REGULATORY ENFORCEMENT AND OFF-BALANCE SHEET ENTITIES: EVIDENCE FROM THE “SHADOW INSURANCE” MARKET By Inna Voytsekhivska A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Business Administration – Doctor of Philosophy 2017 ABSTRACT REGULATORY ENFORCEMENT AND OFF-BALANCE SHEET ENTITIES: EVIDENCE FROM THE “SHADOW INSURANCE” MARKET By Inna Voytsekhivska Captive reinsurance (“shadow insurance”) is a controversial form of non-traditional reinsurance that is associated with opaque statutory reporting in the insurance industry. Captive reinsurance subsidiaries are special purposes entities that are not consolidated under statutory accounting principles (SAP) and can be used to manage a firm’s statutory reserve liabilities and premiums, which are reported net of reinsurance. This paper studies the relation between regulatory enforcement and an insurance firm’s use of off-balance sheet captive insurance entities, as well as the implications of regulatory enforcement and captive reinsurance use for the firm’s credit ratings and the degree of information asymmetry in the market for the firm’s equity. I find that regulatory enforcement is negatively associated with the use of captive reinsurance among life insurers. Among life insurers, I find some evidence that credit rating agencies infer information about a firm’s default risk from its regulatory enforcement environment, and that regulatory enforcement can reduce information asymmetry in the market. Contrary to what I hypothesize, the use of “shadow insurance” is negatively associated with proxies for information asymmetry in equity markets among pure property-casualty insurers. Overall, my findings suggest that regulated firms, credit rating agencies, and equity investors act as if regulatory enforcement increases the credibility of accounting reports and reduces information asymmetry in the market. Also, public awareness of accounting issues may be important for regulatory enforcement and its credibility. These findings should be of interest to regulators, investors, preparers, and other stakeholders impacted by accounting standards and their enforcement. Copyright by INNA VOYTSEKHIVSKA 2017 This thesis is dedicated to my family. iv ACKNOWLEDGEMENTS I would like to thank Dr. Marilyn Johnson for her encouragement and continuing advice throughout my time in the PhD Program in Accounting at Michigan State University, especially during the dissertation stage of the program. Dr. Kathy Petroni provided excellent constructive feedback on numerous theory-related and methodological issues. I am grateful to Drs. Ranjani Krishnan and Chris Hogan for being excellent mentors both in and outside of classroom: I have learned a lot from them about the value of being an excellent researcher, teacher, mentor, colleague, and community member. I am also grateful to Dr. Janice Beecher for inspiring me to do research on regulatory issues and for teaching me regulatory theories from economics, sociology, and political science. I want to express my gratitude to Drs. Andrew Acito, John Jiang, Mike Shields, Isabel Wang, and Dan Wangerin for their guidance, advice, and constructive criticism in the Accounting PhD seminars and/or Accounting PhD workshops. I would also like to thank various insurance industry experts for their help. They provided me with important and useful information that helped me better understand insurance regulation and captive reinsurance market. On a personal note, I owe a lifetime of gratitude to my family and friends. My husband Igor, my daughter Valerie, and my son Anatoliy were the best source of inspiration and encouragement; I would not have finished the PhD program without them. I am deeply grateful to my mother Svitlana and my sister Victoria for their unquestionable love and support. I would also like to thank my fellow students, whose friendship and support helped me navigate through and enjoy the PhD program. v TABLE OF CONTENTS LIST OF TABLES ..................................................................................................................... viii CHAPTER 1: INTRODUCTION ................................................................................................ 1 CHAPTER 2: BACKGROUND AND HYPOTHESIS DEVELOPMENT ........................... 10 2.1. U.S. Insurance Regulation .................................................................................................. 10 2.2. Solvency Regulation in the Insurance Industry.................................................................. 11 2.2.1. Statutory Accounting Principles (SAP) and Loss Reserving ...................................... 11 2.2.2. Risk-Based Capital (RBC)........................................................................................... 13 2.3. Reinsurance and the Captive Reinsurance Market ............................................................. 15 2.3.1. Traditional vs. Captive Reinsurance ............................................................................ 15 2.3.2. Incentives for Captive Reinsurance ............................................................................. 18 2.4. Hypothesis Development ................................................................................................... 20 2.4.1. Regulatory Enforcement and Captive Reinsurance ..................................................... 21 2.4.2. Regulatory Enforcement, Captive Reinsurance, and Credit Ratings ........................... 26 2.4.3. Regulatory Enforcement, Captive Reinsurance, and Capital Markets ........................ 27 CHAPTER 3: SAMPLE SELECTION AND RESEARCH DESIGN ................................... 30 3.1. Sample Selection ................................................................................................................ 30 3.2. Regulatory Enforcement Variable ...................................................................................... 33 3.3. Captive Reinsurance Determinants .................................................................................... 37 3.4. Captive Reinsurance, Regulatory Enforcement, and Credit Ratings ................................. 40 3.5. Captive Reinsurance, Regulatory Enforcement, and Information Asymmetry ................. 42 CHAPTER 4: MAIN RESULTS ............................................................................................... 44 4.1. Descriptive Statistics .......................................................................................................... 44 4.2. Main Results....................................................................................................................... 46 4.2.1. Regulatory Enforcement and Captive Reinsurance ..................................................... 46 4.2.2. Regulatory Enforcement, Captive Reinsurance, and Credit Ratings ........................... 50 4.3.2. Regulatory Enforcement, Captive Reinsurance, and Information Environment ......... 51 CHAPTER 5: SUPPLEMENTAL ANALYSES ...................................................................... 53 5.1. Alternative Specifications of Captive Reinsurance ............................................................ 53 5.2. Alternative Functional Form .............................................................................................. 54 5.3. Endogeneity Robustness Tests ........................................................................................... 55 5.3.1. Instrumental Variable, SURE, and Heckman Estimation............................................ 55 5.3.2. Falsification Test ......................................................................................................... 57 5.3.3. Propensity Score Matching (PSM) .............................................................................. 58 CHAPTER 6: CONCLUSIONS ................................................................................................ 61 APPENDICES ............................................................................................................................. 65 vi APPENDIX A APPENDIX B APPENDIX C APPENDIX D APPENDIX E Reinsurance Example ...................................................................................... 66 Calculation of Regulatory Variables................................................................ 68 Variable List .................................................................................................... 69 Sample Selection ............................................................................................. 72 Main Results…………………………………………………………………..73 REFERENCES…………..….………………………………………………………………………...103 vii LIST OF TABLES TABLE A1 Reinsurance Effect on the Balance Sheet Accounts……...……………….………66 TABLE A2 Reinsurance Journal Entries…………...……………………………...….……….67 TABLE B1 Regulatory Score Calculation…..…………………………..…………….……….68 TABLE C1 Variable List……………………………………………..……………….……….69 TABLE D1 Sample Selection Steps……………………………………………………………72 TABLE E1 Pooled Sample Descriptive Statistics and Correlations…………………...……...73 TABLE E2 Descriptive Statistics and Univariate Tests of Differences………………............80 TABLE E3 Regulatory Enforcement Factor……………………………………………..........82 TABLE E4 Regulatory Enforcement and Captive Reinsurers……………………….…...…...83 TABLE E5 Regulatory Enforcement, Captive Reinsurers, and Credit Ratings………......…..86 TABLE E6 Illiquidity Factor……………………………………….……………………..…..88 TABLE E7 Regulatory Enforcement, Captive Reinsurers, and Information Asymmetry.…....89 TABLE E8 Alternative Specifications………………………….………………………....…..92 TABLE E9 Quantile Tobit Model: Regulatory Enforcement and Captive Reinsurers….…….94 TABLE E10 IV, SURE, and Heckman Estimation: Regulatory Enforcement and Captive Reinsurers……………...………………………………………………..……..…96 TABLE E11 Falsification Test: Regulatory Enforcement and Capital Gains……………….....99 TABLE E12 PSM: Regulatory Enforcement, Captive Reinsurers, and Credit Ratings……....100 TABLE E13 IV: Regulatory Enforcement, Captive Reinsurers, and Credit Ratings……........101 TABLE E14 PSM: Regulatory Enforcement, Captive Reinsurers, and Information Asymmetry………………………………………………………….