

0.244*** (0.058) −0.013

0.245*** (0.063) −0.013

0.218*** (0.068) −0.018*

(0.054) 0.205* (0.115) −0.017

0.995** (0.413) 0.190

0.425 (0.335) 0.123

0.490 (0.686) 0.121

0.073

0.131** (0.056) −0.021

0.164** (0.068) −0.016

0.135** (0.061) −0.020

(0.011) 0.137** (0.056) −0.019

(0.018) 0.179 (0.136) −0.222***

(0.015) 0.121 (0.120) −0.283***

(0.017) 0.140 (0.119) −0.229***

(0.017) 0.111 (0.120) −0.256***

(0.030) −0.119 (0.142) −0.222***

(0.028) −0.146 (0.147) −0.242***

(0.028) −0.076 (0.140) −0.214***

(0.030) −0.097 (0.140) −0.211***

(0.010) −0.053 (0.133) −0.195***

(0.010) −0.053 (0.130) −0.196**

(0.010) 0.153 (0.136) −0.150**

(0.013) 0.347 (0.254) 0.016

(0.128) −1.989* (1.096) −1.931***

(0.162) −4.177*** (1.026) −0.846

(0.200) −5.107*** (1.435) −2.681***

(0.071) 0.227

(0.084) 0.103

(0.070) 0.102

(0.074) 0.117

(0.074) −0.211

(0.083) −0.249

(0.070) −0.111

(0.073) −0.188

(0.074) −0.259

(0.079) −0.261

(0.074) 0.165

(0.133) 0.168

(0.523) −4.283**

(0.538) −13.585***

(0.639) −8.244**

(0.371) −0.010

(0.323) −0.010

(0.296) −0.010

(0.319) −0.009

(0.297) −0.013

(0.295) −0.013

(0.290) −0.013

(0.295) −0.013*

(0.250) −0.009

(0.245) −0.009

(0.242) −0.010

(0.299) −0.016**

(2.159) 0.018

(3.277) 0.004

(4.046) 0.097

(0.008)

(0.007)

(0.008)

(0.008)

(0.008)

(0.008)

(0.008)

(0.008)

(0.006)

(0.006)

(0.006)

(0.008)

(0.052)

(0.055)

(0.061)

0.129*** (0.041) 0.005** (0.002) Y N 103 0.577 10.22 0.436

0.150*** (0.042) 0.004** (0.002) Y N 103 0.598 13.21 0.594

0.128*** (0.042) 0.005*** (0.002) Y N 103 0.582 11.24 0.418

0.130*** (0.041) 0.005*** (0.002) Y N 103 0.598 11.76 0.365

0.140*** (0.041) 0.004*** (0.002) Y N 103 0.186 12.34 0.913

0.164*** (0.043) 0.004** (0.002) Y N 103 0.194 14.55 0.997

0.142*** (0.043) 0.004*** (0.002) Y N 103 0.193 12.13 0.904

0.141*** (0.040) 0.005*** (0.002) Y N 103 0.182 13.71 0.927

0.129*** (0.041) 0.005** (0.002) Y N 103 0.429 10.22 0.045

0.129*** (0.041) 0.005** (0.002) Y N 103 0.452 10.22 0.856

0.129*** (0.041) 0.005** (0.002) Y N 103 0.599 10.22 0.571

Y Y 1133

Y Y 1133

Y Y 1133

Y Y 1133

0.999

0.998

0.999

0.998

Note: The table reports regression coefficients and associated standard errors (in parentheses). The time-span of the regressions, the dependent variables and the estimation method are reported at the top of each column. In Panel A, we add the indicators of human capital formation and cultural infrastructures to the baseline regressions. In Panel B, we test the effect of banking development on the indicators of human capital formation and cultural infrastructures. The set of instruments includes the number of bank branches in the province in 1936 (per 100,000 inhabitants) and the number of savings banks in the province in 1936 (per 100,000 inhabitants). ***Significant at 1%, **significant at 5%, *significant at 10%. The table reports the value of the R2 and the value of the F-statistics for a test of the weakness of the instruments. The table also reports the p-values of a Sargan test, as a test of overidentifying restrictions.

ARTICLE IN PRESS

(0.002) −0.009

Cultural and recreation struct.

Manufacturing (share)

Year 2001 2SLS

Gini (log)

(0.002)

Agriculture (share)

Panel 2001–2011 Arellano-Bond

Gini (log)

Infrastructure for education

Unemployment (log)

(15)

Gini (log)

Secondary degree 2001

Per capita GDP (log)

(14)

Panel B

Year 2001 2SLS

Branch density (log)

(13)

G Model

JEBO-4124; No. of Pages 24

Variables

A. D’Onofrio et al. / Journal of Economic Behavior & Organization xxx (2017) xxx–xxx 17

Please cite this article in press as: D’Onofrio, A., et al., Banking development, socioeconomic structure and income inequality. J. Econ. Behav. Organ. (2017), http://dx.doi.org/10.1016/j.jebo.2017.08.006

Table 8 Banking development and economic structure: human capital formation.

