Efficiencies and Unobserved Heterogeneity in Turkish Banking 1990-2000.docx

《Efficiencies and Unobserved Heterogeneity in Turkish Banking 1990-2000.docx》由会员分享，可在线阅读，更多相关《Efficiencies and Unobserved Heterogeneity in Turkish Banking 1990-2000.docx（37页珍藏版）》请在得力文库 - 分享文档赚钱的网站上搜索。

1、 Efficiencies and Unobservable Heterogeneity In Turkish Banking: 1990-2000 Mahmoud A. El-Gamal and Hulusi Inanoglu* September 2002 Abstract Studies of bank efficiency tend to draw conclusions from pooled estimates, assuming that all banks in a sample use the same technology, or estimates based on a

2、priori classifications of the banks. It is well known that efficiency rankings may be corrupted if banks that use different technologies are pooled together in estimating the technological frontier with respect to which inefficiency is estimated. In this paper we model unobserved heterogeneity in ba

3、nking technologies as a mixture model, and investigate the efficiencies of 53 Turkish banks using likelihood- based stochastic frontier analysis for the period 1990-2000. We use the EC (Estimation- Classification) estimator to obtain data-driven identification of bank-technology-classes in our sampl

4、e, and to estimate the mixture components of banking technologies therein. We have four dimensions of possible a priori heterogeneity in our sample: Large vs. small, state vs. private, foreign vs. domestic, and Islamic vs. conventional banks. In some cases, multiple a priori criteria may be confound

5、ed, e.g. most foreign banks were also small. Simple tests of homogeneity cannot disentangle those confounded effects. Our likelihood-based analysis finds no evidence of heterogeneity along the state vs. private and Islamic vs. conventional dimensions. The estimated classifications and mixture compon

6、ents have intuitive ex post institutional explanations. Keywords: Turkish banking, stochastic frontier efficiency analysis, classification. * Mahmoud A. El-Gamal is a Professor of Economics and Statistics at Rice University, where Hulusi Inanoglu is a Ph.D candidate in the Economics Department. We t

7、hank Robin Sickles, Eli Berman, and seminar participants at Rice University, University of Houston, and Harvard University, for many useful comments and suggestions. Needless to say, all remaining errors are ours. Correspondence: Mahmoud El-Gamal, Dept. of Economics MS 22, Rice University, Houston,

8、TX 77005. Emails: elgamalrice.edu , hulusirice.edu 1 1. Introduction While financial innovations have caused traditional banking to decline in developed countries, banks continue to play a dominant financial intermediation role in most countries. For instance, the share of U.S. commercial banks lend

9、ing as a percentage of total domestic non-financial sector credit has declined from 35% in 1974 to 25% in 1998. Moreover, U.S. commercial banks share of total financial intermediary assets declined from 40% in 1980 to below 25% in 1998, c.f. Mishkin and Eakins (2000, p. 481). On the other hand, comm

10、ercial banks are still the primary financial intermediaries in the financial sector, and U.S. banks relative share in GDP is increasing, c.f. Allen and Santamero (2001). Since the main function of banks is moving funds from lenders to borrowers, a banks loans-to-assets ratio is a good indicator of i

11、ts fulfillment of this basic function. In this regard, while U.S. banks loans amounted to 60% of their total assets, Turkeys commercial banks averaged a loans-to-assets ratio below 40% for the period 1990-2000. Taking into account that the majority of financial flows go through the banking sector, a

12、nd that the banking sector accounts for 75% of the assets of the total financial sector in Turkey, c.f. BRSA Report (2001, p.1), it is fair to say that the Turkish banking sector is not fulfilling its potential in generating loans. Turkish banks have not fared better on the deposits-attraction side

13、of financial intermediation. In this regard, it is estimated that Turkish households are hoarding roughly 15 billion U.S. Dollars-worth in the form of cash and gold,1 the latter serving as an inflation-hedge in a highly inflationary environment.2 Indeed, this pattern is explained by the fact that re

14、al returns for foreign exchange transactions and gold have been positive, while real returns on deposits have been negative. For instance, the real return from USD exchange transaction in 2001 was 27.6%, and the real return from DM exchange transaction was 20.5%, while gold-holdings offered a 30.6%

15、real rate of return. In contrast, the real return on bank deposits was negative 5%, c.f. Milliyet (2002). Mismanagement of a countrys banking sector can cause significant damages beyond the low growth caused by financial disintermediation. For instance, banking crises in Mexico, South Korea, Russia

