The Effect of Diversification of the Commercial Banking

Transcription

1 The Effect of Diversification of the Commercial Banking Gwon, Eun Ji Abstract The separation of commercial banking and investment bank, initiated by the legislation of the Glass-Steagall Act in 1933, was maintained in the United States until the Gramm-Leach-Bliley Act officially eliminated the boundary in However, commercial banks were effectively diversifying its businesses from as early as the 1980s by taking advantaging of regulatory loopholes or developing innovative financial products. Against this backdrop, it might not be a stretch to suggest that the Glass-Steagall Act had existed only in form. This paper sheds light on how business diversification has affected the commercial banking sector s performance and stability by analyzing the data from the 1980s to the recent years. Because of the possibility of the endogeneity problems, this paper used the structural equation models by two stage least squares methods. The expansion of non-traditional services led to higher profitability, as demonstrated by the proportion of non-interest income to total revenue. On the other hand, no meaningful relation was found between the proportion of non-traditional services and credit risk or capital risk. For the banks that adopted the financial holding company system after the repeal of the Glass-Steagall Act, the increase in non-interest income resulted in lower capital risk but higher credit risk. If diversification is an inevitable trend, an appropriate step would be to find proper ways to manage the risks from diversification. To this end, an appropriate regulation is necessary to lower risks from diversification effectively, while maximizing its benefits. Keywords: Glass-Steagall Act, diversification, bank holding companies, nontraditional banking activities, financial holding companies, financial regulation JEL classifications: N22, G21 This preliminary paper prepared for the Asia-Pacific Economic and Business History Conference 2011, held in Berkeley, California, USA. If you would refer or have any comment on this article, please contact the author via below . BK21 Postdoctoral Fellow, Department of Economics, Seoul National University. eunji7811@gmail.com 1

2 1 Introduction From the 1980s, U.S. commercial banks have provided more diversified, non-traditional banking services, as were reflected in their non-interest income ratio to total revenue. This trend in effect made the Glass-Steagall Act powerless even before its official repeal in 1999 ( Barth et al. (2001), Yeager et al. (2007)). This paper analyzes the relation between the extent of diversification and performance in the U.S. commercial banking sector from 1986 to Because of the endogeneity problems that exist with the variables in question, this paper uses structural equation models by two-stage least squares methods. The results suggest that there are positive relation between diversification and performance. Bank profitability has risen in line with the diversification levels. On the other hand, diversification level has little effect on the bank risks. The paper is organized as follows. Section 2 briefly describes the historical trends in bank diversification. Section 3 describes the data and presents the empirical framework. Section 4 presents the results of the structural equation models and robustness tests. Section 5 summarizes the main findings and discusses their implications. 2 Diversification of Commercial Banking After the Glass-Steagall Act of 1933, commercial banks were separated from investment banking. Commercial banks main source of profits is interest income from loans. However, since the late 1970s, the ratio of non-interest income to total revenue has increased 1. Fig. 1 shows the trends in commercial banks non-interest income share as a percentage of total revenue. In the late 1970s, the ratio of non-interest income to total revenue was less than 10%, but in the early 2000s, this proportion increased to more than 30%. In part, this may reflect a change in commercial banks activities from traditional interest income-generating activities to nontraditional activities, such as investment banking activities, and off-balance-sheet activities, which generate non-interest incomes. This trend is also related to changes in corporate financing. Beginning in the 1980s, the securities market and corporate bond market grew rapidly, and many companies tried to finance their funds via these markets instead of with bank loans. As Gorton and Rosen (1995) has suggested, these changes in corporate financing affected bank profitability, and commercial banks tried to find new profit sources in nontraditional activities. 1 Net interest income equals interest income minus interest expenses, and operating revenue equals net interest income plus non-interest income. 2

