Estimating Market Power in the South African Banking Sector

Abstract

Indicators of market power can be ambiguous because cost economies associated with scale and not market imperfections may influence results. This article illustrates that without direct measures of concentration, estimates of costs, scale economies and profitability can be used to identify market power in banking. Secondly, via this method, econometric estimates provide meaningful evidence of market power in the South African banking sector over the study sample period (1979–1998). A reasonable conclusion is that while industrial structure is important, careful consideration needs to be given to cost economies and profitability when assessing market power. In addition, there is a need to consider appropriate policy to reduce market power in banking in South Africa.

Key Words:

• Costs
• scale
• banks
• South Africa

Jel Classifications:

• G21
• C14
• D2

Acknowledgements

The author is grateful to Econometrix (Pty) Ltd, South Africa for access to their Sector Star database and to Kofi Afful for research assistance. He is responsible for any errors which may remain.

Notes

1. Statistics South Africa—the national board of South Africa—was established in 1999. As a result changes were made in the way in which official government data series are captured. Accordingly, the sample ends in 1998 so as to reduce potential measurement error, measurement inconsistency or bias.

2. Brucker (1970) extended the multiple‐product price discriminating model of Schull (1963) in which empirical estimates of a bank’s total demand elasticity are obtained via the transformation of deposit inputs into credit outputs.

3. See Chaffai (1997) and Okeahalam (2004, 2005) for a few examples of studies on this issue which have been carried out in Africa.

4. This does not imply that the decision was wrong. The author served as lead adviser to the Competition Commission of South Africa and chaired the enquiry process of this bid.

5. V also includes materials (non‐labour) used as variable inputs of production such as: stationary, correspondence, advertising, technology, service transport and energy costs.

6. The approach used is similar to that of Ferrier et al. (1993).

7. The time frame 1979–1998 was used in measuring marginal costs and marginal revenue.

8. Clark and Speaker (1994) and Clark (1996) illustrate that specification of a detailed flexible cost function can effectively be used to assess measures of the effects of costs and scale on market power in banks.

9. The standard cross‐validation procedure requires at least two samples—the screening sample (Sample 1) and the calibration sample (Sample 2). The basic approach is to compute regression output coefficients in the screening sample and then use weights from the screening sample to compute the predicted scores in the calibration sample.

10. Berger and Hannan (1998) use a stochastic frontier method to check the robustness of estimates. Temple (1998) demonstrates techniques for testing the robustness of cross‐section and panel data regressions and sensitivity to measurement error.

11. To maximise degrees of freedom and increase statistical efficiency, the SAFSSDB has observations for 20 years. It is unlikely that relationships have remained stable. However given the small number of banks, there has had to be a trade‐off between the length of time‐series observations and the goodness of fit of the model.

12. Market shares in the banking sector can be calculated on the basis of volume, for example, the number of bank accounts or by value, for example, the proportion of total assets (or liabilities). Most data concerning the banking market in the South Africa is based on volumes.

13. The costs of merging Volksabank and Trust Bank to form ABSA in 1992 had a short‐term impact on the profitability of the new group and on the Nedcor group after the acquisition of the Board of Executors (BOE) bank in 2003.

14. The current international investment regulations for South African corporations places a cap of R750 million rands per annum on investment to all countries except to countries in Africa—where the annual limit is 2 billion rands.

15. As a result of exchange control regulations banks incurred significant additional labour costs by hiring staff to process documentation between clients and the South African Reserve Bank.

16. The profitability measures (P_WR^V )for monopoly and monopsony are measured indirectly. The profitability measure for monopoly is the product of measures of input‐specific scale economies and market power and the profitability measure for monopsony is derived from measures of marginal cost via estimates of input‐specific scale economies, therefore, P_VR^V· ΣTcw = P_WR^V .