BEFORE THE GLASS-STEAGALL ACT of 1933, close ties between commercial and investment banks in U.S. financial markets were common. According to Carosso (1970), during the 1880s it was typical for investment bankers to form syndicates with powerful and resource-rich commercial banks, which could ensure the flotation of large issues of securities in the still undeveloped financial markets. During the 1920s these ties developed more distinctively as investment banking by commercial banks and became a well-defined service provided either through a bond or investment department (formed as an internal division in the bank) or through a security affiliate (formed as a separate subsidiary of the bank).
Recent research has established, both at the theoretical and empirical level, that commercial bank involvement in the securities business fulfilled an important and beneficial role in financial markets. Theoretical research shows that it is easier for commercial banks than for arm's-length lenders to overcome asymmetric information problems because banks have an advantage in the collection and gathering of information about their client firms. (1) Empirical research also shows that this advantage, developed through long-term relationships and close monitoring, translated to fewer financing problems and higher market value for client firms. (2) Because of the synergies that the acquisition of information brings, these arguments suggest that there may have been economies of scale and scope in the commercial and investment banking industry, which may have justified the close association between these activities.
While theory suggests that there were scale and scope economies, few researchers have actually investigated them empirically in detail. There are at least two reasons for this. First, modern attempts to model or estimate economies of scale and scope in the banking industry have yielded largely ambiguous results in terms of the parameters in question. Problems ranging from the choice of a functional form of the production technology to the choice of the framework and estimation techniques have precluded researchers from reaching a consensus on this topic. (3) Second, although there have been recent improvements in estimation techniques and in the framework used (for example, Berger and Mester 1997), calculating parameters for economies of scale and scope is an extremely data-intensive exercise because it involves the estimation of many parameters (not just scale and scope elasticities, but many cross-elasticities as well). Because detailed bank financial data for this period are spotty at best, it is almost impossible to conduct a detailed and reliable study of economies of scale and scope, even if the model were to be estimated under the most restrictive functional forms. In light of these difficulties, it is unlikely that studies in economies of scale and scope can reveal the true nature of the cost structure of banks that were involved in the securities business during this period.
While estimating economies of scale and scope parameters may not be possible in a practical sense, measuring the impact of economies of scale and scope on banks' charter values is much more feasible. Clearly, the existence of economies of scale and scope implies that large banks that also offered investment banking services should have enjoyed a cost advantage over small, unit banks. All else being constant, this cost edge should consequently translate into higher market value for the bank.
In this paper I investigate whether this assertion can be empirically verified. I do this by examining whether financial markets attached a premium to the market value of banks with security affiliates or bond departments. To accomplish this I estimate Tobin's q for a random sample of banks operating in 1926-28 and test whether having an affiliate increased the value for the bank, after controlling for a large array of financial variables. I find that a bank's involvement in the securities business was rewarded in financial markets with a premium of about 4 to 7 percent on the market value of the bank. This translates to about $8 million of 1927-28 dollars per bank or nearly $61.5 million per bank in 1999 dollars. (4)
The empirical results of this paper, combined with the findings of previous research, suggest that the increase in value was not a reflection of increased riskiness of a bank's operations. Moreover, the premium cannot be explained by the exploitation of conflict of interests or even monopoly power. Ruling out these alternative hypotheses makes it more likely that the premium reflects economies of scale and scope. In fact, after controlling for the size of the bank, this premium is reduced by about 40 percent, indicating that scale effects do explain a large portion of the premium.
These results can be used to shed light on several important issues in the banking literature. First, they provide empirical support for the beliefs of many Depressionera bankers who thought that there were substantial advantages in combining commercial lending and underwriting services. Although at the time there were no sophisticated theoretical models that could show the existence of economies of scale and scope, bankers did have intuition about the benefits and the synergies associated with the combination of these services.
Second, the results indicate that there were direct costs associated with the enactment of the Glass-Steagall Act for the banking industry. (5) The separation of commercial and investment banking services necessarily reduced the scale and scope of banks with bond departments or affiliates, thereby reducing their market value. Although many argue that Glass-Steagall may have brought benefits to the financial system, these benefits must be measured against the direct and indirect costs that it imposed on the financial system. (6) One must be careful, however, in interpreting and generalizing these costs. While this paper presents evidence that there were direct costs for the banking industry, these costs do not necessarily translate to a loss for society in general. To gauge the overall impact of the enactment of Glass-Steagall, one would have to net out the gains that other financial intermediaries and institutions (for example, independent investment banking firms) obtained as a result of the enactment of the Act. Unfortunately, nearly all investment banking firms of the period were partnerships, and did not have stocks publicly traded. Thus, it is not possible to assess the possible gain in market value for these firms.
Third, these results can shed some more light on the choice of the organizational structure of these banks. In particular, banks that provided investment banking services had a choice as to how to provide them--either through the bond department or through the affiliate system. Kroszner and Rajan (1997) find that toward the late 1920s, most banks involved in the securities business were choosing the affiliate system type of organization because during that period, there was a growing perception that banks with bond departments were more prone to problems of conflict of interest. With the affiliate, they argue, it was harder for bankers to use their information to fool unsuspecting investors into buying securities of poor quality. The evidence presented by Kroszner and Rajan (1997) is supportive of their claim: They find that securities underwritten by banks with bond departments carried a substantial premium in financial markets, even after controlling for bank and client-firm characteristics. Their results imply that the affiliate system was more profitable for the bank than having a bond department. The empirical results of this paper are consistent with this implication--I find that banks with securities affiliates were more valuable than banks with bond departments after 1926, when the perception of conflict of interest problems was at its peak.
Fourth, the results of this paper offer important policy-making implications for the current wave of mergers in the commercial and investment banking industry. As Glass-Steagall restrictions are progressively relaxed, commercial banks are entering the investment banking industry at an increasing rate. (7) The conclusions of this paper suggest that there might be substantial gains from these mergers and that they might not necessarily come from gaining monopoly power.
The rest of this paper is organized as follows. Section 1 briefly documents the common perception among bankers of the pre-Glass-Steagall period regarding the existence of economies of scale and scope between the commercial and the investment banking industry. Section 2 discusses the data and the empirical findings. Section 3 presents a discussion that ties the empirical results to evidence from previous research. Section 4 offers concluding remarks.
1. ECONOMIES OF SCALE AND SCOPE IN THE PRE-GLASS-STEAGALL BANKING INDUSTRY
As mentioned in the previous section, the lack of detailed financial data from the 1920s has precluded researchers from conducting comprehensive and reliable studies of economies of scale and scope in the banking industry during that time. The only recent indirect evidence comes from White (1986), who derives an efficient frontier for banks with security affiliates and shows that there were significant diversification gains in the joint production of commercial and investment banking services. His evidence indirectly confirms the perception among bankers and financiers of that period that there were great advantages to be gained from combining commercial and investment banking services and that these advantages were the result of complementarities that would arise from the combination.
