Profit efficiency in U.S. BHCs: Effects of increasing non-traditional revenue sources

The Quarterly Review of Economics and Finance 50 (2010) 132–140

Contents lists available at ScienceDirect

The Quarterly Review of Economics and Finance journal homepage: www.elsevier.com/locate/qref

Profit efficiency in U.S. BHCs: Effects of increasing non-traditional revenue sources Aigbe Akhigbe a , Bradley A. Stevenson b,∗ a b

The University of Akron, Akron, OH, United States Bellarmine University, Louisville, KY, United States

a r t i c l e

i n f o

Article history: Received 23 September 2008 Received in revised form 29 October 2009 Accepted 6 November 2009 Available online 18 November 2009 JEL classification: G21 D02

a b s t r a c t Using information multiple times across revenue streams, BHCs may increase efficiency due to economies of scope. Our main contribution is to be the first to examine noninterest income after passage of the Gramm-Leach-Bliley Act, when additional opportunities to increase noninterest income arise. We examine profit efficiency and its relationship to noninterest income for BHCs using stochastic frontier analysis and multivariate analysis on BHC data from 2003 to 2006. Contrary to our hypothesis, the results indicate multi-noninterest income types are associated with decreased profit efficiency. These results are robust using the Efficiency Ratio as our measure and are particularly strong for small BHCs. © 2009 The Board of Trustees of the University of Illinois. Published by Elsevier B.V. All rights reserved.

Keywords: Financial Services Modernization Act Noninterest income Fees Bank efficiency Frontier analysis

1. Introduction There is a great deal of published theory explaining why banks exist. One common thread through most of the theoretical literature is the role of asymmetric information and the bank’s ability to overcome asymmetric information. Pyle (1971) shows the importance of the interaction between assets and liabilities while Leland and Pyle (1977) develops a model where banks overcome asymmetric information to improve markets. A quintessential model of banking is developed in Diamond (1984) where he shows financial intermediaries minimize the cost of monitoring which proves useful for reducing incentive problems between borrowers and lenders. In addition, he shows diversification provided by intermediation is the key to reducing costs, even in a risk neutral economy. Using this notion, we propose that outside of the traditional banking model there is room for efficiency gains by diversifying across income types and reusing information that banks generate in their lending activities in the noninterest income arena including areas

more prominent after the Gramm-Leach-Bliley (GLB) Act such as underwriting, venture capital and insurance. Indeed, the main research contribution of this paper is that we are the first to focus on the post-GLB period when more sources of noninterest income are available. However, contrary to this hypothesis we find that less efficient bank holding companies (hereafter, BHCs) are more likely to have expanded into these non-traditional forms of revenue. Many lines of research have looked at what makes banks special and why banks function as they do. Related to traditional banking activity, noting that CD rates paid by banks are equivalent to other comparable securities, Fama (1985) indicates that the reserve “tax” must be borne by the bank’s borrowers.1 Fama asserts that the borrowers bear this “tax” in the form of higher rates because of the monitoring service provided by banks. This monitoring service helps overcome the asymmetric information problem discussed by Leland and Pyle (1977). In James (1987), the CD finding of Fama (1985) is confirmed and excess, positive returns surrounding loan announcements also indicate there is something special about bank

∗ Corresponding author at: W. Fielding Rubel School of Business, Bellarmine University, 2001 Newburg Road, Louisville, KY 40205, United States. Tel.: +1 502 452 8173. E-mail address: [email protected] (B.A. Stevenson).

1 Commercial banks holding reserves in their vault or at the Federal Reserve Bank do not earn income on those reserves. Thus this opportunity cost is often viewed as a “tax”.

1062-9769/$ – see front matter © 2009 The Board of Trustees of the University of Illinois. Published by Elsevier B.V. All rights reserved. doi:10.1016/j.qref.2009.11.003

A. Akhigbe, B.A. Stevenson / The Quarterly Review of Economics and Finance 50 (2010) 132–140

lending.2 In a summary of this line of literature by James and Smith (2000), they note that banks add the most value in lending relationships with information sensitive borrowers. Indeed, the bank loan contract enhances the bank’s monitoring ability due to the collateral, covenants and short maturity found in most bank loans. While these more traditional theories, models, and empirical work focus mostly on lending behavior by banks, as chronicled in DeYoung and Rice (2004), more and more banks are relying on noninterest income that comes from activities other than lending. In their paper, they show that, for 2001, 29.89% of operating income for U.S. commercial banks larger than $1 billion is composed on noninterest income. For commercial banks under $1 billion in 2001, 16.38% of operating income is composed of noninterest income. The question has been asked, “Does noninterest income benefit banks?” DeYoung and Rice (2004) suggest that while larger banks tend to rely more heavily on noninterest income, they also observe that better managed banks rely less on noninterest income. Their results indicate marginal increases in noninterest income are associated with higher and more variable profits and that increases in noninterest income decrease the risk return trade-off for commercial banks. Also, Rogers (1998) shows that when measuring bank efficiency, not including noninterest income lowers the efficiency with higher degrees of noninterest income relative to other banks. This seems to imply that noninterest income is positively correlated with higher bank efficiency. DeYoung and Roland (2001) show an association between increases in fee based activity and an increase in volatility for earnings and revenue as well as higher leverage. Recently, Stiroh and Rumble (2006) find that increasing the diversification of revenue streams for financial holding companies does not increase their performance as measured by profits. More specifically, they found lower risk adjusted profits with increases in noninterest income and that any diversification benefits were outweighed by the increased volatility of noninterest earnings. In addition, looking at small European credit institutions, Mercieca, Schaeck, and Wolfe (2007) find that increasing noninterest income does not improve performance. On the other hand, Baele, De Jonghe, and Vander Vennet (2007) find that higher levels of noninterest income increase the franchise value of European banks. Of interest to us is the passage of the Gramm-Leach-Bliley Act. With the passage of the Gramm-Leach-Bliley of 1999 (GLBA), the players in the financial services industry were allowed to consolidate to a degree not seen since the passing of the Glass-Steagall Act of 1933.3 Since 1933, commercial and investment bank functions have been separated creating a specialized financial services industry in the United States. This degree of specialization in the financial services industry is different from the experience of most other countries (Benston, 1994). With the passage of the GLBA, the U.S. moved closer to its pre-1933 condition, and it fell more in line with other economies such as the European Union where many large U.S. banks now find their competitors. From the perspective of this paper, one consequence of the passage of the GLBA is that it makes additional types of noninterest revenue available to commercial banks. Since commercial banks may now engage in securities underwriting/brokerage, insurance, and other areas such as venture capital, it is important to understand the relationship