G Model JEBO-4124; No. of Pages 24

18

ARTICLE IN PRESS A. D’Onofrio et al. / Journal of Economic Behavior & Organization xxx (2017) xxx–xxx

(sum of immigration and emigration flow) and the net migratory flow of the province (immigration minus emigration flow) in 1991. We focus on lagged values of the migratory flows to minimize endogeneity problems.26 The literature predicts that both urban structure and migratory flows can have significant consequences on inequality within local communities. Regarding the urban structure, for example, recent studies suggest that a more diffuse urban structure and a lower concentration in big cities can help reduce income inequality (Baum-Snow and Pavan, 2013; Behrens and Robert-Nicoud, 2014a,b). This would occur because large cities are more conducive to skill-biased technological change, selection of highly productive entrepreneurs, and segmentation, and may then exacerbate inequality through these channels. For example, Behrens and Robert-Nicoud (2014a) develop a model in which cities generate productivity improvements through agglomeration economies. They show that this can promote firm formation but also the selection of highly productive entrepreneurs, exacerbating inequality.27 Based on these arguments, one could then expect that a more diffuse urbanization can better mitigate inequality within the provinces. Regarding migratory flows, the net effect of migration on income inequality is ambiguous a priori (see, e.g., Blau and Kahn, 2015, and Card, 2009, for detailed discussions). It is often maintained that immigration may exacerbate inequality in local communities because the inflow of relatively poor immigrants tends to widen the income distribution (both directly, due to the inflow of poor individuals, and by exerting a downward pressure on the salaries of unskilled residents). At the same time, provinces with a larger outflow of emigrants can experience an increase or a decrease in inequality thanks to remittances from their (poor or rich) emigrants. As the regressions in Table 6a and b, Panel A, show, after conditioning on the diffusion of urbanization or on the gross or net migratory flows, the sign of the coefficient on bank branch density remains negative in all the specifications. However, interestingly, the newly added regressors tend to absorb the effect of the banking development indicator. In particular, the percentage of the population living in small municipalities has a negative and statistically significant effect on the Gini (see columns 2, 6 and 10). That is, in line with the above-referenced studies, provinces with more diffuse urbanization tend to have lower income inequality. When considering migration, both the gross and the net migratory flow positively affect the Gini: provinces with a larger inflow of immigrants tend to experience larger inequality (columns 7–8 and 11–12). Most interestingly, while the net migratory flow has no bearing for the results on banking development, when we insert in the regressions the gross migratory flow, the coefficient for banking development tends to lose its statistical significance. In Panel B of Table 6a and b, we further obtain that banking development has a positive and significant effect on the percentage of the population living in small municipalities, and a negative effect on gross migratory flows.28 Overall, the results in the table suggest that geographical mobility and urban structure could be a channel whereby banking development affects income distribution. In Table 6b, we further dig into the urban structure-migration channel. In Panel A, columns 1–6, we focus on the role of the migratory flows of the main city of the province. In particular, in columns 1–4 we reestimate the baseline regressions after conditioning on the gross and net migratory flows into the main city of the province. We find that immigration into the main city of the province tends to increase inequality. Further, banking development tends to somewhat lose significance after conditioning on the migratory flows of the main provincial city. These findings can be contrasted with those in column 7, where we find that migration towards other municipalities of the province has no significant impact on income inequality (and, conditioning on it, the effect of banking development remains negative and significant). Overall the results of Panel A of Table 6b are then consistent with those obtained in Table 6a and b when conditioning on the total (gross or net) migratory flow of the province and on the diffusion of urbanization. In particular, they lend some support to the above argument that concentration of the population in big cities may exacerbate inequality and that banking development may help mitigate inequality by promoting a more diffuse urbanization. Finally, in Panel B of Table 6b we investigate possible non-linearities in the urban structure-migration channel. In particular, we reestimate the regressions of Table 6a and b separately for the North, Center and South of Italy. We obtain evidence that the effects of urban structure and migration tend to manifest themselves especially in northern provinces, while we do not find equally compelling evidence for provinces in the Center and the South.29 This may suggest that the detected nonlinear effect of banking development on inequality could also be due to non-linearities in the effect of the urban structure and migratory flows (in line with the hypotheses in Kuznets, 1955).

26 Burgess and Pande (2005) use population density in Indian states as a control variable. Our indicators are instead specifically aimed at capturing the relative importance of small and large municipalities (diffusion of the urban structure) in a province as well as the geographical mobility of the population. As discussed below, a growing strand of literature stresses the impact of large cities on income inequality. 27 Baum-Snow and Pavan (2013) find that between 25 and 35 percent of the increase in inequality in the United States from 1979 to 2007 is explained by city size. They argue that most of this reflects rapid increases in the return to unobserved skill in cities (e.g., urban agglomerations facilitate the development of new more skill-intensive technologies that are then efficiently implemented in the same cities; see also Bacolod et al., 2009). Behrens and Robert-Nicoud (2014a) document that in a 2007 sample of 356 U.S. cities a city of double size is associated with a Gini coefficient higher by 1.2%. 28 Banking development and financial access an promote the diffusion of urbanization. In addition, they can make available resources for sustaining the sizeable costs of migration or influence the transfer of remittances from emigrants to their communities. For a discussion of some of these mechanisms in the Italian context see, e.g., De Rosa (1980). 29 Interestingly, after conditioning on the net migratory flow, for southern provinces the estimated effect of banking development on inequality is positive, which further corroborates the hypothesis of non-linearities in the banking development-inequality nexus.

Please cite this article in press as: D’Onofrio, A., et al., Banking development, socioeconomic structure and income inequality. J. Econ. Behav. Organ. (2017), http://dx.doi.org/10.1016/j.jebo.2017.08.006

