The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
S M 8 8 -5 c 4 > E E o U ■ u c £ .i S > 0 ; A Note on the Relationship Between Bank Holding Company Risk and Nonbank Activity Elijah Brewer III m O 73 > r — 73 73 < m 00 > Z T. o n x o > o 0 A Note on the Relationship Between Bank Holding Company Risk and Nonbank Activity ABSTRACT It has been argued that permitting banking organizations to expand into other lines of business will reduce their total risk through diversification. This note, using a stock market measure of risk, examines the proposition that diversification into nonbank activities decreases bank holding com pany (BHC) risk. In contrast to studies using accounting-based measures of risk, we find that expansion into nonbank activities during the 1979-1985 period substantially decreased BH C total risk. This suggests that limiting further expansion of nonbank activities of BHCs would reduce their ability to engage in risk reducing diversification. FRB CHICAGO Staff Memoranda 1 A Note on the Relationship Between Bank Holding Company Risk and Nonbank Activity Elijah Brewer III Given the recent financial difficulties experienced by many banking organ izations and the large lending commitments made by money center insti tutions to heavily indebted African and Latin American nations, there is widespread concern about the effects of bank holding company (BHC) ac tivity deregulation on bank riskiness. It is argued that further activity de regulation could compromise the safety and soundness of the banking system and extend the safety net designed for depository institutions to nondepository firms and commercial activities. Others have argued that bank subsidiaries are strengthened when BH C operates nonbanks profitably; yet, should those firms incur losses, bank subsidiaries are pro tected by the legal separateness of the B H C ’s corporate structure. In ad dition, going into nonbank activities diversifies the B H C ’s assets and provides an opportunity to reduce BH C risk sensitivity. Has diversification into nonbank activities by BHCs affected the safety of the banking system? In particular, have banking organizations diversified into nonbank activ ities in ways that increased or decreased their overall exposure to risk? The purpose of this note is to examine the proposition that diversification into nonbank activities decreases BH C risk. Previous studies, using accounting-based measures of risk, find no indi cation that increases in nonbank activities increase BH C exposure to risk and weak, though quantitatively small, indications that such expansions in nonbank activities decrease BH C risk. However, these studies employed inappropriate measures of risk. When we employ BH C risk measures de rived from stock price data, we find, among other things, that expansion into nonbank activities substantially decreases BH C risk. The new evidence summarized here suggests that proposals to limit further expansion of nonbank activities of BHCs would reduce their ability to engage in risk reducing diversification. I. B H C risk and nonbank activities Boyd and Graham (1986) and Wall (1987) attempt to assess the effect of nonbank activity on BH C risk. Unfortunately, both these studies rely pri marily on accounting data and not capital market data, the latter of which provides a theoretically more satisfying basis for analysis in that investors’ assessments are reflected in market risk measures. Boyd and Graham (1986) employ two measures of risk: (1) the standard deviation of the rate FRB CHICAGO Staff Memoranda ? o f return on assets and (2) the probability o f bankruptcy, i.e., the number of standard deviations below the mean that B H C profits would have to fall to make B H C book equity negative. This latter risk measure is similar to the one used by Wall (1987). The primary difference is that Wall’s risk measure is based on return on equity rather than return on assets. Empir ical analysis [e.g. Sinkey (1975)] has suggested that the standard deviation of earnings is not a good measure o f risk. Bowman (1979) has shown that there is no theoretical relationship between earnings