TRUST-PREFERRED SECURITIES AND INSURER FINANCING DECISIONS

Transcription

1 2016 The Journal of Risk and Insurance. Vol. 85, No. 1, (2018). DOI: /jori TRUST-PREFERRED SECURITIES AND INSURER FINANCING DECISIONS James I. Hilliard Steven W. Pottier Jianren Xu ABSTRACT We analyze insurance holding company (IHC) issuance of trust-preferred securities (TPS) from 1994 to We find that larger and more financially levered IHCs issued TPS in 1996 and 1997, as well as those that obtained financial strength ratings from A.M. Best. Abnormal stock price returns are positively related to financial distress costs, growth opportunities, and tax burden, but negatively related to size. Consistent with the pecking order theory, intent to use TPS proceeds to retire debt is positively related to abnormal stock returns, whereas intent to use proceeds to retire preferred equity is negatively related to abnormal stock returns. INTRODUCTION In October 1993, Goldman Sachs introduced the first debt equity hybrid security, known as monthly income preferred stock (MIPS). Since then, other similar hybrid securities have been issued under various names, including quarterly income preferred securities (QUIPS) and trust-preferred securities (TPS). 1 An October 21, 1996 Federal Reserve Board (Fed) ruling that TPS qualified as core (Tier 1) regulatory capital led to more bank-related issues. Although insurance holding companies (IHCs) had first issued debt equity hybrid securities in November 1993, they increased their hybrid securities issuance activity following the October 1996 Fed s ruling. TPS have become an important source of capital for banks and insurers alike. Indeed, by 2008 banks had issued approximately $149 billion and insurers had issued approximately $34 billion in TPS. The period represents the peak 2 years James I. Hilliard is at the W. A. Franke College of Business, Northern Arizona University, and a Associate Fellow, Baugh Center for Entrepreneurship and Free Enterprise, Baylor University. Hilliard can be contacted via jim.hilliard@nau.edu. Steven W. Pottier is at the Terry College of Business, University of Georgia. Pottier can be contacted via spottier@uga.edu. Jianren Xu is at the Department of Finance, Mihaylo College of Business and Economics, California State University, Fullerton. Xu can be contacted via jrxu@fullerton.edu. 1 The terms TPS, debt equity hybrid securities, and hybrid securities are used interchangeably throughout this study. 219

2 220 THE JOURNAL OF RISK AND INSURANCE for TPS issues by IHCs with over $14 billion in issues and is the only 2-year period that TPS issues by IHCs exceeded common and preferred stock issues (see Table 1). However, in an effort to strengthen bank balance sheets following the 2008 financial crisis, TPS has been phased out as Tier 1 regulatory capital for larger bank holding companies (BHCs) under the Dodd Frank Financial Reform Act of 2010 (Balla, Cole, and Robinson, 2011). Engel, Erickson, and Maydew (1999) study the general impact of hybrid securities and Benston et al. (2003) study their effect in the banking industry. The present study is the first to extend the inquiry to the insurance industry. Three key differences between banks and IHCs that motivate a specific study for IHCs include the nature of regulation, capital sources, and volatility of cash flows. With respect to regulation, although banks are regulated by the federal government at the holding company level, insurance capital regulation is typically enforced by states at the operating subsidiary (company) level. Furthermore, as state regulators do not explicitly consider holding company capital structure in their regulatory risk-based capital models, private credit rating agencies play an important role in evaluating IHCs. Standard & Poor s (2002) states that insurance groups issue hybrid capital securities to manage economic capital and satisfy constituents other than regulators (including rating agencies like Standard & Poor s). Consequently, rating agency treatment of TPS as equity may substitute for or intensify regulatory action in practice. In addition, capital sources are an important difference between banks and insurers. Deposits are a bank s major source of capital (rather than debt or equity) and loss reserves are a major source of capital for insurers. Finally, differences in how insurance company liabilities arise (stochastically) relative to how BHC liabilities arise (relatively stable over time) provide the third distinction between the two industries that makes this study a contribution to our knowledge about TPS and regulatory spillover in general. This study examines firm-specific features related to an IHC s decision to issue debt equity hybrid securities and the stock market responses for firms that issued them. We examine the determinants of TPS issuance with a logistic regression and find that larger and more financially levered firms are more likely to issue, as are firms that are rated (as opposed to nonrated firms) by A.M. Best Company. The result with respect to the rating variable implies that ratings agency treatment of hybrid securities is important for insurers, even though their issuance does not directly affect regulatory capital. Our event study results suggest that the market generally rewards issuers, in contrast to prior studies on market reactions to conventional equity offerings (see, e.g., Spiess and Affleck-Graves, 1995; Loughran and Ritter, 1997; Born and McWilliams, 1997). IHCs that issued TPS in 1996 and 1997, on average, experienced positive abnormal returns around the date of the Fed s ruling (October 21, 1996) and on each TPS issuer s announcement date. Abnormal returns around the ruling date reflect investor reaction to the potential change in rating agents treatment of TPS capital, and abnormal returns around each firm s announcement date reflect investor reaction to announcements of a planned or actual issue, that is, to expected capital structure changes.