G Model

ARTICLE IN PRESS

JEBO-4124; No. of Pages 24

A. D’Onofrio et al. / Journal of Economic Behavior & Organization xxx (2017) xxx–xxx

19

7.3. Infrastructures and human capital In Table 7, we investigate the effect of material and immaterial infrastructures. As noted, material infrastructures can reduce inequality by easing the access to productive opportunities by the poor (World Bank, 2003; Ferreira, 1995; Lopez, 2003). In Panel A, we find some evidence that a broad indicator of the material infrastructures of the province reduces income inequality (see columns 1 and 5). When we consider finer categories of infrastructures, we detect no significant effect of railway and road infrastructures on inequality, while we obtain some weak evidence that IT infrastructures exacerbate inequality (columns 8 and 12). Most importantly for our purposes, we find that the estimated effect of banking development remains essentially unaltered after conditioning on the various indicators of material infrastructures. Further, in Panel B we uncover no evidence that banking development significantly affects material infrastructures in the provinces. This might reflect the fact that in Italy a significant portion of infrastructures is financed through public (central and local government) budgets rather than through bank financing. In Table 8, we turn to the role of human capital formation. In Panel A, we first add, as a proxy for education, the percentage of the provincial population with at least a secondary school degree (columns 2, 6, and 10).30 This proxy for education in the province enters the regressions for income inequality with a negative sign, in line with the implications of the Galor and Zeira (1993) model on the effect of education on the distribution of income. However, its effect is estimated imprecisely. Next, we experiment by including an indicator for the quality of infrastructures for education (see columns 3, 7 and 11) and, more broadly, for the quality of infrastructures for cultural advancement. These indicators appear to have little power in explaining income inequality in the provinces. In Panel B, we test the impact of local banking development on the indicator for higher education, and on the two indices for education and cultural infrastructures. We find no evidence of a significant impact of banking development on the indicators. Plausibly, this could again reflect the relevant role of public budgets in financing education and school development in Italy (and, correspondingly, the limited influence of bank funding). Overall, the results in Tables 6–8 support the hypothesis that, to the extent that the economic structure acts as a link between banking development and inequality, this link may manifest itself especially through the impact of banks on geographical mobility and urban structure. By contrast, little evidence is found for a possible role of material infrastructures and human capital formation in the finance-inequality nexus. 8. Conclusion In this paper, we have empirically investigated the relationship between financial development and income distribution and the role of the socioeconomic structure in this relationship. In particular, we have analyzed the impact of local banking development, measured by bank branch density, on income inequality in Italian provinces. Exploiting the Italian banking regulation of the 1930s to tackle endogeneity issues, we have found that local banking development has a significant negative effect on the Gini coefficient and other measures of inequality. When considering Italian macro areas (North, Center, South), the result is however significant only for the North, suggesting that the banking development-inequality nexus is not linear and depends on the stage of economic development. In the second part of the paper, we have taken a step towards investigating the channels of influence of banking development on inequality, with an emphasis on structural, socioeconomic aspects. We have studied the influence of urban structure, geographical mobility, material and immaterial infrastructures, human capital formation, and education. The results suggest that urban structure and geographical mobility may play a relatively important role in the finance-inequality relation. The tests instead suggest limited influence of material infrastructures and education in the finance-inequality nexus. We have argued that these results can reflect the predominant role played by the State (relative to the banking sector) in financing the development of infrastructures and the education system. Our empirical findings support the hypothesis that financial development affects the distribution of income. They also suggest relevant mechanisms of influence tied to the socioeconomic structure. More work is clearly needed to ascertain the contribution of socioeconomic, structural factors to the finance-inequality nexus. The findings can also suggest interesting insights for possible spillovers from Italy to other European countries. For example, to the extent that financial development affects geographical mobility and the intensity of migratory flows, this could have relevant spillover effects on other European countries, through the impact on migratory flows towards those countries. We leave this and other relevant issues for future research. Appendix A.

30

Seven and Coskun (2016) insert secondary school enrollment as a control in the empirical analysis.

Please cite this article in press as: D’Onofrio, A., et al., Banking development, socioeconomic structure and income inequality. J. Econ. Behav. Organ. (2017), http://dx.doi.org/10.1016/j.jebo.2017.08.006

G Model

ARTICLE IN PRESS

JEBO-4124; No. of Pages 24

A. D’Onofrio et al. / Journal of Economic Behavior & Organization xxx (2017) xxx–xxx

20

Table A.1 Data sources and variable definitions. Variable

Definition and source (in parentheses)

Main dependent variables Gini index (log)

Logarithm of Gini index at provincial level, computed starting by income data at municipial level. (MEF) Logarithm of Theil index at provincial level, computed starting by income data at municipial level. (MEF)

Theil index (log) Banking development Branch density (log)

Logarithm of bank branch density by province, number of branches normalized by the population. (BI and ISTAT) Herfindahl-Hirschman Index of branch concentration in the province. (BI) Logarithm of ATM density by province, number of ATM normalized by the population. (BI and ISTAT)

HHI of branches ATM density (log) Control variables Per capita GDP (log) Unemployment (log) Agriculture (share) Manufacturing (share) Construction (share) Trade Openess (log) Center South Small municipality 2001 (share) Gross flow 1991 (log) Net flow 1991 (log) Material infrastructure

Street index Railway index IT and broadband index Secondary degree 2001 Infrastructure for education index Cultural and recreation structures index Instrumental variables Savings banks in 1936 Number of branches in 1936

Logarithm of provincial GDP per capita. (ISTAT) Logarithm of provincial unemployment rate. (ISTAT) Share of total workers occupied in the Agriculture sector in the province. (ISTAT) Share of total workers occupied in the Manifacturing sector in the province. (ISTAT) Share of total workers occupied in the Manifacturing sector in the province. (ISTAT) Logarithm of the ratio of trade on GDP in the province. (ISTAT) Dummy that takes the value of one if the province is located in the central area of Italy; zero otherwise. (ISTAT) Dummy that takes the value of one if the province is located in a southern area of Italy; zero otherwise. (ISTAT) The percentage of population living in small municipalities (less than 15,000 inhabitants) in 2001. (ISTAT) Logarithm of gross flow of migrants in the province in 1991. (ISTAT) Logarithm of net flow of migrants in the province in 1991. (ISTAT) Synthetic index of material infrastructure in the province. This data contains informations about: Road Network, Railways, Ports, Airports, Environmental Energy Networks, Broadband Services, Business Structure. (GEOWEB) Index of road network infrastructure in the province. (GEOWEB) Index of railways infrastructure in the province. (GEOWEB) Index of IT and broadband services in the province. (GEOWEB) The percentage of the provincial population with at least a secondary school degree. (ISTAT) Index of Educational facilities in the province. (GEOWEB) Index of Cultural and recreation facilities in the province. (GEOWEB) Number of savings banks in the year 1936 in the province, per 100,000 inhabitants. (BI) Number of bank branches in the year 1936 in the province, per 100,000 inhabitants. (BI)