variability and at least one market-based risk measure, the systematic risk o f the firm. The quality of accounting data for users attempting to measure risk is subject to ques tion. Ronen and Sorter (1972), for example, criticize accounting informa tion by suggesting that it does not explicitly deal with considerations of risk. Accounting data may supply assistance for risk assessment in an im plicit, rather than an explicit, manner. The work of Beaver, Kettler and Scholes (1970) suggests that this is in fact the case. They concluded that accounting-determined measures of risk were indeed impounded in market-based risk measures. Brewer and Lee (1986) find that there is a significant but imperfect correlation between accounting-based measures of equity risk and market-based measures of equity risk. In another study, Brewer and Lee (1988) find that an accounting-based measure o f interest rate risk exposure is significant in explaining the sensitivity of bank equity returns with respect to unanticipated interest rate movements. Modern finance theory then suggests that the riskiness o f BH Cs’ involve ment in nonbank activities can be measured by analyzing stock market re turns. B H C equity returns are sensitive to all the factors that affect the overall stock market as well as to factors specific to the banking industry. For example, banking organizations are sensitive to “earnings risk” through possible defaults on their loans and investments, changes in loan demand, and potential variability in growth and profitability of their non-portfolio operations. Banking organizations’ equity returns are also sensitive to movements in interest rates because they typically fail to match the interest sensitivity of their assets and their liabilities. As a result, movements in interest rates affect the market value of each side of the banking organization’s balance sheet and both its net worth and stock returns. Changes in nonbank activities can affect banking organizations’ stock re turns because these activities could make them more or less exposed to earning and interest rate risks. Using stock market data, we can test how B H C ’s involvement in nonbank activities is related to a market-based risk measure. Such cross-sectional tests will not let us draw strong inferences about the riskiness of specific nonbank activities because we do not know the type o f nonbank activity levels. However, the tests will at least let us determine how, in a particular period, levels of B H C risk and nonbank ac tivity have been related— positively, negatively, or not at all. FRB CHICAGO Staff Memoranda 3 II. Methodology and data To investigate the relationship between nonbank activity and BH C risk, we examine common stock daily returns of BHCs during the period 1979-1985. The methodology utilized to examine differences in risk is similar to that used by Aharony, Jones and Swary (1980) in their study of corporate failure and by Aharony and Swary (1981) in determining the risk-return effects of the Bank Holding Company Act of 1970. Specifically, we use the standard deviation of BH C equity returns as a measure of the risk borne by the BHC. The standard deviation of equity returns measures the risk of equity and does not entirely capture risk of BH C assets, as measured by the variability of the returns on BH C assets. We concentrate on the risk of equity because B H C equity can be viewed as an option on the assets of the BHC. Using option pricing theory, it can be shown that the variability of BH C equity returns is proportional to the variability of the returns on B H C ’s underlying assets. The proportionality factor measures the elasticity of the BH C stock price with respect to the underlying assets of the BHC. Therefore, the standard deviation of BH C equity returns must be assessed in order to properly measure risk of BH C assets. The data used in this note are for 40 bank holding companies whose stock was traded on the New York Stock Exchange, American Stock Exchange, or Over the Counter and which filed Reports of Condition and Reports of Income and Bank Holding Company Annual Report Financial Supple