The perception of economies of scale and scope among bankers and financiers is well documented; although there was no formal survey on bankers' opinions during this period, there are many quotes and anecdotes presented in Peach (1941) that can attest to their opinions. Moreover, contemporary researchers such as Kiplinger (1928), Preston and Finlay (1930a, 1930b), and Osterweis (1932) concur with the view that there were important and advantageous synergies between the commercial and investment banking industries. (8)
Peach (1941) attributes the entrance of commercial banks into the investment banking industry to three scale- or scope-related reasons: First, during this period commercial bankers were in constant contact with the securities business and thus were able to acquire expertise and knowledge of investment banking operations. Second, because of the close and constant contact that these bankers had with the investment banking industry, it became increasingly common for clients of commercial banks to solicit investment banking advice and services from commercial bankers. Third, the strong economic conditions of the period increased the demand for the issuance of securities to finance investment spending.
The first two reasons are essentially economies-of-scope arguments. The knowledge, contacts, and connections of these bankers imply that the human capital and information acquired and developed in the process was an important resource that could be effectively used in both commercial and investment banking operations. The third reason is related to the existence of economies of scale; a high demand for the issuance of securities implies that the volume of investment banking services could only be provided by banks with a large capital base, which allowed them to underwrite large issues more frequently.
In summary, there is extensive financial history literature that directly or indirectly suggests that there may have been economies of scale and scope involved in the combination of commercial and investment banking. The following section investigates in more detail how bank charter values were affected by the existence of scale and scope economies.
2. EMPIRICAL FINDINGS
A random sample of 320 banks was obtained from the December 1927 issue of the Commercial and Financial Chronicle, Banks and Trust Companies section. The only requirement imposed on the selection of banks for inclusion in the sample was that their shares were actively traded to ensure the availability of stock market prices. In addition to obtaining stock market data for the months of December 1926, December 1927, and December 1928 (using the Commercial and Financial Chronicle), I obtained information on financial variables and balance sheet items for this sample using Moody's Directory of Bank and Trust Companies (for 1926, 1927, and 1928). Banks for which data could not be found were eliminated from the sample, reducing the sample size to 251. (9) Because the banks selected had actively traded shares, they tended to be the larger and better-established banks relative to the general population of banks. The average size (in term of assets) for the sample is $70.8 million (using 1927 figures), about 28 times larger than the average bank size for the entire population. (10)
To identify banks that were involved in the securities business (either through a bond department or a securities affiliate), the bank list was cross-searched with several issues of the Security Dealers of North America directory. Moody's Directory of Bank and Trust Companies also reported whether a bank had a securities affiliate, which I also used to check and confirm the affiliation list. (11) This source also reported each bank's year and place of incorporation, and indicated whether it was a member of the Federal Reserve System.
Table 1 presents the summary statistics for the sample, split by the type of organization structure (security affiliate, bond department, or no department) and by year. It indicates that banks with security affiliates tended to be larger and older than banks uninvolved in the securities business or banks that had bond departments. (12) The fact that they were larger is not surprising; Peach (1941), Kiplinger (1928), Preston and Findlay (1930a), White (1986), and Kroszner and Rajan (1997) document this. Kiplinger (1928), for example, argues that only large banks could afford a separate company to handle the securities business. Although he does not say it specifically, his reasoning hinges on economies-of-scale arguments.
The table indicates that there were no appreciable differences among banks in terms of the liquid assets-total assets ratio, or the capital-asset ratio. Although there are slight variations among these variables, the differences are not statistically significant. As far as profitability is concerned, the table shows that return on equity was higher for banks with affiliates than for other banks. However, the relative difference is statistically significant for 1926 and 1927 only. When profitability is measured in terms of total assets, the differences are not significant.
It is worth noting the significant increase in the number of banks with security affiliates between 1926 and 1928, at the expense of the other two categories. This is quite telling of the dynamics of the period--bankers were flocking to this industry at a very fast rate, especially toward the end of the 1920s. The tremendous increase clearly indicates that bankers were very attracted to the securities business, and is consistent with the general thinking of the period that there were valuable and unexploited opportunities in the investment banking business for commercial banks. (13)
The table also presents the average value of q for each class of bank. Following the definition commonly used in banking literature, a bank's q is defined as the market value of common equity stock (stock price times the number of shares, plus surplus and retained earnings) plus the book value of liabilities, all divided by total assets. (14) At the outset it may seem that this definition of a bank's q does not include data for the security affiliate (for banks that had them). However, according to Peach (1941), most banks with security affiliates owned the entire capital stock of the affiliate and carried it in the balance sheet of the parent bank. Bank shareholders were thus, in most cases, also pro rata shareholders of the affiliate company. Therefore, to the extent that banks carried their affiliate's capital stock as investments, q does include the affiliate's financial data.
According to the table, banks with security affiliates seemed to have been, on average, more valuable than other banks. This is one of the main findings of this paper. However, this result does not control for other conditions that may also affect q. The following section analyzes this difference in value in greater detail.
B. Estimation and Results
The value of an affiliate (or a bond department) is measured by regressing a bank's q on two dummy variables (one to identify banks with security affiliates, and the other to identify banks with bond departments), and on several control variables. Tables 2 through 4 present the main empirical findings for 1926, 1927, and 1928, respectively. The results indicate that these coefficients are estimated to be between 0.03 and 0.10, depending on the specification considered, as well as the year. Moreover, in virtually all of the specifications, the coefficients are so precisely estimated that the statistical chances of finding a positive premium where one does not, in fact, exist are less than 1 in 1,000.
Each table presents five regression specifications which differ by the set of control variables included. They are presented in a nested form in order to better see the impact that each added control variable has on the magnitude of the premium. The results are estimated for each year over the 1926-28 period in order to evaluate the robustness of the premium over time.
In regression A the control variables included are state dummies, the liquid-total asset ratio, the equity capital-total asset ratio, the natural log of the bank's age, and dummies indicating whether the bank was a member of the Federal Reserve or whether it was a national bank. State dummies are included because these banks were randomly selected from different states, and it is possible that the different regulatory environment of each state can explain at least part of the premium. If banks choose to open their affiliates or bond departments based on the degree of restrictiveness the state demonstrates regarding capital requirements, branching regulations, etc., the affiliate dummy may be just a proxy for the regulatory environment of the state.
The ratio of cash plus government securities plus other short-term (liquid) assets to total assets is included to control for the level of portfolio risk, since it is possible that the premium may be due to added risk. If it is this that drives the market value of the bank, then the inclusion of the liquid to total assets ratio should reduce the bond department/securities affiliate's premium. The results, however, show that the liquid asset-total asset ratio does not affect q.
