Universal banking deregulation and firms choices of

Transcription

1 Universal banking deregulation and firms choices of lender and equity underwriter I empirically examine whether firms engage in one-stop shopping for loans and equity underwriting, following the relaxation of regulatory restrictions. I find that large rated firms often obtain both these services from the same financial intermediary. Such one-stop shopping remains rare among smaller unrated firms. I show that this difference is because unrated firms borrow from geographically proximate, commercial banking lenders. Conversely, in the nationally integrated equity underwriting market, unrated firms value the intermediary s distribution capabilities more than lending relationships or proximity. The results suggest that universal banking has not improved access of more opaque firms to the equity market This paper is based on chapter 2 of my dissertation at Harvard University. I am grateful to my advisors, Jeremy Stein, Richard Caves, Fritz Foley and Bharat Anand, as well as to Ariel Pakes and George Baker. Seminar participants at Harvard Business School and Harvard University provided useful input. All remaining errors are mine. 1

2 1. Introduction The relaxation of restrictions on commercial bank underwriting in the late 90s removed long-standing barriers that separated the businesses of lending and underwriting in the United States. Academic research on universal banking at the time of the deregulation suggested that the entry of commercial banks into the underwriting market would be particularly beneficial for younger, smaller, lower reputation companies. The belief was that commercial banks would be able to leverage their private information about these companies, to provide them with underwriting services at lower cost (the so- called information-based economies of scope argument, see for instance, Saunders (1985), Puri (1996), Kanatas & Qi (1998) and Calomiris and Karceski (2000)) 2. It was also believed that specialized investment banks would be at a disadvantage against commercial banks, which could now offer both lending and underwriting services to clients. Many mid-sized and small investment banks merged with commercial banks in the years following the deregulation, rather than continue to operate as independent investment banks. Eight years have passed since the restrictions on commercial bank underwriting were removed. The empirical evidence available on the success of commercial banks in the underwriting market presents a mixed picture. On the one hand, commercial banks have made great strides in debt underwriting (Sufi (2004). Commercial banks, with the help of mergers with investment banks, have also been 2 Krozner and Rajan (1994) expressed some skepticism about the extent of these informational scope economies. This was based on their findings using data for the period prior to 1933, that the securities affiliates of commercial banks underwrote safer securities (debt rather than equity) for older, larger, and relatively less indebted firms, relative to comparable investment banks. 2

3 able to convert some of their lending relationships into equity underwriting deals (see for instance, Ljungqvist et al (2006b), Schenone (2004), Bharath et al (2004)). On the other hand, the impact of commercial bank entry into equity underwriting has been strongest for large deals (Ljungqvist et al (2006a)). Moreover, independent investment banks continue to account for a large share of the equity market, and have even gained significant market shares in lending (see Drucker and Puri (2005)). Many of the merged commercial-investment banks have not been able to increase their presence in the equity underwriting market beyond that acquired through the merger 3, even as they hold on to their shares in the lending market. Do public firms that raise loans and issue equity use the same bank for lending and equity underwriting services after the deregulation? Do commercial banks have a natural advantage over independent investment banks in providing these one-stop services? Understanding the answers to these questions becomes important in the context of the international debate on universal banking. It has been recognized that extending the powers of commercial banks by allowing them to do universal banking comes with costs, particularly in less developed countries (see Rajan (2002). In these countries, both the sequence of regulation and the extent of powers given to commercial banks may impact the growth of institutions and public markets. Thus, it is very important to understand whether small, more opaque firms really benefit from universal banking. Secondly, within the US context, there has been much interest recently both about the specialness of commercial banks, and 3 Indeed, some of the mergers have even been reversed. This includes the spin-off of Piper Jaffrey by US Bank in 2003, as well as the closing down of Robertson Stephens by Fleet-Boston in

4 the potential they have for scope expansion into related activities (see Rajan (1998), James and Smith (2002) and Kashyap, Rajan and Stein (2002) for a discussion of some of these issues). Understanding the relative success of commercial banks and investment banks in the lending and underwriting markets has the potential to contribute to this discussion. In this chapter, I hypothesize that firms will differ in the extent to which they engage in one-stop shopping for lending and equity underwriting services. In particular, I argue that public firms that are unrated 4 will continue to use specialized financial intermediaries for lending and equity underwriting, rather than rely on the same bank for both services. Larger rated companies, on the other hand will be more likely to use the same bank or banks for lending and equity underwriting after the deregulation. The argument proceeds as follows. Smaller unrated firms require a great deal of due diligence and monitoring, before they can obtain finance. Lenders to smaller unrated borrowers must invest in obtaining costly information about them, extending credit to them, and monitoring them on an ongoing basis. Traditionally commercial banks have specialized in doing this kind of lending. Previous research has shown that this is likely to be facilitated by geographical proximity and close lending relationships. While lending banks obtain private information about smaller unrated companies, the information is about current operations and liquidation values (Rajan (2002). By contrast, financial intermediaries providing equity underwriting services 4 In this paper, I define unrated companies as those that do not have a public credit rating from S&P or Moody s either in the public bond market as reported by SDC or in the syndicated loan market as reported by Dealscan 4

