Local financial development, socio-institutional environment, and firm productivity: Evidence from Italy

European Journal of Political Economy 35 (2014) 38–51

Contents lists available at ScienceDirect

European Journal of Political Economy journal homepage: www.elsevier.com/locate/ejpe

Local financial development, socio-institutional environment, and firm productivity: Evidence from Italy Luigi Moretti Department of Economics and Management, University of Padova, via del Santo 33, 35123 Padova, Italy

a r t i c l e

i n f o

Article history: Received 8 August 2013 Received in revised form 23 March 2014 Accepted 24 March 2014 Available online 2 April 2014 JEL classification: D2 G2 016 Keywords: Finance–growth nexus Social capital Judicial enforcement Firm labor productivity

a b s t r a c t This paper uses the case of Italy to investigate the effects of local financial and socioinstitutional development on productivity. The analysis employs firm-level productivity data and exploits variations in banking sector development, judicial enforcement, and social capital across Italian provinces. After controlling for potential endogeneity, our empirical results suggest that the real effects of financial development are conditional on the quality of the socio-institutional environment. In particular, we find that the positive effects of greater financial depth on productivity are stronger when the socio-institutional environment is sufficiently developed. Therefore, to exploit potential productivity gains stimulated by financial development, it is necessary to achieve a higher-quality socio-institutional environment, including reducing the duration of civil trials. © 2014 Elsevier B.V. All rights reserved.

1. Introduction When we observe substantial differences in output per worker among firms, it is important to know whether the causes are related to a lack of highly productive investment opportunities in the economy or to barriers that prevent firms from exploiting these opportunities. Insufficient availability of credit is frequently indicated as a barrier to entry, innovation, productivity, and growth, and improved access to financial resources typically has positive effects on the performance of the real sector (for a survey, see Levine, 2005). However, there is evidence that certain of these relationships are non-linear and the effects of financial development might depend on several factors, such as the initial level of financial development (see for instance, Rioja and Valev, 2004a, 2004b; Coricelli et al., 2008). The objective of this paper is to further explore the non-linearity of the effects of financial development. In particular, this paper gauges whether increasing availability of credit has differential effects on firms' productivity in areas with socioinstitutional environments of varying quality (i.e., in areas with different productive opportunities). We focus on the case of Italy, which is particularly appealing because Italy is characterized both by large differences in terms of income and productivity across provinces and by highly segmented local financial markets, as effectively documented by Guiso et al. (2004a, 2004b). Similarly, the efficiency of judicial enforcement, the presence of organized crime, and, in general, the level of social capital and the quality of institutions vary widely among Italian provinces. In particular, there is a sharp divide between the

E-mail address: [email protected].

http://dx.doi.org/10.1016/j.ejpoleco.2014.03.006 0176-2680/© 2014 Elsevier B.V. All rights reserved.

L. Moretti / European Journal of Political Economy 35 (2014) 38–51

39

north and the south of Italy with respect to each of these factors. However, there is also significant variation within these two large regions that we are able to capture by analyzing information at the province level. Our empirical analysis covers the 2000–2007 period, and we employ firm-level data for productivity and province-level data for the characteristics of the financial markets and socio-institutional environments. Our main results show that greater financial depth, measured as the ratio of bank credit to value added, has positive and significant effects on firms' productivity only in provinces with better socio-institutional environments. In terms of policy implications, our main results indicate that socio-institutional development is a necessary condition to achieve returns in productivity from greater credit availability to the private sector. We characterize the socio-institutional environment with a composite index that includes variables for social capital, violent crime, and the quality of judicial enforcement, and we also focus in particular on how (and to what extend) the length of civil trials and bankruptcy proceedings condition the effects of financial development. Reforms aimed at improving judicial efficiency are high on the policy agendas of numerous countries, including Italy, and are typically viewed as a fundamental step in improving the enforcement of contracts. All else being equal, the results show that an increase in the availability of credit has a larger effect on firm productivity in provinces with shorter civil trials and bankruptcy proceedings. Our findings indicate that the sequence of reforms is crucial, which is an idea that was previously suggested by Hausmann et al. (2005), among others, who emphasized that the timing of economic and institutional reforms i) determines the effectiveness of the reforms' impact on the economy and ii) depends on the initial context and the country's characteristics. Furthermore, reforms can be divided into two categories: those that affect barriers to economic growth and those that affect the opportunities for such growth. Removing barriers in an environment in which opportunities for growth are limited or absent will not be effective. Accordingly, reforms that aim to remove barriers to growth should be enacted either jointly with or after reforms that aim to make the environment conducive for the creation of business opportunities. In our analysis, as in Hausmann et al. (2005), reforms that affect the financial sector impact barriers to growth, whereas reforms that affect the socio-institutional context determine the set of growth opportunities. The rest of the paper is organized as follows. Section 2 reviews the relevant literature and pays particular attention to empirical evidence. Section 3 contains our empirical analysis: in Section 3.1, we present the data and the model specification; in Section 3.2, we discuss the econometric strategy and the main estimation results; in Section 3.3, we show some robustness checks related to the use of different indicators for the quality of the socio-institutional environment; and in Section 3.4, we present the results related to the role of an efficient system of judicial enforcement. Finally, Section 4 concludes. 2. Credit, socio-institutional factors, and productivity: a background The idea that the effects of local financial development on productivity are conditioned on socio-institutional development builds on earlier literature that focuses on the following issues: i) the role of financial development, ii) the direct role of socio-institutional development on productivity, and iii) the indirect role of socio-institutional development through its effect on the financial system. The objective of this section is to briefly review the previous contributions on which we base our testable hypothesis. Numerous studies have shown the positive effects of financial development on several aspects of the real sector (see Levine, 2005 for a review). In this paper, we are interested in the relationship between the availability of financial resources in local banking markets and firm efficiency, as proxied by firm productivity.1 Firms' adoption of new production processes or technologies depends on the availability of external finance, which is more easily accessible when banking markets, including those at the local level, are more developed. Credit availability allows firms to borrow and spread costs over time, which increases their efficiency and the likelihood that they will engage in innovation. Furthermore, deeper financial markets are typically associated with better screening of projects and entrepreneurs, which allows credit to be more appropriately allocated toward high-productivity firms (for a discussion, see Pagés, 2010). Firms also typically suffer as a result of institutional underdevelopment (Beck et al., 2005). In this paper, we focus our attention on three socio-institutional dimensions: the quality of the judicial enforcement, the level of social capital, and the presence of criminal activities. These factors vary significantly across Italian territories and are relevant to explaining firms' productivity. The quality of judicial enforcement influences transactions between private actors and between private actors and public institutions (see, for instance, Djankov et al., 2003; Claessens and Klapper, 2005). Better and faster court enforcement reduces uncertainty among the parties to a transaction and therefore reduces opportunistic behaviors and costs. Certainty in transactions allows firms to rely on a wide spectrum of relationship opportunities, even when personal or network interactions are not yet established between the parties. Previous studies have focused on a single country and have exploited local variations in the quality of the judicial enforcement to assess its effects on firm performance and productivity.2 In particular, Ponticelli (2013) argued that firms face fixed costs when adopting technology and their capacity to pay the fixed costs depends on the quality of contractual enforcement by the courts. He studied Brazil and found that the introduction of improvements in bankruptcy legislation had a larger positive effect on firm productivity in counties with less congested courts (i.e., in counties with a higher-quality court enforcement). Like the legal system, active social networks also reduce opportunistic behavior by enforcing informal norms (reputation), facilitating the dissemination of information and ideas, and encouraging cooperation among a network's members (Coleman, 1988). 1 A greater quantity of (and easier access to) credit in local banking markets can also increase firm entry (Black and Strahan, 2002; Bonaccorsi di Patti and Dell'Ariccia, 2004; Cetorelli and Strahan, 2006) and innovation (Benfratello et al., 2008), which can ultimately improve the efficiency of the economic system and economic growth (Guiso et al., 2004a; Vaona, 2008; Hasan et al., 2009; Fernandez de Guevara and Maudos, 2009). 2 See, for instance, Chemin (2012), who studied the effects of judicial reform implemented in India in 2002 and found that increasing the speedy disposal of civil suits resulted in fewer breaches of contract, encouraged investment, and facilitated firms' access to finance; and Laeven and Woodruff (2007), who analyzed the legal-economic framework in Mexico and showed that the legal system affects firm size by reducing idiosyncratic risk faced by firm owners.