133

of these types of income to the efficiency of BHCs who use them compared to their peers. As BHCs continue to engage in these previously restricted activities, given the conflicting evidence regarding increases in noninterest income cited in the above research, why have they chosen to do so? As Puri (1999) suggests in reference to underwriting, it is because banks can reuse information in underwriting that they already collect in lending. In her paper, Puri shows that banks may even better certifiers of quality than investment banks. Drucker and Puri (2005) show that combining lending and underwriting especially benefits security issuers who have a greater degree of asymmetric information. Not only that but it also encourages future relationships with the bank that ensure more business. In addition to securities underwriting, Gramm-Leach-Bliley allows BHCs to further expand into other financial arenas such as insurance, investment management, and brokerage. Given the likelihood that BHCs will continue to expand their noninterest income, this paper addresses the following question: Does the increase in additional noninterest income types occur in instances of increased or decreased efficiency? In other words, does the use of additional noninterest income types put BHCs in a better position relative to their peers? A contribution of this paper relative to previous efficiency studies on U.S. commercial banks is the use of data from the post-GLB period. With additional types of income at their disposal, we can reasonably expect increases in noninterest income from BHCs. Also, while previous work looks at noninterest income as a single item, here it is analyzed in its component parts. The first three income types are those allowed under the GLBA which include security underwriting/brokerage, venture capital revenue, and insurance (collectively these are referred to in this paper as non-traditional income). The fourth is “other” noninterest income that includes items such as check fees and fiduciary activities. Disaggregating the income this way allows us to examine the post-GLB data in a way that can show if one income type has a positive association and another type a negative association with efficiency. We may expect this because certain types of noninterest income may lend themselves more readily to the reuse of information than others. The results of our analysis dispute the notion that increases in noninterest income will go hand in hand with increases in profit efficiency. Increases in underwriting/brokerage, venture capital and insurance, especially underwriting/brokerage income, have a significant negative relationship with profit efficiency. While we do find some evidence that increases in these income types may benefit large and medium BHCs’ revenue efficiency, the overall results of our tests show that the benefits of economies of scope, on their own, are not great enough for BHCs to choose to increase their noninterest income. Our results suggest increases in these areas for any BHC should be undertaken for other, perhaps strategic, considerations besides efficiency. The rest of the paper proceeds as follows. Section 2 presents motivation and hypotheses while Section 3 reviews the methodology. Section 4 discusses the data, Section 5 presents results, and Section 6 concludes. 2. Motivation and hypothesis

2 Findings in subsequent empirical work such as Lummer and McConnell (1989), Slovin, Johnson and Glascock (1992), Best and Zhang (1993), Billett, Flannery, & Garfinkel (1995), Johnson (1997), and Hadlock and James (2002) confirm the findings on bank loan announcements and show that the effect is even more pronounced for informationally sensitive firms. 3 The Gramm-Leach-Bliley Act is also called the Financial Services Modernization Act of 1999.

We gain additional insight into noninterest incomes relationship to efficiency by examining two previously unexplored pieces of evidence. First, we look the previously unexamined post-GLB period. Second, because we examine this period we can also examine noninterest income by type which includes new noninterest revenue streams available after GLB.

134

A. Akhigbe, B.A. Stevenson / The Quarterly Review of Economics and Finance 50 (2010) 132–140

Using recent data for U.S. institutions, we attempt to provide answers and reconcile the following views. In one view, diversification across different revenue streams should allow BHCs to reuse information as suggested by Puri (1999) and thus allow them to be more efficient. In a second view, if, as in Stiroh and Rumble (2006), increasing noninterest income represents too much of a cost due to the volatility of the revenue, then increasing diversification may lead BHCs to be less profit efficient. To help examine these issues we propose the following hypothesis: Hypothesis. The profit efficiency of BHCs will have a positive and significant relationship with the amount of noninterest and GLB types of revenue relative to total revenue due to the BHCs reuse of information and current expertise. It is a common assumption that one benefit of having many financial services under “one roof” will allow banks and other financial firms to reuse information. For instance, if Bank A makes a loan to Firm Z, it can later reuse that information when underwriting Firm Z’s bonds. Alternatively, if I have a retail customer with the bank and I have gathered information from them for loans and through their transaction accounts, my cost to sell them an insurance product may be less. This example may apply more to smaller more retail oriented BHCs, while the lending/underwriting example may apply to larger BHCs. On the other hand, if a small BHC does not do much commercial lending they may not find benefits from reusing information if they begin to underwrite securities. Affirmative results for the above hypothesis would contradict those such as Rumble and Stiroh (2006) in that, at least in terms of efficiency, diversification of income into noninterest income is a beneficial strategy for BHCs. 3. Methodology, variable selection, and expected variable relationships 3.1. Stochastic frontier analysis BHC efficiency can be measured using stochastic frontier analysis (SFA) where financial performance of individual BHCs is compared to the best achievable performance based on other institutions in the sample. The key item needed in the model for the sample used in this paper is the ability to compare bank holding companies (BHCs) by their level of noninterest income. We collect data for BHCs from the Federal Reserve Bank of Chicago’s Bank Holding Company database for 2003–2006 for a total of 4 years of data. The data is the condition and income data that each BHC must report on form FRY-9C which is similar to what is commonly called a “call report” for individual banks. In this study we examine alternative profit efficiency. Alternative profit efficiency is judged on how close the BHC comes to maximum profit levels based on output levels instead of output prices.4 Alternative profit efficiency is the best method to answer our proposed hypotheses because it can show if the addition of underwriting/brokerage, venture capital, and insurance services to the commercial BHCs repertoire improves a BHC’s ability to efficiently generate profit. To check for the robustness of our results, we also examine the relationship of these types of income to the

4 As Berger and Mester (1997) note, using levels as opposed to prices can be motivated by several reasons. For our sample, we assume that there can be substantial variation in the quality of services and the ability of BHCs to set prices due to competition or lack thereof. These assumptions make the alternative profit efficiency measure preferable to standard profit efficiency.

