Do National Differences in Social Capital and Corporate Ethical Behaviour Perceptions Influence the Use of Collateral? Cross-Country Evidence

Studies that relate the use of collateral with moral hazard are, among others, Boot and Thakor (1994), Boot et al. (1991), Rajan and Winton (1995). The second main reason for the use of collateral is adverse selection, as discussed in, among others, Bester (1985), Chan and Kanatas (1985), and Besanko and Thakor (1987).

For instance, according to WBES, the proportion of loans requiring collateral range from 23% in Brazil to 100% in Guinea and Sudan. Data from the survey do not correspond to the same year for all countries. For example, the latest data for Brazil are from 2009, the ones for Guinea from 2016, and those for Sudan from 2014. However, the variation in the figures is not due to time differences. For example, in 2013, 28.9% of the loans in Turkey and 87% of the loans in Tunisia required collateral.

The Enterprise Surveys implemented in European and Central Asian countries are also known as Business Environment and Enterprise Performance Surveys (BEEPS) and are jointly conducted by the World Bank Group, the European Bank for Reconstruction and Development (EBRD), the European Investment Bank (EIB), and the European Commission (EC). Enterprise Surveys in Latin America are jointly funded with the Inter-American Development Bank (IDB). Enterprise Surveys in the Middle East and North Africa are jointly funded with EBRD and the EIB. These Surveys were initiated in 2002 and they cover firm-specific attributes and a broad range of business environment topics. For most countries, an Enterprise Survey is conducted about every 3–4 years. While the surveys were initially conducted by different units within the World Bank, since 2005–2006 most data collection efforts have been centralized within the Enterprise Analysis Unit. This resulted in a unified set of core survey questions and a consistent application of survey methodology across countries. Therefore, to ensure consistency the present study uses data from these standardized surveys in the post 2005–2006 period.

To give some examples, in the case of SOCIAL the figures are as follows: Ethiopia = 42.19 (2007), 42.19 (2008), 42.19 (2009), 42.37 (2010), 42.37 (2011); Jamaica = 50.33 (2007), 50.33 (2008), 50.33 (2009), 50.33 (2010), 50.02 (2011); Chile = 53.94 (2007), 53.63 (2008), 52.24 (2009), 52.52 (2010), 53.52 (2011); Panama = 54.78 (2007), 55.16 (2008), 53.59 (2009), 52.60 (2010), 53.27 (2011). The corresponding figures in the case of ETHICS are as follows: Ethiopia = 3.81 (2007), 3.83 (2008), 3.65 (2009), 3.94 (2010), 3.92 (2011); Jamaica = 3.98 (2007), 3.82 (2008), 3.88 (2009), 3.95 (2010), 3.89 (2011), Chile = 5.46 (2007), 5.34 (2008), 5.47 (2009), 5.56 (2010), 5.54 (2011); Panama = 4.35 (2007), 4.19 (2008), 4.02 (2009), 4.15 (2010), 4.13 (2011). Other countries follow similar patterns.

Micro firms represent only 0.53% of the firms in the dataset. Therefore, we group them together with small firms and we refer to them as small and micro enterprises.

Our approach is consistent with many other studies on the use of collateral that employ Logit/Probit models, like among others, Elsas and Krahnen (2002), Jiménez et al. (2006, 2009), Chakraborty and Hu (2006), Berger et al. (2011), Meles et al. (2017). We also estimate Eq. (1) as a linear probability model through the use of OLS, or a fractional response model (to constrain the feasible dependent values to the [0,1] range). The results remain intact and they are available upon request.

In unreported regressions we include one by one in the baseline regression (Column 1 and 7 of Table 3) another two variables. The first is the ratio of total annual sales to the number of full-time employees that serves as a measure of labour productivity. The second is an indicator that shows whether the sales of main product increase/remain the same/decrease in the last fiscal year. Given that we do not have information on the actual value of sales, we include this indicator either a set of dummy variables to distinguish between the three potential outcomes or as an index that takes the values of − 1 (decrease), 0 (no change), 1 (increase). This does not influence the results. In all the cases, the main results hold. We do not include these variables simultaneously with the other firm-level control variables in the regressions because this reduces substantially the sample to 121 observations in the case of SOCIAL and 115 observations in the case of ETHICAL. The estimations are available from the authors upon request.

The regressions presented earlier control for size; however, the inclusion of size as a control variable does not allow us to examine whether the coefficients of the variables of interest (i.e. SOCIAL, ETHICS) vary by firm size.

In unreported regressions we use two alternative instruments commonly employed in the literature. The first is religion, captured by the percentages of a country’s populations being Catholics, Muslims, and Protestants, those being the three most widely spread religions in the world in 1980 s. The second is colonization, captured by dummy variables that distinguish between British colonies, Latin colonies, and colonies of others (e.g. China, Turkey, Germany). The results hold and all the estimations are available from the authors upon request.

It is worth mentioning that although the 2SLS approach disregards the binary nature of the dependent variable, it remains an option which is viable and less complex, as pointed out by Angrist and Pischke (2010).

To conserve space, we do not report all the specifications. All the results are available from the authors upon request.