3 Figure 1: The Ratio of Noninterest Income to Total Revenue Source: FDIC-Insured Commercial Banks aggregate data from the Historical Statistics on Banking(HSOB) Note: total revenue = interest income + noninterest income Although the Glass-Steagall Act was still law, commercial banks could provide nontraditional services through financial innovation and thus skirt banking laws, or they could use the form of bank holding companies to similar effect. Competition between bank regulators, the Federal Reserve, the Comptroller of the Currency, and the courts also helped commercial banks diversify into nontraditional activities(benston, 1990). The deregulation that occurred beginning in the 1980s also accelerated this trend. The Federal Reserve (Fed) made an amendment to Regulation Y (which regulated commercial bank activities) that facilitated deregulation. In 1987, the Fed allowed some bank holding companies to establish securities affiliates (Section 20 Subsidiaries) to underwrite some kinds of securities, after which the Fed further relaxed the regulations governing what securities activities securities affiliates could engage in. Finally, in 1999, the Glass-Steagall Act was officially repealed with the enactment of the Gramm-Leach-Bliley Act, which was just a rubber stamp on an evolutionary process already well underway (Yeager et al., 2007). There have been many studies of the effect of diversification on commercial banks. In this paper, I focus on studies of the relation between diversification level and bank performance based on accounting data 2. Some studies find that diversification into non- 2 DeYoung and Roland (2001), Smith et al. (2004), Stiroh (2004), and Drucker and Puri (2007) 3

4 traditional activities has the potential to enhance profits and reduce risks ( Kwast (1989), Rogers and Sinkey (1999), and Smith et al. (2004)). Other studies find that diversification can increase bank risk ( DeYoung and Roland (2001), Stiroh (2004), DeYoung and Rice (2004), Stiroh and Rumble (2006), and Stiroh (2006)). The riskiness of a bank is measured using several methodologies and several variables. This is one of the main sources of the differences among the previous studies. To measure returns, some studies use the standard deviation of ROA (or ROE) ( DeYoung and Roland (2001), DeYoung and Rice (2004), and Stiroh and Rumble (2006)), while other studies use the variance in stock market returns (Stiroh, 2006), and still other studies use accounting data ( Rogers and Sinkey (1999) and Yeager et al. (2007)). In this paper, bank risk is measured using accounting data. To use the standard deviation of ROA (or ROE), one must also use other variables averaged for some period, and that process can dilute the apparent relationships between the variables. With respect to regulation, the ability to manage the accounting variables is most important. Thus I chose risk variables using accounting data. What is the most different about previous studies is the methodology used. Endogeneity is an issue when the variables in question are bank profits and risks( Clark (1986), Berger (1995), and Shrieves and Dahl (1992)). However, most studies do not take into account the endogeneity problem. This paper uses the panel instrument variable method and a two stage least squares model with fixed effects. Futhermore, to determine the impact of the repeal of the Glass- Steagall Act, more long-term panel data from 1986 to 2006 are used. 3 Data and Methodology Data Bank holding companies (BHCs) create a way for banks to diversify their activities through their subsidiaries. To analyze the effect of diversification, bank holding companies consolidated financial data are used 3. All BHCs must report their financial data to the Federal Reserve on a consolidated basis in the form of a balance sheet, an income statement, and detailed supporting schedules, including a schedule of off balance-sheet items. Data from the second quarter of 1986 onward are provided on the Federal Reserve Bank of Chicago web page. provide detailed reviews of bank performance and diversification. 3 The BHC data are very similar to those included in the FDIC s Report of Condition and Income For Commercial Bank( Call Report ). However, the Call Report contains only information on individual commercial bank. 4

