Competition and risk in South East Asian commercial banking

Abstract

This article investigates the effects of competition on bank risk taking behaviour in four South East Asian countries (Indonesia, Malaysia, Philippines and Vietnam). Our main finding is that competition does not increase bank risk-taking behaviour and the results appear robust to different model specifications, estimation approaches and variable construction. We also find that concentration is inversely related to bank risk whereas regulatory restrictions positively influence bank risk-taking.

Keywords:

• bank competition
• banking system fragility
• financial stability
• regulation

JEL Classification::

• G21
• F30
• L89
• G38

Notes

¹ See Ghosh (2006) for an excellent exposition of financial restructuring trends in East Asian financial systems, especially Chapters 1–4.

² Laeven (2006) also notes that in some banking systems of East Asia have competition levels below pre-1997 to 1998 crisis levels.

³ Berger et al. (2009) also note that the mixed findings for developing countries are likely to be an artefact of the small sample size.

⁴ ROA is in percent.

⁵ We also computed H ² which equals H ¹ but replacing staff costs over total assets by staff costs over loans plus deposits () as a measure of personnel unit price. H ⁴ equals H ³ but the similar change is also made. This alternative construction of variables is used for comparison of H-statistic which is later given in Table 5.

⁶ Linear dynamic panel regression models include one or more lags of the dependent variable as covariates and contain unobserved individual effects (either fixed or random). Arellano and Bond (1991) use a GMM estimator for such models, known as the difference GMM. The lagged exogenous variables values (levels) constitute legitimate instruments for the first-differenced, lagged dependent variable. However, these lagged variables may provide little information about the first differences (Arellano and Bover, 1995; Blundell and Bond, 1998). Building on the work of Arellano and Bover (1995), Blundell and Bond (1998) developed a system estimator that exploits additional moment conditions on both first differences and levels, with lagged first differences of the series employed as instruments in the levels equation. The system GMM estimator reduces potential bias in finite samples as well as asymptotic imprecision associated with the difference estimator (Blundell and Bond, 1998).

⁷ In addition, because loan-loss reserves are stock items, banks managers may determine the timing of these stocks at their discretion to reduce regulatory costs (Altunbas et al., 2007)

⁸ Z-index captures three important components. First, it includes ROA, which is widely used as a measure of bank performance. Second, it includes ROA volatility, a measure of risk used in bank financial management. Third, the index incorporates the bank equity-to-asset ratio (the reciprocal of the equity multiplier). The Z-index has been widely used as measure of the ‘safety and soundness’ of a banking system (Nash and Sinkey, 1997, p. 96). Z-index has been used to measure banking sector risk by, for example, Nash and Sinkey (1997), De Nicolo (2000), De Nicolo et al. (2004), Boyd et al. (2006), Yeyati and Micco (2007) and Uhde and Heimeshoff (2009).

⁹ Because the Z-index can take large negative values, we adjust the value by taking the logarithm of (Z-index + 150).

¹⁰ Total number of bank–year observations between 1998 and 2008 amounted to 1216, distributed as follows: Indonesia, 447; Malaysia, 261; Philippines, 311; and Vietnam, 197.

¹¹ These include Chou (2000), Montreevat (2000), Tschoegl (2001, 2003), Chua (2003), Coppel and Davies (2003), Focaralli (2003), Bekaert and Harvey (2004), Detragiache and Gupta (2004), Megginson (2005). Other sources are ASEAN Bankers Association (Regional Updates), World Bank (2000), McMillan (2002), Montlake (2003), The US Embassy in Jakarta (2005).

¹² This is estimated for a pooled country sample yielding one H-statistic for each country. We have tried to compute yearly H-statistics following Molyneux et al. (1994) and Yeyati and Micco (2007).