Product Diversification and Performance in the Financial Industry: FHCs’ Expansion into Insurance Activities

Mu-Sheng Chang* Department of Risk, Insurance, and Healthcare Management Fox School of Business and Management Temple University Philadelphia, PA 19122 TEL: (215) 204-1916 Fax: (215)204-4712 [email protected] Elyas Elyasiani Department of Finance Fox School of Business and Management Temple University Philadelphia, PA 19122 TEL: (215) 204-5881 FAX: (215) 204-1697 [email protected]

January 12, 2008

* Corresponding author. Thanks are due to Mary A. Weiss, John Burton, Jr., J. David Cummins, Norman A. Baglini, and Jacqueline Zinn for comments and discussions. Any remaining errors are ours.

Product Diversification and Performance in the Financial Industry: FHCs’ Expansion into Insurance Activities

Abstract This article investigates whether insurance activities (underwriting and agency) enhance the financial performance of financial holding companies (FHCs). Stiroh and Rumble (2006) and Yeager et al. (2007) have argued that the extension of banking to non-banking activities provides no diversification benefits for FHCs eligible to consolidate banking and insurance services. Using quarterly panel observations of 510 FHCs over the period 2003-2005, we obtain two main results: First, in the aggregate sample, risk-adjusted returns of FHCs improved by a shift toward non-interest activities. Second, risk-adjusted profitability of the FHCs is positively associated with insurance agency activities in small-sized FHCs, with insurance underwriting in large-sized FHCs, and unassociated with insurance activities in the very large FHCs (those with assets beyond the 75th percentile). An implication of our finding is that both small and large FHCs can reap diversification benefits as long as they choose the right niche.

Key words: Financial holding company (FHC), insurance underwriting, insurance agency, financial consolidation. JEL classification: G2, G22

1

Product Diversification and Performance in the Financial Industry: FHCs’ Expansion into Insurance Activities

1.

Introduction Recent divestitures by U.S. financial holding companies (FHCs) are often interpreted as an

indication that the lackluster financial performance of these conglomerates are due to the consolidation of commercial banking, investment banking and insurance business (Stiroh and Rumble 2006, p. 2134; Yeager et al. 2007, p. 330). More simply, it is suggested that the expansion of the banking enterprise into non-interest activities such as insurance services is unrewarding. Examples of such spin-offs include the dramatic divestiture by Citigroup of the Travelers property/casualty insurance unit to St. Paul Companies in 2003, and the Travelers Life & Annuity business to MetLife in 2005.1 It is notable, however, that these two large transactions had to do with insurance underwriting, an area that never attracted more than 20% of FHCs before 2006.2 In contrast, insurance agency income which was reported by 74 percent of FHCs in 2005 (by 63 percent in the first quarter of 2003) has not been carefully examined in the domain of FHCs.3 Using the agency theory due to Jensen and Meckling (1976), Aggarwal and Samwick (2003) contend that managers diversify their firms to capture private benefits, rather than achieving a reduction in idiosyncratic risk. Nonetheless, taking into account the firm-specific characteristics, Campa and Kedia (2002) suggest that diversification is a value-enhancing strategy.4 Furthermore, 1

Financial Services Fact Book 2005, published by Insurance Information Institute. “MetLife: Priming for Travelersesque Acquisition?” Grant Catton, Mergers & Acquisitions Report, February 20, 2006. 2 Insurance underwriting is the process whereby insurance companies assume risks (e.g. that a death, sickness, casualty or other event) will occur, for which premiums based upon underwriting standards are charged. Source: Instructions for Preparation of Consolidated Financial Statements for Bank Holding Companies, Reporting Form FR Y–9C, published by Board of Governors of the Federal Reserve System, Reissued March 2003. 3 Source: Federal Reserves. See related information in Financial Services Fact Book 2007, published by Insurance Information Institute. Agency (or brokerage) incomes are from non-underwriting activities, mostly from insurance product sales and referrals, service charges and commissions, and fees earned from insurance and annuity sales. Brokerage or agency incomes are used in contrast to underwriting incomes with respect to insurance activities throughout this article. A bank holding company (BHC) is a company that owns one or more banks and needs to meet certain requirements to become a FHC that can undertake more financial services. 4 In addition, Lamont and Polk (2001) argue that diversification does not destroy value on the grounds that discount

2

Berger et al. (2000) provide evidence that conglomerate production (focus strategies) is valid for some types of financial service providers but not for others. In light of the fact that nearly three quarters of FHCs are involved with insurance agency business, an investigation of insurance agency activities by FHCs in the aftermath of the Gramm-Leach-Bliley Act (GLBA, 1999) is compelling. We explore whether FHCs can enhance their risk-adjusted returns by putting banking and insurance services under one umbrella. Ever since the GLBA went into effect, FHCs have been able to offer banking, insurance, and securities products under a single corporate identity.

5

These products include: securities

underwriting and dealing, insurance underwriting, insurance agency activities, and merchant banking. 6 In view of the benefits conferred on the GLBA, numerous bank holding companies (BHCs) have converted to FHCs, while insurance companies and securities firms have taken advantage of this opportunity to broaden their realm of businesses. We employ the accounting data from the Federal Reserve FR Y-9C report that consists of all FHCs for the years 2003-2005 to examine the financial performance of the FHCs and its relation to insurance activities.7 Since 2003, FHCs have been required to report their underwriting and agency insurance incomes on the FR Y9C report. 8 This provides a unique opportunity for an in depth analysis of FHCs’ insurance activities. We employ a set of fixed-effect panel regressions to investigate whether FHCs engaged in insurance activities demonstrate a higher performance as, measured by risk-adjusted returns. Our results suggest that, at the aggregate level, FHC performance is positively and significantly firms have significantly higher subsequent returns than premiums firms. 5 Event dates surrounding the passage of GLBA can be found in Carow and Heron (2002, p. 471) and Akhigbe and Whyte (2001, p. 123). President Clinton signed GLBA on November 12, 1999. 6 Report to the Congress on Financial Holding Companies under the Gramm-Leach-Bliley Act, November 2003, Board of Governors of Federal Reserve System. 7 In 2005 and before, FR Y-9C form is filed by holding companies with total consolidated assets of $150 million or more. In 2006 and after, only those with total consolidated assets of $500 million or more have to file this report. 8 In 2001 and 2002, only insurance commissions and fees are reported in Schedule HI – Consolidated Income Statement. Before 2000, there is no specific report on insurance income in the statement.

3

associated with the share of total operating income from non-interest activities.9 Specifically, a positive and statistically significant relationship exists between the performance and insurance agency income of the small-sized FHCs as well as between the performance and insurance underwriting income of the large-sized FHCs. It follows that small and large FHCs can both reap the benefits of diversification by expanding into insurance activities as long as they choose the right area to diversify into. These findings stand in contrast to Stiroh and Rumble’s (2006) results that the gains due to diversification into non-interest activities are offset by increased risk. The evidence reached here, based on separating insurance activities into insurance underwriting and insurance agency in the post-GLBA era, is consistent with portfolio diversification theory purporting that expanded powers into non-banking activities reduces risk (Boyd et al. 1993; Saunders and Walter 1994; Kwan and Laderman 1999; Lown et al. 2000 ; Adkisson and Fraser 2003).10 We extend the literature in several ways. First, we use a comprehensive data set covering the FHCs over the period 2003-2005 to analyze the influence of insurance agency versus underwriting activities on FHC performance in the post GLBA era. There does exist a voluminous literature on bank mergers, but academic research into insurance activities within FHCs is still in its infantile period because formation of FHCs is a very recent phenomenon. Second, our results can complement previous studies by offering a more accurate profile of insurance agency activities in the new financial structure. Third, our evidence from the financial sector provides complementary viewpoints on diversification benefits derived from studies on the non-financial sectors.11 The paper proceeds as follows. Section 2 discusses the theoretical background and the hypotheses concerning the association of FHC performance and insurance activities. Section 3 presents the data and 9

In FR Y-9C, operational income is composed of net interest income and non-interest income. See Boyd, Graham, and Hewitt (1993) for a detailed discussion of portfolio diversification theory. Saunders and Cornett (2006) address product diversification benefits in a similar way. Adkisson and Fraser (2003) contend that a well-diversifying FHC can stabilize earnings and profits. 11 Most studies exclude the financial sector when they investigate diversification (Berger and Ofek, 1995; Servaes, 1996; Lamont and Polk, 2001; Campa and Kedia, 2002; Mansi and Reeb, 2002; Thomas, 2002; Schoar, 2002). 10

4

methodology. Section 4 discusses the empirical results coupled with robustness tests. Section 5 summarizes and concludes.

2.

Theoretical Background and Hypotheses Prior to the passage of the GLBA (1999), stringent regulatory barriers restricted the entry of

banks into insurance activities and vice versa.12 Removal of these restrictions by the GLBA, allows FHCs to sell insurance as an agent for a fee, or to act as an insurance underwriter.13 In general, insurance agency activities have low risks in loss potential, relative to insurance underwriting, and, hence, they appeal to a larger number of FHCs (Saunders and Cornett 2006). Table 1 shows a comparison of the FHCs engaged in insurance underwriting and those engaged in insurance agency, as well as a comparison of BHCs and FHCs. As of December 2005, only 465 out of the 2256 banking organizations (21%) had chosen to become FHCs.14 As compared to BHCs, FHCs are relatively larger in terms of asset holdings (Yeager et al. 2007).15 Theoretical Background Most studies of financial consolidation rely on either accounting or market data in evaluating the changes in financial performance due to product expansion. An alternative approach draws on the input-and-output price and quantity data to evaluate the changes in cost and/or profit efficiency due to consolidation. The advantage of accounting data is that it allows a more comprehensive analysis of FHCs because FHCs with assets of $150 million or more have been required to report financial statements to the Federal Reserve since 1978, providing a wealth of 12

Some crucial regulations include the Glass-Steagall Act of 1933, the Bank Holding Company Act of 1956, and its Amendment of 1970. The former Act separated commercial banks and investment banks, while the latter specified permissible activities in BHCs. See Saunders and Cornett (2006, p. 38) for major banks laws. 13 Carow (2001a, p. 138) tabulates important dates concerning bank insurance rights. The latest ruling affecting a bank’s right to sell insurance products before GLBA is the Supreme Court’s ruling on March 26, 1996, that nationally chartered banks may sell insurance. 14 As of December 2006, 299 of 974 (31%) of the banking organizations had chosen to be FHCs. This is due to an increase in the asset-size threshold for filing the FR Y-9C from $150 million to $500 million in 2006. See Section 3. 15 As of December 2004, the FHC share of total holding company assets was 86%.