5 This analysis focuses on the impact of the repeal of Glass-Steagall Act, and I have limited the sample to BHCs established before the repeal year (1999) that were still in existence in 2006, just before the financial crisis. The data do not include information about the year each BHC was established. The Fed provides new identification numbers to new BHCs, or when major structural reorganization takes places in an existing BHC. Thus, I considered the first year for which data were available for the BHC as the year it was founded. I also restricted the sample to surviving BHCs because including those that closed due to failures, mergers and acquisitions would have affected the overall findings. The sample contains 11,902 yearly observations from unbalanced panel data from 1986 to In this paper, when commercial banking is diversified, that means that commercial banking have begun to undertake non-traditional banking activities. Traditional (commercial) banking activities are usually considered to be those in which banks act as intermediaries between savers and borrowers, thus, interest income is closely related to traditional activities. Rogers and Sinkey (1999) describe nontraditional activities as all other fee-generating activities by banks, which range from underwriting activities to cash management and custodial services. In general, non-interest income is related to nontraditional activities. However, non-interest income also includes income from traditional activities, including income from fiduciary activities, service charges on deposit accounts, and trading revenue. A number of banks began as strictly trust companies providing only trustee or fiduciary services ( Kaufman and Mote (1994), p.6), so income from fiduciary activities can be understood as a traditional activity. Trading revenue includes contains the net gains or losses from trading cash instruments and offbalance-sheet derivative contracts. Trading cash instruments and some derivative contracts can be understood as traditional activities. If derivative instruments for interest rate contracts or foreign exchange contracts are used to hedge against interest rate or exchange rate risk, these will not be considered nontraditional activities. However, the brokerage or underwriting of derivative instruments represents a nontraditional activity (Rogers and Sinkey, 1999). Therefore, I use both the ratio of non-interest income and other non-interest income to total revenue as a proxy for each BHC s diversification level. Total revenue equals total interest income plus total non-interest income. Other non-interest income equals total non-interest income minus income from fiduciary activities, service charges on deposit accounts, and trading revenue. Tab. 1 reports the numbers and total assets in the overall BHC data and in the selected sample. For 2006, the selected sample data represent 77% of total BHC assets, 4 BHC data are provided quarterly, so I constructed yearly data using fourth-quarter observations. 5

6 Table 1: Selected Sample Data BHC DATA Selected Sample DATA year Asset($ B) Revenue($ B) Num Nonint Ononint Asset($ B) Revenue($ B) Num Nonint Ononint , ,324 15% 8% % 7% , ,388 16% 9% % 8% , ,420 15% 9% % 7% , ,482 14% 8% % 7% , ,613 16% 9% % 7% , ,620 18% 10% % 8% , ,631 21% 11% 1, % 10% , ,618 24% 12% 1, % 11% , ,357 23% 12% 1, % 11% , ,390 21% 12% 1, % 10% , ,434 23% 13% 1, % 12% , ,513 24% 14% 2, % 12% , ,602 27% 18% 3, % 19% , ,704 31% 20% 4, % 21% , ,782 29% 20% 4, % 20% , ,892 31% 21% 5, % 19% , ,028 34% 23% 5, % 21% , ,185 38% 27% 6, % 24% ,400 1,030 2,301 37% 27% 7, % 24% ,200 1,180 2,310 34% 24% 8, % 22% , % 23% 9, % 21% Source:Bank Hodling Company Data Note: Selected Sample Data contain BHCs established before 1999 that were still in existance in Revenue=interest income + noninterest income ; Nonint = noninterest income / revenue ; Ononint = (noninterest income-trading income-fiduciary income-service charge on deposit) / revenue 6

7 76% of total BHC revenue, and 71% of total BHC numbers. BHC diversification levels measured as non-interest income or other noninterest income as a % of revenue, are almost the same in both datasets. Figure 2: Trends of Diversification: The Ratio of Noninterest Income to Total Revenue Source: Bank Holding Comapany data with selected sample Note: Fidu/Rev = fiduciary income / total revenue ; Servd/Rev = service charge on deposits / total revenue ; Trad/Rev = trading income / revenue ; Ononint/Rev = other noninterest income / total revenu. Fig. 2 displays the trends toward diversification in detail using a stacked area graph. The total area that shows noninterest income as a % of total revenue grew rapidly beginning in the early 1990s, mostly due to the increasing share of other noninterest income (Ononint). It seems not to have been affected by the repeal of the Glass-Steagall Act, Variables and Summary Statistics The focus of this paper is the effect of diversification on bank performance, especially bank profits and bank risks. Using the BHCs financial information, I measure profits as return on assets (ROA). I define this variable as net income before taxes, extraordi- 7