Efficiency Ratio (ER) which measures how much of bank’s noninterest income and net interest income is consumed by noninterest expenses. Higher levels of ER denote lower levels of efficiency.5 For stochastic frontier analysis in this study, one frontier is created per year for all BHCs in the sample and each BHC is included even if it has only 1 year of available data.6 The creation of a frontier for each year allows for the regression coefficients to vary over time and provide a flexible estimation procedure. For each bank we calculate profit as PREROA where PREROA is earnings before extraordinary items, taxes and loan losses divided by total assets. Calculating the efficiency measure allows us to analyze each BHC’s performance relative to their peers. This will enable us to use this efficiency measure as a dependent variable and use regression analysis to explain the performance of BHCs.7 Efficiency scores can range from 0, least efficient, to 1, most efficient. For example, if a particular BHC in the sample has an efficiency score of .86, this means that the BHC is 86% as efficient as the best-practice BHC in the sample year. In our analysis, we also disaggregate the sample based on the size of the BHC. As Akhigbe and McNulty (2003) and DeYoung and Hasan (1998) point out, small BHCs may differ from large BHCs in their production technologies and strategies. To take this into account, three separate size classes are examined—one for BHCs with greater than $10 billion in total assets, one for between $10 billion and $1 billion in total assets, and one for BHCs less than $1 billion in total assets. In practice, two common specifications are used for the functional form. The first is the translog and the other is the Fourierflexible form. The Fourier-flexible form adds onto the translog form by including Fourier trigonometric terms to the model. Some studies (McAllister & McManus, 1993; Mitchell & Onvural, 1996) show that the Fourier-flexible form is a better approximation due to its flexibility and its quality of being a global approximation. As discussed in Berger and Mester (1997), other studies find little difference in results when using one form or another. For this reason and due to the Fourier-flexible form requiring so many variables that too many degrees of freedom are lost, the functional translog form is used and the alternative profit function is as follows8 :

ln(PREROA) = ˛ +

4  i=1

+

5 

k ln

k=1

 4

+

r=1

1  ˇij ln(wi ) ln(wj ) 2 4

ˇi ln(wi ) +

ır ln

4

i=1 j=1

y  k

z5

z  r

z5

1  km ln 2 5

+

5

k=1 m=1

1  ırs ln 2 4

+

4

r=1 s=1

y  k

z5

z  r

z5

ln

ln

y  m

z5

z  s

z5

5 The Efficiency Ratio (ER) = noninterest expense/(net interest income + noninterest revenue). 6 As suggest in Bauer, Berger, Ferrier, and Humphrey (1998), we used stochastic frontier analysis where all years were combined to create uniform coefficients for all years and we used a distribution free analysis method as well to compare with the results of the analysis here for consistency in the rankings of BHCs. Since the results of all methods were very comparable, we only present the SFA results by year in the paper. 7 DeYoung and Hasan (1998) provide a justification for this procedure. See page 580 of their article for an explanation. 8 The scaling and use of certain terms in addition to costs and revenues of the bank follow recommendations of Berger and Mester (1997). Also, we assume the normal symmetry and linear homogeneity restrictions. Since this is our a priori assumption, we normalize all costs and inputs by total assets.

A. Akhigbe, B.A. Stevenson / The Quarterly Review of Economics and Finance 50 (2010) 132–140

+

5 4  

ik ln(wi ) ln

i=1 k=1

× ln(wi ) ln

4 4  y   k

z5

4 5 z    r

z5

+

k=1 r=1

+

ir

i=1 r=1

kr ln

y  k

z5

ln

z  r

z5 (1)

where PREROA = (earnings before extraordinary items, taxes and loan losses)/total assets; w1 = price of non-deposit related borrowing; w2 = price of deposits; w3 = price of labor; w4 = price of property plant and equipment; y1 = real estate loans; y2 = commercial and industrial loans; y3 = personal loans; y4 = securities underwriting, investment and brokerage, venture capital, and insurance commissions and fees; y5 = other noninterest fee income; z1 = off-balance sheet items (letters of credit, loan commitments, etc.); z2 = shareholder’s equity; z3 = liquid assets; z4 = average of non-accrual and past due assets for the BHC’s home state; z5 = total assets. The specification of variables in the model is crucial to the specific questions addressed by the hypotheses in Section 2. Unlike previous studies, multiple categories of noninterest income are used: (1) securities underwriting, investment and brokerage, venture capital, and insurance commissions and fees and (2) other noninterest fee income. By creating separate categories, the efficiency measure will be more robust. Thus, the results presented here give a more complete measure of efficiency than we would obtain otherwise for only one category of noninterest income. Other inputs and outputs that affect BHC efficiency are included. First, off-balance sheet items are included as a substitute product for loans and to capture all outputs by the BHC. Several studies (Clark & Siems, 2002; Lieu, Yeh, & Chiu, 2005; Stiroh, 2000, etc.) indicate that excluding off-balance sheet activities creates a downward bias in efficiency. Second, the price of labor, or human resource expense, proxies for the inputs the BHC uses to generate its product. Third, liquid assets are included to account for a major portion of a BHC’s assets and to account for the trade-off made by BHCs between liquidity and return. Fourth, since different types of lending carry different risks and returns, loans are broken out by type in the calculation of efficiency. Next, the cost of deposits and non-deposit borrowing are inputs for traditional banking activities. In addition, the average non-accrual and past due assets for the BHC’s home state are included as a measure of the current operating environment. Incorporating the state average of non-performing assets accounts for negative (and positive) external shocks to the BHC’s operating environment not under control of the BHC’s management.9 Lastly, shareholder’s equity is included in the model to measure of a BHCs insolvency risk, determine the size of equity funding lending and other activities in addition to

9 See Berger and Mester (1997) for a further explanation on the reasoning for this variable. We are indebted to an anonymous referee for pointing out the need to include this variable in our analysis.

