The Role of Firewalls in Universal Banks: Evidence from Commercial Bank Securities Activities before the Glass-Steagall Act

Randall S. Kroszner and Raghuram G. Rajan, University of Chicago


This summary excludes all footnotes, most tables, and much text of the complete paper, which is available on request from the authors. Please do not cite or quote this summary.

I. Introduction

Before 1933 commercial banks, either through their securities departments or through separately incorporated securities affiliates, competed directly with investment banks in originating and underwriting securities issues. The Glass-Steagall Act of 1933 ended this competition by legislating the separation of commercial and investment banking in the U.S. The commercial banks were accused of succumbing to the conflicts of interest inherent in their dual capacity as underwriter and lender, and hence, of systematically misrepresenting the quality of securities to a gullible public. Recent studies, however, have found no evidence that such systematic fooling did indeed take place (e.g., Kroszner and Rajan 1994). A number of economists and policy makers have argued that since the rationales for Glass-Steagall Act do not appear to be justified, it should be repealed. Support for repeal is growing in the Federal Reserve Board and the Office of the Comptroller of the Currency (see Greenspan 1988 and Benston 1990).

The policy focus has begun to shift away from the merits of a blanket prohibition on the underwriting of corporate securities by commercial banks to debating how to structure underwriting activities by commercial banks such that the potential for conflicts of interest and misuse of public trust is minimized. Concerns about "conflicts of interest and loss of public confidence" have led the regulators to require extensive "firewall" separations to provide "an insulating framework" between a commercial bank and its securities activities (Greenspan 1988).

Commercial banks in the United States, before the Glass Steagall Act, underwrote both through separately capitalized affiliates and through in-house departments. We analyze the consequences of the organizational structure chosen by the commercial banks for their securities businesses. More specifically, we examine how the relative independence of the securities and commercial lending businesses of different banks affects the types of securities and companies they underwrite, the way the market prices the securities, and the subsequent performance of the issues.

We find that firewalls appear to have been valuable in enhancing an underwriter's credibility in the market. Our results suggest that the cost of underwriting securities in-house rather than through an arm's length affiliate is that the issue price is discounted by a market concerned about potential conflicts of interest. This effect seems to outweigh the potential for more informed certification whereby in-house securities departments use their easier access to information about a firm's prospects to identify and certify "jewels in the rough." Even though we do find the organizational segregation of underwriting and lending activities increased the ability of underwriters to certify firms to the market, we find that market pressures, and not regulation, led to the adoption of some form of firewalls. During our sample period, the organization of securities activities in a separate affiliate rapidly became the dominant form for commercial banks and trusts that were lead underwriters and syndicate managers.

II. Organizational Structures of Commercial Bank Involvement in the Securities Business before the Glass-Steagall Act

To investigate the consequences of the different organizational forms, that is, the extent of the firewall separations, we will contrast the types of underwriting activities and the performance of the underwritten securities across the two main organizational structures for the commercial banks' securities businesses: "captive" internal securities departments and separately capitalized and separately incorporated affiliates. The affiliates were chartered under state laws as regular corporations, free of the regulations associated with banking. There were no minimum capital regulations, and some affiliates were incorporated with small amounts of capital (see, e.g., Peach 1941, p.81). Since the affiliates typically shared the name of their parent, affiliates enjoyed the "full benefit of the goodwill of their parent banks" (Peach 1941, p.52).

While many bank and trust companies entered the securities business during the 1920s, the movement was not universal. Some banks and trusts argued that having an internal department which underwrote and distributed securities could compromise the "soundness, integrity, and conservatism" of their investment advice, and such institutions proudly advertised that they did not have such a department (Peach 1941, p. 72). In 1925, for example, the Farmers' Loan and Trust Company of New York announced in the Commercial and Financial Chronicle (May 2, 1925, p. 2228):

Due to our policy and firm conviction that, as a trustee, we should never place ourselves in the position of a buyer and seller of securities at the same time, we have never had a bond department. Our whole security department is organized for the impartial study of securities for the benefit of our customers and not for the sale of bonds to the public.

