Assessing lending market concentration in Bulgaria: the application of a new measure of concentration.

• Author: Lapteacru, Ion
• Position: Report

1. Introduction

   The issue of the measures of concentration is a long-standing debate. The most frequently applied measure of concentration, at least in banking, is the Herfindahl-Hirshman Index (HHI). It is used, for instance, in studies focusing on the relationship between market structure and banks' performance, but also by banking authorities to examine the impact on competition of the eventual mergers and acquisitions (2). Despite its popularity, this concentration measure is much criticised. Its frequent usage is due to the fact that it is simple to calculate.

   Another simply constructed concentration measure is the Entropy Index (EI). It was frequently applied to assess the industrial firms' strategy, and, to our knowledge, never to banking. The construction principle is the same as for the HHI, the weights attached to market shares are only different: HHI assigns higher weights to higher shares whereas the EI assigns to higher shares lower weights. Thus, both indexes are subject to "weight bias".

   Many other measures have been proposed. Among them, market share inequality indexes of Rhoades (1995) reveal the inequality among firms even within markets with similar HHIs. Nevertheless, they are not exempt from some weaknesses. As Rhoades (1995) mentions, two of them are very sensitive about the number of firms, which increase rapidly with the increment in number of firms. Another inequality measure is the difference between the largest and smallest market shares. The latter concentration measure has, as other concentration measures, a weight bias, placing greater weight on larger market share differences.

   Two of the market share inequality measures of Rhoades (1995) have been used by Hannan (1997) along with the inequality part of the HHI, in order to assess their effects on the bank deposit and loan rates. His results are inconclusive in explaining the deposit interest rates. For the loan-rate analysis, it arises that only the number of firms influences the small business loan rates.

   Even if some transformations have been made, all these inequality measures are not completely exempted from the effect of the number of firms. An original research is that of Melnik et al. (2008) where the authors propose a dominance measure in order to disclose the firm with the dominant position. For this reason it has a limited interest for assessing the market concentration (3). Nevertheless, neither the inequality measures of Rhoades (1995) nor the dominance measure of Melnik et al. (2008) gauge the market concentration and so are not of interest in our study.

   In our opinion, a concentration measure must not be influenced by the number of entities existing in the market, only the share they own should determine the market concentration. This could easily be corrected by the normalisation of the HHI and EI, as they take values between zero and one regardless of the number of firms on the market. However, the weight bias that characterises these concentration measures will always be present. As will be discussed below, this sensitivity on the weight attached to market shares makes different values for these two concentration measures and different rates of evolution, even if the sense of the evolution is the same. The measure we introduce in this paper avoids all weight problems.

   Thus, this work presents a twofold interest. First, we propose a new methodology to estimate a concentration index that avoids any weight bias. It is based on the same approach as the Gini index and allows the determination of the shape of the Lorenz curve. We use this new measure to assess the sectoral lending concentration in Bulgaria. The market shares used are therefore the share of each economic activity in a lending portfolio. Thus, this new approach provides information on the loan distribution across economic sectors. The case of a small number of sectors receiving the most funding corresponds to what we call beta-concentration. In the case of a large number of highly financed sectors, the loan portfolio is said to be alpha-concentrated. Second, we assess the evolution of the sectoral concentration of bank loans and determine its factors for Bulgaria, and compare these results with those obtained with the normalised HHI and EI.

   We choose to make our study on the case of the Bulgarian lending market since it underwent a marked evolution of the economic sectoral concentration due to several banking crises which occurred in the nineties and also experienced many banking reforms. Once the non-performing loans ratio began to decline, Bulgarian banks gradually adopted a new approach for financing the economy and it was reflected in the way in which the different sectors of economy are financed.