Despite the lack of significance of the liquid asset-total asset ratio variable, risk could still be an important consideration if banks involved in the securities business increased their market value by engaging in riskier operations that variations in asset composition do not necessarily pick up. However, engaging in risky operations implies that banks in the securities business should have experienced higher failure rates, all else being constant. White (1986) reports that only 7.2 percent of national banks that dealt in the securities business failed during the period from 1930 to 1933. However, the failure rate of all national banks was 26.3 percent during the same period. Although some of the difference in these failure rates is due to the fact that national banks with security affiliates were larger than other banks, it is doubtful that the high failure rate of national banks can be attributed to bank involvement in the securities business.
The capital-total assets ratio is another way of measuring the degree of risk. Clearly, greater equity and lower leverage reduces the probability of insolvency and failure. The capital-asset ratio is included to control for this possibility. The results indicate that while this variable carries substantial explanatory power, the affiliate and bond premium coefficients easily survive its inclusion--they remain large and statistically significant.
There is another, more recent interpretation regarding the importance of including the capital-asset ratio as an independent variable in the regression. Keeley (1990) and Bhattacharya, Boot, and Thakor (1998) argue that this variable could serve as a proxy for market power. More precisely, they find that banks with more market power also tend to have higher capital-asset ratios. The argument is straightforward. Banks with monopoly power clearly have more valuable charters because the capitalized stream of future earnings is higher than that of banks operating in a competitive environment. Thus, monopoly banks have a higher incentive to keep their charters alive. Because increasing the capital-asset ratio reduces the probability of closure, all else being constant, monopoly banks should have higher capital-asset ratios. (15)
It is also possible that this premium is the result of selection bias. More specifically, the affiliate dummy could be simply picking up the experience or reputation of the bank. This could be the case if the more well-established and reputable banks are also those more likely to open an affiliate. To control for this possibility I included the log of age of the bank as an independent variable. (16) If it is true that only well-established banks are the banks that tend to have affiliates and that it is the banks' resilience that created the value, then including the number of years in business as a control variable should reduce the value of the premium significantly.
The regressions also show the effect of including the log of age. While it is generally significant in explaining q, it is somewhat sensitive to the specification considered. In particular, when measures of profitability and size are included, the coefficient declines in magnitude and statistical significance. This suggests that older banks tended to also be larger and more profitable. More importantly, however, the values of the affiliate or bond premium are not affected. It is therefore unlikely that the estimated premium is simply the result of a selection bias.
The next two control variables included are two dummies that capture whether the bank was a member of the Federal Reserve System (coded as 1, if it was a member, 0 otherwise), and a dummy indicating whether the bank was chartered at the national level (coded as 1 if it was a national bank, 0 otherwise). The purpose of including these variables was to better control for the differences in regulation that applied to national and state banks, and for banks that were members of the Federal Reserve System. Although the state dummies control for the regulatory environment at the state level, different regulations affected national banks (and Fed member state banks) as a group.
The regressions (in Tables 2, 3, and 4) show the impact of the two dummies on the results. Interestingly, the coefficients indicate that national banks were generally less valuable than their counterparts, although the estimated coefficients are significant only for 1927. It is tempting (but perhaps premature) to see this as the result of the enactment of the McFadden Act of 1927, which imposed branching restrictions on national banks. However, a more detailed analysis than the one provided here would have to be conducted to arrive at this conclusion. More importantly, however, the affiliate-bond department coefficients are not materially affected by the inclusion of these variables. It is therefore unlikely that the premium can be attributed to regulatory restrictions.
I next investigate the effects of growth on q. The inclusion of this variable relates to the notion that firms with higher q are expected to grow faster. More specifically, banks may have chosen to become involved in the securities business in order to signal that they intended to grow more rapidly than their competitors, perhaps at the expense of taking on more risk. (17) To control for this possibility, I include the growth rate of loans, measured as the percentage change in the quantity of loans for one year to the next. (18) The regressions show mixed results as far as the importance of this variable in explaining q: it is statistically insignificant for 1926, but significant for 1927. (19) Even the lagged version of this variable, used in the 1928 regressions, shows mixed results--its importance and significance varies with the regression specification considered. The magnitude of the affiliate and bond premia however, remains essentially unchanged. Thus, the market seems to have attached a premium to banks involved in the securities business regardless of their growth rate.
The next set of variables included, the return on equity (ROE) and the return on assets (ROA), control for bank profitability. The idea behind their inclusion is to ascertain the extent to which being involved in the securities business enhanced the bank's value by increasing current earnings. Regressions D and E show that while the profitability coefficients are imprecisely estimated (at least for 1926 and 1927), the magnitude of the affiliate/bond coefficient declines by about 8 percent for 1926, 30 percent for 1927, and 12 percent for 1928. This suggests that at least part of the added value was coming from enhanced bank profitability. This result is not inconsistent with the hypothesis of the paper--if scale and scope economies were present, banks which exploited them should have enjoyed a cost advantage and hence, higher earnings. (20) However, the fact that 70 to 90 percent of the affiliate and bond premia survive the inclusion of current earnings measures suggests that investors were also valuing the capitalized value of future higher earnings from becoming involved in the securities business. (21)
I next consider the inclusion of size as a way of ascertaining the importance of economies of scale. Regression E shows the effect of adding total loans to the specification.(22) The magnitude of the security affiliate premium is reduced by 37 percent for 1926, 42 percent for 1927, and 46 percent for 1928. Likewise, the bond department premium is reduced by 27 percent for 1926, 50 percent for 1927, and 34 percent for 1928. This suggests that there are important economies of scale effects--larger banks were more likely to be involved in the securities business. This result was first documented by Preston and Finlay (1930a, 1930b) and Peach (1941) and more recently by White (1986) and Kroszner and Rajan (1997).
There is, however, the other 50 to 60 percent of the premium that neither size nor current profitability measures can explain. It is quite possible that economies of scope can explain it. However, in order to be more certain that economies of scope can explain the remaining part of the premium, it is important to rule out other alternatives which the regressions in Table 2, 3, and 4 may not be able to control.
One important alternative is the possibility that the stock market was overvalued in the late 1920s, prior to the stock market crash of 1929. Many firms, including those involved in the securities business, had experienced a tremendous increase in their stock prices during this period. Hence, it is possible that the estimated affiliate/ bond premium is a reflection of these optimistic sentiments.
While the debate in the literature regarding the extent to which there was an over-valuation of stock prices during this period has not been settled, the more conservative accounts put March of 1928 as the beginning of the overvaluation period (for example, Galbraith 1954). (23) Thus, to the extent that overvaluation is an issue, it is more likely to affect the 1928 regression results, rather than those of 1926 or 1927.
