1. Introduction
Financial development is considered an essential pillar in long-run economic growth (King and Levine, Reference King and Levine1993). As Madsen and Ang (Reference Madsen and Ang2016) argued, it facilitates the accumulation of productive factors, new technologies and knowledge. According to Beck et al. (Reference Beck, Demirgüç-Kunt and Maksimovic2008), access to financial resources is especially important for small and medium-sized firms, a fact heavily conditioned by the existence of a high-quality formal institutional framework. In that regard, formal institutions can be understood as those rules and policies established by the states, providing a framework of guarantees that should encourage people to participate in the economy. We consider as a proxy for formal institutions the strengh of the economic-judicial institutions, closely related to the protection of property rights. However, informal institutions, more related to culture and social capital, are important too, and they have received a great deal of attention in recent decades. A common proxy to measure such institutions is the level of social trust (henceforth simply trust), which is the focus of this paper.
Starting with the pioneering work of Putnam (Reference Putnam1993) on the Italian regions, today there is wide agreement on the positive influence of social features such as trust in the economic sphere (Bjørnskov and Méon, Reference Bjørnskov and Méon2015; Forte et al., Reference Forte, Peiró-Palomino and Tortosa-Ausina2015; Peiró-Palomino and Tortosa-Ausina, Reference Peiró-Palomino and Tortosa-Ausina2013; Zak and Knack, Reference Zak and Knack2001). These effects are extensible to the case of financial development in a broad sense (see Guiso et al., Reference Guiso, Sapienza and Zingales2004, Reference Guiso, Sapienza and Zingales2008b; Ng et al., Reference Ng, Ibrahim and Mirakhor2016). In particular areas of finance such as the credit market, however, evidence is still scant, with a few exceptions such as Wu et al. (Reference Wu, Firth and Rui2014). In addition, most of the trust–finance studies are conducted from the micro-level perspective, considering particular countries or regions. Whereas studies based on specific contexts are useful in shedding light on the role of trust in these particular areas with specific conditions, conclusions can hardly be generalised, since it is difficult to explain the role trust plays, if any, in accounting for the large cross-country disparities in the amount of credit granted to the private sector. Very few papers adopt a transnational approach, and most are based on very limited samples (see, for instance Guiso et al., Reference Guiso, Sapienza and Zingales2008b; Jordaan et al., Reference Jordaan, Dima and Golet2016), and are silent on the particular trust–private credit nexus. In contrast, this paper tackles that issue in detail, and from a cross-country perspective.
Moreover, the causal direction between trust, formal institutions and finance remains largely unexplored, and lack of consensus is widespread. One branch of the trust literature identifies a substitutive relationship, positing the idea that where formal institutions are weak trust is more important to operate in the market. In this vein, Knack and Keefer (Reference Knack and Keefer1997) suggested that trust may be more important for economic growth in poor economies with weaker institutional frameworks, thus suggesting a substitutive relationship. Ahlerup et al. (Reference Ahlerup, Olsson and Yanagizawa2009) corroborated these arguments for a limited sample of 46 countries. Similar arguments were put forward by Guiso et al. (Reference Guiso, Sapienza and Zingales2004) for financial development, suggesting that trust is especially relevant to enforce financial contracts where the legal system cannot guarantee them. The literature is far from conclusive, since more recent studies by McCannon et al. (Reference McCannon, Asaad and Wilson2018) found that contracts and trust are complementary. In this line, Bjørnskov and Méon (Reference Bjørnskov and Méon2013, Reference Bjørnskov and Méon2015) concluded that the observed positive effects of trust on productivity and growth are channelled via better institutional frameworks, particularly through the quality of economic-judicial institutions. These findings would indicate that trust and formal institutions are actually not substitutes.
Building on these arguments, this paper attempts to make a twofold contribution. First, it explores the relationship between trust and private credit for a sample of 119 countries in the period 1993–2015. To our knowledge, this is one of the largest samples in this context, which includes both developing and developed economies. This last point is important, since the variability in both trust levels and credit ratios is increased. As Beugelsdijk et al. (Reference Beugelsdijk, De Groot and Van Schaik2004) and Berggren et al. (Reference Berggren, Elinder and Jordahl2008) suggested, the effect of trust is very sensitive to the sample composition, and increased heterogeneity in the sample reinforces the robustness of the estimates. Second, the paper examines in depth the transmission mechanisms, focusing on the relationship between informal and formal institutions, and analysing whether there is a substitutive effect or if the effects from trust to credit are actually channelled through an improved formal institutional framework, proxied by a nine-dimensional composite measure capturing the quality of the economic-judicial institutions.