III. Theoretical Debates

Political debate on universal banking has focused, since the Pecora Committee hearings in 1933 (see U.S. Senate, 1933-34), on two main issues. First, legislators have been concerned about the adverse effects of extending bank powers on bank risk taking. It has been argued that the wave of bank failures in the U.S. in the early 1930s may have resulted from bank involvement in the excessively risky underwriting business (see the discussion in White, 1986). Second, conflicts of interest may arise when a bank combines lending and deposit taking with underwriting. If a firm suffers an adverse shock without the public realizing it, for example, a commercial bank may have an incentive to underwrite public issues on behalf of the firm and use the proceeds to repay earlier bank loans made to the firm.

Theoretically, however, it is not clear that before the Glass-Steagall Act bank incentives were distorted in such a way that banks wanted to increase the riskiness of their activities. Furthermore, there is no justification for assuming that underwriting activities were riskier than lending. Consistent with the lack of theoretical justification for the argument that bank involvement in underwriting lead to excessive risk taking, White (1986) finds that securities operations of commercial banks did not impair their stability prior to Glass-Steagall. Banks engaged in the securities business had no higher earning variance or lower capital ratios than banks without such operations. In addition, those banks with securities operations were less likely to fail. Although 5000 banks failed during the 1920s, virtually none were the city banks which were the most likely to have securities affiliates (Carosso 1970, p. 242; see also White 1983). In the bank crises between 1930 and 1933, more than a quarter of all national banks failed but less than 10% of those with large securities operations closed (White 1986, p. 40).

Concerning the second issue, Saunders (1985) and Benston (1990) argue that even if underwriters harbor conflicts of interest, rational investors will not be systematically fooled if they are aware of these conflicts. Kroszner and Rajan (1994) compare the performance of bonds underwritten by the affiliates of commercial banks with those of independent investment banks. The bonds underwritten by commercial banks outperformed similarly rated bonds underwritten by investment banks. Furthermore, Kroszner and Rajan find that the difference in performance was especially pronounced amongst lower quality bonds where the incentives and the potential for commercial bank affiliates to dupe the public investor would have been the greatest.

While the theory and evidence imply that the focus of policy on safeguarding the interests of the public investor by prohibiting banks from underwriting is misplaced, they do not imply that conflicts of interest are costless. Public investors rationally will require a higher promised yield from bank underwritten than for similar investment bank underwritten issues. The effect should be most pronounced for small, low quality junior issues underwritten by small, relatively less well known affiliates. We term this the `rational discounting' hypothesis. There is, however, an alternative hypothesis which has contrary implications. Banks may have better access to information through the lending process than do independent investment banks. Consequently, they may be better able to certify firms to the market. This would imply that, compared to issues underwritten by affiliates, investors would discount issues underwritten by independent investment banks because these investment banks are less informed. We term this the `informed certification' hypothesis. Which effect dominates - the investors' suspicion that banks harbor conflicts of interest or the better certification because banks have better information and/or enjoy scope economies in gathering it - is an empirical question.

IV. Data Collection and Sources

We use a variety of contemporary and more recent sources to identify commercial banks engaged in investment banking prior to the Glass-Steagall Act: Carosso (1970), Peach (1941), Preston and Findlay (1930a and 1930b), Moore (1934), White (1986), the Commercial and Financial Chronicle (CFC) and the National Securities Dealers of North America (1929). To be included in our sample, the bank or trust must be listed in the Moody's Banking Manual. Moody's provides information on each bank's balance sheet and on the organization of the securities operations. Some, but not all, affiliates report a separate balance sheet.

In order to determine which of these banks were actively engaged in securities underwriting, as opposed to simply acting as brokers, we examine the two volume American Underwriting Houses and Their Issues (1928 and 1930). This source groups all new public securities issues between January 1, 1925 and December 31, 1929 by underwriter. The listings include common and preferred stock and short and long bonds of private corporations and governments. The entry for each security lists the month of issue, issue size, the coupon, the price, and the underwriter(s). The monthly new capital flotations section of the CFC reports the implied yield to maturity for bonds. The lead underwriter or syndicate manager is listed first, if more than one house is involved. As in our earlier work (Kroszner and Rajan 1994), we include in our sample only those securities in which the commercial bank, trust, or its securities affiliate is the lead underwriter or syndicate manager. This process yields a total of 893 securities underwritten by 24 commercial banks and trusts in-house and 33 securities affiliates.