135

deposits, and adjust for a BHC’s risk aversion (Hughes, Lang, Mester, & Moon, 1996a, 1996b, 1997). In the frontier analysis, the BHCs error for 2003 will be compared to the highest error term for 2003, etc.10 As in DeYoung and Hasan (1998) and Akhigbe and McNulty (2003), the measure of profit efficiency, PROFEFF, is calculated as: PROFEFF= (actual PREROA/potential PREROA) if PREROA > 0, PROFEFF= 0 if PREROA < 0. Potential PREROA is the actual PREROA plus inefficiency. The most efficient banks will have PROFEFF near one while the least efficient will have PROFEFF closer to zero. Inefficiency is determined performing stochastic frontier analysis (SFA) in LIMDEP 9.0. In SFA, the error term from Eq. (1) above is assumed to be composed of two parts, ui and vi . In this method, vi is assumed to be statistical noise that follows a normal distribution with a mean of zero. However, ui , which is measured inefficiency, is assumed to be half-normally distributed. The inefficiency is computed by the Jondrow, Lovell, Materov, and Schmidt (1982) formula where,11 Eˆ = [u|ε] =

 1 + 2

 (z)

1 − ˚(z)



−z ,

ε = v ± u, z =

ε 

While distributional assumptions besides half-normal are sometimes made for ui , Bauer et al. (1998) note that regardless of the assumption made, while individual firm efficiencies may vary, the benefit of using SFA is that the rankings are consistent no matter what distributional assumptions are used. 3.2. Regression analysis and expected outcomes The dependent variable used in the regression analysis is the profit efficiency measure calculated in the frontier analysis described above.12 PE = f (% GLB income, % underwriting, % venture capital, % insurance, % other noninterest income, Section 20, asset size, relative portfolio risk, real estate loans/ total loans, demand deposits, 2004, 2005, 2006) where PE = our estimate of alternate profit efficiency; % GLB income = underwriting/brokerage, venture capital, and insurance income as a % of total income; % Underwriting/brokerage = underwriting/brokerage income as a % of total income; % Venture capital = venture capital income as a % of total income; % Insurance = insurance income as a % of total income; % Other noninterest income = non-GLB income as a % of total income; Section 20 = 1 if the BHC had previously had a Section 20 subsidiary and 0 otherwise;

10 Because this ranks BHCs within a year relative to the frontier for the year, even when their inputs and outputs are static, a BHC’s efficiency measure may shift from year to year. To account for this variation year dummies are used in subsequent regressions to minimize this problem. 11 For a more detailed presentation of the estimation technique presented here see either Jondrow et al. (1982) or Stochastic Frontier Analysis by Kumbhakar and Lovell (2003, pp. 74–80). 12 DeYoung and Hasan (1998) provide a justification for this procedure. See page 580 of their article for an explanation. The basic justification is that the error term is orthogonal the variables in the frontier analysis and then transformed. Thus there should not be an issue using the profit efficiency term in subsequent regressions with similar variables.

136

A. Akhigbe, B.A. Stevenson / The Quarterly Review of Economics and Finance 50 (2010) 132–140

Asset size = log of total assets; Relative portfolio risk = non-accrual assets and assets 90 days past due divided by total assets for the BHC less the state average for the same measure; Demand deposits = demand deposits divided by total deposits; 2004, 2005, 2006 = 1 if the year of the observation is in that year and 0 otherwise. In relation to the hypotheses, the noninterest/GLBA income variables are expected to have a positive and significant relationship to profit efficiency. This would support the assertion that efficient BHCs reuse information and benefit from diversification. A dummy variable is included to indicate if the BHC had Section 20 subsidiary prior to the passage of GLBA. Section 20 subsidiaries allowed commercial banks to engage in underwriting activities limited to certain classes of securities and with caps on revenue generated by this activity. The expectation is that if a bank previously had a Section 20 subsidiary that they should have “learned the ropes” of the business before GLBA and therefore should be more efficient relative to other commercial banks without prior experience. Next, total assets are included in the regression to control for the effect of size on efficiency due to economies of scale, and nonperforming assets to total assets for the BHC less the average of the same for the BHC’s state (labeled relative portfolio risk) is included as a measure of the quality of firm assets.13 If there is a size effect in relation to bank efficiency, total assets should have a positive and significant relationship. Relative portfolio risk should have a negative and significant relationship, since this variable indicates poor performance on the bank’s portfolio of assets relative to their operating environment. In addition, demand deposits are conjectured to aid in profit efficiency since they are a lower cost and more reliable source of funds than non-deposit funding as noted in Koch and MacDonald (2003). Lastly, a measure of the proportion of real estate loans to total loans is included to measure the banks type of lending activity. As mentioned above, robustness checks are also performed by using the ER, or Efficiency Ratio, to measure the efficiency of each BHC. 4. Data The data for this study is annual bank holding company data from 2003 until 2006 available online from the Federal Reserve Bank of Chicago and contained in the required FR Y-9C report. Since the Gramm-Leach-Bliley Act was not passed until 1999, data on each category of underwriting/brokerage, venture capital, and insurance was not included on the FR Y-C9 until 2003.14 Table 1 has summary statistics on the data for banks used in the profit efficiency analysis. In Table 1, the first column contains the whole set while the second and third column contains banks with above 10% of their revenue in GLB income and below 1% in GLB income, respectively. High GLB banks tend to be significantly larger (measured in assets, deposits, and loans), have significantly higher human resource costs (HR price), significantly higher capital costs (PPE

13 Empirical evidence, for example in Giradone, Molyneux, and Gardener (2004), exists that indicates that size does not have a relationship with efficiency. However, other research such as Berger, Hancock, & Humphrey (1993) indicates that banks of different size do have measurable differences in efficiency. Since evidence points in both directions, size is included in this paper. 14 Data breaking down underwriting/brokerage and venture capital was available in 2001.

Table 1 Descriptive statistics for the profit efficiency sample. Variable

(1) Whole Sample

(2) High GLB

(3) Low GLB ***

Total assets

$7,286,491.08 [$65,411,530.51]

$103,845,680.41 [$303,298,489.71]

$1,043,682.68*** [$6,477,472.11]

Profit efficiency

0.621 [0.147]

0.615 [0.161]

0.620 [0.148]

Portfolio risk

.0051 [.0065]

0.0041** [0.0053]

0.0051** [0.0072]

State risk

.0055 [.0025]

0.0059 [0.0030]

0.0055 [0.0024]

GLB income

$53,437.60 [$808,326.38]

$1,762,773.54*** [$5,035,864.76]

$258.42*** [$3,113.68]

Other Noninterest Inc.