40

L. Moretti / European Journal of Political Economy 35 (2014) 38–51

More specifically, stronger ties among members of a social network influence a firm's productivity through increased control of workers' efforts, the enforcement of repeated interactions, and better matching in the job search process (see Granovetter, 2005).3 Finally, certain types of criminal activities may negatively affect a firm's organizational choices and performance. The presence of organized criminal activities in a territory increases firms' risks, adds extortion costs, and prevents firms from fully organizing their production processes because of retail market limitations (e.g., by obliging firms to acquire goods from crime-connected firms) or market distortions (e.g., caused by the participation of crime-connected firms in markets). The previous literature has found that criminal activities negatively affect various economic outcomes, including firm productivity.4 For example, Albanese and Marinelli (2013) found that, in Italy, organized crime significantly reduces productivity and explains approximately 10% of the productivity differential between firms in provinces with high levels of mafia-related crimes and those in provinces with low levels of such crimes. The empirical literature has also identified these social and institutional factors as drivers of local financial development, which shifts the focus onto their roles in the functioning of credit markets. The substantial differences that we observe in Italy in terms of the depths of the provincial banking markets cannot be attributed to the origin of the legal system, as proposed by La Porta et al. (1997b, 1998) in studies that examined cross-country differences. Italy has been an integrated country since 1861, and since that time financial regulations and creditors' rights are common throughout the country. Nonetheless, important differences regarding credit availability persist across Italian provinces; one reason for these differences might be the varying quality of the judicial enforcement, as Jappelli et al. (2005) argue. Better judicial enforcement reduces opportunistic behaviors and increases credit availability and lending by facilitating more secure relationships between firms and banks. Another important driver of financial development at the local level is social capital, which is a measure of trust, which facilitates transactions, including financial transactions. Guiso et al. (2004b) showed that, in Italy, households in areas with high levels of social capital invest less in cash and more in stocks, are more likely to use checks, have better access to institutional credit, and make less use of informal credit. Moreover, high levels of criminal activities in certain areas of Italy offer additional insight into the explanations for different levels of local banking market development. Bonaccorsi di Patti (2009) found a positive correlation between crime rates and difficulties in accessing credit in Italian provinces and argued that organized crime affects the loan market because it increases both the fragility of firms (e.g., through extortion) and the expected losses from loan defaults (e.g., because of fraud and fraudulent bankruptcy). Lower levels of social capital and the presence of criminal organizations can also increase the power of personal or group interests, which in turn might negatively influence the availability of credit and favor politically connected incumbent firms. At the cross-country level, it has been shown that group interests or different levels of democracy shape the development of financial systems (Rajan and Zingales, 2003; Becerra et al., 2012; Campos and Coricelli, 2012). In a within-country analysis of Italy, social capital—defined as the combination of rules, networks, and trust among people that facilitates the achievement of collective goals and the functioning of political institutions (Putnam, 1993)—was shown to reduce the influence of personal or group interests and to improve political accountability (see Nannicini et al., 2013). This finding indicates that social institutional factors not only have independent effects on financial development but also might interact with and reinforce one another.5 In this paper, we will not explicitly consider the role of interest groups because it would require a particular collection of data and a dedicated analysis. In sum, the previous literature indicates the following: i) a greater availability of credit positively affects firm efficiency and productivity; ii) socio-institutional factors—such as social capital, the efficiency of the judicial enforcement, and the presence of crime—directly influence firm efficiency through their effects on the costs and opportunities to develop profitable business activities; and iii) a better socio-institutional environment favors financial transactions and financial development through a variety of channels. In this paper, we extend the previous literature by focusing on a new channel, i.e., the possible interaction between credit and socio-institutional development as a determinant of firm productivity.6 We test the hypothesis that conditional factors for the effects of financial development are the socio-institutional characteristics of the economy. Specifically, if contracts are secured by an efficient court system or by moral reputation, if social networks help circulate information and create more efficient cooperation and institutions, and if the absence (or low level) of criminal activities and 3 A higher level of social capital is also found to be associated with higher innovation (e.g., Fountain, 1997; Akçomak and ter Weel, 2009), firm size (La Porta et al., 1997a; Bloom et al., 2012), and, ultimately, economic growth, prosperity, and competitiveness (e.g., Fukuyama, 1995; Beugelsdijk et al., 2004; Beugelsdijk and Van Schaik, 2005). Certain notable interaction effects have been found by Ahlerup et al. (2009), who showed that the effect of social capital on growth is more pronounced when there is lower institutional development. 4 In the Italian context, it has been found that the presence of organized crime also has negative effects on the labor market (Peri, 2004; Mauro and Carmeci, 2007), foreign investment (Daniele and Mariani, 2011), and overall economic growth (Detotto and Otranto, 2010; Pinotti, 2012). 5 According to Knack and Keefer (1997, 1254) “trust and civic norms may improve economic outcomes indirectly, through political channels. They may improve governmental performance and the quality of economic policies by affecting the level and character of political participation”. See also Del Monte and Papagni (2007), who found that higher levels of social capital in Italy can reduce corruption, and Haber et al. (2008) for a selection of works on the links between political institutions and financial development, including within-country studies on Brazil, Mexico, and the U.S. 6 This interaction can also give some insights into the non-linearity of the finance–growth nexus at the cross-country level (e.g., Rioja and Valev, 2004a, 2004b; Coricelli et al., 2008). For within-country studies on non-linearities in the finance–growth nexus, see, among others, Kendall (2012), who found a non-linear relationship between finance and growth in India and emphasized the role of deepening human capital in reducing financial constraints; and Guariglia and Poncet (2008), who found that provinces in China that were characterized by higher FDI stocks relative to GDP tend to suffer less from the negative effects of financial development in terms of growth and productivity. Our focus on the interaction effect also expands the evidence relating to the roles of both institutions and finance in firms' outcomes as, for example, shown by Johnson et al. (2002), who used a survey of firms in post-communist countries and found that the securing of property rights has a positive effect on investment decisions, even after controlling for the availability of bank loans. Our analysis differs from Johnson et al. (2002) in several ways. First, our outcome variable is firm productivity. In addition, we do not take into account the securing of property rights (which is not a risk in an advanced market economy such as Italy and is the same throughout the country) but instead study the roles of judicial efficiency, the presence of criminal activities, and the level of social capital. Finally, we look at the interaction effects between the availability of external finance and an indicator of socio-institutional environment.

L. Moretti / European Journal of Political Economy 35 (2014) 38–51

41

organized crime corresponds to an absence (or low level) of additional costs and market distortion, then greater quantities of and easier access to credit can allow firms to better organize their production and better exploit productive opportunities, which will lead to increased productivity. However, if the levels of these socio-institutional factors are such that they limit the widespread creation of productive opportunities, additional credit will not have a clear and positive effect on firm productivity. In addition, the new financial resources that would be allocated might not be channeled efficiently toward the most productive firms. Using productivity as a proxy for the efficiency of the system and employing firm-level data, we provide evidence of the interaction between local financial development and the socio-institutional environment. In the next section, we present our data and then follow the previous literature by describing our choice and construction of the indicators of financial development and socio-institutional environment, in addition to the econometric strategies used to better identify the relationships under analysis. 3. The empirical analysis 3.1. Data and estimated equation To estimate whether local financial development has an effect on firm productivity—and whether this effect depends on the development of the social and institutional environment—we assume that the economy's production is represented by a Cobb– Douglas function, which can be specified in per-worker terms and expressed in logarithmic form. For each firm c (located in province p) and each year t, the productivity of labor (VA/L) is represented as the ratio between the value added and the number of employees, and the capital per worker (K/L) is represented as the ratio between fixed capital stock and the number of employees. We can then add additional firm-level control variables and our provincial-level variables of interest to this function. Our estimated model is as follows:     ln VAcpt =Lcpt ¼ β0 þ β1 FDpt þ β2 SI pt þ β3 FDpt  SI pt þ β4 C cpt þ β5 ln K cpt =Lcpt þ εcpt