5

information on these organizations.16 In contrast, market data are available only for the largest firms in the industry, limiting the sample size.17 Regulators of financial institutions rely heavily on accounting figures in their evaluation of a these institution’s risk and profitability conditions (Reichert and Wall 2000).18 Using market data on stock return reactions of financial institutions to the passage of the GLBA, several studies find that financial consolidation of the type allowed by this legislation can potentially generate diversification gains (Yildirim et al., 2006; Mamun et al., 2005; Carow and Heron, 2002; Akhigbe and Whyte, 2001).19 For example, Carow (2001b) finds that life insurance companies and large banks showed significant stock price increases around the time of the CiticorpTravelers Group merger (April 6, 1998).20 Saunders and Walter (1994) have also found that life and property insurance companies offer the biggest potential for money center banks to reduce their

16

Please see Section 3 for a detailed description of data on FHCs. Yildrim et al. (2007) address the restriction of using market data due to the fact they cover only the publicly traded firms. The advantage of market data is that they can readily quantify market value creation in response to financial convergence at a specific point in time. Moreover, stock returns incorporate forward-looking expectations, while the accounting measures fail to do so (Kwan and Laderman 1999). However, two primary drawbacks in using accounting data are that the current market value may not be considered, and a smoothing problem can occur over time. See Houston et al. (2001) for a detailed discussion of shortcomings of using accounting and market data. 18 The results from the studies using market data can help stock investors, but may not be directly relevant to regulators. Kwon and Laderman (1999, p. 28) contend that stock returns are composed of an idiosyncratic and a systematic component, but only the idiosyncratic component is relevant to regulators. 19 Mamun et al. (2005) report that money center banks and super regional banks benefited most from this deregulation, using a sample of three money center and four super regional banks. Mamun et al. (2005) suggest that the GLBA created diversification opportunities and was fairly successfully in containing possible risks. Carow and Heron (2002) suggest that diversification gains may arise from economies of scope, market power, and/or from an implicit extension of government guarantees to banking affiliates. Yildirim et al. (2006) further indicate a significant risk shift and overall reduction in riskiness for the financial sector around the event period. Carow and Heron (2002) and Yildirim et al. (2006) document that abnormal stock returns are positive and significant for investment banks and insurance companies but insignificant for banks. Akhigbe and Whyte (2001) document hat the stocks of all financial institutions responded positively when the Act was passed. 20 Carow (2001b) defines large banks as those with assets greater than $10 billion and small banks as those assets less than $10 billion. Her analysis provides evidence that investors expect large banks and insurance companies to receive significant benefits from the resulting financial convergence. The returns of small banks, health insurers, and property/casualty insurers are insignificantly different from zero in her event study. 17

6

systematic risk.21 In brief, market-based studies find that combinations of insurance and banking do produce diversification benefits.22 More recent studies employing accounting data do not find significant synergies between banking and non-banking activities. This stands at odds with portfolio diversification theory that expanding non-banking powers would reduce the risk of BHC failure (Boyed et al., 1993). For example, drawing on accounting data during 1997-2002, Stiroh and Rumble (2006) suggest that diversification benefits enjoyed by FHCs are offset by their ongoing shift toward non-interest activities because the latter are more volatile than the traditional interest-generating activities but not necessarily more profitable.23 However, Stiroh and Rumble (2006) break down the aggregate non-interest income only to four categories of fiduciary, service, trading, and other income, failing to explicitly consider FHC’s insurance activities. Yeager et al. (2007) too, find no evidence to support the view that financial consolidation between banking and insurance underwriting activities leads to significant improvement in profitability and productivities. 24 It is notable that the prevalence of insurance agency activities is not discussed in these studies.25 Thus, little can be inferred from these studies about the empirical relevance diversification gains due to engagement in insurance agency activity under financial convergence.

21

They construct portfolio market returns for an assortment of combinations between banks and other financial firms. Delong (2001) claims that general mergers in the banking industry neither create nor destroy shareholder wealth by examining the abnormal returns upon 280 merger announcements during 1988-1995. Mergers that focus on activity and geography enhance stockholder value by 3 percent. In other words, mergers of banks related to dissimilarity (diversification) fail to create positive market reaction. Due to her sample based on the pre-GLBA and mergers which do not involve with insurance or securities, little can be inferred from her findings for diversification benefits in the post-GLBA era. 23 Their sample is composed of a panel of FHC/year observations by calculating averages and standard deviations over the quarters in each year. 24 Yeager et al. (2007) treat the status of a holding company as a dummy variable in the model, that takes the value one if the firm is an FHC and zero otherwise. This study includes holding companies with at least 32 observations in the period 1996-2004. 25 Insurance underwriting is not separated from insurance agency in some study. For instance, Reichert and Wall (2000) argue that insurance activities become less influential in portfolio diversification by construct optimal portfolios of banking, insurance, and securities activities using the accounting data for the years 1974-97. The optimal portfolio mix is time-varying. Banks became a larger part of the efficient financial services portfolio in the 1990s, than they were in the 1970s and 1980s. 22

7

Several studies present a bright side for financial consolidations (Adkisson and Fraser, 2003; Lown et al., 2000; Boyd et al., 1993). For example, Adkisson and Fraser (2003) find that the very large first fliers (70 FHCs with assets exceeding $10B) did demonstrate higher profitability, with a large portion of profits being attributable to their non-banking subsidiaries. 26 Similarly, by simulating consolidations of banks with insurance companies, Lown et al. (2000) and Boyd et al. (1993) found that such consolidations may indeed reduce risk.27 Existing studies on efficiency effects of mergers among financial institutions rely on evidence either among banks or among insurers. For instance, Berger et al. (2000), employing data on U.S. insurance companies, find that diversified conglomerate insurers dominate in some types of financial services, while strategic focus insurers dominate in other types.28 Shaffer (1993), and Akhavein, Berger, and Humphrey (1997) suggest that bank mergers can potentially lower costs and increase profit efficiency.29 However, these studies do not determine whether consolidations of banking into non-banking enhance the efficiency of FHCs.

26

They compared 70 ‘first fliers’ which became FHCs in 1999 with otherwise comparable non-converting BHCs. However, the sample employed only accounts for a small part of the 457 FHCs which had sprang up by the end of 2000. 27 Lown et al. (2000) simulate mergers across the financial services industries using the accounting data of publicly traded financial firms in S&P Computstat database over the period 1984-98. They report that the largest diversification benefits would result if BHCs combined with life insurance firms. Such mergers would produce a lower standard deviation of ROE and a higher Z-score in terms of risk measure, i.e. a smaller probability of bankruptcy. They also cite the expansion of banks into the life insurance business in Europe as evidence. Boyd et al. (1993) use annual accounting data and market data to compute profitability and risk measures (Z-scores) for simulated mergers of BHCs with life insurance and property/casualty insurance firms over the sample period 1971-1987. See Stiroh (2004) for the calculation of Z-score. The Z-score, equal to (mean ROA+ mean Equity to assets) / standard deviation of ROA), measures how many standard deviation’s profits must fall below its mean to bankrupt the firm and is related to the probability of failure. Lower values indicate increased risk. 28 They contrast the performance of diversified conglomerates to that of the strategic focus firms by employing data on costs, revenues, and profits to gauge scope economies for insurer products over the period 1988-1992. In addition, Cummins et al. (1999) argue that mergers and acquisitions in the life insurance industry have had a beneficial effect on efficiency. 29 Paristiani (1997) observes that mergers during the 1980s were not beneficial to banks in terms of X-efficiency but did achieve moderate improvements in scale efficiency. His analysis suggests that in-market mergers (i.e., distance and overlap variables) yield no significant improvement in post-merger performance. In contrast, some other researchers find that bank mergers may not result in more efficiency. Berger and Mester (1997) and Berger et al. (1996) report that bank mergers may indeed bring about a burden rather than benefits. As banks grow larger, they continue to be able to control costs, but they find it harder to create revenue efficiently. Rhoades (1998) reviews nine case studies and reaches the conclusion that most mergers have not resulted in any significant post-bank-merger improvements in efficiency.

8

Hypotheses Sources of diversification benefits in financial conglomeration may include economies of scope and an increase in market power (Berger et al., 1999; Berger et al., 2000; Carow and Heron, 2002). Moreover, these organizations can serve as department stores offering an array of financial products including investment, savings, credit, and insurance products, to their customers. These one–stop shopping centers would provide considerable convenience for their clients and would reduce their search, information, monitoring, and transaction costs, compared to shopping from specialty firms (Skipper and Kwon, 2007). Kwan and Laderman (1999) find that securities activities, insurance agency, and insurance underwriting are all riskier and more profitable than banking activities.30 Moreover, as Mishkin (1999, p. 686) and Berger et al. (1999, p. 178) have pointed out, financial consolidation across activities has the potential to increase diversification in addition to lowering the portfolio risk and the likelihood of financial firm failure.31 The level of net interest income is dependent on changes in the interest rates and the level of intermediation, while non-interest income can be generated through such activities as fiduciary, service, trading, insurance, etc., which are less interest rate sensitive. For instance, fiduciary income can derive from commissions and fees on the sales of annuities that are executed by a firm in a fiduciary capacity.32 Service income may come from the charges from the accumulation or disbursement of funds deposited to Individual Retirement Accounts (IRAs). Trading revenues can be produced by trading derivative contracts of interest rates, exchange rate, equity, and commodity. Assuming that financial consolidation has the potential to produce diversification benefits, we formulate our first hypothesis as: 30

They employ accounting data to explore the possible effects of combining banking and non-banking financial activities on banking organizations’ profitability and risk. Their measures of profitability and risk are based on variance of ROE or ROA, coefficient of variance (standard deviation of returns/mean of returns) of ROE or ROA, or probability of bankruptcy. 31 Consistent with these studies, Yildrim et al. (2006) find a significant decline in systematic risk of the financial sector around the event period surrounding the passage of the GLBA (1999). 32 Fiduciary services are usually rendered by the trust departments of the BHCs.