…………...102 viii CHAPTER 1: INTRODUCTION In this paper, I examine regulatory enforcement of complex accounting standards pertaining to off-balance sheet transactions in the insurance industry (hereafter, shadow insurance). U.S. insurance firms are regulated at the state level, and there is considerable heterogeneity across states in enforcement. I examine whether heterogeneity in regulatory enforcement is associated with insurers’ use of off-balance sheet entities for statutory reporting. I also examine whether credit ratings and proxies for information asymmetry in equity markets reflect a firm’s regulatory enforcement climate and its use of shadow insurance.1 Prior academic literature provides evidence on the role of regulatory enforcement. Some studies find that regulatory enforcement can significantly affect managers, auditors, underwriters, and market participants (Dechow et al. 1996; Beatty et al. 1998; Beneish 1999; Farber 2005). Regulatory enforcement is associated with a lower cost of capital (Leuz and Hail 2006), as well as a reduction in information asymmetry and incentives for tax avoidance (Guedhami and Pittman 2008; El Ghoul et al. 2011; Johnson and Petacchi 2014; Bens et al. 2016; Kubick et al. 2016). Other studies demonstrate some of the limits of regulatory enforcement (La Porta et al. 2006; Djankov et al. 2008) and that regulated firms anticipate regulatory enforcement (Dechow et al. 2016). Regulatory agencies can be captured by the industry they monitor (deHaan et al. 2015; Cornaggia et al. 2016), and regulatory enforcement can be limited due to information asymmetry between regulated entities and regulators. 1 In the United States all insurance companies file regulatory reports to their state regulators under Statutory Accounting Principles (SAP). Public insurance companies also file financial reports to the Securities and Exchange Commission (SEC) under the U.S. GAAP. SAP and US GAAP differ in their requirements for the consolidation of variable interest entities (VIEs). Under SAP insurance companies have an option to not consolidate VIEs that must be consolidated under US GAAP. Shadow insurance represents transactions with related VIEs that are not consolidated, and hence off-balance sheet, under SAP. 1 One source of information asymmetry between regulators and regulated entities stems from accounting standards that allow off-balance sheet activities. For example, in the 2002 Enron scandal, “a Senate investigation found “systemic and catastrophic failure” by the Securities and Exchange Commission in its regulation of Enron Corp” that used off-balance sheet entities to hide losses and risks from the equity investors and various external monitors (Wall Street Journal 2002). Off-balance sheet activities in the financial sector have been implicated in the financial crisis of 2007-2008. The 2008 American International Group (AIG) bailout revealed that the company had liquidity issues in its non-insurance entity, which was not consolidated under Statutory Accounting Principles (SAP) and hence not subject to insurance regulation. The AIG was “doing stuff that was totally outside of what insurance regulators were looking for or able to look for” (Sage Business Researcher 2017). Since 2008 various regulators in the United States and abroad have taken steps to regulate off-balance sheet transactions in the financial sector. However, the shadow banking and insurance sectors are growing in the United State and abroad (e.g., UK, China, Ireland, Germany), and there is little evidence on whether regulatory enforcement addresses the use of off-balance sheet transactions in the financial sector (Sage Business Researcher 2017). This study contributes to the debate about regulatory enforcement of accounting standards by examining captive reinsurance, a complex form of non-traditional reinsurance that is associated with statutory reporting opacity in the insurance industry. Captive reinsurers are subsidiaries organized by insurance companies to provide reinsurance services to the captive’s parent and its affiliates (i.e., sister subsidiaries). Similar to traditional reinsurance, captive reinsurance allows insurers to reduce reinsurance contracting costs, to decrease tax liabilities, and to manage statutory reserve liabilities through reinsurance accounting (see Appendix A for 2 an example). Reinsurance accounting standards have some similarities across statutory accounting principles (SAP) and US GAAP.2 Under both frameworks, if an insurer transfers insurance risk to other companies, it can use reinsurance accounting to reduce its net loss reserves; deposit accounting standards (i.e., no reinsurance credit) apply to reinsurance contracts that result in risk retention. One important difference, however, is that under SAP insurance reports are net of reinsurance, while under US GAAP financial reports are gross of reinsurance. As a result, insurance companies could have incentives to use reinsurance accounting to manage their reported numbers (e.g., net written premiums, net loss reserves) under SAP. Unlike traditional third-party reinsurance, affiliated captive reinsurance can be opaque due to SAP consolidation standards and confidential reporting. In contrast to U.S. GAAP, variable interest entities such as affiliated captive reinsurance subsidiaries are not consolidated under SAP, allowing insurance firms to report on a stand-alone basis. As a result, the risk associated with captive reinsurance stays off-balance sheet. In most jurisdictions, captive reinsurers are not required to publicly release their financial statements and must privately (confidentially) report only to the state regulator that issued the original captive license (NAIC White Paper 2013)3. This statutory reporting opacity, in addition to the accounting complexity of captive reinsurance transactions, could limit policyholders’, creditors’, and equity holders’ ability to determine the degree of risk retention by the captive’s parent and its affiliates. Hence, stakeholders must rely on state insurance regulation for protection. Exceptions include retroactive reinsurance contracts, liability for overdue reinsurance receivables, and contracts that transfer underwriting risk but not timing risk. 3 Only Iowa’s captive standards allow the public release of captives’ financial statements. In other captive jurisdictions, captives’ financial statements can be released only under subpoena, unless the company permits its insurance commissioner to release the captive’s financial data. In a few jurisdictions, captive standards do not allow the insurance commissioner to release captives’ financial records even under subpoena (NAIC White Paper 2013). 2 3 Advocates of captive reinsurance argue that state insurance regulation provides adequate protection against risks imposed by captive reinsurance (Harrington 2015). Insurance regulators can disallow the transfer of reserves to affiliated captives, as captive reinsurance transactions require regulatory approval. However, regulatory approval of captive reinsurance is inherently subjective, and there is significant variation across states in enforcement incentives and opportunities. Critics allege that captive reinsurance allows an insurance company to engage in regulatory arbitrage of reserving standards by shifting risk into a market with high opacity and low regulatory oversight (Harrington 2015). Low regulatory oversight is driven by regulators’ incentives to encourage economic development in their state and results in lax enforcement of reinsurance accounting standards (Schwarcz 2015). Overall, regulation of captive reinsurance has raised the question whether group regulation of insurers, which report on stand-alone basis, is adequate or results in lax regulatory enforcement environment. Motivated by the debate about the adequacy of state insurance regulation, my thesis examines the association between regulatory enforcement and the use of captive reinsurers, as well as the implications of regulatory enforcement and captive insurance use for a firm’s credit ratings and the degree of information asymmetry in the market for the firm’s equity. Based on the prior academic literature and insights from the new institutional economics (NIE) framework (Richter 2015), I hypothesize that regulatory enforcement reduces incentives for captive reinsurance (H1) due to increases in firms’ expected regulatory compliance and non-compliance costs. Regulatory enforcement can be state-dependent. That is, regulators can use public information signals about the firm to condition their enforcement of accounting standards (Mills and Sansing 2000; Beck et al. 2000; Mills et al. 2010). Since surplus adequacy is important for 4 solvency monitoring, I hypothesize that a firm’s surplus position (constraint) moderates the association between regulatory enforcement and captive reinsurance. Insurers need “free” surplus to grow (i.e., increase capacity) and to manage their risk positions (Society of Actuaries 2000). Furthermore, capacity constraints can result in increased insurance prices and reduced insurance product availability (Doherty and Posey 1997), which are considered in insurance regulation. Hence, I expect that insurers with “constrained” surplus are more likely to receive regulatory attention, and thus, I hypothesize a stronger negative association between regulatory enforcement and captive reinsurance among insurers’ with the surplus constraint (H2). Based on the theory of hard and soft information (Bertomeu and Marinovic 2016), I argue that regulatory enforcement is a form of “soft” information that increases the credibility of insurers’ financial reports to other stakeholders. I predict that regulatory enforcement is positively associated with credit ratings (H3). I also hypothesize that regulatory enforcement helps to reduce information asymmetry in the market (H4). Finally, I examine whether the use of shadow insurance is associated with information asymmetry in