V. Results
1. Underwriting activities of departments and affiliates.

The commercial banks were eroding the market share of independent investment banks in the bond underwriting business (US Senate 1931, p.299). Commercial banks underwrote roughly 22% of this market in 1927 and 45% by 1929. Separate affiliates were becoming the dominant organizational form for commercial bank underwriting. In particular, the share of bond issuance underwritten by separate affiliates of commercial banks tripled between 1927 and 1929, while the share underwritten through the securities departments of the banks fell by half. For these three years, separate affiliates underwrote three-quarters of the total volume of issues underwritten by commercial banks' securities departments and affiliates. In the sample we have collected from 1925 to 1929, the pattern is the same. The evolution away from the department structure and widespread adoption of the affiliate structure is consistent with the `rational discounting' hypothesis according to which the greater independence and affiliate credibility improves its competitive position in the underwriting market.

2. The relative pricing of bonds.

Both the `rational discounting' hypothesis and the `informed certification' hypothesis make predictions about the difference in promised yields between otherwise similar department and affiliate underwritten bonds. The possibility of conflicts of interest make it difficult for departments to convey the quality of bonds accurately to investors. According to the `rational discounting' hypothesis, the added uncertainty about the quality of bonds underwritten by departments would lead investors to demand a higher yield. By contrast, according to the `informed certification' hypothesis, departments are better informed about the quality of bonds they underwrite. If conflicts of interest do not impede the credible communication of this information to the public, ceteris paribus, the informed certification should lead to lower yields of department underwritten bonds.

We first compare unconditional mean and median yields promised on bonds underwritten by the two organizational structures. We define the initial net yield as implied yield to maturity at the offering date minus the long-term government bond yield in the month of issue. The mean and median initial net yields for the affiliates are 2.02% and 1.99%, which are roughly 35 and 50 basis points lower than for the departments, and the differences are statistically significant. Although this difference in the unconditional means is consistent with the rational discounting hypothesis, we must correct for differences in the quality of issues underwritten by the two organizational structures before drawing conclusions.

Table 5 attempts to make such adjustments by regressing initial net yield on a variety of ex ante observable proxies for creditworthiness and an indicator variable for the organizational structure of the underwriter. Larger and more established firms on average tend to be of better credit quality than smaller and younger firms. As proxies for this, we include the log of the asset size (in thousands) of the firm being underwritten and the log of one plus the firm age in years. We also include an indicator which is one if the firm's securities are listed on an exchange as a proxy for other aspects of quality associated with such a listing, e.g., minimum disclosure requirements and, perhaps, reputation. All of the regressions include year of issue indicators to capture any changes in overall quality or market pricing over time and a constant, but for clarity, we do not report these coefficient estimates.

Column (i) of Table 5 includes the three quality controls and a department indicator which is one if the underwriter is an internal securities department. Unsurprisingly, the firm size and firm age proxies are strongly negatively correlated with initial net yield. The exchange listing indicator coefficient is positive but not statistically significant. The department indicator is positive and economically and statistically significant. This result suggests that the public discounted the prices of department underwritten securities, instead of imputing higher value to them as the `informed certification' theory would suggest.

In column (ii) of Table 5, we add the debt to total assets ratio at the time of issue as another ex ante observable control for credit quality. The department indicator remains positive and statistically significant. Curiously, higher leverage is (marginally) statistically significantly associated with lower initial net yields. Column (iii) helps to clarify this puzzle. We include indicators for issuer type, since certain industries tend to have both high leverage but also high ex ante creditworthiness. When indicators for railroads and public utilities are included, the debt to assets ratio coefficient changes sign to become positive but not statistically significant. While diminished somewhat in magnitude, the department indicator continues to be positive and highly statistically significant. After these quality adjustments, the initial net yield on a department underwritten bond is roughly 19 basis points higher than one underwritten by an affiliate. As a comparison, the difference in net yields between a bond rated Aa and a bond rated A in our sample is about 12 basis points. The effect of underwriting through a department thus is economically sizeable.

