Client profiles and access to retail bank services in South Africa

Abstract

In many developed, and some developing countries, technological innovation has enabled banks to provide intermediary services and supply products via a variety of methods. This article uses econometric techniques to assess the impact of the product suite, credit score and the spatial availability of bank branches on the level of access to banking services in Gauteng Province, South Africa over the period 1999 to 2004. The results indicate that the allocated product suite and the credit score do not clearly influence the overall level of access. The spatial availability of branches has the most significant impact on overall access and at the margin, the number of bank branches is less important than the geographic distribution of the branches. Given the socio-economic profile of retail clients in South Africa, increases in investment in product suite inputs provides less productive benefits for overall access than bank branches do. This is contrary to the view that may hold for most developed economies.

Acknowledgements

This article is part of the Southern Africa Financial Institutions Industry Structure (SAFIIS) Project funded under BARDA-Grant 017/2002–2008. I wish to acknowledge Econometrix (Pty) Ltd, South Africa for allowing me access to their Sector Star, Econosnaps, and Finance Monitor database products. I am grateful to Michael Adeyemo, Fabian Ogbonna and Bongiwe Kanzi for able research assistance particularly in the cleaning and cross validation of the branch input cost data. They are not responsible for any errors or omissions which may remain.

Notes

¹Although I use the terms access to finance generically, access to finance is not the same thing as access to banking services. Access to finance encompasses a number of things including, but not limited to, the ability to obtain credit from formal and informal sources of supply. Access to bank services can mean the same thing but is used herein to imply access to services and products supplied by the formal financial services sector. This study is an assessment of access to formal finance, of which one component is retail banking.

²A possible methodology to assess this proposition is to carry out the type of tests which are used in measurements of income convergence. But in this instance, this would entail establishing some benchmark for the level of access and then regressing a time series of the average growth rate of the level of access on the initial level of access. The presence of a negative and significant coefficient would support the view that lower initial levels of access to services correspond to higher rates of growth in access. If this is true, it would support some form of ‘access convergence’. While the absence of time series data of access levels prevents the adoption of this method, an assessment of this hypothesis is nevertheless a key objective of this article.

³See Appendix 1 for full list of municipal areas. I am aware that it is possible to use other measures to assess concentration. For example, in her study of agglomeration economies, Viladecans-Marsal (2004), uses the Gini index, the index of relative concentration and Moran's I-statistic of spatial association.

⁴In terms of GDP, provinces in South Africa are considerably smaller, than states in the USA and regions in Europe. For example, according to data from International Monetary Fund (IMF) World Economic Indicators World Fact Book 2005, the GDP of South Africa was US$256.8 billion. The only anomaly is Gauteng Province. The GDP of Gauteng Province which is the largest GDP of the nine provinces, is approximately 33.4% (US$66.4) billion of the total GDP of South Africa. North West province has the smallest GDP, US$12.44 billion. In comparison, in 2005, the GDP of the smallest states of the USA, Alabama, Arkansas, Louisiana and Mississippi was US$132.8 billion, US$74.7 billion, US$110.9 billion and US$72 billion, respectively.

⁵The average cost per client is the sum of the total annual cost per product suite multiplied by the number of clients in each product category divided by the number in each category.

⁶The distance from home to the branch may not be the shortest distance. For clients that are in full-time employment, the distance from the branch to the place of work may be shorter and perhaps the more accurate measure for branch access – however, while the bank keeps records of the home address it does not do so for work address – which is subject to greater change.

⁷The original DCIAS file has data on gender and race. However, the bank refused to provide data on the race of the client and the gender explanatory variable has incomplete data entries and missing observations. Accordingly, it was not possible to test the impact of either of these two explanatory variables.

⁸Clients are located in city, township and rural areas. Accordingly, the average distance is larger than the average distance (8.7 km) which was used in the original DCIAS database in Okeahalam (2005b).

⁹The results of the correlation tests are available and will be provided upon request.