the capital markets (H5). To test these hypotheses, I develop a measure of regulatory enforcement based on general theories of regulation from political science, sociology, law, and economics, which argue that regulatory enforcement is characterized by enforcement capacity (i.e., the regulator’s ability to exert control over regulated entities), enforcement style (i.e. how a regulator interacts with regulated entities and enforces standards), and the broader political environment (Carrigan and Harrington 2015). Primary tests of my first two hypotheses (H1 and H2) are based on a Tobit model that examines the use of captive reinsurance by sixty two large, public U.S. insurance groups between 2006 and 2015. The sample includes pure property-casualty (P/C), pure lifehealth (L/H), and diversified (i.e., both P/C and L/H) insurers. Since the debate about the risks 5 imposed by captive reinsurance has been in the life insurance sector, I divide my sample into two parts: “Life” insurers include both pure L/H insurance groups and diversified insurers (i.e., thirty two insurance groups), and “P/C” insurers include only pure property-casualty insurance groups (i.e., thirty insurance groups). The tests of my credit rating hypothesis (H3) are based on the S&P credit ratings for these insurance groups between 2006 and 2015. The tests of the information asymmetry hypotheses (H4 and H5) are based on the market liquidity tests for these insurance groups between 2006 and 2015. Among life insurers, I find evidence supporting H1, H3, and H4 and some preliminary evidence supporting H2. Consistent with my first hypothesis (H1), regulatory enforcement is negatively associated with the use of captive reinsurance subsidiaries among life insurers. That is, regulatory enforcement can constrain a parent’s incentives to use off-balance sheet entities. I find some evidence of a direct positive relation between regulatory enforcement and credit ratings (H3). However, there is a weak evidence of a negative association between regulatory enforcement and market illiquidity (H4). I do not find support for H5 (i.e., a positive association between captive reinsurance and market illiquidity) among life insurers. In my tests of H2, consistent with the general theories of regulation and prior research, I find that a firm’s characteristic (i.e., in this setting, leverage) interacts with regulatory enforcement (i.e., regulatory enforcement is state-dependent). However, my model is based on ad-hoc threshold levels of leverage and some independent variables are highly correlated. Due to multi-collinearity between the variables of interest and possible model misspecification, I cannot draw conclusions regarding the role of an insurer’s financial position in regulatory enforcement of reinsurance standards. All results on H2 should be interpreted with caution. 6 Among P/C insurers, I do not find support for H1 and H3 – H5, but there is some preliminary evidence supporting H2. However, similar to life insurers, the model testing H2 is possibly misspecified and there is multicollinearity in the model. Furthermore, my statistical tests could lack power (e.g., due to the unreliable measure of captive reinsurance or regulatory enforcement). Alternatively, regulatory enforcement can vary across firm types. Since the use of captive reinsurance by the P/C insurers has not been scrutinized by the regulators or media, the role of regulatory enforcement could be limited in the P/C insurer sub-sample. I find that P/C insurers with a surplus constraint (H2) due to high premiums (i.e., premiums-to-surplus above the 90th quantile), which are facing greater regulatory enforcement, are less likely to use captives. This result, once again, is inconclusive and needs to be interpreted with caution. Finally, contrary to what I predict, the use of captive reinsurance subsidiaries is negatively associated with proxies for information asymmetry in equity markets (H5) among P/C insurers. This result indicates that investors or other external monitors may substitute regulatory monitoring efforts for firms that may receive less regulatory attention. This dissertation makes at least three contributions to the literature. First, it contributes to the debate about the role of regulatory enforcement in an accounting context. My findings are of interest to regulators, insurance policyholders, creditors, and academics. I complement the literature on firms’ off-balance sheet financing (Shevlin 1987; Ely 1995; Beatty et al. 1995; Altamuro 2006; Dechow and Shakespeare 2009; Feng et al. 2009; Zechman 2010) and the effect of different regulators on firm level choice and outcomes (Bushman and Williams 2012; Huizinga and Laeven 2012; Acharya et al. 2013; Nicoletti 2015; Costello et al. 2016; Gallemore 2016). Regulatory enforcement constrains incentives to use off-balance entities and is contingent on a firm’s financial position. Sub-sample analyses are consistent with prior papers in economics 7 that find that institutional features and incentives can result in seemingly inconsistent implementation of regulation (Mishkin 2000; Weinberg 2002; Agarwal et al. 2014). My analyses indicate that regulatory enforcement can vary across firm types (i.e., property-casualty versus life insurers) and time periods (i.e., low versus high public scrutiny). The analyses also suggest that public scrutiny or awareness of accounting issues may be important for regulatory enforcement of accounting standards and regulatory enforcement credibility. Second, this paper is related to the growing literature on the incorporation of hard and soft information by credit rating agencies (Bozanic and Kraft 2014; Kraft 2015). Information varies in its degree of hardness or softness, and differential resources used in the oversight and verification of hard information can result in the differential hardness of financial reports (Bertomeu and Marinovic 2016). Consistent with the theory of hard and soft information, I find some evidence that regulatory enforcement provides “soft” information on regulatory efforts to monitor firm risk. I find that the role of regulatory enforcement as an information source can vary across firm types and time periods. I find stronger evidence for my hypothesis (H3) in the time period characterized by increased public scrutiny of the use of captive reinsurance by life insurers (i.e., 2012 – 2015). This finding implies that public awareness of accounting issues and regulatory efforts could be important for regulatory enforcement and its credibility. Finally, this paper contributes to the debate on the role of the institutional environment in capital markets (Leuz and Wysocki 2016). I find some evidence that regulatory enforcement is negatively associated with information asymmetry. This result, once again, emphasizes the importance of the institutional environment. Contrary to my expectations, I find that the use of captive reinsurance doesn’t result in increased market illiquidity. Instead, there is some evidence that captive reinsurance is negatively associated with market illiquidity among pure P/C insurers. 8 There are important limitations to this study. I examine a small sample of large public U.S. insurers that have affiliates in their structure. My findings may not generalize to other settings (e.g., smaller insurance companies and private insurers). In addition, my research design does not definitively rule out endogeneity concerns. While I do include main regulator fixed effects to control for common regulatory enforcement and firms’ self-selection into a regulatory environment, I cannot rule out an omitted variable bias. I do, however, use various robustness checks including a falsification test, an instrumental variable estimation, simultaneous equations modelling, and propensity score matching. Finally, some of the models include highly collinear independent variables and could be misspecified. All results on H2 should be interpreted with caution. The remainder of the dissertation is organized as follows. Chapter 2 provides background on insurance regulation and captive reinsurance and develops the hypotheses. Chapter 3 describes the sample and research design. Chapter 4 presents the main results, and Chapter 5 provides supplemental analyses. Chapter 6 concludes and discusses limitations and future work. 9 CHAPTER 2: BACKGROUND AND HYPOTHESIS DEVELOPMENT I begin this chapter by discussing the nature of U.S. insurance regulation. Then, I describe the goals of insurance regulation and how they are reflected in insurance industry accounting standards. Next, I discuss the differences between traditional and captive reinsurance and explain the incentives for vertical integration of the reinsurance function through captive reinsurance. Finally, I discuss prior literature and develop my hypotheses. 2.1. U.S. Insurance Regulation U.S. insurance regulation has its historic origins in the 1800s (Klein 2005). As a result of the Paul v. Virginia ruling by the U.S. Supreme Court in 1869, the U.S. insurance industry is regulated at the state level. To increase efficiency and reduce redundancy inherent in a system of 56 regulatory bodies, U.S. insurance regulators coordinate their efforts, often through the National Association of Insurance Commissioners (NAIC). Many statutory accounting standards are uniform across states, and insurance regulators freely share information (except for captive reinsurer data) with each other through a centralized financial database. Although insurance firms must meet the regulatory requirements of each state in which they operate, states can defer insurance regulation to the firm’s domiciliary insurance regulator. Insurance group supervision is based on the NAIC model law adopted in 1969, with subsequent revisions (NAIC 2012a). This model law applies to groups of two or more affiliates, at least one of which is an insurer. Each insurance group is assigned a lead (main) regulator, who coordinates enforcement efforts and cooperation among the domiciliary insurance regulators who monitor the various insurance affiliates in the group. Insurance group supervision is based on the “windows and walls” approach (NAIC 2012a). Regulators have “windows” to scrutinize an insurance group’s activities (e.g., a shared information database, NAIC Schedule Y which 10 provides a holding company organizational chart, and the right to examine the subsidiaries in the insurance group) and “walls” to protect surplus by requiring regulatory approval of material related-party transactions such as reinsurance and management agreements, cost sharing, investment purchases and intercompany investments, extraordinary dividends, and tax-allocation agreements. The protection of surplus is one of the key objectives of insurance regulation. 