Are Bigger Banks More Profitable than Smaller Banks?

Transcription

1 Journal of Applied Finance & Banking, vol., no.3, 0, 59-7 ISSN: (print version), (online) International Scientific Press, 0 Are Bigger Banks More Profable than Smaller Banks? Matthew C. Chang, Chien-Chung Nieh and Ya-Hui Peng 3 Abstract In this study, we apply Panel Threshold Model (Hansen, 999) to examine whether there is optimal asset scale for interest spread to affect banks' profs. The empirical results demonstrate the existence of three thresholds, which divide banks into four groups into four groups according to asset scale. When asset scales of banks are in the 3rd capal group, banks prof by the widening in loan-depos interest spread. For the other three groups, however, the relationship between prof and loan-depos spread is negative. Banks' return on equy (ROE) is posively correlated wh net commission income, net invest income, net non-operating income, and net interest income. JEL classification numbers: G, G8 Keywords: Panel threshold model, Interest spread, Bank profabily Department of Finance and Banking, Hsuan Chuang Universy, No. 48, Hsuan Chuang Rd., Hsinchu 300, Taiwan, a04979@gmail.com Department of Banking and Finance, Tamkang Universy, No. 5 Ying-chuan Rd., Tamsui District, New Taipei Cy 5, Taiwan. 3 Department of Finance and Banking, Hsuan Chuang Universy, No. 48, Hsuan Chuang Rd., Hsinchu 300, Taiwan. Article Info: Revised : October 4, 0. Published online : November 30, 0

2 60 Are Bigger Banks More Profable than Smaller Banks? Introduction In the 970s, developed countries in the western world started for financial deregulation, and most countries follow the trend of financial liberalization. Facing the rapid changes in financial environment, Taiwan also deregulated her restrictions on financial instutions and markets. In 005, there are 44 commercial banks in Taiwan. Under the competive environment, banks tend to narrow down the loan-depos spread to widen their market shares, and further makes banks less profable and increasing in the non-performing loan ratio. Taiwan is one of the most overbanking and fragmented country in Asia. Although the leading commercial banks control /3 of Taiwan's banking assets, some of the smaller banks are likely to be absorbed by stronger players or forced out of the industry because of fierce competion, eroding pricing power and prof margins, and many of the banks' best corporate customers are decamping to mainland China. As Taiwanese banks have been forbidden to follow their customers, the authories in Taipei are pressing for banking reform. Domestic bank consolidation in Taiwan began in 00 when the Financial Holding Company Act was enacted. This law encourages consolidation in the industry by allowing the formation of financial holding companies. In 00, the government announced a renewed push to consolidate s banking sector, encouraging combinations between financial holding companies as well as acquision of smaller banks. The goal of the reform is to reduce the existing 4 financial holding companies in Taiwan to seven by the end of 006. Domestic consolidation in Taiwan is likely to occur eher through merger among the big state-owned banks, or by private-sector acquision of state-owned banks. The state-sector restructuring will be driven by the government which has announced the goal to reduce the number of state-run banks from to six by end of 005 wh the establishment of an M&A task force for the banking industry. Market forces will drive private-sector acquisions of state-owned banks, and takeover action will happen only when further regulatory reform becomes affirmative. Few

3 M.C. Chang, C.C. Nieh and Y.H.Peng 6 private-sector banks would be willing to risk an acquision whout a government guarantee that the state would absorb future losses from emerging non-performing loans. In this paper, we explore whether the impact of loan-depos spread on banks' profabily is different for banks wh different sizes. Hameetemam et al. (000) find negative relationship between bank size and net prof before income tax. Similarly, Allen and Rai (996) show that small banks have advantages for economy of scale. On the contrary, Hunter and Timme (986) indicate posive relationship for size of bank assets and non-tradional banking prof. Haslem et al. (983) also show that the better a bank performs, the higher the more stable net interest margin. Generally speaking, leratures distinguish banks' business into tradional depos and loan, and non-tradional business. Diamond (984) and Gorton and Rosen (995) point out that a bank's role is an intermediate to transfer money from deposors to borrowers. Relatively, Rogers and Sinkey (999) define non-tradional activies of banks as generating income from activies, including stock underwring, cash management and wealth management. Rogers (998) measure banks' revenue, prof, and cost efficiency, and shows that banks' prof efficiency and cost efficiency are increased and revenue efficiency is reduced if non-tradional banking income is included in the model. It implies that banks engaging in non-tradional banking increase prof efficiency and cost efficiency but reduce revenue efficiency. Rogers and Sinkey (999) apply regression analysis to investigate 8,93 commercial banks in the U.S. from 989 through 993. They show that the degree banks involved in non-tradional is posive to their scales. In all tradional indicators of profabily, return on equy (ROE) is generally considered as the most important and representative indicators. Venkartraman and Ramanujam (986) divide performance into three types: financial, business, and organizational. Financial performance, which commonly includes return on total assets (ROA), ROE and earnings per share (EPS), is the measurement to achieve

