Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Social Capital, Investment and Economic Growth: Evidence for Spanish Provinces Emili Tortosa-Ausina† Jesús Peiró-Palomino∗ Abstract This article analyzes the role of social capital on economic growth in the Spanish provinces during the 1983–2005 period. Whereas most studies have been using survey data in order to measure social capital, we use a measure whose construction is based on similar criteria as other measures of capital stock. In addition, compared with more standard measures of social capital and trust, our measure is available with a high level of disaggregation and with annual frequency for a long time span. Following a panel data approach, we determine upon the case of the Spanish provinces that social capital impacts positively on GDP per capita, highlighting that social features are important for explaining the differences in wealth that we can observe for the different Spanish provinces. We also try to determine the transmission mechanisms from social capital to growth, finding a highly positive relation between social capital and private physical investment. Keywords: social capital, GDP per capita, convergence, physical capital, investment. Communications to: Jesús Peiró-Palomino, Departament d’Economia, Universitat Jaume I, Campus del Riu Sec, 12071 Castelló de la Plana, Spain. Tel.: +34 964388615, fax: +34 964728591, e-mail: peirojuji.es ∗ Universitat † Universitat Jaume I. Jaume I and Ivie. 1 1. Introduction Traditionally, economic growth has been one of the topics which have attracted more interest in the economic literature. The first steps in the matter are attributed to Solow (1957), who proposed a model including physical capital investment, labor and technological change. Subsequently, the economic growth literature has considered a large set of possible explanatory variables of a different nature, such as human capital or regional, political, religious and social variables. Among the most important contributions subsequent to Solow’s work, we find those which found evidence in favor of human capital as an additional explanatory factor (Barro, 1991; Mankiw et al., 1992; Nonneman and Vanhoudt, 1996). Despite the considerable efforts for determining which are really the robust factors behind the economic growth (Levine and Renelt, 1992; Sala-i-Martin, 1997; Crespo-Cuaresma et al., 2011), a consensus has not been reached yet. In the last few years, a new variable has been considered by several studies on this issue: social capital. This concept was introduced by Coleman (1988) more than two decades ago. In the context of empirical economic growth and convergence, the analyst would be confronted with evaluating if social features such as trust, associationism, social participation or publicspiritedness influence on the economic performance of one region, or how important might the social component be. However, despite the importance of this type of issue is growing, scholars face up two important problems. The first one is what some authors refer to as the vagueness of the concept (Torsvik, 2000). Social capital is a topic which stirs the interest not only of economists but also sociologists and policymakers and, accordingly, the concept is characterized by an interdisciplinary nature. Although this might be a priori good, in practical terms it impedes a consensus to be reached about the impact of social capital—both where and how it truly impacts. The second problem scholars face when approaching the concept, and perhaps the most relevant from a point of view of measuring how it affects growth and convergence, is that data on social capital are relatively scarce, and that the data provided by different institutions usually carry different meanings—and, therefore, the implications for growth and convergence may also vary from one measure to the other. As we will see throughout the study, this is ultimately the main reason for considering a social capital measure which has been constructed with similar underpinnings to those used for building other databases such as physical or human capital. Considering the particular discipline of economics, over the last few years some studies have been analyzing how social capital affects different dimensions of economic activity in different countries and regions. In this line of research, Beugelsdijk and Van Schaik (2005) have reported evidence for Europe using a sample of 54 European regions, considering the information of the European Values Survey (EVS) indicators for both trust and associational 2 activity. They suggest that trust is not related to growth but rather point out a strong relationship between active groups and growth.1 Previously, Knack and Keefer (1997) had found for a sample of 29 countries and using data from World Values Survey (WVS) a strong relationship between trust and civic cooperation and growth but, contrary to Beugelsdijk and Van Schaik (2005), the relation between associational activity and growth was not significant. Some other studies have been focused on developing countries such as Narayan and Pritchett (1999) who, using a sample of 1,376 household villages in rural Tanzania, along with data from the Social Capital and Poverty Survey (SCPS),2 determined that higher levels of social capital provide higher incomes.3 Portela and Neira (2002) compared social capital in Spain with European countries finding lower endowments in Spain, employing trust and associational indicators provided by WVS. They also found significant effects from social capital to growth in a sample of European countries. The studies cited in the previous paragraph are some examples of contributions reporting interesting conclusions which might be useful to understand—for the particular case of Spain on which we focus—why some provinces are systematically richer than others based on terms of GDP per capita. Pérez (2007) concluded that all provinces without any exception have experienced an important economic growth in the period 1955–2005, but disparities among them are still considerable today. It constitutes virtually a consensus for scholars who have studied the Spanish case, the fact that convergence process in GDP per capita among the different Spanish regions4 slowed down in the 1980s, whereas labor productivity followed a convergent path. Dolado et al. (1994), Raymond Bara and García Greciano (1994), Rabadán et al. (1996), Goerlich and Mas (2001), Goerlich et al. (2002), or Peña Sánchez (2008), constitute relevant contributions. Pérez (2007) determined that the persistent encountered disparities are consequence of the distinct capacity of the provinces for attracting activity, which is finally responsible for the generation of income and employment. In that sense, studies such as Becattini (1979) or Trigilia (2005) concluded that the existence of social capital in one territory is one of the factors for the activity attraction and local development. Because the presence of social capital in one territory can also impel the generation of other kinds of capital, such as human or physical capital (Dearmon and Grier, 2011), it is feasible that the cited differences in the location of activity in the Spanish provinces can be explained, at least partially, by differences in the endowments of social capital. In addition, since Spanish economic development has been accompanied by 1 Specifically, they included “active groups” whose components participated actively, and “passive groups” where people is only part of the group but not active participants. 2 This database, along with the European Values Survey (EVS), the World Values Survey (WVS) and other databases will be commented with more detail below. 3 They also concluded that social capital owned by a single household spreads its effects to the entire village. 4 We find studies basically with two distinct levels of disaggregation: NUTS 2 if the disaggregation is at Comunidades Autónomas level and NUTS 3 is that disaggregation is at Provinces level. 3 greater levels of physical investment (Pérez, 2007) and physical capital accumulation has been considered by the literature as an important determinant of economic growth, we test as an attempt of highlight how social capital affects the economy, if a causality relationship from social capital to physical investment takes place at province level. This paper is structured as follows. In the next section we present a revision of the vast literature around the social capital concept and its measurement. Afterwards, we explain the advantages of using Pérez et al. (2005) model of social capital generation and accumulation. In section four, data, sources and some descriptive statistics are presented. In section five we test if social capital has a significant role on economic development. We also provide some evidence in favor that investment could be one of the transmission mechanisms from social capital to growth. At the end of the section robustness of the results is tested providing jackknife estimations. Finally, section six concludes. 2. Social capital literature review 2.1. Two different approaches for the same concept The concept of social capital can be examined from different perspectives. A great number of contributions deal with the concept itself and its impact on a variety of fields. It is widely accepted among scholars that social capital contributes to reduce transaction costs and affects positively economic development, among other beneficial effects. However, nowadays, despite the large number of studies (Helliwell and Putnam, 1995; Knack and Keefer, 1997; Beugelsdijk and Van Schaik, 2005), there is no agreement as to which definition, approach or methodology is the most appropriate to determine its effects. Durlauf (2002) has expressed his disagreement with the techniques used in some of the most relevant studies on the matter. The majority of the most influential contributions are dated in the past decade. Robert Putnam, in his seminal study entitled “Making Democracy Work” (1993), analyzed the effect of social capital for explaining the differences in economic development and institutional performance in the Northern and Southern regions of Italy. The main conclusion he arrived to was that social capital partly explains the large differences between the North and South of Italy in terms of institutional performance and economic development. 