8 nary items, and other adjustments. 5 The risk measures that I consider are capital risk (capital adequacy) and credit risk (credit quality). Capital risk (CAP) is defined by the ratio of total equity capital to total assets, and credit risk (NPL) is defined by the ratio of non-performing loans to total loans 6. The main explanatory variable is diversification level : noninterest income share (Nonint) or other noninterest income share (Ononint) to total revenue. If there exist portfolio diversification benefits, diversification level will be positively related to profits and negatively related to risks. Another main explanatory variable DFHC is designed to measure the impact of the repeal of the Glass-Steagall Act. In 1999, the Gramm-Leach- Bliley Act (GLBA) permitted BHCs to become a financial holding companies (FHCs) and thus to engage in a greatly diversified range of activities, including securities, insurance, and merchant banking activities. DFHC is a dummy variable with a value of 1 if a BHC became a financial holding company and 0 otherwise. If the repeal of the Glass-Steagall Act (GLBA) has had an effect on bank performance, such an effect may be visible in BHCs that became financial holding companies. To distinguish the effect of a firm s becoming a financial holding company from the effect of its newly diversified range of activities in the wake of its transformation into a FHC, I multiply DFHC by the variables indicating the diversification level. Each equation has instrumental variables, Effratio, Depo, Loan, and Allow. Effratio is the ratio of non-interest expenses to operating revenue. Operating revenue equals total revenue less interest expenses. Depo is the ratio of deposits to assets. Loan is the ratio of loans to assets. Allow is the ratio of allowances for loans to loans. All of the equations contain common control variables. To control for the size effect, log of total assets and the square value of log asset are used. To control for macroeconomic and geographical fluctuation, year dummy variables and state dummy variables are used. Tab. 2 reports summary statistics for all of the variables used in the structural equation models. The Total BHC sample includes 11,902 observations from 1986 to 2006, and the FHC only sample includes 1,472 observations from 2000 to 2006, which contain only FHCs transformed from within Total BHC sample. In the Total BHC sample, the mean non-interest income share is 14%, with considerable variation (from -5% to 92%). This variation is mainly due to other non-interest income share; the mean of 5 In the BHC data, Schedule HI(consolidated income statement) is net income before taxes, extraordinary items, and other adjustments. It is the sum of net interest income and net non-interest income minus provisions. 6 Non-performing loans are those loans marked past due 90 days or more and still accruing and nonaccrual on the consolidated balance sheet. 8

9 Table 2: Summary Statistics TOTAL FHCs mean median min max s.d. N mean median min max s.d. N ROA , ,472 CAP , ,472 NPL , ,472 Nonint , ,472 Ononint , ,472 Trad , ,472 Fidu , ,472 Servd , ,472 ln(asset) , ,472 Effratio , ,472 Depora , ,472 Loanra , ,472 Allow , ,472 Note: Total contains selected sample data from 1986 to FHCs contains Financial Hodling Comapany Data in the Total, from 2000 to