ð1Þ

where FD is the financial development measured for each province p for each year t, SI is the socio-institutional environment measured at the provincial level for each year, FD ∗ SI is the interaction between these two measures, and C is a vector of additional control variables (e.g., age, size, and leverage) for each firm c and each year t. The error term captures all factors that influence the productivity of labor but are not captured by the model specification's variables and consists of the following: i) firm-specific time-invariant effects, ii) an idiosyncratic component of time-varying firm-specific effects, and iii) time-varying macro effects that influence all firms. 3.1.1. Sample and dependent variable We employed firm-level data from the Aida-Bureau Van Dijk database, a comprehensive and harmonized database that contains information on the balance sheets and performances of Italian firms.7 Our final sample included firms operating in manufacturing, construction, transportation services, tourism, and market services during the 2000–2007 period and consisted of 532,315 observations for 173,500 firms (because the number of firms varied during the 8 years of analysis, we had to rely on an unbalanced panel). We computed the dependent variable as the (natural log of the) ratio between a firm's real value added and the number of its employees (ln(VA/L)). As expected, the summary statistics for firm labor productivity show lower values for firms located in southern Italy than for firms located in the rest of the country (see Table 1, panel A, for summary statistics of firm-level variables). 3.1.2. Firm-level control variables We also extracted certain firm-level control variables from the Aida database, including the capital labor ratio (ln(K/L)), which is (the natural log of) the ratio between real fixed capital and number of employees; firm size (ln(size)), which is the (natural log of) total assets; the (natural log of) firm age (ln(age)); and a measure of firm leverage (leverage, which is short-term debt plus long-term debt over total assets). All these factors can have a direct and significant effect on a firm's productivity and they need to be included in the model specification to reduce problems related to omitted time-varying variables. 3.1.3. Local financial development Our indicator of local financial development (FD) measures the depth of the banking market and is commonly used in empirical analyses of the finance–growth nexus (see, for example, Levine, 2005; Beck, 2008). This indicator is defined as the ratio of bank loans (credit) to the productive sector and value added (loan data are from the Bank of Italy and value added data are from the Italian National Institute of Statistics — ISTAT). Higher values of the indicator indicate higher availability of external finance that can be used to finance projects and upgrade productivity. The indicator is computed at the province level for each year during the 2000–2007 period, and it reveals cross-provincial and over-time differences in terms of credit provided to the non-financial firms by banks relative to the size of the provincial economy. The variation of this variable at the provincial level (see Fig. 1A) and over time (Fig. 1B) allows us to exploit differences in terms of financial development not only between the center-north and the south of Italy but also the intra-region segmentation of the credit markets. On average, the credit-tovalue-added ratio is approximately 33% in the southern provinces, compared with 55% in the northern provinces, and important 7

See Online Appendix A.1 for the data-cleaning criteria.

42

L. Moretti / European Journal of Political Economy 35 (2014) 38–51

Table 1 Summary statistics. Variable Panel A: firm-level variables Full sample ln(VA/L) ln(K/L) ln(size) Leverage ln(age) Center-north ln(VA/L) ln(K/L) ln(size) Leverage ln(age) South ln(VA/L) ln(K/L) ln(size) Leverage ln(age)

Obs.

Mean

sd

p25

p50

p75

532,315 532,315 532,315 532,315 532,315

3.885 3.015 7.871 0.673 2.595

0.532 1.454 1.286 0.247 0.791

3.550 2.041 6.932 0.544 2.079

3.814 3.045 7.721 0.739 2.708

4.152 3.979 8.695 0.863 3.135

449,034 449,034 449,034 449,034 449,034

3.899 2.954 7.907 0.671 2.634

0.528 1.439 1.291 0.247 0.784

3.562 1.992 6.960 0.542 2.197

3.824 2.983 7.751 0.738 2.773

4.163 3.902 8.736 0.861 3.178

83,281 83,281 83,281 83,281 83,281

3.809 3.344 7.677 0.683 2.384

0.548 1.491 1.243 0.243 0.796

3.482 2.343 6.786 0.551 1.946

3.756 3.411 7.558 0.744 2.485

4.093 4.365 8.460 0.873 2.996

824 824 823 824

0.332 0.722 0.621 0.571

0.189 0.108 0.163 0.166

0.183 0.673 0.525 0.469

0.326 0.744 0.645 0.576

0.451 0.797 0.743 0.693

536 536 536 536

0.415 0.772 0.692 0.603

0.166 0.059 0.118 0.152

0.307 0.733 0.625 0.507

0.399 0.776 0.705 0.606

0.509 0.812 0.777 0.722

288 288 287 288

0.177 0.627 0.489 0.512

0.120 0.115 0.151 0.173

0.080 0.547 0.401 0.404

0.155 0.643 0.507 0.523

0.242 0.715 0.602 0.623

Panel B: province-level variables Full sample FD SI Civil trials Bankruptcy Center-north FD SI Civil trials Bankruptcy South FD SI Civil trials Bankruptcy

Note: Panel A: Data refer to variables at the firm level for an unbalanced panel of 173,500 Italian firms in the period between 2000 and 2007. Data are from the AIDA-Bureau Van Dijk database. ln(VA/L) is the log of real value added per worker; ln(K/L) is the log of real fixed capital per worker; ln(size) is the log of real total assets; leverage is given by short-term debt plus long-term debt over total assets; and ln(age) is the log of firm age. Panel B: FD is defined as (the standardized values of) the ratio between bank loans to the productive sector and the value added for each province in each year in the 2000–2007 period. SI is our main indicator of the quality of socio-institutional environments and is computed as the simple average of (i) one minus the standardized values of the duration of bankruptcy proceedings, (ii) the standardized values of voter turnout, and (iii) one minus the standardized values of the number of murders per capita, for each province in each year in the 2000–2007 period. Civil trials is defined as one minus the standardized values of the duration (in days) of civil trials; higher values indicate lower duration. Bankruptcy is defined as one minus the standardized values of the duration (in days) of bankruptcy proceedings; higher values indicate lower duration.

intra-regional differences persist. Although it is not surprising that the metropolitan provinces of Milan and Rome have loans-tovalue-added ratios that are much higher than those of their neighboring provinces, there are similar differences with respect to other regions. These variations among provinces and over time are at the core of our identification strategy (Table 1, Panel B reports summary statistics).8 3.1.4. Measure of the socio-institutional environment Choosing proxies for social capital and institutional environment was not an easy task. We decided to build a summary index that would capture three different dimensions of socio-institutional development in Italian provinces: judicial efficiency, social capital, and violence. According to our literature review, these factors directly influence a firm's organizational and production cost structures and thereby have implications for firm efficiency. In this section, we propose a proxy for each of these three dimensions. However, to test the stability of our summary index against different variable definitions and measurement error problems (some of the variables required missing values to be imputed),9 we will rely on alternative proxies in Sections 3.3 and 3.4. Our measure of the quality of judicial enforcement is the average length of bankruptcy proceedings (in days) in each province for each year (data are from ISTAT). This measure indicates the time required to enforce a contract or obtain repayment in the event of default insolvency; this measure shows variability over time and among provinces located in the same area of the 8 To facilitate an easier comparison of the values of financial development and the values of the indicator of the socio-institutional environment (see the following section), we standardized this variable to the interval [0,1], using a min–max standardization approach. 9 For proxies of social capital and crime, we imputed missing values by alternatively using linear interpolation methods or “nearest neighbor” methods, depending on the stability of the variable basis.