This table describes the definitions of the variables used in the paper. Four main data sources are used in the empirical analysis: (i) hand-collected data from the municipality-level database on tax revenue compiled by the Department of Finance of the Italian Ministry of Economy and Finance (MEF); (ii) the Statistical Bulletin of the Bank of Italy (BI); (iii) the province-level database of the Italian National Statistics Office (ISTAT); and (iv) Geoweb Starter, a database containing local, provincial and regional statistical information produced by the Istituto Guglielmo Tagliacarne (GEOWEB).

Table A.2 Summary statistics.

Torino Vercelli Novara Cuneo Asti Alessandria Aosta Imperia Savona Genova La Spezia Varese Como

Gini coefficient

Theil coefficient

Branches per 1000 inhab

GDP per capita

Unemployment rate

Mean

Std. Dev.

Mean

Std. Dev.

Mean

Std. Dev.

Mean

Std. Dev.

Mean

Std. Dev.

35.60 31.90 34.30 34.70 34.30 33.60 35.00 36.10 35.40 37.20 33.60 35.40 36.40

0.30 0.20 0.30 0.20 0.30 0.10 1.00 0.30 0.20 0.30 0.30 0.30 0.20

0.266 0.216 0.251 0.252 0.245 0.233 0.249 0.262 0.255 0.282 0.225 0.265 0.283

0.009 0.002 0.009 0.001 0.008 0.003 0.028 0.003 0.003 0.002 0.004 0.006 0.004

0.501 0.755 0.587 0.888 0.731 0.695 0.77 0.553 0.659 0.596 0.616 0.553 0.631

0.006 0.008 0.014 0.005 0.009 0.005 0.014 0.007 0.011 0.009 0.01 0.015 0.01

28228 28803 27960 30036 25032 27524 34227 25231 27698 27654 25400 29460 27646

985 1003 806 1007 1106 1168 903 1360 897 1123 811 1149 1513

6.436 4.658 6.115 2.922 4.729 4.979 3.655 6.276 4.773 5.366 5.735 4.351 4.541

2.323 0.806 1.407 0.506 1.26 0.51 0.703 2.19 0.492 0.787 1.282 1.424 0.826

Please cite this article in press as: D’Onofrio, A., et al., Banking development, socioeconomic structure and income inequality. J. Econ. Behav. Organ. (2017), http://dx.doi.org/10.1016/j.jebo.2017.08.006

G Model

ARTICLE IN PRESS

JEBO-4124; No. of Pages 24

A. D’Onofrio et al. / Journal of Economic Behavior & Organization xxx (2017) xxx–xxx

21

Table A.2 (Continued)

Sondrio Milano Bergamo Brescia Pavia Cremona Mantova Bolzano Trento Verona Vicenza Belluno Treviso Venezia Padova Rovigo Udine Gorizia Trieste Piacenza Parma Reggio Emilia Modena Bologna Ferrara Ravenna Forli’-Cesena Pesaro Urbino Ancona Macerata Ascoli Piceno Massa-Carrara Lucca Pistoia Firenze Livorno Pisa Arezzo Siena Grosseto Perugia Terni Viterbo Rieti Roma Latina Frosinone Caserta Benevento Napoli Avellino Salerno L’aquila Teramo Pescara Chieti Campobasso Foggia Bari Taranto Brindisi Lecce Potenza Matera Cosenza Catanzaro Reggio Calabria Trapani Palermo

Gini coefficient

Theil coefficient

Branches per 1000 inhab

GDP per capita

Unemployment rate

Mean

Std. Dev.

Mean

Std. Dev.

Mean

Std. Dev.

Mean

Std. Dev.

Mean

Std. Dev.

33.30 40.90 34.90 35.70 34.90 33.60 34.10 38.10 34.60 36.50 34.00 32.20 35.50 34.40 36.50 31.20 34.10 31.80 34.20 35.70 36.70 34.50 35.60 36.50 32.70 34.30 34.10 34.20 34.10 34.20 33.70 34.00 35.80 33.10 36.80 34.20 35.00 32.90 36.30 35.40 34.00 33.10 35.40 32.60 40.90 35.20 31.60 35.00 34.70 37.30 34.10 35.90 33.70 33.50 36.10 33.20 34.10 36.30 37.00 33.00 33.80 35.60 33.50 34.20 35.10 35.00 33.90 36.10 37.80

0.10 0.50 0.30 0.30 0.20 0.10 0.20 1.00 0.40 0.10 0.20 0.10 0.30 0.20 0.30 0.20 0.20 0.30 0.20 0.20 0.20 0.40 0.30 0.30 0.20 0.30 0.30 0.10 0.20 0.20 0.20 0.10 0.20 0.10 0.20 0.20 0.10 0.20 0.10 0.40 0.20 0.20 0.40 0.30 0.10 0.60 0.30 0.50 0.30 0.50 0.40 0.40 0.40 0.20 0.10 0.40 0.40 0.70 0.30 0.50 0.50 0.20 0.30 0.30 0.80 0.50 0.70 0.50 0.30

0.234 0.36 0.264 0.275 0.252 0.237 0.247 0.296 0.245 0.279 0.251 0.222 0.27 0.243 0.279 0.202 0.241 0.205 0.24 0.263 0.278 0.248 0.266 0.276 0.22 0.245 0.246 0.244 0.238 0.242 0.231 0.239 0.266 0.224 0.28 0.235 0.249 0.227 0.267 0.249 0.239 0.222 0.241 0.203 0.336 0.24 0.199 0.235 0.229 0.275 0.226 0.25 0.217 0.229 0.262 0.221 0.226 0.259 0.272 0.217 0.223 0.249 0.222 0.227 0.233 0.238 0.221 0.248 0.275