The results show that the estimated affiliate-bond coefficients are indeed larger for 1928 than those estimated for the previous two years. While the premium ranges between 4 and 7 percent for 1926 and 1927, it jumps by nearly 40 percent (from 7 to about 10 percent) in 1928. Thus, although this increase is consistent with the overvaluation hypothesis, it is unlikely that the premium is being driven solely by the overvaluation of stock prices, especially for 1926 and 1927.
Despite this finding, it is still possible to argue that 1926 and 1927 still reflected some degree of exuberance in the market as prices and volume were rising rapidly. This may result in an inflated measure of the long-run cost of Glass-Steagall, as the q-ratios used in the regressions are computed using market value of stock prices. Unfortunately, it is difficult to fully control for this possibility, because there is no clear and precise definition of overvaluation. One possibility is to look at the behavior of financial ratios such as multiples of price-earnings or price-dividends as indicators of market price relative to fundamental values. It is typically considered to be a symptom of exuberant market conditions when stock prices are rapidly outstripping fundamental values during a particular time period. Following Campbell and Shiller's (1988) methodology, White (1990) looks at the behavior of dividends for firms in the Dow Jones index from 1922 to 1930. His results are self-explanatory: "From 1922 to 1927 dividends and prices moved together, but while dividends continued to grow rather smoothly in 1928 and 1929, stock prices soared far above them" (p. 72). Thus, although White's (1990) results do not necessarily prove that 1926 and 1927 were years of normal stock price behavior, they at least suggest that they were not out of line from reasonable expectations of fundamental values.
Another important alternative is the possibility that banks involved in the securities business exploited conflicts of interest to the detriment of unsuspecting investors. If this is correct, it would imply that these banks may have been able to extract extra profits from their investment banking activities (and therefore increase their market values) by selling overpriced, poor-quality securities to investors. Recent research, however, has not been able to find any evidence supporting this allegation (for example, Kroszner and Rajan 1994; Puri 1994; Dhawan 1994; Benston 1990). Despite this lack of evidence, Kroszner and Rajan (1997) argue that banks with bond departments may have been more susceptible to conflict of interest problems than banks organized with separate security affiliates. Thus, to the extent that conflict of interest problems existed, they were more likely to occur among banks with bond departments than among banks with security affiliates. Their findings suggest that, all else being constant, banks with bond departments should have been less, not more, profitable than banks with security affiliates because securities underwritten through bond departments carried larger discounts (and thus higher yields) than securities underwritten through affiliates. This implies that banks with security affiliates should be more valuable than banks with bond departments.
By adding different dummies for security affiliates and bond departments, the regressions control for the different types of organizational structures. The results show that the coefficient of the bond department dummy is about 20 to 25 percent lower than that of the securities affiliate for 1927 and 1928. However, it is virtually the same when considering the results for 1926. (24) This suggests that to the extent there was a conflict of interest problem, it seems to have been reflected in the bank's market value after 1926, not before. This result is consistent with the historiography of the perception of conflicts of interests during this period. For example, Senator Carter Glass began denouncing commercial bank involvement in investment banking in 1928 (for example, Perkins 1971; Kelly 1985). This can explain why, by the end of the 1920s, banks with bond departments were switching to security affiliates at an increasing rate.
3. CONCLUDING REMARKS
According to Peach (1941), in 1922 there were 277 banks that were either operating a securities affiliate or were directly engaged in the securities business through a bond department. By 1929 this figure had reached 591, implying an annual growth rate of almost 11 percent. Thus, banks were flocking to this industry at a very fast rate. It is evident that investment banking was an attractive business for commercial banks, handsomely rewarded in financial markets with a premium of about 4 to 7 percent. Undoubtedly, the attractiveness stemmed from higher profitability. The evidence presented here suggests that this added profitability and value did not seem to be coming from added risk, more monopoly power, or conflict of interest problems. That size can explain about one-half of the premium suggests that economies of scale played an important role.
There is some indirect evidence that economies of scope can explain the remainder of the premium. Theoretically, economies of scope arise from the joint use of a given resource. For banks with affiliates, this common resource was the human capital and information that bankers developed about their corporate clients, which Peach (1941), among others, documents.
Empirically, the work of White (1986) and Ramirez (1999) also indirectly supports this argument. White (1986), for example, shows that there may have been diversification gains from the combination of commercial and investment banking activities. Ramirez (1999), using firm-level data from the 1920s through the 1950s, finds that during the 1920s firms that attached themselves to banks with security affiliates tended to have fewer liquidity problems than companies that did not. During the 1930s, after the New Deal financial market reforms were implemented, this effect disappears: Attachment to a bank no longer implied being freed of liquidity constraints. This change suggests that the Glass-Steagall Act eroded bankers' ability to provide liquidity needs for long-term financing, which they were able to do previously through their security affiliates or bond departments. Although one can never be sure, given all of this evidence it is difficult to find convincing alternative explanations to scale and scope arguments.
One direct implication of this result is that we can now estimate the direct cost to the banking industry of having banks disallowed from involvement in investment banking--about $8 million of 1927 dollars per bank, roughly equivalent to about $61.5 million per bank today. This cost in lost value to the banking industry, however, does not necessarily translate to an estimate of the social cost of Glass-Steagall. To address this issue, one would have to net out the gain in value that other financial intermediaries and institutions (namely, independent investment banking firms) may have experienced as a result of the enactment of this Act. Nevertheless, it is quite telling that today the American financial system seeks to have a closer integration of commercial and investment banking operations than the Glass-Steagall Act allowed. Certainly there appear to be substantial private profits from such integration; such integration is desired by both commercial and investment banks. Based on the results of this paper, it is difficult to see such profits as being derived totally from either greater ease of monopolization, greater risk, or from diverting returns away from depositors.