L. Moretti / European Journal of Political Economy 35 (2014) 38–51

43

B) Time trend of the financial

indicator (average value 2000–2007)

development indicator

Average Financial Development index .3 .35 .4

A) Map of the financial development

2000

2001

2002

2003 2004 Year

2005

2006

2007

0.46 - 1 0.32 - 0.46 0.18 - 0.32 0 - 0.18

D) Time trend of the socio -

environment indicator (average value 2000–2007)

institutional environment indicator

.66

Average Socio-institutional index .68 .7 .72 .74

C) Map of the socio-institutional

2000

2001

2002

2003 2004 Year

2005

2006

2007

0.79 - 1 0.74 - 0.79 0.69 - 0.74 0 - 0.69

Fig. 1. Map and time trend of the financial development (FD) and socio-institutional environment (SI) indicators. Note: FD is defined as (the standardized values of) the ratio between bank loans to the productive sector and the value added for each province in each year between 2000 and 2007. SI is our main indicator of the quality of socio-institutional environments and is computed as the simple average of (i) one minus the standardized values of the duration of bankruptcy proceedings, (ii) the standardized values of voter turnout, and (iii) one minus the standardized values of the number of murders per capita, for each province in each year during the 2000–2007 period.

country. However, the more significant divide is between the southern and northern parts of the country, with much longer proceedings in the southern provinces. As a measure of social capital, we employed voter turnout, which is a measure of civic-mindedness and captures individuals' willingness to participate in the determination of institutions (see Del Monte and Papagni, 2007; Guiso et al., 2008; Nannicini et al., 2013). We defined our variable as the percentage of eligible voters who cast a ballot in the elections for the European

44

L. Moretti / European Journal of Political Economy 35 (2014) 38–51

Parliament (with data from the Italian Ministry of the Interior).10 The data show large territorial variations in voter turnout, with higher participation in the central and northern provinces and important differences that persist within the regions. Finally, our measure of violence is the number of murders and attempted murders per province per year (data from ISTAT).11 As emphasized by Detotto and Otranto (2010), this measure suffers less from under-reporting problems relative to other crime indicators, which might be an issue in the southern regions of the country (i.e., the areas with higher crime rates). As expected, this measure shows important territorial variations among Italian provinces and tends to be persistent over time. According to Peri (2004), these variations are caused by the territorial presence of criminal organizations (i.e., the Mafia, Camorra, and 'Ndrangheta), particularly in the south. The three variables described above are brought together in a single index of the quality of local socio-institutional environments (SI) using the following procedure: (i) each variable is standardized using a min–max standardization approach (in which the maximum and minimum are the maximum and minimum values of the considered variable between 2000 and 2007, respectively); (ii) rescaling the measures of violence and the inefficiency of the judicial system such that the maximum value indicates the highest-quality environment (as does the measure for social capital already); and (iii) computation of the arithmetic mean of each of these three standardized variables for each province and year. Similar to the variation of the indicator of financial development, the summary index SI shows that provinces with higher social capital, less violence, and more efficient judicial enforcement are located in the northern and central areas of the country, whereas the southern provinces show much lower levels of the SI indicator. However, there are large differences among the provinces within these macro-regions (see Fig. 1C) and over time (Fig. 1D).

3.2. Estimation results We estimated our model specification (Eq. (1)) using four different estimation approaches. The first estimation approach includes industry, province, and time dummies to control for effects that might affect productivity with respect to firms that have similar production processes (i.e., firms operating in the same industry) or that operate within the same province, which is of particular importance because it allows us to reduce the possible bias caused by the correlation between the province-level financial development variable of interest and the error term. Furthermore, it allows us to control for macro shocks that might affect productivity in a given year. This augmented model was then estimated using a pooled ordinary least squares (pooled-OLS) estimator with clustered standard errors at the provincial level, which allowed for heteroskedasticity between the error terms for firms within the same province. The second estimation approach controls for time-invariant, firm-specific characteristics that affect productivity but are not captured by firm-level control variables or the industry, province, and time dummies. In particular, we exploited the time dimension of our data and estimated the model specifications using a within-group estimator (i.e., firm-fixed effects).12 The third and fourth estimation approaches account for the fact that regressors might be correlated with the firm-specific, time-varying, idiosyncratic component of the error term and that this might be a source of endogeneity. For example, a shock at the provincial level might affect both firm productivity and a bank's decision whether to increase the supply of credit. We address this potential endogeneity problem using a 2SLS pooled estimator and a GMM estimator. In the 2SLS estimation, we used the provincial-level characteristics of the banking market in 1936 to instrument the current level of Italian provincial financial development, as proposed by Guiso et al. (2004a). In particular, we used the 1936 values of bank branches per inhabitant, the share of bank branches owned by local banks over total branches, the number of savings banks, and the number of cooperative banks per capita (data are from Guiso et al., 2004a and ISTAT).13 Moreover, following Benfratello et al. (2008), we interacted all these variables with year dummies because these variables are instruments for the values of the indicator of financial development, which varies over time in our analysis.14 To test for the potential endogeneity of the banking market variable (FD), the Durbin–Wu–Hausman test was performed in its regression-based form (see Wooldridge, 2002: 118–122). Finally, we used the first-difference GMM estimator developed by Arellano and Bond (1991) and Arellano and Bover (1995). The variables are first-differenced to eliminate firm-specific, time-invariant effects, and then the first-differences of endogenous variables are instrumented using suitable lags of their levels. In this case, the estimated model is slightly different because we introduced the lagged value of productivity as a regressor; thus, we assumed that firm productivity followed a persistent process.15 Table 2 reports the estimation results using pooled OLS (Columns 1 to 3), panel within-group estimator (Columns 4 to 6), pooled 2SLS (Columns 7 to 9), and the GMM estimator (Columns 10 to 12). We first estimated a model specification including the 10 We decided not to use data from Italian general elections because Italian citizens are required by law to vote. Data are related to the European elections held in 1994 (June 12), 1999 (June 13), 2004 (June 12), and 2009 (June 7). 11 This measure is elaborated by ISTAT on the basis of data reported by the police to the judicial authorities and collected by the Ministry of the Interior. Original data are available for the period prior to 2003. 12 When we use the within-group fixed effect estimator, we employ a restricted sample that is limited to firms that have been present in our panel for at least three years. 13 Guiso et al. (2004a) explained the reasons for using these instruments for the current values of the Italian local banking markets and tested the related excluded restrictions. In brief, the regulations imposed by the “Legge bancaria” of 1936 (which imposed constraints on opening new branches for different types of banks) shaped the Italian banking system until a process of deregulation occurred at the end of the 1980s. National banks were more tightly regulated and, among the local banks, cooperative banks faced tighter constraints. 14 Note that the same set of instruments interacted with the SI index are used as instruments for the values of the interaction term FD ∗ SI. 15 In the first-difference GMM estimation, we used the five lagged values of all regressors as instruments, except for the time dummies, firm's age and SI index which entered as standard instruments. The Sargan test of all model specifications does not indicate the presence of important model misspecifications.

L. Moretti / European Journal of Political Economy 35 (2014) 38–51

45

linear terms of financial development (FD) and the SI indicator (Columns 1, 4, 7, and 10). Then, we augmented the model with the interaction term FD ∗ SI (Columns 2, 5, 8, and 11). Finally, we introduced both the interaction term FD ∗ SI and the squared values of SI (SI2, Columns 3, 6, 9, and 12). This last specification is particularly helpful for understanding the non-linearity of both FD and the SI environment on firm productivity. In fact, if FD has a higher impact on firm productivity for a higher level of SI while the impact of an improvement in SI on firm productivity is higher for a lower level of SI (see for instance, Ahlerup et al., 2009), then introducing only one interaction term (FD ∗ SI) in the model specification might not show the actual impact of these two variables. However, the inclusion of the squared term (SI2) might allow us to capture both of the opposing effects. The pooled OLS estimation in Column 1 shows that FD has an average positive effect on firm productivity, which confirms the general findings described in Section 2, i.e., that a higher depth of the banking market can improve, on average, firm productivity and efficiency. However, when we introduce the interaction terms and allow for non-linear effects (Columns 2 and 3), the estimated coefficient of FD becomes negative, although the interaction term FD ∗ SI is positive and statistically significant, which indicates a higher positive effect of financial development on productivity at higher levels of socio-institutional development. This interaction effect represents a novel result and is confirmed by within-group estimations (Columns 4 to 6). It is also confirmed when we control for potential endogeneity problems by using both 2SLS-pooled and difference-GMM estimators (Columns 7 to 9 and 10 to 12, respectively). To better appreciate the interaction effect of financial development and socio-institutional development, we took the partial derivative of the model specification with respect to FD. On the basis of the estimated coefficients in Table 2, Column 6, the following exercise of comparative statics illustrates the economic meaning of the estimated coefficients. A standard deviation increase in the value of the financial development indicator is associated with a 2.8% increase in labor productivity for an average level of socio-institutional development. An increase of a standard deviation in the financial development indicator is associated with a 1.5% increase in productivity when the value of the socio-institutional indicator is in the 25th percentile and with an increase of approximately 4.7% when the value of the socio-institutional indicator is in the 75th percentile of its distribution.16 Based on the estimated coefficients in Table 2, column 6, Fig. 2 provides a graphical representation of the marginal effects of financial development for different levels of the socio-institutional indicator. This graph shows that for values of the socio-institutional indicator above 0.6 the marginal effects are positive and continue to increase as values of the socio-institutional indicator increase. For values of the socio-institutional indicator below 0.6, the effect of financial development on productivity is actually negative. However, only for 4% of our estimated sample the values of socio-institutional indicator fall below 0.6. Overall, these results show that financial development has appreciably different effects on productivity across different levels of socio-institutional quality and that a firm's efficiency is effectively influenced by a larger availability of external finance, particularly in the areas in which social and institutional factors contribute to create better productive opportunities. However, availability of external finance does not enhance productivity when productive opportunities are limited by the low-quality of the institutions. It might also be interesting to understand the effects of SI on productivity. According to our evidence, the effect of SI is non-linear with respect to both SI and FD. In fact, as shown in Table 2 (Columns 3, 6, 9, and 12), the estimated coefficient of the single term SI is positive and typically statistically significant, the coefficient of its squared term (SI2) is negative, and the interaction between SI and FD is positive. Given these non-linearities, it is difficult to interpret the effect of SI. However, the finding that must be emphasized for purposes of this study is that improvements in SI have a greater impact than improvements in FD in the southern provinces (i.e., at lower levels of SI and FD), whereas improvements in FD are more effective in the northern provinces than in the southern provinces (i.e., at higher levels of FD and SI).17