9

H1: FHCs’ performance is associated positively with non-interest income activities. Since the deregulatory provisions of the GLBA were intended to improve competitiveness and performance of financial institutions, the main question here is whether insurance agency activities can contribute to FHC profitability. As Saunders and Walter (1994) have pointed out, certain insurance products—for example, credit life insurance, mortgage insurance, and auto insurance—tend to have natural synergistic links to bank lending. Insurance agency activities are also comparatively low loss risk and generate income that does not hinge on interest rates. Kwan and Laderman (1999) have argued that insurance agency activities require a low level of investment and, thus, can help reduce some categories of BHC risk.33 These activities are said to be riskier and also more profitable than banking services. The costs of developing the expertise associated with insurance agency are likely to be low. Besides, using a similar customer base, FHCs may cross-sell banking and insurance products and enjoy scope economies. Hence, our second hypothesis can be formulated as: H2: Risk-adjusted returns of FHCs are positively associated with insurance agency activities. The claims loss risk for an insurance underwriter is contingent on the actuarial predictability of losses relative to premiums earned (Saunders and Cornett, 2006).34 Underwriting risk exposure is due to unexpected increases in losses and administrative expenses, and/or unexpected declines in investment returns. Failure to properly predict these expenses can lead to enormous losses.35 Risks faced by insurance underwriters are referred to as event risk, timing risk, and credit risk, or simply as “ETC”. Event risk concerns the cause of loss, while timing risk reflects the manifestation and

33

Boyd and Graham (1988) and Boyd et al. (1993) suggest that insurance agency can reduce BHc’s probability of bankruptcy. 34 Saunders and Cornett (2006, p. 76) address that the predictability depends on a number of characteristics of the perils insured, e.g., property versus liability, severity versus frequency, long tail versus short tail, and product inflation versus social inflation. 35 Saunders and Cornett (2006) contend that some liability insurance lines with long-tail loss are especially difficult to estimate expected losses. Insurance underwriters also face market pressure brought about by their competitors that suppress price in an attempt to gain market share.

10

emergence of the loss. Credit risk exists because the underwriter bears the risk of not being able to collect premiums from the insured and/or to collect reinsured coverage from the reinsurer.36 As Yeager et al. (2007) note, insurance underwriting is substantially concentrated in few FHCs.37 Kwan and Laderman (1999) contend that insurance underwriting activities are riskier and more profitable than banking. However, in contrast to a bank, which may lose up to the entire amount of a loan, an underwriting insurer is liable only for losses covered by the policy. 38 Insurance underwriting takes a substantial expertise to handle uncertain risk. In view of sophisticated expertise, considerable risk involved with insurance underwriting, and potential higher profitability, we formulate our third hypothesis as: H3: Risk-adjusted returns of FHCs are positively associated with insurance underwriting activities.

3.

Methodology, Data, and Sample

Methodology Our goal is to examine the link between risk-adjusted returns of a FHC and its diversification into insurance activities. The procedure employed here is built on Stiroh and Rumble (2006). The model used for empirical specification is a fixed-effect panel regression described by equation (1) below:

Yit = α + β1SH NET ,it + β 2 SH NON ,it + β 3 SH LOAN ,it + γ X it + ei + ut + ε it In this model, Y is a measure of performance, SHNET is the share of net interest income in total, SHNON is the share of non-interest income, SHLOAN is the share of loans,39 and X is a vector of other

36

Insurance companies have a huge amount of reinsurance recoverable and are subject to credit risk in the event that reinsurers go insolvent. 37 Yeager et al. (2007, p. 321) document that just two firms – MetLife and Citigroup – accounted for 96% of the $453 billion in insurance assets as of December 2004. This statistic fails to account for Citigroup’s divesture of its life insurance business to MetLife in 2005. In addition, Out of 748 FHC quarterly observations that have insurance underwriting income, 628 (84 percent) have also insurance agency income. 38 We thank Norman Baglini for sharing with us his insight and experience in the insurance industry (October 2, 2007). 39 See page 19 for the rationale of using the share of loans.

11

control variables for FHCi in quarter t. The model incorporates three error components ei, ut, and ɛit, for state fixed effects, quarter fixed effects, and random error term, respectively. We use three performance measures: return on equity (ROE), return on assets (ROA), and risk-adjusted returns. ROE and ROA are defined as net income relative to equity and to total assets, respectively. Following Stiroh and Rumble (2006) and Yeager et al. (2007), risk-adjusted return (RAR) is calculated as ROE (or ROA) divided by respective standard deviations: RARROE = ROE/SDROE and RARROA = ROA/SDROA. The RAR measure accounts for changes in risk. We calculate the standard deviation of returns over the previous four, six, or eight quarters. These riskadjusted returns are denoted as: RARROE(4), RARROE(6), and RARROE(8), respectively.40 To measure activity diversification for a FHC, following Stiroh and Rumble (2006), we follow a two–step procedure. First, we calculate the proportions of the operational income from net interest and non-interest income sources. Since net operating income (NOI) comprises net interest income (Net-II) and non-interest income (Non-II), i.e., NOI = Net-II + Non-II, we define the share of each component as SHNET = Net-II/ NOI and SHNON = Non-II/NOI.41 Six categories of non-interest income are considered; fiduciary, service, trading, insurance agency, insurance underwriting and other revenues. 42 We construct the share of each of the above

40

In addition, since returns on FHCs have a pattern of starting low in the first quarter, growing incrementally in the second and third quarters, and culminating in the last quarter, we calculate the standard deviation of returns over the four, six, or eight quarters of the previous years, excluding the current year. Therefore, RARs on equity, denoted by RARROE(4a), RARROE(6a), and RARROE(8a) , respectively. This phenomenon is based on our observation of the data set. 41 In addition, following Stiroh and Rumble (2006), we calculate a measure of diversification (DIV) in the FHC’s operating income in a way similar to a Herfindahl index. This measure is defined as one minus the sum of the squared shares of net interest income (Net-II) and non-interest income (Non-II) in the total net operating income (NOI): DIV = 1 – (SHNET2+SHNON2). A higher value suggests a more diversified structure: a zero value indicates that all income is from one single source, while 0.5 denotes an even split between Net-II and Non-II. Some weaknesses of using this measure are detailed in the next section. 42 In FR Y-9C form, fiduciary, service, and trading income are reported in the records of BHCK4070, BHCK4483, and BHCKA220, respectively. Insurance underwriting and agency incomes are reported in the records of BHCKC386 and BHCKC387, respectively. Trading revenue, reported from cash instruments and derivative instruments, are classified into four types of exposures: interest rate, foreign exchange, equity security and index, and commodity and others. The service category is referred to as the service charges on deposit accounts in domestic offices. See footnote 2 for the data source.

12

income flows in total non-interest income as: SHfiduciary, SHservice, SHtrading, SHagency, SHunderwrting, and SHother, respectively.43 These shares show how focused a given FHC is on a particular non-interest activity. In order to examine how interest income affects performance, we follow Stiroh and Rumble (2006) to disaggregate the total loans into the shares of real estate, commercial and industrial (C&I), and consumer loans, SHresl estate, SHC&I, and SHconsumer, respectively.44 These loan shares show how heavily a FHC is dependent on a specific loan category. Control variables (vector X) include the log of total assets (size), equity to assets ratio (leverage), loan to assets ratio (illiquidity), asset growth,45 and loss to assets ratio (credit risk).46 The rationale for inclusion of the equity ratio, loan ratio, and asset growth rates is the belief that disparate financial structures may be associated with the performance of FHCs. A FHC can minimize equity and maximize loans with an aim to grow more rapidly. However, poor quality lending practices can have an adverse impact on returns. Accordingly, the loss ratio, i.e., allowance for loan and lease losses divided by total assets, is incorporated to control for this effect. Derivatives and off-balance-sheet items help to capture the factors that are likely to affect performance. Fixed-effect variables consist of quarter fixed effects, state fixed effects, and the number of quarters the FHC is in operation. The cyclicality of returns makes it necessary to include a quarter dummy in the model to control for differences in the banking environment over time. The state dummy variables can control for geographic differences among FHCs. As Stiroh and Rumble (2006) note, the state dummy variables can be considered as proxies for multi-state FHCs. The number of 43

In addition, non-interest income includes investment banking, advisory, brokerage, and underwriting fees and commissions (BHCKB490), venture capital revenues (BHCKB491), net servicing fees (BHCKB492), net securitization income (BHCKB493), net gains (losses) on sales of loans and leases (BHCK8560), net gains (losses) on sales of other real estate owned (BHCK8561), net gains (losses) on sales of other assets (excluding securities) (BHCKB496), and other noninterest income (BHCKB497). 44 The rationale of Stiroh and Rumble (2006) is that for net interest income, it is impossible to isolate the corresponding income flows. Thus, the alternative solution is to analyze the activities of the main lending lines in the balance sheet. 45 The dependent variable is the performance measure of FHCs. Growth opportunities (e.g., Tobin’s Q or market to book ratio) can be used as an alternative performance measure based on market data. 46 We tried control for derivatives and off-balance-sheet items but found they are insignificant and limited to some FHCs.