In doing so, we run OLS and also 2SLS and 3SLS regressions to control for endogeneity. We provide disaggregated results for the nine components of the economic-judicial quality indicator, which allows us to identify more specific transmission mechanisms. The results show that trust positively affects the volume of private credit. We also find that the impact is transmitted via the quality of economic-judicial institutions – although, only some of the nine indicators are actually significant channels. Moreover, neither evidence of substitutive effects between informal and formal institutions nor uneven effects of trust at different development levels are found. Overall, our results suggest that a reliable economic-judicial system is the main institution responsible for guaranteeing credit transactions, somehow questioning the direct positive effects on economic outcomes attributed to trust by several papers in the literature. In that regard, the paper furthers our understanding of how trust actually operates.
The reminder of the paper is structured as follows. Section 2 provides a review of the literature. Section 3 describes the empirical framework. The results are reported in Section 4 and, finally, Section 5 concludes.
2. Literature review
Links between trust and finance
The underpinnings of the links between trust and the development of the credit market are similar to those in other areas where trust plays a significant role. The expected benefits of trust are associated with the ideas of cooperation and the reduction of transaction costs. Gambetta (Reference Gambetta and Gambetta2000) defined the concept as the expectation that somebody will act in a way beneficial or, at least, a way that is not detrimental to us, encouraging us to cooperate. Previously, Arrow (Reference Arrow1972) argued that trust is an intrinsic element in economic transactions and a lack of trust may explain in a large extent economic backwardness in some parts of the world. Similarly, Coleman (Reference Coleman1990) highlighted the role of trust in limiting opportunistic behaviours in transactions, and Knack and Keefer (Reference Knack and Keefer1997) concluded that cooperative behaviours reduce moral hazard and agency problems as well as monitoring costs.
Most of the theoretical arguments in the trust–finance nexus relate to the existence of information asymmetries between parties in financial transactions, as financial agreements need adequate levels of transparency (Porta et al. Reference Porta, Lopez-de Silanes and Shleifer2006). As in low-trust countries information asymmetries are more likely to arise, and we may expect credit to be negatively affected. As argued by Ferrary (Reference Ferrary2003), when agents such as banks grant credit, they take the risk that the borrowers will not return the amount granted. Even though financial institutions have methods of risk evaluation that reduce information asymmetries, they rely on trust levels too, as the quality of the gathered information and the correct evaluation of the risk will depend, to some extent, on the existing level of trust where the transaction takes place. Accordingly, we might expect credit entities in high-trust environments to be more prone to lend financial resources and with more favourable conditions, as they will perceive these transactions as less risky (see, for instance, Hasan et al. Reference Hasan, Hoi, Wu and Zhang2017).
Demand is also sensitive to trust levels. Guiso et al. (Reference Guiso, Sapienza and Zingales2004) developed a model in which an investor's demand for stock depends not only on the expected return, but also on the risk aversion of the investor. This is associated, among other aspects, with the probability that the counterpart of the transaction will behave honestly. As Guiso et al. (Reference Guiso, Sapienza and Zingales2008b) concluded, in low-trust environments people are more likely to be suspicious of the information provided, which may prevent their participation in the financial markets. In that respect, Moro and Fink (Reference Moro and Fink2013) concluded that trust facilitates access to more soft information, thus reducing information asymmetries and facilitating financial transactions.
Links between trust and institutional quality
Much less is known on the transmission mechanisms of the trust effects. As commented on in the Introduction and having support in the related literature, the quality of the formal institutions postulates as a good candidate. Boix and Posner (Reference Boix and Posner1998) and Knack (Reference Knack2002) summarised some of the links between informal institutions and government performance. Following their arguments, trust makes citizens better informed and facilitates the articulation of their demands. Also, as more people participate in the electoral system, politicians will be quickly replaced if they do not fulfil their promises. Moreover, government members in a more trustworthy country are more likely to be trustworthy, and will avoid taking advantage of their position for personal benefits (Bjørnskov and Méon, Reference Bjørnskov and Méon2013).