Column (iv) of Table 5 checks the robustness of our earlier results by adding three proxies which might affect the commercial bank's reputation and credibility: age, capitalization, and regulation. The first is log of one plus the age in years since incorporation of the parent commercial bank. The second is log of parent bank's capital, defined as the book value of equity plus surplus plus undivided profits. The third is an indicator variable that is one if the parent institution has a national charter and zero if it has a state charter. This variable may proxy for both differences in regulatory treatment and differences in the public's perception of the quality of monitoring by state versus national authorities. We find that older and larger banks, that is, as banks with greater non-monetary (i.e. reputational) and monetary capital, tend to have lower initial net yields on the securities they underwrite, although only the age variable is statistically significant. Holding bank age and bank capitalization constant, the securities underwritten by national banks have higher initial net yields. Adding the three variables increases both the size and the precision of the estimated coefficient for the securities department indicator, confirming our previous results.

On average, banks with securities affiliates were larger than those that were underwriting through departments. Larger banks are associated with underwriting larger and older clients firms (e.g., the correlation between the amount of capital of the underwriting bank and the asset size of the issuing firm is a statistically significant 0.56, and the correlation between the capital of the underwriting bank and the age of the issuing firm is 0.13). The securities department indicator thus could be proxying for an omitted quality variable. We perform a further robustness test to adjust for the size, and perhaps quality, difference across the organizational structures. In our sample, the bank with the largest capital which underwrites through a bond department has total capital of $69 million. We then exclude all issues underwritten by affiliates with greater total capital (the sum of bank capital and, if available, affiliate capital) than $69 million. We are left with 16 affiliates. The banks in this sub-sample underwrote a total of 183 bond issues. Of these 70 are underwritten by affiliates and 113 by departments belonging to organizations of comparable size.

Comparing the activity choices in this sub-sample, affiliates have a larger percentage of their underwriting activities in junior securities: more equity (6% of issues versus 2%) as well as more preferred (14% versus 10%). Not only did the departments focus on more senior debt securities, but an analysis of the characteristics of firms issuing long-term bonds suggests that they underwrote safer firms. The mean assets of firms underwritten by the departments is $7.5 million which is statistically higher than the mean of $6.2 million for the affiliates. The mean debt to assets ratio before the issue is 0.15 for departments while it is 0.17 for affiliates, and 53% of department underwritten firms are listed while 46% of affiliate underwritten firms are. The mean age since founding of the company is 17 for both. These data suggest that, except along the dimension of age, departments underwrite marginally safer, higher quality firms than comparably sized affiliates. If the department indicator is a proxy for omitted quality variables in the regressions reported in columns (i) to (iii) of Table 5, the effect should weaken or reverse in this sub-sample.

Column (v) in Table 5 reports the results of running the regression in column (iii) using the sub-sample of 183 bonds from the comparably sized underwriters. Contrary to what would be expected if the bond department indicator was a proxy for omitted quality variables, the coefficient on bond department is roughly the same magnitude (0.210 versus 0.192 for the full sample) and remains statistically significant at the 1% level. The stability of most of the coefficient estimates also gives us some confidence in the robustness of the specification.

A final possibility is that a only a small subset of departments which underwrite many bonds could be responsible for the premium we estimate, rather than it being a systematic effect. Since the regressions in Table 5 in a sense weight the yield premium or price discount associated with a house by the number of bonds underwritten by that house, these specifications could overstate the effect. To determine whether this is driving our results, first we calculate the average yield premium (or price discount) on issues underwritten by each house as follows. We estimate the model in column (iii) of Table 5, except that we drop the department indicator as an explanatory variable. For each department and affiliate, we average the residual from that regression across all securities underwritten by that department or affiliate. This number is the average yield premium for the house. We then run a cross-sectional regression of the average yield premium for each house on the department indicator, the log of the bank's capital, and the log of one plus the age of the bank. If the organizational structure is important, the department indicator should be positively correlated with the average yield premium, even after correcting for the age and capital of the bank. The coefficient estimate for the department indicator is positive - a department obtains a yield for an issue which is 17 basis points higher than an affiliate - and statistically significant at the 10% level. Neither the age nor capital of the bank are statistically significant.

