Japanese Banking Sector Competition - Dissertation Example

? Investigating Japanese Banking Sector Competition – A Panel Data approach This paper investigates the degree of competitiveness in the Japanese Banking Sector. Using a longitudinal data set containing information on 1018 banks covering time span of 1988-2012 and applying a modified version of the Panzar-Rosse methodology suitable for the present purpose, several interesting insights are derived. We find that the overall Japanese Banking industry has a monopolistically competitive structure. But, the nature of competitiveness is not the same across all bank sizes. The most important finding of the paper is that the degree of competitiveness has been falling over the last two decades and this can be at least partially an important explanation of the worsening performance of the Japanese economy over this time span. Introduction Albeit the Japanese economy has been long hailed as the model of success in Asia, the transition towards a market oriented economy has not been smooth. Despite of the ravages left in the aftermath of the 2nd world war, the 1960’s and 1970’s saw Japan attain enormously high rates of economic growth (Johnson, 1982). The regulated financial sector working in tandem with the government and business corporations led to a stable and steady integrated economic system which allowed the economy to flourish. The Japanese banking system had a critical role to play in this phase. Not only did the banks act as corporate governing bodies, they also played roles or rescuers when enterprises where in financial difficulties. By providing loans to enterprises that were investing in sectors with strong growth potential these banks shared the risks in similar vein to venture capitalists (Wade, 1999). But in the decades of the 1970’s and 1980’s the fast growing economy compelled by the global environment of market integration had to modify its structure and attempt to adjust to the new environment. Growing domestic businesses gradually had a lower requirement to borrow from the domestic banking system. Circumvention of financing from external sources coupled with developing asset markets through the accumulation over the earlier decades led to alterations in the capital flows and liberalization of the financial sector followed (Noguchi, 1998). In the latter half of the 1980’s decade such liberalization resulted in a lack of adequate tightness in monetary conditions which in turn led to an asset oriented initial upturn and economic boom but finally the asset bubble got burst and this opened the floodgates for Japan’s economic woes. Due to the depressed market conditions the 1990’s have been famously coined as the “lost decade” (Takahashi, 2011). Since the early 2000’s the Japanese economy has been in the process of trying to recover through market oriented reforms but no remedy to the ailment which continued to make the economy weaker and the recent global financial crisis has only worsened the situation (see figures 1 to 3). Figure 1 Figure 1 above shows the path of real GDP over time. Evidently the climb is steeper and more steady until 1990 since when evidence of volatility is observed and the slope is flatter as well. A substantial dip is also visible in the mid 2000’s. Figure 2 Figure 2 reveals that the percentage of annual GDP growth has actually been quite volatile. However, more noticeably there is a downward trend in the series and the growth rate has decelerated to negative values over the last few years. Figure 3 Finally figure 3 shows the average growth rate for the four decades since 1970. Evidently the performance was substantially lower in the decade of the 1990’s and to add to the woes of the economy, the growth rate has been even lower in the 1st decade of the 2000’s. The economy is still in quest for attaining a system that has the advantage of institutional complementarities as it once had in its golden era of growth. Substantial amounts of research has established that the degree of competitiveness has important bearing on economic growth and financial stability of an economy (see Levine, Loayza and Beck, 2000; Collender and Sherrill, 2003 for instance). With the recent financial crisis bringing to the forth the importance of the banking system in maintenance of financial stability, monetary policy transmission, credit provision and thus for steady economic growth, we question whether it could be possible that the degree of contestability (or the lack thereof) could be playing a role in the current malaise of the Japanese economy. The objective of this paper, thus, is to evaluate the nature of contestability in the current banking system in Japan. We shall explore the degree of current competitiveness as well as whether observable significant differences exist over time in competitiveness. We use the Panzar-Rosse approach (to be detailed in succeeding section) and apply it to a real time updated Panel dataset containing information on 1019 Japanese banks. Tsutsui and Kameska (2005) adopted a similar model to evaluate the competition dynamics of the security industry in Japan during the last quarter of the 20th century1. However, no study this author is aware of has been carried out using this framework and real time updated data on the Japanese Banking sector. The structure of the study is as