4 6 Are Bigger Banks More Profable than Smaller Banks? business financial objectives. Brewer (990) concludes the negative relationship between risk and investment in non-banking subsidiary of banks. Although Demirguc-Kunt and Huizinga (999) point out that return-adjusted return on capal (RAROC) is a better measurement that ROA because most banks in developing are protected by government and thus make their capal low and result in banks' ROE over-stated. However, the Taiwanese government does not involve in banks' capal management after the financial reform and the opening up of financial services industry. Therefore, we apply ROE as the proxy for profabily of banks. Olson and Sollenberger (978) decompose revenue changes of a bank into operating income and non-interest income. Furthermore, Ho and Sauders (98) point out that the bank interest margin (i.e., spreads) depends on market structure, competion, the average amount per transaction, and assets and liabilies nominal interest rates. Haslem et al. (983) conclude that better performed banks have higher net interest margin and stable growth rate. In addion, better performed banks reduce their operating costs by controlling their non-interest expense. Flannery (98) indicates that most commercial banks in the U.S. tend to borrow in long-term and loan in short-term, and short-term and long-term interest rate fluctuations do not significantly influence banks' profs. Samuelson (945) show banks prof more in interest rates rising than in interest rates descending, and commercial banks are more profable than savings banks. Similarly, Hancock (985) demonstrates that raise in interest rates will increase the profs of banks, namely, to lending rate elasticy is higher than depos rate elasticy. However, Silverberg (973) holds the oppose view that interest rates will reduce bank profs and increase the risk. Hamweck and Kilcollin (984) show that small banks have short-term advantage when interest rates rise. Moreover, descending in interest rates may result in slow loan growth and loan losses. For large banks, net interest income is less sensive to the change of interest rates. For smaller banks, however, the net interest income decreasing may result in reducing loan growth

5 M.C. Chang, C.C. Nieh and Y.H.Peng 63 and loan losses during the period of decline in interest rates. Thus small banks are difficult to maintain profabily during market interest rates decline than raise. Data Other than tradional studies that decompose banks profs into tradional and non-tradional profs, we break banks profs down into net interest income, net fee income, net investment income, net operating income. In addion, we add ratio of overdue loans, rediscount rate, asset size, and loan-depos interest spread as control variables. The dependent variable is ROE, which is banks profabily proxy concerned by investors. We then explore if there is a threshold of asset size, and further a threshold of interest margin to influence banks profabily. We obtain the data from Taiwan Economic Journal (TEJ), and the study period is from the st of 000 through the 4th quarter 009. We focus on the commercial banks, thus the industrial banks are excluded. Therefore, this study sample includes 34 banks wh a total of 40 quarters. The data are cross-sectional and longudinal time series, namely balance panel data. 3 Methodology: Threshold Autoregression (TAR) In this paper, spreads on bank deposs and loans interest rates may differently affect banks profabily for different asset sizes. In general, the larger size of a bank s assets, the greater economy of scale and profabily of the bank. The depos and loan interest rate spreads have a posive impact on bank profabily if other condions remain unchanged. However, a bank can hardly depos and loan if s asset scale is not large enough when the spread of the bank deposs and loans interest rate is wide. Furthermore, as a bank increases s asset size, the bank loan to low-creded and thus increasing in asset may result in

6 64 Are Bigger Banks More Profable than Smaller Banks? diminishing marginal utily-scale. We expect there is a threshold of the size of bank assets. The relationship between spreads of banks deposs and loans interest rates and profabily is negative below the threshold, while the relationship is posive between the threshold and the threshold. Due to over-increased asset size, the relationship is negative for the asset size beyond the threshold. According to Hansen (999), we set the empirical threshold model as follows: i h s if s v () i h s if s (, ), h ( SP, NC, NII, NIR, NOI, NNOR, NPL, RR ) where v is ROE, s is asset scale and the measure of threshold, is the threshold value. h is the control variable vector, which includes: () SP : spread on bank deposs and loans interest rate; () NC : net fee income; (3) NII : net interest income; (4) NIR : net investment income; (5) NOI : other net revenue income; (6) NNOR : net non-revenue income; (7) NPL : over-due loan ratio; (8) RR : market interest rate. In addion, i is the fixed effect to capture the heteroskedastivy of banks under different management, is the error term, and the expected value is 0, and () is identical to: is the variance and i. ~ iid(0, ) p ' h s I s s I s ()