ments (F R Y-9). Balance sheet data are from the Board of Governors of the Federal Reserve System. Data for individual banks are grouped by holding company. Stock market data are from Interactive Data Services, Inc. To obtain our measure of risk, we use daily data to estimate for each month in the sample period the standard deviation of returns on a B H C ’s stock. These monthly estimates were then averaged together to generate annual estimates o f BH C stock price volatility for each year of the sample period. We chose to test the proposition that diversification into nonbank activity decreases BH C risk by identifying those factors which affect the standard deviation of BH C stock returns. Boyd and Graham relate BH C risk to the ratio of capital-to-asset (C A P IT A L ) and the log of total asset (TA), a measure of BH C size. A recent study by Brewer and Lee (1986) relates market-based measures of risk to BH C balance sheet data. They find that three key variables were consistently related to BH C risk sensitivity: C A P IT A L , loans-to-asset (LOANS), and purchased funds-to-asset FRB CHICAGO Staff Memoranda 4 (F U N D S ) . 1 We relate these ratios to estimates of the standard deviation of B H C stock returns. Besides C A P IT A L , F U N D S , L O A N S , and T A , we use an additional vari able to measure bank holding company involvement in nonbank activities. The measure of this variable is one minus the ratio o f estimated B H C ’s total bank assets to its total consolidated assets (N O N B A N K ). An estimate o f bank assets is obtained from Reports of Condition by summing deposits, federal funds purchased, other borrowings, and other liabilities. These dollar amounts were aggregated over all banks owned by each B H C to generate estimates of B H C ’s total bank assets. These estimates are then divided by BH C consolidated total assets to compute the proportion of B H C consolidated total assets atributable to bank activities. Our measure of BH C involvement in nonbank activities is one minus the proportion of BH C consolidated total assets attributable to bank activities. Another measure of nonbank activity is used to check the robustness of our results. The measure of this variable is one minus the ratio o f the B H C ’s total bank assets obtained from Reports of Condition to its total consolidated assets (N O N B A N K 1 ) .2 The above discussion suggests the following models: STD(R„) = Oo + ax C A P IT A L ,-, + a2 F U N D S ,-, + a3 LOANS,, + < T A ,-, + a5 NONBANK,-, + e,-, 24 STD(R„) = b0 + b x CAPITAL,-, + Z7 F U N D S > + 64 TA,-, + b5 NONBANK1,-, + vjt + b3 L O A N S , where STD(R,,) is the standard deviation of stock return on B H C j in period t; C A P I T A L ,- is the market value o f equity-to-total asset ratio o f B H C j in , period t; ejt and v are error terms; and the other variables are defined as jt before.3 The seven years of data beginning in 1979 and ending in 1985 are pooled, yielding 280 observations. Using this pooled data, the relationship between B H C nonbank activity and the standard deviation of BH C equity returns was estimated using both ordinary least squares (OLS) regression and the Fuller-Battese technique for estimating regression coefficients when dealing with cross-section time series data .4 Time dummy variables, Dum79-Dum84, are included in the equations estimated by O LS to control for the effects on the standard deviation of equity returns of changes in time-specific factors that are not captured by C A P IT A L , F U N D S , L O A N S , T A , and N O N B A N K (N O N B A N K 1 ) . 