The measurement and determinants of x-inefficiency in commercial banks in Sub-Saharan Africa

Abstract

This paper uses the translog stochastic cost and profit frontier approach to measure the degree of x-inefficiency in a panel of 89 commercial banks drawn from nine Sub-Saharan African countries, covering the period 1992–99. The paper then models the determinants of x-inefficiency in terms of bank-specific factors and general macroeconomic variables. It is found that profit x-inefficiency is slightly higher than cost x-inefficiency, which suggests that revenue x-inefficiency is rather small. The evidence also shows that the degree of cost x-inefficiency is exacerbated by bad loans, high capital ratios and financial liberalisation. In contrast, it is shown that larger banks are more efficient and the level of foreign bank penetration reduces x-inefficiency. These findings have important implications for bank managers and regulators in Sub-Saharan Africa.

Keywords:

• x-inefficiency
• commercial banks
• Sub-Saharan Africa

Acknowledgements

We thank Chris Adcock, the Editor, Christopher J. Green and two anonymous referees of this journal for constructive comments on a previous version of this paper. We also thank the Department for International Development (DFID) for funding the research under the “Finance and Development Research Programme”, Contract No. RSC106506. However, we retain responsibility for surviving errors.

Notes

Broadly, the concept of x-inefficiency has been widely used in banking studies as a proxy for managerial performance (Allen and Rai 1996; Berger and DeYoung 1997) i.e. inefficiencies that result from managerial deficiencies, rather than sub-optimal output decisions.

Hence, the relevance of this research stems from the argument that the ultimate objective of regulatory authorities is to minimise the inefficiency in the banking sector, in order to realise at least three benefits (Berger, Hunter, and Timme 1993). First, more efficient banks intermediate more funds, offer a wider range and better quality of services to clients at competitive prices. Second, banks become more profitable and therefore investors expect higher dividends. Third, banks are able to attract more capital and reserves, in order to increase safety and soundness of the banks, and hence reduce the risk of bank failures.

Notable exceptions are the studies by Brownbridge (1998) and Maimbo (2002); however, neither do these studies measure x-inefficiency nor do they use econometric methods to model the determinants of x-inefficiency.

Unlike econometric work that focuses on estimated coefficients, the coefficients become subordinate in the process of computing the inefficiency measure (Greene 1993). What is of primary concern is the error term.

The number of share , where n is the number of inputs.

Worthington (1998) argues that the cost concept is more relevant in the determination of bank x-inefficiency because banks are constrained from achieving maximum profits due to regulatory restrictions, such as the minimum reserve and capital adequacy requirements. Moreover, it is argued that management has substantial control on the cost of inputs, whereas the output side is beyond their control. On the other hand, it may be argued that the profit concept is more appropriate because management performance involves controllable and non-controllable variables, and any measure that excludes either of them does not fully capture management performance. Given the arguments for and against the cost and profit measures of inefficiency, we use both.

Berger, Leusner, and Mingo (1997) examine two approaches to estimating cost x-inefficiency, namely the intermediation versus the production approach. For the intermediation approach, costs include both operating and interest expenses, whereas the production approach is based on operating costs only.

Berger and Mester (1997) show that the alternative profit function offers some distinct advantages that make it more applicable to panel data from different countries. Moreover, the alternative profit function takes into account variations in the quality of bank outputs across countries.

The non-parametric approach lacks an a priori functional form, which makes hypothesis testing difficult. Moreover, non-parametric techniques do not isolate the random error term and will therefore lead to inaccurate estimates of inefficiency.

We also tried to measure market power using a market share variable, denoted as BMKT, measured by loans and deposits equally weighted. However, the SHHI performed better than BMKT.

A bank is defined as foreign if at least 50% of its shares is foreign owned (see, among others, Lensink and Murinde 2006).

In most OECD countries, the corporate governance of banks has been subjected to greater scrutiny in order to mitigate agency-related problems and the pertinent agency costs.

Berger, Hunter, and Timme (1993) suggest that agency costs can be mitigated by adopting a corporate governance structure that separates decision-control from decision-management.

Brownbridge and Harvey (1998) found that foreign links and ownership contributed to better management and performance in the African commercial banking sector.

Williamson and Maher (1998) have indicated that almost their entire sample of 34 countries that undertook financial liberalisation between the beginning of the 1980s and mid-1997 subsequently experienced some form of systemic financial crisis. Moreover, Kaminsky and Reinhart (1998) have found that in 18 of the 26 banking crises studied, the financial sector had been liberalised during the preceding 5 years.

Due to missing data problems, we dropped three countries from the first possible sample of 12, namely Uganda, Tanzania and Zimbabwe.

For brevity, the descriptive statistics as well as the estimation and testing results for the cost and profit translog function and share equations are not reported here, but they are available from the authors. The diagnostic test results (not reported here but available from the authors) show that each of the explanatory variables in the translog model is stationary on the basis of the Augmented Dickey–Fuller (ADF) test.

This finding is consistent with earlier studies by Mester (1997), Berger and DeYoung (1997), and DeYoung (1998), which indicate that inefficiency is exacerbated by bad loan problems.