7 M.C. Chang, C.C. Nieh and Y.H.Peng 65 where I(.) is an indicator function. That is, where p h s, or i v x (3) i si s s and ',, ' ', ', and x h, s. si s Since there may exists three threshold, () may be modified to: i h d if d i h d if d (4) i h 3d if d 3 i h 4d if d 3 where 3. Furthermore, we may extend the model to mutli-threshold:,,, n ). ( Empirical Results 4. Single threshold model As Table shows, the asset size threshold ( ) is The threshold divide the observations into two blocks: when the asset size ( r ) is less than , the t-statistic of ' in homogeneous and heterogeneous standard deviation is and respectively, and they are significant at % level. It shows highly significant and negatively correlated. Similarly, when the ' asset size ( r ) is greater than , the t-statistic of in homogeneous and heterogeneous standard deviation is and -3. respectively, and they are also significant at % level. Since ' and ' are both significantly negative at % level, demonstrates that the relationship for the impact of depos and loan

8 66 Are Bigger Banks More Profable than Smaller Banks? interest rate spread on a bank's prof is negative, no matter s asset size is below or beyond the threshold. Table : Single Threshold Model Threshold Estimate Homogeneous Hetergeneous Value t-statistic t-statistic S.D. S.D. ˆ ' *** *** ˆ ' *** -3. *** Note: ***, **, * respectively indicates significance at, 5, 0 percent. ' The parameter ˆ and ˆ ' denote for the regressor coefficient below and beyond the threshold respectively. 4. Multiple threshold model We estimate the thresholds by (4). As Table shows the three thresholds which are and , which divide the observations into four blocks: the first is the case as asset size is less than , the second is the case as asset size is between , , and In the first, the second, and the ' ' third case, we find that ˆ, ˆ, and ˆ 4 are all significantly negative at 5% level ' ( ˆ and ˆ ' at % level) for both homogeneous and heterogeneous standard deviation estimation. However, we find that 3 ˆ is significantly negative at % level for both homogeneous and heterogeneous standard deviation estimation. Figures and tables should be placed in the middle of the page between left and right margins. Reference to the figure in the text should use "Figure". In the final formatting of your paper, some figures may have to be moved from where they appeared in the original submission. Figures and tables should be sized as they are to appear in print. Figures or tables not correctly sized will be returned to the author for reformatting.

9 M.C. Chang, C.C. Nieh and Y.H.Peng 67 Table : Multiple Threshold Model Fixed Tresholds Threshold Estimate Thresholds Homogeneous Hetergeneous Value t-statistic t-statistic S.D. S.D. ˆ ' *** *** ˆ ' ** ** 3 ˆ *** *** ˆ *** *** 4 Note: ***, **, * respectively indicates significance at, 5, 0 percent. ' The parameter ˆ ', ˆ, 3 ˆ and 4 ˆ denote for the regressor coefficient for the four thresholds respectively. Therefore, the empirical results show that when the asset size is less than , the depos and loan interest rates spread for bank profabily has a negative correlation between the asset. On the other hand, if the asset size is greater than and less than 5.005, the relationship is posive. Furthermore, the relationship becomes negative again when the asset size is greater than Robustness Table 3 demonstrates the comparison for single, double, and triple-threshold models, and we apply the 00 times bootstrap. As Table 3 shows, the hypothesis of no double-threshold is not rejected at 0% significance level. The hypotheses of single and triple-threshold, however, are rejected at 0% significance level. Furthermore, we find that the F-Statistic of the triple-threshold model is largest among the three models, and implies that we may reject the null hypothesis of no threshold effects for triple thresholds at a higher significance level.