On the government side, Boix and Posner (Reference Boix and Posner1998) argued that the implementation costs of policies and regulations are lower where trust is high, as there is no need to create and maintain complex systems of supervision. This may happen because trust shapes the expectations of citizens about the behaviour of their peers.Footnote 1 In addition, there are government decisions that may be beneficial in the long term but imply sacrifices in the short term. Such decisions are likely to be better accepted by citizens in high-trust countries, as they rely on government criteria and the promise of future benefits of these policies.
Trust also facilitates cooperation between political elites from different government levels, allowing them to work together efficiently and not pass difficulties to other colleagues. Similarly, where trust and norms of reciprocity are strong, opposing sides are more likely to resolve their differences (Knack, Reference Knack2002) and reforms are more easily implemented. For instance, Heinemann and Tanz (Reference Heinemann and Tanz2008) found that high-trust countries are better at introducing reforms strengthening economic freedom. In a similar vein, Berggren et al. (Reference Berggren, Sven-Olov and Hellström2016) concluded that trust has a crucial influence in the speed of implementation of reforms, and Berggren and Bjørnskov (Reference Berggren and Bjørnskov2017) found that trust facilitates the introduction of positive political and legal reforms, but it acts as a deterrent to the negative ones.
Empirical evidence
Empirical studies widely support these arguments. Regarding the trust–finance nexus, Calderon et al. (Reference Calderon, Chong and Galindo2002) reported positive links between trust and financial efficiency and other indicators such as bank assets, private credit by banks, bank overhead costs and the net interest margin. Georgarakos and Pasini (Reference Georgarakos and Pasini2011) and Ng et al. (Reference Ng, Ibrahim and Mirakhor2016) found that trust is a determinant of the stock market. Afandi and Habibov (Reference Afandi and Habibov2016) concluded that trust is positively associated with households’ use of banking services in a sample of transition economies, especially where levels of education are low and in countries with low levels of legal enforcement. More recently, Blau (Reference Blau2017) found that trust is an important determinant of financial market liquidity.
At the firm level, Moro and Fink (Reference Moro and Fink2013) found a positive relationship between trust, the amount of short-term credit granted and SMEs’ risk of being credit constrained. Wu et al. (Reference Wu, Firth and Rui2014) assessed the link between trust and the provision of trade credit, concluding that trust helps private firms to overcome institutional difficulties to finance their activities. Recently, Li et al. (Reference Li, Wang and Wang2017) found that firms headquartered in regions of high trust tend to have less crash risk.
Related evidence can be found in the context of microfinance programmes in developing countries. Not only does credit flow better where trust abounds, but trust is also positively associated with higher repayment rates (Karlan, Reference Karlan2007; Sharma and Zeller, Reference Sharma and Zeller1997; Van Bastelaer and Leathers, Reference Van Bastelaer and Leathers2006). Although cross-country studies are scant, there are many examples in particular geographical settings. Apart from the well-known study by Guiso et al. (Reference Guiso, Sapienza and Zingales2004) for the Italian case, finding that trust is associated with greater use of financial instruments such as checks or formal credit, others as Mwangi and Ouma (Reference Mwangi and Ouma2012) report similar evidence in the cases of Kenya, Talavera et al. (Reference Talavera, Xiong and Xiong2012) in China, and Heikkilä et al. (Reference Heikkilä, Kalmi and Ruuskanen2016) in Uganda, where trust is especially relevant to access to credit for poor people in rural areas.
More limited evidence is available on the transmission channels, giving rise to one of the objectives of this paper, i.e. discovering whether there is a causal relationship between trust and formal institutional quality, and whether these effects are eventually transferred to the amount of credit granted. As previously argued, some authors have dealt with this issue, but in related fields. Knack and Keefer (Reference Knack and Keefer1997), Ahlerup et al. (Reference Ahlerup, Olsson and Yanagizawa2009) and Guiso et al. (Reference Guiso, Sapienza and Zingales2004) pointed to a substitutive relationship between trust and institutional quality. Others, such as Berggren and Jordahl (Reference Berggren and Jordahl2006), questioned the direction of causality. They suggested that the security provided by a reliable regulatory framework fosters the generation of trust in people's interactions. By contrast, more recent papers advocate the stability of trust (see, for instance, Nunn and Wantchekon, Reference Nunn and Wantchekon2011; Uslaner, Reference Uslaner2008). In this line, other contributions including Bjørnskov and Méon (Reference Bjørnskov and Méon2013, Reference Bjørnskov and Méon2015) supported the idea that institutional quality is actually a transmission mechanism of the positive externalities of trust.