10 other non-interest income share is 7%, and again the range is wide at -9% to 88%. The FHC only sample statistics are similar to those of Total BHC, but FHCs have more noninterest revenue, more other noninterest revenue, and more assets than BHCs 7. It is natural that FHCs are larger in size and have more diversified activities. Empirical Framework According to the studies by Clark (1986), Berger (1995), and Shrieves and Dahl (1992), bank profits are determined simultaneously with bank risks. Thus, if these variables are used in the regression models, the endogeneity problem must be taken into account. However, many of the studies of diversification effects have not considered this problem. In this paper, to resolve the endogeneity problem, I have constructed the following structural equation models, because the equations include endogenous variables, simultaneous equations for Model I and Model II are calculated using two-stage least squares for panel data models 8. Structural equation model I (1) ROA i,t = α i + β 1 CAP i,t + β 2 NP L i,t + β 3 Nonint i,t + β 4 DF HC i,t + β 5 DF HC i,t Nonint i,t + β 6 ln (Asset i,t ) + β 7 (ln (Asset i,t )) 2 + β 8 Effratio i,t + ε i,t (2) CAP i,t = α i + β 1 ROA i,t + β 2 NP L i,t + β 3 Nonint i,t + β 4 DF HC i,t + β 5 DF HC i,t Nonint i,t + β 6 ln (Asset i,t ) + β 7 (ln (Asset i,t )) 2 + β 8 Depo i,t + β 9 Loan i,t + ε i,t (3) NP L i,t = α i + β 1 ROA i,t + β 2 CAP i,t + β 3 Nonint i,t + β 4 DF HC i,t + β 5 DF HC i,t Nonint i,t + β 6 ln (Asset i,t ) + β 7 (ln (Asset i,t )) 2 + β 8 Loan i,t + β 9 Allow i,t + ε i,t Structural equation model II (4) ROA i,t = α i + β 1 CAP i,t + β 2 NP L i,t + β 3 Ononint i,t + β 4 T rad i,t + β 5 F idu i,t + β 6 Servd i,t + β 7 DF HC i,t + β 8 DF HC i,t Ononint i,t + β 9 DF HC i,t T rad i,t + β 10 DF HC i,t F idu i,t + β 11 DF HC i,t Servd i,t + β 12 ln (Asset i,t ) + β 13 (ln (Asset i,t )) 2 + β 14 Effratio i,t + ε i,t 7 Not reported in Tab. 2; if FHC s statistics are compared with those of BHC during the same period, from 2000 to 2006, the results are the same. 8 The results of the Hausman specification test for Models I and II obtained by comparing the two-stage least squares estimates to the ordinary least squares estimates for panel data models also confirm that the two-stage least squares estimates are more efficient. 10

11 (5) CAP i,t = α i + β 1 ROA i,t + β 2 NP L i,t + β 3 Ononint i,t + β 4 T rad i,t + β 5 F idu i,t + β 6 Servd i,t + β 7 DF HC i,t + β 8 DF HC i,t Ononint i,t + β 9 DF HC i,t T rad i,t + β 10 DF HC i,t F idu i,t + β 11 DF HC i,t Servd i,t + β 12 ln (Asset i,t ) + β 13 (ln (Asset i,t )) 2 + β 14 Depo i,t + β 15 Loan i,t + ε i,t (6) NP L i,t = α i + β 1 ROA i,t + β 2 CAP i,t + β 3 Ononint i,t + β 4 T rad i,t + β 5 F idu i,t + β 6 Servd i,t + β 7 DF HC i,t + β 8 DF HC i,t Ononint i,t + β 9 DF HC i,t T rad i,t + β 10 DF HC i,t F idu i,t + β 11 DF HC i,t Servd i,t + β 12 ln (Asset i,t ) + β 13 (ln (Asset i,t )) 2 + β 14 Loan i,t + β 15 Allow i,t + ε i,t 4 Empirical Results Tab. 3 presents the regression results for Models I and II. Because of the endogeneity problem, panel instrument variable model with two-stage least squares regression method is used 9. To properly account for time-independent effects and individual firmspecific properties, a fixed effect model is used 10. In Tab. 3, equation (1) shows the relationship between bank profits and bank activities. Non-interest income share (Nonint) has a positive and significant relationship with bank profits. As expected, more diversified banks tend to be more profitable. Shifting one percentage point of revenue from interest income to non-interest income will increase ROA at the median BHC in the sample by about 2% (=0.03*0.01/0.015). FHC status (DFHC) has little effect on bank profits. FHC non-interest income share (DFHC*Nonint) also has no significant relation to ROA. This means that the repeal of the Glass-Steagall act did not affect bank profitability. Because, as pointed out by Yeager et al. (2007), BHCs already conducted considerable non-traditional activities before the repeal of the law, diversification level affected bank profitability more than organizational form. Equation (2) reveals the relation between capital risk and bank activities. Non-interest income share (Nonint) has a positive but insignificant relationship to bank capital ratio. The capital ratio is usually regulated by BIS criteria, so diversification has little effect on capital risk. Although FHC status has a significant negative relationship with the capital ratio, and higher FHC non-interest income share is correlated with higher capital 9 To test the appropriateness of the instrumental variables, underidentification test (LM test) is performed for all equations. In all equations, the null hypothesis in the underidentificaion test is rejected, indicating that the instruments are not underidentified. For the equation (1) and (4), which are overidentified equations, the Sargan-Hansen overidentified test is performed. The Sargan-Hansen test does not reject the null hypothesis that the instruments are valid, implying that the instruments are exogenous. Those statistics are omitted in Tab Based on the Hausman specificaion test of the Models, the null hypothesis that the individual effects are uncorrelated in rejected, so I can confirm that the fixed effect model is more efficient 11