3 TRUST-PREFERRED SECURITIES AND INSURER FINANCING 221 TABLE 1 TPS and Other Financing Aggregates for U.S. Publicly Traded Insurers ( ) ($ Millions) Year TPS LTD EQUITY TPS/LTD (%) TPS/(LTD þ EQUITY) (%) , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , Total 34, ,465, , Note: TPS represents trust preferred securities, LTD denotes long-term debt, and EQUITY is common and preferred equity. This table reports the size of issuances of TPS, LTD, and EQUITY each year from 1996 to The relative importance of TPS issuance is demonstrated by reporting the ratios of the size of TPS issuance to that of LTD, and LTD and EQUITY issuances combined, respectively. Data on TPS issues are obtained from SNL. Data on LTD and EQUITY issues are obtained from Compustat.

4 222 THE JOURNAL OF RISK AND INSURANCE Cross-sectional evidence suggests that the market rewards larger and more leveraged TPS issuers on the Fed s ruling date. We also find that TPS announcement-date abnormal returns are negatively related to firm size and profitability, and positively related to financial leverage, the extent of cash and fixed (bond) investments, and interest coverage. Consistent with our hypotheses, these findings indicate that higher abnormal returns following TPS issue announcements are related to higher expected financial distress costs, superior growth opportunities, and greater tax burdens. In addition, we find that TPS issuers that plan to use the issue proceeds to retire debt or finance growth experience higher abnormal returns, whereas issuers that plan to use the issue proceeds to retire preferred stock or as capital for a subsidiary experience lower abnormal returns. Finally, by examining the effect of the Fed s ruling on insurers, the study provides evidence of the spillover effect of banking regulation to other financial services firms, as the ruling was directed primarily at banks but appears to have impacted insurers through the shadow regulatory function of rating agencies (White, 2013). 2 The article proceeds as follows. The Hybrid Security Background and Literature Review section reviews the existing literature, and summarizes the structure of and the market for debt equity hybrid securities. The Hypotheses and Variables section develops capital structure hypotheses in relation to TPS. Data, variables, and methodology are described in the Data and Methodology section. Empirical results are presented in the Empirical Results section, including event study results, determinants of TPS issuance, and cross-sectional analysis of abnormal returns. The final section summarizes the findings of this study and provides concluding remarks. HYBRID SECURITY BACKGROUND AND LITERATURE REVIEW The first debt equity hybrid security was issued on October 27, 1993 by Texaco (Benston et al., 2003). By the end of 1999, almost 300 corporations, including insurance and securities firms, had issued at least $65 billion of these securities (Sachs, 2000). TPS are the most common type of debt equity hybrid securities. A TPS deal involves three parties: the holding company, the trust, and the investors. For example, an IHC issuing such a security creates a special-purpose vehicle (SPV) and holds its nominal common equity. The SPV issues preferred stock in the form of TPS to investors, and lends the proceeds to the parent IHC in the form of junior-subordinated, 30-year term, deferrable-interest debentures under economic terms similar to the TPS. This loan is the sole asset of the SPV. The SPV is structured as a statutory business trust formed under state law (typically in a corporation-friendly state like Delaware) and taxed as a partnership. Interest payments on the loan are tax deductible to the IHC and flow through as dividend income to the holders of the TPS. The basic structure of a TPS 2 Demsetz and Lehn (1985) hypothesize that regulators partly substitute for market monitoring of firms. Similarly, we conjecture that rating agencies serve a monitoring role in addition to capital markets and regulators, especially in light of the growing expectation of integrated financial services in the mid to late 1990s (culminating in the Financial Services Modernization Act of 1999).

5 TRUST-PREFERRED SECURITIES AND INSURER FINANCING 223 FIGURE 1 Trust-Preferred Security Arrangement Notes: Illustration of the basic structure of trust preferred securities. Note that the insurance holding company (IHC) holds all the common equity of the trust. First, the trust issues preferred stock to external investors. Second, the proceeds from preferred stock issue are loaned to IHC in exchange for junior subordinated debt. Third, the IHC pays interest on debt. Fourth, the trust pays preferred stock dividends to external investors. deal is illustrated in Figure 1. Starting in 2002, some trust preferred security offerings were pooled and securitized by investment banks as collateralized debt obligations (CDOs). Under these pooled transactions, there are two SPVs related to a TPS offering: one is the statutory business trust formed by the IHC, and the other is a SPV formed by an investment bank that pools and securitizes TPS issued by the trust entities of different IHCs. Securitized insurance trust preferred has opened up a new funding source for small to medium-sized IHCs. 3 Banks and insurers were among the major investors in TPS (Cordell, Hopkins, and Huang, 2011). As state insurance regulators are primarily concerned with operating company solvency in their own states, the regulation of capital sources in the insurance industry is more diffuse than in the banking industry, and capital innovations at the holding company level may receive less regulatory attention than similar innovations at federally regulated banks (Herrig and Schuermann, 2005). Rating agencies are concerned about the flexibility provided by the various funding sources, noting that debt financing imposes more constraints and increases the riskiness of the firm, whereas equity funding provides more flexibility and reduces the firm s risk (A.M. Best, 2003). As hybrid securities exhibit both some of the flexibility of equity and some of the constraints of debt, it was not a priori clear how rating agencies would treat them in their credit analysis. A.M. Best (2003) stated that they were giving equity credit for such issues on a case-by-case basis. In the following years, A.M. Best (2005, 2014a) issued reports suggesting that hybrid securities would be treated like equity provided the characteristics of the securities are more equity-like. 4 Further, insurance operating companies, in contrast to IHCs, began issuing a different type of hybrid security known as surplus notes in the early 1990s. State insurance regulators include surplus notes in policyholder surplus, as they are subordinate to policyholder claims. Interest and principal payments on surplus notes are also subject to regulatory approval. These securities are issued by mutual insurers and stock 3 A.M. Best (2012) describes the rating process and transaction structure of securitized surplus notes and TPS. TPS CDOs usually include rated and unrated TPS and several tranches. 4 These equity characteristics include permanence of capital, ability to defer interest or dividends, and subordination in an insurer s capital structure.