16、and Indonesia are estimated to have cost each of those countries between 21% and 50% of its GDP in the past decade. Similarly, the Banking Regulation Supervision Agency (BRSA) declared at the end of 2001 that recent Turkish banking crises have cost the country more than USD 20 billion (ibid.). Turki

17、sh banking began in 1856, with the establishment of Ottoman Bank. Established by foreign capital, Ottoman Bank was granted the authority of printing money, and mainly served as a vehicle for facilitating internal and external borrowing. Ottoman period banking was dominated by foreign banks, until th

18、e declaration of the Turkish Republic in 1923, c.f. B.A.Ts 40th Year Book (p. 8). 1 In 2000, Turkey imported 205 tons of gold, accounting for 6% of total international gold imports, and earning Turkey a ranking of the seventh largest importer of gold in the world. 2 Average annual inflation rate is

19、77 % during 1990-2000. 2 Akguc (1989) divides the history of Turkish banking after 1923 into five phases: (i) the period of national banks (1923-1932), (ii) the period of state-owned banks (1933- 1944), (iii) the period of developing private banking (1945-1959), (iv) the planned period (1960-1980),

20、and (v) the period of liberalization and open economy (After 1980). A detailed review of those five periods is provided in B.A.Ts 40th Year Book. In its early periods of development, the Turkish banking sectors main goal was to regain national control of domestic capital. During the import-substitut

21、ion era prior to 1980, the national plan called for negative real interest rates to finance the development of domestic industries. This period was also characterized by limited competition and state control of the banking sector. For instance, only seven new banking licenses were issued between 196

22、0 and 1980, two of which for foreign banks. Therefore, the issue of efficiency of the banking system only came to the fore during the period of liberalization and open economy starting in 1980. The 1980 structural change and reform program called for free market and export- oriented policies, includ

23、ing liberalization of the financial system. New policies included the abolition of directed credit, liberalization of deposit and credit interest rates, liberal exchange rate policies, and the adoption of international best standard banking regulations, c.f. BRSA (2001, p.1). The liberalization prog

24、ram allowed a number of domestic and foreign banks to enter the market, with a marked increase in competition. On the other hand, the regulatory framework of that period gave banks an opportunity to make “easy profits” through non-core-banking activities. With increased freedom of capital movement a

25、nd foreign currency transactions, banks borrowed funds from abroad, and invested them in deficit-financing government bonds which paid high interest rates. The high interest payments were, in turn, financed through more public borrowing, and inflationary monetary expansions, leading to very high inf

26、lation in the early 1990s. Starting in the late 1980s, there has been a surge of research on the efficiency of Turkish banking, especially to investigate the effects of the liberalization program. That body of research has mainly focused on the types and degrees of heterogeneity in the Turkish banki

27、ng industry. In the presence of such heterogeneity in banking technology, measures of “inefficiency” may be corrupted. Using the data-driven Estimation- Classification methods of El-Gamal and Grether (1995), we find significant differences in banking technology between small and large banks, and bet

28、ween foreign (mostly small) and domestic banks. The likelihood functions for our estimation are derived from classical trans-log Stochastic Frontier cost functions, with controls for the quality of loans and risk exposure of banks, following the specification in Mester (1996). In fact, there are a n

29、umber of a priori dimensions along which heterogeneity could have been detected. Interestingly, our data-driven estimation-classification did not select some of the obvious a priori classifications. 3 Apart from methodology, our study also has the advantage of using a panel data set that is superior

30、 to those used in previous studies. First, our dataset includes all banks that were in operation throughout the decade. Moreover, the 49 conventional banks in our sample account for more than 93% of the total assets of the conventional banking system. Second, our dataset includes special finance hou

31、ses (SFHs), which function as Islamic banks in Turkey. The four special finance houses in our dataset managed at least 90% of the total assets of all Islamic banks in Turkey during the sample period. While Islamic banking in Turkey started in 1985, there have been virtually no rigorous empirical stu

32、dies of the Turkish banking sector including the SFHs. As a matter of fact, there has been very little empirical study of Islamic banking in general, despite the significant growth of this sector in a number of countries over the past two decades.3 Empirical studies of Islamic banks have mostly reli

33、ed on descriptive statistics, and theoretical analyses, rather than rigorous statistical estimation methodologies. For instance, Aggarwal and Yousef, (2000) surveyed the financial instruments used by Islamic banks, and found that most of their instruments mimicked conventional banking debt-based fin