12 Table 3: Regression Results Structural Equation Model I Structural Equation Model II (1) ROA (2) CAP (3) NPL (4) ROA (5) CAP (6) NPL ROA 1.122*** *** 1.110*** *** (0.038) (0.037) (0.039) (0.037) CAP 0.176*** *** (0.019) (0.040) (0.019) (0.041) NPL *** *** (0.026) (0.042) (0.026) (0.042) Nonint 0.030*** ** (0.001) (0.004) (0.002) Ononint 0.031*** ** (0.002) (0.004) (0.002) Trad 0.016* (0.008) (0.026) (0.015) Fidu 0.022*** 0.035*** (0.004) (0.010) (0.006) Servd 0.029*** *** (0.003) (0.010) (0.005) DFHC ** (0.000) (0.001) (0.001) (0.000) (0.001) (0.001) DFHC*Nonint *** 0.008*** (0.001) (0.005) (0.003) DFHC*Ononint *** 0.009** (0.002) (0.006) (0.004) DFHC*Trad (0.009) (0.029) (0.016) DFHC*Fidu *** 0.015*** (0.003) (0.010) (0.006) DFHC*Servd *** (0.005) (0.015) (0.008) Num 11,902 11,902 11,902 11,902 11,902 11,902 Adjusted R Note: This table shows the coefficients of the Models I and II using the panel instrument variable models(two stage least squares) with fixed effects. The coefficients of the control variables and istrument variables are not reported. Standard errors are in parentheses. DFHC is an indicator variable equal to one if the bank holding company is an Financial Holding Company in a given year, and equal to zero otherwise. * p < 0.1, ** p < 0.05, *** p <

13 ratios (indicating less capital risk), the economic significance of the coefficients is small. If the FHC s non-interest income share increases by one percentage point more, this represents an increase in the median FHC capital ratio of just 0.2% (=0.018*0.01/0.087). Equation (3) shows the effect of diversification on credit risk. Non-interest income share and FHC non-interest income share are positively and significantly related to nonperforming loan share. This means that FHCs experience greater credit risk than do other BHCs. For the median BHC, a 1% point increase in non-interest income share is associated with an increase of just 0.7% (=0.005*0.01/0.007) in NPL. However, the median FHC s NPL increases about 2% (= ( )*0.01/0.006). As previously mentioned, non-interest income includes some kinds of traditional income. Thus, in Model II, more detailed non-interest income shares are used. I focus on the other non-interest income share, which excludes some aspects of traditional incomes, including trading revenue, fiduciary revenue, and service charges on deposits from the non-interest income. The results are similar to the Model I. Diversified BHCs exhibit greater profitability but little change in capital risk and credit risk. A 1%-point increase in the other non-interest income share enhances median bank profitability about 2% (=0.031*0.01/0.015), but increases bank credit risk by only 0.6% and has no effect on capital risk. However, if a FHC s other non-interest income share increases, credit risk does increase about 2% (= ( )*0.01/0.006). Next, the results of some robustness tests are presented in Tab. 4 and Tab. 5. (A) uses the basic sample (used in Tab. 3) for a longer period (extended until 2009). (B) uses balanced panel data that contains BHCs that existed throughout the whole period, (C) uses all of the BHC data regardless of prior sample criteria. The results are quite similar to those in Tab. 3. BHC s ROA is positively and significantly related to diversified activities, both those included in the non-interest income share and the other non-interest income share, for all cases. The non-performing loan ratio is also positively related to diversification, but its economic significance is very low. The capital ratio has a little relation with diversification levels. For subsample (B) and in the overall BHC data (C), survivor bias is present. (B) contains BHCs that had been in existence for 21 years, so they are theoretically healthy banks with a sufficient capital bumper (capital ratio) and less non-performing loans. In contrast, (C) contains short-lived banks that failed or merged, and exhibited poor balance sheet structure. Because of survivor bias, diversification level in (B) is more positively related to capital ratio and less related to non-performing loans. However, in (C), diversified BHCs are less related to capital ratio and more positively related to non-performing loans. 13