6 224 THE JOURNAL OF RISK AND INSURANCE companies, but not IHCs. 5 By focusing on TPS, which are issued exclusively by publicly traded firms, our study includes a more homogeneous sample and provides evidence on the stock market response to the issuance of hybrid securities by insurance firms. 6 Most pertinent to the present study, though, Benston et al. (2003) and Harvey, Collins, and Wansley (2003) analyze BHC abnormal returns after the 1996 Fed s ruling that permitted TPS and related issues. In an analysis of TPS issuers and nonissuers, Benston et al. find that larger and more financially levered banks were more likely to issue TPS. In addition, they find that firms with more growth opportunities and greater need to reduce income taxes were more likely to issue TPS in 1996 and Benston et al. examine 65 TPS issuers and find positive abnormal returns for early issuers (firms that issued in 1996) and insignificant abnormal returns for late issuers (firms that issued in 1997) around the Fed s ruling date. However, around the firm s Securities and Exchange (SEC) filing date, both 1996 and 1997 TPS issuers experience significantly positive abnormal returns. 7 They also find that abnormal returns are negatively related to prior-year equity ratios and positively related asset growth and tax liability. The negative relationship between abnormal returns and the equity ratio suggests that less-well capitalized firms were rewarded for increasing their Tier-1 capital, whereas the positive relationship between tax liability and abnormal returns suggests that investors valued the increased cash flows resulting from the hybrid securities tax shield. Harvey, Collins, and Wansley (2003) examine a sample of 23 BHCs that issued TPS between October 1996 and September In contrast to Benston et al. (2003), Harvey, Collins, and Wansley find that banks that ultimately issued TPS had significantly positive abnormal stock returns surrounding the Fed s ruling, but generally abnormal returns not significantly different from zero around the issue announcement dates. They suggest that financial markets anticipated and fully incorporated issue expectations around the time of the Fed s announcement, leading to no evidence of abnormal returns upon issue announcement. Although these two key articles appear to reach contradicting conclusions, note that Benston et al. focus only on the first 2 years after the Fed s announcement, whereas Harvey, Collins, and Wansley focus on the first 4 years, and the samples differ substantially. However, both articles provide evidence of a favorable stock market response at either the Fed s ruling date or issue announcement dates. 5 See A.M. Best (2003, 2014b) for an overview of surplus notes. 6 In an unpublished manuscript, Zhang and Cox (2006) examine the determinants of surplus notes issuance and find that larger insurers and mutual insurers are significantly more likely to use surplus notes. These authors examine surplus notes holdings for a single financial statement year. Therefore, our study is also more extensive in that we provide analysis of TPS issuances over a 20-year ( ) period. 7 In contrast, note that for purposes of our empirical analysis, early issuers are firms that issued TPS in 1996 and 1997, and late issuers are firms that issued TPS in Harvey, Collins, and Wansley s (2003) sample is much smaller than Benston et al. (2003) because Harvey, Collins, and Wansley include only banks that have debentures with daily price quotes available on the Bloomberg quotation system.