5 Other authors have focused their interest on testing whether Putnam’s results can be generalized using a sample of countries (Schneider et al., 2000), finding some conflicting results. We can find two distinct views for explaining the origins of social capital. Jackman and Miller (1998) compiled and discussed the different social capital approaches. They argued that the pioneering social capital studies employed an endogenous approach of the concept. That view 5 Putnam also analyzed the American case concluding that social capital has been declining in the last decades (Putnam, 1995). 4 implies that social capital is born inside the individual and the organizations. Be A and B two representative individuals in one determined society, Coleman (1988) defined trust as the expectation created in A of being corresponded by B when A makes something for B. This would imply that a stock of social capital in a given society can be created by the accumulation of reciprocal trust relationships. Coleman (1988) also argued that information is needed in providing a basis for trusting the others.6 Another relevant factor is the penalties imposed if one individual acts in opportunistic way.7 Opportunistic behavior may imply an exclusion and the impossibility to participate in the aggregated benefits that social capital provides.8 Thereby, trust to the long term is also viewed as an instrument to reach a cooperative solution in a context of the Prisoner’s Dilemma (Torsvik, 2000).9 In contrast, there is an exogenous view of the concept, which stresses that social capital is not a personal cooperative decision but a structural element of the society created by a confluence of certain cultural values, religion, political system, past and current institutions and social structure. A considerable number of studies underlying the role of “cultural matters” in explaining economic development has included these aspects in the traditional models of economic growth. Whereas both of the mentioned views are incompatible for some authors like Jackman and Miller (1998), others have not made that distinction, combining different endogenous and exogenous aspects. Let’s see, for instance, Knack and Keefer (1995), Knack and Keefer (1997), Keefer and Knack (1997), Keefer and Knack (2002), Putnam (1995), Helliwell and Putnam (1995), Akçomak and Ter Weel (2009), La Porta et al. (1997), Fukuyama (1995), or Granato et al. (1996a). In Knack and Keefer (1997) an index for measuring the effect of social capital on economic development is constructed. It consist of three endogenous variables: (i) trust; (ii) associational activity; and (iii) a civic index. These are traditionally, three of the most used variables as proxies for social capital in the literature. They found strong significant effects for trust and civic cooperation but no significant relationship between associational activity and economic performance. They also found evidence that low social polarization and formal institutional rules that constrain the government from acting arbitrarily are related to the creation of trust and cooperative norms. So, they argued that exogenous factors are the determinants of social capital proxies. The contradiction for Jackman and Miller (1998) is that the idea of the pioneering studies of social capital, leaded by Coleman (1988), considered trust and cooperative 6 In a society with certain and clear information, making decisions is easier and securer because individuals can check all the important variables they need to know to make a decision. 7 The nature of these penalties may be formal (laws and regulations) or informal (social cost imposed to opportunistic actors). The last one would be closely related with social capital. 8 Exclusion has a damaging effect not only on the excluded but on the global society. 9 In the classic iterated Prisioner’s Dilemma game, participants cooperate because they know that long-term benefits of cooperation are higher than short-term benefits derived from deviations of the cooperative solution. The nature and the mechanisms of the endogenous view are very close to this theory. 5 behaviors as a product of an internal decision, not a consequence of environmental factors. In Knack and Keefer (1995) they focused exclusively on the role of the institutions on the economic performance and the catching up effects, concluding that institutions which guarantee the property rights are crucial for achieving better economic outcomes. Within the exogenous view, we can find other authors whose research is focused on social capital as a result of political regimes and policies (Paldam and Svendsen, 2000; Paxton, 2002; Torcal and Montero, 2000; Granato et al., 1996b). Paldam and Svendsen (2000) concluded that antidemocratic, communist societies lead to groups which act in their interest, and against general interest. Rose (2000) studied for the case of Russia the different mechanisms used by Russian citizens in order to cover those needs that the state cannot supply. Paxton (2002) found evidence for explaining that the relationship between trust and democracy is reciprocal, whereas Torcal and Montero (2000), in their research for the Spanish case, concluded that the Spanish political transition in 1975 from dictatorship to democratic system did not create a larger amount of social capital stock. This fact leads authors to conclude that Spain is anchored in a low intensity social capital equilibrium. This low social capital stock is inherited, following these authors, from generation to generation as a result of past political experiences. Social capital implications can also reach the credit market. One of the most important contributions is Guiso et al. (2004), who concluded that in countries or regions with high social capital endowments their inhabitants can gain better access to credit since there is an increase in the number of credit instruments used. Social capital also increases the transparency of the available information and, putting all together, economic activity may be positively affected. Until now, we have been focusing on the different views that the social capital literature considers, and the different fields where the positive effects of social capital have been demonstrated. However, nothing about how social capital is spread inside one society has been said. It is here when the concept of the network emerges. A high number of studies has stressed its role (Coleman, 1988; Woolcock and Narayan, 2000; Paldam and Svendsen, 2000; Paxton, 2002; Torsvik, 2000). The network is considered as the instrument which provides the connections among the individuals, and allows for the diffusion of social capital. According to Pérez et al. (2005), high trust societies are characterized by a high-density, well connected network.10 The concept of network should be understood as the relationships and ties between members in the society we are considering. If individuals in one society are rich in terms of social capital but the network is not wide enough, then the positive effects that social capital provides will not be achieved. The above overview has shown that there is not a consensus on how social capital should be understood. Only one thing seems clear: regardless of the approach followed, either endogenous or exogenous, in those areas where social capital is abundant, contracts and agreements 10 Societies with isolated groups may be harmful for the creation of a social capital stock (Paxton, 2002). 6 may be enforced with lower transaction costs. However, in spite of the advances in the knowledge of this issue, more evidence on the effects of social capital is needed—at least from the point of view of some disciplines such as economics. Yet it is not an easy task because the analysts are firstly confronted with the difficulties in quantifying social capital itself, which have generated a relevant literature focused on the creation mechanisms of social capital, of which Fedderke et al. (1999), Torsvik (2000) or Woolcock (1998) constituted relevant examples. Accordingly, in recent years there has been a growing interest by scholars for determining and quantifying how important social capital is in order to achieve certain levels of economic development. 2.2. The measurement of social capital Therefore, from the previous section it may be easily inferred that one of the major problems in the study of social capital is its measurement. As indicated in the preceding paragraphs, several authors have highlighted the difficulties in the measurement because its intangible nature. Two of the measures traditionally used are the trust and associational activity indicators contained in the World Values Survey (WVS) database,11 using what scholars have referred to as “the generally speaking question”. Specifically, the question asked by the WVS in its surveys is: “Generally speaking, would you say that most people can be trusted, or that you cannot be too careful in dealing with people?”, with two possible answers: “most people can be trusted”, or “can’t be too careful”. The WVS also provides a membership association indicator. These indicators have been used in some of the major contributions in the field (Granato et al., 1996a; Knack and Keefer, 1997; Zak and Knack, 2001). Another indicator of widespread use is that contained in the European Value Studies (EVS) database,12 where we can find regional human values data for a set of European countries. The question is again the “generally speaking question”. In this case, though, answers are not dichotomous but can vary from 1 (“you can’t be too careful”) to 10 (“most people can be trusted”), providing more complete indicator. Group membership are deduced from two questions: “During the last 12 months, have you done any of the following: (i) worked in a political party or action group?” and “(ii) worked in another organization or association?”