3.3. Robustness checks The objective of this section is to further evaluate the robustness of the baseline results described above. It is natural to ask whether the variables have identical weight in the construction of our index of SI development or whether it is possible to use a different aggregation method. Recall that the SI index used thus far was built as the simple average of (i) the standardized values of voter turnout (as a measure of social capital), (ii) one minus the standardized values of the duration of bankruptcy proceedings (as a measure of the quality of the judicial enforcement), and (iii) one minus the standardized values of the number of murders per capita (as a measure of the presence of criminal activities). Now, we bring these three variables together by extracting the first principal component and constructing an alternative indicator of socio-institutional development

16

The values of the effects are obtained from the following equations: (i) (β1 + β3meanSI) ∗ sdFD = (−0.830 + 1.353 ∗ 0.722) ∗ 0.189 = 0.028; (ii) (β1 + β325°p. SI) ∗ sdFD = (−0.830 + 1.353 ∗ 0.673) ∗ 0.189 = 0.015;

(iii) (β1 + β375°p. SI) ∗ sdFD = (−0.830 + 1.353 ∗ 0.797) ∗ 0.189 = 0.047. Note that coefficients are from Table 2, Column 6, and descriptive statistics are reported in Table 1, Panel B. 17 Estimated coefficients of firm level control variables typically assume the expected signs. Estimation results are not reported in the tables because they are beyond the scope of this paper. However, they are available in the Online Appendix A.2.

46

L. Moretti / European Journal of Political Economy 35 (2014) 38–51

Table 2 Financial development, socio-institutional environment, and firm productivity: Estimation results. Dependent

FD SI

ln(VA/L): labor productivity Pooled

Pooled

Pooled

Panel

Panel

Panel

Pooled

Pooled

Pooled

GMM

GMM

GMM

OLS

OLS

OLS

FE

FE

FE

2SLS

2SLS

2SLS

FD

FD

FD

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

(11)

(12)

0.269* (0.148) 0.246** (0.111)

−0.810** (0.380) −0.254 (0.215) 1.393*** (0.469)

0.225 (0.149) 0.163 (0.103)

−0.660* (0.383) −0.250 (0.205) 1.138** (0.452)

−1.186*** (0.397) −0.533*** (0.182) 2.060*** (0.424)

−0.641** (0.326) −0.323** (0.141) 1.477*** (0.465)

X

X

X

X

X 417,991 0.173

X 417,991 0.174

X X X X 532,315 0.233 0.000

−0.675** (0.336) 0.536** (0.211) 1.429*** (0.457) −0.618*** (0.208) X

X 417,991 0.173

X X X X 532,315 0.231 0.067

−1.164*** (0.358) 1.358*** (0.498) 1.998*** (0.384) −1.424*** (0.356) X X X X 532,315 0.233 0.000

0.349** (0.152) 0.106*** (0.038)

X X X X 532,315 0.232

−0.830** (0.391) 1.367*** (0.375) 1.353*** (0.472) −1.271*** (0.290) X

0.665*** (0.165) 0.134 (0.123)

X X X X 532,315 0.232

−0.973** (0.405) 1.490*** (0.461) 1.597*** (0.508) −1.390*** (0.355) X X X X 532,315 0.232

X 148,098

X 148,098

X 148,098

0.000 0.881 0.412

0.000 0.107 0.208

0.000 0.128 0.240

FD ∗ SI SI2 Firm-level var. Industry FE Province FE Year dummy Observations R-squared Endogeneity AR(1) AR(2) Sargan

Note: Robust standard errors and clustered at the province-level are in parentheses. Significance: ***p b 0.01, **p b 0.05, *p b 0.1. The dependent variable is the log of the real value added per worker (ln(VA/L)). FD is the (standardized) ratio between bank loans to the productive sector and the value added. SI is our main indicator of the quality of socio-institutional environments and is computed as the simple average of (i) one minus the standardized values of the duration of bankruptcy proceedings, (ii) the standardized values of voter turnout, and (iii) one minus the standardized values of the number of murders per capita. When noted, estimated equations also include the following: firm-level control variables; province fixed effects; industry fixed effects; and year dummies. When we use the panel within-group estimator (columns 4 to 6), we employ a restricted sample that is limited to firms that have been present for at least three years. In columns 7 to 9, the 1936 values of bank branches per inhabitant, the share of bank branches owned by local banks over total branches, the number of saving banks, and the number of cooperative banks per capita, all interacted with time dummies, are used as instruments for the values of FD. The same set of instruments interacted with SI are used as instruments for FD ∗ SI. In columns 10 to 12, the set of instruments includes lagged values of FD, lagged values of FD ∗ SI (only in columns 11 and 12), lagged values of fixed capital per employee, lagged values of firm size, and lagged values of firm leverage and its squared values. Values of firm age, year dummies, values of the SI (and, in column 12, SI2) are also used as instruments. Endogeneity is the regression-based form of the Durbin– Wu–Hausman test. If the null hypothesis is not rejected, OLS estimations are preferred; p-values are reported. AR(1) and AR(2) test the presence of first- and second-order serial correlation in the transformed error; p-values are reported. Sargan is a Sargan test of the validity of the overidentifying orthogonality conditions; p-values are reported.

(SI_pc). To choose the number of components, we applied the rule of the eigenvalue greater than one and it resulted that we must use the first principal component in the analysis. Table 3, Column 1, reports the estimation results obtained using SI_pc in our model with firm-level fixed effects. The sign and statistical significance of the estimated coefficients are similar to those found in the previous analysis. Another concern has to do with the robustness of the SI index to alternative proxies. As an alternative measure of social capital, we followed Putnam (1993) and employed the density of voluntary associations in the province, which is defined as the number of voluntary associations per capita (data from ISTAT). This variable is another measure of civil capital and gauges altruistic behavior at the province level (Guiso et al., 2008). As shown in previous studies on Italy, this variable reveals important differences between the southern provinces and the central and northern provinces of the country (e.g., Del Monte and Papagni, 2007; Guiso et al., 2008; Nannicini et al., 2013). As an alternative measure of judicial enforcement, we used the average duration (in days) of first-instance civil trials (weighted over the number of pending cases) in each province for each year (data from ISTAT). It has been previously documented that this measure is a good proxy for differences in the judicial enforcement of contracts among Italian provinces (e.g., Jappelli et al., 2005; Coviello et al., 2013). The average duration of civil trials between 2000 and 2007 was 968 days, with a minimum of 205 days and a maximum of 2221 days.18 Finally, as an alternative proxy for organized crime, we used the number of extortions per capita. Extortion is a crime typically associated with the presence of organized crime, as emphasized in Daniele and Mariani (2011), and it is much more concentrated in southern areas of Italy.19 We thus had two alternative proxies for each of the three dimensions of our synthetic indicator: voter turnout and the number of voluntary associations as proxies for social capital; the duration of bankruptcy proceedings and the duration of civil trials as proxies for judicial efficiency; and the number of murders and the number of extortions per capita as proxies for criminal