7 TRUST-PREFERRED SECURITIES AND INSURER FINANCING 225 This study extends the work of Benston et al. (2003) and Harvey, Collins, and Wansley (2003) in three important ways. First, our study provides evidence on the market response to the Fed s ruling by IHCs that issue TPS; considering that the Fed s ruling does not apply to IHCs, any market response suggests a spillover effect of a bank regulation. Second, closely related to the first extension, we provide evidence of the importance of the treatment of TPS by insurer credit rating agencies, specifically A.M. Best Company, as shadow regulators. Third, we examine the ability of different theories of optimal capital structure to explain both the determinants of TPS issuance and the abnormal stock returns among TPS issuers for IHCs. Although this third extension has been studied for banks, it has not been studied for IHCs. This study also updates the data about TPS issuance through 2013 and provides empirical evidence regarding the share price response to additional variables including announced use of funds raised through TPS issuance and the relative importance of TPS issuance to other sources of capital for IHCs. The next section proposes hypotheses related to the benefits and costs available to IHCs from issuing hybrid securities, and considers how the Fed s ruling might have spilled over to the insurance industry. HYPOTHESES AND VARIABLES Modigliani and Miller (1958) identify the basic conditions under which managers and investors would be indifferent to the source of capital funding: absence of flotation costs, financial distress costs, taxes, and agency costs. Since then, extensive theoretical and empirical work has examined the decisions regarding capital funding sources. 9 In this section, hypotheses regarding IHCs, choice of hybrid securities as part of a firm s capital structure are developed. The following hypotheses predict that the stock market response to the Fed s ruling and issuance announcement depends upon the relaxation of the respective assumptions of Modigliani and Miller: (1) firm size, due to comparative advantage in evaluating financing choices and reducing flotation costs, greater investment opportunities, lower insolvency risk, and lower information costs; (2) reduction in expected costs of financial distress; (3) increased capacity to exploit growth opportunities; (4) tax benefits of debt equity hybrids; and (5) information asymmetry, in which managers know more about the investment opportunities than shareholders. All hypotheses are based on the assumption that marginal market participants form rational expectations of the benefits and costs of issuing hybrid securities, and in turn, these expectations are reflected in stock market prices. The first four hypotheses predict zero or positive abnormal stock returns, whereas the last one predicts negative abnormal returns. The intuition and theoretical background for each hypothesis are presented below. Size Firm size is the most common control variable in empirical studies in corporate finance and insurance. Size is a proxy for many firm characteristics, including 9 For a survey, see Harris and Raviv (1991).

8 226 THE JOURNAL OF RISK AND INSURANCE economies of scope and scale, managerial expertise, investment opportunities, and complexity of operations. Also, large firms generally have a lower probability of insolvency due to greater diversification and greater access to capital markets (Cummins and Sommer, 1996). Larger firms generally release more information and are followed by more financial analysts and, consequently, should have lower information asymmetry costs (Bhushan, 1989). Lower information asymmetry costs, greater investment opportunities, and better capital market access imply lower marginal costs and higher marginal benefits for larger firms that issue TPS. Managers of small companies are less likely to have the expertise necessary to analyze, develop, and issue hybrid securities. Thus, firms with relatively small capital needs will face larger flotation costs per dollar of issue, reducing the attractiveness for small firms. Finally, the development of pooled and securitized TPS offerings suggests that firm size was a barrier for some insurers (Sherman, 2003). Therefore, we expect that large firms, as measured by natural log of total assets, are more likely to issue hybrid securities than small firms. 10 Financial Distress Costs The costs of insurer financial distress include increased regulatory attention and increased cost of capital. As shown by Munch and Smallwood (1980), regulatory intervention imposes administrative costs, increased barriers to entry, and decreased efficiency. Unless firms issue hybrid securities to retire equity, their issuance increases the holding company s equity capital while decreasing relative leverage. By reducing the likelihood of financial distress, such an increase in capital may contribute positively to financial strength ratings of operating subsidiaries as holding company capital structure and financial flexibility are important inputs to the private rating process. Further, the additional capital, if distributed among operating subsidiaries, may contribute to a favorable regulatory capital situation for subsidiaries. Thus, TPS issuers should benefit from a reduction in the probability and expected costs of insolvency. This is similar to Merton s (1973) hypothesis, which states that the value of a particular debt issue to a firm depends on, among other things, the probability that a firm will default. We predict that insurers with a higher likelihood of financial distress are more likely to benefit from a reduction in expected financial distress costs. In addition, the reduction in expected financial distress costs is likely to be directly related to the proportion of debt financing in the insurer s capital structure. Firms with higher financial leverage, as measured by debt to total assets would, thus, be expected to gain the most from hybrid security issuance. Other potential measures of expected financial distress costs include the ratio of preferred stock to total assets and the interest coverage ratio, measured as the ratio of earnings before interest and taxes (EBIT) to the sum of interest expense and preferred dividends. 10 Balla, Cole, and Robinson (2011) point out that flotation costs became less binding over time as TPS became more frequently collateralized.

9 TRUST-PREFERRED SECURITIES AND INSURER FINANCING 227 Growth Opportunities Additional financial capital may expand capacity to take advantage of growth opportunities. In particular, Jung, Kim, and Stulz (1996) find that equity issuance adds shareholder value for firms with good investment opportunities, but not otherwise. This is especially true for IHCs during the late 1990s, a time of financial innovation and structural change that culminated with the passage of the U.S. Financial Services Modernization Act (also known as the Gramm Leach Bliley Act [GLB]) in GLB promoted, among other things, consolidation and integration within the financial services industry. In addition, financial innovation and new product development further expanded the investment opportunities available to IHCs. The flexibility inherent in hybrid securities makes them particularly attractive for financing future growth opportunities, so shares of IHCs with greater growth opportunities would realize more positive returns when announcing an issue. Firms with low growth opportunities will invest assets in less liquid and more risky securities such as equities rather than internal growth opportunities. Thus, equity investments to total invested assets should be negatively related to abnormal returns from issuing hybrid securities. Other balance sheet measures of growth opportunity include the following two variables divided by total cash and invested assets: cash and fixed (bond) investments. Finally, three profitability measures are used to measure growth opportunities: pretax operating income divided by total revenue (pretax return on revenue) and divided by total assets (pretax return on assets) and pretax investment yield (pretax net investment income to total cash and invested assets). However, Irvine and Rosenfeld (2000) find that the presence of growth opportunities is not an important determinant of BHC shareholder value effect of issuing hybrid securities, as they act more like subordinated debt issues than equity issues. If investors agree with bank regulators and treat IHC hybrid securities as equity, firms with greater growth opportunities will experience higher event-day returns. If hybrids continue to be viewed more like debt, consistent with Irvine and Rosenfeld, there will be little relationship between event-day returns and proxies for growth opportunities. Tax Benefits Hybrid securities, especially TPS, generate greater tax savings than equity as interest payments made to the trust subsidiary are usually deductible at the holding company level. IHCs seeking to exploit these tax advantages would use hybrid securities either to redeem common or preferred stock, or in place of stock for corporate financing. As shown by Graham (2000) and Titman and Tsyplakov (2007), firms can dynamically balance tax shields and expected bankruptcy costs to maximize shareholder value. Thus, in general, insurers with relatively higher income tax rates would be more likely to issue TPS, and their shares should experience larger gains when such issues are announced. Note that this hypothesis holds whether investors view hybrid securities as debt or equity, as debt-like tax benefits can be realized in either case. In general, profitable firms are more likely to benefit from the potential tax savings of TPS, and hence, we rely on the measures of profitability, as well as the interest coverage ratio, discussed earlier to test this hypothesis.