10 68 Are Bigger Banks More Profable than Smaller Banks? Table 3: Robustness Test of Thresholds Single Threshold Double Thresholds Triple Thresholds and , , and F-Statistic F-Statistic.87 F-Statistic p-value 0.00 p-value 0.7 p-value 0.00 Crical 0% 5.84 Crical Crical 0% value value % value 5% % 7.0 5% % % % Note: ***, **, * respectively indicates significance at, 5, 0 percent. F-Statistics and p-values are obtained from bootstrapping 00 times. As Table 4 shows, we find that net fee income, net interest income, net investment income, other net revenue income, and net non-revenue income have posive impacts on banks ROE at 5% significance level. In addion, over-due loan ratio has negative impact on ROE. These findings are straight forward and shows consistency for the triple-threshold model. Table 4: Tests for Control Variables: Triple Thresholds Value Homogeneous S.D. Hetergeneous S.D. t-statistic t-statistic ˆ ˆ *** ˆ *** ˆ *** ˆ *** ˆ ** ˆ -.95 ** ˆ ***, **, * respectively indicates significance at, 5, 0 percent

11 M.C. Chang, C.C. Nieh and Y.H.Peng 69 5 Conclusion This study aims to analyze the profabily of banks by Hansen (999) through asset size as threshold to empirically examine the existence of asymmetric non-linear relationship. We take ROE as a proxy for bank profabily, bank assets as the threshold variable. The empirical results show that there are triple thresholds, the threshold values are , , and 5.005, the size of the asset banks is below the threshold or greater than 5.005, the relationship between bank's depos and loan interest spreads and bank's profabily is negative. However, if is between and 5.005, the relationship is posive. Thus, demonstrates the optimum size of banks. In addion to depos and loan interest spreads, fee income, net investment income, net non-operating income, other net operating income, net interest income also influence bank profs, and overdue loans reduce banks' profabily. References [] L. Allen and A. Rai, Operational efficiency in banking: an international comparison, Journal of Banking and Finance, 0, (996), [] A. Demirguc-Kunt and H. Huizinga, Determinants of commercial bank interest margins and profabily: some international evidence, World Bank Economic Review, 3, (999), [3] E. Brewer, The risk of banks expanding their permissible nonbanking activies, Financial Review, 5, (990), [4] K.S. Chan, Consistency and liming distribution of the least squares estimator of a continuous threshold autoregressive model, Annals of Statistics,, (993),

12 70 Are Bigger Banks More Profable than Smaller Banks? [5] D.W. Diamond, Financial intermediation and delegated monoring, World Bank Economic Review, 3, (984), [6] M.J. Flannery, 983, Interest rates and bank profabily: addional evidence, Journal of Money, Cred and Banking, 5, [7] G. Gorton and R. Rosen, Coporate control, portfolio choice and the decline of banking, Journal of Finance, 50, (995), [8] B.E. Hansen, Threshold effects in non-dynamic panels: estimation, testing and inference, Journal of Econometrics, 93, (999), [9] D. Hameeteman and B. Scholtens, Size, growth, and variance among the world's largest non-merged banks, International Journal of the Economics of Business, 7, (000), [0] D. Hancock, Bank profabily interest rates and monetary policy, Journal of Money, Cred and Banking, 5, (985), [] G.A. Hamweck and T.E. Kilcollin, Bank profabily and interest rate risk, Journal of Economics and Business, 36, (984), [] T.S.Y. Ho and A. Saunders, The determinants of bank interest margins: theory and empirical evidence, Journal of Financial and Quantive Analysis, 6, (98), [3] J.A. Haslem, J.P. Bedingfield and A.J. Stagliano, Bank performance measures and relative profabily, Bankers Magazine, 40, (983), [4] E.I. Kaparakis, S.M. Miller and A.G. Noulas, Short-run cost inefficiency of commercial banks: a flexible stochastic frontier approach, Journal of Money, Cred, and Banking, 6, (994), [5] Y. Li, The Asian financial crisis and non-performing loans: evidence from commercial banks in Taiwan, International Journal of Management, 0, (003), [6] R.W. McShane, I.G. Sharpe, A time series/cross section analysis of the determinants of Australian trading bank loan/depos interest margin: 96-98, Journal of Banking and Finance, 9, (985), 5-36.

13 M.C. Chang, C.C. Nieh and Y.H.Peng 7 [7] R.L. Olson and H.M. Sollenberger, Interest margin variance analysis: a tool for current times, Magazine of Bank Administration, 5, (978), [8] K. Rogers and J.F. Sinkey, An analysis of nontradional activies at US commercial banks, Review of Financial Economics, 8, (999), [9] K.E. Rogers, Nontradional activies and efficiency of US commercial banks, Journal of Banking and Finance,, (998), [0] P.A. Samuelson, The effect of interest rate increases on the banking system, American Economic Review, 35, (945), 6-7. [] S.C. Silvefrberg, Depos costs and bank portfolio policy, Journal of Finance, 8, (973), [] N. Venkatraman and V. Ramanujam, Measurement of business performance in strategy research: a comparison of approaches, Academy of Management Review,, (986),