TABLE 1 SUMMARY STATISTICS Banks with Variable Year Security Aff. 1926 1.081 (0.094) Tobin's q 1927 1.115 (0.082) 1928 1.156 (0.155) 1926 0.394 (0.103) Liq/Assets 1927 0.353 (0.091) 1928 0.374 (0.092) 1926 0.101 (0.040) Cap/Assets 1927 0.113 (0.056) 1928 0.120 (0.048) 1926 35.77 (21.19) Age 1927 32.57 (24.39) 1928 31.30 (24.06) 1926 17.652 (1.830) Ln (loans) 1927 17.460 (1.688) 1928 17.570 (1.694) 1926 286 (379) Total Assets 1927 229 (344) 1928 258 (385) 1926 0.137 (0.056) ROE 1927 0.127 (0.052) 1928 0.108 (0.056) 1926 0.014 (0.007) ROA 1927 0.013 (0.006) 1928 0.012 (0.008) 1926 0.522 (0.509) Percent Nat. Banks 1927 0.488 (0.505) 1928 0.482 (0.504) 1926 26 No. of Obs. 1927 45 1928 56 Banks with Bond Variable Year Departments 1926 1.080 (0.086) Tobin's q 1927 1.096 (0.096) 1928 1.127 (0.119) 1926 0.392 (0.164) Liq/Assets 1927 0.374 (0.134) 1928 0.393 (0.120) 1926 0.125 (0.062) Cap/Assets 1927 0.125 (0.042) 1928 0.126 (0.045) 1926 27.19 (21.843) Age 1927 28.05 (19.838) 1928 30.63 (23.657) 1926 16.377 (1.454) Ln (loans) 1927 16.214 (1.418) 1928 16.273 (1.440) 1926 64.3 (107) Total Assets 1927 59.1 (118) 1928 67.1 (130) 1926 0.113 (0.040) ROE 1927 0.104 (0.035) 1928 0.106 (0.041) 1926 0.012 (0.005) ROA 1927 0.012 (0.006) 1928 0.013 (0.006) 1926 0.323 (0.471) Percent Nat. Banks 1927 0.311 (0.467) 1928 0.293 (0.459) 1926 68 No. of Obs. 1927 61 1928 58 Banks Variable Year with Neither 1926 1.022 (0.115) Tobin's q 1927 1.037 (0.099) 1928 1.049 (0.091) 1926 0.387 (0.131) Liq/Assets 1927 0.383 (0.131) 1928 0.402 (0.114) 1926 0.131 (0.072) Cap/Assets 1927 0.129 (0.054) 1928 0.124 (0.043) 1926 24.36 (21.970) Age 1927 23.74 (21.863) 1928 23.97 (20.005) 1926 15.311 (1.290) Ln (loans) 1927 15.256 (1.231) 1928 15.216 (1.170) 1926 29.4 (80.7) Total Assets 1927 26.7 (78.5) 1928 28.4 (101.0) 1926 0.108 (0.046) ROE 1927 0.099 (0.035) 1928 0.095 (0.035) 1926 0.012 (0.005) ROA 1927 0.012 (0.006) 1928 0.011 (0.004) 1926 0.363 (0.482) Percent Nat. Banks 1927 0.351 (0.479) 1928 0.350 (0.478) 1926 157 No. of Obs. 1927 145 1928 137 NOTES: This table presents average statistics for the following variables: "Tobin's q" is the market value of the bank (as defined in the text) divided by total assets; "Liq/Assets" is the ratio of cash plus government securities plus other marketable securities to total assets; "Cap/Assets" is the ratio of the book value of common stock divided by total assets; "Age" is the number of years the bank has been in business since incorporation; "Ln (loans)" is the natural log of loans; "Total Assets" is reported in millions of dollars; "ROE" is net earnings divided by the book value of equity; "ROA" is net earnings divided by total assets; "Percent Nat. Banks" is the percentage of banks chartered at the national level. Standard deviations are in parentheses. TABLE 2 REGRESSION RESULTS FOR 1926 Var Reg. A Reg. B Reg. C Constant 0.881 (a) 0.876 (a) 0.885 (a) (0.044) (0.046) (0.056) Security Aff. 0.059 (a) 0.059 (a) 0.044 (a) (0.019) (0.019) (0.020) Bond Dept. 0.055 (a) 0.054 (a) 0.053 (a) (0.013) (0.012) (0.015) Liq/Assets 0.055 (c) 0.055 (c) 0.005 (0.037) (0.037) (0.034) Cap/Assets 0.553 (a) 0.498 (a) 0.667 (a) (0.142) (0.138) (0.184) Ln(Age) 0.027 (a) 0.029 (a) 0.020 (a) (0.008) (0.009) (0.010) Fed Mem -0.006 -0.008 -0.019 (0.015) (0.014) (0.016) Nat Bank -0.022 (c) -0.020 -0.022 (0.015) (0.015) (0.017) Loan Growth 0.033 -0.037 (0.028) (0.039) ROE 0.269 (c) (0.173) ROA Ln (loans) State Dum? Yes Yes Yes R-sqd. 0.235 0.24 0.263 Root MSE 0.099 0.099 0.078 F-Statistic 4.68 (a) 4.21 (a) 25.32 (a) No. Obs. 224 224 145 Var Reg. D Reg. E Constant 0.917 (a) 0.650 (a) (0.046) (0.114) Security Aff. 0.046 (a) 0.029 (0.021) (0.022) Bond Dept. 0.052 (a) 0.038 (a) (0.015) (0.016) Liq/Assets 0.005 0.040 (0.034) (0.040) Cap/Assets 0.509 (a) 0.610 (a) (0.202) (0.201) Ln(Age) 0.018 (b) 0.009 (0.010) (0.010) Fed Mem -0.018 -0.025 (c) (0.015) (0.015) Nat Bank -0.024 -0.024 (c) (0.017) (0.016) Loan Growth -0.025 -0.025 (0.039) (0.038) ROE ROA 1.965 1.679 (1.715) (1.660) Ln (loans) 0.016 (a) (0.006) State Dum? Yes Yes R-sqd. 0.264 0.291 Root MSE 0.079 0.077 F-Statistic 26.09 (a) 19.35 (a) No. Obs. 145 145 NOTES: Dependent variable: Tobin's q. Independent variables: "Affiliate/Bond" is a dummy variable equal to 1 if the bank had a security affiliate or a bond department, 0 otherwise; "Liq/Assets" is the ratio of cash plus government securities plus other marketable securities to total assets; "Cap/Assets" is the ratio of the book value of tier 1 equity (common stock, surplus, and undivided profit) divided by total assets; "Ln(Age)" is the natural log of the bank's age (years since incorporation); "Fed Mem" equals 1 if the bank was a member of the Federal Reserve System, 0 otherwise; "Nat Bank" equals 1 if the bank was a national bank, 0 otherwise; "Loan growth" is percentage change in loans from 1927 to 1928; "ROE" is net earnings divided by the book value of equity; "ROA" is net earnings divided by total assets; "Ln(Loans)" is the natural log of total loans; "State Dum?" indicates whether dummy variables where introduced for each state. Robust (White (1980)-corrected) standard errors are reported in parenthesis. a = p < [absolute value of 0.051]; b = p < [absolute value of 0.10]; c = p < [absolute value of 0.15]. TABLE 