13

quarters variable is likely to control for any survivor effect, namely that long-lived FHCs may outperform others in terms of financial returns (Stiroh and Rumble, 2006). The coefficient of the non-interest share is expected to be positive because of diversification benefits in the shift to non-interest activities.47 The insurance agency share is predicted to have a positive sign because of potential scope economies, low investment, low loss risk, and higher profitability. We anticipate a positive sign for the insurance underwriting variable because it involves high loss risks, considerable expertise, and potential high profitability. Table 2 summarizes our predicted signs for the variables. We use quarterly data on a panel of FHCs over the sample period 2003-2005 to estimate the model.48. Our sample period features a stable and mature FHC market with the number of FHC hovering around 460.49 The sample period is characterized by 12 fed fund rate increase. Beginning the sample period in 2003 has the advantage that it avoids the early period of rapid growth in FHCs when they were still finding their niches and reaching maturity. Data and Sample

The list of FHCs and data were obtained from the web site of the Board of Governors of the Federal Reserve System. The consolidated financial statements for BHCs (FR Y-9C) have to be reported on a quarterly basis.50 In 2005 and before, this form was to be filed by BHCs with total consolidated assets of $150 million or more. Since 2006, the Fed has changed the rule, requiring 47

Kwan and Laderman (1999) find that securities activities, insurance agency, and insurance underwriting are all riskier and more profitable than banking activities. Some studies on diversification discount propose diversification is insignificantly or negatively related to excess firm value (Berger and Ofek, 1995; Servaes, 1998; Mansi and Reeb, 2002). However, their samples exclude financial services sector. 48 Stiroh and Rumble, 2006, p. 2155) employ a sample period of 1997-2002 and annual data. Their yearly observations come at the expense of increased noise when mean and volatility are calculated over four quarters of data. Compared to their sample period that is characterized by a volatile stock market and stock market crash, ours can provide complementary analysis of FHCs. 49 There were merely 94 FHCs in the first quarter of 2000, while the number of FHCs surged to 431 in the first quarter of 2003 (Yeager et al. 2007). 50 FR Y-9C report is required by Regulation Y and the Banking Holding Companies Company Act of 1956 as amended. Inflation, industry consolidation, and normal asset growth of BHCs led to an increase in the asset-size threshold for filing the FR Y-9C from $150 million to $500 million, effective with the March 2006, report date. http://www.federalreserve.gov/reportforms/ReportDetail.cfm?WhichFormId=FR_Y-9C&Historical=yes.

14

this report to be filed by BHCs with total consolidated assets of $500 million or more. These reports contain comprehensive accounting data for all U.S. BHCs and FHCs. Our sample consists of 510 U.S. FHCs, each of which has at least four quarters of data out of the twelve quarters in the sample period: 2003Q1 to 2005Q4.51 Due to changes of format in Schedule HI of the FR Y9-C that requires the reporting of underwriting income and income from other insurance and reinsurance activities in 2003, we are able to investigate insurance activities in more detail. 52 Given the changed threshold of assets from $150 to $500 million for reporting institutions in 2006, the sample period is chosen to cover the 2003-2005 time period. This allows us to collect information about insurance underwriting activities, which usually involve large transactions and easily receive the attention of the public media, as well as insurance agency activities, an area of activity that the majority of FHCs are engaged in.53

4.

Empirical Results

Our principal objective is to explore the relationship between the risk-adjusted performance of FHCs and diversification of their non-interest income activities. Table 3 presents descriptive statistics for the main variables of interest. Three steps will be taken here. First, we examine whether risk-adjusted return of FHCs are related to the share of non-interest income in total at the aggregate level and report the results in Table 4. Second, we disaggregate non-interest income into five categories to analyze the relationship at the component level and report the results in Tables 551

Our sample contains top-tier FHCs only. Due to the tiered nature of the U.S. financial institutions, i.e., one holding company may own another, the inclusion of only the top-tiered holding company can avoid multiple counting of the same activities. Those records with BHCK9802 = "1" or "3" are selected for top-tier FHCs. The sample excludes FHCs headquartered in the Cayman Islands, Gibraltar, and Puerto Rico (12 FHCs). 52 In 2001 and 2002, it only requires the report of insurance commission and fees. In 2000 and before, there is no specific report of insurance income. 53 Survivor bias is the effect of considering only the performance of FHCs that are alive and present in the database in the sample period. The number of available FHCs quarterly files in our dataset is around 460 over the period 2003Q12005Q4. Our final sample, including 510 FHCs with 5459 quarterly observations, filtered out only 106 observations that didn’t meet the criteria. This is quite representative of the entire population. Therefore, survivor bias is expected to be minimal.

15

6.54 Third, further analysis will be conducted to explore whether insurance activities play different roles in the sub-samples based on the size of FHCs’ assets and report the results in Table 7. Robustness tests using the period 2003-2006 are presented in Table 8. By and large, the vast majority of FHCs’ revenues derive from net interest income. Specifically, net interest income accounts for 82 percent of total income for the median FHC (Table 3). In other words, the median of non-interest income share across FHCs is only 18 percent. The 1st , 2nd, and 3rd quartiles of the trading share or the insurance underwriting share are zero, suggesting that trading and insurance underwriting activities are concentrated in a small number of FHCs. The distribution of assets ($ millions) across FHCs deserves further attention because the median of assets is merely $507 million as compared to the mean of $17,858 million. A small number of very big FHCs make the mean drastically higher than the median assets, implying the FHC size distribution is skewed considerably skewed. 55 In addition, due to a high correlation between diversification index (DIV) and SHNON (or SHNET), we can only incorporate them alternatively.56 Furthermore, off-balance-sheet items are highly correlated with assets, so only the asset variable is included in the model.57 Consistent with our first hypothesis, a positive and significant relationship exists between risk-adjusted performance and non-interest share in total income indicating that increased expansion into non-interest activities enhances risk-adjusted returns (Table 4).58 This finding is contrary to the arguments made by Stiroh and Rumble (2006) who put forward a dark side of diversification by

54

We present the results based on ROE only. Measures based on ROA are qualitatively similar. Saunders and Cornett (2006) point out that the U.S. banking system is unique in that it comprises not only very big banks but also a large number of relatively small community banks. 56 Whether this is an orthogonal issue deserves further study. In addition, This DIV variable can not distinguish whether non-interest or net interest activities contribute to performance. For example, a FHC with 25 percent of its revenues from non-interest activities and a FHC with 75 percent would have the same DIV measure. A correlation matrix among main variables is in appendix. 57 This selection of the asset variable can avoid multicollinearity problem and increase our sample size. 58 We also left out the behemoth FHCs with assets more than 85th percentile from the sample to check whether some outliers are influential. However, the statistical results do not change quantitatively. 55

16

delineating a negative association between performance and non-interest activities. 59 Using quarterly observations on a panel of FHCs, we present the positive side. The negative and significant coefficient of the share of net interest income (SHNET) in all regressions in Table 5 indicates that the poor performance of FHCs over the sample period (20032005) can be attributed partially to an increased reliance on net interest income. Regarded as the core of the banking business, net interest income is the dominant source of revenues in FHCs in that SHNET at the 25th and 50th percentile is almost 76 and 88 percent of entire operating income, respectively. This stands in contrast to the claim that increased non-interest income is the sources the poor performance of a FHC. The coefficients of insurance agency activities are mostly insignificant or negative, rejecting our hypothesis that risk-adjusted returns of FHCs are positively associated with insurance agency activities. The prevalence of insurance agency among FHCs appears not to be in line with this outcome. However, our risk-adjusted profitability measure is positively related to the share of noninterest income from insurance underwriting. As expected, this positive relationship is consistent with our last hypothesis but not with the findings in Yeager et al. (2007). In addition, the shift to non-interest activities can be beneficial as evidenced by a positive relationship between performance and the service share. This positive relationship holds especially in the models in which standard deviation of ROE is based on the previous 8 quarters (Table 5). Service income comes from charges on deposit accounts.60 Customers appear to be willing to pay reasonable fees in exchange for good services. Disaggregation based on whether a FHC is engaged in insurance activities

59

Stiroh and Rumble (2006, p. 2159) address their finding may be due to the sample period examined (1997-2002) that is a bad draw for the FHCs because there were stock market crash, telecomm bust, slowdown in initial public offerings, and Russian debt crisis. 60 For instance, some FHCs like Wachovia and Bank of America have 20-30 percent of their non-interest income from service charges in our sample.

17

To check whether insurance activities affect the performance of FHCs, we disaggregate all FHCs into subsamples of those the firms merely pertaining to insurance underwriting, insurance agency, or both. Estimation results for those FHCs involved with some type of insurance activities are reported in Table 6. According to the test statistics in this table, risk-adjusted returns continue to be consistently and positively associated with the insurance underwriting share and still insignificantly related to insurance agency. Disaggregation based on the FHC’s asset size

To investigate whether the association between risk-adjusted profitability and diversification is size dependent, we divide the sample into three groups with cutoff points at the 25th and 75th percentiles of FHCs’ assets.61 In other words, we divide the sample, according to FHC assets, into three subgroups of small (below the 25th percentile), large (the 26th - 75th percentile), and very large (above 75th percentile). This step facilitates our analysis on the contribution of non-interest activities to performance for each group with similar corporate financial structure. We estimate the model once for each sub-sample and report the results in table 7. In this table, risk-adjusted performance of FHCs is found to be positively related to the insurance agency share in the small-sized subgroup, an outcome consistent with our second hypothesis.62 The implication of this finding is that small FHCs can enhance their risk-adjusted returns by expanding into insurance agency activities. As mentioned earlier, the loss risks involved when a business acts as an insurance agency are low. According to our findings, small community FHCs benefit from consolidating banking and insurance agency activities. Contrary to Stiroh and Rumble’s (2006) argument that the increased shift to non-interest activities does not bring about better returns, our finding presents evidence that synergies do exist between banking and insurance 61

The descriptive statistics (Table 3) for the assets show that the fourth-quartile group has a very wide dispersion across FHCs in terms of size. That is, the maximum assets are about 1,000 times larger than the 75th largest FHC. 62 Statistical results do not change quantitatively as standard deviation is calculated excluding the current year. Therefore, RARs on equity, denoted by RARROE(4a), RARROE(6a), and RARROE(8a), are not reported in the tables 7-8.