18

This variable is missing for the province of Teramo in 2002. We refrained from imputing the missing value. This measure is elaborated by ISTAT on the basis of data reported by the police to the judicial authorities and collected by the Ministry of the Interior. Original data are available for the period after 2003. 19

0

10

.5

5

-.5 -1

0

Percentage of observations

47

0

.2

.4

.6

.8

1

Marginal effects of Financial development (FD)

15

L. Moretti / European Journal of Political Economy 35 (2014) 38–51

Index of socio-institutional environment (SI) Marginal effects of FD

Percentage of observations

Fig. 2. Marginal effects of financial development and distribution of the indicator of socio-institutional environment. Note: marginal effects are based on the estimated coefficients in Table 2 column 6. The dependent variable is the log of the real value added per worker (ln(VA/L)). FD is defined as (the standardized values of) the ratio between bank loans to the productive sector and the value added for each province in each year between 2000 and 2007. SI is our main indicator of the quality of socio-institutional environments and is computed as the simple average of (i) one minus the standardized values of the duration of bankruptcy proceedings, (ii) the standardized values of voter turnout, and (iii) one minus the standardized values of the number of murders per capita, for each province in each year during the 2000–2007 period.

activities. Combining a proxy for each dimension and extracting the first principal component, we constructed seven synthetic indicators of socio-institutional environment as an alternative to our SI_pc index.20 In Table 3 (Columns 2 to 8), we report the estimation results obtained using these seven alternative indicators to our main indicator of socio-institutional environment, SI_pc. In the bottom line of the table, for each column, we report the three variables used for the construction of each alternative indicator. The signs and significance of the coefficients remain stable with respect to our main estimation results, which indicates that, although measuring the socio-institutional environment with a synthetic index is not an easy task and involves some degree of discretion, our results are stable both to different proxies and to different aggregation methods.21

3.4. Extensions In terms of policy implications, our results suggest that improvements in the factors that determine the socio-institutional environment can increase the effects of financial development on productivity. In our framework, the use of a synthetic indicator indicates that improvement can be achieved by reducing violence, increasing involvement of individuals in civil society, reducing the duration of civil trials and bankruptcy proceedings, or some combination of the foregoing approaches. Furthermore, improving one of these dimensions can positively affect the other dimensions and reinforce positive effects. To provide further evidence of policy implications, we focus in this section on the effects across firms of different sizes and on the role of judicial efficiency. Earlier empirical literature emphasizes that small firms have greater financial constraints because they have less availability of collateral and greater opacity (see, among others, Beck et al., 2005; Pagés, 2010). Improving access to credit should therefore benefit small firms more than large firms. Using our model, we can analyze the effect of higher credit availability on the productivity of firms of different sizes at different levels of socio-institutional development. We interact FD, SI and FD ∗ SI with a measure of firm size (proxied by firm's total assets). It should be clear that this model specification includes many interaction terms that might affect the accuracy of the estimates. For this reason, we offer the most transparent evidence using OLS, robust regressions to control for the influence of outliers, and firm-level fixed effects estimations. Estimation results (Table 4, Columns 1 to 3) show that in socio-institutionally developed provinces, a larger quantity of credit benefits larger firms less (positive sign of the coefficient of FD ∗ SI and negative sign of the coefficient of FD ∗ SI ∗ Size), whereas in

20 It is notable that the original variables are closely correlated with the corresponding principal component, and this suggests that the summary measures capture a common socio-institutional pattern. Moreover, the summary indexes are strongly correlated with one another, suggesting that the choice of different variables and different aggregation methods does not affect the nature of the indicator. See also Online Appendix A.2 for correlation among the individual components of the indicators. 21 For reasons of space, we reported only (i) the within-group estimation and (ii) the results obtained using the principal component as a method of variables' aggregation. The results, available in the Online Appendix A.2, are also robust to the use of the estimators OLS, 2SLS, GMM, and robust regressions (which downweigh observations with large residuals using the Huber weight function) in addition to the indexes aggregated by means of simple average.

48

L. Moretti / European Journal of Political Economy 35 (2014) 38–51

Table 3 Robustness checks: alternative indicators of socio-institutional environment. Dependent

FD SI_pc FD ∗ SI_pc SI_pc2 Firm-level var. Year dummy Observations R-squared SI index includes

ln(VA/L): labor productivity Panel

Panel

Panel

Panel

Panel

Panel

Panel

Panel

FE

FE

FE

FE

FE

FE

FE

FE

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

−0.660 (0.412) 0.643** (0.265) 1.098** (0.426) −0.704*** (0.162) X X 417,991 0.174 Bankruptcy; voter turn.; murders.

−0.668 (0.411) 1.085*** (0.349) 1.874** (0.720) −2.104*** (0.675) X X 417,991 0.173 Bankruptcy; volunt. ass.; murders.

−0.491* (0.288) 0.683*** (0.147) 0.950*** (0.340) −0.730*** (0.130) X X 417,991 0.175 Bankruptcy; voter turn.; extortions.

−0.611* (0.314) 1.343*** (0.349) 1.876*** (0.605) −2.445*** (0.657) X X 417,991 0.174 Bankruptcy; volunt. ass.; extortions.

−1.100*** (0.330) 0.216 (0.214) 1.570*** (0.339) −0.607*** (0.162) X X 417,603 0.175 Civil trials; voter turn.; murders.

−1.203*** (0.403) 0.766** (0.353) 2.484*** (0.647) −2.087*** (0.557) X X 417,603 0.174 Civil trials; volunt. ass.; murders.

−1.059*** (0.286) 0.580*** (0.153) 1.575*** (0.323) −0.885*** (0.129) X X 417,603 0.175 Civil trials; voter turn.; extortions.

−1.213*** (0.343) 1.099*** (0.376) 2.750*** (0.634) −2.757*** (0.661) X X 417,603 0.175 Civil trials; volunt. ass.; extortions.

Note: Robust standard errors clustered at the province level are in parentheses, with the following significance levels: ***p b 0.01, **p b 0.05, *p b 0.1. The dependent variable is the log of real value added per worker (ln(VA/L)). FD is the (standardized) ratio between bank loans to the productive sector and value added. SI_pc is the first component of the indicator of the quality of the socio-institutional environment and is obtained using the variables indicated in the bottom line of the table. Firm-level control variables and year dummies are included in the estimated equations. When we use the panel within-group estimator (columns 1 to 8), we employ a restricted sample that is limited to firms that have been present for at least three years. Note that the value of the duration of civil trials is missing for the province of Teramo for 2002.

less socio-institutionally developed environments, a larger quantity of credit benefits smaller firms less (negative sign for the coefficient of FD and positive sign for the coefficient of FD ∗ Size). This result seems to suggest that in institutionally developed areas, productive business opportunities can be exploited by reducing credit barriers, which is particularly beneficial for smaller firms, which typically have more difficulty accessing external sources of credit and suffer more from institutional constraints (Beck et al., 2005). Conversely, in areas with low institutional development, higher credit availability does not help smaller firms increase their productivity because improving access to credit might not be able to overcome institutional constraints on firm efficiency. The role of judicial enforcement also has relevant policy implications. First, reforms aiming to improve the efficiency of the judicial enforcement, in particular by reducing the duration of civil trials, are high on the reform agenda in Italy. Second, it seems reasonable to expect that, compared with improvements in social capital or the fight against criminal organizations, improvements in the duration of civil trials might be accomplished over a relatively short time horizon. Third, improvements in the efficiency of the judicial enforcement can complement social capital in securing economic transactions. Finally, in terms of the robustness of our analysis, evaluating the effects of judicial efficiency is a good check of robustness; in fact—in contrast to our measures of social capital and crime—the measures of the duration of civil trials and bankruptcy proceedings do not contain any imputed observations. The estimation results in Table 4 (Columns 4 to 8) show the effects of local financial development on firm productivity, which is conditional on the duration of civil trials. Using the estimated coefficients in Table 4, Column 6, we can estimate that an increase of one standard deviation in the financial development indicator is associated with an increase of 0.9% in the average productivity of the province for average values of the duration of civil trials. However, the same increase in the financial development is associated with an increase in productivity of 3% in provinces with shorter average trial durations.22 With respect to the interaction between the indicator of financial development and the indicator of the duration of bankruptcy proceedings (Table 4, columns 9 to 13), estimation results show similar effects even when the statistical significance is weaker. This result could be caused by the presence of outlying values because the robust regression shows statistically significant results.