5 must assess the viability of future cash flows, price the issue (in case of an initial public offering) and market it to investors. The nature of this activity is quite different from that of making loans to smaller unrated companies, and monitoring them on an ongoing basis. The underwriting market is a nationally integrated one. Equity underwriters are therefore likely to be chosen 5 more on the basis of prior underwriting relationships, and distribution network in the industry at a national level. Smaller unrated firms are likely to attach a lot of importance to these factors relative to larger rated companies, since their future cash flows are inherently more uncertain. It can be challenging for a single bank to develop or acquire skills required for both lending and providing equity underwriting services to smaller unrated companies. I argue that these companies may therefore be better served by making a separate choice of lender and equity underwriter. Conversely, I hypothesize that larger rated companies will be more likely to use the same bank for lending and equity underwriting. The activities of lending and underwriting are closely related for these companies, for the following reasons. These companies are well known to investors and are less dependent either on monitoring capabilities of lenders or certification and marketing efforts of underwriters, relative to smaller unrated companies. Second, lenders can syndicate the credit extended to these companies to a dispersed network of institutional investors, to a greater extent than smaller, unrated companies (see Sufi (2007) and 5 Prior research (see Drucker and Puri (2005), Ljungqvist et al (2006a) and (2006b)) has confirmed the importance of underwriting relationships, reputation and research support in determining choice of equity underwriter. Industry expertise has also been recently found to be a determinant of choice of underwriter (see Asker and Ljungqvist (2006)). However, this research has not examined if smaller unrated companies and rated companies differ in importance attached to these factors. 5

6 Dennis and Mullineaux (2000)). In this sense, loans to these companies are similar to public debt. Hence, I argue that commercial banks will be more able to convert lending relationships into winning equity underwriting transactions, and investment banks will be able to leverage on their equity underwriting relationships to provide loans to these companies. The second question I ask in this chapter is: Do commercial banks and investment banks differ in terms of their ability to provide one-stop finance? I contend that among larger rated companies, commercial banks will have an advantage in certain types of loans deriving from their greater ability to provide liquidity insurance. Recent research (Kashyap, Rajan, and Stein (2002) and Strahan and Gatev (2005)) has shown that the access to the deposit market gives commercial banks an advantage over other types of financial intermediaries in making commitment-based loans or lines of credit. This suggests that the ability of commercial banks and investment banks to provide loans to larger rated companies will vary by type of loan. Commercial banks will have an advantage in making commitment-based loans, while they will face greater competition from investment banks in making term loans. I further hypothesize that this advantage will carry forward to the equity underwriting market - commercial banks will be more likely to provide one-stop finance to firms that are relatively heavy users of commitmentbased loans, whereas they will be less likely to provide one-stop services to firms that do not use lines of credit intensively. To test these hypotheses empirically, I assemble a dataset of loans and equity underwriting issues of US industrial public companies over the period

7 (excluding companies in the finance, utility and government sectors). The sample includes loans and equity issues 6 from both larger rated companies and smaller unrated companies. Most of the loans in the sample are syndicated loans. I use the revealed preference of firms in terms of their choice of lead lender and lead equity underwriter after the deregulation, that is, over the period , to provide empirical evidence for the above hypotheses. I conduct two sets of analyses. First, I estimate discrete choice models of lender and equity underwriter choice to test if the determinants of choice vary across larger rated companies and smaller unrated companies, as predicted. The data includes information on over 10,000 loans and 3,000 equity underwriting deals between 1998 and Second, I focus on a smaller set of loans and equity issues by companies that are active in both the lending and equity underwriting markets in the period I estimate a logit model examining whether or not companies use the same bank for loans and equity issues made after 1998, and how this varies by company type. Using these two sets of analyses, I develop five basic pieces of empirical evidence to support the arguments in the chapter. First, I find that in the lending market, smaller unrated companies are more likely to choose geographically proximate lenders relative to larger rated firms (consistent with previous research, 6 This chapter does not examine public debt issues, since smaller unrated companies rarely issue public debt. In unreported results, I have studied the impact of the deregulation on one-stop shopping including the debt underwriting market. Most of the debt issues are from larger, rated companies, who are also heavy users of commitment-based loans. I find that partly because of this, commercial banks have been more successful than investment banks in offering one-stop shopping for loans and debt issues to a large number of these companies. 7