3 REGRESSION RESULTS FOR 1927 Var Reg. A Reg. B Reg. C Constant 0.927 (a) 0.906 (a) 0.848 (a) (0.052) (0.054) (0.064) Security Aff. 0.088 (a) 0.095 (a) 0.068 (a) (0.017) (0.017) (0.023) Bond Dept. 0.055 (a) 0.056 (a) 0.042 (a) (0.014) (0.014) (0.019) Liq/Assets 0.068 (b) 0.054 0.022 (0.041) (0.044) (0.057) Cap/Assets 0.557 (a) 0.567 (a) 0.763 (a) (0.130) (0.154) (0.167) Ln(Age) 0.018 (a) 0.019 (a) 0.014 (0.008) (0.008) (0.010) Fed Mem 0.001 0.003 0.007 (0.015) (0.016) (0.018) Nat Bank -0.045 (a) -0.054 (a) -0.044 (a) (0.015) (0.016) (0.019) Loan Growth 0.042 (a) 0.048 (a) (0.018) (0.021) ROE 0.486 (0.356) ROA Ln (loans) State Dum? Yes Yes Yes R-sqd. 0.282 0.304 0.304 Root MSE 0.089 0.089 0.090 F-Statistic 6.25 (a) 7.29 (a) 7.81 (a) No. Obs. 247 230 165 Var Reg. D Reg. E Constant 0.902 (a) 0.432 (a) (0.078) (0.220) Security Aff. 0.074 (a) 0.043 (a) (0.021) (0.021) Bond Dept. 0.044 (a) 0.022 (0.018) (0.017) Liq/Assets 0.027 0.094 (c) (0.057) (0.062) Cap/Assets 0.452 (a) 0.738 (a) (0.265) (0.297) Ln(Age) 0.013 0.001 (0.011) (0.009) Fed Mem 0.008 -0.005 (0.018) (0.016) Nat Bank -0.046 (a) -0.045 (a) (0.019) (0.018) Loan Growth 0.043 (b) 0.050 (a) (0.022) (0.023) ROE ROA 3.125 2.081 (2.537) (2.519) Ln (loans) 0.027 (a) (0.010) State Dum? Yes Yes R-sqd. 0.297 0.357 Root MSE 0.090 0.087 F-Statistic 6.78 (a) 7.79 (a) No. Obs. 165 165 Notes: Dependent variable: Tobin's q. Independent variables: "Affiliate/Bond" is a dummy variable equal to 1 if the bank had a security affiliate or a bond department, 0 otherwise; "Liq/Assets" is the ratio of cash plus government securities plus other marketable securities to total assets; "Cap/Assets" is the ratio of the book value of tier 1 equity (common stock, surplus, and undivided profit) divided by total assets; "Ln(Age)" is the natural log of the bank's age (years since incorporation); "Fed Mem" equals 1 if the bank was a member of the Federal Reserve System, 0 otherwise; "Nat Bank" equals 1 if the bank was a national bank, 0 otherwise; "Loan growth" is percentage change in loans from 1927 to 1928; "ROE" is net earnings divided by the book value of equity; "ROA" is net earnings divided by total assets; "Ln(Loans)" is the natural log of total loans; "State Dum?" indicates whether dummy variables where introduced for each state. Robust (White (1980)-corrected) standard errors are reported in parenthesis, a = p < [absolute value of 0.05]; b = p < [absolute value of 0.10]; c = p < [absolute value of 0.15]. TABLE 4 REGRESSION RESULTS FOR 1928 Var Reg. A Reg. B Reg. C Constant 0.967 (a) 0.951 (a) 0.908 (b) (0.052) (0.051) (0.064) Security Aff. 0.106 (a) 0.104 (a) 0.091 (a) (0.022) (0.022) (0.021) Bond Dept. 0.077 (a) 0.079 (a) 0.066 (a) (0.018) (0.018) (0.019) Liq/Assets 0.001 -0.016 -0.099 (0.066) (0.066) (0.076) Cap/Assets 0.690 (a) 0.696 (a) 0.927 (a) (0.198) (0.194) (0.206) Ln(Age) 0.019 (b) 0.022 (a) 0.016 (0.010) (0.010) (0.011) Fed Mem -0.001 -0.005 -0.004 (0.020) (0.020) (0.019) Nat Bank -0.033 (c) -0.035 (b) -0.018 (0.022) (0.023) (0.022) Loan Growth 0.054 (b) 0.007 (lagged) (0.030) (0.030) ROE 0.702 (a) (0.237) ROA Ln (loans) State Dum? Yes Yes Yes R-sqd. 0.308 0.323 0.374 Root MSE 0.106 0.105 0.097 F-Statistic 6.15 (a) 6.43 (a) 6.59 (a) No. Obs. 228 228 173 Var Reg. D Reg. E Constant 0.975 (a) 0.436 (a) (0.057) (0.177) Security Aff. 0.096 (a) 0.052 (a) (0.022) (0.023) Bond Dept. 0.067 (a) 0.044 (a) (0.020) (0.018) Liq/Assets -0.091 0.001 (0.076) (0.071) Cap/Assets 0.541 (a) 0.816 (a) (0.206) (0.220) Ln(Age) 0.013 -0.005 (0.011) (0.012) Fed Mem -0.002 -0.016 (0.019) (0.018) Nat Bank -0.020 -0.014 (0.022) (0.020) Loan Growth 0.003 0.001 (lagged) (0.029) (0.027) ROE ROA 4.817 (a) 3.799 (a) (1.696) (1.431) Ln (loans) 0.032 (a) (0.010) State Dum? Yes Yes R-sqd. 0.368 0.425 Root MSE 0.097 0.093 F-Statistic 6.72 (a) 7.34 (a) No. Obs. 173 173 Notes: Dependent variable: Tobin's q. Independent variables: "Affiliate/Bond" is a dummy variable equal to 1 if the bank had a security affiliate or a bond department, 0 otherwise; "Liq/Assets" is the ratio of cash plus government securities plus other marketable securities to total assets; "Cap/Assets" is the ratio of the book value of tier 1 equity (common stock, surplus, and undivided profit) divided by total assets; "Ln(Age)" is the natural log of the bank's age (years since incorporation); "Fed Mem" equals 1 if the bank was a member of the Federal Reserve System, 0 otherwise; "Nat Bank" equals 1 if the bank was a national bank, 0 otherwise; "Loan growth (lagged)" is percentage change in loans from 1927 to 1928; "ROE" is net earnings divided by the book value of equity; "ROA" is net earnings divided by total assets; "Ln(Loans)" is the natural log of total loans; "State Dum?" indicates whether dummy variables where introduced for each state. Robust (White (1980)-corrected) standard errors are reported in parenthesis. a = p < [absolute value of 0.05]; b = p < [absolute value of 0.10]; c = p < [absolute value of 0.15].
I would like to thank, without implicating, Charles Calomiris, Marshall Reinsdorf, Pu Shen, Ling Hui Tan, participants at the 4th Washington Area Finance Conference, as well as two anonymous referees for their helpful comments and suggestions. Daniel Lastra provided very valuable research assistance.