. The two possible answers are “Yes” or “No”. Other measures have also been utilized as proxies for social capital including political participation, institutional variables, confidence in government, compound civic indicators like (Knack and Keefer, 1997), or different items or questionnaires used to measure specifically social capital levels in a concrete region. A relevant study in this respect is that by Narayan and Pritchett (1999), who constructed a measure from a “Social Capital and Poverty 11 See 12 See http://www.worldvaluessurvey.org. http://www.europeanvaluesstudy.eu. 7 Survey questionnaire” to test the role of social capital viewed from a domestic perspective. Unfortunately, the measures reviewed in the preceding paragraphs have certain disadvantages which can jeopardize their use under some circumstances. First, they have a limited coverage both in the dimensions of space (number of countries or regions included) and time (years in the sample). Second, in the particular case we are dealing with, in which we attempt to understand how social capital might have affected the growth profiles of the fifty Spanish provinces, the measures just reviewed do not provide the required level of disaggregation (provinces, NUTS 3 in European terminology, which would also include the autonomous cities of Ceuta and Melilla).13 In order to expand both the space and time dimensions of our data we will consider a new measure of social capital constructed by the Ivie.14 This measure is available not only for Spanish provinces and regions, but also for a large sample of countries and long time span, which is updated on a regular basis. In addition, it has some additional features which make its use quite attractive in this particular setting. We summarize its main characteristics in the next section. This measure has already been used in recent studies applied to different contexts but with aims related to ours such as Pastor and Tortosa-Ausina (2008) or Miguélez et al. (2011). 3. Using an economic approach of social capital As indicated above, an important stem of the literature has been devoted to measure the impact of social capital on growth and, to a lesser extent, convergence using proxy data for social capital drawn from surveys. In contrast to other measures of social capital such as those reviewed in the preceding section, the measure we use is more elaborated one. We devote this section to stress those of its features which are more relevant for our study in terms of its suitability. Appendix A provides the most important mathematical relations of the measure. For a more complete explanation, see (Pérez et al., 2005). As discussed previously, data from surveys provided by WVS or EVS are not available neither for the different Spanish provinces nor for the analyzed time period . In contrast to the surveys described above, which are only available for certain years,15 the measure of social capital we use provides data yearly, enabling the construction of a balanced panel data, which can yields stronger conclusions. This is one of the reasons why we use the Ivie’s measure but its use has other powerful advantages. One of the most interesting features of the measure is that it deals with social capital as 13 As indicated in the introduction, some studies such as Beugelsdijk and Van Schaik (2005) have analyzed social capital issues for European regions; however, the level of disaggregation employed was far less detailed than that corresponding to Spanish NUTS 3. 14 Instituto Valenciano de Investigaciones Económicas (www.ivie.es), in collaboration with the Banco Bilbao Vizcaya Foundation (FBBVA, www.fbbva.es). 15 There are different waves but frequency is not annual. 8 an asset in which people can invest. Solow (2000) disagreed with the idea that social capital can be one of the drivers of economic activity, in part because he thought that social capital cannot be seen as capital. Specifically, he claimed that the word capital is related to a stock of production factors which is expected to yield productive services for some time. Dasgupta and Serageldin (2001) suggested the plausibility of the construction of an index of aggregated social capital, concluding also that additional research should follow in that direction. Glaeser et al. (2000) expressed that the traditional measures for social capital are not the most suitable when we are in the economic field. They developed a model of individual social capital accumulation, acknowledging the existence of difficulties in the aggregation at the society level. Therefore, this model cannot provide an answer when studying the differences among provinces, which are not individuals but communities of individuals and, consequently, aggregation becomes essential. In the same line, Durlauf (2002) criticized the lack of a theoretical framework for the determinants of social capital formation and accumulation and also pointed out the weakness of the studies which tests the importance of social capital from a macroeconomic perspective. The social capital measure we use provides an answer in this respect. The model of social capital accumulation considered is based for its construction on similar ideas as models of physical capital accumulation. This implies that social capital is understood as an additional input in the production process and a stock of it is available for each society, which depreciates over time as any other type of capital stock. Individuals invest in social capital because they expect positive returns in the future derived from that investment. Ivie’s approach considers that social capital provides services, and those services translate into a reduction of transaction costs. That cost reduction conforms the final benefits of investing in social capital. Another advantage of this approach is the importance that the measure devotes to the economic aspects in the generation of social capital as opposed to other measures focusing on social and cultural characteristics. Our approach considers the economic relationships such as trade, employment, financing or income distribution as determinants of the incentives for investing in social capital. Pérez et al. (2005) claim that the cited economic variables have not been sufficiently considered by the literature of social capital, and that they could be even more important than other social or cultural variables widely considered. The cited author also gives several explanatory reasons justifying the dominance of social variables over economic variables in the measurement of social capital. The main conclusion they arrive to is that social capita generation cannot be exclusively confined to non-economic relationships, and that economic relationships must also be taken into account, especially when we are dealing with advanced economies with expectations of a continuous progress, which is precisely the case of Spanish provinces. The above arguments manifest that the use of Ivie’s approach can be more suitable in the 9 specific context we are working in, where we attempt to evaluate the implications of social capital on regional economic growth and performance. This economic approach to measure social capital overcomes some of the biggest difficulties highlighted by the literature: the vagueness, the measurement, the aggregation, the treatment of social capital as an asset in which people can invest and the consideration of economic variables in the social capital formation process. It can also offers additional insights in order to understand the role of one concept characterized by a multifaceted perspective better, and its use will allow for comparison with previous results from studies which have been using more traditional measures as described above. 4. Data and descriptive statistics Data for the estimations are disaggregated by provinces. There are 50 provinces in Spain excluding Ceuta and Melilla.16 They correspond to NUTS 3 in European nomenclature. Selecting the explanatory factors of economic growth is not an easy matter. As we previously commented in the introduction, a number of references have focused on determining the true determinants of economic growth. The only three variables which appear always significant are: initial level of income, investment rate and human capital. Sala-i-Martin (1997), in an effort to further investigate more possible robust variables, elaborated a robustness analysis using a modification of the extreme bounds test initially developed by Leamer (1985), concluding that a considerable set of variables could be used as robust growth determinants.17 Unfortunately, a measure of social capital is not included in that robustness analysis, in part because the major of the studies incorporating social capital are more recent. In our case, in order to explain the different levels of GPD per capita ( GDPpc) that Spanish provinces present, we consider as an explanatory variables the three “robust variables”: the initial level of GDP per capita ( GDPpc83), which reflects the β convergence effect18 , the private physical capital investment ( PK ), human capital ( HK ) and, following Pérez et al. (2005), we also include the employed population ( Employ). Last variable can play an important role in Spain because in the last decades, but specially in the period 1995 - 2005, Spanish economic growth has been based on a strong generation of employment with important differences among provinces. Together with this set of variables, we introduce to the model our variable of interest, social capital (SK ). On the second stage of the study, we test the impact of social capital on private physical cap16 Ceuta and Melilla are not provinces but autonomous cities, with different characteristics to those of provinces and consequently are not considered in the analysis. 17 Concretely, besides the three cited variables, Sala-i-Martin (1997) found 9 different “groups” of robust variables: regional variables, political variables, religious variables, types of investment, primary sector production, openness degree, type of economic organization and former Spanish Colonies. 