22

The values of the effects are obtained from the following equations: (i) (β1 + β3meanCivil Trials) ∗ sdFD = (−0.538 + 0.940 ∗ 0.621) ∗ 0.189 = 0.009. (ii) (β1 + β375°p. Civil Trials) ∗ sdFD = (−0.538 + 0.940 ∗ 0.743) ∗ 0.189 = 0.030.

Where Civil Trials is defined as one minus the standardized values of the duration (in days) of civil trials; higher values indicate lower duration.

Table 4 Extensions: differential effects across firm sizes and the role of judicial enforcement. Dependent

ln(VA/L): labor productivity Pooled OLS

FD FD ∗ Size SI SI ∗ Size

FD ∗ SI ∗ Size SI2

Panel FE

Pooled OLS

Robust reg.

Panel FE

Pooled 2SLS

GMM FD

Robust reg.

Panel FE

Pooled 2SLS

GMM FD

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

(11)

(12)

(13)

−2.494*** (0.841) 0.203** (0.093) 1.985*** (0.501) −0.071* (0.040) 2.603** (1.091) −0.147 (0.125) −1.311*** (0.341)

−2.499*** (0.182) 0.208*** (0.022) 2.109*** (0.132) −0.067*** (0.011) 2.504*** (0.239) −0.140*** (0.029) −1.408*** (0.072)

−3.763*** (1.188) 0.361*** (0.119) 0.714 (0.594) 0.061 (0.062) 3.080** (1.497) −0.227 (0.162) −1.127*** (0.269)

−0.649** (0.274)

−0.718*** (0.036)

−0.538** (0.263)

−1.825*** (0.558)

−1.326 (1.064)

0.137 (0.195)

0.185*** (0.028)

0.098 (0.178)

−0.129 (0.299)

−0.123 (1.093)

−0.376** (0.172) 1.130*** (0.311) −0.121 (0.182)

−0.417*** (0.035) 1.211*** (0.043) −0.125*** (0.036)

−0.323** (0.156) 0.940*** (0.318) −0.114 (0.168)

−0.440** (0.182) 2.317*** (0.542) −0.500*** (0.181)

−0.185 (0.143) 2.090** (1.048) −0.659* (0.354) 0.219 (0.176) 0.182 (0.250) −0.155 (0.159) X X X X 532,315 0.232

0.189*** (0.035) 0.111*** (0.036) −0.104*** (0.031) X X X X 532,315 0.264

0.176 (0.143) 0.175 (0.229) −0.157 (0.139) X

0.070 (0.157) 0.828*** (0.277) −0.251 (0.160) X X X X 532,315 0.232 0.000

0.023 (0.052) 0.385 (0.729) −0.105 (0.235) X

Civil trials FD ∗ Civil trials Civil trials2 Bankruptcy FD ∗ Bankruptcy Bankruptcy2 Firm-level var. Industry FE Province FE Year dummy Observations R-squared Endogeneity AR(1) AR(2) Sargan

Pooled OLS

X X X X 532,315 0.234

X X X X 532,315 0.268

X

X 417,991 0.178

X X X X 531,847 0.233

X X X X 531,847 0.265

X

X 417,603 0.174

X X X X 531,847 0.233 0.000

X

X 147,576

0.008 0.612 0.183

X 417,991 0.173

X 148,098

L. Moretti / European Journal of Political Economy 35 (2014) 38–51

FD ∗ SI

Robust reg.

0.006 0.980 0.168

49

Note: Robust standard errors and clustered at the province level (except in Robust reg.) are in parentheses. Significance: ***p b 0.01, **p b 0.05, *p b 0.1. The dependent variable is the log of real value added per worker (ln(VA/L)). FD is the (standardized) ratio between bank loans to the productive sector and value added. SI is our main indicator of the quality of socio-institutional environment and is computed as the simple average of (i) one minus the standardized values of the duration of bankruptcy proceedings, (ii) the standardized values of the voter turnout, and (iii) one minus the standardized values of the number of murders per capita. Size is the log of firm's real total assets (and it is included also in the firm-level control variables). Civil trials is defined as one minus the standardized values of the duration of civil trials; higher values indicate lower duration. Bankruptcy is defined as one minus the standardized values of the duration of bankruptcy proceedings; higher values indicate lower duration. When noted, estimated equations also include the following: firm-level control variables, province fixed effects, industry fixed effects, and year dummies. Robust regression is an iteratively reweighted least squares procedures (IRLS), which down-weighs observations with large residuals using the Huber weight function. When we use the panel within-group estimator (Columns 3, 6, 11), we employ a restricted sample that is limited to firms that have been present for at least three years. In Columns 7 and 12, the 1936 values of bank branches per inhabitant, the share of bank branches owned by local banks over total branches, the number of savings banks, and the number of cooperative banks per capita, all interacted with time dummies, are used as instruments for FD. The same set of instruments interacted with SI are used as instruments for FD ∗ SI. In Columns 8 and 13, the set of instruments includes lagged values of FD, lagged values of FD ∗ SI, lagged values of firm's fixed capital per employee, lagged values of firm's size, and lagged values of firm's leverage and its squared values. Values of firm's age, year dummies, values of SI, and SI2 are also used as instruments. Endogeneity is the regression-based form of the Durbin-Wu-Hausman test. If the null hypothesis is not rejected, OLS estimations are preferred; p-values are reported. AR(1) and AR(2) test the presence of first- and second-order serial correlation in the transformed error; p-values are reported. Sargan is a Sargan test of the validity of the overidentifying orthogonality conditions; p-values are reported. Note that the value of the duration of civil trials is missing for the province of Teramo for 2002.