10 228 THE JOURNAL OF RISK AND INSURANCE Information Asymmetry and Agency Costs Myers and Majluf (1984) developed the pecking order theory, in which information asymmetry between managers and investors dictates that managers will issue equity when they have information superior to that of investors. As investors recognize this, share prices fall upon equity issue. Hybrid securities can mitigate this problem as they provide the disciplining features of debt with some of the flexibility of equity. Speiss and Affleck-Graves (1995), among others, show that equity issues are typically related to stock market losses. Stock price drops on equity issuance are generally related to either the dilution of existing shares or negative investor response to managerial signaling. Although Harvey, Collins, and Wansley (2003) show that debt issues (which do not increase the probability of financial distress) are related to stock market gains, other research suggests that stock market responses to debt issuance are generally negative, but at least not as negative as the responses to equity issues (see, e.g., Eckbo, 1986; Mikkelson and Partch, 1986). In contrast to evidence of a negative market response to debt issues by firms in general, Akhigbe, Borde, and Madura (1997) demonstrate that property liability insurers experience positive stock price reactions upon debt issues. The positive stock price reaction to debt issues likely arises from the tax benefits of debt and signals that managers believe that the firm is capable of meeting its debt service obligations. Managers decisions to use debt or equity may partially reveal their private information regarding the present value of assets and the firm s ability to provide investors with their required return. Whether hybrid securities generate abnormal returns is an empirical question and depends on whether investors view hybrid securities as more like debt or equity, among other unobserved factors. Agency costs related to different incentives of firm managers, current owners, and new investors also may influence the decision to issue TPS and the market response to TPS issues. Although credit rating agencies and financial analysts may help mitigate costs related to information asymmetry and incentive conflicts among stakeholders, firm insiders may still possess information that outsiders do not, and the true or complete motivation for TPS financing by decision makers (such as firm management and directors) is not observable. To help capture these agency costs and information costs, we add a set of indicator variables related to the planned (i.e., stated) use of TPS proceeds by IHCs that issue TPS to the list of variables that may explain cross-sectional differences in abnormal returns of TPS issuers. 11 We expect that abnormal returns will be positive when the proceeds are used to retire debt because TPS preserves the tax shield of debt, decreases financial distress costs, and increases financial flexibility. We expect abnormal returns will be negative when the proceeds are used to retire common or 11 The stated use of proceeds is identified from Factiva for 75 of the insurers that issued TPS. A TPS insurer may state more than one planned use of proceeds. Seventy-three issuers stated that they would use the proceeds for general corporate use, 29 to retire debt, 8 for acquisition, 7 for capital for subsidiary, 5 to redeem preferred stock, 4 to retire common stock, and 3 to finance growth.