18 β convergence reflects that those provinces in the sample less economically developed at the beginning of the period, grow in higher degree. That idea derives from the assumption of diminishing returns of the productive factors. 10 ital investment. One more, there is not an agreement about which are really the determinants of this type of investment and authors studying this matter have used different explanatory variables. Studies such as Knack and Keefer (1997) or Zak and Knack (2001) consider the price of investment goods, which is theoretically one of the potential drivers of investment although this variable does not appear, for instance, in Dearmon and Grier (2011), who incorporate other macroeconomic indicators like the lag of the inflation, the lag of government spending as % of GDP and the lag of GDP growth besides a human capital indicator, arguing that last may present spillover effects which can affect investment. In our case, we have considered as an explanatory variables the legal interest rate (r), a price index (CPI ), the lag of GDP (lagGDP), the lag of private physical capital investment (lagPK ), human capital ( HK ) and the unemployment rate (Unem). The decision of incorporating lagGDP is sustained by the argument that investment activity may be influenced by the GDP obtained in the previous year because it could determine the saving level. Another reason is that the decision of investing is one which has to be deliberated, so it needs time to become reality. Both commented ideas can make that the income achieved in a certain year is not immediately invested. By its part, lagPK is expected to be positive because investment in one year may constitute a positive incentive in the recipient provinces, which can make them more attractive in investment in the future periods. We find some evidence in Bernanke (1983), who found a significant relationship between the lag of investment and investment in the following period. Because we consider that expectations about the future evolution of the economy play an important role on the investment decision, we include Unem as an additional regressor. The unemployment rate has been traditionally one the main worries for Spanish population and it could be a good indicator of how the economic environment is perceived. Unem is not only an expectation measurer in the sense that higher levels of unemployment are expected to be explained by economic weakness, so accordingly, investment activity will be reduced. Because of the large impact of the construction sector in Spain during the analyzed period, which is partly responsible for the current crisis affecting the country, we consider also using as a dependent variable the physical private investment subtracting the amount corresponding to the residential investment ( PKNR). Figure ?? shows that in provinces such as Málaga, Alacant and Illes Balears, the residential component is around the 50% of the total private physical investment. To our knowledge, there is not any literature focused on the role of social capital in investment separating that residential component. We carry out this because in the specific case we are dealing with, Spain, that distinction turns practically necessary. The reason is that the “Spanish construction bubble”, which exploded in 2008, was been conformed since 1995, after the 1993-1994 economic crisis, so an important part of the analyzed period is affected by this fact. If social capital is one of the determinants of investment, it could be quite interesting to determine if its effects remain significant when residential component is removed and how 11 much important the possible differences are. Figures ?? to 7 and table 1 provide some descriptive statistics. As a previous step before the subsequent regression analysis, it could be useful to have a look at the mentioned statistics. We have depicted in scatter plots the different observations of the variables GDPpc, PK, PKNR and SK. This kind of graph shows the simple correlation among variables and gives us an idea about any potential relationship among them, besides permitting the detection of outliers which can mar or results. Figure 2 shows a positive relation between social capital and GDP per capita. Spearman’s ρ is 0.62, significant at 1% level. Figures 3 and 4 denote a similar pattern. In this case, Spearman’s ρ is 0.83 for both cases, which are also significant at 1% level. These preliminary results are valid in this previous stage but we test in the next section through a panel data regression analysis if these results are sustained. Finally, before closing this section, looking at the maps could be utile in order to understand how variables are distributed inside Spanish territory. In figure 5 we can observe significant disparities in the GDP per capita among provinces. In 1983, Mediterranean’s, North’s provinces and Madrid presented the highest levels whereas South’s and interior’s provinces presented the lowest. The same can be observe in 2005, the end of the period so, differences are highly persistent. Figure 6, which represents the total physical capital investment by provinces, displays a close pattern, also showing relevant and persistent differences. Again, Madrid and coastal North’s and Mediterranean’s provinces present the highest levels. Sevilla and Málaga, which are not among the richest provinces in terms of GDP per capita, are high-investment provinces. We can also notice how physical capital investment are more polarized at the end of the period. Provinces like Córdoba, Cáceres, Badajoz, Ciudad Real and Asturias loss weight in favor of coastal provinces and Madrid. This fact leads to an immediate question: Which factors are behind this polarization process? Because the necessity of additional insights about the determinants of these disparities becomes evident, this paper provides a response by considering social capital, which has not been accounted for in previous works for Spain, at least from a province perspective and analyzing the effects on investment too. Figure 7 depicts how the biggest stocks of social capital in 1983 were in Madrid, Mediterranean’s and North’s provinces besides Córdoba and Sevilla. In this case, polarization is not as evident as in physical investment, but differences are also accused. It is remarkable how social capital declined between 1983 and 2005 in Cantabric’s coastal provinces (A Coruña, Asturias and Cantabria) and also in Pontevedra and León. In 2005, the biggest stocks of social capital are in Madrid, Mediterranean provinces, Zaragoza, Sevilla and Vizcaya. Data for the estimations are provided by different institutions. In Appendix B an explanation of all variables used in the study and the data sources is available. 12 5. 5.1. Results Social capital and GDP per capita Our framework for the regression analysis is a standard growth equation, including those variables defined in Section 4. The dependent variable is GDP per capita. All variables are expressed in logarithms. The availability of data considered for the hole period 1983 - 2005 disaggregated by provinces allows for a panel data estimation. It can provide stronger conclusions than previous studies based on cross-sectional estimations. We estimate the following model: GDPpcit = αi + β1 GDPpc83it + β2 PKit + β3 Employit + β4 HKit + β5 SKit + µit (1) All variables have been introduced in 4 and defined in Appendix B. In order to test if data present any specification problem, heteroscedasticity, serial autocorrelation and spatial autocorrelation are tested. The first one is tested by using modified Wald’s test considering Greene and Zhang’s (2003) suggestion, which makes that the test works properly under the assumption that errors are non-normal distributed. Serial autocorrelation is tested with Wooldridge’s (2002) autocorrelation test and finally, Pesaran’s (2004) spatial autocorrelation test allows for testing whether our data suffer cross-sectional dependence. For all three tests we reject the null hypothesis of no specification problem, so, we have to carry out the estimations correcting the mentioned problems in order to provide valid statistic inference. With the purpose of controlling the unobservable heterogeneity, fixed effects by provinces are used. We test its convenience through the Hausman’s test. In this respect, different studies for the Spanish case have reached dissimilar conclusions. For instance, Miguélez et al. (2011) found that fixed effects were not important, estimating the model with random effects whereas Peña Sánchez (2008) used fixed effects in his study. It is curious that both studies utilized the same level of disaggregation, NUTS 2, so, it seems to be that there is not a consensus in the literature in respect of this matter. In our case, because standard Hausman’s test does not work properly under the aforementioned specification problems, we perform the test adopting the refinement proposed by Wooldridge (2002), which provides valid statistical inference for these cases. Results corroborate that fixed effects are important. Ignoring cross-sectional correlation in the estimation of panel data models can lead to severely biased statistical results (Hoechle, 2007). To our knowledge, cross-sectional dependance has not been tested in other panel data studies evaluating the role of social capital. Thereby, due to Pesaran’s spatial autocorrelation test denotes that our data suffer crosssectional autocorrelation, we estimate the model using Driscoll and Kraay (1998) standard errors, which are robust to all three commented specification problems. Table 2 shows the 13 results. We incorporate social capital and the controls sequentially. In all cases, coefficients show the expected signs and are highly significant. Consistently with economic growth theory, the sign of the initial level of GDP per capita is negative. That means that those provinces with lower level of GDP per capita in 1983 have experienced a higher growth. Coefficient is small becoming evident that GDP per capita convergence process in Spain has not advanced considerably during the sample period.19 Comparatively, the variable PK has had the biggest impact on the dependent variable in the considered period. Concretely, a 10% increase on private physical capital generates, ceteris paribus, a 2% increase on GDP per capita in average. This result is in line with Peña Sánchez (2008), who obtains the same coefficient but using labor productivity as a dependent variable and NUTS 2 level of disaggregation. The positive and significant sing of the variable HK denotes that the role of the education on the economic performance is also quite important. The coefficient is surprisingly high and it can be compared with PK. The variable Employ is also positive and significant at 5%. It is not surprising that in those areas with higher levels of employment, residents enjoy upper levels of GDP per capita. Our variable of interest, SK, is positive and highly significant and its coefficient is the smallest. Concretely, a 10% increase in social capital remaining the other variables constant, yields a 0.5% increase on GDP per capita. Previous evidence for Spain is scarce. Pérez et al. (2005), who carried out a growth accountant exercise for Spain, found a positive contribution from social capital to growth for the period 1964 - 2001, also finding negative contributions in certain periods20. This fact supports the idea that social capital is a significant variable for explaining the level of economic performance inside the Spanish provinces. 