8 for instance see Sufi (2007)). In the equity underwriting market, I show that smaller unrated firms attach greater importance to industry expertise than do larger rated firms. Geographical proximity is not a primary consideration of equity underwriter choice, even for smaller unrated companies. Second, smaller unrated firms are more likely to choose commercial banks as lead arrangers for all type of loans, relative to larger rated companies. This is consistent with the traditional role of commercial banks in providing finance to informationally-opaque firms. The third piece of evidence concerns the importance placed by smaller unrated companies on own and cross-product pre-deregulation relationships in choosing lenders and underwriters after Smaller unrated companies highly value own-product relationships in both the lending and underwriting markets (that is, previous loan relationships are an important determinant of lender choice and previous equity underwriting relationships are the most important determinant of equity underwriter choice). Banks are not able to leverage cross-product relationships to win business from smaller unrated companies in either the lending or equity underwriting market 7. Fourth, I show that larger rated companies are less tied to pre-deregulation own-product relationships, and more likely to use cross-product relationships as an important determinant of choice in both the lending and underwriting markets, 7 Of course, this could partly be because investment banks and commercial banks like Deutsche Bank and Citibank that focus on larger companies are not interested in making loans to smaller, unrated companies. My argument is that commercial banks that do make loans to these companies, like Bank of America, and have underwriting operations are not very successful in converting these lending relationships to winning equity deals. 8

9 relative to smaller, unrated companies. In the lending market, the importance placed on own product and cross product relationships varies by type of loan. Companies raising term loans choose lenders both on the basis of previous underwriting relationships, and previous lending relationships. This suggests that investment banks are able to leverage prior underwriting relationships to extend into lending, when it comes to term loans. By contrast, companies raising lines of credit tend to stay with previous commercial banking lenders in the loan market and equity underwriting relationships do not impact their choice of lender. This is reflected in the equity underwriting market as well. Here, companies that use lines of credit intensively are more likely (than companies that do not) to switch underwriters by choosing pre-deregulation lending banks as equity underwriters. However, companies that do not use lines of credit intensively are more loyal to previous equity underwriters. Finally, using a sample of companies that raise both loans and equity issues after the deregulation, I confirm that overall, smaller unrated firms are less likely to engage in one-stop shopping, than larger rated companies. Greater opacity of the firm, by other measures of information asymmetry used in the literature, increases the likelihood that companies will make a separate choice of lender and equity underwriter. The rest of the chapter proceeds as follows. Section 2 provides some background by describing the deregulation and the response of commercial banks to it. In Section 3, I discuss the hypotheses tested in the paper, and empirical implications. Section 4 discusses the data sources used for the paper and definitions 9

10 of key empirical variables, and presents summary statistics. Section 5 contains results from discrete choice models of lender and underwriter choice. In Section 6, I estimate the logit model to test if a firm uses the same bank for lending and underwriting after the deregulation. In Section 7, I offer some concluding remarks. 2. Background In this section I briefly review the Glass Steagall Act, theoretical and empirical research on universal banking, the gradual relaxation of restrictions on commercial bank underwriting, and the response of commercial banks to the deregulation. Commercial banks were prohibited from being principally engaged in the underwriting of stocks, bonds, and other securities under the Glass-Steagall Act of The Act was originally motivated by fears that the conflicts of interest resulting from commingling of investment and commercial banking functions would harm investors. The fear was that commercial banks would be tempted to bring to the market poor securities to pay off their own loans, concealing or misrepresenting information about these securities. Specialized intermediaries thus carried out investment banking and commercial banking until the 1990s. At this time, increased disintermediation, global competition and increased competition from private finance companies increased pressure on regulators to ease restrictions on banking (see Calomiris and Karceski (2000)). Meanwhile, academic research in the early 1990s (Kroszner and Rajan (1994), Ang and Richardson (1994) and Puri (1996)) comparing the activities 10

11 of commercial banks and investment banks prior to Glass Steagall failed to find any evidence that commercial banks were bringing low quality issues to the market. Instead, there was some evidence that issues underwritten by commercial banks performed better than comparable securities underwritten by investment banks. Theoretical research instead emphasized the benefits to issuers from permitting commercial banks to underwrite securities. It was believed that commercial banks early contact with corporations (Diamond (1984), Puri (1994)) and superior information (James (1987), Rajan (1992)), about the firm and its management obtained through close lending relationships would enable them to offer better underwriting services (Saunders (1985), James and Weir (1990), Kanatas and Qi (1998), Puri (1996)). Specifically, it was believed that these scope economies in information access and processing would give commercial banks an advantage over investment banks in underwriting more "information-intensive" securities, in other words, junior securities and securities of younger, smaller, and less well-known firms. Universal banking would thus particularly benefit small and young firms, providing them more access to the public securities markets than they otherwise would have had, and at lower cost. Under these circumstances, the Fed began to gradually relax restrictions on the activities of commercial banks. In January 1989, it allowed commercial banks to underwrite corporate debt, and shortly thereafter-corporate equity. Banks were required to house these underwriting activities under independent units, or Section 20 subsidiaries, to be approved by the Fed on a case-by-case basis. They were also required to maintain several operating limitations or firewalls between the 11