follows: in the next section we introduce relevant literature and the theoretical and empirical upshots present to set the context for the present endeavour. In subsequent sections we shall provide information on the nature of the dataset being used for the purpose, the empirical framework and the results of the analysis respectively. Finally, the conclusion will summarize the main objectives and results of the study. Empirical Framework Theoretically, following Lerner (1934) the extent of competitiveness of the banking industry can be evaluated by looking at by how much the marked up price exceeds the marginal cost. However, because for that difficult to obtain detailed data regarding the costs and prices of various products offered are necessary. To by pass such difficulties, alternative measures of competition have emerged in literature. Bikker (2004) classifies these into two groups – structural measures and non-structural measures. The structural approach stems from the prediction of higher concentration leading to greater collusion among firms. Competition is attempted to be proxied by concentration measures such as the Hirschman-Herfindahl index. This approach has been criticised in a number of studies that argue that concentration is a poor reflection of competition (Claessens and Laeven, 2004; Shaffer and DiSalvo, 1994). The non-structural approach follows the direction suggested by the New Industrial Organization Approach or NIOA and is built on a profit maximizing framework which is a considerably more acceptable compared to heuristic alternatives (Bikker et al, 2009). The approach we focus on here is the Panzar-Rosse (Panzar and Rosse, 1982, 1987) method which is the most prominent among the non-structural clan of approaches to measuring banking competition. In this approach, a reduced form equation representing the representative bank’s revenue is explored and in particular the parameter of interest is the sum of the elasticities of revenue to factor costs. Since under perfect competition increases in factor prices do not affect output and thus revenue, under monopolistic competition, revenue is affected by such changes in factor prices. So, under perfect competition or perfect contestability the sum of the elasticities is anticipated to be unity, it is expected to be negative under monopoly and lie between zero and unity under monopolistic competition. The reduced form revenue equation expresses the relationship between the total revenue and factor prices with appropriate firm specific controls for bank size and risks. If we assume a one output k-input production technology, the reduced form revenue equation can be represented in the following log-log form: In Panzar and Rosse (1977) it has been shown that the sum of the revenue elasticities to the factors, reflects the competitiveness of the market. In particular, it is shown that for a competitive firm in its long run equilibrium, H equals unity, for a monopolistically competitive firm H will lie between zero and unity and it will be zero or lower for a profit maximizing monopolist or a colluding oligoplist firm. This forms the basis of the framework to be followed in this study. We shall use fixed effects estimation to estimate a variant of equation (1). The 1st regression specification we shall estimate is: Where IR is the log of interest revenues, LP is the natural log of price of labour, KP represents the natural log of price of capital and DP represents the natural log of deposit price. Eq stands for the natural log of equity and is included as a control for general risk level, while loan, the natural log of loans is included to control for bank specific risks. Finally, since the larger the bank, the larger is the revenue, to control for bank size we have included the natural log of total assets, TA. Under this specification, the Panzar-Rosse statistic will be: . If H = 1, we can conclude that the market is perfectly competitive. If H = 0, that will imply a monopoly and finally if , we shall conclude that the market is monopolistically competitive. Note that represents firm specific fixed effects while represents period specific fixed effects in equation (3). Although a number of studies have used a scaled version of the revenue function as the dependent variable to control for the size effect, such scaling leads to erroneous inferences about the degree of competition as shown by Bikker, Shaffer and Spierdijk (2009. So, we incorporate the size effect by using total assets as a proxy for size as a control variable. It has also been argued that such an estimation of the Panzar-Rosse statistic is not free from errors. Particularly the predicted values of the statistic hold in equilibrium. If the firm is not in equilibrium even if it is competitive, the value of the statistic may not equal unity. Similar arguments hold for the monopoly market as well. Since the required behaviour holds only when profits are maximized at the optimum, off equilibrium behaviour may not lead to an H value lying at or below zero. In order to evaluate the possibility of disequilibrium we also estimate the following specification: The only difference is the incorporation of the log of Returns on assets (ROA) instead of interest revenue (IR) as the dependent variable. The explanatory variables are the same as the previous specification. The