2.2. Solvency Regulation in the Insurance Industry State insurance regulation is primarily focused on protecting policyholders from losses by ensuring the solvency, i.e., capital adequacy, of insurance companies (Galloway and Galloway 1986). Regulation of solvency is accomplished by conservative statutory accounting standards, cross-state review of insurers’ financial position with the domiciliary state taking a lead position, risk-focused examinations, actuarial certification of policy reserves, and capital requirements. Insurance regulators often impose capital adequacy, solvency, and liquidity requirements and restrictions on market entry, business activities, and investments (Galloway and Galloway 1986). 2.2.1. Statutory Accounting Principles (SAP) and Loss Reserving Statutory accounting principles (SAP) are established by state insurance regulators and are required by state law. Since insurance is regulated at the state level, insurance regulators can set statutory accounting standards in their state that deviate from standards set in other state insurance jurisdictions. Also, insurers can request a permission from the domiciliary state regulator to depart from SAP. In general, insurers use SAP to report to insurance regulators and the Internal Revenue Service (IRS). Public insurers use US GAAP to report to the Securities and Exchange Commission (SEC). Some statutory accounting principles (SAP) differ from US GAAP. SAP and US GAAP provide similar information about an insurer’s performance. But in contrast to US GAAP, SAP 11 aims to reveal solvency and thus focuses more on the balance sheet than the income statement. Under SAP, insurers report more detailed financial statements, including a section on the Capital and Surplus account. The financial statements are presented on a net of reinsurance basis. Some assets are “non-admitted” under SAP and are assigned a zero value. Examples of non-admitted assets include: prepaid expenses, furniture and equipment, accounts receivable overdue 90 or more days, portions of deferred tax assets, and goodwill. Also, SAP results in a mismatch between acquisition expenses and revenues because acquisition costs are expensed as incurred, while premium revenues are deferred. As a result, insurers need “free” surplus to finance their future sales. Under SAP, surplus equals “admitted” assets less liabilities. Statutory loss reserves usually represent the largest liability on insurers’ balance sheets. Insurers can have multiple technical (statutory) loss reserves. In general, loss reserves are estimates of liabilities for future policyholder benefits or claims that reflect expectations, managers’ discretion, and the quality of the formula used to estimate the losses. For property-casualty (P/C) insurers, loss reserves represent an estimate of incurred losses: both reported losses and unreported losses (i.e., incurred but not reported; future policyholder claims). For life-health (L/H) insurers, loss reserves can include reserves for incurred losses as well as actuarial reserves for future policyholder benefits on long-duration products (e.g., annuities, whole life insurance products). Loss reserves on longduration products (e.g., FAS 60) are determined using a prospective method as the difference between the present value of future insurance benefits and the present value of expected future premiums (i.e., discounted loss reserves are affected by interest rates and mortality assumptions). Both under-reserving and over-reserving are costly to the firm. Under-reserving increases insolvency and bankruptcy risk and is especially costly in the case of correlated risks. Over- 12 reserving is costly to the firm as it limits surplus available for investment and can result in increased premiums (i.e., premiums are based on a firm’s loss experience: over-reserving implies greater loss experience and hence higher insurance premiums in the future). Since loss reserves affect a firm’s reported profitability, surplus, and taxes, managers have incentives to manage loss reserves through loss reserve accounting or reinsurance accounting. Prior research finds that P/C insurers use discretion in loss reserves accounting to achieve reporting goals (Petroni 1992; Petroni and Shackelford 1995; Adiel 1996). However, in the life insurance sector, in contrast to the P/C lines of business, there is limited discretion in loss reserve accounting standards. As a result, life insurers would need to rely on discretion in other accounting standards (e.g., reinsurance accounting) in order to relieve a capital constraint due to excessive reserves. 2.2.2. Risk-Based Capital (RBC) To monitor the capital adequacy of insurers, regulators use a Risk-Based Capital (RBC) ratio in combination with other monitoring tools (e.g., Insurance Regulatory Information Systems (IRIS) ratios, Financial Analysis and Surveillance Tracking (FAST) scores, inspections). The RBC ratio is not designed to compare insurers and is used by regulators to identify weakly capitalized companies. Approximately, the RBC ratio compares insurers’ adjusted surplus to the Company Action Level (CAL) risk-based capital.4 Required minimum capital is calculated using a riskbased formula established by the regulator and reflects material risks to which an insurer has 4 In general, the RBC framework focuses on asset risk (affiliates and others), underwriting risk, and other risks to calculate the CAL risk-based capital (American Academy of Actuaries 2014). Life, property-casualty, health, and fraternal insurers have different RBC formulas. For example, life CAL capital is based on risks from affiliates, investment, interest, claims, and general business risks and excludes immaterial risks (i.e., short term), tail risks, and risks that cannot be pre-funded (e.g., liquidity risk). Life RBC calculates a post-tax amount while P/C and Health RBC formulas measure a pre-tax amount. Life RBC is measured at the legal entity level, and there are no requirements to calculate the group level RBC. 13 exposure. The RBC formula establishes the CAL capital which acts like a trigger in solvency regulation. When Total Adjusted Capital (TAC) falls below CAL, it triggers a regulatory action. Regulatory actions (e.g., an RBC plan, liquidation, regulatory control) depend on the ratio between TAC and CAL. For example, life RBC ratio is approximately as follows (American Academy of Actuaries 2014): 𝑅𝐵𝐶 = 𝑇𝐴𝐶 𝐶𝐴𝐿 + 𝑓𝑟𝑒𝑒 𝑠𝑢𝑟𝑝𝑙𝑢𝑠 + 𝑎𝑠𝑠𝑒𝑡 𝑣𝑎𝑙𝑢𝑎𝑡𝑖𝑜𝑛 𝑟𝑒𝑠𝑒𝑟𝑣𝑒 + 0.5 𝑑𝑖𝑣𝑖𝑑𝑒𝑛𝑡 𝑙𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑦 = 1 1 𝐶𝐴𝐿 2 2 𝐶𝐴𝐿 The RBC formula assumes that insurers’ policyholder loss reserves are sufficient to cover expected losses under moderately adverse conditions (e.g., 83th percentile for normally distributed risks) and establishes a requirement for additional capital necessary to sustain losses that would arise under more adverse conditions. Statutory reserves and minimum capital requirements are expected to be sufficient to protect insurer solvency 95% of the time. To improve their RBC ratios, insurers can enhance their surplus through structured finance, investments from a parent company (e.g., cash infusion into surplus), and reinsurance (Appendix A). Alternatively, insurers can restructure their liabilities by reducing excess liabilities, writing liabilities that use the properties of the RBC’s covariance formula, using pooling or reinsurance, or by reducing growth in surplus-intensive insurance products. Insurers can reorganize their affiliates and can restructure their asset portfolio to include higher quality assets or increase portfolio diversification. Overall, product design, asset type and allocation, and liability management strategies (e.g., reinsurance) can be used to improve the RBC ratio. 14 2.3. Reinsurance and the Captive Reinsurance Market 2.3.1. Traditional vs. Captive Reinsurance Insurance firms accept risk from market participants in exchange for a premium. If an insurer underprices a risk or accepts risks with high uncertainty, the insurer can incur distress costs such as volatile income and insolvency. To manage their risk exposure, insurance firms often transfer a part of their risk to other market participants such as reinsurers or investors. Thus, reinsurance is insurance coverage for insurers. Reinsurance allows insurers to increase their underwriting capacity (i.e., increase “free” surplus) and stabilize underwriting results. Reinsurance can also allow insurers to manage their statutory reserve levels, since reinsurance allows firms to reduce their net loss reserves and, thus, increase surplus (Appendix A). Insurance firms can use both traditional (third-party) and non-traditional (affiliated) reinsurance for risk management purposes. Since there is no requirement for consolidation under SAP, the use of affiliated reinsurance results in the same statutory reporting outcomes as equivalent unaffiliated reinsurance. The non-consolidation of affiliated reinsurance allows insurers to transfer reserves to their affiliates, if affiliated reinsurance meets the reinsurance accounting standards. Regulators can allow affiliated reinsurance because, statistically, affiliated reinsurance can be used to manage risk, even though insurance affiliates belong to the same economic entity. For example, risk pooling can allow actuaries to better estimate expected