Technical Change in Banking: Evidence From Transition Countries

Abstract

This paper investigates the effect of technical change on the costs of banking firms operating in 11 Central and Eastern European countries using Fourier‐flexible cost function specification for the period 1995–2002. A common cost frontier with country‐specific variables is employed in order to take into account macro‐economic and regulatory conditions that vary over country and time. Our findings suggest that the rate of reduction in costs resulting from technical change increased during the sample period. Banks operating in Hungary, Czech Republic and Poland benefited more from technical change than their counterparts. In terms of cost reduction, large banks benefited more from technical progress. This indicates that large banks are more able to change their optimal input mix in response to changes in technology.

Key Words:

• Technical Change
• Banking
• Transition Economies
• JEL Classifications: G21, D21

Notes

1. Several studies have examined the banking efficiency in transition economies. However, only a small number of studies have studied the cost and profit efficiency using cross‐country dataset. See Kraft and Tirtiroglu (1998), Opiela (2000), Hasan and Marton (2003), Weill (2002), Weill (2003), Kasman (2005), Jemric and Vujcic (2002), Grigorian and Manole (2002), Bonin et al. (2003).

2. Recent empirical studies on the efficiency of banking industry show that the different assumptions on the one‐sided component of the composite error term do not lead to the very same results in terms of efficiency levels. It is possible to avoid distributional assumptions on the error terms by using panel data. The ‘distribution free’ model developed by Berger (1993) uses a balanced panel data for estimating efficiency. Using it in our study, however, dramatically decreases the number of banks in the sample. Hence, we use stochastic frontier approach to estimate the inefficiency score for each bank in the sample.

3. The Fourier is a semi‐nonparametric approach and is used to tackle the problem arising when the true functional form of the relationship is unknown. Gallant (1981, 1982), Mitchell and Onvural (1996), and Berger et al. (1997) note that the Fourier is a global approximation, which can represent a broader range of cost structure than other functional forms. The Fourier is shown to dominate the conventional translog functional form that is commonly applied in banking literature.

4. Following Mester (1996), equity is included in the cost function specification to control for risk. This variable is fully interactive with the output and input price variables.

5. The formula for z_i is 0.2π − µ · a, ln Y_i where µ ≡ (0.9 · 2π − 0.1 · 2π)/(b − a) and [a, b] is the range of ln Y_i .

6. Since the input prices show little variation across banks, we exclude the Fourier terms for the input prices in order to have limited number of Fourier terms.

7. Albania, Bosnia, Yugoslavia, and Macedonia are excluded from the sample due to data limitations.

8. For international comparison, it is preferable to convert total assets, total costs, loans, deposits, and other earnings into an international currency. Hence, we convert them into US dollars at current exchange rates. Some of the countries in the sample experienced significant exchange rate fluctuations within the sample period. Since trend in total costs and in other variables does not change much whether measuring them in local currency or not, the impact of converting these variables into an international currency on the size of the time trend coefficients would be limited.

9. We also used the intermediation approach suggested by Sealey and Lindley (1977), where the inputs, labour, physical capital and deposits are used to produce earning assets. Following Altunbas et al. (1999), total off‐balance sheet items are included as an output in the total cost function. The problem was that almost one‐third of the banks in the sample did not produce (or report) off‐balance sheet products during the sample period. Hence, we replaced values like $1000 instead of zero outputs. The empirical results are available from the authors upon request.

10. We did not have enough information on personnel expenses and fixed assets for several countries and did not have any for Bulgaria. Thus, in order to include these countries into the analysis, we followed Hasan and Marton (2003) and calculated a common price for labour and capital, using operating costs. Bankscope does not include number of employees for most banks either. Hence, the common price for labour and capital is computed as the ratio between operating costs and total assets.

11. See Dietsch and Lozano‐Vivas (2000) on the importance of controlling environmental variables in cross‐country studies.

12. We tried both the standard translog and the Fourier‐flexible specification (see McAllister and McManus 1993; Mitchell and Onvural 1996). Since formal tests indicate that the Fourier terms are jointly significant, the results reported here are those for the following Fourier‐flexible specification.

13. The sign of the income per capita is positive, which shows that the higher is the development level of the economy, the higher are the operating and financial costs. The signs of the coefficient of the population density and density of demand variables are also negative, suggesting that higher population and deposits densities contribute to lower banking costs. The Herfindhal index has a positive sign. If this index is indeed a measure of market power, the positive sign suggests that a higher market power induces banks to spend more on staff or on other personal expenses. The sign of the coefficient of intermediation ratio is negative, suggesting that higher amounts of loans per unit of deposits decrease banking costs. Financial depth variable also has a negative sign suggesting that financial financial depth decreases banking costs. The average capital ratio has a negative sign indicating that it is less costly for well‐capitalized banks to provide banking services. The coefficient on GDP growth is negative, indicating that the higher is the growth rate, the lower is the banking costs. The size variable has a negative sign, while the coefficient on capitalization has a positive sign.

14. As mentioned earlier, we also estimated the common cost frontier using off‐balance sheet items instead of deposits as an output. The regression results are very similar. The off‐balance sheet items were reported predominantly by the largest banks in the sample.

15. We also estimated the cost function with the environmental variables replaced by a set of country dummies as suggested by one of the referees. These would control for any permanent cross‐sectional variation between countries, with all of the time‐series variation being attributed to the ‘technological change’ terms. In this context, these terms represent combined effects of technological and environmental changes. Separating these two effects out, however, is difficult. Regression with a set of country dummies produces similar results with respect to the direction of the technological change. Regression results are available upon request from the authors.

16. This result is meaningful because the ratio of average costs to average assets also decreased during the sample period (from 12% in 1995 to 8% in 2002).