Banking market integration in the SADC countries: evidence from interest rate analyses

Abstract

This study investigates the state, development and drivers of banking market integration in the member countries of the Southern African Development Community (SADC). A Principal Component Analysis (PCA) of national retail interest rates indicates increasing integration in loan and deposit markets. These integration processes are not developing uniformly and we can identify a convergence club. When investigating the interest rate pass-through from central bank onto retail rates for this convergence club, we find both, genuine and monetary-integration driven processes though the latter dominate. We thus conclude that a selective expansion of the Common Monetary Area (CMA) is possible.

Keywords:

• banking market integration
• monetary integration
• Southern African Development Community
• interest rate pass-through
• principle components

JEL Classification::

• E42
• E43
• E52
• E58
• F36

Acknowledgements

We would like to thank the participants of the 2nd Emerging Markets Finance Conference at the Cass Business School in London in May 2008 and the Biennial Conference of the Economic Society of South Africa in Johannesburg in September 2007 and an anonymous referee for their helpful comments and suggestions. All remaining errors are our sole responsibility.

Notes

¹ Most of the regional alliances are focusing on common economic and trade policies, such as the common market for Eastern and Southern Africa and economic community of West African States, whereas the 14 member countries of the African Financial Community form two monetary unions, the West African Economic and Monetary Union and the Central African Economics and Monetary Community with a single central bank in each union and a single currency or peg to the French franc/euro, respectively.

² In choosing this point of view, we also implicitly acknowledge the new strand of literature on endogenous optimum currency areas which starts with Rose (2001) and hypothesizes a potentially large positive effect of a common currency on trade. For a review of the related literature since the 1950s, see Dellas and Tavlas (2009).

³ Seychelles was a member of SADC from 1997 to 2004, when it pulled out of the arrangement, but has since then started negotiations to rejoin. Hence in the analysis it is treated as a SADC member.

⁴ For a summary of exchange rate systems and restrictions on capital account transactions in SADC countries, see Table A1 in the Appendix.

⁵ Advanced countries typically have deposits to GDP and loans to GDP ratios around or even above 100%.

⁶ For applications of sigma and beta convergence measures to interest rates, see Adam et al. (2002), Baele et al. (2004) and Kleimeier and Sander (2006). For an application to Southern Africa, see Honohan (1992), Nielsen et al. (2005) and Wang et al. (2007). We calculate sigma convergence four each of our four regions as the monthly cross-country coefficient of variation across the region's countries. When calculating beta convergence, we follow Kleimeier and Sander (2006). We estimate the regression Δz_{c,t +12} = βz_{c, t} where z_{c, t} is defined as the difference in month t between the interest rate of country c and the South African interest rate. The change in this deviation over the coming 12 months is measured by Δz_{c,t  + 12} = z_{c, t +12} − z_{c, t} . A negative β coefficient thus indicates the existence of beta-convergence.

⁷ Only the beta convergence coefficients for lending rates of Madagascar and Mauritius relative to the South African lending rate are not statistically significant at 10% or stronger.

⁸ In addition to the contribution to the understanding of SADC banking markets, our study is thus also contributing to the growing literature which assesses regional integration based on PCA (Nellis, 1982; Fernandez-Izquierdo and Lafuente, 2004; Figueira et al., 2005; Siliverstovs et al., 2005; Perignon et al., 2007; Gilmore et al., 2008; Becker and Hall, 2009).

⁹ Note that the PCA makes no assumptions about the underlying properties of the data series X. Thus there is no need, for instance, to determine the stationarity properties of each series.

¹⁰ For a discussion of weaknesses and alternative rules, see Jolliffe (2002, pp. 112–31). For applications, see, for instance, Nellis (1982), Figueira et al. (2005) and Meric et al. (2008).

¹¹ Additional results of the PCA – as well as detailed results for the PTA – can be found in the working paper version of this study available as METEOR Research Memorandum RM07047 at http://edocs.ub.unimaas.nl/.

¹² Though the results did not show a monotonically increasing pattern for the entire period across the interest rates, we have chosen the 2000 to 2005 periods for this analysis to coincide with a period where there is some evidence of consistent increase in the level of integration in the entire SADC as revealed by the R ².

¹³ In contrast to our symmetric PT model, asymmetric PT models have, for example, been applied by Sander and Kleimeier (2004, 2006) or Gambacorta and Iannotti (2007). Asymmetric PT models are most useful when investigating the internal structure of the banking market as they provide detailed information about the – possibly asymmetric – short- and mediumterm adjustment of interest rates. For the long run, however, they estimate the same multiplier as presented here in Equation 4. As we will mainly focus on this long-run multiplier as our indicator for integration, we prefer the simpler symmetric PT model.

¹⁴ In contrast, I(0) or stationary series have a limited variance and fluctuate around their mean. By differentiating an I(1) series, an I(0) series is obtained. We employ various tests to establish whether the interest rate series are I(1) or I(0). For each rolling sample period we conduct two types of unit root tests: an Augmented Dickey–Fuller (ADF) test and a modified Dickey–Fuller (DF) test, based on Generalized Least Squares (GLS) detrended series (commonly called the DF-GLS test), as proposed by Elliott et al. (1996). For the full period, we additionally estimate mean-shift, trend-shift and recursive unit root tests which are valid even in the presence of a structural break (Banerjee et al., 1992). Results can be found in the working paper. Our tests indicate that the series are I(1), but there are a few exceptions. As we, however, do not encounter cases where both the monetary policy rate and the retail rate are I(0), we always estimate the PT in first differences.

¹⁵ We employ two types of tests in order to determine whether or not monetary policy and retail rates are cointegrated. First, we estimate the usual Durbin–Watson (DW), Dickey–Fuller (DF) and ADF tests. Cointegration is considered to exist if at least two test statistics are significant at the 10% level or one test statistic at the 5% level. Second, we follow Kremers et al. (1992) and consider cointegration to exist when the coefficient of the lagged error correction term (ECT _t−1) is significant at the 5% level in Equation 6. If one or both of these test procedures indicates cointegration, we estimate the PT as in Equation 6. Only if both of these test procedures reject cointegration, the PT is estimated as in Equation 5. To avoid switching too frequently between these two PT models, i.e. for our overlapping sample periods, exceptions are made for single periods. Details can be found in the working paper.

¹⁶ For specific national multipliers, see Table A3 in the working paper.

¹⁷ With respect to the short-run multipliers (not shown in Figure 2) note that the long-run values of the average multipliers are almost reached after 3 months. However, the 3-months multipliers show a slightly higher variability across countries.

¹⁸ By focussing on the transmission from South Africa's central bank rates to national interest rates, we follow Aziakpono (2006). He finds that policy convergence exists among CMA countries while market convergence is lacking possibly due to the absence of arbitrage opportunities resulting from both poor institutional development and lack of investment opportunities.

¹⁹ Buigut and Valev (2006) also favour these two countries.