$111,494.27

$1,266,226.86***

$15,653.29***

[$984,705.66]

[$3,825,175.28]

[$192,404.54]

HR price

$55,610.84 [$21,633.42]

$81,882.89*** [$43,989.07]

$54,704.78*** [$19,370.27]

PPE price

30.40% [31.31%]

43.95%*** [38.56%]

31.37%*** [34.62%]

Price of deposits

1.75% [0.66%]

1.55%*** [0.76%]

1.75%*** [0.65%]

Amount of deposits

$3,810,345.58

$41,400,786.72***

$752,861.73***

[$31,889,537.72]

[$127,319,291.92]

[$4,162,407.33]

Total loans

$3,576,468.76 [$29,128,192.35]

$36,834,151.64*** [$111,919,585.24]

$655,112.04*** [$4,090,224.32]

ROE

12.16% [7.25%]

10.90%*** [5.63%]

12.09%*** [7.79%]

ROA

1.08% [0.77%]

1.32%* [1.71%]

1.06%* [0.78%]

PREROA

1.47% [0.91%]

1.82%** [2.36%]

1.45%** [0.89%]

Efficiency Ratio (ER)

64.96%

72.07%***

64.41%***

Number in sample

[12.90%]

[10.41%]

[13.47%]

7744

167

5312

Numbers in brackets are the standard deviation for the variable. Profit efficiency is the efficiency of each bank measured against a best-practice bank in the sample. This measure is used as the dependent variable in later regressions. Portfolio risk is equal to non-accrual assets and assets 90 days past due divided by total assets. State risk is equal to non-accrual assets and assets 90 days past due divided by total assets for all BHCs within the banks state. GLB income consists of revenue from underwriting/brokerage, venture capital, and insurance. HR price is the total of salary and benefit expense divided by the number of full-time equivalent employees. PPE price is the expense on fixed assets divided by the dollar amount of fixed assets. Price of deposits is the interest expense on deposits divided by the dollar amount of deposits. PREROA is ROA calculated as earnings before extraordinary items, taxes and loan losses divided by total assets. The Efficiency Ratio is (noninterest expense − amortization of intangible assets)/(net interest income + noninterest income) where the ratio indicates the amount of revenue and income consumed by overhead. High GLB indicates GLB income of more than 10% of total income and low GLB indicates GLB income below 1% of total income. All dollar amounts are in thousands except HR price which is the actual value. Dollar amounts are in thousands except for HR price which is in actual dollars. * p-Value of the difference in the means between GLB high and low of <.10. ** p-Value of the difference in the means between GLB high and low of <.05. *** p-Value of the difference in the means between GLB high and low of <.01.

price), significantly lower ROE, and significantly higher Efficiency Ratios (meaning they are less efficient). Also, the profit efficiency variable is not significantly different between the two groups. Lastly, the large GLB group does have a significantly higher ROA and PREROA which is somewhat counter to the other measures.

A. Akhigbe, B.A. Stevenson / The Quarterly Review of Economics and Finance 50 (2010) 132–140 Table 2 Regressions with the alternative profit efficiency measure as the dependent variable. Variable

GLB aggregated ***

Intercept

0.41898 [14.53]

% GLB income/total income

−0.00658*** [−2.44]

GLB disaggregated 0.44916*** [4.88]

% Underwriting/brokerage

−0.01259*** [−3.39]

% Venture capital

0.01336 [0.74]

% Insurance

0.00023 [0.07]

% Other Noninterest Inc./TI

−.02135*** [−6.48]

−.02105*** [−6.39]

Section 20

0.02918* [1.78]

0.03087* [1.87]

Asset size

0.01087*** [7.24]

0.01127*** [7.32]

Relative portfolio risk

0.00290 [0.01]

−0.02474 [−0.09]

Real estate loans/total loans

−0.03600*** [−3.24]

−0.03597*** [−3.25]

Demand deposits

0.04280 [1.57]

0.04518* [1.65]

2004

−0.00111 [−0.26]

−0.00120 [−0.28]

2005

0.01337*** [3.06]

0.01328*** [3.04]

2006

0.04532*** [7.75]

0.04524*** [7.73]

N R2 F-Stat F-Stat probability

7744 0.0286 22.77 <.0001

7744 0.0294 19.49 <.0001

The dependent variable in these regressions is the alternative profit efficiency measure calculated using stochastic frontier analysis as described in Section 3 of the paper and is the efficiency of each bank measured against a best-practice frontier in the sample. Each variable is presented with its corresponding coefficient and tvalue with the t-value in brackets listed below the coefficient. GLB income consists of revenue from underwriting/brokerage, venture capital, and insurance. Other noninterest income is noninterest fee income besides GLB income as defined above and does not include any trading gains and losses, etc. Section 20 is a dummy variable that equals one if the bank holding company had a Section 20 subsidiary before the passage of the GLBA. Relative portfolio risk is equal to non-accrual assets and assets 90 days past due divided by total assets for the BHC less the average for the BHC’s state. Real estate loans/total loans proxies for the lending and traditional focus of the bank holding company. Due to skewness of the values in asset size, percentage of income, and demand deposits, the variables used in the regressions are taken as logs. * p-Value <.10. ** p-Value <.05. *** p-Value <.01.

137

The first column indicates that higher levels of GLBA income are associated with decreased profit efficiency (coefficient of −.00658 and a t-statistic of −2.44). This does not lend weight to the argument that scope economies are exploited by BHCs to enhance efficiency. The analysis also indicates that increasing asset size is associated with higher profit efficiency, and thus there are economies of scale. Three other variables are significant in column one. First, other noninterest income has a coefficient of −.02135 indicating an increase in other noninterest income is associated with a decrease in profit efficiency. Second, Section 20 is positive (0.02918) indicating that increased experience with the GLBA types of income seems enhance efficiency. Next, the real estate loans/total loans coefficient is negative (−0.04232). Based on our hypothesis, this indicates that firms focusing on traditional mortgage lending are less efficient than those with broader portfolios. The year dummy variables for 2005 and 2006 are significant working from a base year of 2003 with positive coefficients. The second column provides more insight into the negative association between GLBA income and profit efficiency. The −3.39 t-statistic on the coefficient of −.01259 for underwriting indicates that increased underwriting activity is associated with BHCs that are less profit efficient. With venture capital and insurance being insignificant, our conclusion is underwriting/brokerage is the driving force behind the negative coefficient of the GLB income/total income variable in the first column. All but one of the other variables remains the same in terms of the sign and size of coefficients and in terms of significance. Demand deposits becomes significant and is in the expected direction with a positive sign (coefficient = .04518) and indicates that cheaper sources of funds are associated with higher profit efficiency. 5.2. Multivariate analysis with the Efficiency Ratio The Efficiency Ratio (ER) measures how much income the BHC is devoting to expenses. Higher ratios imply that for every dollar of income, more of it is being consumed by noninterest “overhead” expenses. Unlike Table 2, in Table 3, positive coefficients indicate lower efficiency and negative coefficients indicate higher efficiency associated with an increase in that variable. The results with the ER reinforce our previous results. In the first column, the overall GLB income/total income variable is positive (indicating lower efficiency) and significant as is the underwriting/brokerage variable in the second column. There are two noteworthy differences in the outcomes of this set of regressions from the regressions in Table 2. First, the insurance variable is positive (coefficient = 0.1504) and significant (t-stat = 5.66) indicating that BHCs that generate increased levels of revenue from insurance tend to be less efficient. Second, instead of demand deposits having a positive relationship with efficiency, they are associated with decreased levels of efficiency. 5.3. Multivariate profit efficiency and Efficiency Ratio analysis by BHC size