losses (i.e., rely on the law of large numbers), and hence, premiums can more accurately price underwriting risk. Nevertheless, regulators still need to examine reinsurance contracts to assess risk transfer and use regulatory tools to monitor a firm’s reinsurance choices. To monitor firms’ reinsurance transactions, insurance regulators set licensing standards for reinsurers. If reinsurance is purchased from a reinsurer authorized by the regulator, assurance is automatically assumed. Unauthorized reinsurance requires regulatory approval. A captive 15 reinsurance license usually differs from a traditional reinsurance license, and as a result, captive reinsurance is usually unauthorized and requires regulatory approval. A captive reinsurer is a reinsurance subsidiary licensed under captive insurance laws. Captive reinsurers are formed to provide reinsurance services to their parent and affiliates. Captives typically are financed through a parental guarantee, a letter of credit from a bank, or a surplus note issued to investors.5 Thus, captive reinsurers can be “isolated” or can have exposure to capital markets. The transaction that formed the captive must be filed with the parent’s regulator(s), but is considered approved if not disapproved within a specified period of time. To receive a credit for reinsurance (i.e., to reduce net loss reserves), each captive reinsurance transaction is reviewed by the ceding insurer’s domiciliary regulator and the captive’s regulator to ensure that the transaction meets regulatory reinsurance standards. Captive reinsurance agreements used by an insurance group can be reviewed by the lead state regulator and other domiciliary state regulators monitoring the group. The requirement for regulatory approval of captive reinsurance is due to the differences between captive and traditional reinsurers in their capital, reporting, and disclosure requirements. Captive insurance companies were originally created by non-insurance companies to insure risks that could not be covered by conventional insurance at reasonable cost. Since harm from insolvency of such a captive directly impacts only the non-insurance parent, captive solvency regulation is often lax. For regulatory reporting purposes, captive reinsurers can use US GAAP rather than SAP. As a result, captives can recognize assets on their balance sheets that are “nonadmitted” under statutory accounting rules. For instance, captive reinsurers can use contingent 5 Society of Actuaries 2014 (see References). 16 notes, parental guarantees, and deferred tax assets to support their reserves. In 2014, data released by the state of Iowa for eight captive reinsurers indicated an aggregate US GAAP surplus of $1.5 billion, in contrast to an estimated $2.7 billion deficit under statutory accounting rules (Koijen and Yogo 2016). Finally, captive reinsurers privately report only to their (captive) regulator, and captive insurance laws can limit access to a captive’s financial reports even to traditional insurance regulators. The significant growth in the use of captive reinsurance among life insurers has attracted attention from the media and regulators. In 2012, life insurers transferred an estimated $364 billion of liabilities to captive reinsurers in comparison to $11 billion transferred in 2002 (Koijen and Yogo 2016). The accounting complexity and reporting opacity of captive transactions has also resulted in public scrutiny of captive reinsurance due to disagreement about the magnitude of risk imposed by the captive reinsurance industry.6 Strict capital standards and increased financial disclosure for captive reinsurers could reduce these risks but would also decrease capacity in the insurance industry. For example, it is estimated that in the absence of “shadow insurance”, life insurance prices would rise by eighteen percent and the market would shrink by twenty three percent (Koijen and Yogo 2014). Thus, insurance regulators need to tailor their regulatory enforcement of reinsurance standards to achieve a balance between solvency and product pricing that is acceptable to insurance market participants. Also, a firm’s economic motives for captive reinsurance can be considered in the regulatory process. 6 See, e.g., studies and articles by NY Department of Financial Services (2013), the Financial Stability Oversight Council (2016), and the Federal Insurance Office (2013a). 17 2.3.2. Incentives for Captive Reinsurance The choice to write a reinsurance contract with unaffiliated entities versus an affiliated captive reinsurer will depend on differences in the expected net benefits of the two options. In traditional reinsurance, an insurance firm shares its risk with an unaffiliated insurance company in exchange for a fee. Third-party reinsurance contracting involves various direct and indirect transaction costs. Ex ante reinsurance transaction costs involve reinsurer search and evaluation costs, negotiation, contract design, and reinsurance fees. Ex post transaction costs arise from reinsurers’ credit risk and increased moral hazard risk. Insurers can use affiliated captive reinsurance to reduce transaction costs. Incomplete contracting and the associated moral hazard risk due to reinsurance can be reduced by writing reinsurance contracts with affiliates. Reinsurance contracting with affiliates can be more efficient than third-party reinsurance due to lower information asymmetry between affiliated contracting parties, implying more efficient pricing of underwriting risk and counterparty credit risk. In addition, transactions with affiliates can lower search and negotiation costs and can change the bargaining position of the firm in the reinsurance market. However, ex ante contract design costs may increase, as risk sharing with affiliates has to be properly structured in order to achieve a statistical risk distribution. An example is additional reinsurance underwriting costs arising from the purchase of specialized services (e.g., actuaries, lawyers, accountants, captive managers). In addition to reduced contracting costs, affiliated captive reinsurance can help firms reduce their tax liability by reducing the variability of pre-tax firm values when a firm’s tax function is convex (Smith and Stulz 1985). Also, reinsurance premiums are deductible expenses and thus can be used to lower pretax income. Historically, the IRS disallowed reinsurance premium deductions in parent-child or brother-sister captive structures due to the economic family theory. As courts rejected this theory, the IRS has allowed deduction of reinsurance 18 premiums when there is sufficient third party risk. While there is no bright line percentage test for third party risk, in the Harper Group (1991) ruling, the Tax Court found that 29% unrelated risk should be sufficient for risk distribution. Even if there are no outside policyholders, reinsurance among affiliates can qualify for tax deductibility if risk shifting and risk distribution are present (Humana Inc, 881 F.2d 247 (1989); Kidde Industries v US, 40 Fed Cl (1997)). IRS Revenue Ruling 2002-90 established a safe harbor of twelve affiliates for deductibility of captive reinsurance premiums. Finally, captive reinsurance jurisdictions usually offer favorable tax treatment of captive reinsurer income. Many captive reinsurer jurisdictions impose no premium taxes (e.g., Arizona) and no income or capital gain taxes (e.g., Bermuda, Cayman)7. In a concurrent working paper, Hepfer et al. (2016) find that life insurers with captive reinsurers have lower ex post GAAP ETRs. In addition to contracting and tax costs, regulatory costs can create incentives for the vertical integration of reinsurance. Strict or inflexible regulation can impose high compliance and non-compliance costs on regulated firms, and these firms then have an incentive to take advantage of arbitrage opportunities. Regulatory arbitrage is a change in the structure of firm activities that reduces the cost of regulation. That is, regulated firms will attempt to find loopholes in the regulatory system to circumvent unfavorable regulation. If insurers find reserving or pricing standards unfavorable, they can exit the market, bear the regulatory cost, or use a loophole in regulation to reduce this cost. For instance, if firms are required to hold excess reserves for certain products, they can stop selling those products, accept regulatory reserve requirements, or reduce reserves through a loophole. In the last case, insurers 7 If a captive reinsurance subsidiary is located offshore, insurers might need to pay additional taxes. For instance, insurance companies need to pay federal excise taxes on reinsurance premiums paid to foreign reinsurers (i.e., 1%), unless the firm chooses the 953(d) election for its CFC (controlled foreign corporation) captive reinsurer. 19 can use discretion in loss reserve accounting or reinsurance accounting standards, which differ across risks and hence firm types. To decrease the cost of financing excess reserves through reinsurance, insurers can reinsure with affiliated captive reinsurers. However, in contrast to reinsurance with third parties or affiliated traditional insurers, affiliated captive reinsurance requires regulatory approval. 2.4. Hypothesis Development There is a growing literature in accounting that examines the role of the regulatory environment in firms’ decision making. There is an extent literature on the role of auditors, analysts, and credit rating agencies as external monitors. Currently, there is a growing interest in understanding enforcement and monitoring by the SEC and the IRS, as well as how those regulatory bodies affect firm behavior (Kedia and Rajgopal 2011; Ettredge et al. 2011; Robinson et al. 2011; Hoopes et al. 2012; Cassell et al. 2013; Hanlon et al. 2014; Bens et al. 2016). These studies show that firms anticipate regulatory enforcement and change their behavior in response to a regulator’s enforcement choice. Also, regulatory enforcement can change the information risk of accounting reports. For example, Dechow et al. (2016) find that firm managers opportunistically sell shares in anticipation of SEC comment letters on revenue recognition. Johnson and Petacchi (2014) show that earnings response coefficients increase and stock return volatility decreases around earnings announcements following the resolution of the SEC comment letters. The resolution of tax-related SEC comment letters is negatively associated with future tax avoidance (Kubick et al. 2016). Similarly, there is empirical evidence that IRS enforcement reduces information asymmetry and incentives for tax avoidance (Guedhami and Pittman 2008; El Ghoul et al. 2011). 