5.2. Social capital and investment In order to determine how social capital is affecting economic performance, we test if investment could be one of the mechanisms. Authors such as (Hall and Jones, 1999; Dearmon and Grier, 2011; Zak and Knack, 2001) found a positive relationship between social capital and physical capital investment. These authors also use, as the majority of the literature, data from surveys. We perform a similar study but using the social capital measure considered above. The dependent variable in this case is the total private investment in physical capital (PK). The explanatory variables have been already defined in the previous section and Appendix B provides a full explanation. Variables are expressed in logarithms except CPI, r and Unem. We 19 As we commented in the introduction, a vast literature about the convergence process in Spain have determined that the Spanish convergence process slowed down in the 1980’s. 20 In the cited growth accountant exercise, authors disaggregate economic growth in different explanatory factors. This kind of exercise highlighted a positive relation between social capital and growth, showing how economic performance was affected by social capial variations, both negatives and positives. 14 estimate the following model: PKit = αi + β1 lagGDPit + β2 lagPKit + β3 CPIit + β4 r + β5 Unemit + β6 SKit + µit (2) As we mentioned, due to the importance of the residential investment in Spain in the analyzed period, specially in coastal provinces and Illes Balears, we have developed the same study but excluding this residential component (PKNR). Explanatory variables are exactly the same. Thus, the estimated model is: PKNRit = αi + β1 lagGDPit + β2 lagPKNRit + β3 CPIit + β4 r + β5 Unemit + β6 SKit + µit (3) Table 3 shows the estimation results. Heteroscedasticity, serial and spatial autocorrelation tests indicate that, once more, we reject the hypothesis of no specification problems so, results are estimated, as in the previous analysis, by using Driscoll and Kraay standard errors. Robust Hausman’s test manifests again that fixed effects are important. We find a positive and highly significant relationship between social capital and investment. Specifically, when the investment considered is the total value, an increase of 10% of social capital is corresponded by a 0.7% increase in total private physical investment. The regression allows us for determining other important relationships. Was expected, lagPK is positive and significant and jointly with lagGDP, has the biggest impact on the dependent variable (0.7% and 0.5% respectively as a response to a 10% increase). It is surprising the result obtained for r because, albeit the sign is the expected, is not significant. CPI has a positive but not significant effect and Unem is negative and its effect is significant at 5%. HK is positive and has a weakly significant effect. Thus, from that analysis we can extract that social capital is relevant for explaining the investment levels in the different Spanish provinces. Its effect is similar to human capital but statistical inference results are stronger for social capital in terms of significance level. Concentrating on the results for investment when the part corresponding to the residential component is detracted, results are quite similar. The only change is that Unem becomes no significant. It can be explained because activities not related with construction are not linked with employment in the same degree as construction activities are.21 Nevertheless, the sign remains negative showing that high levels of unemployment detract investment because denote economic weakness. HK is positive and 10% significant and its effects are slightly higher than in the previous analysis.22 Social capital is again positive and highly significant. Results show, ceteris paribus, that a 10% increase in SK generates an 0.9% increase in investment in non-residential private invest21 Traditionally, construction sector is linked with employment because the nature of the activity. investment activities are characterized by greater necessities of human capital. 22 Non-residential 15 ment. That coefficient, although is only moderately higher than in late analysis, has involvements in the sense that if activities which provide more aggregated value are those not related with construction sector, this finding may imply that in provinces where the cited sector has a smaller weight in the economy, the role of social capital is more substantial than in other provinces where construction sector is one of the pillars of the economy. In this part of the paper we have highlighted that social capital is one of the drivers of private physical capital investment. There is no previous evidence for the Spanish case, so comparison with preceding results is not possible. Due to private physical capital investment is, according to the first result of the paper, the most relevant explanatory factor of GDP per capita, the result obtained in this part is, from our point of view, quite interesting. It supports the hypothesis that the increase in the goodness level in the different provinces depends on their capacity for attracting investment, which can generate more activity, employment and wealth. So, differences in social capital endowments among provinces can be one of the factors which determine the large differences existing in our period of reference. 5.3. Robustness analysis In order to test whether the above results are robust we perform a jackknife estimation, which is a common non-parametric technic. It consists of running regressions removing one different province in the sample each time so, in our case, 50 different regressions are performed. From these estimated results, a distribution of the variance is constructed. Table 4 reports the results for all three regressions. For the case of social capital and GDP per capita, significance of the explanatory variables does not vary in any case. When jackknife is applied to estimate social capital and investment, there are small changes in some variables. For SK, significance drops from 1% to 5%. HK increases its significance until 5% whereas the variable r, which was not significant in the standard analysis, is now significant at 5% level when the dependent is PK and remains no significant for PKNR. This result shows that interest rate is important when investing in residential assets but is not relevant for non-residential investment. Unem loses any significance when the dependent variable is PK and remains no significant for PKNR. The rest of the variables does not suffer any variance. The jackknife technic has showed that the results in the first part of the paper were robust and also has corrected biases presented in the second block of estimations. The important idea we have to extract from these results is that social capital is robust in both cases. 16 6. Concluding remarks Spanish provinces have historically presented considerable disparities in terms of GDP per capita. Although differences reduced significantly during the period 1955 - 1980, literature agrees that this process of convergence came to an end in the decade of 1980. Over the last few years there, has emerged a growing interest in determining whether social capital has an impact on economic growth. The encouraging results found by authors such as Putnam et al. (1994); Keefer and Knack (1997); Beugelsdijk and Van Schaik (2005) for the international case drive us to analyze if social capital is behind the commented disparities in the Spanish provinces. Data on social capital have been traditionally based on surveys. The availability of a new data base of social capital, which provides data for the period 1983 - 2005 at province level, enabled us to estimate the results using a panel data approach. Albeit panel data studies are growing up, they are still scarce for the study of these “social features” due to the data shortage commented along the study. The social capital measure utilized, not only is available either with higher degree of disaggregation and a wider time span than the classic ones, but also is a measure which solves some of the problems highlighted by the literature in terms of measure, aggregation and the way the measure is constructed besides the elements considered in its construction. Another contribution of this article is the methodology. To our knowledge, previous studies for the Spanish case did not control for the presence of cross-sectional dependence. Neglecting that problem may lead to invalid statistic inference. After we noticed that our data presented cross-sectional dependence, we used Driscoll and Kraay standard errors for the estimations. Estimating using that refinement corrects heteroscedasticity and both serial and spatial autocorrelation. According to our results, we determine in the first part of the study that social capital has a positive influence on the level of GDP per capita of Spanish provinces. This result is highly relevant in the sense that confirms that social capital is, among other factors, behind the differences presented by the Spanish provinces in terms of GDP per capita. That disparities are characterized by its high persistence along time. The same occurs with social capital, confirming the idea developed by Torcal and Montero (2000), who acknowledged in their analysis for Spain the difficulties to escape from a low social capital endowment situation. We have studied a twenty-three year period and maybe is not enough time to reverse the cited circumstance. These results might have interesting policy implications in the sense that if social capital is one of the mechanisms to achieve a higher stage of economic performance, policies should pursue the generation of greater endowments of social capital in those provinces where this asset is relatively scarcer. 