14 Table 4: Robustness Tests for Structural Equation Model I ROA CAP NPL (A) (B) (C) (A) (B) (C) (A) (B) (C) Nonint 0.029*** 0.032*** 0.046*** *** *** 0.007*** *** (0.001) (0.003) (0.001) (0.003) (0.005) (0.003) (0.002) (0.004) (0.002) DFHC ** *** *** *** * (0.000) (0.001) (0.000) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) DFHC*Nonint *** 0.013*** 0.036*** 0.020*** 0.021*** (0.002) (0.003) (0.002) (0.004) (0.006) (0.004) (0.003) (0.005) (0.003) Num 13,897 5,625 33,527 13,897 5,625 33,527 13,897 5,625 33,527 Num of groups , , ,028 Adjusted R Note: (A) regresses selected sample data in Tab. 3 to the latest period, from 1986 to (B) contains BHCs that existed throughout the whole period, (C) uses all of the BHC data regardless of prior sample criteria. Other variables used in the Structural Equation Model I are not reported here. Standard errors are in parentheses. * p < 0.1, ** p < 0.05, *** p <

15 Table 5: Robustness Tests for Structural Equation Model II ROA CAP NPL (A) (B) (C) (A) (B) (C) (A) (B) (C) Ononint 0.031*** 0.033*** 0.051*** *** *** 0.007*** *** (0.001) (0.003) (0.002) (0.004) (0.005) (0.003) (0.003) (0.004) (0.002) Trad *** * * 0.053*** (0.009) (0.012) (0.007) (0.026) (0.033) (0.014) (0.019) (0.023) (0.010) Fidu 0.019*** 0.019*** 0.021*** 0.025** 0.076*** 0.028*** ** (0.004) (0.007) (0.005) (0.011) (0.016) (0.010) (0.008) (0.013) (0.007) Servd 0.017*** 0.030*** 0.015*** *** (0.003) (0.004) (0.004) (0.009) (0.011) (0.008) (0.007) (0.008) (0.006) DFHC * *** ** *** (0.000) (0.001) (0.001) (0.001) (0.002) (0.001) (0.001) (0.001) (0.001) DFHC*Ononint * *** 0.034*** 0.084*** 0.035*** 0.033*** (0.002) (0.005) (0.002) (0.006) (0.007) (0.005) (0.004) (0.009) (0.004) DFHC*Trad *** * (0.009) (0.018) (0.013) (0.027) (0.046) (0.027) (0.020) (0.034) (0.019) DFHC*Fidu *** ** ** *** (0.003) (0.005) (0.005) (0.010) (0.012) (0.009) (0.007) (0.009) (0.006) DFHC*Servd (0.005) (0.008) (0.007) (0.014) (0.021) (0.013) (0.010) (0.015) (0.009) Num 13,897 5,625 33,527 13,897 5,625 33,527 13,897 5,625 33,527 Adjusted R Note: (A) regresses selected sample data in Tab. 3 to the latest period, from 1986 to (B) contains BHCs that existed throughout the whole period, (C) uses all of the BHC data regardless of prior sample criteria. Other variables used in the Structural Equation Model II are not reported here. Standard errors are in parentheses. * p < 0.1, ** p < 0.05, *** p <

