Macroeconomic and market determinants of interest rate spreads in low- and middle-income countries

Abstract

Numerous variables exogenous to the operations of commercial banks have been widely touted in academic literature and popular discourse to be important factors causing the typically high Interest Rate Spreads (IRS) in developing countries. Using data for a group of 33 countries, this article applies dynamic panel estimation techniques to investigate the macroeconomic and market determinants of banking sector IRS in low- and middle-income countries. The empirical results suggest that only one market specific factor, the banking sector reserve requirement, significantly and positively affects IRS. Conversely, several macroeconomic and macro-policy variables such as inflation, government crowding-out and the discount rate are important determinants of IRS. Results are also examined to ascertain whether the determinants of spreads vary across regional groupings of countries.

Notes

¹ Chirwa and Mlachila (2004).

² Other widely varying measures of bank efficiency are also used in the literature. For example, Yildirim (2002) and Ataullah and Le (2006) estimate the efficiency of banks using data envelopment analysis, and Forster and Shaffer (2005) use an efficiency ratio adopted from the US Federal Financial Institutions Examination Council.

³ Quaden (2004) further notes that the increased efficiency of financial institutions should ‘facilitate the re-allocation of capital towards new developing sectors and firms that have a high growth potential.’ This is supported by Lucchetti et al. (2000) who argue that efficient financial institutions tend to use technologically-driven cost reduction methods, the use of which is a ‘necessary condition for the efficient allocation of resources.’

⁴ This model was tested using data from 17 administrative regions in Spain over the period 1986 to 2001. One of the conclusions made is that there is a significant and negative effect of the variable that proxies intermediation costs on gross fixed capital formations, ‘showing the negative effect of augmenting transformation costs on investment’ (Valverde et al., 2004).

⁵ Robinson (2002), Jayaraman and Sharma (2003) and Tennant (2006).

⁶ See, for example, Demirguc-Kunt and Huizinga (1998).

⁷ Demirguc-Kunt and Huizinga (1998) explain by noting that, ‘a reduction in NIMs can, for example, reflect a reduction in bank taxation or, alternatively, a higher loan default rate. In the first instance, the reduction in NIMs reflects an improved financial market function, while in the second case the opposite may be true. Also, note that variation in an accounting ratio such as NIM may reflect differences in net interest income (the numerator) or differences in (say) nonlending assets (in the denominator).’

⁸ Brock and Franken (2003) cite Catao (1998), Aizenman and Hoffmaister (1999), and Corvoisier and Gropp (2001) as examples. See also Moore and Craigwell (2000).

⁹ Sologoub (2006) notes that the ideal measure of the interest rate spread of a bank is the difference between the average interest earned on loans and the average interest paid on deposits.

¹⁰ With a slightly different focus, Jacques (1995) examines the ability of the spreads between 3-month and 6-month Treasury bill rates to predict inflation, Adao and Luis (2000) investigate the probability of convergence of options spreads in Spain, Italy and Germany, and Poya and Matthews (2004) examine the link between the term spread and Gross Domestic Product (GDP) growth in the Korean economy, and explores the usefulness of the spreads as an indicator of recessions.

¹¹ Brock and Franken (2002) cite Ho and Saunders, McShane and Sharpe, and Brock and Rojas-Suárez as examples.

¹² See, for example, Demirguc-Kunt et al. (2003).

¹³ In examining the relationship between bank efficiency and size in Latin America, Forster and Shaffer (2005) noted that, ‘robust associations were found between absolute size and efficiency, whereas no such associations were found between relative size and efficiency. These findings together suggest that economic development may benefit from policies that are tolerant of large banks, and tend to rule out market power of dominant banks as a likely cause of the observed empirical associations.’

¹⁴ Yildirim (2002), while not specifically examining IRS, similarly notes that macroeconomic conditions had a profound influence on the efficiency of Turkish commercial banks.

¹⁵ As quoted in Jayaraman and Sharma (2003).

¹⁶ Robinson (2002), however, notes that, ‘discussions of banking behaviour which rely only on ex ante measures downplay the importance of portfolio composition, capital adequacy and asset quality.’ It must be noted though that whilst this limitation is acknowledged, it does not impact very heavily on this study, which focuses on the market and macroeconomic determinants of IRS, rather than the individual bank characteristics mentioned by Robinson (2002).

¹⁷ It must be noted though, that for a number of countries, various specificities are included in the IFS’ definition of the average commercial bank lending and deposit rates. The comparison of spreads across countries is therefore not perfect, but is the best that can be achieved using aggregated data in large cross-country studies.

¹⁸ Sologoub (2006).

¹⁹ Randall (1998), Jayaraman and Sharma (2003) and Tennant (2006).

²⁰ This measure is similar to that used by Vergil (2002) to examine the effects of exchange rate volatility on trade.

²¹ Tennant (2006).

²² See the Appendix of list of countries included in the study.

²³ Generalizations about Latin American countries have to be viewed with caution, as our dataset only includes two such countries. Forster and Shaffer (2005) have, however, noted that efficient banking in the Latin American region has been hindered by ‘volatile economies, lack of quality loan demand, high fraud rates, poor legal systems and low levels of savings.’

²⁴ Cross-section F-statistic: 0.9931 (probability: 0.4825); cross-section chi-square statistic: 34.667 (probability: 0.3419).

²⁵ The Durbin–Watson statistic of 1.8936 is greater than Du (1.86923).

²⁶ Numerous other attempts were made to correct for autocorrelation, but none were successful. It should be noted that when lagged values of CROWD and SCALE were included along with the AR(1) term, the Durbin–Watson statistic falls within the zone of indecision, and results were generally weaker than when contemporaneous values were used.

²⁷ Demirguc-Kunt and Huizinga (1998) define their concentration variable as ‘the ratio of the three largest banks’ assets to total banking sector assets’.

²⁸ Although countries such as Belize and Guyana are not islands, they are included as SIDS because being small coastal states they share many of the same characteristics. These countries also are members of the Alliance of Small Island States (AOSIS).