5. Results 5.1. Multivariate profit efficiency analysis Unlike the pair-wise comparisons in the descriptive statistics, multivariate analysis allows us to use of the entire data set and control for other factors that may influence our outcome. Table 2 contains two columns. The first column sums the GLBA income types while column two leaves them separated into: underwriting/brokerage, venture capital, and insurance. In each case the variable is a proportion of total revenue.

Table 4 is broken down by size into BHCs with assets over $10 billion (hereafter, large), assets between $10 billion and $1 billion (hereafter, medium), and assets below $1 billion (hereafter, small). Only small BHCs continue to have a strong negative association between underwriting/brokerage and profit efficiency. For insurance, small BHCs have a positive (coefficient = 0.00665) but less significant association relative to underwriting/brokerage. While the regressions using alternative profit efficiency show a loss in significance on the income type coefficients for large and medium BHCs, the results for the three income types presented

138

A. Akhigbe, B.A. Stevenson / The Quarterly Review of Economics and Finance 50 (2010) 132–140

Table 3 Regressions with the Efficiency Ratio as the dependent variable. Variable Intercept

% GLB income/total income

GLB aggregated ***

1.13874 [46.88]

GLB disaggregated 1.24383*** [16.04]

Table 4 Regressions with the alternative profit efficiency measure as the dependent variable where the sample is disaggregated by size. Variable

>$10 B

<$10 B, >$1 B

<$1 B

0.94750*** [21.38]

0.81458*** [10.90]

0.44248*** [2.60]

% Underwriting/brokerage

−0.00038 [−0.20]

−0.00335 [−1.16]

−0.01268*** [−2.97]

% Venture capital

−0.00142 [−0.29]

−0.00644 [−0.48]

−0.00055 [−0.02]

% Insurance

0.00048 [0.21]

0.00106 [0.43]

0.00665* [1.94]

% Other Noninterest Inc./TI

−0.00139 [−0.55]

−0.00104 [−0.37]

−0.01875*** [−5.40]

Section 20

−0.00176 [−0.44]

0.03512 [1.36]

−0.06595** [−2.03]

Asset size

−0.00017 [−0.12]

0.00295 [1.23]

0.01152*** [3.19]

Relative portfolio risk

0.52025 [1.34]

−0.09150 [−0.30]

0.13512 [0.54]

Real estate loans/total loans

−0.00890 [−1.13]

−0.03579*** [−3.87]

−0.03700*** [−3.05]

Demand deposits

0.00180 [0.06]

−0.06207** [−2.18]

0.06588** [2.45]

2004

0.00246 [0.62]

−0.02806*** [−6.81]

−0.01025** [−2.38]

2005

−0.02123*** [−5.34]

−0.01480*** [−3.62]

0.00190 [0.44]

2006

0.01683*** [−3.85]

−0.01173*** [−2.82]

0.19473*** [28.36]

N R2 F-Stat F-Stat probability

372 0.1415 4.93 <.0001

1,341 0.0547 6.41 <.0001

6,031 0.1723 104.41 <.0001

Intercept

0.02959*** [13.04]

% Underwriting/brokerage

0.03681*** [11.79]

% Venture capital

−0.00228 [−0.15]

% Insurance

0.01504*** [5.66]

% Other Noninterest Inc./TI

0.06230*** [22.44]

0.06181*** [22.27]

Section 20

−0.01223 [−0.88]

−0.01669 [−1.20]

Asset size

−0.02152*** [−17.01]

−0.02249*** [−17.36]

Relative portfolio risk

1.48230*** [6.67]

1.52310*** [6.85]

Real estate loans/total loans

0.03648*** [3.90]

0.03547*** [3.80]

Demand deposits

0.06129*** [2.67]

0.05759** [2.51]

2004

0.00699* [1.91]

0.00698* [1.91]

2005

0.00908** [2.47]

0.00915** [2.49]

2006

0.03180*** [6.45]

0.03202*** [6.50]

N R2 F-Stat F-Stat probability

7744 0.1092 94.81 <.0001

7744 0.1113 80.72 <.0001

The dependent variable in these regressions is the Efficiency Ratio described in Section 3 of the paper and is the ratio of noninterest expense/(net interest income + noninterest income). Higher ratios indicate lower efficiency. Each variable is presented with its corresponding coefficient and t-value with the t-value in brackets listed below the coefficient. GLB income consists of revenue from underwriting/brokerage, venture capital, and insurance. Other noninterest income is noninterest fee income besides GLB income as defined above and does not include any trading gains and losses, etc. Section 20 is a dummy variable that equals one if the bank holding company had a Section 20 subsidiary before the passage of the GLBA. Relative portfolio risk is equal to non-accrual assets and assets 90 days past due divided by total assets for the BHC less the average for the BHC’s state. Real estate loans/total loans proxies for the lending and traditional focus of the bank holding company. Due to skewness of the values in asset size, percentage of income, and demand deposits, the variables used in the regressions are taken as logs. * p-Value <.10. ** p-Value <.05. *** p-Value <.01.

in Table 5 continue to be significant for all BHC sizes when using the Efficiency Ratio as the dependent variable. As in the overall sample, an increase in underwriting/brokerage and insurance are associated with a decrease in efficiency. The main difference is in the disaggregated sample, compared to Table 3, is venture capital is associated with increased efficiency for large BHCs but decreased efficiency for medium BHCs. Does the lack of significance for large and medium BHCs in the profit efficiency regressions and their significance in the Efficiency Ratio regressions tell us something? Since the Efficiency Ratio measures the use of overhead by the BHC and profit efficiency measures