20 Related research in banking and health care examines how different (multiple) regulators shape firms’ reporting. The general inference is that regulators are heterogeneous due to institutional features and incentives and that enforcement depends on regulators’ objectives. Regulators can be heterogeneous in their enforcement actions (Agarwal et al. 2014) and can selectively decouple and exhibit leniency in their enforcement (Heese et al. 2016). 2.4.1. Regulatory Enforcement and Captive Reinsurance To develop my regulatory enforcement hypotheses, I rely on the NIE framework and the general theories of regulatory enforcement from economics, political science, law, and sociology. The NIE provides a general framework where formal institutions and their enforcement determine the incentives and constraints of economic actors and thus shape economic outcomes (North 1992; Williamson 2000; Acemoglu et al. 2004). “In the jargon of the economist, institutions define and limit the set of choices of individuals. Institutional constraints include both what individuals are prohibited from doing and, sometimes, under what conditions some individuals are permitted to undertake certain activities. ... They are perfectly analogous to the rules of the game in a competitive team sport” (North 1990, pp. 3-4). The NIE framework allows me to draw inferences about the role of accounting regulation, a formal institution, on a firm’s governance choice (i.e., the vertical integration of the reinsurance function). Accounting regulation consists of accounting standards and their enforcement. Accounting standards define and limit the set of choices firms can take. Accounting rules are a mapping of transaction characteristics into an accounting report (Gao 2013), and transaction characteristics can vary with a transaction type (Commons 1924). Accounting standards can be flexible to reflect transaction complexity and allow managerial judgement. To shape firm behavior and compliance, accounting standards can include various 21 constraints or opportunities. For example, accounting standards can have exemptions or specific guidance that differs across entities. Accounting rules can also define the circumstances under which a certain activity will be permitted. In general, accounting rules result in a binary outcome: ex post compliance or non-compliance with a standard. Prior literature on compliance and non-compliance spans many research areas and is too extensive to review here (e.g., see Scholz 1984; Edelman and Suchman 1997; Oded 2013). In general, standards determine compliance and non-compliance costs and thus can incentivize entities’ compliance or non-compliance with standards. Strict (precise) standards result in a narrow range of compliance choices and a wide range of non-compliance behaviors, and thus, result in a wide range of probable penalties (i.e., assuming the penalty is not fixed). Imprecise standards can result in a “shadow region” chosen by the firm, which can be arguably either in the compliance or non-compliance region of acceptable behaviors. Also, ex ante conservative rules can limit managers’ opportunities to “inflate transaction characteristics” (Gao 2013). Therefore, strict (imprecise) accounting standards reduce (increase) the incentives for non-compliance by increasing (decreasing) the expected non-compliance costs, which can be substantial (Karpoff et al. 2008). The enforcement of accounting rules is also important as it determines expected noncompliance costs. Enforcement is defined as regulators’ monitoring, inspection, and actual enforcement activities (e.g., warnings, fines) that aim to achieve regulatory outcomes (OECD 2014). Regulatory detection efforts (i.e., monitoring and inspection) and enforcement incentives affect the optimal probability of detecting a violation and imposing a penalty for noncompliance. That is, regulatory enforcement affects a firm’s expected non-compliance costs (NCC) by determining the probability of detection (p(D)) and the probability of enforcing (p(E)) 22 a certain penalty (P). Thus, ex ante non-compliance costs can be determined as follows: E (NCC) = P * p (D) * p (E). Non-compliance costs vary with the penalty choice (P), which can be informal or formal. Regulatory penalties can range from regulatory non-approval and warnings to fines and criminal prosecutions. Penalty structure can include fixed, variable, or both components. The probability of detection (p(D)) and the probability of enforcement (p(E)) depend on regulators’ monitoring, inspection, and enforcement activities, which can be targeted, i.e., statedependent (Landsberger and Meilijson 1982; Greenberg 1984; Harrington 1988). For instance, regulators can condition their enforcement choice on a firm’s reputation for non-compliance (Malik 2014), the severity of violation, or regulatory objectives, etc. Competing priorities and external environment pressures can change the incentives to enforce accounting standards. Regulators’ beliefs in the adequacy of accounting standards or in their ability to verify noncompliance with those rules could also shape regulators’ incentives to enforce standards. Depending on the standard type, penalty structure, and institutional environment, accounting standards and regulatory enforcement can be either substitutes or complements (Laux and Stocken 2014). For example, when penalties are variable (fixed), accounting standards and regulatory enforcement are substitutes (complements) in their effect on misreporting. In the context of captive reinsurance, accounting standards (i.e., reinsurance accounting) are very subjective and imprecise, allowing an opportunity to establish a lower “shadow threshold” for compliance purposes. To qualify for reinsurance accounting, a reinsurance contract has to transfer insurance risk. A reinsurance contract might not qualify for reinsurance accounting if the contract transfers insurance risk at the individual contract level, but not at the aggregate level when all reinsurance contracts are considered, and vice versa. Also, reinsurance 23 contracts can have provisions that might limit risk transfer. Thus, regulators need to examine all reinsurance contracts with a reinsurer to ensure the compliance with risk transfer requirements. This, in turn, allows insurers to establish a “shadow threshold” in their compliance with reinsurance accounting standards. As the result of subjectivity inherent in reinsurance accounting standards, the expected non-compliance costs associated with reinsurance accounting standards will vary with firm characteristics. While direct penalties (i.e., fines) are likely small, the indirect “penalty” (i.e., regulatory disapproval of a captive transaction and hence the requirement to put the transferred loss reserves back on the balance sheet of the parent company) could be large. Prior literature finds that ambiguous standards create discretion to report aggressively (Beatty and Weber 2006; Dechow et al. 2010; Blacconiere et al. 2011, Bratten et al. 2013) unless regulatory enforcement constrains aggressive reporting by increasing non-compliance costs. Since captive reinsurance transactions require regulatory approval, the probability of regulatory detection and enforcement of compliance with reinsurance accounting standards can be high. I hypothesize that when regulatory enforcement is strong and hence the probability of detection and enforcement is high, the expected non-compliance costs (i.e., fines, disapproval of captive reinsurance transactions, bad reputation) are high and thus a firm is less likely to use captive reinsurance transactions. In other words, regulatory enforcement is negatively associated with the use of captive reinsurers. My first hypothesis is as follows (i.e., in the alternative form): Hypothesis 1: Ceteris paribus, the number of affiliated captive reinsurers is negatively associated with regulatory enforcement. 24 Since captive reinsurance transactions are complex and opaque, regulatory approval of captive reinsurance is inherently subjective. Regulatory enforcement actions towards captive reinsurance are likely state-dependent and thus can vary across firm types. Since captive reinsurance affects a firm’s solvency, regulators will likely scrutinize captive reinsurance based on a firm’s leverage position. Anecdotal evidence suggests that captive reinsurance has been associated with regulatory arbitrage incentives to circumvent statutory reserve standards in the life insurance sector. Insurance regulators use assets-to-surplus to measure life insurers’ leverage (Federal Insurance Office 2013B). High leverage indicates that a life insurer has a greater exposure to estimation errors (e.g., long-tail risks) and thus has to rely on having adequate reserve funds. Life insurers would need to have sufficient reserves before captive reinsurance could be used. I hypothesize that very high or very low leverage could attract regulatory attention to the use of captive reinsurance and thus could result in anticipated greater regulatory enforcement of reinsurance accounting standards. Surplus constraints can also result from rapid growth. Prior research and numerous case studies have shown that rapid growth rates can result in financial problems, including bankruptcy, among P/C insurers (Fu 2012). Insurance regulators monitor premium-to-surplus ratios among P/C insurers (NAIC 2016). Insurers that issue long-tail risks are expected, in general, to maintain lower gross premiums-to-surplus and net premium-to-surplus ratios because it is more difficult to estimate losses for products with the long-tail risk (i.e., there is a greater variability of losses on these products) (NAIC 2016). Thus, a high premium-to-surplus ratio could attract regulatory attention to the use of captive reinsurance by P/C insurers (i.e., captive reinsurance affects reported net reserves and net premiums under SAP). 