11 TRUST-PREFERRED SECURITIES AND INSURER FINANCING 229 preferred equity because financial distress costs are likely to increase and financial flexibility decrease. We expect abnormal returns to be negative when the proceeds are used for growth, acquisitions, or subsidiary capital due to potential agency costs and information asymmetry. We do not have a prediction for the share price response to general corporate purposes. Furthermore, rating agencies play an important role in reducing monitoring and information costs for regulators, investors, and policyholders (Doherty and Phillips, 2002). The role of rating agencies relies on imperfect information regarding a firm s risk and true value. The rating process and eventual assignment of a financial strength rating to an insurer reduces the information asymmetry between insiders (especially managers) and outsiders (investors and regulators). Rating agencies often have access to proprietary information and internal company reports that firms do not publicly disclose in an unbundled form. Firms that submit to the rating process do so to achieve benefits associated with the communication of private knowledge in the most cost-efficient manner (Doherty and Phillips, 2002). Thus, if the ratings agencies were to treat TPS as equity when determining an insurer s financial strength rating, we expect that highly rated insurers would be more likely to issue TPS because of institutional features of insurance product markets and financial markets. For instance, Sommer (1996) finds that the price consumers are willing to pay for insurance is inversely related to an insurer s insolvency risk. Therefore, for insurers, capital structure and product pricing are strongly linked. Akhigbe, Borde, and Madura (1997) argue that because the demand for insurance products (reflected in policyholders reservation prices) is directly related to the insurer s credit ratings, the stock market response to insurer security offerings should be less negative than that of nonfinancial firms. 12 These authors also discuss the potential monitoring and information benefits of external financing. We use a binary variable that takes the value of 1 if the firm obtains an A.M. Best financial strength rating, and 0 otherwise. DATA AND METHODOLOGY Our sample of IHC hybrid security issues is obtained by searching the Factiva and SNL Financial databases. Announcement dates and issue sizes are confirmed through searches of the Factiva database, SEC filings, annual reports to shareholders, and SNL. 13 Similar to Benston et al. (2003), in the case of multiple filings by a single firm, only the first issue with sufficient stock price data is included in the sample of issuers. From these sources, we identified 98 publicly traded insurance firms that issued hybrid securities between 1993 and We did not identify any publicly disclosed first-time TPS issues by U.S. IHCs after In Table 2, we report the number of firms 12 Akhigbe, Borde, and Madura (1997) find support for their hypothesis that insurer stock returns around equity issue announcements are less negative than those of comparison industrial firms, but they are not significantly different from zero in their sample. 13 We used SNL Insurance Quarterly, a paper publication, through 2004, and the online product SNL Unlimited: Industry-Specific Financial Market Data for

12 230 THE JOURNAL OF RISK AND INSURANCE TABLE 2 Sample of First-Time TPS Issuers ( ) All Issuers Issuers in A.M. Best, Compustat Announcement Year N Percent N Percent Total Note: All Issuers are all insurers that announced TPS issuance for the first time from 1993 to 2007, and Issuers in A.M. Best, Compustat are the subsample of all issuers with financial statement data available in Best's Holding Company Guide or Compustat. Issuers in A.M. Best, Compustat are based on Best's Holding Company Guide (Best, ) through data year 2000, and Compustat thereafter. each year that announced issuance for the 98 firms that issued TPS, and the subset of 89 firms with financial statement data. 14 Table 2 highlights the relative prominence of hybrid security issues in 1996 and 1997, in which over 40 percent of the firm announcements occurred during the sample period. Note that Tables 1 and 2 represent different samples. Table 1 illustrates all new capital issued by publicly traded insurers between 1996 and Table 2 summarizes firsttime TPS issues, both for all first-time issuers and then for the subset of issuers included in both A.M. Best and Compustat databases. As Table 2 includes only firsttime issuers, some issues reflected in Table 1 will not be identified in Table 2. The two tables together suggest that first-time issuance ended in 2007 and TPS issuance in general ended in Although some banks continued to issue TPS instruments after 2008, insurers did not. One possible explanation for the cessation of insurer issues is that those insurers that would benefit from the tax advantages and flexibility 14 As shown in Table 2, among the 89 TPS issuers (i.e., IHCs) with financial statement data, 32 issuers are related to Among these 32 issuers, 13 participated in pooled/ securitized transactions.

13 TRUST-PREFERRED SECURITIES AND INSURER FINANCING 231 associated with TPS had already issued before 2008 and would not benefit from subsequent issues. Furthermore, the investor appetite for hybrid securities issued by insurers may have waned relative to banks during the financial crisis, as investors sought out more traditional investment vehicles. In Tables 1 and 2, we observe time-series variation in TPS issuance by insurers. Specifically, after the Federal Reserve ruling (which applied to banks), there was an influx of TPS issuance by insurers. This continued in 1997, with TPS issuance representing 14 percent of traditional debt and equity issues in 1996 and The insurers with the most to gain from issuing TPS would have taken advantage of the opportunity as soon as it was feasible. After 1997, however, TPS issuance drops off substantially, with no aggregate issues over $5 billion in any year, and no more than 4.6 percent of traditional debt and equity issues. Table 1 also illustrates that TPS issuance dropped off entirely during the time of the financial crisis of 2008 and never recovered. Finally, Table 1 illustrates a dramatic growth in use of long-term debt during the decline in TPS issuance, exceeding $100 billion per year from 2005 to 2007 and exceeding $200 billion in The shift from TPS and equity issues to debt issues during this time may reflect investor preferences shifting toward more stable returns over this time period. Furthermore, the elimination of TPS allowance as Tier 1 capital for banks (included in the Dodd Frank Act of 2010), along with the financial crisis of 2008, and changing financial market conditions generally may have changed the cost benefit analysis of TPS issuance for insurers to favor more traditional capital sources. A further explanation of the cessation of TPS issuance by insurers is that approximately 80 percent of the IHCs still in existence in 2013 had issued TPS prior to 2009, presumably as part of their estimated optimal capital structure. Hence, further issuance may cause the IHCs to deviate from their optimal capital structure. Although the Dodd Frank Act did not apply directly to IHCs, it may have indicated a new perspective among investors and regulators about the relative benefits of TPS. Finally, both macroeconomic and industry factors jointly influence both demand and supply of capital. 15 A rigorous study of the time-series patterns of TPS, along with the changes in demand for both debt and equity capital by insurers, would be complex and is suggested for future research. The universe of potential TPS issuers is constructed from an initial sample of 239 unique U.S. publicly traded IHCs listed in Best Holding Company Guide (A.M. Best, ), using data reported for operations in Of these 239 firms, 227 have stock return data available in the Center for Research in Security Prices (CRSP) database. We include an additional 16 insurers that issued TPS during the sample period and have data available in CRSP, but are not in A.M. Best ( ). Thus, our sample of firms for purposes of our event study analysis numbers 243, which includes 151 firms that did not issue TPS and 92 firms that 15 Baker, Greenwood, and Wurgler (2003) and Zhou et al. (2012) provide evidence that macroeconomic factors influence the time-series pattern of corporate debt issues.