12 commercial banking units of these banks and the Section 20 subsidiaries. These firewalls limited information transfer, and resource and financial linkages between the parent commercial bank and the Section 20 subsidiary. The Fed also capped the revenue that could be derived from all ineligible securities (defined to include corporate debt and equity as well as some other products) at ten percent of the revenue from eligible securities of the Section 20 affiliate. In December 1996, the Fed further eased the regulations by raising the revenue cap to 25 percent and eliminating certain firewalls. This allowed for much larger levels of underwriting activity, and information sharing and joint operations between the Section 20s and their parent commercial banks. Finally, the Financial Modernization Act of 1999, also known as the Gramm-Leach-Bliley Act, removed most of the regulatory barriers to entry into underwriting by commercial banks. Between 1990 and 1996, many commercial banks entered the underwriting market by setting up Section 20 subsidiaries (see Chaplinsky and Erwin (2005) and Puri (1996)). In this period, their presence was felt mostly in the debt markets and they did not have much success in the equity underwriting market. The constraints on exploiting potential informational scope economies between lending and equity underwriting were significantly reduced by the easing of the revenue caps and removal of the operating firewalls in December, Many commercial banks entered into mergers with investment banks between 1997 and 2002, with a desire to expand their presence in the underwriting market, and provide the full range of financial services to clients. Two of the mergers, between Citibank and Salomon Smith Barney and between JP Morgan and Chase Manhattan, were 12

13 mergers between banks catering to large corporations. The majority of the mergers involved acquisitions of several small and some mid-sized investment banks by middle market commercial banks. 3. Hypotheses and empirical implications 3.1 Overview of the hypothesis The logic of the argument can be illustrated with the help of a simple example. Consider two public companies R and U in the electronics industry in California, and their choices of lender and underwriter after the deregulation. Company U does not have a bond market or loan market credit rating. U is likely to be a smaller, younger company. The amount of public information available about U is limited, and it requires a significant amount of due diligence and monitoring. U therefore is reliant on loans for finance, and is likely to use equity financing primarily for large expansion or growth projects, which are difficult to finance using loan funds 8. Let us now consider U s choice of lender and underwriter after As suggested by previous literature, I hypothesize that U will be likely to obtain loans from a geographically proximate, commercial bank. Let us say that U obtains loans from Bank of America, a commercial bank that has a significant branch presence in California. In the equity underwriting market that is more nationally integrated, my hypothesis suggests that U will be likely to choose an underwriter who is better able to assess its future prospects and has a distribution network in the 8 James (1996) provides some evidence these firms face investment constraints, when it comes to making very large investments. It seems likely that some of these firms will finance these projects by issuing equity, when market conditions permit them to do so. 13

14 electronics industry. Thus when U wants to issue equity, it is likely to choose say, CSFB that has a large market share in providing equity underwriting services to companies in the electronics industry. Above, I have been assuming that U is a company that accesses the loan and equity markets for the first time after the deregulation. Instead, let us now suppose that U had issued loans and equity issues prior to the deregulation. After the deregulation, my hypothesis suggests that U will be likely to stay with its lender and equity underwriter. On the other hand, consider firm R, also from the electronics industry. It is publicly rated and is likely to be larger, well established and known to investors. Suppose that R has a prior equity underwriting relationship with CSFB. Suppose also that it has a prior lending relationship with Bank of America. Consider its choice of lender and equity underwriter after the deregulation. My hypothesis suggests that Firm R will now be prepared to borrow from its existing equity underwriter, CSFB, since it is less sensitive to geographical proximity while obtaining loans. R will be willing to switch to using its existing lender Bank of America as its equity underwriter. R may even develop a one-stop lending and equity underwriting relationship with another bank, say Citigroup. This is because R places less emphasis on either industry expertise or prior equity underwriting relationships in the underwriting market, relative to U. Overall, after the deregulation firm R will be more likely to use the same bank for lending and equity underwriting, relative to U. 14

15 The second part of the hypothesis on commercial banks advantage in liquidity provision implies that firm R will be more likely to use a commercial bank as lender if the loan is a line of credit. On the other hand, if the loan is a term loan R may use either its existing lender Bank of America or its existing equity underwriter CSFB. Further, if firm R is a heavy user of lines of credit, my hypothesis suggests that R will be more likely to use its commercial bank for equity underwriting as well. Conversely, if a larger share of firm R s loans is term loans; it will be more likely to use an investment bank as its one-stop bank. I discuss the testable implications of the hypothesis below. 3.2 Testable Implications Company type and preference for geographical proximity of lenders and national industry expertise of underwriters The hypothesis implies that smaller unrated firms will prefer to borrow from geographically proximate lenders, while choosing underwriters based on distribution network and expertise in the issuer s industry at a national level. By contrast, larger, unrated firms will be less sensitive both to geographical proximity of lenders and industry expertise of underwriters. The role of geographical proximity is familiar from previous research on the loan market (see Petersen and Rajan (2002) and Sufi (2007)). Petersen and Rajan (2002) also find that the increased availability of information technology methods has enabled small businesses to borrow at greater distances. My sample focuses only on public companies. It is thus possible, that with the information revolution in 15