34、ancing. That is in contrast to the theory of Islamic banking, which portrayed the focus of the latter to be on equity-based financing and profit-sharing arrangements. In this regard, Al-Deehani, Abdelkarim, and Murinde (1999) proposed a model in which, under certain assumptions, an increase in inves

35、tment accounts financing would enable Islamic banks to increase both their market values, and their shareholders rates of return at no extra financial risk to the bank. Their empirical analysis of the annual accounts of 12 Islamic banks supported their theoretical predictions of increased Islamic ba

36、nks market values without a change in their cost of capital. More recently, Iqbal (2001) reviewed the performance of various groups of Islamic and conventional banks within various countries, using trend and ratio analyses. His sample consisted of twelve Islamic banks with a “control group” of twelv

37、e conventional banks from ten different countries, over the period 1990-98. Similarly, Samad (1999) compared the performance of one Malaysian Islamic bank to seven Malaysian conventional banks over the period 1992-1996 using financial ratios. Bashir (1999) performed a similar risk and profitability

38、examination of two Sudanese banks. Thus, the extant efficiency analyses of Islamic banking have been limited to the examination of simple financial ratios. In this regard, our paper provides the first rigorous efficiency analysis of Islamic banks within a conventional banking industry. In Section 2,

39、 we provide a literature review of the most commonly used econometric methods of efficiency analysis, as well as a review of previous efficiency studies of Turkish conventional banking. In Section 3, we describe our dataset and provide basic ratio analyses. In Section 4, we report the results of our

40、 efficiency analysis for the pooled dataset, as well as the results allowing for heterogeneity. Finally, we provide some concluding remarks in Section 5. 3 Recent studies estimate the size of the Islamic finance industry world-wide at $200 billion, with an estimated annual growth rate of 15%, c.f. W

41、arde (2000, p.1), Lewis and Algaoud (2001, p.7). 4 2. Literature Review 2.1 Econometric Efficiency Analysis Econometric efficiency analysis methods may be grouped into three main categories: parametric, nonparametric and semi-parametric. Naturally, parametric techniques impose the strongest function

42、al and distributional assumptions, while nonparametric techniques impose the weakest. Parametric efficiency analysis methods date back to Aigner, Lovell and Schmidt (1977) and Meeusen and van den Broeck (1977), who independently proposed a stochastic frontier approach (SFA). The model was originally

43、 specified for cross-sectional data, and included an error term with two components: a normally distributed random effects component, and a half-normally distributed technical inefficiency component. Numerous recent studies use the maximum- likelihood approach for estimating stochastic frontiers all

44、owing for panel data sets, as well as missing data, as developed by Battese and Coelli (1992). Schmidt and Sickles (1984) proposed a semi-parametric approach for estimating stochastic frontiers in panel data models without imposing distributional assumptions on error terms. Cornwell, Schmidt and Sic

45、kles (1990) later improved on that distribution free approach by allowing for time heterogeneity in slopes and intercepts. Adams, Berger, and Sickles (1997) further approached a fully non-parametric setting by imposing only minimal assumptions on the functional form. The stochastic frontier efficien

46、cy analysis approach has been extensively applied to the study of financial institutions performance. Berger and Humphrey (1997) provided an excellent survey of 130 studies that applied frontier efficiency analysis to financial institutions in 21 countries. Fully nonparametric techniques, such as Da

47、ta Envelopment Analysis (DEA) and Free Disposal Hull (FDH) are also available (ibid.). Those techniques require neither the specification of a functional form, nor the assumption of cost minimization or profit maximization. However, nonparametric techniques have the fundamental drawback of confoundi

48、ng potential random measurement errors with inefficiency measures. Thus, for purposes of efficiency analysis, the SFA approach seems to dominate nonparametric approaches by allowing for measurement error. In this paper, we utilize a fully parametric stochastic frontier analysis in order to utilize t

49、he likelihood-based EC-estimator of El-Gamal and Grether (1995) for modeling unknown heterogeneity. The issue of separating heterogeneity effects from efficiency has been a concern in many studies of U.S. and European banking. For instance, Brown and Glennon (2000) performed tests of homogeneity in U.S. banking technologies and rejected that null hypothesis. They performed their test by grouping the banks using cluster analysis, and then estimating the efficiencies for six different clusters. Altunbas et al. (2001) estimated three separat