50

L. Moretti / European Journal of Political Economy 35 (2014) 38–51

4. Conclusions In this study, we tested whether the depth of local banking markets is among the determinants of the differences in productivity among the Italian provinces. In particular, we verified whether the effects of financial development on firm productivity are conditional on the quality of socio-institutional environments. Exploiting the large territorial variability regarding financial and socio-institutional development and taking into account that the socio-institutional environment is also a determinant of local financial development, we found that larger local banking markets are associated with higher labor productivity when the socio-institutional environment is more developed (i.e., in the northern and central areas of Italy). According to our results, we can interpret improvements in the socio-institutional environment as enhancing a virtuous cycle; they can increase the level of local financial development and also create better productive opportunities that can be exploited through larger credit availability. Improvements in the socio-institutional environment could be the target of public policies. However, incentives to create a more cooperative spirit in society typically do not bear fruit quickly, and they frequently require the support of educational policies. Other dimensions of the socio-institutional environment can also be targeted, such as the efficiency of the judicial system. For example, our results show that a reduction in the duration of civil trials allows financial development to have larger positive effects on productivity. In Italy, a reform of the judicial system to reduce the duration of civil trials is one of the most-debated reforms of the last decade. Better enforcement of contracts and improving the security of transactions will be important, particularly in times of crisis, when uncertainty increases and trust among people does not necessarily provide a substitute for judicial enforcement of contracts. This study also suggests that further research on the interaction between financial development, socio-institutional environments, and the real economy could focus on the role played by interest groups in shaping financial development in countries with different levels of social capital across territories (and therefore different levels of political accountability). Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.ejpoleco.2014.03.006. References Ahlerup, P., Olsson, O., Yanagizawa, D., 2009. Social capital vs institutions in the growth process. Eur. J. Polit. Econ. 25 (1), 1–14. Akçomak, I., Ter Weel, B., 2009. Social capital, innovation and growth: evidence from Europe. Eur. Econ. Rev. 53 (5), 544–567. Albanese, G., Marinelli, G., 2013. Organized crime and productivity: evidence from firm-level data. Riv. Ital. Economisti 3, 367–394. Arellano, M., Bond, S., 1991. Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. Rev. Econ. Stud. 58 (2), 277–297. Arellano, M., Bover, O., 1995. Another look at the instrumental-variable estimation of error-components model. J. Econ. 68 (1), 29–52. Becerra, O., Cavallo, E., Scartascini, C., 2012. The politics of financial development: the role of interest groups and government capabilities. J. Bank. Financ. 36 (3), 626–643. Beck, T., 2008. The econometrics of finance and growth. Policy Research Working Paper Series 4608. The World Bank. Beck, T., Demirguc-Kunt, A., Maksimovic, V., 2005. Financial and legal constraints to growth: does firm size matter? J. Financ. 60 (1), 137–157. Benfratello, L., Sembenelli, A., Schiantarelli, F., 2008. Banks and innovation: microeconometric evidence on Italian firms. J. Financ. Econ. 90 (2), 197–217. Beugelsdijk, S., Van Schaik, T., 2005. Social capital and growth in European regions: an empirical Test. Eur. J. Polit. Econ. 21 (2), 301–324. Beugelsdijk, S., De Groot, H., Van Schaik, A., 2004. Trust and economic growth: A robustness analysis. Oxf. Econ. Pap. 56 (1), 118–134. Black, S.E., Strahan, P.E., 2002. Entrepreneurship and bank credit availability. J. Financ. 57 (6), 2807–2833. Bloom, N., Sadun, R., Van Reenen, J., 2012. The organization of firms across countries. Q. J. Econ. 127 (4), 1663–1705. Bonaccorsi di Patti, E., 2009. Weak institutions and credit availability: the impact of crime on bank loans. Bank of Italy Occasional Papers no. 52. Bonaccorsi di Patti, E., Dell'Ariccia, G., 2004. Bank competition and firm creation. J. Money, Credit, Bank. 36 (2), 225–251. Campos, N.F., Coricelli, F., 2012. Financial liberalization and reversals: political and economic determinants. Econ. Policy 27 (71), 483–513. Cetorelli, N., Strahan, P.E., 2006. Finance as a barrier to entry: bank competition and industry structure in local U.S. markets. J. Financ. 61 (1), 437–461. Chemin, M., 2012. Does court speed shape economic activity? Evidence from a court reform in India. J. Law Econ. Org. 28 (3), 460–485. Claessens, S., Klapper, L.F., 2005. Bankruptcy around the world: explanations of its relative use. Am. Law Econ. Rev. 7 (1), 253–283. Coleman, J., 1988. Social capital in the creation of human capital. Am. J. Sociol. 94, S95–S120. Coricelli, F., Masten, A.B., Masten, I., 2008. Non-linear growth effects of financial development: does financial integration matter? J. Int. Money Financ. 27 (2), 295–313. Coviello, D., Moretti, L., Spagnolo, G., Valbonesi, P., 2013. Courts' efficiency and procurement performance. Marco Fanno Working Paper no. 164. University of Padova. Daniele, V., Mariani, U., 2011. Organized crime, the quality of local institutions and FDI in Italy: a panel data analysis. Eur. J. Polit. Econ. 27, 132–142. Del Monte, A., Papagni, E., 2007. The determinants of corruption in Italy: regional panel data analysis. Eur. J. Polit. Econ. 23 (2), 379–396. Detotto, C., Otranto, E., 2010. Does crime affect economic growth? Kyklos 63, 330–345. Djankov, S., La Porta, R., Lopez-De-Silanes, F., Shleifer, A., 2003. Courts. Q. J. Econ. 118 (2), 453–517. Fernandez de Guevara, J., Maudos, J., 2009. Regional financial development and bank competition: effects on firms' growth. Reg. Stud. 43 (2), 211–228. Fountain, J.E., 1997. Social capital a key enabler of innovation and technology. In: Branscomb, L.M., Keller, J. (Eds.), Investing in Innovation: Toward a Consensus Strategy for Federal Technology Policy. MIT Press, Cambridge MA. Fukuyama, F., 1995. Trust: The Social Virtues and the Creation of Prosperity. Free Press. Granovetter, M., 2005. The impact of social structure on economic outcome. J. Econ. Perspect. 19 (1), 33–50. Guariglia, A., Poncet, S., 2008. Could financial distortions be no impediment to economic growth after all? Evidence from China. J. Comp. Econ. 36 (4), 633–657. Guiso, L., Sapienza, P., Zingales, L., 2004a. Does local financial development matter? Q. J. Econ. 119 (3), 929–969. Guiso, L., Sapienza, P., Zingales, L., 2004b. The role of social capital in financial development. Am. Econ. Rev. 94 (3), 526–556. Guiso, L., Sapienza, P., Zingales, L., 2008. Long term persistence. NBER Working Papers no. 14278. Haber, S., North, D., Weingast, B. (Eds.), 2008. Political Institutions and Financial Development. Stanford University Press, CA. Hasan, I., Koetter, M., Wedow, M., 2009. Regional growth and finance in Europe: is there a quality effect of bank efficiency? J. Bank. Financ. 33 (8), 1446–1453. Hausmann R., Rodrik D., Velasco A., (2005). Growth Diagnostics. Mimeo, Inter-American Development Bank. Jappelli, T., Pagano, M., Bianco, M., 2005. Courts and banks: effects of judicial enforcement on credit markets. J. Money, Credit, Bank. 37 (2), 223–244.

L. Moretti / European Journal of Political Economy 35 (2014) 38–51

51

Johnson, S., McMillan, J., Woodruff, C., 2002. Property rights and finance. Am. Econ. Rev. 92 (5), 1335–1356. Kendall, J., 2012. Local financial development and growth. J. Bank. Financ. 36 (5), 1548–1562. Knack, S., Keefer, P., 1997. Does social capital have an economic payoff? A cross-country investigation. Q. J. Econ. 112 (4), 1251–1288. La Porta, R., Lopez-de-Silanes, F., Shleifer, A., 1997a. Trust in large organizations. Am. Econ. Rev. 87 (2), 333–338. La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R.W., 1997b. Legal determinants of external finance. J. Financ. 52 (3), 1131–1150. La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R.W., 1998. Law and finance. J. Polit. Econ. 106 (6), 1113–1155. Laeven, L., Woodruff, C., 2007. The quality of the legal system, firm ownership, and firm size. Rev. Econ. Stat. 89 (4), 601–614. Levine, R., 2005. Finance and growth: theory and evidence. In: Aghion, P., Durlauf, S. (Eds.), Handbook of Economic Growth. Elsevier, pp. 865–934. Mauro, L., Carmeci, G., 2007. A poverty trap of crime and unemployment. Rev. Dev. Econ. 110, 681–712. Nannicini, T., Stella, A., Tabellini, G., Troiano, U., 2013. Social capital and political accountability. Am. Econ. J. 5, 222–250. Pagés, C. (Ed.), 2010. The Age of Productivity. IDB-Palgrave McMillan, NY. Peri, G., 2004. Socio cultural variables and economic success: evidence from Italian provinces 1951–1991. B.E. J. Macroecon. 4 Article 12. Pinotti, P., 2012. The economic costs of organized crime: evidence from southern Italy. Bank of Italy working paper no. 868. Ponticelli, J., 2013. Court enforcement and firm productivity: evidence from a bankruptcy reform in Brazil. Chicago Booth Research Paper No. 14–08, Fama-Miller Working Paper. Putnam, R., 1993. Making Democracy Work: Civil Traditions in Modern Italy. Princeton University Press. Rajan, R.G., Zingales, L., 2003. The great reversals: the politics of financial development in the twentieth century. J. Financ. Econ. 69 (1), 5–50. Rioja, F., Valev, N., 2004a. Does one size fit all? A re-examination of the finance and growth relationship. J. Dev. Econ. 74, 429–447. Rioja, F., Valev, N., 2004b. Finance and the sources of growth at various stages of economic development. Econ. Inq. 42, 127–140. Vaona, A., 2008. Regional evidence on financial development, finance term structure and growth. Empir. Econ. 34 (1), 185–201. Wooldridge, J.M., 2002. Econometric analysis of cross section and panel data. MIT Press.

Further Reading Klapper, L., Laeven, L., Rajan, R., 2006. Entry regulation as a barrier to entrepreneurship. J. Financ. Econ. 82 (3), 591–629.