16 lending, larger rated and smaller unrated public firms may not differ very significantly in their preference for geographically proximate lenders. Most academic research on the underwriting market (with the exception of Asker and Ljungqvist (2006)) has not explicitly considered industry expertise as a determinant of equity underwriter choice. This is somewhat at odds with the fact that in practice, personnel within investment banks are organized into industry groups and many mid-sized investment banks specialize by industry Company type and likelihood of choosing a commercial bank lender Commercial banks have traditionally specialized in providing loans to opaque companies. Hence, my hypothesis suggests that smaller unrated companies will be more likely to choose a commercial bank lender for all kinds of loans, even after controlling for branch network and relationships. On the other hand, commercial banks primarily act as liquidity providers to larger rated companies. Hence these companies will have a preference for commercial banks over investment banks only if the loan they are seeking is a line of credit Smaller unrated companies and importance of pre-deregulation relationships There is large literature in finance showing that lending relationships are more valuable to opaque companies (see for instance, Petersen and Rajan (1995)). Hence I hypothesize that smaller unrated firms will be more loyal to existing lenders in the loan market. My hypothesis also suggests that in the equity underwriting market, smaller unrated firms that have more uncertain future cash flows will be more loyal to existing underwriters (relative to larger rated companies). This also implies that they 16

17 will be less likely to switch to previous lending banks. In testing empirically whether previous lending relationships help win equity deals, it is important to focus only on commercial banks that have an equity underwriting capability. This is because a large number of lending relationships of smaller unrated companies are with regional commercial banks that do not participate in the underwriting market Larger rated companies and importance of pre-deregulation relationships The hypothesis suggests that larger rated companies will be less loyal to own-product relationships in both the lending and equity underwriting markets, compared to smaller unrated companies. Further, I hypothesize that banks should also be able to leverage cross-product relationships to gain new business from these companies. Previous research (Kashyap, Rajan, and Stein (2002) and Strahan and Gatev (2005)) has identified that commercial banks have an advantage over other types of financial intermediaries in providing commitment-based loans. This research also suggests that any synergy commercial banks have in making term loans is indirect 9. This suggests that the relative success of commercial banks and investment banks will depend on the type of loan. Specifically, the hypothesis suggests that in the loan market 1) Larger, rated companies will be more likely to continue previous lending relationships with commercial banks when the loan is a line of credit, 2) Companies will be more likely to switch to investment banks in the lending market if the loan deal includes a term loan, and in the equity market 3) Commercial banks will be 9 Another recent paper has also found evidence consistent with this theory (see Wittenberg-Moerman (2006)) 17

18 able to leverage previous loan relationships to provide equity underwriting services to larger, rated companies that are heavy users of lines of credit and 4) Investment banks will be able to retain previous underwriting clients that use term loans more heavily Company type and propensity to engage in one-stop shopping The first four empirical implications focus on the determinants of choice of lender and underwriter and how these vary across larger rated companies and smaller unrated companies. The hypothesis also implies that this variance in choice determinants will impact the extent to which firms engage in one-stop shopping for lending and equity underwriting after the deregulation. Among firms that both raise loans and issue equity in the period after the deregulation, larger rated firms should be more likely than smaller unrated companies to use the same bank for the loan and the equity issue. 3.3 Alternative theories The hypothesis and empirical implications I test in this paper basically suggest that synergies between lending and equity underwriting for smaller unrated companies are not very high. It is possible that these firms choice of lenders and equity underwriter are motivated instead by concerns about conflict of interest, as described in Section 1. Some papers, including Krozner and Rajan (1994), suggest that fears about conflict of interest compel commercial banks to confine themselves to underwriting safe issues. This could also be a factor in explaining my findings, namely that commercial banks have greater success in underwriting equity issues of larger rated companies. However, the basic conclusion in the paper would be the 18

19 same, namely that net benefits from information-based scope economies are not large enough to persuade smaller, unrated companies to choose commercial banking lenders for providing equity underwriting services. 4. Data 4.1 Data sources The primary data in this paper comes from two sources - Dealscan for data on loans and SDC for data on all equity underwritten issues. I focus on transactions of US-based industrial firms, that is, I leave out companies in the financial, real estate, utilities and government sector. I focus on loans and equity issues made by publicly traded companies 10, during the period From Dealscan I obtain data on all loan transactions and banks identified as Agent or Lead Arranger for these transactions during the period I exclude transactions where this role is unidentified or has more than four banks identified as such. I use the loan deal as the unit of analysis, and calculate the size of the loan and share of each lead bank at the deal level. Each loan deal can contain multiple tranches, with varying tranche amounts and maturities. However, a deallevel analysis is appropriate because the lead bank(s) is chosen and the terms are negotiated at a deal level (see Sufi (2007)), and usually the same lead bank(s) is involved in all the tranches. The data for the entire period includes 22,239 loan deals, by 6,064 US industrial public companies, employing the services of 953 different lead arrangers. 10 I identify public firms using the corresponding variable in Dealscan. In addition, I classify any company that has issued debt or equity (including an initial public offering) during the period as a public company. 19