18

agency for smaller-size FHCs. In addition, risk-adjusted returns are also positively related to the net interest share in the small-sized FHCs. This is an interesting finding because small FHCs also heavily depend on interest-generating activities. However, the relationship is not conclusive between insurance agency and profitability in large and the very large FHCs.63 The results for the relationship between FHC performance and insurance underwriting are reported in Table 7. According to these results, this relationship is positive and significant for the large-sized FHCs, consistent with our third hypothesis that the performance of FHCs is positively associated with insurance underwriting activities. Nevertheless, this outcome does not hold for the small and the very large FHCs. It appears that small-sized FHCs are incapable of handling the potential loss risks from insurance underwriting activities or too small to benefit from scale economies in this line of business. Very large FHCs also tend to have well-diversified portfolios already and may have little gains from further diversification into underwriting. The insignificant relationship in the very large subgroup may explain in part the recent divestiture activities by powerhouses such as Citigroup and JPMorgan Chase.64 This outcome is consistent with the finding of Yeager et al. (2007) that synergies do not exist between commercial banking and insurance underwriting.65

63

The performance of FHCs varies inversely with the insurance agency share in large FHCs as standard deviation of ROE is computed by the previous 4 or 6 quarters. This relationship becomes insignificant in the models that standard deviation of ROE is based on the previous 8 quarters. In addition, the performance of FHCs is negatively related to insurance agency in the very large FHCs as risk-adjusted performance is based on the previous 4 quarters. The relationship between performance and insurance agency in the very large FHCs is insignificant in the models based on the previous 6 and 8 quarters. 64 Rieker, Matthias, “JPM Insurance Sale Endnote on a Nontrend,” American Banker, February 9, 2006. 65 Yeager et al. (2007) investigate whether within-firm profitability and productivity change after BHCs became FHCs over the years 1996-2004. Their sample of FHCs has mean assets of 16,867 millions; ours has 17,858 millions over the period 2003-2005. As another explanation of this finding, Jay Fishman, CEO of The Travelers Companies, Inc., has stated that although banking and insurance firms are both in the financial sector, there is a major inherent difference in management between them. As a result, financial consolidation may become a catastrophe if the executives fail to coordinate well between these two divisions. His comments were made in response to authors’ questions in the 2007 American Risk and Insurance Association annual meeting in Quebec City, Canada, on August 5, 2007. Citibank paid a high price to learn this lesson.

19

Risk-adjusted profitability is positively and significantly connected to trading revenues only in the very large FHCs (Table 7). These institutions increase their profitability by trading derivatives in their off-balance-sheet accounts.66 Trading income is the most volatile part of noninterest income (Stiroh 2004).

However, trading activities are concerned with high risk of

derivative contracts and cluster in 12 percent of FHCs only.67 Hence, it is reasonable to see a mostly insignificant relationship between the trading share and performance in the small and large-sized FHCs. Robustness

We employ the sample period 2003-2006 to examine whether insurance activities can robustly enhance performance by using 2006 FHC dataset as a base. The sample contains FHCs that have more than $500 million consolidated assets in 2006. Some of these FHCs were small-sized in 20032005, growing to be qualified for the report after 2006, but those FHCs that have assets less than $500 millions from 2003 to 2006 are not included in this sample. Table 8 reports panel regressions of FHC’s performance on corporate characteristics over the period 2003-2006. Based on this table, performance is positively and significantly associated with insurance underwriting in the large-sized FHCs. This finding confirms that insurance underwriting can provide profitable opportunities for some FHCs. Similar to our findings in the previous section, this positive relationship between insurance underwriting and profitability does not hold in the small- and very large subgroups.68 Of particular interest is the finding that insurance agency is positively associated with risk-adjusted performance in the large-sized FHCs. Insurance 66

Very large FHCs are highly exposed to off-balance-sheet activity risk. The correlations between the assets and various off-balanced-sheet items range from 85 to 95 percent. An off-balance-sheet account is usually employed to reduce risk through hedging with derivative securities and other means (Saunders and Cornett, 2006). Despite the high risk involved that may result in tremendous losses, trading activities are prevalent in the very large FHCs. 67 88 percent of the quarterly observations in our sample have non-zero trading share. There are just 11 percent with positive trading shares and 1 percent with negative trading shares. 68 We further check this relationship for the FHCs with assets between $697.47 and $5628.36 millions in the 2003-2005 sample used in the previous section, finding a positive and statistically significant relationship still exists between insurance underwriting and profitability.

20

agency activities offer opportunities of increasing the profitability in the large FHCs. However, this result looks different from the finding in the previous section that a positive relationship between insurance agency and profitability exists in the small-sized FHCs.69 Thus, we further tested whether this positive relationship exists year by year, finding that it only holds in 2006. We infer that benefits of consolidating banking and insurance agency activities in large FHCs are unstable and may vary across firms.70

5.

Conclusion

Using the accounting data of FHCs over the period 2003-2005, we find evidence that the bank expansion into non-interest activities can effectively improve risk-adjusted performance. Our panel regressions suggest that insurance agency activities help small-sized FHCs improve riskadjusted returns, and insurance underwriting activities are significantly conducive to risk-adjusted profitability for large FHCs. Insurance activities do not have a significant impact on the performance of the very large FHCs. Our results confirm the contribution of insurance activities to the diversified financial conglomerates. The Citi Group case should be considered a special case for which diversification did not work well from time to time, possibly because of the conflict in management styles and personalities of the executives, rather than a “poisonous” example of diversification that has an adverse effect on every type of financial consolidation. After all, FHCs need diverse skills and expertise to survive.

69

Small-sized FHCs in the sample 2003-2005 are not included in the sample 2003-2006 if their assets are not more than $500 millions in 2006. 70 The 25th and 75th percentile assets in the sample 2003-2005 are $ 274 and $ 1,546 millions, respectively (Table 7). The same two percentile assets in the sample 2003-2006 are $697 and $5,628, respective (Table 8). This change is due to the increase of asset threshold from $150 m to $500 m in 2006.

21

Further research can be conducted by pursuing the following areas to uncover more facts about the development of FHCs. We need to identify insurance activities as life or property and liability. This paper can be further improved by testing for random effects and heteroskedasticity to make sure the model is well specified. The use of Tobin’s Q as a performance measure can help investigate the robustness of the results.

Table 1 FHCs engaged in underwriting and agency activities

FHCs engaged in insurance underwriting FHCs engaged in insurance agency FHCs engaged in both Number of FHCs Number of BHCs Total of FHCs and BHCs

2003 65 14% 322 71% 54 12% 453 100% 1,668 2,121

2004 57 346 50 477 1,752 2,229

2005 12% 56 73% 342 10% 54 100% 465 1,791 2,256

12% 74% 12% 100%

2006* 54 227 50 299 675 974

18% 76% 17% 100%

Note: The analysis is based on the data of the fourth quarter of each year. Before 2005, FR Y-9C form is filed by holding companies with total consolidated assets of $150 million or more. * In 2006 and after, only those with total consolidated assets of $500 million or more have to file this report. Source: Consolidated Financial Statements for Bank Holding Companies- FR Y9C, Board of Governors of the Federal Reserve System.

Table 2 Summary of Expected Signs and Rationale

Variable

Expected signs

Non-interest share

+

Insurance agency share

+

Insurance underwriting share

+

Rationale Expansion into non-interest activities can lead to better performance. Insurance agency activities are involved with low loss risk. Some have natural synergistic links to banking services. Insurance underwriting activities are involved with high loss risk, sophisticated expertise, and potential higher profitability.

22

Table 3 Descriptive Statistics The firm-level sample includes all FHCs with at least four quarters of data over the period 2003-2005. The profit ratios are ROE and risk-adjusted returns (RARs) of ROE (i.e., RARROE = ROE / SDROE). The standard deviation (SD) of ROE is calculated by 4 previous quarters, and RARs on equity is denoted by RARROE(4). SHNET is the share of operating income from net interest income. SHNON is the share of operating income from non-interest income. The assets are in million dollars. The equity ratio is equity/assets. The loan ratio is loans/assets. The loss to assets ratio is allowance for loan and lease losses/assets. SHNON contains the share of total non-interest incomes for each of the non-interest income flows, denoted by SHfiduciary, SHservice, SHtrading, SHunderwrting, and SHagency, respectively. SHLOAN contains the share of total loans for real estate, commercial and industrial (C&I), and consumer lendings, denoted by SHresl estate, SHC&I, and SHconsumer, respectively.

ROA ROE RARROA* RARROE* DIV SHNET SHNON SHfiduciary SHservice SHtrading SHunderwrting SHagency Assets ($m) Equity / assets Loans / assets Loss / assets SHreal estate loan SHC&I loan SHconsumer loan Valid N (listwise)

N 5459 5459 4946 4933 5459 5459 5459 5459 5459 5459 5459 5459 5459 5459 5459 5459 5459 5459 5459 4563

Min. -2.5 -51.8 -15.5 -28.0 .0 2.3 .3 -.7 .0 -1306.2 -235.5 -34.1 122 .71 .60 .01 .00 .00 .00

1st Q 0.34 3.97 1.04 1.03 .21 75.89 12.35 0.00 22.51 0.00 0.00 0.00 274 7.44 58.42 0.68 62.40 10.02 2.82

Mean .718 7.97 2.04 2.05 .29 79.63 20.37 7.70 38.19 .45 .74 5.30 17,858 9.22 65.06 .84 70.13 16.27 7.64

Median 0.61 7.09 1.89 1.90 .28 82.42 17.57 0.87 35.79 0.00 0.00 0.79 507 8.63 67.70 0.83 72.34 14.63 5.65

3rd Q 0.95 10.87 2.83 2.86 .36 87.64 24.10 10.70 53.75 0.00 0.00 4.51 1,564 10.20 74.43 0.97 80.61 20.48 10.01

Max. 13.7 67.9 26.9 17.8 .5 99.7 97.7 97.4 191.8 76.6 86.8 76.1 1,547,789 76.5 93.6 2.8 217.5 96.5 100.0

S. D. .65 5.49 1.29 1.33 .11 12.94 12.94 13.12 21.14 18.33 6.56 10.98 102,595 4.18 14.39 .29 15.83 9.75 7.33

Note: There are some outliers in the variables of SHtrading, SHunderwrting, and SHagency. A further work will be conducted as they are removed from the sample. A preliminary check found that a couple of FHCs have negative income in some quarters.