17 A second conclusion we can extract from this paper is the importance of social capital to foster investment. We have shown that investment is the most relevant factor in the increase of the output per capita. Investment is an activity which needs trust. We all know that the major of the investment activities are made borrowing financial resources. The presence of social capital in a given society or region makes easier and cheaper this kind of activities. The theory of social capital declares the importance of the social features in the reduction of transaction costs. If banks can save costs in supervision and checking the reliability of clients, the last can obtain cheaper credit. Trust also can extend the relationships between banks and clients and this fact induces lower transaction cost in future economic transactions. So, in the current economic context in which credit does not flow as a few years ago, social capital might be an additional instrument to recover the credit, specially in those activities not related with construction, and drive the economy ahead in the following years. The enormous importance of the construction sector in Spain, especially in the last decade of the considered period, leads us to disaggregate the investment extracting the residential component. Effects of social capital are quite similar but, when we detach the last component the coefficient is higher and the significance is prevailed. Thereby, in those areas where investment in residential assets is more relevant, the effects of social capital fostering investment are slimly lower. To sum up, this paper has evidenced the importance of social capital as a considerable factor for explaining the economic development and the persistent differences among Spanish provinces. We have also reported evidence that investment may be one of the candidate transmitters of the stimulant effects from social capital to GDP per capita. Additionally, using the social capital measure that Ivie provides we give an answer to some problems related with the treatment of social capital besides correcting some econometric issues ignored in other contributions in the matter. Our future research agenda is comprised basically by two immediate goals. First, we try to extend the present analysis at country level. It could be interesting in comparison with other cross-country studies that use cross-sectional samples using data from WVS and EVS. Second, given the controversy around the explanatory factors of economic growth and the amount of possible variables as a regressors through parametric approaches (Sala-i-Martin, 1997), we should consider some additional flexible techniques within the nonparametric econometrics field. These remain as open and intriguing questions to be explored in immediate research initiatives. 18 Appendix A. An economic approach to measuring social capital. Some basic ideas The social capital measure we use is based on three initial hypothesis: 1. Cooperation in a society is favored by the economic incentives derived from a higher expected incomes, result of a continuous growth. 2. The incentives for cooperation are reinforced/weaken by two factors: • The effective opportunity of participation in the final incomes. • The culture of reciprocity fulfillment. 3. The effects of cooperation are increased in societies with a high density network. The investment in social capital is denoted by ( Is ). A member in a given society invests in social capital if the expected benefits of cooperation are positive ( Is > 0). If the economy follows a continuous growth trend, the income achieved is higher than the simple reposition of the production factors and, moreover, the results are crescent in time. y > rk + w̄ (4) where y is the income, rk is the cost of the physical capital and w̄ is the salary of labor. Other assumptions of this approach are: • The individuals observe the difference in the incomes that they obtain under certain time and place conditions and other conditions less favorable. • This difference determines the incentives for cooperation and trust (investment in social capital). An individual incurs in two types of costs to obtain incomes: • Cost of contribution with productive resources (we expect a retribution equal to the reposition costs). • Cost in terms of effort of cooperation inside an incomplete information environment. Cost of cooperation include both time and psychical costs. Following the above statements, benefits are expressed as: π = y − (rk + w̄) − w̄C ( Is ) 19 (5) where C ( Is ) is the cost of cooperation measured in wage terms. If one individual owns social capital, she/he would expect to obtain additional income using it for her/his economic transactions. The T horizon defines her/his expectations according to the duration of her/his economic relations inside a society or network. If her/his expectations are not fulfilled, her/his social capital will be depreciated at ρ rate. In a given moment, our representative individual invests in social capital if, T π= 1 ∑ (1 + ρ)t (yt (1 − G) − rkt − w̄t (1 + C( Ist ))) > 0 (6) t =0 where (1 − G ) is the Gini’s Index and measures the inequality in the society. The next step is to focus on the services that social capital provides, ( f ks). The capability of social capital to contribute to an increase of total output depends on its capacity to generate services, i.e. a reduction in transaction costs. (7) f ksi = ci ksi where ci is the degree of connection of the network and ksi is the social capital stock of the individual i. If a given individual is perfectly connected, it implies ci = 1, the contribution of social capital is maximum. The opposite holds for ci = 0. The economic value of the services of social capital is defined in terms of its use cost ui . (8) u i = ρi + di where ρi is the financial opportunity cost, and di is the depreciation cost. Therefore, the value of the services of social capital can be expressed as: vksi = ui f ksi = (ρi + di)ci ksi (9) The final step is the aggregation of the social capital of the individuals. Services cannot be directly added because of their varying nature. Therefore, authors follow a multiplying process, weighting each social capital unit by its own use cost weighted respect the total use cost. The weight is calculated as: vi = vksi vks j (10) ∑N j=1 Regarding the above consideration, the services of social capital are aggregated as follows: N N i=1 i=1 vi vi KS = N ∏ f ksvi i = N ∏ c i ksi 20 (11) Appendix B. Variables and data sources • GDPpc: Real GDP per capita (e). Serie deflated using 2000 as base year. [Source: Instituto Nacional de Estadística (INE)]. • GDPpc83: Real GDP per capita in 1983 (e). Year 2000 used as a base year. [Source: Instituto Nacional de Estadística (INE)]. • PK: Private physical capital investment (Thousand e). Serie deflated using 2000 as base year. [Source: BBVA Foundation]. • PKNR: Private physical capital investment detracting the residential component (Thousand e). Serie deflated using 2000 as base year. [Source: BBVA Foundation]. • Employ: Employed population (persons). [Source: Instituto Nacional de Estadística (INE)]. • HK: Rate of active working population at least with secondary studies. Values between (0 - 1). [Source: Instituto Nacional de Estadística (INE) and Ivie-Bancaja]. • SK: Social capital services stock. Data are provided by Ivie’s (1964 - 2001) database and its updating until 2005. Both series have been connected using Spain 1983 = 100 as a reference point. [Source: Ivie] • CPI: Consumer Price Index. Year 2001 = 100. [Source: Instituto Nacional de Estadística (INE)]. • Unem: Unemployment rate in percentage. [Source: Instituto Nacional de Estadística (INE)]. • r: Legal Interest Rate in percentage. [Source: Instituto Nacional de Estadística (INE)]. References Akçomak, İ. and Ter Weel, B. (2009). Social capital, innovation and growth: Evidence from Europe. European Economic Review, 53(5):544–567. Barro, R. (1991). Economic growth in a cross section of countries. The Quarterly Journal of Economics, 106(2):407. Becattini, G. (1979). Dal settore industriale al distretto industriale. alcune considerazioni sullunità di indagine delleconomia industriale. Rivista di Economia e Politica Industriale, 5(1):7– 21. 