Asset size

The dependent variable in these regressions is the profit efficiency measure calculated using frontier analysis as described in Section 3 of the paper and is the efficiency of each bank measured against a best-practice frontier in the sample. Each variable is presented with its corresponding coefficient and t-value with the t-value in brackets listed below the coefficient. GLB income consists of revenue from underwriting/brokerage, venture capital, and insurance. Other noninterest income is noninterest fee income besides GLB income as defined above and does not include any trading gains and losses, etc. Section 20 is a dummy variable that equals one if the bank holding company had a Section 20 subsidiary before the passage of the GLBA. Relative portfolio risk is equal to non-accrual assets and assets 90 days past due divided by total assets for the BHC less the average for the BHC’s state. Real estate loans/total loans proxies for the lending and traditional focus of the bank holding company. Due to skewness of the values in asset size, percentage of income, and demand deposits, the variables used in the regressions are taken as logs. * p-Value <.10. ** p-Value <.05. *** p-Value <.01.

efficient use of revenues and costs, it may tell us that while medium and large BHCs suffer operationally, they compensate for this by being more efficient on the revenue side. 5.4. Summary of results and comments The regression analysis for alternative profit efficiency and ER for the sample as a whole contradicts the hypothesis that increases in GLB income would positively correlate with increased efficiency due to economies of scope. However, the analysis by BHC size suggests that while increases in GLB income for small BHCs is negatively related to efficiency, increases for large and medium BHCs

A. Akhigbe, B.A. Stevenson / The Quarterly Review of Economics and Finance 50 (2010) 132–140 Table 5 Regressions with the Efficiency Ratio as the dependent variable where the sample is disaggregated by size. Variable

Intercept

Asset size >$10 B

<$10 B, >$1 B

<$1 B

0.86902*** [4.38]

1.66547*** [9.05]

1.27332*** [8.16]

***

***

***

% Underwriting/Brokerage

0.05882 [6.82]

0.03956 [5.55]

0.02582 [6.59]

% Venture Capital

−0.05764** [−2.59]

0.07193** [2.16]

0.01082 [0.34]

% Insurance

0.04339*** [4.44]

0.01364** [2.21]

0.01503*** [4.78]

% Other Noninterest Inc./TI

0.00483 [0.42]

0.04403*** [6.45]

0.07314*** [22.94]

Section 20

−0.03932** [−2.18]

−0.01835 [−0.29]

−0.00154 [−0.05]

Asset size

−0.00551 [−0.87]

−0.02720*** [−4.62]

−0.02357*** [−7.09]

Relative portfolio risk

−2.53009 [−1.45]

1.81356** [2.39]

1.58337*** [6.84]

Real estate loans/total loans

−0.10263*** [−2.92]

−0.01379 [−0.61]

−0.03700*** [−3.05]

Demand deposits

−0.01062 [−0.08]

−0.01930 [−0.28]

0.07055*** [2.86]

2004

0.02972* [1.66]

0.00446 [0.44]

0.00612 [1.55]

2005

0.04102** [2.30]

0.00365 [0.36]

0.00873** [2.18]

2006

0.05889*** [3.01]

0.01268 [1.24]

0.04189*** [6.64]

N R2 F-Stat F-Stat probability

372 0.2090 7.90 <.0001

1,341 0.0858 10.39 <.0001

6,031 0.1177 66.90 <.0001

The dependent variable in these regressions is the Efficiency Ratio described in Section 3 of the paper and is the ratio of noninterest expense/(net interest income + noninterest income). Higher ratios indicate lower efficiency. Each variable is presented with its corresponding coefficient and t-value with the t-value in brackets listed below the coefficient. GLB income consists of revenue from underwriting/brokerage, venture capital, and insurance. Other noninterest income is noninterest fee income besides GLB income as defined above and does not include any trading gains and losses, etc. Section 20 is a dummy variable that equals one if the bank holding company had a Section 20 subsidiary before the passage of the GLBA. Relative portfolio risk is equal to non-accrual assets and assets 90 days past due divided by total assets for the BHC less the average for the BHC’s state. Real estate loans/total loans proxies for the lending and traditional focus of the bank holding company. Due to skewness of the values in asset size, percentage of income, and demand deposits, the variables used in the regressions are taken as logs. * p-Value <.10. ** p-Value <.05. *** p-Value <.01.

are associated with decreases in operational efficiency as measured by the ER, but this drag on efficiency may be offset by increased efficiency on the revenue side. Overall, can we say increased noninterest income in the postGLB era helps increase efficiency for BHCs? The answer appears to be a cautious no. Based on the current use of these sources of income by BHCs it appears that less efficient BHCs take in more of this type of income. Does that mean banks should not pursue noninterest income including the GLB income types? No. Using these business lines may affect customer relationships and the ability to attract business, especially for large international banks. Also,

139

the results in the disaggregated sample by size suggest that there may be some revenue benefits to this type of income for large and medium BHCs. Finally, as banks learn these businesses over time and develop larger scale operations, we may see very different results. 6. Conclusion The main goal of this research was to expand previous research on noninterest income by looking at a new period, the post-GLB period, and to look in greater detail at different types of noninterest income. Specifically, the research presented here looks at three types of noninterest income (underwriting/brokerage, venture capital, and insurance) and general noninterest income from 2003 to 2006 that should allow banks to benefit through economies of scope due to the reuse of information. Previous research such as DeYoung and Rice (2004), Stiroh and Rumble (2006) and Mercieca et al. (2007) suggests that better performing banks use less noninterest income. However, other research such as Baele et al. (2007) and Vander Vennet (2002) has shown that increasing levels of noninterest income have made European banks more efficient and increased their value. In addition, other research such as Drucker and Puri (2005) and Puri (1999) shows that engaging in other intermediary activities like underwriting allows banks to reuse information for their benefit and their clients benefit. This begs the question, “If the reuse of information is beneficial to banks, does the increase in noninterest income generated by these BHCs allow them to experience gains in efficiency relative to their peers?” This study helps to answer this question by being the first to examine the efficiency of BHCs during the 2003–2006 period, and how their efficiency is related to varying levels of noninterest income. Our hypothesis is increasing levels of underwriting/brokerage, venture capital, and insurance income will be used more by BHCs that show greater profit efficiency due to economies of scope in information. By and large, the empirical results do not affirm this hypothesis. Increases in noninterest income, especially underwriting/brokerage income, have a significant negative relationship with profit efficiency. When examining BHCs by size, however, there is some minor evidence to suggest large and medium BHCs offset some decreased cost efficiency with increased efficiency on the revenue side when increasing their use of GLB income. Given our findings, any bank that increases their proportion of noninterest and non-traditional income should do so carefully and possibly base their decision on other considerations besides efficiency. Acknowledgements The authors would like to thank seminar participants at participants at The University of Akron and at the 2007 Midwest Finance Association conference as well as an anonymous referee for helpful comments on earlier versions of this paper. References Akhigbe, A., & McNulty, J. E. (2003). The profit efficiency of small U.S. commercial banks. Journal of Banking and Finance, 27, 307–325. Baele, L., De Jonghe, O., & Vander Vennet, R. (2007). Does the stock market value bank diversification? Journal of Banking and Finance, 31, 1999–2023. Bauer, P., Berger, A., Ferrier, G., & Humphrey, D. (1998). Consistency conditions for regulatory analysis of financial institutions: A comparison of frontier efficiency methods. Journal of Economics and Business, 50, 85–114. Benston, G. (1994). Universal banking. Journal of Economic Perspectives, 8, 121–143. Berger, A., Hancock, D., & Humphrey, D. (1993). Bank efficiency derived from the profit function. Journal of Banking and Finance, 17, 317–347. Berger, A., & Mester, L. (1997). Inside the black box: What explains differences in the efficiencies of financial institutions? Journal of Banking and Finance, 21, 895–947.