23

Table 4 Panel Regressions of FHCs’ Performance on Corporate Characteristics The regression specification is:

Yit = α + β1SH NON ,i + γ X it + ei + ut + ε it

The firm-level sample includes all FHCs with at least four quarters of data over the period 2003-2005. The dependent variable is risk-adjusted return (RAR) of ROE (i.e., RARROE = ROE / SDROE). The standard deviation (SD) of ROE is calculated by 4, 6, or 8 previous quarters, respectively. If SD includes the current year, RARs on equity are denoted by RARROE(4), RARROE(6), and RARROE(8), respectively. SHNON is the share of operating income from non-interest income. Xit contains the logarithm of total assets, the equity to assets ratio, the loan to assets ratio, the growth of assets, i.e., (assetst/assetst-1)-1, and the loss to assets ratio, i.e., allowance for loan and lease losses divided by assets. The specification also includes state fixed effects (ei), quarter fixed effect (ut), and the number of quarters the FHC is observed. A p-value is in parentheses below each coefficient. *** Significant at 1 percent; ** Significant at 5 percent; *Significant at 10 percent.

RARROE(i) Variable Constant SHNON ln(Assets) Equity / assets

i=4

i=6

i=8

2.239***

2.067***

2.002 ***

(.000)

(.000)

(.000)

.010***

(.000)

(.000)

-.016

-.006

-.012

(.185)

(.567)

(.206)

-.020***

.002 (.237)

Asset growth Loss / assets

-.005*

-.015*** (.000) .003** (.018) -.004**

-.018*** (.000) .005*** (.000) -.003

.438 (.177) .017*** (.000) -.036** (.016) -.026*** (.000) .004*** (..055) -.007*

i = 6a .579 ** (.029) .015 *** (.000) -.030 *** (.009) -.025 *** (.000) .006*** (.000) -.003

i = 8a .357 (.194) .012*** (.000) -.021* (.072) -.022*** (.000) .006*** (.000) -.007**

(.092)

(.108)

(.184)

(.058)

(.244)

(.022)

.033

-.030

-.020

.002

-.055

.006

(.637)

Adjusted R2 Observations

.008 **

(.000)

(.000)

Loans / assets

.007***

i = 4a

(.643)

(.725)

(.981)

(.417)

(.931)

.373

.558

.537

.325

.482

.477

4568

4284

3991

4374

4029

3729

24

Table 5 Panel Regressions of FHCs’ Performance on Non-interest Income The regression specification is:

Yit = α + β1SH NET ,it + β 2 SH NON ,it + β 3 SH LOAN ,it + γ X it + ei + ut + ε it

The firm-level sample includes all FHCs with at least four quarters of data over the period 2003-2005. The dependent variable is risk-adjusted return (RAR) of ROE (i.e., RARROE = ROE / SDROE). The standard deviation (SD) of ROE is calculated by 4, 6, or 8 previous quarters, respectively. If SD includes the current year, RARs on equity are denoted by RARROE(4), RARROE(6), and RARROE(8), respectively. SHNET is the share of operating income from net interest income. Xit contains the log of total assets, the equity to assets ratio, the loan to assets ratio, the growth of assets, i.e., (assetst/assetst1)-1, and the loss to assets ratio, i.e., allowance for loan and lease losses divided by assets. SHNON contains the share of total non-interest incomes for each of the non-interest income flows, denoted by SHfiduciary, SHservice, SHtrading, SHunderwrting, and SHagency, respectively. SHLOAN contains the share of total loans for real estate, commercial and industrial (C&I), and consumer lendings, denoted by SHresl estate, SHC&I, and SHconsumer, respectively. The specification also includes state fixed effects (ei), quarter fixed effect (ut), and the number of quarters the FHC is observed. A p-value is in parentheses below each coefficient. *** Significant at 1 percent; ** Significant at 5 percent; *Significant at 10 percent.

Variable Constant SHNET ln(Assets) Equity / assets

i=4

Loss / assets SHfiduciary SHservice SHtrading SHunderwrting

SHreal estate loan SHC&I loan SHconsumer loan Adjusted R2 Observations

i = 8a

2.566***

2.638***

2.806***

1.887***

1.476***

(.194)

(.194)

(.000)

(.000)

(.194)

-.010***

-.008***

-.010***

-.018***

(.000)

(.000)

(.000)

(.000)

-.015

-.006

-.010

-.039

(.215)

(.575)

(.313)

(.012)

-.020*** .006 -.005*

-.015*** (.000) .003* (.067)

-.018*** (.000) .004*** (.000)

-.026*** (.000) .004* (.082)

(.000) -.026** (.029) -.026*** (.000) .006*** (.000)

(.000) -.017 (.147) -.023*** (.000) .006*** (.000)

-.004

-.003 (.184)

(.059)

(.213)

(.022)

.065

-.006

.005

.010

-.029

.021

(.360)

(.923)

(.938)

(.908)

(.673)

(.777)

-.002

(.079)

(.200)

-.001

.000

(.485)

(.720)

-.002* (.082) .001** (.194)

-.001

-.003

-.015***

(.103)

-.003*

-.007*

-.017***

(.066)

-.005***

(.451)

(.001)

.000

.002

(.987)

(.152)

-.007**

-.004*** (.008) .003** (.013)

.000

.003

.002

.005

.002

.005

(.933)

(.352)

(.557)

(.300)

(.567)

(.145)

.003 (.249)

SHagency

i = 6a

3.007***

(.617)

Asset growth

RARROE(i) i=8 i = 4a

(.194)

(.000)

Loans / assets

i=6

.005** (.029)

.006*** (.003)

-.003

-.002

-.001

(.115)

(.266)

(.387)

.003**

.004**

.003**

.001 (.809) -.003* (.096) .004**

.004* (.091)

.004* (.084)

-.002

-.002

(.292)

(.240)

.004*

.003**

(.060)

(.019)

(.017)

(.089)

(.010)

(.063)

.002

.001

.001

.002

.002

.002

(.415)

(.524)

(.654)

(.402)

(.339)

(.476)

.002 (.449)

.005* (.085)

.004* (.083)

.004 (.222)

.005* (.064)

.004 (.116)

.374

.559

.539

.325

.486

.480

4568

4284

3991

4374

4029

3729

25

Table 6 Panel Regressions of FHCs’ Performance on Non-interest Income - subsamples The regression specification is:

Yit = α + β1SH NET ,it + β 2 SH NON ,it + β 3 SH LOAN ,it + γ X it + ei + ut + ε it

The firm-level sample includes all FHCs with at least four quarters of data over the period 2003-2005. The dependent variable is risk-adjusted return (RAR) of ROE (i.e., RARROE = ROE / SDROE). The standard deviation (SD) of ROE is calculated by 4, 6, or 8 previous quarters, respectively. If SD includes the current year, RARs on equity are denoted by RARROE(4), RARROE(6), and RARROE(8), respectively. SHNET is the share of operating income from net interest income. Xit contains the log of total assets, the equity to assets ratio, the loan to assets ratio, the growth of assets, i.e., (assetst/assetst1)-1, and the loss to assets ratio, i.e., allowance for loan and lease losses divided by assets. SHNON contains the share of total non-interest incomes for each of the non-interest income flows, denoted by SHfiduciary, SHservice, SHtrading, SHunderwrting, and SHagency, respectively. SHLOAN contains the share of total loans for real estate, commercial and industrial (C&I), and consumer lendings, denoted by SHresl estate, SHC&I, and SHconsumer, respectively. The specification also includes state fixed effects (ei), quarter fixed effect (ut), and the number of quarters the FHC is observed. A p-value is in parentheses below each coefficient. *** Significant at 1 percent; ** Significant at 5 percent; *Significant at 10 percent.

Variable Constant

FHCs with insurance underwriting RARROE(6) RARROE(8) .420 (.646)

SHNET ln(Assets) Equity / assets Loans / assets

-.032 *** (.000)

-.010 *** (.000)

-.013 *** (.000)

-.011 *** (.000)

-.013 *** (.000)

-.105

-.004

-.009

-.006

-.009

(.007)

(.745)

(.405)

(.661)

(.406)

-.034 * (.061) .018 *** (.003) -.011 ** (.048)

-.016 *** (.001) .002 *** (.277) -.007 ** (.046)

-.023 *** (.000) .004 *** (.012) -.006 ** (.043)

-.017 *** (.001) .003 * (.091) -.008 ** (.018)

-.023 *** (.000) .005 *** (.002) -.006 ** (.021)

-.211

-.030

.027

.014

.000

.017

(.908)

(.727)

(.826)

(.997)

(.792)

.000

.001

.001

.000

(.583)

(.663)

(.992)

.011 *

.011 (.065)

(.075)

(.839)

-.001

.001

.001

(.859)

(.882)

(.383)

(.044)

(.163)

(.007)

.005

.003

.005

.004

(.295)

(.373)

(.196)

(.215)

SHtrading

.021 **

.012 (.246)

SHunderwrting

(.043)

.009 *** (.006)

.011 *** (.001)

.003 *** (.200)

.002 **

.005 ** (.018)

.002

.006 *** (.018)

.003 ***

.007 *** (.001)

-.007

-.004

.000

.000

.000

.001

(.254)