14 232 THE JOURNAL OF RISK AND INSURANCE TABLE 3 Logistic Regression Analysis of Early ( ) TPS Issuers Explanatory Variable Coefficient Intercept (2.1095) LN(Assets) (0.1176) Long-term debt to total assets (3.9973) Net investment income to invested assets (0.1329) Rating Dummy (0.7859) Observations 212 R Note: This regression is modeling the probability of being an early TPS issuer that issued TPS for the first time from January 1, 1996 to December 31, There are 212 observations in total, among which 38 are 1996 or 1997 issuers, and 174 are not. LN(Assets) is the natural log of firm s assets. Rating Dummy equals 1 if an insurer obtains a financial strength rating by A.M. Best, and 0 otherwise. Standard errors are reported in parentheses., and denote statistical significance at the 1, 5 and 10% levels, respectively. did issue TPS between 1994 and Because the last financial statement year available in A.M. Best ( ) is 2000, we obtain financial statement data from Standard & Poor s Compustat for years 2001 through 2007 for our sample firms. 17 We conduct a logistic regression analysis to identify determinants of TPS issuance. These results are reported in Table 3 and discussed in the Empirical Results section. An indicator variable is set equal to 1 for firms issuing TPS in 1996 and 1997, and 0 otherwise, with financial data as reported in 1995 comprising the explanatory variables. 18 For the 1995 data year, among the 239 IHCs, 221 have complete financial data in A.M. Best ( ); among the 221, we drop 9 companies that issued debt equity hybrid securities prior to 1996, leaving a final sample of 212 insurers. Among the 40 insurers that announced debt equity hybrid securities in 1996 or 1997, 38 had financial data in A.M. Best ( ) for data year We use 1995 financial data because the largest volume of TPS occurred in 1996 and 1997, and our event study 16 Note that some of these firms did not have sufficient stock return data for the final samples used in the event study analysis. 17 Available financial statement data on insurers are more extensive in Best Holding Company Guide (A.M. Best, ) than in Standard & Poor s Compustat, and for that reason, we rely on it for such data through year As explained shortly, over 40 percent, the TPS issuance are in 1996 and 1997, and that influenced our decision to use 1995 as our data year of primary interest.

15 TRUST-PREFERRED SECURITIES AND INSURER FINANCING 233 results point to 1996 and 1997 as the years when hybrid security issuance generated significant positive abnormal stock returns. By using 1995 data, we can isolate the determinants of issue based on the most recent annual data available to investors when the Fed s ruling was announced. 19 In the logistic regression, a firm is classified as a TPS issuer if it issued a trust-preferred, monthly income preferred or other similar security between January 1, 1996 and December 31, 1997, and appeared in A.M. Best ( ) for data year The event study portion of our empirical analysis consists of two major parts, reported in Table 4. In the first part, abnormal returns surrounding the Fed s ruling date of October 21, 1996 are examined. In the second part, the abnormal returns surrounding the announcement date are examined. The announcement date is the first public announcement of a planned or actual issue of a hybrid security. Most often, the announcement came in the form of a press release or business press article about the company. However, in some instances, the first public announcement is contained in an SEC filing. To capture information leakage before the event and delayed effects for less liquid stocks, returns around three windows are studied. 20 It is important to note that standard event study results for events that affect multiple firms from the same industry on the same day are potentially subject to crosscorrelation bias, which can inflate test statistics and result in over-rejection of the null hypothesis. This is true for the ruling-date analysis, so we correct the problem by aggregating firm returns for each group into portfolios to address potential crosscorrelation bias. The standard market model tests for deviations from expected return when controlling for firm risk and broad market moves, using the capital asset pricing model (CAPM). Daily stock returns, including dividends, for each firm of interest on each day in the estimation window are regressed against the market return on the same day. The daily return on the CRSP equal-weighted index proxies for the market return. The estimation period begins 255 trading days prior to the event and ends 15 days prior to the event. Parameter estimates for Equation (1) are estimated using ordinary least squares (OLS): AR it ¼ R it ^a i þ ^b i R it ; ð1þ where ^a and ^b are estimated from the standard CAPM over the estimation period. AR it is security i s abnormal return on trading day t and R it is security i s observed 19 As shown in Table 1, represents the peak 2-year period of TPS issues by IHCs with over $14 billion in issues, an amount exceeding common and preferred stock issues for those 2 years. 20 To ensure robustness of our results, we tested additional windows but do not report them here in order to conserve space. Results of these additional tests are available upon request from the authors but none changes the findings related to either the Fed s ruling or announcement of an issue.