0.004 0.013 0.007 0.009 0.005 0.002 0.005 0.017 0.008 0.004 0.006 0.004 0.008 0.006 0.008 0.003 0.005 0.007 0.003 0.007 0.005 0.008 0.007 0.007 0.002 0.004 0.006 0.001 0.005 0.005 0.004 0.004 0.003 0.003 0.006 0.003 0.004 0.005 0.003 0.004 0.004 0.003 0.002 0.004 0.004 0.005 0.004 0.007 0.002 0.01 0.006 0.005 0.007 0.004 0.004 0.002 0.003 0.006 0.002 0.004 0.003 0.005 0.005 0.005 0.011 0.01 0.01 0.005 0.007

0.7 0.652 0.703 0.776 0.63 0.809 0.82 0.841 1.064 0.812 0.774 0.909 0.76 0.615 0.708 0.744 0.879 0.772 0.604 0.781 0.845 0.8 0.74 0.866 0.631 0.883 0.916 0.855 0.794 0.754 0.704 0.547 0.686 0.673 0.703 0.613 0.685 0.694 0.827 0.698 0.669 0.574 0.658 0.538 0.508 0.347 0.406 0.233 0.321 0.269 0.31 0.339 0.504 0.61 0.548 0.458 0.475 0.361 0.385 0.303 0.305 0.327 0.428 0.42 0.28 0.287 0.248 0.398 0.338

0.017 0.014 0.017 0.023 0.011 0.015 0.009 0.01 0.006 0.014 0.013 0.024 0.018 0.009 0.011 0.012 0.005 0.012 0.012 0.022 0.022 0.017 0.02 0.012 0.009 0.013 0.019 0.01 0.013 0.015 0.015 0.009 0.012 0.01 0.014 0.015 0.012 0.02 0.024 0.036 0.005 0.016 0.008 0.015 0.005 0.01 0.019 0.004 0.015 0.005 0.007 0.006 0.002 0.011 0.014 0.005 0.007 0.006 0.008 0.009 0.004 0.005 0.006 0.01 0.006 0.005 0.005 0.006 0.006

29498 37498 31447 31171 26906 28546 32110 35191 30256 30181 30281 29885 28938 29843 30052 26745 28548 26396 31154 29838 32126 30684 33412 34609 27182 29681 31537 25685 29114 24933 23609 22808 28213 25658 31060 26831 28005 26704 28153 25726 24638 23387 22073 21905 32359 23721 22982 15873 16794 16485 17795 17960 21038 21570 21521 21701 20433 15349 18232 17814 16406 16656 18478 17507 16887 18681 16006 16053 17500

1776 637 985 965 1220 1205 813 1055 728 485 873 863 859 754 624 1064 1071 630 522 1082 553 1120 1520 800 1768 667 1400 1075 746 716 1171 727 1439 635 420 546 889 576 899 851 668 776 861 1097 633 655 618 248 810 252 563 732 1024 577 810 612 1209 663 404 242 262 808 203 294 440 325 258 543 742

4.638 4.603 3.205 4.228 4.817 4.724 4.369 2.621 3.409 4.108 4.302 3.17 4.409 4.73 4.227 4.206 4.49 4.756 4.219 2.363 3.012 3.434 4.321 3.19 5.387 4.218 5.281 4.448 4.387 4.763 7.351 9.535 5.138 5.505 4.433 5.54 4.714 5.155 4.241 4.698 5.562 5.559 9.762 6.867 7.456 9.442 8.763 9.592 10.598 14.319 9.843 12.695 7.787 6.397 7.508 7.273 9.191 11.914 11.101 10.443 13.518 15.674 10.746 11.671 11.447 12.43 11.463 11.228 17.561

1.101 1.063 0.502 1.24 0.841 1.35 1.432 0.184 0.544 0.578 1.011 1.231 1.281 1.439 1.035 1.046 1.232 1.242 0.54 0.401 0.816 1.647 1.652 1.087 1.767 1.386 0.916 1.05 0.944 0.732 1.861 1.494 1.953 1.112 0.596 0.902 0.779 0.543 0.737 0.56 1.126 1.078 1.837 1.222 1.227 1.293 0.864 0.794 0.807 1.22 1.33 1.345 1.532 1.296 1.345 1.931 0.89 1.72 1.359 1.241 1.068 1.267 0.734 2.413 0.877 1.55 0.826 1.042 1.308

Please cite this article in press as: D’Onofrio, A., et al., Banking development, socioeconomic structure and income inequality. J. Econ. Behav. Organ. (2017), http://dx.doi.org/10.1016/j.jebo.2017.08.006

G Model

ARTICLE IN PRESS

JEBO-4124; No. of Pages 24

A. D’Onofrio et al. / Journal of Economic Behavior & Organization xxx (2017) xxx–xxx

22 Table A.2 (Continued)

Messina Agrigento Caltanissetta Enna Catania Ragusa Siracusa Sassari Nuoro Cagliari Pordenone Isernia Oristano Biella Lecco Lodi Rimini Prato Crotone Vibo Valentia Verbania

Gini coefficient

Theil coefficient

Branches per 1000 inhab

GDP per capita

Unemployment rate

Mean

Std. Dev.

Mean

Std. Dev.

Mean

Std. Dev.

Mean

Std. Dev.

Mean

Std. Dev.