(.540)

(.794)

(.630)

(.974)

(.630)

SHreal estate loan

.028 *** (.001)

SHC&I loan

.029 *** (.000)

.022 *** (.006)

SHconsumer loan

.017 ** (.037)

.041 *** (.000)

Adjusted R2 Observations

2.065 *** (.000)

(.400)

SHfiduciary

SHagency

2.152 *** (.000)

-.018 ** (.002)

SHservice

2.424 *** (.000)

.017 *** (.003)

Loss / assets

2.478 *** (.000)

-.001 (.938)

Asset growth

1.992 **

-.072 * (.055)

FHCs with either type RARROE(6) RARROE(8)

(.036)

-.025 *** (.000)

FHCs with insurance agency RARROE(6) RARROE(8)

.041 *** (.000)

.003 ** (.205)

.004 ** (.035)

.006 ** (.016)

.007 *** (.001)

-.001

.000

.001

.002

(.736)

(.978)

(.671)

(.435)

.001

.004

(.674)

(.219)

.006 * (.053)

.008 *** (.004)

.590

.482

.553

.563

.550

.480

614

599

3135

2959

3221

3043

26

Table 7 Panel Regressions of FHCs’ Performance on Non-interest Income 2003-2005 The regression specification is: Y = α + β SH it 1 NET ,it + β 2 SH NON ,it + β 3 SH LOAN ,it + γ X it + ei + ut + ε it The firm-level sample includes all FHCs with at least four quarters of data over the period 2003-2005. The dependent variable is risk-adjusted return (RAR) of ROE (i.e., RARROE = ROE / SDROE). The standard deviation (SD) of ROE is calculated by 4, 6, or 8 previous quarters, respectively. If SD includes the current year, RARs on equity are denoted by RARROE(4), RARROE(6), and RARROE(8), respectively. SHNET is the share of operating income from net interest income. Xit contains the log of total assets, the equity to assets ratio, the loan to assets ratio, the growth of assets, i.e., (assetst/assetst1)-1, and the loss to assets ratio, i.e., allowance for loan and lease losses divided by assets. SHNON contains the share of total non-interest incomes for each of the non-interest income flows, denoted by SHfiduciary, SHservice, SHtrading, SHunderwrting, and SHagency, respectively. SHLOAN contains the share of total loans for real estate, commercial and industrial (C&I), and consumer lendings, denoted by SHresl estate, SHC&I, and SHconsumer, respectively. The specification also includes state fixed effects (ei), quarter fixed effect (ut), and the number of quarters the FHC is observed. A p-value is in parentheses below each coefficient. *** Significant at 1 percent; ** Significant at 5 percent; *Significant at 10 percent. Assets ≤ 25th percentile ($274m)

Subsamples

25th < Assets ≤ 75th

Assets > 75th percentile ($1564m)

RARROE(i) Variable Constant SHNET

i=4

i=6

2.521

-1.282

-1.441

(.585)

(.795)

(.693)

.024*** (.010)

ln(Assets) Equity / assets Loans / assets Asset growth Loss / assets SHfiduciary SHservice SHtrading SHunderwrting SHagency SHreal estate loan SHC&I loan SHconsumer loan Adjusted R2 Observations

.021** (.045)

i=8

.014* (.062)

-.154

-.065

.076

(.668)

(.866)

(.788)

-.013

-.010

(.556)

(.675)

-.008

-.010**

(.113)

(.044)

-.017

-.012

(.135)

(.317)

-.053

.058

(.804)

(.795)

-.005

-.005

(.273)

(.345)

-.059*** (.001) -.012*** (.001) -.016* (.084) .269* (.091) -.010** (.017)

i=4

5.121*** (.000) -.014*** (.000) -.102** (.026) -.042*** (.000) .006*** (.004) .003 (.444) .021 (.840)

i=6

i=8

i=4

i=6

i=8

4.437*** (.000) -.010*** (.000) -.070* (.074) -.038 (.000) .008** (.000) .003 (.338) -.129 (.151)

5.363 (.000) -.012* (.000) -.138*** (.000) -.051*** (.000) .008*** (.000) .003* (.400) -.033 (.718)

3.484***

3.921***

3.293***

(.000)

(.000)

(.000)

-.015*** (.000)

-.015*** (.000)

-.015*** (.000)

-.006

-.026

-.035

(.832)

(.281)

(.136)

-.021*** (.001) .001 (.852) -.014*** (.001) .146* (.255)

.000

-.001

-.001

-.004

(.785)

(.435)

(.447)

(.170)

-.014** (.012) .006*** (.070) -.014*** (.000)

-.008 (.140) .009*** (.002) -.009** (.016)

.006

-.082

(.961)

(.474)

-.001* (.600)

-.004 (.154)

-.002

-.003

-.004

-.001

.000

.002

.000

.002

.001

(.484)

(.386)

(.207)

(.436)

(.748)

(.164)

(.876)

(.533)

(.817)

.003

.003

.003

(.893)

(.850)

(.810)

.014

.033

.020

(.549)

(.157)

(.197)

-.010*

-.006

-.009*

(.086)

(.192)

(.058)

.008*** (.006) -.003* (.050) -.005* (.080) -.008** (.027) -.006 (.129) .676 2245

.009*** (.002) -.002 (.302) -.006** (.020) -.011*** (.002) -.009** (.027) .603 2109

-.015

-.016

-.013

(.246)

(.232)

(.167)

.256

.362

.495

.004 (.208) -.004* (.063) -.003 (.420) -.006 (.126) -.005 (.307) .461

1012

884

766

2360

.013**

.012*

.009**

(.031)

(.060)

(.045)

.004

.006

.008

(.603)

(.437)

(.195)

-.015

-.012

-.007

(.142)

(.235)

(.157)

.006

.009**

(.262)

(.071)

.002

.004

(.588)

(.263)

-.006* (.078) .008*** (.007) .009*** (.009) .015*** (.000)

.010** (.040) .007* (.051)

-.003

-.001

(.404)

(.822)

.010*** (.000) .009*** (.002) .019*** (.000)

.009*** (.001) .008*** (.006) .018*** (.000)

.447

.639

.564

1196

1155

1116

27

Table 8 Panel Regressions of FHCs’ Performance on Non-interest Income 2003-2006 The regression specification is: Y = α + β SH it 1 NET ,it + β 2 SH NON ,it + β 3 SH LOAN ,it + γ X it + ei + ut + ε it The firm-level sample includes FHCs that have more than $500 million assets in 2006 and at least four quarters of data over the period 2003-2006. The dependent variable is risk-adjusted return (RAR) of ROE (i.e., RARROE = ROE / SDROE). The standard deviation (SD) of ROE is calculated by 4, 6, or 8 previous quarters, respectively. If SD includes the current year, RARs on equity are denoted by RARROE(4), RARROE(6), and RARROE(8), respectively. SHNET is the share of operating income from net interest income. Xit contains the log of total assets, the equity to assets ratio, the loan to assets ratio, the growth of assets, i.e., (assetst/assetst-1)-1, and the loss to assets ratio, i.e., allowance for loan and lease losses divided by assets. SHNON contains the share of total non-interest incomes for each of the non-interest income flows, denoted by SHfiduciary, SHservice, SHtrading, SHunderwrting, and SHagency, respectively. SHLOAN contains the share of total loans for real estate, commercial and industrial (C&I), and consumer lendings, denoted by SHresl estate, SHC&I, and SHconsumer, respectively. The specification also includes state fixed effects (ei), quarter fixed effect (ut), and the number of quarters the FHC is observed. A p-value is in parentheses below each coefficient. *** Significant at 1 percent; ** Significant at 5 percent; *Significant at 10 percent. Assets > 75th percentile Assets ≤ 25th percentile Subsamples 25th < Assets ≤ 75th ($697.47m) ($5628.38m) RARROE(i) i=4 i=6 i=8 i=4 i=6 i=8 i=4 i=6 i=8 Variable 2.905 3.983 1.628 1.552** 1.314** 2.360*** .630 1.612 1.808 Constant SHNET ln(Assets)

(.585)

(.135)

(.547)

-.006

-.001

.001

(.256)

(.783)

(.865)

-.320 (.137)

Equity / assets Loans / assets Asset growth Loss / assets SHfiduciary SHservice SHtrading SHunderwrting SHagency SHreal estate loan SHC&I loan SHconsumer loan Adjusted R2 Observations

-.442** (.023)

-.358* (.068)

.027

-.008

-.031

(.237)

(.682)

(.127)

.017***

.014**

.011***

(.000)

(.000)

(.003)

.010

.001

.004

(.108)

(.799)

(.493)

.231

.207

(.327)

(.316)

.374* (.068)

-.003

-.002

.000

(.253)

(.510)

(.922)

(.019) -.009*** (.000)

(.039) -.008*** (.001)

.045

.048

(.209)

(.147)

-.022*** (.000)

-.037*** (.000)

(.000) -.008*** (.000) .063** (.042) -.058*** (.000)

.000

.002

.003

(.824)

(.308)

(.131)

-.011*** (.000) .284*** (.001) -.006*** (.005)

-.011***

-.008***

-.017** (.021) .008* (.067) -.012***

(.000)

(.155) -.017*** (.000)

.035

.009

(.348)

(.801)

-.019*** (.009) .009** (.023) -.010**

-.020*** (.006) .005 (.181) -.004

(.029)

(.039)

.007

-.180

-.157

-.185

(.684)

(.932)

(.347)

(.383)

(.297)

(.390)

-.002

.000

.000

.004

.005

(.263)

(.829)

(.923)

(.233)

(.102)

.002

.000

.000

.002

(.921)

(.781)

(.123)

.003** (.010)

.007

.003

-.004

-.003

-.001

-.001

(.427)

(.707)

(.607)

(.303)

(.711)

(.743)

-.011

-.010

-.004

(.125)

(.191)

(.604)

(.000)

.002

.003

.003

.001

(.529)

(.224)

