INSTALACIÓN DE SENSORES

In this section, we examine how financial inclusion affects bank stability after controlling for bank and country level variables. The results are reported in Table 4.4, where we use two different measures of bank stability. In columns 1-2 and 3-4, we regress ln(Z-score) and –ln(sd(ROA)) on financial inclusion, respectively. For the latter case, we follow Beck et al. (2013) and transform return volatility to make it directly proportional to banking stability. To check for the robustness of our results, we use two variant measures of market power denoted as C-Lerner and E-Lerner. Before choosing which estimator we should use for equation (4.1), we conduct an endogeneity test for the financial inclusion measures, which is reported at the bottom of Table 4.4. For rejecting the null hypothesis of exogeneity, we employ the IV-GMM estimator. In case we cannot reject the exogeneity of financial inclusion, we use the OLS estimator as it is more efficient. In both cases, we calculate heteroskedasticity and autocorrelation consistent (HAC) standard errors which are reported in square brackets. We test the validity of our instrumental variables as in GMM procedures using the under-identification LM test by Kleibergen and Paap (2006) and the over-identification test by Hansen (1982). The results on these tests show that the

80 _{Since financial inclusion is generally related to per capita income, these two variables tend to be correlated.}

We computed the variance inflation factors (VIF) for each of our model estimates. VIF is equal to 1/ (1-r2_),

where r2 _{is from the regression of an independent variable on the rest of the independent variables. The}

average VIF never exceeds 3, indicating that multicollinearity is not a cause for concern for our results (Anginer, Demirgüç-Kunt and Zhu, 2014). Furthermore, following previous studies on the determinants of financial development (e.g., La Porta, Lopez-de-Silanes, Shleifer and Vishny, 1997; Beck, Demirgüç-Kunt and Levine, 2003), as a robustness test we exclude per capita GDP in all estimations and the results are broadly consistent with the main findings of this study. The results are available from the authors upon request.

instruments used are valid as the p-value of the former (latter) requires a value lower (higher) than 0.05 to reject the null hypothesis at the 5% level.

Table 4.C1 shows the First-stage regressions of financial inclusion on instruments used in this study.81 We find that all instruments have statistically significant effects on financial inclusion- the direct effect and interaction. Moreover, the signs and magnitudes of the coefficients are economically important, as financial inclusion increases more in markets with greater financial freedom and higher density of newly registered companies. A system with higher financial freedom and entry density would facilitate access to finance through augmenting banking competition and create a milieu for efficient financial intermediation between households, financial institutions, firms and entrepreneurs.82 For example, to assess the economic significance of a whole set of instruments, consider column 4, at the mean for entry density (2.61), the marginal effect of financial freedom equals 0.002 (0.002-0.0001*2.61 = 0.002). This effect implies that one standard deviation above average increase in financial freedom (70.83) leads to a 0.14 unit increase in financial inclusion (equals a little less than ½ standard deviation of financial inclusion).

It is clear from the results that a more inclusive financial system is associated with greater banking stability, as indicated by its positive and significant (at the 1% level) coefficients (once again, a greater estimated Z-score indicates more stability i.e., less risk taking). Since we use the natural logarithm of Z-score, the coefficients can be interpreted as semi-elastic. In column 1, a one standard deviation increase in the index of financial inclusion, which equals 0.30, is associated with an increase in the ln(Z-score) of 189% (6.3*0.30). Put differently, our financial inclusion index lies between zero and unity, where

81 _{The significant negative relationship between market power and financial inclusion in Table 4.A3 is}

consistent with the existing literature and should serve as another indication of the robustness of our index (see for example Carbó-Valverde, Rodriguez-Fernandez and Udell, 2009; Ryan, O’Toole and McCann, 2014).

82 _{Entry density is one of the channels through which financial development fosters economic growth}

(Klapper, Laeven and Rajan, 2006).

a one standard deviation increase would be a substantial increase for any given economy; for a ¼ standard deviation increase in the index of financial inclusion leads to a 45% (6.4*0.07) increase in the ln(Z-score) (based on averaging the results across columns 1 and 2). The effect is economically important as it suggests that financial inclusion enhances banks to have a secure deposit base as well as wider lending opportunities. Therefore, with the inclusive financial sector, banks enjoy greater financial stability. This result also corroborates with the additional risk measures used in this study. The negative of return volatility –ln(sd(ROA)), in columns 3 and 4, is also positively related to financial inclusion, suggesting that an increase in the index of financial inclusion is associated with a reduction in return volatility.

These results are consistent with the view that a system with inclusive financial services tends to reinforce banking stability (e.g., Han and Melecky, 2013; Khan, 2011; Morgan and Pontines, 2014) and that a higher degree of financial inclusion mitigates excessive risk-taking of an individual bank. Since greater financial inclusion reduces distance between financial institutions and low-end customers it is able to decrease the probability of loan defaults, and hence bank fragility. This result is also supported by Agarwal and Hauswald (2007) and DeYoung, Glennon and Nigro (2008), who use US data and find that loan default probability increases with the distance between lender and borrowers. Recent empirical evidence also finds positive impacts from geographic diversification and reducing distance between banks and borrowers (e.g., Berger and DeYoung, 2001; Bos and Kolari, 2005; Deng and Elyasiani, 2008; Rossi, Schwaiger and Winkler, 2009). One would also expect that when financial institutions expand activities towards areas where more unbanked populations are located, they may be more likely to engage in diversified lending and have a wider source of cheaper funding through retail deposits rather than relying on volatile wholesale funding thus increasing the soundness of banks. Therefore, it can be argued that financial inclusion is good for banking stability.

Our results on control variables are also consistent with existing literature. As might be expected, larger banks, and banks with better management, higher equity capital and pricing power are more stable. Regarding country-level macro controls, the results show that banks operating in countries with higher economic growth and less income level significantly increase banking stability.