20 SDC provides data on equity issues. I define all equity issues made by a company on the same day as one deal in the analysis. SDC contains 8,414 equity underwriting deals over the period by 5,349 industrial, public US companies. There are over 400 different underwriters involved in the lead manager role in these deals. Of these 5,349 companies, I am able to hand-match 3,373 companies to Dealscan. I am also able to match 4,028 companies of the 5,349 companies from SDC and 4,150 of the 6,064 companies from Dealscan to Compustat to obtain information on assets, sales, and other financial ratios. I also use Compustat and Hoovers database to keep track of changes in company name changes and company mergers over time. Further, I match the firms in SDC to IBES to obtain information about the research coverage of banks and the industries covered by them. Finally, I evaluate the quality of the research coverage provided by different banks by obtaining analysts rankings from the Institutional Investor Magazine in each year. This is because analyst research has been shown in previous research to be an important determinant of equity underwriter choice (see Drucker and Puri (2005) and Ljungqvist et al (2006a and b)). Many commercial banks and investment banks undergo mergers during this period. I obtain information on merger events involving commercial banks from FDIC and NIC, involving investment banks from the Security Information Database and involving non-bank finance companies from web sources. Information on the location of branches of commercial banks in each state and year is aggregated from FDIC (I use web sources to obtain current data on offices of investment banks and 20

21 finance companies, and assume that these offices existed in all the years). All financial institutions are aggregated to their parent company and inherit the characteristics of the aggregated company. I am interested in studying the impact of relaxation of the restrictions on commercial bank participation in the underwriting market on firms choice of lender and equity underwriter. As Section 3 describes, the deregulation and the entry of commercial banks into underwriting occurred gradually over But, commercial banks had not developed a significant presence in the equity underwriting market until The change in regulations in December 1996 and the subsequent mergers between commercial banks and investment banks were particularly significant events in the deregulation process. Further, most of the initial mergers occurred in late With this in mind, I treat the period as the period of interest for conducting the empirical analysis, and use data over the period to define pre-deregulation relationships. 4.2 Key empirical variables Here, I discuss the definitions of key variables used in subsequent analyses to test the empirical predictions outlined in the previous section. In defining the variables, I take into account all the mergers between banks in the sample; all financial institutions are aggregated to their parent company and inherit the characteristics of the aggregated company, using the effective date of the merger (geographic location, total assets, relationships, industry expertise etc.). 21

22 A critical variable RATED (RATED=1 if the company has either a bond market or a syndicated loan market rating, and is 0 otherwise) indicates the availability of a credit rating. Firms that do not have either a public bond market rating or a syndicated loan market rating from S&P or Moody s are categorized as unrated firms 11. The same company may thus be categorized as rated or unrated at different points over time. In practice, I combine information on ratings from two sources public bond market rating of the company from SDC (in the same year as or one year before or after the loan or equity issue) and syndicated loan market rating from Dealscan (again, in the same year as or one year before or after the loan or equity issue). Another important empirical measure is the classification of banks as commercial banks (CB) or investment banks (IB). In this paper, I use the term commercial bank to mean a federally regulated commercial bank that has domestic branches, and access to the domestic deposit network. In some specifications, I further categorize commercial banks into pure commercial banks (PCB) that do not participate in the underwriting market and universal banks (CBIB) that participate in both markets. I classify banks whose commercial banking operations are mostly outside the US, such as Credit Suisse First Boston and Union Bank of Switzerland, as investment banks. I also treat banks like Merrill Lynch (that has recently set up a 11 See Kisgen (2004) and Petersen and Faulkender (2006) for a detailed discussion of public bond market ratings. These papers show that obtaining a rating enables a firm to become better known to the public. Sufi (2006) describes in detail the introduction of syndicated loan market ratings in He shows that obtaining a loan rating expanded financing choices for bank borrowers that were too small to access the public bond markets. 22

23 local commercial banking subsidiary, but continues to primarily be an investment bank) as investment banks. Geographical proximity of the bank to a borrower is another key variable in the analysis. Dealscan and SDC contain information on the state in which the firm is located, but do not provide any further detail. I match this information with data on the branch network of banks in each state from FDIC to measure geographical proximity of the bank under consideration and the firm. I use two measures of proximity - 1) MANY_BRANCHES_IN_STATE, an indicator variable measuring how widespread the branch network of a bank in the company s state is. It takes the value 1 if a bank has more than 20 branches in a given state, and is 0 otherwise 2) BRANCH_IN_STATE that takes the value 1 if the bank under consideration has between 1 and 20 branches in the state, and is 0 otherwise. The choice of 20 branches is arbitrary, and the results are robust to using a different number. Typically, the large investment banks and finance companies have an office in most of the major cities in the US. But only commercial banks have extensive branch networks covering entire states. I hypothesize that equity underwriting expertise of each bank in the industry of the company INDEXP is an important determinant of choice of underwriter, particularly for unrated companies. SDC and Dealscan contain data on primary 4- digit SIC codes of the companies. To arrive at INDEXP, I first classify the companies into 40 different industries. The classification is largely 12 based on the 12 I modify the Fama-French industry classification for some sic codes to align it with definitions of industry used by IBES and Institutional Investor 23