35.30 35.50 35.20 35.10 37.20 37.60 35.60 35.30 31.90 34.70 33.70 34.30 31.80 32.60 35.40 33.20 36.50 34.80 34.70 33.80 33.20

0.30 0.30 0.20 0.20 0.30 0.50 0.30 0.50 0.50 0.60 0.40 0.30 0.50 0.40 0.20 0.30 0.20 0.30 0.90 0.60 0.10

0.239 0.24 0.236 0.235 0.272 0.272 0.24 0.247 0.194 0.237 0.238 0.235 0.195 0.236 0.266 0.228 0.272 0.251 0.233 0.216 0.232

0.007 0.007 0.005 0.005 0.008 0.004 0.004 0.008 0.009 0.01 0.01 0.007 0.007 0.011 0.005 0.005 0.003 0.006 0.014 0.009 0.005

0.359 0.369 0.368 0.389 0.341 0.396 0.317 0.438 0.445 0.364 0.734 0.388 0.494 0.708 0.697 0.722 0.983 0.573 0.217 0.246 0.548

0.005 0.006 0.011 0.008 0.008 0.011 0.006 0.01 0.045 0.008 0.011 0.013 0.025 0.016 0.019 0.039 0.017 0.011 0.004 0.007 0.013

17894 14642 16785 15649 16697 17849 18394 19024 18970 22244 29032 19833 17571 27504 29052 26162 30361 27687 14428 15260 23316

247 662 535 393 310 418 818 396 1150 586 1134 267 638 1146 1289 828 1696 331 407 392 1060

12.463 16.735 15.728 16.161 11.852 8.269 10.721 14.53 10.683 11.017 4.354 8.263 12.85 5.77 3.756 4.438 5.931 6.499 12.498 13.636 4.848

1.766 2.157 0.877 0.636 0.381 0.947 1.326 4.075 1.401 1.07 1.282 0.417 1.958 1.606 1.107 1.09 1.725 0.924 1.187 0.886 1.286

Table A.3 Robustness checks. Variables

(1)

(2)

(3)

Without unemployment

Branch density (log) Per capita GDP (log)

Manufacturing (share) Construction (share) Trade openness (log) Instruments Saving banks 36 N. branches 36 Region fixed effects Year fixed effects Observations R-squared F instruments Overid p value

(5)

(6)

Without GDP

Year 2001 2SLS Gini (log)

Average 2001–2011 2SLS Gini (log)

Panel 2001–2011 Arellano-Bond Gini (log)

Year 2001 2SLS Gini (log)

Average 2001–2011 2SLS Gini (log)

Panel 2001–2011 Arellano-Bond Gini (log)

−0.162** (0.064) 0.155** (0.065)

−0.147*** (0.051) 0.294*** (0.069)

−0.142*** (0.053) 0.255*** (0.061)

−0.152** (0.070)

−0.181** (0.083)

−0.131* (0.068)

−0.036 (0.023) −0.018 (0.093) −0.226*** (0.075) 0.033 (0.348) −0.008 (0.007)

−0.113** (0.045) −0.377*** (0.142) −0.269*** (0.084) −0.457 (0.335) −0.009 (0.008)

−0.041** (0.017) −0.294** (0.123) −0.194** (0.085) −0.458 (0.281) −0.005 (0.005)

0.138*** (0.043) 0.005** (0.002) Y N 103 0.578 9.900 0.452

0.129*** (0.041) 0.005** (0.002) Y N 103 0.224 10.55 0.679

Unemployment (log) Agriculture (share)

(4)

0.208 (0.147) −0.180*** (0.070) 0.207 (0.359) −0.011 (0.008)

−0.069 (0.149) −0.179** (0.077) −0.176 (0.301) −0.013* (0.008)

0.130*** (0.049) 0.005*** (0.002) Y N 103 0.589 8.218 0.361

0.161*** (0.048) 0.005** (0.002) Y N 103 0.229 11.55 0.884

−0.041 (0.136) −0.183** (0.076) −0.244 (0.251) −0.009 (0.006)

Y Y 1133

0.998

Y Y 1133

0.965

Note: The table reports regression coefficients and associated standard errors (in parentheses). The time-span of the regressions, the dependent variables and the estimation method are reported at the top of each column. The set of instruments includes the number of bank branches in the province in 1936 (per 100,000 inhabitants) and the number of savings banks in the province in 1936 (per 100,000 inhabitants). ***Significant at 1%, **significant at 5%, *significant at 10%. The table reports the value of the R2 and the value of the F-statistics for a test of the weakness of the instruments. The table also reports the p-values of a Sargan test, as a test of overidentifying restrictions.

Please cite this article in press as: D’Onofrio, A., et al., Banking development, socioeconomic structure and income inequality. J. Econ. Behav. Organ. (2017), http://dx.doi.org/10.1016/j.jebo.2017.08.006

G Model JEBO-4124; No. of Pages 24

ARTICLE IN PRESS A. D’Onofrio et al. / Journal of Economic Behavior & Organization xxx (2017) xxx–xxx