3. Empirical framework
Sample and data description
Our sample comprises 119 countries (listed in Table A.1, Appendix A) and covers the period 1993–2015. As a measure of credit, we take domestic credit to the private sector as a share of GDP, provided by the World Bank. The trust indicator corresponds to the percentage of respondents who claim that most people can be trusted when asked the widely supported ‘generally speaking question’. The data are taken from Bjørnskov and Méon (Reference Bjørnskov and Méon2013), who collected trust scores for a wide sample of countries considering various surveys and barometers for different years over the period 1981–2010 (World Values Survey, LatinoBarometro, Asian and East Asian Barometers, AfroBarometer and the Danish Social Capital Project).
Although the trust measure we use is the most common in the literature, it is not completely free from criticism (see for instance Delhey et al., Reference Delhey, Newton and Welzel2011). However, recent studies such as Johnson and Mislin (Reference Johnson and Mislin2012) and Aksoy et al. (Reference Aksoy, Harwell, Kovaliukaite and Eckel2018) have used the traditional trust question, finding that responses are highly correlated with experimental trust. This reinforces the consensus on using that survey indicator to measure trust in macro-level analyses. As control variables, we selected several indicators common in the related literature (see, among others, Djankov et al. Reference Djankov and Shleifer2007; Guiso et al. Reference Guiso, Sapienza and Zingales2004). We include variables reflecting the macroeconomic stability of the country, such as GDP pc growth and the inflation rate. We also account for educational levels by considering total years of schooling of the population over 25, provided by the Barro–Lee dataset, and for the level of development of the country, measured by GDP per capita. In addition, we control for credit market conditions with two variables, namely credit information and credit rights. The former measures rules affecting the scope, accessibility and quality of credit information available through public or private credit registries. The latter accounts for the degree to which collateral and bankruptcy laws protect the rights of borrowers and lenders.
Finally, we include in the models the quality of both political and economic-judicial institutions, as in Bjørnskov and Méon (Reference Bjørnskov and Méon2015). Following these authors, political institutions are proxied by the polity2 indicator provided by the Polity IV Dataset, which is a measure of the degree of autocracy/democracy. The quality of the economic-judicial system is measured using the property rights and legal quality indicator, which is one of the dimensions of the Economic Freedom Index by the Fraser Institute. In particular, that dimension subsumes nine different indicators, namely (1) judicial independence; (2) impartiality of the courts; (3) protection of property rights; iv) military interference in rule of law and politics; (5) integrity of the legal system; (6) legal enforcement of contracts; (7) regulatory restrictions on the sale of properties; (8) reliability of police; and (9) business costs of crime. As a robustness proof, the indicators of rule of law and control of corruption, in this case provided by the World Bank Governance Indicators, are used. Following Berggren et al. (Reference Berggren, Bergh and Bjørnskov2012), we also account for the stability of the institutional measures over time, which was computed as the coefficient of variation of each of the institutional quality variables considered.
Finally, we selected some variables to be used as instruments in order to control for the potential endogeneity of trust. These are the minimum temperature in the coldest month of the year and the pronoun-drop characteristic in the predominant language in the country. These two variables are widely supported in the trust literature and have been considered by a long list of scholars including, but not restricted to, Guiso et al. (Reference Guiso, Sapienza and Zingales2008a), Tabellini (Reference Tabellini2008a) and Bjørnskov and Méon (Reference Bjørnskov and Méon2013, Reference Bjørnskov and Méon2015). Languages that drop the personal pronoun are associated with cultures with less respect for individual rights and characterised by distrust. In the case of climatic features, in places with cold winters people are more collaborative and trustful, a character trait rooted in times long past when collaboration and trust were essential for survival (Bjørnskov and Méon Reference Bjørnskov and Méon2013, Reference Bjørnskov and Méon2015). As a robustness check, we use as an alternative instrument the degree of ethnic fractionalisation, taking data from Alesina et al. (Reference Alesina, Devleeschauwer, Easterly, Kurlat and Wacziarg2003). In this respect, contributions such as Alesina and La Ferrara (Reference Alesina and La Ferrara2002) and Zerfu et al. (Reference Zerfu, Zikhali and Kabenga2009), among others, pointed to a negative association between ethnic diversity and trust.