16 Diversified activities on the part of FHCs are consistently related to higher capital ratios. This may reflect that the transition from BHC to FHC requires the satisfaction of certain capital requirements; as a result, FHCs which more diversified activities should expected to have higher capital ratios. This suggests that with proper regulation criteria can affect BHC risk management. 5 Conclusion This paper examines the effect of diversification on bank performance. The data analysis shows that diversification contributes to profits. The more diversified BHCs are to the nontraditional activities, the more profits they produce. On the other hand, diversification has little effect on bank risk, such as capital risk and credit risk. The relation between diversification and risks is not economically significant. After the repeal of the Glass-Steagall Act in 1999, FHCs could provide almost all kind of financial services. Their diversified activities lessened capital risk, but heightened credit risk. These results are related to regulation structures. Commercial banks capital ratios have been strictly regulated, and if a BHC wants become a FHC, it must satisfy the criteria for capital adequacy. This explains why FHC s diversified activities may lower capital risk. If the credit risk, such as non-performing loan ratio, is regulated similarly accordingly, it may not be increased by diversified activities. The Glass-Steagall Act, introduced to prevent bank crisis by separating commercial banking and investment banking was not a success. Before the repeal of the Glass- Steagall Act, many banks were already providing more diversified non-traditional services on the back of financial innovations and deregulations. If diversification is an inevitable trend, an appropriate step would be to find proper ways to manage the risks from diversification. To this end, an appropriate regulation is necessary to lower risks from diversification effectively, while maximizing its benefits. 16

17 Bibliography Barth, James R., Gerard Caprio, and Loss Levine, Banking systems around the globe: Do regulations and ownership affect performance and stability?, in F. S. Mishkin, ed., Prudential Supervision: What Works and What Doesn t, Univ. of Chicago Press, 2001, pp Benston, George J., The Separation of Commercial and Investment Banking: The Glass- Steagall Act Revisited and Reconsidered, Oxford University Press, USA, August Berger, Allen N., The Relationship between Capital and Earnings in Banking, Journal of Money, Credit and Banking, 1995, 27 (2), Clark, J., Single-equation, multiple-regression methodology: Is it an appropriate methodology for the estimation of the structure-performance relationship in banking?, Journal of Monetary Economics, November 1986, 18 (3), DeYoung, R. and K. P. Roland, Product Mix and Earnings Volatility at Commercial Banks: Evidence from a Degree of Total Leverage Model, Journal of Financial Intermediation, January 2001, 10 (1), DeYoung, Robert and Tara Rice, Noninterest Income and Financial Performance at U.S. Commercial Banks, The Financial Review, February 2004, 39 (1), Drucker, Steven and Manju Puri, Banks in Capital Markets, in Espen B. Eckbo, ed., Handbook in Corporate Finance: Empirical Corporate Finance, North-Holland, 2007, pp Gorton, Gary and Richard Rosen, Corporate Control, Portfolio Choice, and the Decline of Banking, The Journal of Finance, 1995, 50 (5), Kaufman, George G. and Larry R. Mote, Is banking a declining industry? A historical perspective, Economic Perspectives, 1994, 18 (May), Kwast, M., The impact of underwriting and dealing on bank returns and risks, Journal of Banking & Finance, March 1989, 13 (1), Rogers, Kevin and Joseph F. Sinkey, An analysis of nontraditional activities at U.S. commercial banks, Review of Financial Economics, June 1999, 8 (1), Shrieves, R. and D. Dahl, The relationship between risk and capital in commercial banks1, Journal of Banking & Finance, April 1992, 16 (2),

18 Smith, Rosie, Christos Staikouras, and Geoffrey E. Wood, Non-Interest Income and Total Income Stability, Social Science Research Network Working Paper Series, April Stiroh, Kevin J., Diversification in Banking: Is Noninterest Income the Answer?, Journal of Money, Credit and Banking, 2004, 36 (5), , A Portfolio View of Banking with Interest and Noninterest Activities, Journal of Money, Credit and Banking, 2006, 38 (5), and Adrienne Rumble, The dark side of diversification: The case of US financial holding companies, Journal of Banking & Finance, August 2006, 30 (8), Yeager, Timothy J., Fred C. Yeager, and Ellen Harshman, The Financial Services Modernization Act: Evolution or Revolution?, Journal of Economics and Business, July 2007, 59 (4),