25 Since surplus monitoring is important in insurance solvency regulation, insurance regulators are likely to be critical of captive reinsurance for firms with potentially constrained surplus due to insolvency risk considerations. Thus, my second hypothesis is as follows (in the alternative form): Hypothesis 2: Ceteris paribus, the negative association between captive reinsurance and regulatory enforcement is stronger for firms with potentially constrained surplus. 2.4.2. Regulatory Enforcement, Captive Reinsurance, and Credit Ratings In addition to insurance regulators, credit rating agencies evaluate risks associated with captive reinsurance. Since captive transactions are complex and opaque, credit rating agencies will likely use subjectivity in their assessment of captive reinsurance risks. For example, in 2014, the S&P credit rating agency issued proposed guidance on its approach to evaluation of captive reinsurance. S&P examines captive reinsurance based on the economic view of the entity (Society of Actuaries 2014). The adjustments to the entity’s statutory statements depend on regulatory approval of captive reinsurance credits. Similarly, Moody’s adjusts credit ratings to reflect aggressive accounting, management quality, governance risk, etc. (Moody’s 2007). Thus, credit rating agencies may infer “soft” information on firm risk based on regulatory enforcement environment. Both hard (quantifiable) and soft (qualitative) information can be used in the evaluation of firm risk and performance. Kraft (2015) finds that credit rating agencies use both hard and soft information to better evaluate firms’ default risks. However, the responsiveness of credit ratings can be limited. Post-issuance credit rating monitoring can be lax, especially in the presence of off-balance sheet items (Bonsall et al. 2015). 26 Credit rating agencies can rely on the monitoring and information intermediation efforts of other external monitors. Information acquisition efforts can depend on information available from other regulators (Bozanic et al. 2016). Cheng and Subramanyam (2008) find that the credit ratings of non-financial firms are associated with analyst following. However, in the captive reinsurance setting, analysts are an unlikely source of additional information due to the opacity of captive reinsurers. Regulators, on the other hand, have (limited) access to information on the “shadow insurance” market as well as the power to enforce reinsurance accounting standards and disallow captive reinsurance. Thus, credit rating agencies can infer soft information on a firm’s default risk based on regulatory enforcement. Regulatory enforcement can decrease risks imposed by regulated entities through regulatory detection and actual enforcement actions. Thus, my third hypothesis is as follows (in the alternative form): Hypothesis 3: Ceteris paribus, credit ratings are positively associated with regulatory enforcement. 2.4.3. Regulatory Enforcement, Captive Reinsurance, and Capital Markets Accounting standards and their enforcement are also important in equity markets. Accounting regulation can influence the level of information asymmetry between managers and investors (Healy and Palepu 2001, Beyer et al. 2010), as well as analyst information processing (Asbaugh and Pincus 2001, Wang et al. 2008, Tan et al. 2010). The quality of financial information is a function of both accounting standards and regulatory enforcement (Sunder 1997, Kothari 2000). Regulatory enforcement can reduce the instances of financial reporting-related fraud and thus increase the reliability of financial reports (Ball 2001). For example, the increased reliability 27 of reports can reduce financial analysts’ uncertainty about the accounting methods used and thus make the task of forecasting earnings easier (Hope 2003). Also, regulatory enforcement can increase the credibility of reports based on imprecise standards (Kolev 2013; Bens et al. 2016). Furthermore, by preventing aggressive reporting, regulatory enforcement results in reports that inform investors about the lower bound (worst-case scenario), which increases liquidity because buyers are informed about credible minimum bids (Lunawat et al. 2014). Based on the prior literature, I hypothesize that regulatory enforcement can increase the credibility of financial reports and thus reduce information asymmetry. My fourth hypothesis is as follows (in the alternative form): Hypothesis 4: Ceteris paribus, market illiquidity is negatively associated with regulatory enforcement. In addition to regulatory enforcement, market liquidity and the information environment are affected by the quality of accounting information. However, accounting reporting quality is a complex construct with multiple dimensions that include such attributes as earnings quality, disclosure quality, comparability, consistency, reliability, relevance, and various types of complexity (e.g., accounting, linguistic). In the case of shadow insurance, I argue that the use of captive reinsurance transactions can change firm’s earnings quality, the reliability and comparability (i.e., the precision of across-firm information) of accounting reports, and information complexity. Captive reinsurance transactions impact a firm’s estimation of loss reserves and hence the firm’s earnings quality (i.e., captive reinsurance can be used to reduce the variability of estimated losses and hence to increase the “smoothness” of reported earnings). Captive reinsurance also can change the level of reliability, comparability, and complexity by 28 changing uncertainty about users’ information endowment. “Shadow” insurance transactions are complex and can represent information risk. Prior literature finds that higher reporting quality is positively associated with firm valuations and liquidity (Lang et al. 2012) and smaller analyst forecast errors and dispersion (Behn et al. 2008). Furthermore, accounting reporting comparability is associated with better information processing by analysts (De Franco et al. 2011, Horton et al. 2013, Peterson et al. 2015) and lenders (Kim et al. 2013; Fang et al. 2016), market trading around earnings restatements (Campbell and Yeung 2016), reduced insider ability to exploit private information (Brochet et al. 2013), and improved market outcomes (Neel 2017). Information complexity, on the other hand, is negatively associated with analyst forecasting ability (Barth et al. 2001, Gu and Wang 2005, Hodder et al. 2008). Also, opportunities to exploit private information can result in lower analyst coverage (Bushman et al. 2005). Since captive reinsurance transactions can decrease report reliability and comparability, increase information complexity, and can be used by management to obfuscate value-relevant information, I hypothesize that the use of captive reinsurance subsidiaries can result in higher market illiquidity as well as a lower analyst following.8 The fifth hypothesis as follows (i.e., in the alternative form): Hypothesis 5: Ceteris paribus, market illiquidity is positively associated with captive reinsurance. 8 Captive reinsurance can represent an information risk due to statutory reporting opacity, which could result in the obfuscation of value-relevant information on firm risk and cash flows affected by captive reinsurance (e.g., parental guarantees or letters-of-credit used to form a captive). 29 CHAPTER 3: SAMPLE SELECTION AND RESEARCH DESIGN Chapter 3 discusses sample selection and research design. First, I describe the sample used in this dissertation. Then, I describe the measurement of my independent variables, i.e., regulatory enforcement and captive reinsurance. Next, I present the empirical model used to test H1 and H2, which predict a negative association between regulatory enforcement and captive reinsurance, especially for firms with potentially constrained surplus. Then, I present the empirical model used to test H3, which predicts a positive association between regulatory enforcement and credit ratings. Next, I present the empirical model used to test H4 and H5, which predict a negative association between regulatory enforcement and information asymmetry and a positive association between captive reinsurance and market illiquidity, accordingly. 