3. Pricing by sophisticated analysts: initial ratings.

Investors appear to have discounted the price of bonds underwritten by the departments relative to those underwritten by the affiliates. A possible concern is that the price may have been determined by relatively unsophisticated investors like depositors. One way to check the validity of our prior findings is to examine the initial ratings that a rating agency such as Moody's ascribes to the bonds. Presumably, Moody's analysts are more sophisticated than the average individual investor. If investors base their pricing on ratings or use the same information that the rating agencies do, we will find results similar to those reported in the previous sub-section. The results of our ordered logit regressions, with the rating categories as the dependent variable, confirm our findings on the initial net yields.

VI. Conclusions

Our results on the ex ante measures of the choice of underwriting activities, the initial market pricing of bonds, and the initial rating of bonds consistently support a more important role for rational discounting than informed certification in explaining the nature of commercial bank underwriting before the Glass-Steagall Act. Adjusting for ex ante observable characteristics of issuer quality, we find that yields are higher and ratings are lower for bonds underwritten by internal securities departments of banks and trusts relative to bonds underwritten by their separately incorporated and capitalized affiliates.

The concern about the potential for conflicts of interest when commercial lending and underwriting are combined within a bank appear to outweigh the certification benefits that might arise from improved information flows and economies of scope when the two functions are combined within an institution. While the bank may still retain the ability to refer business cultivated by its lending activities to its securities affiliate, the ability of the lending department to control the activities of the affiliate may be weaker. This then enhances affiliate credibility and improves its ability to certify firms.

The evolution toward separate affiliates also may shed some light on the policy issues of whether firewalls should be required if Glass-Steagall were to be repealed. Our evidence indicates that firewalls can be effective at improving underwriter credibility. It also appears that such value was internalized by the banks, since they voluntarily chose the arm's length structure. At least as far as conflict of interests rationales are concerned, it does not appear necessary to mandate firewalls to mitigate potential problems. Certainly, further detailed work on the organizational structure of banks in the 1920s and today is needed before strong conclusions for current policy can be drawn.

REFERENCES

Ang, James S. and Richardson, Terry, "The Underwriting Experience of Commercial Bank Affiliates Prior to the Glass Steagall Act: A Re-examination of Evidence for Passage of the Act," mimeo, University of Florida, 1993.

Benston, George J., The Separation of Commercial and Investment Banking, Oxford: Oxford University Press, 1990.

Carosso, Vincent P., Investment Banking in America: A History, Cambridge: Harvard University Press, 1970.

Dewing Arthur S., The Financial Policy of Corporations, 5th ed., 2 vols, New York: Ronald Press Company, 1953.

Diamond, Douglas, " Financial Intermediation and Delegated Monitoring", Review of Economic Studies, 393-414, 1984.

Edwards, George W., "The Myth of the Security Affiliate," Journal of the American Statistical Association, June 1942, 37, 225-32.

Fama, Eugene, "What's different about banks", Journal of Monetary Economics, 15, 29-39,1985.

Greene, William. Econometric Analysis, second edition. New York: Macmillan, 1993.

Greenspan, Alan. "Statement by Alan Greenspan, Chairman, Board of Governors of the Federal Reserve System, Before the Senate Committee on Banking, Housing, and Urban Affairs, December 1, 1987," Federal Reserve Bulletin, 1988.

James, Christopher, "Some evidence on the uniqueness of bank loans", Journal of Financial Economics, 19, 217-235, 1987.

James, Christopher and Wier, Peggy," Borrowing relationships, intermediation, and the cost of issuing public securities", Journal of Financial Economics, 28, 149-171, 1990.

Kanatas and Qi, "Underwriting by commercial banks: Conflicts of Interest vs. Scope Economies", Mimeo, University of South Florida, 1994.

Kaufman, George and Mote, Larry. "Glass-Steagall: Repeal by Regulatory and Judicial Reinterpretation," Banking Law Journal, September-October 1990, 388-421.

Kroszner, Randall and Rajan, Raghuram. "Is the Glass-Steagall Act Justified? A Study of the U.S. Experience with Universal Banking before 1933." American Economic Review, September 1994.