(.280)

(.589)

(.016)

.004

.012**

.075* (.063)

-.019***

(.000)

(.401)

.008

(.000)

(.174)

.033

.001

.012*

-.017***

(.000)

(.728)

.010***

(.000)

.006

.000

-.005

(.243)

(.932)

(.291)

.018***

.012

.019**

(.004)

(.102)

(.008)

-.001

-.002

.002

(.853)

(.627)

(.670)

.000

.001

.001

(.000)

(.949)

(.799)

(.673)

-.001

.000

.000

.005

.014*** (.000) .004**

.014*** (.000) .006***

.008**

(.064)

(.144)

(.030)

(.400)

(.842)

(.962)

(.885)

(.113)

.001

-.002

-.003

.004

.000

.002

.003

.000

-.002

(.845)

(.745)

(.637)

(.421)

(.997)

(.626)

(.556)

(.962)

(.648)

.016***

(.015)

.014

.007

.011

-.002

-.004

-.003

.007

(.151)

(.432)

(.199)

(.744)

(.477)

(.508)

(.137)

(.004)

(.001)

.017***

.524

.713

.672

.597

.747

.704

.467

.642

.574

818

789

758

1702

1631

1539

847

822

793

28

References Adkisson, J. A. and D. R. Fraser, 2003, The ‘first filers’: An examination of the first financial holding companies, Journal of International Banking Regulation 4: 328-336. Akhavein, Jalal D., Allen N. Berger, and David B. Humphrey, 1997, The effects of mega-mergers on efficiency and prices, working Papers -- US Federal Reserve Board's Finance & Economic Discussion Series. Akhigbe, A. and A.M. Whyte, 2004, The Gramm-Leach-Bliley Act of 1999: risk implications for the financial services industry, Journal of Financial Research 27: 435-446. Akhigbe, A. and A.M. Whyte, 2001, The market’s assessment of Financial Services Modernization Act of 1999, The Financial Review 36: 119–38. Aggarwal, R.K. and A.A. Samwick, 2003, Why do managers diversify heir firms? Agency reconsidered, The Journal of Finance 58: 71–118. Barth, J.R., R.D. Brumbaugh Jr., and J.A. Wilcox, 2000, The repeal of Glass-Steagall and the advent of broad banking, The Journal of Economic Perspectives 14: 191-204. Berger, A. N., J. D. Cummins, M. A. Weiss, and H. Zi, 2000, Conglomeration versus strategic focus: evidence from the insurance industry, Journal of Financial Intermediation 9: 323-362. Berger, A. N., Rebecca S. Demsetz, and Philip E. Strahan, 1999, The consolidation of the financial services industry: Causes, consequences, and implications for the future, Journal of Banking & Finance 23: 135194. Berger, A. N., and Loretta J. Mester, 1997, Inside the black box: what explains differences in the efficiencies of financial institutions?, Journal of Banking & Finance 21: 895-947. Berger, A. N., D. B Humphrey, and L. B. Pulley, 1996, Do consumers pay for one-stop banking? Evidence from an alternative revenue function, Journal of Banking and Finance 20: 1601–1621. Berger, P. G., and E. Ofek, 1995, Diversification’s effect on firm value, Journal of Financial Economics 37: 39-65. Boyd, John H., Stanley L.Graham, and R. Shawn Hewitt, 1993, Bank holding company mergers with nonblank financial firms: effects on the risk of failure, Journal of Banking & Finance 17: 43-63. Campa, J.M. and S. Kedia, 2002, Explaining the diversification discount, The Journal of Finance LVII (4): 1731–1762. Carow, Kenneth A., 2001a, The wealth effects of allowing bank entry into the insurance industry, The Journal of Risk and Insurance 68: 129-150. Carow, Kenneth A.,2001b, Citicorp-Travelers Group merger: Challenging barriers between banking and insurance, Journal of Banking and Finance 25: 1553-1571. Carow, Kenneth A.,and Randall A. Heron, 2002, Capital market reactions to the passage of the Financial Services Modernization Act of 1999, Quarterly Review of Economics & Finance 42: 465-485. Cheng, David, Benton E. Gup, and Larry D. Wall, 1989, Financial determinants of bank takeovers, Journal of Money, Credit, and Banking 21: 524-536. Cocheo, Steve, 2000, Financial holding companies debut- and they’re not all giants, ABA Banking Journal 92, Jun, Issue 6. Cornett, M. M. and H. Tehranian, 1992, Changes in corporate performance associated with bank acquisition, Journal of Financial Economics 31: 211-234. Cornett, M. M., E. Ors, and H. Tehranian, 2002, Bank performance around the introduction of a Section 20 Subsidiary, Journal of Finance 57: 501–522. Cummins, J. David, Sharon Tennyson, and Mary A. Weiss, 1999, Consolidation and efficiency in the US life insurance industry, Journal of Banking & Finance 23: 325-357. Cummins, J. David, and Mary A. Weiss, 2000, Analyzing Firm Performance in the Insurance Industry Using Frontier Efficiency and Productivity methods, in Handbook of Insurance, ed. Georges Dionne, Boston, MA: Kluwer Academic Publishers. Cummins, J. David, and Xiaoying Xie, 2005, Efficiency and Value Creation in Acquisitions and Divestitures: Evidence from the US Property-Liability Insurance Industry, working paper, the Wharton School.

29

Delong, Gayle L., 2001, Stockholder gains from focusing versus diversifying bank mergers, Journal of Financial Economics 59: 221-252. Ferrier, G., S. Grosskopf, K. Hayes, and S.Yaisawarng, 1993, Economies of diversification in the banking industry: A frontier approach, Journal of Monetary Economics (31): 229–249. Fluck, Z., and A.W. Lynch, 1999, Why do firms merge and then divest? A theory of financial synergy, Journal of Business (72): 319–346. Kwan, Simon H and Elizabeth S. Laderman, 1999, On the portfolio effects of financial convergence-a review of the literature, Economic Review – Federal Reserve Bank of San Francisco 2: 18-31. Holyoke, Thomas T., 2002, Expanding the limits of power: the Federal Reserve and the implementation of functional regulation in the Gramm-Leach-Bliley Era, Review of Policy Research 19. Houston, J.F., C.M. James, and M.D. Ryngaert, 2001, Where do merger gains come from? bank mergers from the perspective of insiders and outsiders, Journal of Financial Economics 60: 285-331. Lamont, O.A., and C. Polk, 2001, The diversification discount: cash flows versus returns, The Journal of Finance LVI (5): 1693–1721. Lewellen, W. G., 1971, A pure financial rationale for conglomerate merger,” Journal of Finance (26): 521– 537. Lown, Cara S., Carol Osler, Philip E.Stahan, and Amir Sufi, 2000, The changing landscape of the financial services industry: What lies ahead?, Economic Policy Review – Federal Reserve Bank of New York 6: 3954. Mamun, Abdullah, M. Kabir Hassan, and Neal Maroney, 2005, The wealth and risk effects of the GrammLeach-Bliley Act (GLBA) on the US banking industry, Journal of Business Finance & Accounting 32: 351-388. Mansi, Sattar, and David Reeb, 2002, Corporate discount: who get discounted?, Journal of Finance LVII (5): 2167-2183. Mishkin, Frederic S., 1999, Financial consolidation: dangers and opportunities, Journal of Banking & Finance 23, 675-691. Pilloff, Steven J., 1996, Performance changes and shareholder wealth creation associated with mergers of publicly traded banking institutions, Journal of Money, Credit, and Banking 28: 294-310. Peristiani, Stavros, 1997, Do mergers improve the X-efficiency and scale efficiency of U.S. banks? evidence from the 1980s, Journal of Money, Credit, and Banking 29: 326-337. Rajan, R., H. Servaes, and L. Zingales, 2000, The cost of diversity: the diversification discount and inefficient investment, Journal of Finance LV (1): 35-80. Reichert, Alan K. and Larry D. Wall, 2000, The potential for portfolio diversification in financial services, Economic review - Federal Reserve Bank of Atlanta, third quarter. Report to Congress on Financial Holding Companies under the Gramm-Leach-Bliley Act, Nov. 2003. Rhoades, Stephen A., 1998, The efficiency effects of bank mergers: An overview of case studies of nine mergers, Journal of Banking & Finance 22: 273-291. Saunders, A. and D. J. Cornett, 2006, Financial institutions management: a modern perspective, IrwingMcGraw Hill, 5th. Saunders, A. and I. Walter, 1994, Universal banking in the U.S.?, New York: Oxford University Press. Schoar, A., 2002, Effects of corporate diversification on productivity, Journal of Finance LVII (6): 23792403. Servaes, H., 1996, The value of diversification during the conglomerate merger wave, Journal of Finance LI (4): 1201-1225. Shaffer, Sherrill, 1993, Can megamergers improve bank efficiency?, Journal of Banking and Finance 17: 423-436. Skipper, Harold D and W. Jean Kwon, 2007, Risk management and insurance: perspectives in a global economy, Blackwell Publishing, Malden, MA, USA. Stiroh, Kevin J., 2004, Diversification in banking: is non-interest income the answer?, Journal of Money, Credit, and Banking 36: 853-882. Stiroh, Kevin J., and Adrienne Rumble, 2006, The dark side of diversification: The case of US financial holding companies, Journal of Banking & Finance 30: 2131-2161.

30

Thomas, Shawn, 2002, Firm diversification and asymmetric information: evidence from analysts' forecast and earnings announcements, Journal of Financial Economics 64: 373-396. Yeager, Timothy J., Fred C. Yeager, Ellen Harshman, 2007, The Financial Services Modernization Act: Evolution or Revolution?, Journal of Economics and Business 59: 313-339. Yildirim, H. Semih, Seung-Woog (Austin) Kwag, and M. Cary Collins, 2006, An examination of the equity market response to the Gramm-Leach-Bliley Act across commercial banking, investment banking, and insurance firms, Journal of Business Finance & Accounting 33: 629-649.

31