(1.) See, for example, Baliga and Polak (1995), Calomiris and Hubbard (1990), Diamond (1991), Rajan (1992), Sharpe (1990), and Webb (1991).
(2.) For evidence suggesting that banks eased corporate financing problems see Ramirez (1999) for the United States, Hoshi, Kashyap, and Scharfstein (1991) for Japan, and Calomiris (1995) for Germany. For evidence of higher market value see James (1987), De Long (1991), and Ramirez and De Long (1995).
(3.) There is a large body of research on this topic in the banking literature. For a comprehensive survey see Berger and Humphrey (1997), Humphrey (1990), and section 5 of Berger, Demsetz, and Strahan (1999).
(4.) The figures for 1999 are computed using the producer price index, all commodities series, available from the Bureau of Labor Statistics web page (http://www.bls.gov).
(5.) At the outset, it may seem more appropriate to measure the impact of Glass-Steagall by estimating the magnitude of the securities affiliate premium in 1931-33, just before the Act was enacted. There are at least two reasons why this may not be as suitable as it seems. First, the banking industry was experiencing arguably one of its most serious crises in history during this period. Second, although the Glass-Steagall bill had not fully matured, by 1931 the passing of far-reaching financial legislation which included the separation of commercial and investment banking, more branching restrictions, and even deposit insurance was all but imminent. With these many significant events, turmoils, and regulatory uncertainty in the banking industry, it is not possible in a practical sense to reliably identify and measure the added value for a bank of being involved in the securities business during this period.
(6.) The loss of market value is not the only cost that Glass-Steagall may have imposed. Ramirez (1999) argues that this Act may have also increased the cost of capital for corporations.
(7.) See, for example, section 2 of Berger, Demsetz, and Strahan (1999), and section 1 of Calomiris and Karceski (1998) for detailed evidence.
(8.) Kroszner and Rajan (1997) also note the prevalence of this perception among bankers during this period.
(9.) Not all banks reported complete financial information for all three years. Hence, the number of observations used in the regressions varies somewhat depending on the specification considered.
(10.) Statistics for the entire population of banks were computed using data from All Bank Statistics: United States, 1896-1955 (Board of Governors of the Federal Reserve System 1959).
(11.) The proportion of banks identified as being involved in the securities industry (either through a bond department or through an affiliate) is 37 percent for 1926, 42 percent for 1927, and 45 percent for 1928. This relatively high number is not surprising given that the population from which the sample was drawn is the set of banks that had actively traded shares, not the general population of banks.
(12.) Difference of means tests show that these differences (in size and age) were significant at the 5 percent level for every year.
(13.) Between 1926 and 1927 alone, nineteen banks in the sample switched to having a securities affiliate. Of these, thirteen switched from having a bond department, and six were previously not involved in the securities business before 1927. This pattern was very typical of the period, as Peach (1941) documents. It is worth pointing out that for this sample of nineteen banks, the average q increased by 4.32 percent, a figure in the neighborhood of what the regressions show.
(14.) This measure of q is used, for example, by Keeley (1990) and Allen and Rai (1996).
(15.) Although this argument is intuitively appealing, it is important to recognize that this period was characterized by relatively less regulation and freer entry. It is therefore unlikely that monopoly power can explain the affiliate/bond premium.
(16.) I include the log of age (defined as the number of years the bank had been operating since the charter was granted) in order to control for a nonlinear effect. Using levels instead does not change the results. Not all banks reported the year in which they were organized, hence the number of observations is reduced by two in the 1926 sample, and by four in the 1927 sample.
(17.) It is possible that banks with high growth rates assume greater risk as credit review procedures become less rigorous.
(18.) I use total loans as a proxy for bank output, as is commonly done in the banking literature. This approximation implicitly assumes that the services that banks provide are proportional to the value of loans.
(19.) I was not able to get loan figures for seventeen banks in 1926, hence the number of observations was reduced to 230 for the 1927 regressions. I also lost a few observations for 1928 due to the unavailability of financial data as well as two mergers.
(20.) This is not the only possible interpretation, however. Higher relative returns may also suggest higher exposure to risk. Normally it is possible to sort out which of these two possibilities is the more likely one by also looking at additional risk measures like the ratios of loan losses and past due loans to net loans. Unfortunately, these data are not available for this period, as there were no specific disclosure requirements like those of today.
(21.) Unfortunately, a significant number of banks did not report earnings figures. This reduced the number of observations to 145 for 1926, 165 for 1927, and 173 for 1928.
(22.) I use the log of loans in the specification in order to allow for a possible nonlinear effect. Although using the level (and its square) produces a similar set of results, the coefficients are harder to interpret intuitively.
(23.) For an overview of the debate see White (1990) and Cecchetti (1994).
(24.) The security affiliate coefficient is not statistically significant in regression E of Table 2 at the standard levels. While the coefficient is positive, the standard errors are large enough to push level of significance to 20 percent only. The bond department coefficient in regression E of Table 3 is significant at the 10 percent level by a one-tailed test.
Allen, Linda, and Anoop Rai. "Bank Charter Values and Capital Levels: An International Comparison." Journal of Economics and Business 48 (1996), 269-84.
Baliga, Sandeep, and Ben Polak. "Bank versus Bonds: A Simple Theory of Comparative Financial Institutions." Cowles Foundation Discussion Paper no. 1100. Cowles Foundation, Yale University, 1995.
Becht, Marco, and Carlos D. Ramirez. "Does Bank Affiliation Mitigate Liquidity Constraints? Evidence from Germany's Universal Banks in the Pre-World War I Period." Working Paper, George Mason University. 2000.
Benston, George J. The Separation of Commercial and Investment Banking, The Glass-Steagall Act Revisited and Reconsidered. New York: Oxford University Press, 1990.
Berger, Allen N., Rebecca S. Demsetz, and Philip E. Strahan. "The Consolidation of the Financial Services Industry: Causes, Consequences, and Implications for the Future." Journal of Banking and Finance 23 (1999), 135-94.
Berger, Allen N., and David B. Humphrey. "Efficiency of Financial Institutions: International Survey and Directions for Future Research." European Journal of Operational Research 98 (1997), 175-212.
Berger, Allen N., and Loretta J. Mester. "Inside the Black Box: What Explains Differences in the Efficiencies of Financial Institutions?" Journal of Banking and Finance 21 (1997), 895-947.
Bhattacharya, Sudipto, Arnoud W.A. Boot, and Anjan V. Thakor. "The Economics of Bank Regulation." Journal of Money, Credit, and Banking 30 (1998), 745-70.
Board of Governors of the Federal Reserve System. All-Bank Statistics: United States 1896-1955. 1959.