140

A. Akhigbe, B.A. Stevenson / The Quarterly Review of Economics and Finance 50 (2010) 132–140

Best, R., & Zhang, H. (1993). Alternative information sources and the information content of banks loans. The Journal of Finance, 48, 1507–1522. Billett, M., Flannery, M., & Garfinkel, J. (1995). The effect of lender identity on a borrowing firm’s equity return. The Journal of Finance, 50, 699–718. Clark, J., & Siems, T. (2002). X-efficiency in banking: Looking beyond the balance sheet. Journal of Money, Credit, and Banking, 34, 987–1013. DeYoung, R., & Hasan, I. (1998). The performance of de novo commercial banks: A profit efficiency approach. Journal of Banking and Finance, 22, 565–587. DeYoung, R., & Rice, T. (2004). Noninterest income and financial performance at U.S. commercial banks. The Financial Review, 39, 101–127. DeYoung, R., & Roland, K. (2001). Product mix and earnings volatility at commercial banks: Evidence from a degree of total leverage model. Journal of Financial Intermediation, 10, 54–84. Diamond, D. (1984). Financial intermediation and delegated monitoring. Review of Economic Studies, 51, 393–414. Drucker, S., & Puri, M. (2005). On the benefits of concurrent lending and underwriting. The Journal of Finance, 60, 2763–2799. Fama, E. (1985). What’s different about banks? Journal of Monetary Economics, 15, 29–39. Giradone, C., Molyneux, P., & Gardener, E. (2004). Analysing the determinants of bank efficiency: The case of Italian banks. Applied Economics, 36, 215–227. Hadlock, C., & James, C. (2002). Do banks provide financial slack? The Journal of Finance, 57, 1383–1419. Hughes, J., Lang, W., Mester, L., & Moon, C. (1996a). Efficient banking under interstate banking. Journal of Money, Credit, and Banking, 28, 1045–1071. Hughes, J., Lang, W., Mester, L., & Moon, C. (1996b). Safety in numbers? Geographic diversification and insolvency risk. Proceedings, Federal Reserve Bank of Chicago, 202–218. Hughes, J., Lang, W., Mester, L., & Moon, C. (1997). Recovering risky technologies using the almost ideal demand system: An application to U.S. banking. Journal of Financial Services Research, 18, 5–27. James, C. (1987). Some evidence on the uniqueness of bank loans. Journal of Financial Economics, 19, 217–235. James, C., & Smith, D. (2000). Are banks still special? New evidence on their role in the corporate capital-raising process. Journal of Applied Corporate Finance, 13, 52–63. Johnson, S. (1997). The effect of bank reputation on the value of bank loan agreements. Journal of Accounting, Auditing, and Finance, 12, 83–100.

Jondrow, J., Lovell, C., Materov, I., & Schmidt, P. (1982). On the estimation of technical efficiency in the stochastic frontier production function model. Journal of Econometrics, 19, 233–238. Koch, T. W., & MacDonald, S. (2003). Bank management (5th edition). Mason, OH: South-Western. Kumbhakar, S., & Lovell, C. (2003). Stochastic frontier analysis. Cambridge, UK: The Press Syndicate of the University of Cambridge. Leland, H., & Pyle, D. (1977). Informational asymmetries, financial structure, and financial intermediation. The Journal of Finance, 32, 371–387. Lieu, P., Yeh, T., & Chiu, Y. (2005). Off-balance sheet activities and cost inefficiency in Taiwan’s banks. The Services Industry Journal, 25, 925–944. Lummer, S., & McConnell, J. (1989). Further evidence on the bank lending process and the capital-market response to bank loan agreements. Journal of Financial Economics, 25, 99–122. McAllister, P., & McManus, D. (1993). Resolving the scale efficiency puzzle in banking. Journal of Banking and Finance, 17, 389–405. Mercieca, S., Schaeck, K., & Wolfe, S. (2007). Small European banks: Benefits from diversification? Journal of Banking and Finance, 31, 1975–1998. Mitchell, K., & Onvural, N. (1996). Economies of scale and scope at large commercial banks: Evidence from the Fourier flexible function form. Journal of Money, Credit, and Banking, 28, 178–199. Puri, M. (1999). Commercial banks as underwriters: Implications for the going public process. Journal of Financial Economics, 54, 133–163. Pyle, D. (1971). On the theory of financial intermediation. The Journal of Finance, 26, 737–747. Rogers, K. (1998). Nontraditional activities and the efficiency of US commercial banks. Journal of Banking and Finance, 22, 467–482. Rumble, A., & Stiroh, K. (2006). The dark side of diversification: The case of U.S. financial holding companies. Journal of Banking & Finance, 30, 2131–2161. Slovin, M., Johnson, S., & Glascock, J. (1992). Firm size and the information content of bank loan announcements. Journal of Banking and Finance, 16, 1057–1071. Stiroh, K. (2000). How did bank holding companies prosper in the 1990s? Journal of Banking and Finance, 24, 1703–1745. Stiroh, K., & Rumble, A. (2006). The dark side of diversification: The case of US financial holding companies. Journal of Banking and Finance, 30, 2131–2161. Vander Vennet, R. (2002). Cost and profit efficiency of financial conglomerates and universal banks in Europe. Journal of Money, Credit, and Banking, 34, 254–282.