¹⁰The records provided us with information in four risk categories, high, neutral 1, neutral 2 and low and no record. We have grouped neutral 1 and 2 together and therefore, have three credit profile cohorts.

¹¹On average clients live 14.3 km from their nearest bank branch. The data for the distance from work were incomplete.

¹²Some of the black townships are urban areas. Accordingly, a distinction is made by separating urban non – township areas from urban-township areas.

¹³An exclusion restriction can be used in a simultaneous equation system – to indicate that some of the exogenous variables are not in some of the equations. This idea can normally be expressed by saying the coefficient that is next to the exogenous variable is zero. Expressed in this way this restriction may be testable and may make a simultaneous equation system identified. See, Stock et al. (2002) for a survey.

¹⁴See Dolton and Makepeace (1987) for details of interpretation of the coefficient estimate of this regressor.

¹⁵Information from the Banking Council of South Africa data indicates that a leading bank indicates that it has spent 40 million rands to set up the Mzanzi lowest cost account product suite and needs to open 700 000 such accounts to break – even implying an average cost 57 rands per account. A minimum deposit of 20 rands is also required to open the account and this level of deposit needs to be maintained by the client to enable the bank to breakeven

¹⁶Calem and Nakamura (1998) find that although branch banking enables banks to differentiate themselves via the choices of branch locations, bank branching reduces local market power by widening the geographic scope of competition among banks. Therefore, bank services at noncentral or convenient locations tend to be priced more competitively in locations served by y branch networks.

¹⁷Berger and Udell (2002) find (at the firm level) that the ‘soft’ relationship between the loan officer and a client – based on factors such as character and inter-personal track record is more important than the relationship between a firm and the bank. However, they also illustrate that such ‘soft data’ such as the relationship between a loan officer and a client is difficult to quantify, assess and transmit through normal channels in bank firms. However, as Chandler and Coffman (1979) indicate this ‘soft’ judgment concept is not new.

¹⁸An assessment of the parameters of unmeasured heterogeneity was made by estimating the constants from the binary and continuous equations, the SD and location and size of the points of support. All parameters of location of the points of support and the relevant weights were significant at the 5% level. Three points of support were used on the basis of the likelihood ratio test and the product suite and branch availability coefficients are robust at this level. An attempt was also made to increase the number of points, however, when this was done, some of the other parameters of unmeasured heterogeneity became insignificant and inconclusive points of support.

¹⁹Put simply, neoclassical growth theory contends that over time, there will be a reduction in the level of differences in per capita income (beta convergence) and also a reduction in spatial incomes (sigma convergence). Despite the political changes of 1994, this has not been the case in South Africa and this fact fits well with the findings herein, in that a proportion of the benefits of the product suite is dependent on increasing the level of access to computers and virtual media – which is dependent on increasing per capita incomes and/or investment in low income areas. Although not tested explicitly here, these conclusions indirectly support the recent findings e.g. Drennan et al. (2004) regarding the relative lack of international evidence in support of the beta income convergence hypothesis (Barro and Sala-i-Martin, 1991) and sigma income convergence (Quah, 1993). Tests of beta and sigma (unconditional and conditional) income convergence and divergence have been conducted primarily with US data – but have also been replicated extensively in many parts of the world – except Africa which is odd -given the high Gini coefficient measures in most African countries. Indeed, Drennan et al. (2004) use unit root tests to illustrate that in the United States, at the metropolitan level, sigma income convergence has not taken place over the period 1970 to 2000 – they find evidence of divergence.

²⁰From the perspective of Mzanzi market share, Postbank leads the way with 27.3% of the market, followed by Standard bank with 25.9%, ABSA with 23.9% and FNB with 16.2%. Nedbank holds 6.7% of the Mzansi market. New accounts have brought some 240 million rand (US$37 million) into the formal banking sector. Fifty-six percent of Mzanzi account holders are female and the majority of account holders are in the 25–54 years age group.