23

References Acciari, P., Mocetti, S., 2013. The geography of income inequality in Italy. In: Questioni di Economia e Finanza (Occasional Papers) 208. Bank of Italy, Economic Research and International Relations Area. Acemoglu, D., Zilibotti, F., 1997. Was Prometheus unbound by chance? Risk, diversification, and growth. J. Polit. Econ. 105, 709–751. Aghion, P., Bolton, P., 1992. Distribution and growth in models of imperfect capital markets. Eur. Econ. Rev. 36, 603–611. Aghion, P., Bolton, P., 1997. A theory of trickle-down growth and development. Rev. Econ. Stud. 64, 151–172. Aghion, P., Caroli, E., García-Pe nalosa, C., 1999. Inequality and economic growth: the perspective of the new growth theories. J. Econ. Lit. 37, 1615–1660. Ang, J.B., 2010. Finance and inequality: the case of India. South. Econ. J. 76, 738–761. Ascher, W., Krupp, C., 2010. Distributional implications of alternative financing of physical infrastructure development. In: Ascher, W., Krupp, C. (Eds.), Physical Infrastructure Development. Balancing the Growth, Equity, and Environmental Imperatives. Palgrave, New York. Autor, D.H., 2014. Skills, education, and the rise of earnings inequality among the “other 99 percent”. Science 344, 843–851. Bacolod, M., Blum, B.S., Strange, W.S., 2009. Skills and the city. J. Urban Econ. 65, 127–135. Bagchi, S., Svejnar, J., 2015. Does wealth inequality matter for growth? The effect of billionaire wealth, income distribution, and poverty. J. Comp. Econ. 43, 505–530. Banerjee, A.V., Newman, A.F., 1993. Occupational choice and the process of development. J. Polit. Econ. 101, 274–298. Baum-Snow, N., Pavan, R., 2013. Inequality and city size. Rev. Econ. Stat. 95, 1535–1548. Beck, T., Demirguc-Kunt, A., Levine, R., 2007a. Finance, inequality, and the poor. J. Econ. Growth 12, 27–49. Beck, T., Demirguc-Kunt, A., Martinez Peria, M.S., 2007b. Reaching out: access to and use of banking services across countries. J. Financ. Econ. 85, 234–266. Beck, T., Demirguc-Kunt, A., 2008. Access to finance: an unfinished agenda. World Bank Econ. Rev. 22, 383–396. Beck, T., Levine, R., Levkov, A., 2010. Big bad banks? The winners and losers from bank deregulation in the United States. J. Finance 65, 1637–1667. Behrens, K., Robert-Nicoud, F., 2014a. Survival of the fittest in cities: urbanisation and inequality. Econ. J. 124, 1371–1400. Behrens, K., Robert-Nicoud, F., 2014b. Urbanisation Makes the World More Unequal. Vox, CEPR’s Policy Portal. Benfratello, L., Schiantarelli, F., Sembenelli, A., 2008. Banks and innovation: microeconometric evidence on Italian firms. J. Financ. Econ. 90, 197–217. Berger, A.N., Demirgüc¸-Kunt, A., Levine, R., Haubrich, J.G., 2004. Bank concentration and competition: an evolution in the making. J. Money Credit Bank. 36, 433–451. Blau, F., Kahn, L., 2015. Immigration and the distribution of incomes. In: Chiswick, B.R., Miller, P.W. (Eds.), Handbook of the Economics of International Migration. Elsevier, Amsterdam, pp. 793–843. Bofondi, M., Gobbi, G., 2006. Informational barriers to entry into credit markets. Rev. Finance 10, 39–67. Bonaccorsi Di Patti, E., Gobbi, G., 2001. The changing structure of local credit markets: are small business special? J. Bank. Finance 25, 2209–2237. Bourguignon, F., 2001. The distributional effects of growth: micro vs. macro approaches. In: Sede de la CEPAL en Santiago (Estudios e Investigaciones) 34859. Naciones Unidas Comisión Económica para América Latina y el Caribe (CEPAL). Buera, F., 2009. A dynamic model of entrepreneurship with borrowing constraints: theory and evidence. Ann. Finance 5, 443–464. Burgess, R., Pande, R., 2005. Do rural banks matter? Evidence from the Indian social banking experiment. Am. Econ. Rev. 95, 780–795. Calderon, C., Serven, L., 2004. The Effects of Infrastructure Development on Growth and Income Distribution. World Bank Policy Research Working Paper., pp. 3400. Card, D., 2009. Immigration and inequality. Am. Econ. Rev. 99, 1–21. Charlton, A., Stiglitz, J., 2008. Capital market liberalization and poverty. In: Ocampo, J.A., Stiglitz, J. (Eds.), Capital Market Liberalization and Development. Oxford Press, pp. 121–138. Claessens, S., Feijen, E., 2007. Financial Sector Development and the Millennium Development Goals. World Bank Working Paper No. 89. Claessens, S., Perotti, E., 2007. Finance and inequality: channels and evidence. J. Comp. Econ. 35, 748–773. Clarke, G., Xu, L.C., Zou, H., 2006. Finance and income inequality: what do the data tell us? South. Econ. J. 76, 578–596. De Bonis, R., Pozzolo, A.F., Stacchini, M., 2012. The Italian Banking System: Facts and Interpretations. Economics and Statistics Discussion Papers esdp12068. University of Molise, Dept. SEGeS. De Rosa, L., 1980. Emigranti, Capitali e Banche, Napoli. Banco di Napoli (in Italian). Degryse, H., Ongena, S., 2005. Distance, lending relationships, and competition. J. Finance 60, 231–266. Deidda, L.G., 2006. Interaction between economic and financial development. J. Monet. Econ. 53, 233–248. Deininger, K., Squire, L., 1998. New ways of looking at old issues: inequality and growth. J. Dev. Econ. 57, 259–287. Demirguc-Kunt, A., Levine, R., 2009. Finance and inequality: theory and evidence. Annu. Rev. Financ. Econ. Annu. Rev. 1 (November (1)), 287–318. Demirguc-Kunt, A., Levine, R., 2008. Finance and Economic Opportunity. Policy Research Working Paper Series 4468. The World Bank. Dollar, D., Kraay, A., 2002. Growth is good for the poor. J. Econ. Growth 7, 195–225. Faber, B., Gaubert, C., 2016. Tourism and Economic Development: Evidence from Mexico’s Coastline. Working Paper. University of California, Berkeley. Farina, F., Franzini, M., 2015. La Casa, il Benessere e le Disuguaglianze. Egea Editore, Milan (in Italian). Ferreira, F.H.G., 1995. Roads to Equality: Wealth Distribution Dynamics with Public-Private Capital Complementarity. LSE-STICERD Discussion Paper TE/95/286. London School of Economics, London. Galor, O., Moav, O., 2004. From physical to human capital accumulation: inequality and the process of development. Rev. Econ. Stud. 71, 1001–1026. Galor, O., Zeira, J., 1993. Income distribution and macroeconomics. Rev. Econ. Stud. 60, 35–52. Gine, X., Townsend, R., 2004. Evaluation of financial liberalization: a general equilibrium model with constrained occupation choice. J. Dev. Econ. 74, 269–307. Greenwood, J., Jovanovic, B., 1990. Financial development, growth, and the distribution of income. J. Polit. Econ. 98, 1076–1107. Greenwood, J., Smith, B.D., 1997. Financial markets in development, and the development of financial markets. J. Econ. Dyn. Control 21, 145–181. Guiso, L., Pistaferri, L., Schivardi, F., 2013. Credit within the firm. Rev. Econ. Stud. 80, 211–247. Guiso, L., Sapienza, P., Zingales, L., 2003. The Cost of Banking Regulation. Working Paper. Chicago Graduate School of Business. Guiso, L., Sapienza, P., Zingales, L., 2004. Does local financial development matter? Q. J. Econ. 119, 929–969. Herrera, A., Minetti, R., 2007. Informed finance and technological change: evidence from credit relationships. J. Financ. Econ. 83, 223–269. Jauch, S., Watzka, S., 2015. Financial development and income inequality: a panel data approach. Empir. Econ. 51, 1–24. Kaldor, N., 1955. Alternative theories of distribution. Rev. Econ. Stud. 23, 83–100. Kappel, V., 2010. The Effects of Financial Development on Income Inequality and Poverty. CER-ETH Working Paper 10/127. Kim, D.-H., Lin, S.-L., 2011. Nonlinearity in the financial development-income inequality nexus. J. Comp. Econ. 39, 310–325. Kpodar, K., Singh, R.J., 2011. Does Financial Structure Matter for Poverty? Evidence from Developing Countries. Policy Research Working Paper Series 5915. The World Bank. Kuznets, S., 1955. Economic growth and income inequality. Am. Econ. Rev. 45, 1–28. Law, S., Tan, H.B., 2009. The role of financial development on income inequality in Malaysia. J. Econ. Dev. 34, 153–168. Lee, J., 1996. Financial development by learning. J. Dev. Econ. 50, 147–164. Levine, R., 2005. Finance and growth: theory and evidence. In: Aghion, P., Durlauf, S.N. (Eds.), Handbook of Economic Growth, vol. 1A. Elsevier North Holland, Amsterdam, pp. 865–934. Lewis, W.A., 1954. Economic development with unlimited supplies of labor. Manch. Sch. 22, 139–191. Li, H., Squire, L., Zou, H., 1998. Explaining international and intertemporal variations in income inequality. Econ. J. 108, 26–43. Lopez, H., 2003. Macroeconomics and Inequality. World Bank Working Paper, Washington D.C.