Table B.1 of Appendix B gives a short description of the variables used and their sources. All figures correspond to the average of all the available information over the period 1993–2015. Following recent arguments in the literature, trust is considered time-invariant. Extensive evidence for this consideration can be found in Tabellini (Reference Tabellini2008b), Uslaner (Reference Uslaner2008), Nunn and Wantchekon (Reference Nunn and Wantchekon2011) and Bjørnskov and Méon (Reference Bjørnskov and Méon2015), to name a few.Footnote 2 Therefore, for countries with information for only one year in our period we considered that single figure, while for those with available information for several years (either from different surveys or different waves of the same survey) we considered the average (see Bjørnskov and Méon, Reference Bjørnskov and Méon2013).
Table 1 reports summary statistics for all the variables of interest. In general, the variability is high for all them. In the case of trust, our sample includes countries such as Cape Verde and Brazil, where less than 6% of the population trust each other, which contrasts with Norway and Denmark, where trust levels are above 65%. According to Nunn and Wantchekon (Reference Nunn and Wantchekon2011), individuals whose ancestors were victims of the slave trade are less trusting today. These findings complement those by Uslaner (Reference Uslaner2008), suggesting that trust levels are inherited from one generation to the next and current generations have trust levels that are more similar to those of their grandparents than those in area where they were raised. This might partly explain low trust levels in countries populated by a larger proportion of slave descendents.
Table 1. Descriptive statistics
There are also notable disparities in the credit ratios, our dependent variable, whose standard deviation (41.76%) is almost as large as its mean value (49.85%). Table 1 also lists information for the variables used in the robustness checks and for the instruments. Figure 1(a) reports the scatter plot for trust and credit, showing a clear positive correlation. Figure 1(b) displays notched boxplots for credit in countries with trust levels below and above the median, revealing a large difference in favour of the second group. The median of credit, represented by the horizontal line inside the box, is almost double in high-trust economies. The notches around the median represent 95% confidence intervals for the median. As the intervals for the two groups do not overlap, we can conclude that they differ statistically. In addition, the group of low-trust countries presents much lower variability in credit, but there are three outlying observations, corresponding to Cyprus, Malaysia and Portugal. The distribution in both groups, however, is relatively normal.
Figure 1. Social trust and credit
Models and econometric approach
We estimate different model specifications, which differ in their explanatory variables. The dependent variable is in all cases the ratio of credit over GDP, as defined in the previous section. The explanatory variables include trust, considered as a single regressor in Model 1 (M1), and the control variables, which are added sequentially in Models 2–6 (M2–M6). Finally, the most comprehensive Models 7, 8 and 9 (M7, M8 and M9) incorporate interaction terms between trust, legal quality and GDP per capita to assess the existence of substitutive or complementary effects.
The models are estimated via OLS, although two-stage (2SLS) and three-stage (3SLS) instrumental variable estimations are also performed in order to provide results robust to endogeneity and to analyse the transmission mechanisms from trust to credit through the quality of the economic-judicial institutions (legal quality). In both 2SLS and 3SLS estimations, trust is first regressed on a vector of predetermined instruments. Formally, the more complex 3SLS system is defined as follows:
$$Social\; trust_i = \delta + {\bi \theta} {\bf V}_{\bf i} + \phi _i$$
$$Legal{\rm \;} quality_i = \lambda + \rho Social{\rm \;} trust_i + \omega _i$$
$$Credit/GDP_i = \alpha + \beta Legal{\rm \;} quality_i + {\bi \gamma} {\bf C}_{\bf i} + \varepsilon _i$$
where i represents a particular country; 
${\bf V}$ and 
${\bf C}$ are vectors of instruments and control variables, respectively; δ, λ and α are the constant parameters; θ, and γ are vectors of parameters for the instruments and the controls, respectively; and ρ and β are the parameters associated with the variables of interest. Finally, ϕ, ω and ε are the disturbances at stages I, II and III, respectively. The fitted values for trust from the first regression are included in the second regression. Analogously, the fitted values of legal quality at stage II are included at stage III.
4 Results
Informal, formal institutions and credit: OLS estimates
Table 2 reports the results from the OLS regressions. The estimated coefficient for trust in Model 1 is positive and significant. In particular, an increase in a sample standard deviation of trust would increase credit GDP ratio by about 18 points, ceteris paribus. This coefficient logically diminishes when control variables are included in the analysis, although it remains relatively large. Model 2 includes macroeconomic control variables such as GDP per capita growth and inflation level, as well as relevant variables for credit market quality and transparency, such as credit information and credit rights. Of these controls only the inflation rate and credit information are significant, showing the expected negative and positive effect, respectively. The sign and significance of trust remain unaltered, as they do in Models 3 and 4, where political quality and education are added.