5 In using Fuller-Battese, rather than O LS with time dummies, the existence of other time as well as cross-sectional effects can be determined by the sample. FRB CHICAGO Staff Memoranda 5 III. Empirical Results The results of estimating different versions of equations (1) and (2) using both O LS and Fuller-Battese techniques are shown in Table l.6 The esti mated values o f the parameters represent their cross-sectional average val ues.7 Where an increase in a financial ratio would be expected to increase risk, that ratio should have a positive coefficient. The first set of equations using OLS, (la) and (2a), includes the market capital-to-asset ratio (C A P ITAL), total assets (TA), and a measure of nonbank activity (N O N B A N K or N O N B A N K 1). The coefficient on the capital ratio has a negative sign and is significantly different from zero at the 0.01 level in both equations. The coefficients of N O N B A N K and N O N B A N K 1 are also negative and significantly different from zero at a high confidence level.8 Three of the five time dummies are statistically significant. These results indicate that relative to 1986, BH C riskiness was higher, on average, in 1983 and lower in 1979 and 1981. Equations (lb) and (2b) present coefficient estimates of taking other possi ble factors into account. The coefficients of these additional variables did not prove to be significantly different from zero. The third set of O LS re sults, (lc) and (2c), excludes TA . These results were marginally better than those in equations (lb) and (2b). The coefficient on the purchased funds ratio has a positive sign and is significantly different from zero in both equations (lc) and (2c). In the regression equations based on the FullerBattese estimator, the loans-to-asset ratio is statistically significant.9 In creases in L O A N S tend to raise BH C risk. The sign o f the coefficient on nonbank activity indicates that increases in nonbank activity tend to lower BH C total risk. This result is partially corroborated by evidence presented in Boyd and Graham (1986) and Wall (1987). Using accounting-based measures of risk, Boyd and Graham find a negative but insignificant association between nonbank activity and BH C risk during the 1978-1983 period. Wall (1987) has also used accounting data to investigated the relationship between nonbank activity and BH C risk of failure for a sample of 267 BHCs during the 1976-1984 period. He finds insignificant evidence that nonbank activities reduce BH C risk. Our conclusions are more substantial.1 We find that BHCs with above-average 0 nonbank activities will have below-average risk. The next question is whether or not the implied differences in risk are large. One way this can be established is by looking at the impact of a one-standard-deviation change in nonbank activity on the standard deviation of BH C equity re turns. Table 2 shows how a one-standard-deviation change in both meas ures of nonbank activity translates into a change in the standard deviation of BH C equity returns. Using the results of Table L a one-standard- FRB CHICAGO Staff Memoranda 6 deviation increase in nonbank activity causes the standard deviation of BH C equity returns to fall 8 - 1 1 basis points, or about 5 - 7 percent. IV . Implications The results presented here have two important public policy implications. To begin with, they point out the risk-reducing benefits associated with nonbank activities. Since nonbank activities appear to make BH Cs less risky, then regulators might want to require BHCs with nonbank subsid iaries to hold lower levels of capital. However, it is not enough to show that nonbank activities make BHCs less risky, we also need to evaluate which types of nonbank activity reduce BH C riskiness. What does all this say about recent proposals to substantially expand the nonbank powers of BHCs into such areas as insurance underwriting, in vestment banking, and real estate? We see some evidence that aboveaverage nonbank activity has been associated with below-average risk. If results of these cross-sectional test indicate how future cross sections might look after major expansions of nonbank powers, there might be little reason for concern about increases in BH C risk. However, to the extent that the proposed activities are further removed from banking and much riskier than those permitted during our sample period, the results reported in this paper might provide little indication of the future relationship between nonbank activity and BH C risk. Nonetheless, there appears to be some potential for risk reduction via increases in the percentage of B H C assets devoted to nonbank activities. FRB CHICAGO Staff Memoranda 7 Footnotes 1 Purchased funds are defined as the sum of large time deposits of $100,000 or more, deposits in foreign offices, federal funds purchased and securities sold under agreements to repurchase, commercial paper, and other borrowings with an ori ginal maturity of one-year or less. 2 This variable was used by Boyd and Graham (1986). 