approach to testing whether the market is in equilibrium was used by Matthews, Murinde and Zhao (2007). The market will be in equilibrium if we can obtain evidence that: We shall refer to equations (3) and (4) as the 1st and 2nd specifications. In spite of the mentioned difficulties we retain the Panzar-Rosse methodology for two reasons. The difficulty of the H-statistic revealing the true nature of competition when the market is not in equilibrium is taken care of using the error correction term. Additionally, using total assets as a control along with unscaled revenue function as suggested by Bikker et al (2009) we ensure that the validity of the conclusions derived on the basis of the H-statistic are not compromised. The major advantage of this approach is that apart from being able to evaluate the degree of contestability of the market using the H statistic, each of the signs on the estimated coefficients is expected to reveal the different influences the variables included as controls have on interest revenue. The revenue elasticities individually also may be of interest for policy since they reflect expected changes in interest revenue of banks if the price of a factor or input changes. The model’s simplicity is also appealing for reasons of parsimony which may also reflect why it has been the most popular approach to modelling banking competition in recent literature. Data description This section briefly discusses the source and the nature of the data used for the purpose of the study. Variable definitions are provided along with summary statistics. For the purpose of the present study, we have used yearly data ranging from 1988 to 2012 for 1018 Japanese banks available from the bankscope database. Although the data are of a yearly frequency, there are some gaps for some banks. So, we have an unbalanced panel. The following table presents the variables used in the present study along with their definitions available from bankscope. Table 1: Variable Names and Definitions Variable Name Definition Interest Revenue Net interest revenues Deposit Price Interest Expenses/Total Deposits and Money market funding Capital Price Operating and Other Expenses/ Total Assets Labour Price Personnel Expenses/Total Assets Equity Ratio Equity/Total Assets Loans Loans/Total Assets ROA Returns on Assets = total revenue/total assets Observe that our primary independent variables of interest are all ratios. Capital price and labour price has been scaled by total assets while deposit prices essentially the interest expenses scaled by total deposits and money market funding. Equity and loans both are expressed as a ratio of total assets as well. Table 2: Summary Statistics   Interest Revenue Deposit Price Labour Price Capital Price Equity Ratio Loans  Mean  629.0536  0.004498  0.009952  0.029690  0.053065  18784.98  Median  9.165000  0.002693  0.008487  0.022145  0.046645  300.1000  Maximum  21300.00  0.336809  0.167969  0.926829  0.758799  747566.0  Minimum -1000  0.000000  0.000000  0.000000 -1.6 -6  Std. Dev.  1720.874  0.010898  0.011834  0.044292  0.081915  52948.07  Skewness  5.481215  17.26764  8.408791  9.068599 -4.13261  5.593572  Kurtosis  47.82807  468.4790  83.04606  115.8689  106.5753  47.71536 Table 2 above presents the summary statistics for the data. A critical point to note here is that there is a substantial variation in the interest revenue as observed by the range of the observations as well as by the huge standard deviation. Figure 4: interest revenues across small, medium and large banks Noting the substantial range of interest revenues prompted us to explore whether differences in size led to significant variations in interest revenues or not. As figure 4 shows, there are indeed significant differences. The interest revenues from largest banks (IRL) almost always has remained above interest revenues from medium banks (IRM) which in turn has almost always remained above the interest revenue from small banks (IRS). Thus, we should expect to find significant differences of mean among the net interest revenues of these three groups. Table 3: Mean Differences in interest revenues across small, medium and large banks         Std. Err. Variable Count Mean Std. Dev. of Mean IRL 2910 6.527523 1.54737 0.028685 IRM 3481 2.768656 1.093684 0.018537 IRS 2882 0.500225 0.975089 0.018163 All 9273 3.243226 2.702624 0.028066 Table 3 above confirms this fact. We also carried out a separate a Welch F test which under the null hypothesis of no mean differences led to a computed statistic value of 16010.90 which is significant even at the 1 percent level. Noting these differences leads us to expect different behaviour and possible differences in competitiveness or contestability across these groups. This motivates a separate analysis for each group. Results and Discussion This section will present and discuss the results of running the empirical methodology outlined in the empirical framework section. Table 4 contains the fixed effect estimation results of both specifications 3 and the disequilibrium test from specification 4 reported in the last row of the table for each specification for the entire sample and for subsamples of small medium and large banks (columns 3 through 5). The classification of banks into these three classes was done by examining the total assets as proxy for