21 Bernanke, B. (1983). The determinants of investment: Another look. The American Economic Review, 73(2):71–75. Beugelsdijk, S. and Van Schaik, T. (2005). Social capital and growth in European Regions: an empirical Test. European Journal of Political Economy, 21(2):301–324. Coleman, J. (1988). Social capital in the creation of human capital. American Journal of Sociology, pages 95–120. Crespo-Cuaresma, J., Foster, N., and Stehrer, R. (2011). Determinants of regional economic growth by quantile. Regional Studies, 45(6):809–826. Dasgupta, P. and Serageldin, I. (2001). Social capital: A Multifaceted Perspective. World Bank Publications. Dearmon, J. and Grier, R. (2011). Trust and the accumulation of physical & human capital. European Journal of Political Economy, 27:507–519. Dolado, J., González-Páramo, J., and Roldán, J. (1994). Convergencia económica entre las provincias españolas: evidencia empírica. Documentos de trabajo (Banco de España) ; 9406. Driscoll, J. and Kraay, A. (1998). Consistent covariance matrix estimation with spatially dependent panel data. Review of Economics and Statistics, 80(4):549–560. Durlauf, S. (2002). On the empirics of social capital. The Economic Journal, 112(483):459–479. Fedderke, J., De Kadt, R., and Luiz, J. (1999). Economic growth and social capital: a critical reflection. Theory and Society, 28(5):709–745. Fukuyama, F. (1995). Social capital and the global economy: a redrawn map of the world. Foreign Affairs, 75(4). Glaeser, E., Laibson, D., and Sacerdote, B. (2000). An economic approach to social capital. The Economic Journal, 112:437–458. Goerlich, F. and Mas, M. (2001). La Evolución Económica de las Provincias Españolas: (1955-1998). Fundación BBVA. Goerlich, F., Mas, M., and Pérez, F. (2002). Concentración, convergencia y desigualdad regional en españa. Papeles de Economía Española, 93 (2002):17–36. Granato, J., Inglehart, R., and Leblang, D. (1996a). Cultural values, stable democracy, and economic development: a reply. American Journal of Political Science, 40(3):680–696. 22 Granato, J., Inglehart, R., and Leblang, D. (1996b). The effect of cultural values on economic development: theory, hypotheses, and some empirical tests. American Journal of Political Science, 40(3):607–631. Greene, W. and Zhang, C. (2003). Econometric Analysis, volume 5. Prentice hall New Jersey. Guiso, L., Sapienza, P., and Zingales, L. (2004). The role of social capital in financial development. American economic review, 94(3):526–556. Hall, R. and Jones, C. (1999). Why do some countries produce so much more output per worker than others? Quarterly Journal of Economics, 114(1):83–116. Helliwell, J. and Putnam, R. (1995). Economic growth and social capital in Italy. Eastern Economic Journal, 21:295–307. Hoechle, D. (2007). Robust standard errors for panel regressions with cross-sectional dependence. Stata Journal, 7(3):281. Jackman, R. and Miller, R. (1998). Social capital and politics. Annual Review of Political Science, 1(1):47–73. Keefer, P. and Knack, S. (1997). Why don’t poor countries catch up? A cross-national test of an institutional explanation. Economic Inquiry, 35(3):590–602. Keefer, P. and Knack, S. (2002). Polarization, politics and property rights: Links between inequality and growth. Public Choice, 111(1):127–154. Knack and Keefer (1997). Does social capial an economic payoff?A cross country investigation. The Quarterly Journal of Economics, 112:1251–1288. Knack, S. and Keefer, P. (1995). Institutions and economic performance: cross-country tests using alternative institutional measures. Economics & Politics, 7(3):207–227. La Porta, R., Lopez-de Silanes, F., Shleifer, A., and Vishny, R. (1997). Trust in large organizations. The American Economic Review, 87(2):333–338. Leamer, E. (1985). Sensitivity analyses would help. The American Economic Review, 75(3):308– 313. Levine, R. and Renelt, D. (1992). A sensitivity analysis of cross-country growth regressions. The American Economic Review, pages 942–963. Mankiw, N., Romer, D., and Weil, D. (1992). A contribution to the empirics of economic growth. The Quarterly Journal of Economics, 107(2):407. 23 Miguélez, E., Moreno Serrano, R., and Artís Ortuño, M. (2011). Does social capital reinforce technological inputs in the creation of knowledge? evidence from the spanish regions. Regional Studies, 48.8(13):1019–1038. Narayan, D. and Pritchett, L. (1999). Cents and sociability: household income and social capital in rural Tanzania. Economic Development and Cultural Change, 47(4):871–897. Nonneman, W. and Vanhoudt, P. (1996). A further augmentation of the solow model and the empirics of economic growth for oecd countries. The Quarterly Journal of Economics, 111(3):943–953. Paldam, M. and Svendsen, G. (2000). Missing social capital and the transition in Eastern Europe. Working Paper. Pastor, J. M. and Tortosa-Ausina, E. (2008). Social capital and bank performance: An international comparison for OECD countries. The Manchester School, 76(2):223–265. Paxton, P. (2002). Social capital and democracy: An interdependent relationship. American Sociological Review, 67(2):254–277. Peña Sánchez, A. (2008). Las disparidades económicas regionales en españa. Revista de Estudios Regionales, (82):105–132. Pérez, F. (2007). Claves del Desarrollo a Largo Plazo de la Economía Española. Fundación BBVA. Pérez, F., Montesinos, V., Serrano, L., and Fernández de Guevara, J. (2005). La Medición del Capital Social: Una Aproximación Económica. Bilbao, Fundación BBVA. Pesaran, M. (2004). General diagnostic tests for cross section dependence in panels. CESifo Working Paper Series No. 1229; IZA Discussion Paper No. 1240. Portela, M. and Neira, I. (2002). Capital social: concepto y estudio econométrico sobre el capital social en españa. Estudios Económicos de Desarrollo Internacional, 2(2):1. Putnam, R. (1995). Bowling alone: America’s declining social capital. Journal of Democracy, 6:65–78. Putnam, R., Leonardi, R., and Nanetti, R. (1994). Making Democracy Work: Civic Traditions in Modern Italy. Princeton Univ Pr. Putnam, R. D. (1993). Making Democracy Work: Civic Traditions in Modern Italy. Princeton University Press, Princeton, NJ. 24 Rabadán, I., Salas, R., and Documenta, F. (1996). Convergencia y redistribución interterritorial en España: efecto de los impuestos directos, cotizaciones sociales y transferencias. Fundación BBV Documenta. Raymond Bara, J. and García Greciano, B. (1994). Las disparidades en el pib per cápita entre comunidades autónomas y la hipótesis de convergencia. Papeles de Economía Española, (59):37–59. Rose, R. (2000). Getting things done in an antimodern society: social capital networks in Russia. Social Capital: A Multifaceted Perspective, pages 147–172. Sala-i-Martin, X. (1997). I just ran two million regressions. The American Economic Review, pages 178–183. Schneider, G., Plumper, T., and Baumann, S. (2000). Bringing Putnam to the European regions. European Urban and Regional Studies, 7(4):307. Solow, R. (1957). Technical change and the aggregate production function. The Review of Economics and Statistics, 39(3):312–320. Solow, R. (2000). Notes on social capital and economic performance. Social capital: A Multifaceted Perspective, 6(10). Torcal, M. and Montero, J. (2000). La formación y consecuencias del capital social en España. Revista Española de Ciencia Política, 1(2):79–121. Torsvik, G. (2000). Social capital and economic development. Rationality and Society, 12(4):451. Trigilia, C. (2005). Sviluppo Locale: Un Progetto per l’Italia. GLF editori Laterza. Woolcock, M. (1998). Social capital and economic development: toward a theoretical synthesis and policy framework. Theory and Society, 27(2):151–208. Woolcock, M. and Narayan, D. (2000). Social capital: implications for development theory, research, and policy. The World Bank Research Observer, 15(2):225. Wooldridge, J. (2002). Econometric Analysis of Cross Section and Panel Data. The MIT press. Zak, P. and Knack, S. (2001). Trust and growth. The Economic Journal, 111(470):295–321. 25 Table 1: Descriptive Statistics 26 Province Almería Cádiz Córdoba Granada Huelva Jaén Málaga Sevilla Huesca Teruel Zaragoza Asturias Illes Balears Las Palmas Sta. Cruz de Tenerife Cantabria Ávila Burgos León Palencia Salamanca Segovia Soria Valladolid Zamora Albacete Ciudad Real Cuenca Guadalajara Toledo Barcelona Girona Lleida Tarragona Alacant Castelló València Badajoz Cáceres A Coruña Lugo Ourense Pontevedra Madrid Murcia Navarra Álava Guipúzcoa Vizcaya La Rioja GDPpc 1983 2005 7,290 15,330 7,750 13,010 6,990 11,070 6,330 11,590 7.980 13,960 7,000 10,550 7,600 13,390 6,980 13,160 8,590 16,810 11,500 17,760 9,720 18,420 8,360 14,750 12,080 18,280 11,290 15,810 9,440 14,590 9,920 16,360 7,030 13,440 10,690 19,330 8,860 14,890 10,230 16,720 7,010 14,140 9,270 17,460 9,250 16,230 9,950 18,100 6,310 12,910 6,880 12,610 7,580 13,150 6,840 12,990 8,440 14,680 7,520 13,440 10,760 19,860 11,820 19,950 12,090 19,490 13,510 20,110 9,720 14,450 10,930 18,130 9,250 15,290 5,010 11,560 5,230 11,590 9,170 14,400 9,320 13,270 6,290 12,090 7,980 13,550 12,470 22,120 9,050 13,810 12,700 21,450 16,800 23,780 12,440 21,740 12,360 20,970 10,320 18,460 PK 1983 823,228 1,610,586 953,348 1,223,938 579,276 650,704 2,181,167 1,798,013 572,749 323,480 1,343,488 1,875,148 968,191 1,242,225 1,399,457 1,010,428 325,410 813,416 1,267,558 458,520 739,047 344,360 250,870 840,467 385,672 436,873 971,295 438,970 395,289 907,573 6,307,167 1,770,484 898,672 2,843,645 2,817,847 872,711 3,506,459 936,581 1.205,645 1,632,378 660,475 520,469 1,276,661 8,431,737 1,823,147 763,947 692,147 928,477 1,830,828 520,019 2005 2,740,018 4,532,782 2,410,765 3,078,974 2,272,457 2,024,585 7,541,517 6,702,257 1,705,614 967,757 4,530,717 4,686,708 5,926,796 4,089,235 5,512,771 2,904,759 752,477 2,246,669 2,096,107 946,299 1,512,273 1,034,366 400,218 2,470,537 799,689 1,622,143 2,087,364 958,033 1,214,615 3,258,054 24,876,603 4,107,277 2,201,314 6,197,371 7,646,879 2,717,139 9,866,996 2,000,653 1,857,649 5,220,250 1,342,710 1,213,252 3,526,668 34,926,663 5,861,135 3,915,349 2,015,551 3,813,429 6,368,429 1,825,959 PKNR 1983 2005 424,872 1,825,803 867,345 3,023,796 586,944 1,862,800 569,212 1,998,060 299,629 1,506,026 399,429 1,408,666 802,142 4,014,079 1,008,679 4,888,329 388,995 1,046,041 250,341 815,867 787,386 3,430,601 1,348,882 3,135,047 523,996 3,713,662 845,945 3,193,413 863,002 3,367,001 638,457 1,921,967 187,130 542,461 532,728 1,441,889 829,685 1,840,156 312,101 637,787 537,380 901,996 207,518 692,543 170,881 283,739 510,532 1,825,826 273,992 551,533 280,263 1,334,066 758,337 1,412,041 239,141 853,040 271,827 747,728 551,557 2,309,667 4,920,395 17,751,762 868,079 2,593,299 716,152 1,468,780 2,264,170 3,196,789 1,040,362 4,286,020 499,910 1,985,540 2,121,432 6,871,262 660,865 1,582,951 973,895 1,120,060 