24 categorization used by Fama and French (1997). I then define INDEXP, a measure of equity underwriting expertise of each bank in each of these industries, as the ratio of market share of the bank in equity issues from that industry relative to the market share of the leading equity underwriter in issues from that industry over the period Own and cross product relationships are also important in determining choice. My intention is examine to what extent relationships established before 1998, are useful in determining choice of lender and equity underwriter during the sample period. For this purpose, I use data for the period to determine the share of each bank by amount in all the loans (equity underwriting issues) made by each company over this period. I use 8 years of data to measure prior relationships, rather than use a shorter time period, because smaller unrated companies issue equity (and to a lesser extent, loans) quite infrequently. This leads to measures of loan relationships PRELOANREL and prior underwriting relationships PREUNDREL. I am also interested in examining whether the choice of commercial bank or investment bank as lender varies by type of loan. Previous research (see James, and Smith, 2000) has shown that many of the loans obtained by larger companies tend to be lines of credit. This is also true in my data. However, several loan deals include other types of loans in addition to lines of credit. In this paper, I define a loan deal as a line of credit (LC=1/NOTLC=0) if the loan deal contains only a line of credit. Any loan deal that contains a term loan (or unknown type of loan) by itself or in conjunction with a line of credit is treated as other loan (NOTLC=1/LC=0). I then 24

25 include interactions of loan type LC/NOTLC with bank type CB. I also classify companies on the basis of the composition of loans they make over the period Companies are classified as heavy users of lines of credit if over 50% of the loans they make over are lines of credit (indicated by dummy variable HIGHLC=1). They are classified as less intensive users of lines of credit (HIGHLC=0) if less than 50% of the loans made over are lines of credit. Again the choice of 50% is arbitrary, and results are qualitatively similar if I use a higher percentage. I also include in the analysis on lender choice a variable LOANMARKETSHARE that measures the overall market share of the bank in the loan market over the three years prior to the year of the loan. To determine LOANMARKETSHARE, I measure market share of each bank separately for rated and unrated companies. This is because there is significant variation in the market shares of banks in these two segments. Further, I make a reasonable assumption that a rated (unrated) company primarily cares about prior market experience of the bank in providing loans to other rated (unrated) companies. Similarly, in the analysis on equity underwriter choice, I include a variable EQUITYMARKETSHARE that measures the market share of each bank in the equity underwriting market over the three years prior to the year of the issue. I also include in the equity underwriter choice models, measures of analysts research and quality. Specifically, I focus on whether the bank under consideration has analysts in place researching the industry affirm belongs to, and the quality of this coverage. For each equity issue, ANALCOV is an indicator variable that takes the value 1 if 25

26 the bank has analysts that cover any other company in the industry in the 12 months prior to the date of the equity issue, and is 0 otherwise. ALLSTAR is an indicator value that takes the value 1 if Institutional Investor Magazine has ranked the analysts covering the industry as All-Star in the year prior to the year of the issue, and is 0 otherwise. Other variables of interest include firm sales in $million, firm assets in $million, size of the loan in $million, maturity of the issue, size of the equity issue in $million, ratio of research and development expenses of the firm to assets and ratio of market value to book value of the firm. 4.3 Summary statistics Table 1 provides summary statistics on the companies and loan deals in the sample during the period In the table, I show summary statistics separately for 1) all the rated and unrated companies that have raised loans in this period and 2) companies that have raised not only loans, but also equity issues between 1998 and Overall, there are 11,132 loan deals by 3,968 companies, in this period. Of these 3,968 companies, 1,192 companies also issue equity. Rated firms account for 53% of the loan deals in the sample. On average, public rated firms are much larger in terms of sales and assets than public unrated firms. Their loans are also larger in size and have longer maturities. They are also more likely to use more than 1 lead arranger for the loan. The table also indicates that (as shown in Sufi (2007)), lead arrangers retain a larger share of the loan for unrated borrowers, and make more sole lender loans to 26

27 them. The table shows that the composition of loans is somewhat different across rated and unrated companies; 62% of loans for unrated companies that also issue equity are of type LC, while the corresponding number for rated companies is 51%. 27