Table 2. Informal, formal institutions and credit: OLS estimates
Notes: Dependent variable is credit/GDP.
Robust standard errors are in parentheses.
10%, 5% and 1% significance levels are denoted by *, ** and *** respectively.
However, when the quality of economic-judicial institutions (legal quality) is included in Model 5, the coefficient of trust turns non-significant. This is in line with previous findings by Bjørnskov and Méon (Reference Bjørnskov and Méon2013), showing that once the quality of certain types of formal institutions is accounted for, the effects of trust on economic performance (mostly measures of productivity and economic growth) are diluted, suggesting that the effect of trust is transmitted via institutional quality. This result also seems to be consistent for the case of credit. In Model 6 we control for the level of development by means of GDP per capita, which is found to be positive and significant, while in Model 7 we test whether the effect of trust depends on the level of development by including the interaction term between trust and GDP per capita. The results show a negative interaction term, only weakly significant (10% level) on average. More comprehensive results are displayed in Figure 2(a), showing that the interaction is indeed non-significant across different levels of development (the 95% confidence bands overlap with the zero in all cases). Therefore, unlike what is found in the case of trust and growth (see Ahlerup et al., Reference Ahlerup, Olsson and Yanagizawa2009; Knack and Keefer, Reference Knack and Keefer1997), uneven impacts of trust are not found at different development levels in the case of credit. As most of the effects of trust are indirect, it is difficult to ascertain the reasons, although we can still suggest that the uneven impacts of trust on growth for low and high income economies found in the previous literature are rooted in transmission mechanisms other than credit.
Figure 2. Analysis of the interaction effects
In Model 8, we interact trust and institutional quality in the attempt to identify the existence of substitutive effects. A negative and significant sign would suggest a more prominent role of trust where institutional quality is low, indicating substitutability. However, the estimation yields a non-significant coefficient, both on average and across the entire range of values of legal quality, as shown in Figure 2(b). These results are novel in the field of finance but can again be compared with other analyses from the field of development and growth, finding a substitutive effect between trust and formal institutional quality. We suggest that this theory is not supported when trying to explain the role of trust in granting credit. These results hold in the most comprehensive Model 9, which incorporates GDP per capita.
Given the novelty of the findings, a completely satisfactory explanation is admittedly challenging. Though, we can still provide a potential explanation linking our findings to those of Williams and Vorley (Reference Williams and Vorley2015). They argue that whereas formal institutions can change rapidly, informal institutions are highly persistent over time. For example, several transition economies in Eastern Europe have witnessed remarkable improvements in their formal institutions, but the informal ones have not evolved at the same pace. In these scenarios, which may well be present in other transition countries all over the world, institutional misalignments emerge. An improved formal framework can guarantee economic transactions, but the well-ingrained cultural values defining the informal one might be much less prone to encourage entrepreneurial actions or engage in financial contracts, being unable to substitute, per se, the formal rules. In addition, as far as these informal rules have an impact on the formal ones, these scenarios constrain further improvements in the formal institutional setting.
Dealing with endogeneity and exploring the transmission mechanisms: 2SLS and 3SLS estimates
Trust is remarkably stable in time (Bjørnskov, Reference Bjørnskov2007) and direct reverse causality from credit to trust is unlikely. In any case, other mechanisms and indirect relationships may lead to endogeneity. In this section we provide results from 2SLS regressions, where trust is instrumented at the first stage with exogenous regressors. In particular, and as we commented on in Section 3, we use as instruments the minimum temperature in the coldest month of the year and a dummy variable that equals one if the predominant language in the country keeps the personal pronoun. Table 3 shows the results for 2SLS regressions for Models 1–6. The results hold in all cases. In fact, the coefficient for trust is larger in the instrumented models but, analogously to the baseline OLS estimations, is only significant in those where legal quality is not included. In all cases both the weak instruments F-test and the Sargan test provided at the bottom of the table suggest the validity of the instruments.