16 234 THE JOURNAL OF RISK AND INSURANCE TABLE 4 Cumulative Average Abnormal Returns Event Windows All Issues (1) Panel A: Ruling Date (Portfolio Event Study) Early Issues Following Ruling (2) Early Issues (3) Late Issues (4) Nonissuers (5) Differences in CAARs (3) (4) ( 1,þ1) 0.57% 0.92% 1.03% 0.18% 0.12% 1.21% (0.859) (1.264) (1.417) ( 0.154) ( 0.209) (4.609) ( 5,þ5) 0.17% 0.35% 1.49% 1.32% 0.01% 2.81% (0.136) (0.251) (1.065) ( 0.586) ( 0.007) (5.555) ( 20,þ20) 4.13% 4.72% 6.19% 0.05% 0.46% 6.24% (1.676) (1.752) (2.297) (-0.010) (0.214) (5.705) N Dropped Panel B: Announcement Date (Standard Event Study) Event Windows All Issues (1) Early Issues Following Ruling (2) Early Issues (3) Late Issues (4) Differences in CAARs (3) (4) ( 1,þ1) 0.16% 0.12% 0.02% 0.50% 0.48% ( 0.433) ( 0.132) ( 0.377) ( 0.528) (2.951) ( 5,þ5) 0.65% 2.65% 1.39% 0.11% 1.28% (1.272) (2.993) (1.821) (0.182) (7.800) ( 20,þ20) 0.67% 3.21% 0.58% 1.38% 0.80% (0.260) (1.665) (0.226) (0.377) ( 1.053) N Dropped Note: Panel A reports cumulative average abnormal returns of insurance firms surrounding Fed s ruling date regarding trust preferred securities (TPS), and Panel B reports cumulative average abnormal returns of insurance firms surrounding their announcement of planned or actual issue of TPS. All Issues denote all first-time TPS issues by insurers from 1994 to Early Issues Following Ruling represent TPS issues immediately after the Fed s ruling date untill the end of 1997, which is from October 21, 1996 to December 31, Early Issues are TPS issues between January 1, 1996 and December 31, Late Issues are TPS issues between January 1, 1998 and December 31, N represents the number of TPS issues in the sample. Dropped shows the number of TPS issues that are dropped from the sample because of insufficient stock return data. Significance test statistics are reported beneath cumulative average abnormal returns. For the ruling-date studies, the portfolio Patell Z-statistic is reported to address cross-correlation biases associated with a common event day. For the announcement-date studies, the firm-level Patell Z-statistics are reported. Test statistic for difference between CAARs are based on a Welch s t-test and sample standard deviations of each portfolio s abnormal returns.,, and denote statistical significance at the 1, 5, and 10% levels, respectively.

17 TRUST-PREFERRED SECURITIES AND INSURER FINANCING 235 return on trading day t. Abnormal return significance for each window is tested using the Patell (1976) test. 21 To test the individual hypotheses, three different event window cumulative abnormal returns are used as the dependent variable in OLS regressions, with a vector of firmspecific variables as the explanatory variables. The firm-specific variables are proxies for various firm characteristics that could explain the demand for and shareholder response to hybrid security issuance. Given the potential for correlation among financial statement variables to induce multicollinearity and reduce the ability to identify significant explanatory variables, and considering that two or more variables are available to test most hypotheses, we used a stepwise variable selection method with a probability value of 0.15 for a variable to enter into and to stay in a regression model, with the exception of log of assets, which is included in all regression models. Although an alternative is to include all explanatory variables in each model, the significance and sometimes the coefficient sign of some variables change even though the overall model fit may improve slightly. We believe our approach allows for a finer examination of the importance of individual explanatory variables with virtually no loss of information. As a result of our approach, although all (13) potential explanatory variables initially considered for inclusion in each regression model are the same, the specific variables in each final model differ to some extent from one model to another. Although the specific variables in the final models do differ, the characteristic measured or hypothesis tested may be the same because two or more variables are available to test most hypotheses. EMPIRICAL RESULTS Empirical results are divided into three subsections. First, we examine the determinants of TPS issuance among IHCs. Second, we report event study results for the Fed s ruling date and TPS announcement dates using the whole sample of TPS issuers and various subsamples. Finally, we examine the relationship between cumulative average abnormal returns (CAARs) and firm-specific factors for early (i.e., ) TPS issuers surrounding the Fed s ruling and announcement dates. Determinants of Issue Results A logistic regression analysis is conducted on the universe of 212 eligible firms in the sample to determine the characteristics of firms that led to TPS issue. In particular, we 21 When multiple firms in the same industry are subject to a single event, the standard market and market-adjusted return models may not be appropriate, as common effects experienced by all firms in the industry may bias the residual distribution. When testing for abnormal returns among firms affected by new regulation or legislation, the residual independence violation may inappropriately inflate standard error estimates and render these models inconsistent. There are several alternative tests that correct for these potential problems. Here, we test the abnormal return on a portfolio of potentially affected securities. This approach provides more consistent results and is used to evaluate the abnormal returns from the Fed ruling.