Financial liberalization, financial regulation and bank efficiency: a multi-country analysis

Abstract

This article investigates the impact of financial reforms on bank efficiency. More specifically, we distinguish between two different types of financial reforms, i.e. financial liberalization measures and measures of the quality of bank regulation and supervision (i.e. financial regulation), and study their relationship to bank efficiency separately. Moreover, we analyse whether the impact of financial liberalization on bank efficiency is conditional on the quality of regulation and supervision of the banking system. We apply stochastic frontier analysis to calculate bank efficiency at the individual bank level and use a new and detailed database that measures different aspects of financial reforms. The data-set consists of 87 312 bank-year observations covering 61 countries for the period 1996–2005. Overall, we show that the impact of financial liberalization policies on bank efficiency is conditional on the extent to which bank regulation and supervision has been adopted and developed.

Keywords:

• financial liberalization
• financial regulation
• banks
• efficiency
• stochastic frontier analysis

JEL Classification:

• G18
• G21

Notes

¹ In the remainder of the text, we will use the terms ‘financial regulation’ and ‘bank regulation and supervision’ interchangeably.

² Examples are Chile and Argentina. In the early 1980s, these countries experienced the negative effects of financial liberalization. The same holds for Mexico (in 1994–1995) and the countries affected by the Asian crisis (1997/98).

³ Adopting risk-based capital adequacy ratios based on the Basel standard is considered to be better in terms of the quality of bank regulation and supervision and their impact on bank efficiency, because adopting these capital adequacy ratios reduces moral hazard and risk-taking behaviour of banks. This in turn reduces the overall riskiness of the bank and increases its ability to efficiently allocate financial resources.

⁴ Appendix 1 in Abiad et al. (2010) provides a more detailed account of the definition and measurement of the three components of this dimension of financial reforms.

⁵ Each of the three components receives an equal weight in the final score of the bank regulation and supervision dimension. While other weightings could have been used, we have no a priori theory guiding us to choose a particular weighting scheme. For this reason, we have chosen to stick to using equal weightings. Here, we follow Abiad et al. (2010) who created the original data-set, along with several other researchers who also have made use of this data-set.

⁶ A more detailed description of the coding system can be found in Abiad et al. (2010).

⁷ Data envelopment analysis does not allow for measurement error and luck factors. This technique attributes any deviation from the best-practice bank to technical inefficiency.

⁸ Thus, the total costs a bank faces are never lower than the costs of the frontier. For a graphical representation of the frontier and its dynamics, see Berger et al. (1993). The authors show how inefficiency is determined by both technical and allocative inefficiencies.

⁹ We use the functional form of the model as described in Equation 4, because this allows us to estimate the model in Stata using the uhet option in the frontier command (StataCorp, 2009). Using the exponential specification ensures that σ_u,i,t never becomes negative.

¹⁰ A complete list of all variables and abbreviations used in the analysis in this article can be found in Table A1.

¹¹ Table A2 provides the details of the principal component analysis. The table shows that the first component covers 41% of the variation of the underlying measures. Moreover, the eigenvalue for the first component is 1.63, which is well above that of the second component (1.04). We therefore only use the first component in the estimations presented in the tables below.

¹² This is true, even if we incorporate the results of Greene (2005a) to exploit the sparse nature of the Hessian.

¹³ In the original Abiad et al. (2010) data-set, the banking supervision variable can obtain the values 0, 1, 2 and 3. For optimization reasons regarding the calculation of the cost function and the inefficiency equation we have divided the observations for this variable by 10 so that they are in the same range as the other variables in our model.

¹⁴ The results for the cost frontier are available on request from the authors.

¹⁵ The results for this robustness test are not presented to save space, but are available on request from the authors.

¹⁶ Examples are the Banking Code of the Financial Service Authority in the United Kingdom, the Banking Code of the Committee Maas in the Netherlands and the Code of Banking Practice of the Australian Bankers’ Association.