Leland, Hayne and Pyle, David, "Informational Asymmetries, Financial Structure, and Financial Intermediation," Journal of Finance, June 1977, 32, 371-87.

Lummer, Scott and John McConnell, 1989, Further Evidence on the Bank Lending Process and the Capital Market Response to Bank Loan Agreements, Journal of Financial Economics, Vol 25, 99-122.

Macey, Jonathan and Miller, Geoffrey, Banking Law and Regulation, Boston: Little, Brown and Company, 1992.

Moore, Terris, "Security Affiliate versus Private Investment Banker - A Study in Security Organizations," Harvard Business Review, July 1934, 12, 478-84.

Peach, W. Nelson, The Security Affiliates of National Banks, Baltimore: Johns Hopkins Press, 1941.

Preston, H. H. and Findlay, A. R., (1930a) "Investment Affiliates Thrive," American Bankers Association Journal, May 1930, 22, 1027-28, 1075.

_______and _______, (1930b) "Era Favors Investment Affiliates," American Bankers Association Journal, June 1930, 22, 1153-54, 1191-9.

Puri, Manju, "The long term default performance of bank underwritten securities issues" Journal of Banking and Finance, forthcoming, 1993 (a).

Puri, Manju, "Commercial Banks in Investment Banking: Conflict of Interest of Certification Role?" Unpublished working paper, New York University, 1993 (b).

Rajan, Raghuram G., "A theory of the costs and benefits of universal banking", C.R.S.P. Working Paper No. 346, University of Chicago, 1992.

Rajan, Raghuram G., "An enquiry into the economics of extending bank powers", Mimeo, University of Chicago, 1994.

Saunders, Anthony, "Conflicts of Interest: An Economic View", in I. Walter ed. Deregulating Wall Street: Commercial Bank Penetration of the Corporate Securities Market (John Wiley and Sons, New York), 1985.

Simon, Carol, "The Effect of the 1933 Securities Act on Investor Information and the Performance of New Issues," American Economic Review, June 1989, 79, 295-318.

White, Eugene, The Regulation and Reform of the American Banking System, Princeton: Princeton University Press, 1983.

_______. "Before the Glass-Steagall Act: An Analysis of the Investment Banking Activities of National Banks," Explorations in Economic History, 1986, 23, 33-55.

American Underwriting Houses and Their Issues, Vols I and II. New York: National Statistical Service, 1928 and 1930.

Banking and Monetary Statistics, Washington, DC: Board of Governors of the Federal Reserve System, 1976 [1943].

Bond Quotation Record, New York, various issues, 1921-1930.

Commercial and Financial Chronicle, New York, various issues, 1921-1940.

Robert D. Fisher Manual of Valuable and Worthless Securities, Volumes I-X, New York, 1927-1945.

Fitch Bond Book, New York: Fitch Publishing Company, 1921-1929.

Moody's Manuals: Industrial, Public Utility, Government and Municipalities, and Banking and Financial, New York, various volumes, 1921-1941.

National Securities Dealers of North America, New York: Herbert D. Siebert Company, February 1929.

National Monthly Corporation Bond Summary, New York: National Quotation Bureau, various volumes, 1921-1941.

National Monthly Stock Summary, New York: National Quotation Bureau, various volumes, 1921-1941.

Poor's Industrial Section and Ratings Manuals, New York, 1921-1941.

U.S. Senate, Committee on Banking and Currency, 71st Congress, 3rd Session, Operation of the National and Federal Reserve Systems: Hearings on S.R. 71, Washington, DC: Government Printing Office, 1931.

U.S. Senate, Committee on Banking and Currency, 72nd Congress, 1st Session, Operation of the National and Federal Reserve Systems: Hearings on S.R. 4115, Washington, DC: Government Printing Office, 1932.

U.S. Senate, Committee on Banking and Currency, 72nd Congress, 2nd Session and 73rd Congress, 2nd Session, Stock Exchange Practices: Hearings on S.R. 84 and S.R.56 and S.R. 97, Washington, DC: Government Printing Office, 1933-1934.