   The rest of the paper is structured as follows. Section 2 provides a brief overview of the empirical studies on the use of the different concentration measures, and some reasons to be in search of other measures of concentration. In the third section we explain the similarities and dissimilarities between the Herfindahl-Hirschman and Entropy indexes, showing how they give different results for the same objective. In the fourth section, we explain the construction of a new concentration index that by its nature gives more information on the way in which the concentration occurs. Section 5 examines the determinants of loan concentration for Bulgarian banks by applying the new concentration indexes and compares the results with those obtained with the normalised HHI and EI. Section 6 summarizes the main ideas and concludes.

2. Brief literature review

   Many studies have begun to consider the entropy index as a measure of concentration (or diversification, depending on the context) from the seventies (4), when the first works gave preference to this index with respect to the HHI. Jacquemin and Berry (1979, p. 363), for instance, affirm that, due to different weights that are assigned to market shares, small differences in large shares make little difference to the value of the EI, while the HHI responds more considerably. Conversely, very small shares are much ignored by the HHI. They estimate the role of the two concentration measures on the growth of the firms' assets and conclude on the advantage of the EI with respect to the HHI. This feature of the EI, which makes it more sensitive to small shares, led Gemba and Kodama (2001) to use this concentration measure in their research on diversification dynamics of the Japanese industry.

   The majority of papers on diversification applied the EI as measure of diversification and not the HHI, as the former can be decomposed directly "into additive elements which define the contribution of diversification at each level of product aggregation to the total" (Jacquemin and Berry, 1979 p. 361). According to this, Raghunathan (1995) refines the entropy concept to measure the diversification as well as the entropy formula to estimate the firms' diversification. However, in his improved concept of entropy, the new measure is still a weighted index.

   The EI's feature that consists of weighting the very small shares more, making it very sensitive to them and for which the cited authors used it in their concentration (or diversification) market research becomes a weakness for Stigler (1968, p. 33). He gives preference to the HHI as the more eloquent of the two as a measure of market concentration. For similar reasons, Grant et al. (1988) used the HHI but with the objective of studying industry diversification. Despite this, the HHI is the most applied index used to measure market concentration or market structure.

   In banking it is the only index used. There are, for example, many studies that test the Structure-Conduct-Performance (SCP) hypothesis in the banking industry. Some papers examine the SCP hypothesis throughout the relationship between the market concentration, measured by the HHI, and the profit rates of banks. Pilloff (1999), Pilloff and Rhoades (2002) and Rhoades (1995) find a result that is consistent with current practice of bank antitrust; that is, a concentration market implies higher profit rates. This is not really the case for Berger (1995), who finds that a positive relationship between banking market concentration and banks' profits disappears when a measure of the cost efficiency of banks is added to the regression as an independent variable, which is consistent with the efficient-structure hypothesis. The HHI was also applied in regressions to estimate its effects on the interest rates that banks pay on deposits (Hannan, 1997; Sharpe, 1997) or charge on loans (Hannan, 1991, 1997; Hannan and Liang, 1995) (5). The results are not conclusive in the case of the deposit interest rate and they are clearly in favour of the SCP hypothesis in the case of the loan rates. In the case of the deposit rates, also applying the HHI as measure of concentration, Hannan and Berger (1991) find that banks in more concentrated markets are less likely to change deposit rates. In a previous paper, these authors, with the same concentration measure, examine the price-concentration relationship and confirm the SCP hypothesis (Berger and Hannan, 1989).

   Other series of papers examine, for example, the role of market concentration or competition, measured by the HHI, on the banks' efficiency scores. Maudos et al. (2002) find, on the European banks' example, that within a more concentrated market the banks' cost efficiency scores improve and the profit efficiency scores deteriorate. Other authors use the concentration HHI measure to control the cost function from which is extracted the cost efficiency scores (among others, Dietsch and Lozano-Vivas, 2000, and Staikouras et al., 2008).

   The question which we raise in the next section is how well the HHI and EI describe the market structure. That is, how the assignation of different weights to market shares changes the values on market concentration. To assess these effects...