3 The market value of equity was computed based on averages of outstanding common shares and prices during each year of the sample period. 4 See Drummond and Gallant (1983) for a discussion of cross-sectional time-series models. 5 For a discussion of the existence of “other effects” see Balestra and Nerlove (1966). 6 The average values as a percent of total assets of the financial variables used in Table 1 are: 1979 C A P IT A L FUNDS LOANS NONBANK NONBANK1 1980 1 98 1 1982 1983 1984 1985 0 .0 3 1 5 0 .4 1 8 6 0 .5 2 7 9 0 .1 1 4 0 0 .0 6 5 3 0 .0 2 9 3 0 .4 2 4 2 0 .5 2 6 4 0 .1 1 1 3 0 .0 6 2 8 0 .0 3 4 0 0 .4 4 3 1 0 .5 2 9 3 0 .1 0 5 0 0 .0 5 6 0 0 .0 3 0 7 0 .4 2 6 6 0 .5 3 4 7 0 .1 2 4 5 0 .0 7 5 0 0 .0 4 0 0 0 .3 9 4 5 0 .5 2 8 6 0 .1 2 0 8 0 .0 6 9 1 0 .0 3 8 7 0 .3 7 2 2 0 .5 6 8 7 0 .1 1 5 0 0 .0 6 2 9 0 .0 4 7 2 0 .3 6 3 7 0 .5 5 7 4 0 .1 4 9 5 0 .0 9 6 6 7 Specific tests were made to determine whether pooling across time was permis sible. The null hypothesis of homogeneity of slope coefficients across time cannot be rejected for both equations (1) and (2), F(30,238) equals 0.74 and 0.83, re spectively. 8 Similar results were •obtained when equations (la) and (2a), excluding time dummies, were estimated using the Fuller-Battese technique. 9 When equations (lb) and (2b), excluding time dummies, were estimated using the Fuller-Battese technique, CAPITAL, LOANS, and the nonbank activity measure (NONBANK or NONBANK1) were significantly different from zero. FRB CHICAGO Staff Memoranda 8 1 0 We did some tests using the standard deviation of returns on assets over the 1979-1985 period as the dependent variable. The equations below are represen tative of these tests:" (1) STD(7?04) = 0.6335 - 2.7462 C A P IT A L j - 0.1477 F U N D S j (1.342) (0.747) (0.457) + 0.0209 LO AN Sj - 2.2269 N O N B A N K j _ (0.034) (1.942)** R2 - 0.0818 F-Statistic = 1.869 (2) STD(/?CM7) - 0.4909 - 3.2056 C A P IT A L j - 0.0910 FU ND Sj (1.105) (0.894) _____ (0.286)________ + 0.0608 LOAN Sj - 2.1908 N O N B A N K lj _ (0.101) (1.985)** R2 = 0.0858 F-Statistic = 1.915 **Signigicantly different from zero at the 5% level. "Numbers in parentheses beneath the regression coefficients are the correspond ing t-statistics. Where a bar (-) over a variable denotes an average value over the 1979-1985 pe riod. The number of observation, in-each equation is 40. These results are much less clear-cut than the ones presented in Table 1. While the standard deviation of returns on assets exhibits a negative relationship with nonbank activity, the significance levels are relatively lower than those reported in Table 1. In addition, we Find no significant relationship between the standard deviation of returns on assets and the other independent variables. As a result, we do not have much confidence in these findings. We performed additional tests using accounting or market data covering 1979 through 1983, a sample period not too different from one of Boyd and Graham’s (1986) subperiods, and obtained results somewhat weaker than those for the full sample period, 1979 through 1985. IRB CHICAGO Staff Memoranda 9 References Aharony, Joseph, Charles P. Jones, and Itzhak Swary, “An Analysis of Risk and Return Characteristics of Corporate Bankruptcy Using Capital Market Data,” Journal o f Finance, 35 (September 1980), 1001-1016. Aharony, Joseph and Itzhak Swary, “Effects of the 1970 Bank Holding Company Act: Evidence from Capital Markets,” Journal o f Finance, 36 (September 1981), 841-854. Balestra, Pietro and Marc Nerlove, “Pooling Cross-Section and Time-Series Data in the Estimation of a Dynamic Model: The Demand for Natural Gas,” Econometrica, 34 (July 1966), 585-612. Beaver, William H., Paul Kettler, and Myron Scholes, “The Association Between Market Determined and Accounting Determined