Table 3. Informal, formal institutions and credit: 2SLS estimates
Notes: Dependent variable is credit/GDP. Trust is treated as endogenous and is instrumented by pronoun-drop characteristic and minimum temperature. Robust standard errors are in parentheses. 10%, 5% and 1% significance levels are denoted by *, ** and *** respectively.
According to the literature and our previous results, the effects of trust on credit are likely to be channelled through the quality of economic-judicial institutions, which is analysed by means of 3SLS estimations. As in the 2SLS setting, at the first stage trust is instrumented with the minimum temperature and the pronoun-drop characteristic. At the second stage, legal quality is regressed on trust and finally, these fitted values are introduced in the original Model 6.
Table 4 contains the results for the three stages. For the sake of clarity we only include results for the variables of interest, but all the control variables of Model 6 in Tables 2 and 3 are included at the third stage. At stage I, the instruments have the expected signs. The effect of the minimum temperature is negative and that for the pronoun-drop feature is positive. The results for the second stage show that trust, exogenously determined at stage I, has a positive influence on legal quality. In particular, the model predicts that one sample standard deviation increase in trust yields an improvement of 1.89 points in legal quality, ceteris paribus. Considering that this variable is measured on a 0–10 scale, the impact is relatively important. The effect is ultimately transmitted to credit at stage III. According to the estimated coefficients, one sample standard deviation increase in legal quality yields, ceteris paribus, a predicted 14.30 average increase in the credit/GDP ratio. Again, this is a non-negligible impact.
Table 4. Informal, formal institutions and credit: 3SLS estimates
Notes: The 3SLS estimation includes but does not report all the control variables of Model 6. Complete results are available upon request. Robust standard errors are in parentheses. 10%, 5% and 1% significance levels are denoted by *, ** and *** respectively.
Finally, the index of legal quality subsumes nine different indicators. In order to better identify more specific transmission mechanisms, Table 5 reports disaggregated results for each of them. In all cases results refer to the third stage of a 3SLS estimation where the composite legal quality variable is replaced by its more specific components. This additional exercise reveals that, interestingly, not all the indicators are transmission channels. In particular, judicial independence, impartial courts, the protection of property rights, the integrity of the legal system, the reliability of police and the business costs of crime are significant factors, being the largest coefficient for the protection of property rights. However, indicators of military interference in the rule of law and politics, the legal enforcement of contracts and regulatory restrictions on the sale of real property are non-significant factors.
Table 5. Informal, formal institutions and credit: 3SLS estimates for individual indicators of the composite legal quality index
Notes: Dependent variable is credit/GDP. Results correspond to the third stage of the 3SLS estimation, which includes but does not report all the control variables of Model 6. Complete results are available upon request. Robust standard errors are in parentheses. 10%, 5% and 1% significance levels are denoted by *, ** and *** respectively.
Robustness tests
Some additional tests were performed as robustness checks, all of them provided in Appendix C. First, we used rule of law and control of corruption as alternative indicators of institutional quality, taken from the World Bank Governance Indicators (WGI). Table C.1 reports both OLS and 3SLS results, and they are fairly similar to those obtained with the original variable of legal quality. The similarity in the results suggests that these measures capture similar or even overlapping institutional conditions.Footnote 3
Second, we used an alternative instrument, namely ethnic fractionalisation. Table C.2 reports 3SLS estimations. Ethnic fractionalisation has a negative coefficient in the first stage, showing a negative association with trust. For stages II and III, we find qualitatively analogous results with those in the original analysis, suggesting that results are robust to changes in the instrument.
Third, we argue throughout the paper that trust is stable over time. However, there are some countries showing particular tendencies. For instance, trust declined in the USA and improved in Denmark over the analysed period. We then removed these two countries. In addition, for several countries only one observation for trust is available over the period. These cases were removed too. The final sample has 94 observations. We tested whether the results held in three scenarios: OLS Model 4, OLS Model 6 and the 3SLS estimation. The results, provided in the first three columns of Table C.3, hold in all cases, suggesting that coefficients are not driven by the removed observations.