bank size. The bottom 30% from the perspective of total assets was taken as the group of small banks, and the top 30% total asset holding banks were classified as large banks. The remaining 40% of the total sample constitutes the “medium sized bank subsample. Table 4: Estimation results Variable All Banks Small Banks Medium Banks Large Banks Intercept -1.51*** -2.85*** -1.16*** -0.76 Labour 0.34*** 0.12 0.35*** 0.52*** Capital 0.19*** 0.15*** 0.18*** 0.089** Deposit -0.03 -0.05 0.01 -0.02 Loan 0.36*** 0.2 0.16 0.58*** Equity 0.10*** 0.12*** 0.09*** 0.05 Total Assets 0.66*** 0.82*** 0.83*** 0.41*** Adjusted R-squared 0.99 0.95 0.98 0.99 Pr(H0: Redundant Fixed effects) 0 0 0 0 Joint insignificance statistic (F) 2355.93*** 76.80*** 329.49*** 578.98*** Pr(H0: E=0) 0.57 0 0.09 0.82 Notes: *** implies significance at 1%, ** implies significance at 5% and * implies significance at 10%; specification I and II are equilibrium and disequilibrium specifications respectively. Before looking at the estimated coefficients note from the 2nd last row in the table that for all specifications the p-value associated with the test of insignificance of the firm specific and time specific dummies (fixed effects parameters) is zero for all the considered specifications. Thus the redundancy of running a fixed effects model is rejected. This justifies adoption of the fixed effects method. Looking through the details of table 1, the other point to notice is that the variable ‘deposit’ has a negative coefficient for all specifications but the medium bank subsample. All other variables have positive coefficients. Thus, we find that for all banks across all samples considered an increase in the price of deposits will lead to a reduction in the interest revenue. However the Furthermore, it is significant for none of the specifications. Deposits do not seem to have any statistically significant influence on the interest revenues of banks even after controlling for size differentials. The unit price of capital has a significant (at least at the 5% level) and positive impact on interest revenue for the entire subsample of all banks as well as for all the subsamples. Wages or the unit price of labour has a highly significant (at 1% level) and positive impact on all banks taken together and the medium and large sized banks. For small banks it does not have a significant impact under the specification. Thus from looking at table 1 we conclude that among the variables of interest, cost of capital has the most significant impact across all subsamples and the whole sample for the specifications. Labour costs turn out to have a highly significant impact on all samples apart from the small banks. Turning to the coefficients on our controls we also find some interesting information. First of all, loans, our proxy for bank specific risk is highly significant and positive but only for the large banks (under both specifications). But this impact is strong enough to turn the coefficient on the whole sample significant. This implies that for the largest banks taking higher idiosyncratic risks lead to higher interest revenue as expected. A 1% rise in risk leads to a 0.58 percent rise in revenue. But idiosyncratic risks play no significant role in case of revenues of the medium or small sized banks. Looking at the immediately following column of equities, which reflects the general level of risk, we find that this variable is significant for small and medium sized banks, but for large banks general risk seems to have no impact on the revenue. For the overall sample, general level or risk does have a positive impact implying again if the general level of risk bearing rises, then there is an increase in the interest revenue. Therefore, we find that while firm specific or idiosyncratic risk plays a larger role for large banks, for smaller and medium sized banks, general risk is more significant. Finally, we find that total assets are positive and significant for all banks. The larger the bank in terms of total assets, greater is the interest revenue. However, looking at the coefficients minutely, we find that there is a diminishing rate of returns operational on the bank size aspect. Smallest banks have the greatest revenue gain by increasing their total assets while largest banks have smaller impacts of increasing their total assets. Finally, before finalizing these conclusions it is imperative to note that the conclusions are valid only under equilibrium. Looking through the last row of the table, we find that we find no evidence to reject equilibrium at any meaningful level of significance in case of the entire sample of banks and the large bank subsample. It is rejected strongly for small banks and at the 10% level for medium banks. Thus, overall, we should be happy with the validity of these conclusions. Now we turn to table 5. This table presents the calculated values of the H statistics and their significances for the entire sample and the three subsamples under the two null hypotheses. This table presents the main results of this study. The first broad conclusion to note is that both the null hypothesis are rejected and thus the general contestability of the Japanese banking sector has to be admitted to belong to a regime of monopolistic competition. However, the table brings out a number of finer points. Table 5: Testing the contestability in the overall market and in the sub-markets according to bank size H0 All Banks Small Banks Medium Banks Large Banks H=1 30.39*** 41.16*** 35.69*** 8.08*** H=0 31.59*** 3.3* 51.57*** 16.28*** Notes: The numbers in the columns of the table represent the chi-square statistic computed under the respective null hypothesis; *** implies significance at 1%, ** implies significance at 5% and * implies significance at 10%. On the other hand the rejection of the null hypothesis of monopoly (H=0) is not as straightforward. The rejection of monopoly is substantially strong in case of the entire sample, and it is relatively strong for medium and large banks. For small banks however, we can reject the null hypothesis of monopoly only at the 10% level. Therefore, we find indication that although the banking industry falls in the category of monopolistic competition, contestability in these markets is closer to what is observed under monopoly, compared to the extents likely to be found under perfectly competitive equilibrium. Therefore, from the 5th table we can conclude that the degree of contestability in the Japanese banking sector is on the lower end of the spectrum even for monopolistically competitive markets. Table 6: The H statistic over time and across bank types Decade All Banks Small Medium Large 1990-1999 0.5*** 0.192 -0.15* 1.36* 2001-2011 0.24*** 0.196*** 0.37*** 0.3** Notes: *** implies significance at 1%, ** implies significance at 5% and * implies significance at 10%. Finally, we turn to table 6 which presents the computed values of the H statistics (sum of the revenue elasticities of labour, capital and deposit costs) separately for the decade of the 1990’s and the 1st decade of 2000. The first column indicates that while for the overall banking sector the value was 0.5 for the decade of the 1990’s (the lost decade) it came down to 0.24 for the next decade. Since both these values are significant, we can safely conclude that over the two decades that have been problematic for the Japanese in terms of economic growth, the degree of competition or contestability in the banking sector have actually gone down substantially. The other three columns of the table present similar computations for the subsamples of small, medium and large banks. It should be noted here that the subsamples for the decade of 1990-1999 contain substantially fewer observations compared to the other decade. The lack of significance of the estimated coefficients for the subsamples in 1990-1999 could be the result of this low number of observations. So, we will refrain from inferring from these but note that there is clear indication of the fact that for large banks the degree of competition has gone down but it has improved for smaller and medium sized banks over time. But due to the dominance of the large banks in the market, the overall contestability in the markets has fallen substantially. References Bikker, J.A (2004) Competition and efficiency in a unified European banking market, Edward Elgar Bikker, J. A., Shaffer S and Spierdjik, L (2009) “Assessing Competition with the Panzar-Rosse Model: The Role of Scale, Costs, and Equilibrium”, DNB Working Paper No. 225/2009 Claessens, S., and Laeven, L (2004) “What drives bank competition? Some international Evidence”, Journal of Money, Credit, and Banking 36, 563-583. Collender, R. N. and Shaffer, S. (2003) “Local bank office ownership, deposit control, market structure, and economic growth” Journal of Banking and Finance 27, 27-57 De Haas, R., and Lelyveld, I. V., (2006) “Foreign banks and credit stability in Central and Eastern Europe; A panel data analysis”, Journal of Banking & Finance, 30, 1927-1952 Johnson, C., (1982) MITI and the Japanese Miracle: the Growth of Industrial Policy, 1925-1975, Stanford, Stanford University Press. Levine, R., Loayza, N. and Beck, T, (2000) “Financial intermediation and growth: Causality and causes”, Journal of Monetary Economics 46, 31-77. Lerner, A.P., (1934) “The concept of monopoly and the measurement of monopoly power”, Review of Economic Studies 1, 157175 Matthews, K., Murinde, V. and T. Zhao (2007) “Competitive conditions among the major British banks”, Journal of Banking & Finance, 31, 2025-2042 Noguchi, Y (1998) “The 1940 System Japan under the Wartime Economy,” American Economic Review, 88, pp404-416 Panzar, J.C. and J.N. Rosse (1982) “Structure, conduct and comparative statistics”, Bell Laboratories Economic Discussion Paper 248 Panzar, J. and Rosse, J., (1987) “Testing for `monopoly' equilibrium”, Journal of Industrial Economics 35, 443-456 Rosse, J. and Panzar, J., (1977) “Chamberlin vs Robinson: an empirical study for monopoly rents”, Bell Laboratories Economic Discussion Paper Shaffer, S. and DiSalvo, J., (1994) “Conduct in a banking duopoly”, Journal of Banking and Finance 18, 1063-1082 Takahashi, W (2011) "The Japanese financial sector's transition from high growth to the 'lost decades': a market economy perspective", Kobe University RIEB Discussion paper DP2011-29 Wade, R., (1990) Governing the Market: Economic Theory and the Role of Government in East Asian Industrialization, Princeton, New Jersey: Princeton University Press. Tsutsui, Y. and Kamesakab, A., (2005) “Degree of competition in the Japanese securities industry”, Journal of Economics and Business 57, 360-374 Read More