Risk Measures,” The Accounting Review, 45 (October 1970), 654-682. Bowman, Robert G., “The Theoretical Relationship Between Systematic Risk and Financial (Accounting) Variables,” Journal o f Finance, 34 (June 1979), 617-644. Boyd, John H. and Stanley L. Graham, “ Risk, Regulation, and Bank Holding Company Expansion into Nonbanking,” Quarterly Review, Federal Reserve Bank of Minneapolis, 10 (Spring 1986), 2-17. Brewer III, Elijah and Cheng Few Lee, “How the Market Judges Bank Risk,” Perspectives, Federal Reserve Bank of Chicago, 10 (November/December 1986), 25-31. Economic Brewer III, Elijah and Cheng Few Lee, “The Sensitivity of Bank Stock Returns to Interest Rate Risk Using Schedule J Data,” Unpublished Paper, (January 1988). Ronen, Joshua and George H. Sorter, “ Relevant Accounting” Journal o f Business, 45 (April 1972), 258-282. Sinkey, Joseph F., “A Multivariate Analysis of the Characteristics of Problem Banks,” Journal o f Finance, 30 (March 1975), 21-36 Wall, Larry D. “Has Bank Holding Companies’ Diversification Affected Their Risk of Failure?,” Journal o f Economics and Business, 39 (November 1987), 313-326 FRB CHICAGO Staff Memoranda 10 T b e1 al The R l t o s i Between t e Standard D v a i n o BHC Stock Return eainhp h eito f and Nonbank A t v t ciiy ( 9 9 -1 8 ) 17 95 EQUATION INTERCEPT CAPITAL (1a) 1.5389 ( 3 .5 0 8 ) * " -7 .6 9 6 0 ( 3 .8 3 7 ) " ' -- (1b) 1.7483 ( 3 .0 2 4 ) " * -7.6 355 ( 3 .7 9 8 ) " * (1c) 1.6305 (5 .5 4 0 )*** (2a) FUNDS LOANS TA NONBANK NONBANK1 D U M 79 D U M 80 DU M 81 D U M 82 DU M 83 D U M 84 /?2 F-Statistic N A. O rd in a ry Least S q u ares -- 0.0331 (1.3 85) -2.0 406 ( 4 .1 9 8 ) " * -- -0.4427 ( 3 .7 2 7 ) '" -0 .1 3 4 4 (1 .1 2 4 ) -0 .2 1 8 9 (1 .8 6 7 )* 0.0960 (0.8 14) 0.1858 (1 .6 2 7 )* -0 .0 5 7 0 (0 .4 9 7 ) 0.2323 1 0 .3 8 2 " * 280 0.4007 (1.1 76) 0 5098 (1.352) -0.0 1 0 0 (0.2 37) -1 .6 3 1 0 ( 2 .9 3 1 ) " * -- -0.4587 (3 .7 2 7 )*** -0 .1 4 6 6 (1 .1 8 5 ) -0 .2 3 4 6 ( 1 .9 2 3 ) " 0.0805 (0.6 64) 0.1906 (1 .6 5 0 )* -0 .0 5 7 3 (0.5 0 0 ) 0.2331 8 .7 1 0 " * 280 -7.5 3 1 3 (3 .8 4 6 )*** 0.3342 (1 .7 3 0 )* 0.4746 (1.372) -- -1.6751 ( 3 .1 9 9 ) " * -- -0.4503 ( 3 .8 2 6 ) * " -0 .1 3 8 7 (1 .1 6 6 ) -0 .2 2 6 9 (1 .9 3 3 )** 0.0875 (0.7 47) 0.1933 (1 .6 8 5 )' -0 .0 5 5 6 (0 .4 8 7 ) 0.2358 9 .6 0 9 '* ' 280 1.3302 ( 3 .0 1 7 ) " * -8.2 306 ( 4 .1 1 9 ) " * -- -- 0.0395 (1 .6 3 5 )' ~ -1.8 179 ( 3 .7 2 8 ) " ' -0.4 318 ( 3 .6 1 5 ) * " -0 .1 2 4 5 (1 .0 3 5 ) -0 .2 0 6 4 ( 1 .7 5 2 )' 0.1009 (0.8 50) 0.1919 (1 .6 7 0 )* -0.0 5 1 7 (0 .4 4 8 ) 0.2223 9 .8 5 9 * " 280 (2b) 1.6289 ( 2 .7 7 5 ) " * -8.0 8 8 6 ( 4 .0 4 2 ) " ' 0.4708 (1.379) 0.5962 (1.576) -0.0 129 C .299) -- -1.3 449 ( 2 .4 2 0 ) " -0.4 529 ( 3 .6 6 1 ) '" -0 .1 4 0 9 (1 .1 3 4 ) -0 .2 2 6 9 (1 .8 5 2 )* 0.0814 (0.6 68) 0.1968 (1 .6 9 6 )* -0 .0 5 2 2 (0 .4 5 4 ) 0.2255 8 .3 8 4 " * 280 (2c) 1.4750 ( 5 .2 4 3 ) " -7.9 6 0 5 ( 4 .0 8 0 ) " 0.3866 (2 .0 1 0 )** 0.5503 (1.594) -- -1.4 063 ( 2 .7 2 9 ) " -0.4 422 (3 .7 4 1 )'* -0 .1 3 0 9 (1.0 9 5 ) -0 .2 1 7 2 (1 .8 4 2 )* 0.0903 (0.7 67) 0.2002 (1 .7 3 7 )* -0 .0 5 0 3 (0 .4 3 8 ) 0.2281 9 .2 4 4 * " 280 (1c) 1.2490 ( 3 .6 0 6 ) '" -7 .9 1 9 4 ( 3 .3 8 0 ) " * 0.1304 (0.4 28) 1.1630 (2 .6 9 5 ) * " -- -1.4 8 6 6 ( 2 .8 9 8 ) " * -- -- 280 (2c) 1.1046 (3 .6 0 6 )* '* -8.0 248 ( 3 .4 1 3 ) " * 0.1680 (0.551) 1.2430 ( 2 .8 5 4 ) * " -- -- -1.2 696 ( 2 .4 7 1 ) " * -- 280 B. F u lle r-B a tte s e Note: The numbers in parentheses below the regression coefficients are the absolute values of the corresponding t-ratios. 'Significant at the 10 percent level. "S ig n ifica n t at the 5 percent level. '"S ig n ific a n t at the 1 percent level. Table 2 The Impact of Nonbank Activity on the Standard Deviation of Common Stock Returns Sample Average NONBANKt N0NBANK1 ™ STD(fy) Sample Sta d r nad Dvain eito 0.1200 0.0697 0.0147 007 .61 006 .61 0.0057 tBased on E u t o ( c q a i n 1) nBased on E u t o ( c q a i n 2) Change i s a d r d v a i n o Common n t n a d eito f Stock R t r s due t a one s a d r eun o tnad d v a i n i c e s i nonbank at v t eito nrae n ci i y Fle-ats ulrBtee OLS 001 .01 0.0009 0.0010 0.0008
@FedFRASER