28 Table 1 : Summary statistics on loan deals The table contains summary statistics on loan deals raised during the period Panel A contains information about borrowing companies. Panel B contains information about the individual loan deals In (1), I compare variables across rated and unrated loan deals for all companies active in the loan market between In (2), the set of companies considered is restricted to those that issue equity deals between 1998 and In both (1) and (2), I compare different variables for rated and unrated loan deals, where: - Rated =1 if the company has either a public bond market rating or a syndicated loan market rating from S&P or Moody's at the time of the loan deal - Unrated =1 if the company lacks both ratings at the time of the loan deal Panel A: Company characteristics of companies making loan deals between Companies that issue equity Variable All companies (1) deals (2) Total Rated Unrated Total Rated Unrated Number of companies 3,968 1,485 2,518 1, Number of companies for which Compustat information is available 2,951 1,162 1,816 1, Sales ($mn) - median 1,066 2, , Sales ($mn) - mean 5,111 7, ,767 5, Sales ($mn) - standard deviation 14,147 17,000 3,252 8,848 9,941 4,679 Assets ($mn) - median 1,128 2, , Assets ($mn) - mean 889 9, ,324 5, Assets ($mn) - standard deviation 3,252 29,549 3,532 10,202 11,502 5,104 Panel B: Loan characteristics Variable All companies (1) Companies that issue equity deals (2) Total Rated Unrated Total Rated Unrated Total number of loan deals ,132 5,961 5,171 3,735 2,211 1,524 Size of loan ($mn) - median Size of loan ($mn) - mean Size of loan ($mn) - standard deviation 923 1, , Maturity (months) - median Maturity (months) - mean Maturity (months) - standard deviation Number of lead arrangers - median Number of lead arrangers - mean Amount kept by lead arranger - No. of deals for which information is available 3,095 1,715 1, Median (%) 18% 14% 29% 18% 14% 29% -Mean (%) 26% 20% 39% 25% 21% 35% - Standard deviation (%) 23% 17% 27% 21% 19% 23% Total number of loan deals ,132 5,961 5,171 3,735 2,211 1,524 - Loan deals with pre-deregulation loan 67% 75% 57% 59% 71% 42% - Loan deals that are syndicated loans 92% 99% 84% 93% 99% 84% - Loan deals by top 91 lead arrangers 94% 98% 90% 95% 98% 91% Type of loan deal - Line of credit only (%) 60% 59% 62% 56% 51% 62% - Other loan - Term loan and line of credit(%) 23% 20% 25% 24% 24% 22% - Term loan only (%) 9% 9% 8% 10% 11% 9% - Unknown (%) 8% 11% 5% 11% 13% 7% 28

29 Table 2 provides summary statistics on the equity underwriting market during During this period there are 3,168 equity issues by 2,205 companies. Around 40% of the equity issues are from rated companies; these issues are much larger in size. Unrated companies account for around 60% of the equity issues. Many of the unrated issues are from firms that do not have any bank loans. Consistent with the hypothesis, the ratio of the size of the issue to assets of the company is much larger for unrated companies, suggesting that the difference between current and future cash flows is magnified for these companies 13. The table also shows that unrated companies have a higher ratio of research and development expenses 14 to assets, suggesting that their cash flows are more uncertain. 13 In a recent paper, Chang, Dasgupta and Hilary (2006) show that companies that are followed by fewer analysts, make large, infrequent equity issues, when the market conditions are favorable. This factor may also explain why equity issues of smaller, unrated companies are large, relative to assets. 14 This is consistent with the industry distribution of the issues from rated and unrated companies. Most of the issues by unrated firms are from growth sectors such as the high-tech and life sciences industries, while issues from rated companies are mostly from older, established industries. 29

30 Table 2 : Summary statistics on equity underwriting deals The table contains summary statistics on equity issues raised during the period Panel A contains information about issuing companies. Panel B contains information about the individual equity issues. I compare different variables for rated and unrated equity issues, where: - Rated=1 if the company has either a public bond market rating or a syndicated loan market rating from S&P or Moody's at the time of the loan deal - Unrated =1 if the company lacks both ratings at the time of the equity deal Many unrated companies raising equity do not have any bank loans. So I further classify unrated companies into: - those that have raised at least one bank loan between (2) - and those that do not raise loans over the entire period (3). SEO refers to a seasoned equity issue, while IPO refers to an initial public offering. Panel A: Company characteristics of companies raising equity Total Rated Variable (1) Unrated Has bank No bank loans loans (2) (3) Number of companies issuing equity 2, Number of companies for which Compustat information is available 1, Sales ($mn) - median Sales ($mn) - mean 1,077 2, Sales ($mn) - standard deviation 4,310 7,336 1,073 1,013 Assets ($mn) - median 182 1, Assets ($mn) - mean 1,438 4, Assets ($mn) - standard deviation 13,494 24, ,810 Panel B: Equity issue characteristics SEO IPO Total Rated Unrated Total Rated Unrated Variable Has bank loans No bank loans Has bank loans No bank loans (1) (2) (3) (1) (2) (3) Total number of issues , , Size of issue ($mn) - median Size of issue ($mn) - mean 1,404 3, Size of issue ($mn) - standard deviation 5,040 7,621 1, , ,155 Ratio of issue size to assets - median Ratio of issue size to assets - mean Ratio of issue size to assets - standard deviation Number of deals for which information on is available 1, Ratio of R&D expenses to assets - median Ratio of R&D expenses to assets - mean Ratio of R&D expenses to assets - standard deviation Number of equity deals from companies with prederegulation equity relationships Underwriting deals with more than 1 lead underwriter Number of loan deals by top 84 underwriters % 51% 48% 30% % 37% 31% 40% 28% 22% 12% 40% 1, , % 99% 97% 93% 92% 100% 95% 89% 30