Calomiris, Charles W. "The Cost of Rejecting Universal Banking: American Finance in the German Mirror, 1870-1914" In Coordination and Information: Historical Perspectives on the Organization of Enterprise, edited by Naomi R. Lamoreaux and Daniel M.G. Raff. Chicago: University of Chicago Press, 1995.
Calomiris, Charles W., and R. Glenn Hubbard. "Firm Heterogeneity, Internal Finance, and `Credit Rationing.'" The Economic Journal 100 (1990), 90-104.
Calomiris, Charles W., and Jason Karceski. "Is the Bank Merger Wave of the 1990s Efficient? Lessons from Nine Case Studies." In Mergers and Productivity, edited by S. Kaplan. Chicago: University of Chicago Press, 1998.
Calomiris, Charles W., and Carlos D. Ramirez. "The Evolution of American Corporate Finance." In The American Corporation: A Second Look, edited by Carl Kaysen. Oxford University Press, 1995.
Campbell, John Y., and Robert J. Shiller. "Stock Prices, Earnings, and Expected Dividends." Journal of Finance 43 (1988), 661-76.
Carosso, Vincent P. Investment Banking in America: A History. Cambridge, Mass.: Harvard University Press, 1970.
Cecchetti, Stephen G. "Stock Market Crash of October 1929." In New Palgrave Dictionary of Money and Finance, vol. 3, pp. 573-77, London: Macmillan Press Ltd., 1994.
Clark, Jeffrey A. "Economies of Scale and Scope at Depository Financial Institutions: A Review of the Literature." Economic Review, Federal Reserve Bank of Kansas City (September/October 1988), 16-33.
Commercial and Financial Chronicle. 1927-28, Bank and Trust Companies Section, various issues, New York.
De Long, J. Bradford. "Did J. P. Morgan's Men Add Value? An Economist's Perspective on Financial Capitalism." In Inside the Business Enterprise: Historical Perspectives on the Use of Information, edited by Peter Temin. Chicago: University of Chicago Press, 1991.
Dhawan, Juhi. "Was Universal Banking Prevented by Regulation? The U.S. Experience prior to the Glass-Steagall Act." Brown University Working Paper, 1994.
Diamond, Douglas. "Monitoring and Reputation: The Choice between Bank Loans and Directly Placed Debt." Journal of Political Economy 99 (1991), 689-721.
Galbraith, John K. The Great Crash, 1929. Boston: Houghton Mifflin Company, 1954.
Hoshi, Takeo, Anil Kashyap, and David Scharfstein. "Corporate Structure, Liquidity, and Investment: Evidence from Japanese Industrial Groups." Quarterly Journal of Economics 106 (1991), 33-60.
Humphrey, David B. "Why Do Estimates of Bank Scale Economies Differ?" Economic Review, Federal Reserve Bank of Richmond (September/October 1990), 38-50.
James, Christopher. "Some Evidence on the Uniqueness of Bank Loans." Journal of Financial Economics, 19 (1987), 217-35.
Keeley, Michael C. "Deposit Insurance, Risk, and Market Power in Banking." American Economic Review 80 (May 1990), 1183-1200.
Kelly III, Edward J. "Legislative History of the Glass-Steagall Act." In Deregulating Wall Street, edited by Ingo Walter. New York: John Wiley & Sons, 1985.
Kiplinger, W. M. "National Banks Become Security Dealers." American Bankers Association Journal 20 (September 1928), 211, 239-40.
Kroszner, Randall S., and Raghuram G. Rajan. "Is the Glass-Steagall Act Justified? A Study of the U.S. Experience with Universal Banking before 1933." American Economic Review 84 (1994), 810-832.
--. "Organization Structure and Credibility: Evidence from Commercial Bank Securities Activities before the Glass-Steagall Act." Journal of Monetary Economics, 39 (1997), 475-516.
Moody's Directory of Bank and Trust Companies, 1928-29. New York.
Osterweis, S. "Security Affiliates and Security Operations of Commercial Banks." Harvard Business Review 11 (1932), 124-31.
Peach, W. Nelson. The Security Affiliates of National Banks. Baltimore: The Johns Hopkins University Press, 1941.
Perkins, Edwin J. "The Divorce of Commercial and Investment Banking: A History." The Banking Law Journal 88 (1971), 483-528.
Preston, H.H., and Allan R. Finlay. "Investment Affiliates Thrive." American Bankers Association Journal 22 (May 1930a) 1027-1028, 1075.
--. "Era Favors Investment Affiliates." American Bankers Association Journal 22 (June 1930b), 1153-54, 1191, 1199.
Puri, Manju. "The Long-Term Default Performance of Bank Underwritten Securities Issues." Journal of Banking and Finance (March, 1994), 397-418.
Rajan, Raghuram. "Insiders and Outsiders: The Choice between Informed and Arm's Length Debt." Journal of Finance 47 (1992), 1367-1400.
Ramirez, Carlos D. "Did Glass Steagall Increase the Cost of External Finance for Corporate Investment? Evidence from Bank and Insurance Company Affiliations." Journal of Economic History (June 1999), 372-96.
Ramirez, Carlos D., and J. Bradford De Long. "Bank Influence and Firm Performance: The Impact of Depression-Era Financial Markets Reforms" In Anglo-American Finance Systems: Institutions and Markets in the 20th Century, edited by Michael Bordo and Richard Sylla. Irwin Publishers, 1995.
Roe, Mark J. Strong Managers, Weak Owners. Princeton, N.J.: Princeton University Press, 1994.
Security Dealers of North America. New York: Herbert D. Seibert & Co., 1927-29.
Sharpe, Steven A. "Asymmetric Information, Bank Lending, and Implicit Contracts: A Stylized Model of Customer Relationships." Journal of Finance 45 (April 1990), 1069-87.
Stiglitz, Joseph, and Andrew Weiss. "Credit Rationing in Markets with Imperfect Information." American Economic Review 71 (June 1981), 393-410.
Webb, David. "Long-Term Financial Contracts Can Mitigate the Adverse Selection Problem in Project Financing." International Economic Review 32 (1991), 305-20.
White, Eugene N. "Before the Glass-Steagall Act: An Analysis of the Investment Banking Activities of National Banks." Explorations in Economic History 23 (1986), 33-55.
--. "The Stock Market Boom and Crash of 1929 Revisited." Journal of Economic Perspectives 4 (1990), 67-83.
White, Halbert A. "Heteroskedasticity-Consistent Covariance Matrix Estimator and a Direct Test for Heteroskedasticity." Econometrica 48 (1980), 817-38.
CARLOS D. RAMIREZ is assistant professor of economics at George Mason University. E-mail: firstname.lastname@example.org…