Please cite this article in press as: D’Onofrio, A., et al., Banking development, socioeconomic structure and income inequality. J. Econ. Behav. Organ. (2017), http://dx.doi.org/10.1016/j.jebo.2017.08.006

G Model JEBO-4124; No. of Pages 24

24

ARTICLE IN PRESS A. D’Onofrio et al. / Journal of Economic Behavior & Organization xxx (2017) xxx–xxx

Manova, K., 2013. Credit constraints, heterogeneous firms, and international trade. Rev. Econ. Stud. 80, 711–744. Matsuyama, K., 2000. Endogenous inequality. Rev. Econ. Stud. 67, 743–759. Minetti, R., 2011. Informed finance and technological conservatism. Rev. Finance 15, 633–692. Minetti, R., Murro, P., Paiella, M., 2015. Ownership structure, governance, and innovation. Eur. Econ. Rev. 80, 165–193. Mocetti, S., Porello, C., 2010. La mobilità del lavoro in Italia: nuove evidenze sulle dinamiche migratorie. Bank of Italy Occasional Papers 61. Mookherjee, D., Ray, D., 2003. Persistent inequality. Rev. Econ. Stud. 70, 369–393. Petersen, M., Rajan, R., 1994. The benefits of lending relationships: evidence from small business data. J. Finance 49, 3–37. Petersen, M., Rajan, R., 2002. Does distance still matter? The information revolution in small business lending. J. Finance 57, 2533–2570. Presbitero, A.F., Rebelotti, R., 2014. Geographical distance and moral hazard in microcredit: evidence from Colombia. J. Int. Dev. 26, 91–108. Ravallion, M., 2001. Growth, inequality and poverty: looking beyond averages. World Dev. 29, 1803–1815. Rousseau, P.L., Wachtel, P., 2002. Inflation thresholds and the finance-growth nexus. J. Int. Money Finance 21, 777–793. Seven, U., Coskun, Y., 2016. Does financial development reduce income inequality and poverty? Evidence from emerging countries. Emerg. Mark. Rev. 26, 34–63. The Economist, 2012. For Richer, For Poorer (Special Report, October issue). Townsend, R.M., Ueda, K., 2006. Financial deepening, inequality, and growth: a model-based quantitative evaluation. Rev. Econ. Stud. 73, 251–280. Williamson, J., 1965. Regional inequality and the process of national development. Econ. Dev. Cult. Change 14, 3–45. World Bank, 2003. Inequality in Latin America and the Caribbean. World Bank Latin American and Caribbean Studies. Yilmazkuday, H., 2011. Thresholds in the finance-growth nexus: a cross-country analysis. World Bank Econ. Rev. 25, 278–295.

Please cite this article in press as: D’Onofrio, A., et al., Banking development, socioeconomic structure and income inequality. J. Econ. Behav. Organ. (2017), http://dx.doi.org/10.1016/j.jebo.2017.08.006