1,020,453 3,625,406 472,655 945,050 360,958 883,513 743,921 2,429,082 4,979,612 23,942,998 821,142 4,057,558 547,693 2,775,697 472,142 1,387,599 775,098 2,480,161 1,645,878 4,622,732 290,827 1,259,599 Employ 1983 2005 116,930 312,950 238,530 503,630 186,250 328,600 184,600 361,780 102,730 204,080 158,880 266,550 285,280 636,930 341,850 820,680 74,780 97,650 50,100 62,200 265,050 443,880 361,580 451,400 220,700 504,450 201,300 496,100 196,130 451,030 161,380 261,400 56,980 70,100 116,680 170,880 190,800 207,030 52,580 77,600 105,330 156,250 52,780 74,450 29,750 40,050 136,530 247,580 74,680 76,530 98,280 175,100 130,380 207,700 66,150 84,900 38,600 96,450 152,900 276,680 1,355,080 2,627,930 178,200 359,730 130,530 193,280 158,600 356,100 363,950 813,580 143,100 266,550 618,850 1,171,250 178,280 282,780 130,480 175,250 399,530 520,330 179,830 154,830 177,180 138,250 340,330 441,400 1,449,530 3,067,550 267,350 619,280 162,050 290,850 83,450 149,900 217,850 340,780 353,430 543,380 86,050 150,080 HK 1983 2005 0.13 0.21 0.14 0.29 0.12 0.32 0.14 0.32 0.13 0.36 0.11 0.39 0.17 0.35 0.15 0.32 0.15 0.28 0.12 0.31 0.18 0.24 0.14 0.27 0.17 0.36 0.15 0.33 0.13 0.27 0.17 0.25 0.12 0.35 0.19 0.27 0.16 0.28 0.15 0.20 0.12 0.22 0.17 0.29 0.16 0.38 0.16 0.26 0.12 0.21 0.14 0.40 0.16 0.33 0.16 0.38 0.13 0.30 0.14 0.35 0.21 0.24 0.21 0.34 0.12 0.21 0.19 0.27 0.17 0.35 0.13 0.38 0.20 0.33 0.16 0.42 0.14 0.42 0.11 0.30 0.09 0.36 0.09 0.29 0.14 0.37 0.20 0.21 0.16 0.31 0.21 0.23 0.20 0.23 0.18 0.20 0.20 0.21 0.16 0.23 SK 1983 0.45 0.64 1.14 0.58 0.23 0.31 0.85 1.09 0.60 0.27 2.29 2.66 1.53 0.76 0.82 1.11 0.45 0.61 1.26 0.21 0.69 0.43 0.19 0.68 0.51 0.97 0.39 0.51 0.14 0.84 13.16 1.82 1.29 0.87 2.00 0,74 4.14 0.53 0.69 3.72 0.85 0.90 3.38 33.94 1.34 1.06 0.72 1.71 3.30 0.64 2005 9.60 7.84 7.86 8.50 2.87 2.83 1.53 25.76 3.44 1.85 30.28 13.17 37.29 19.34 16.79 11.32 2.53 10.26 5.96 2.16 5.99 2.85 1.83 12.68 0.98 6.12 4.08 2.77 7.82 10.26 219.97 26.06 9.34 17.80 49.50 20.11 77.51 4.78 3.90 22.94 4.36 2.87 18.70 259.10 31.92 14.11 8.05 18.51 25.36 9.10 Unem 1983 2005 15.20 9.16 28.60 17.65 18.90 14.78 23.05 12.91 21.88 15.88 14.88 15.92 21.20 11.66 29.20 13.86 10.10 6.89 5.20 4.68 16.25 5.76 13.73 10.24 13.88 7.21 21.38 12.83 17.83 10.50 12.70 8.51 11.85 8.82 15.90 6.73 10.90 10.82 15.23 7.50 15.60 9.09 9.70 6.80 9.75 5.16 18.28 9.23 9.85 10.00 15.30 10.00 19.10 10.58 8.35 6.49 21.00 7.05 10.55 9.12 24.60 6.97 9.68 7.33 5.48 5.89 15.48 7.06 19.10 9.61 12.00 7.31 18.40 8.59 18.40 17.51 13.55 12.98 11.58 9.86 6.20 6.67 7.53 10.49 10.93 11.02 17.55 6.80 16.68 8.01 15.80 5.65 14.20 7.09 19.35 5.66 22.03 8.45 11.13 6.18 GDP per capita in (e). PK and PKNR in thousands (e). HK is 0 - 1 fenced. SK is the volume of social capital services. CPI 2001 = 100. Unem is expressed in percentage. CPI 1983 49.43 49.08 45.72 45.03 46.96 46.16 52.44 49.40 48.33 47.11 53.81 45.67 48.18 44.45 40.56 49.25 52.56 50.78 45.45 48.44 51.17 46.46 54.68 46.15 48.52 47.04 45.53 52.74 48.09 48.63 45.93 43.09 42.94 45.37 52.09 51.38 50.50 48.76 46.04 46.45 47.23 46.52 43.68 48.13 48.81 42.01 45.39 47.28 47.61 44.11 2005 114.86 113.01 112.67 113.05 112.07 113.07 114.34 112.75 113.16 113.89 113.42 113.26 113.22 110.21 109.75 112.55 113.08 113.15 113.38 112.17 111.69 112.70 111.94 112.52 112.97 114.28 113.49 112.36 113.69 113.43 115.75 115.93 114.63 112.08 113.04 113.20 113.52 110.60 112.19 114.72 113.20 112.87 113.77 113.38 114.80 113.93 113.72 112.75 113.79 114.24 Table 2: Economic growth determinants 1983–2005 (Intercept) GDP83 GDPpc GDPpc GDPpc GDPpc GDPpc 2.503∗∗∗ (0.055) −0.262∗∗∗ (0.047) −3.560∗∗∗ (0.369) −0.090∗∗∗ (0.018) 0.425∗∗∗ (0.025) −3.182∗∗∗ (0.377) −0.095∗∗∗ (0.019) 0.316∗∗∗ (0.032) 0.223∗∗∗ (0.055) −1.843∗∗∗ (0.317) −0.044∗∗∗ (0.014) 0.256∗∗∗ (0.027) 0.187∗∗∗ (0.363) 0.213∗∗∗ (0.025) −0.648∗∗ (0.309) −0.058∗∗∗ (0.015) 0.204∗∗∗ (0.023) 0.090∗∗ (0.043) 0.193∗∗∗ (0.029) 0.054∗∗∗ (0.001) 1150 31.55∗∗∗ 0.13 1150 543.20∗∗∗ 0.77 1150 711.68∗∗∗ 0.79 1150 963.63∗∗∗ 0.84 1150 632.94∗∗∗ 0.85 χ = 1149.77∗∗∗ F = 231.24∗∗∗ F = 26.78∗∗∗ F = 25.31∗∗∗ PK Employ HK SK N F WhitinR2 Wald’s Heteroscedasticity Test Wooldridge’s Serial Autocorrelation Test Pesaran’s Spatial Autocorrelation Test Robust Hausman’s Test ∗ , ∗∗ and ∗∗∗ denote significance at 10%, 5%, and 1% significance levels, respectively. Standard errors reported in brackets. All test referred to the last not corrected regression. Estimations reported on the table are corrected by using Driscoll and Kraay standard errors. 27 Table 3: Social capital and investment 1983–2005 (Intercept) PK PKNR 4.609∗∗∗ 4.689∗∗∗ (0.696) −0.003 (0.012) 0.596∗∗∗ (0.161) (0.733) −0.008 (0.010) 0.583∗∗∗ (0.158) 0.691∗∗∗ (0.048) r lagGDP lagPK 0.076∗ (0.041) −0.004∗∗ (0.002) 0.002 (0.011) 0.074∗∗∗ (0.022) 0.676∗∗∗ (0.046) 0.104∗ (0.061) −0.002 (0.002) −0.001 (0.015) 0.090∗∗∗ (0.029) 1100 772.31∗∗∗ 0.88 χ = 259.03∗∗∗ F = 93.41∗∗∗ F = 21.18∗∗∗ F = 1.7e + 05∗∗∗ 1100 487.23∗∗∗ 0.86 χ = 643.42∗∗∗ F = 215.75∗∗∗ F = 26.60∗∗∗ F = 2.4e + 05∗∗∗ lagPKNR HK Unem CPI SK N F WhitinR2 Wald’s Heteroscedasticity Test Wooldridge’s Serial Autocorrelation Test Pesaran’s Spatial Autocorrelation Test Robust Hausman’s Test ∗ , ∗∗ and ∗∗∗ denote significance at 10%, 5%, and 1% significance levels, respectively. All test are referred to the not corrected model. Estimations reported on the table are corrected by using Driscoll and Kraay standard errors. 28 Table 4: Robustness analysis. Jackknife estimations (Intercept) GDP83 PK Employ HK GDPpc PK PKNR −0.648 (0.406) −0.058∗∗∗ (0.013) 0.204∗∗∗ (0.028) 0.090∗∗ (0.044) 0.193∗∗∗ (0.029) 4.609∗∗∗ 4.689∗∗∗ (0.412) r lagGDP lagPK (0.383) 0.076∗∗ (0.036) −0.008∗∗ (0.004) 0.583∗∗∗ (0.127) 0.690∗∗∗ (0.030) lagPKNR SK 0.054∗∗∗ (0.011) −0.004∗ (0.003) 0.002 (0.006) 0.074∗∗ (0.031) N F WhitinR2 1150 152.00∗∗∗ 0.85 1100 899.52∗∗∗ 0.88 Unem CPI ∗ , ∗∗ 0.104∗∗ (0.047) −0.003 (0.004) 0.596∗∗∗ (0.146) 0.676∗∗∗ (0.032) −0.002 (0.004) −0.001 (0.007) 0.090∗∗ (0.039) 1100 734.65∗∗∗ 0.86 and ∗∗∗ denote significance at 10%, 5%, and 1% significance levels, respectively. 29 Figure 1: Investment components by provinces. Mean values 1983 - 2005 A Coruña Alacant Albacete Almería Asturias Badajoz Barcelona Burgos Cantabria Castelló Ciudad Real Cuenca Cáceres Cádiz Córdoba Girona Granada Guadalajara Guipúzcoa Huelva Huesca Illes Balears Jaén La Rioja Las Palmas León Lleida Lugo Madrid Murcia Málaga Navarra Ourense Palencia Pontevedra Salamanca Segovia Sevilla Soria Sta. Cruz de Tenerife Tarragona Teruel Toledo Valladolid València Vizcaya Zamora Zaragoza Álava Ávila 0 5.000.000 10.000.000 Non−residential investment 30 Residential investment 15.000.000 Figure 2: GDP per capita and social capital 2.5 1.5 Log GDP pc 3.5 Scatterplot Spearman’s rho = 0.62 Prob > |t| = 0.0000 0 2 Log SK 4 6 Figure 3: Physical capital investment and social capital 11 13 Log PK 15 17 Scatterplot Spearman’s rho = 0.83 0 2 Log SK Prob > |t| = 0.0000 4 6 Figure 4: Physical capital investment (non-residential) and social capital 12 14 Log PKNR 16 18 Scatterplot Spearman’s rho = 0.83 0 2 Log SK 31 4 Prob > |t| = 0.0000 6 Figure 5: GDP per capita by provinces a) 1983 GUIPUZCOA VIZCAYA ASTURIAS A CORUÑA CANTABRIA LUGO ALAVA LEON PONTEVEDRA NAVARRA BURGOS HUESCA LA RIOJA OURENSE PALENCIA GIRONA ZAMORA VALLADOLID LLEIDA SORIA ZARAGOZA TARRAGONA SEGOVIA GUADALAJARA SALAMANCA AVILA TERUEL MADRID CASTELLON CUENCA TOLEDO CACERES BARCELONA VALENCIA ILLES BALEARS ALBACETE CIUDAD REAL BADAJOZ ALICANTE CORDOBA HUELVA MURCIA JAEN SEVILLA GRANADA ALMERIA + 10.800 9.200 − 10.800 7.300 − 9.200 5.000 − 7.300 MALAGA CADIZ ST. CRUZ DE TENERIFE LAS PALMAS b) 2005 VIZCAYA GUIPÚZCOA ASTURIAS A CORUÑA CANTABRIA LUGO ÁLAVA NAVARRA LEÓN PONTEVEDRA BURGOS LA RIOJA PALENCIA OURENSE HUESCA ZAMORA SORIA CÁCERES BARCELONA ZARAGOZA SEGOVIA TARRAGONA GUADALAJARA SALAMANCA ÁVILA GIRONA LLEIDA VALLADOLID TERUEL MADRID CASTELLÓ CUENCA TOLEDO VALÈNCIA ILLES BALEARS CIUDAD REAL ALBACETE BADAJOZ ALACANT CÓRDOBA JAÉN MURCIA HUELVA SEVILLA GRANADA ALMERÍA MÁLAGA CÁDIZ ST. CRUZ DE TENERIFE LAS PALMAS 32 + 18.300 14.800 − 18.300 13.300 − 14.800 10.500 − 13.300 Figure 6: Physical capital investment by provinces b) 1983 GUIPÚZCOA VIZCAYA ASTURIAS A CORUÑA CANTABRIA LUGO ÁLAVA NAVARRA LEÓN PONTEVEDRA BURGOS HUESCA LA RIOJA OURENSE GIRONA PALENCIA ZAMORA VALLADOLID LLEIDA SORIA BARCELONA ZARAGOZA SEGOVIA TARRAGONA GUADALAJARA SALAMANCA TERUEL ÁVILA MADRID CASTELLÓ CUENCA TOLEDO CÁCERES VALÈNCIA CIUDAD REAL ILLES BALEARS ALBACETE BADAJOZ ALACANT CÓRDOBA MURCIA JAÉN HUELVA SEVILLA GRANADA + 1.600.000 930.000 − 1.600.000 580.000 − 930.000 250.000 − 580.000 ALMERÍA MÁLAGA CÁDIZ ST. CRUZ DE TENERIFE LAS PALMAS b) 2005 VIZCAYA GUIPÚZCOA ASTURIAS A CORUÑA CANTABRIA LUGO ÁLAVA NAVARRA LEÓN PONTEVEDRA BURGOS HUESCA LA RIOJA OURENSE GIRONA PALENCIA ZAMORA LLEIDA VALLADOLID SORIA TARRAGONA SEGOVIA SALAMANCA GUADALAJARA ÁVILA CÁCERES BARCELONA ZARAGOZA TERUEL MADRID CASTELLÓ CUENCA TOLEDO VALÈNCIA ILLES BALEARS CIUDAD REAL ALBACETE BADAJOZ ALACANT CÓRDOBA JAÉN MURCIA HUELVA SEVILLA GRANADA MÁLAGA CÁDIZ ST. CRUZ DE TENERIFE LAS PALMAS 33 ALMERÍA + 4.700.000 2.600.000 − 4.700.000 1.800.000 − 2.600.000 400.000 − 1.800.000 Figure 7: Social capital by provinces a) 1983 VIZCAYA ASTURIAS A CORUÑA GUIPÚZCOA CANTABRIA LUGO ÁLAVA NAVARRA LEÓN BURGOS PONTEVEDRA OURENSE HUESCA LA RIOJA PALENCIA GIRONA ZAMORA VALLADOLID LLEIDA SORIA ZARAGOZA TARRAGONA SEGOVIA SALAMANCA GUADALAJARA ÁVILA TERUEL MADRID CASTELLÓ CUENCA TOLEDO CÁCERES BARCELONA VALÈNCIA ILLES BALEARS ALBACETE CIUDAD REAL BADAJOZ ALACANT CÓRDOBA HUELVA MURCIA JAÉN + 1.77 0.94 − 1.77 0.69 −0.94 0.48 −0.69 0.14 −0 .48 SEVILLA GRANADA ALMERÍA MÁLAGA CÁDIZ ST. CRUZ DE TENERIFE LAS PALMAS b) 2005 VIZCAYA ASTURIAS A CORUÑA GUIPÚZCOA CANTABRIA LUGO ÁVILA NAVARRA LEÓN BURGOS PONTEVEDRA OURENSE PALENCIA HUESCA LA RIOJA GIRONA ZAMORA VALLADOLID LLEIDA SORIA ZARAGOZA TARRAGONA SEGOVIA SALAMANCA GUADALAJARA ÁVILA CÁCERES BARCELONA TERUEL MADRID CASTELLÓ CUENCA TOLEDO VALÈNCIA ILLES BALEARS CIUDAD REAL ALBACETE BADAJOZ ALACANT CÓRDOBA MURCIA JAÉN HUELVA SEVILLA GRANADA MÁLAGA CÁDIZ ST. CRUZ DE TENERIFE LAS PALMAS 34 ALMERÍA + 24.15 12.00 − 24.15 7.83 − 12.00 2.87 − 7.83 0.98 − 2.87