Fourth, we tested whether the results for trust hold when the original sample is split into two time periods. Our period of analysis is characterised by years of economic expansion until 2007 and a subsequent period of crisis. As credit markets were deeply affected by the crisis in some countries in the sample, we perform the analysis for the two periods separately. Several arguments suggest that trust might be even more important for credit during financial crisis periods, which experience a general tightening of credit conditions in most countries. Bergh and Bjørnskov (Reference Bergh and Bjørnskovforthcoming) show that in high-trust economies, long-term market interest rates and credit ratings are less sensitive to bad performance of macroeconomic fundamentals such as inflation or growth. Results are also reported in Table C.3 (columns four to nine). OLS results for Model 4 show that the effect of trust holds in both pre-crisis and crisis periods. In fact, the trust coefficient is slightly larger in the latter. Analogously, the trust impact vanishes in Model 6, when legal quality is included in the model. Finally, 3SLS estimates confirm that legal quality acts as a transmission mechanism. The estimated 3SLS coefficient is slightly greater in the crisis period, which provides some preliminary evidence in favour of a more prominent effect of trust in difficult times.
5 Concluding remarks and prospects for future research
Insofar as the private credit to GDP ratio is a macroeconomic fundamental that can explain development gaps across countries, it is important to provide answers to some questions that remain unresolved, i.e. whether trust as a measure of informal institutions could explain cross-country credit disparities and, if so, which the mechanisms are. In fact, it seems that most of the beneficial effects of trust in the macroeconomic sphere are indirect. Whereas previous contributions focused on innovation, investment, education or productivity levels, this paper has provided some evidence for the credit market case by analysing a sample of 119 countries during the period 1993–2015.
For several model specifications, the results suggest that trust explains cross-country disparities in private credit. The results also indicate that the trust benefits on credit are transmitted indirectly via the quality of economic-judicial institutions, measured by a composite index elaborated by the Fraser Institute. As a remarkable novelty, the paper provides disaggregated results for the particular components of the measure, showing that only some of them are indeed actual transmission mechanisms. In addition, the data provide no evidence of substitutive effects between informal and formal institutions, and no significant differences can be attributed to development differentials. The results hold when a variety of robustness tests are performed.
These findings have a series of implications. First, as trust improves economic-judicial institutions, it may be tentatively argued that trust should be fostered in countries where it is low. Unfortunately, it is well-supported by the literature that trust, like other cultural values, is remarkably time-invariant (see Guiso et al., Reference Guiso, Sapienza and Zingales2016), and is passed on from one generation to the next. Does this imply that today's differences in private credit are explained by historical facts that determined trust levels? Only in part, since institutional quality does not depend only on trust, but also on other factors. In fact, Bjørnskov (Reference Bjørnskov2010) shows that institutional quality has undergone great changes in some countries in recent periods and, bearing in mind that trust has remained stable, this would indicate that there is also room for improvement in low-trust countries. Good examples are the transition countries from Eastern Europe. However, economies with higher trust levels set out from a more favourable scenario, a fact that should be kept in mind in understanding why two countries in similar settings but differing in trust levels can show credit disparities.
In our view, there are three main research avenues where future contributions will be especially welcome. First, more work is needed to identify clear transmission mechanisms. We made some progress in this paper, providing results for individual indicators of the legal quality index. They showed that the protection of property rights has the largest effect on credit, although having impartial courts and reliable police forces are important conditions too. Second, more evidence is needed to explain why informal institutions such as trust cannot be a substitute for formal rules. On that point, we could only provide some intuition based on the possible misalignment between formal and informal institutions, but more solid arguments are necessary. Third, it would be interesting to use other trust measures that could provide additional insights. We hope our findings will encourage future studies on these issues.
In any case, the most interesting lesson from this paper is that, in line with other macroeconomic spheres, most of the positive externalities of trust are indirect. In addition, it is a compelling necessity to improve the quality of particular forms of legal institutions in order to spur the provision of credit in those places where the credit flow is low. Without a formal and reliable economic-judicial framework, credit transactions are constrained.
Acknowledgements
Authors acknowledge the financial support of the Spanish Ministry of Economy and Competitiveness (ECO2017-84858-R and ECO2017-85746-P). Paula Cruz-García also acknowledges the financial support of the Spanish Ministry of Education (FPU2014/00936). Jesús Peiró-Palomino is also grateful to the University Jaume I (UJI-B2017-33). The insightful comments from three anonymous referees are also acknowledged. All the remaining errors are ours.
Appendix A. Sample and main variables
Table A.1. Countries, credit and institutions
Appendix B. Variable description and data sources
Table B.1. Variables and sources
Appendix C. Results of the robustness tests
Table C.1. Robustness tests, alternative measures for formal institutions
Table C.2. Robustness tests, alternative instrument
Table C.3. Robustness tests, alternative sample and temporal periods