3.1. Sample Selection I start hand-collecting firms’ organizational data from CorporateAffiliations database (i.e., provided by LexisNexis). This database has information on corporate hierarchies, management, and board of directors for 1.9 million companies in the United States and internationally (as of 2017), both private and public. However, organizational data for private companies is limited, and as the result, I include only public companies in my sample. For public companies, CorporateAffiliations database has detailed organizational data (e.g., parent, subsidiary, affiliate, branch, division, group, plant) and identifies non-operating entities in a firm’s structure. This serves as a good starting point in identifying captive reinsurance subsidiaries, which are often non-operating legal (“shell”) entities. Nevertheless, I check all subsidiaries in a firm’s structure to ensure the completeness of my data. 30 I start with the one hundred twenty largest by -- gross earned premiums -- public U.S. insurance groups in 2006 and collect ten years of panel data on these insurance groups (2006 – 2015. I identify insurance companies using the SIC codes provided in the CorporateAffiliations database. Insurance companies that have the SIC code 6331 are codified as P/C insurers, while insurers that have the SIC code 6311 are codified as life insurers. I include both P/C and life insurers for completeness, even though the public scrutiny of captive reinsurance has been in the life insurance sector. Pure P/C insurers in my sample have only the SIC code 6331, while all other companies (i.e., pure life insurers and diversified insurers) are classified as “life” insurers. I start my sample in 2006 for a couple of reasons. First, there has been a substantial growth in the use of captive reinsurance among insurers only in the past ten or fifteen years. Second, I want to include firm observations both pre- and post- the public scrutiny of captive reinsurance (used by life insurers) starting in 2012. Finally, to ensure the reliability of my captive reinsurer data, I collect information in the most recent time period. There is some missing data on captive reinsurer licenses granted in the 1990s and earlier, especially in off-shore jurisdictions such as Bermuda, Cayman, Turks and Caicos, Guernsey, and Barbados. In addition to the CorporateAffiliations database, I use other data sources to ensure that I can identify all captive reinsurance subsidiaries in a firm’s structure. For each sample insurer I examine Exhibit 21 in the 10-K filings, which are available from the SEC’s Edgar database. I also use the NAIC listing of insurance groups, a free report available on the NAIC’s website. This report lists only insurance subsidiaries in the firms’ structure. Since I am interested only in insurance subsidiaries, this is a useful information source. In this dissertation I study the role of regulatory enforcement by insurance regulators, and thus, I include only insurance subsidiaries in my analyses (e.g., I do not include banking or non-financial entities, which can be subject to 31 regulation and enforcement by non-insurance regulators). Similarly, I supplement my data collection with information on the firm’s organizational structure as identified by the insurance rating agency A.M. Best (i.e., the A.M. Best Corporate Structure file is available online). Finally, since my sample includes the largest (by gross earned premiums) public insurance groups in the United States, I was able to find some statutory filings data, including Schedule Y, online. For example, some companies provide both their 10-K and statutory filings on their website. Also, I use regulatory examination reports to identify subsidiaries in the firm’s organizational structure and the existence of captive reinsurers (i.e., insurance regulatory examination reports have the Schedule Y and identify intercompany agreements, including reinsurance agreements; they are available online on state regulatory websites). Regulatory data on the insurance groups’ organizational structure and the financial resources of the groups’ regulators is obtained from the NAIC reports (i.e., NAIC products “Summary Listing of Companies”, “Insurance Department Resources Report”, and “State Insurance Regulation: Key Facts and Market Trends”). To identify captive reinsurers, I check the subsidiaries’ licenses (i.e., type, effective date, and parent) on regulatory websites. Then, I remove insurers that were acquired by another company or went through financial distress (i.e., in liquidation) between 2006 and 2015. I lose thirty seven insurance groups. Finally, I obtain financial data from the WRDS Compustat database, which has US GAAP and some statutory insurance data. Compustat has statutory data on insurers’ surplus and net income (i.e., from a footnote disclosure). Compustat has detailed US GAAP data but does not have some detailed data reported under SAP (e.g., premiums by state or line of business, reserve estimates by policy period and their revisions, and type of reinsurance). Also, there is missing statutory data. I lose twenty one insurance groups due to the missing required financial statements data. The final sample includes sixty two insurance groups 32 between 2006 and 2015 (see Appendix D). The credit ratings data is obtained from the S&P Global Market Intelligence database. The credit rating tests are based on sixty two insurance groups between 2006 and 2015. Finally, stock price and market liquidity data are obtained from CRSP, analyst forecast data from I/B/E/S, and institutional ownership data from Thomson Reuters. 3.2. Regulatory Enforcement Variable My proxies for regulatory enforcement are based on the characteristics of an insurance groups’ domicile states (i.e., the states where subsidiaries, excluding captive reinsurers, are domiciled). While insurance regulators can oversee insurers licensed to sell products in their state (i.e., both domestic and foreign), they usually defer regulation to the domiciliary regulator. Furthermore, under Section 531(a) of the Dodd-Frank Act, regulatory approval of reinsurance credit is deferred to domiciliary insurance regulators (Government Publishing Office 2010). I use seven proxies to capture regulatory enforcement, which depends on regulatory enforcement capacity, enforcement style, and the broader political environment (Carrigan and Harrington 2015). Regulatory enforcement capacity refers to regulators’ ability to exert control over regulated entities and depends on regulators’ legal autonomy, direct capacity to operate public enterprises, capacity to collect information, and financial resources (Hood and Margetts 2007). Regulatory enforcement authority (i.e., legal autonomy) gives regulators the legal power to enforce standards. State insurance departments have the legal authority to set and enforce insurance standards as established by the 1945 McCarran-Ferguson Act (NAIC 2011). Insurance regulators have the full legal capacity to enforce standards, as well as the capacity to publicly provide insurance services. However, in the United States, insurance regulators prefer to defer insurance services to the private market, if possible (Kunreuther et al. 2013). I measure 33 regulatory enforcement capacity with two proxies: Overlapping Regulators, which proxies for the capacity to collect information, and Regulatory Resources, which measures financial resources available to the regulator. Overlapping Regulators equals the number of unique domiciliary insurance regulators with jurisdiction over an insurance group. Regulatory Resources are measured as the regulatory budget (per $1,000 of premiums) of the insurance group’s domiciliary regulators. Regulatory enforcement style refers to a continuum of behaviors on how regulators interact with regulated entities (Carrigan and Harrington 2015). There are two broad enforcement styles: a deterrence mode of regulation (legalistic style) and an accommodative mode of regulation (cooperative style) (Coglianese and Kagan 2007). A legalistic style is based on the stringent and inflexible interpretation of standards, while a cooperative style accommodates the regulated entities’ arguments and is flexible in its interpretation of rules (Kagan 1989). Regulators may use both styles of enforcement (i.e., flexible enforcement style) but may still prefer one enforcement style over the other (Hutter 1989). I measure regulatory enforcement style with two proxies. Legalistic enforcement style (Strict Regulators) is measured as the number of domiciliary insurance regulators who monitor an insurance group and have a reputation for strict enforcement. Based on the NAIC state report cards (NAIC “State Insurance Regulation: Key Facts and Market Trends”), New York, Florida, Taxes, and California jurisdictions rank the highest in terms of insurance premiums sold, consumer complaints, regulators’ inquiries, and total budgets (per premium) available for regulation. Also, there is some anecdotal evidence that these four states (CA, FL, NY, and TX) are “strict” in their enforcement. For example, according to the R Street Institute – an American conservative and libertarian think tank dedicated to “free markets” – CA, FL, NY, TX, HI, LA, 34 and MT had a “D” grade as insurance jurisdictions in 2015, which implied limited “free markets” in these states (R Street 2015). Hawaii, Louisiana, and Montana, however, have smaller budgets per premium than California, Florida, New York, and Texas, and thus, are not coded as Strict Regulators. Since cooperative enforcement style relies on trust between the regulator and regulated entities, I measure a cooperative enforcement style (Captive Law Regulators) as the number of domiciliary insurance regulators in the group’s oversight who have captive insurance laws in their jurisdiction (Insurance Information Institute 2015). I assume that the presence of captive laws signals regulators’ willingness to accommodate firms’ reinsurance preferences. However, the presence of captive laws does not imply a “favorable” treatment of reinsurance transactions for all firms. Finally, I use three proxies to measure the broader political environment. Regulatory enforcement depends on regulators’ reelection incentives (Besley and Case 1985; Alt et al. 2011; Beland 2015). Elected Regulators equals the number of domiciliary insurance regulators in the firm’s structure who are elected to their office (NAIC 2015b). Insurance commissioners are elected in California, Delaware, Georgia, Kansas, Louisiana, Mississippi, Montana, North Carolina, North Dakota, Oklahoma, and Washington. Also, regulatory enforcement can be sensitive to interest group pressures and regulatory priorities (Carrigan and Harrington 2015). In insurance, regulators need to balance solvency regulation and rate regulation: regulators cannot decrease insolvency risk without affecting product pricing. Stringent solvency requirements can decrease insolvency risk but will also result in higher premiums charged to consumers and reduced product availability. As the result, regulators need to trade off an acceptable level of insolvency risk against product availability and 35 affordability. Insurance buyers are aware of and care about insurance product availability and affordability (Jaffee and Russell 1998). I expect that regulators may have incentives to be strict in enforcing captive reinsurance standards due to their regional geographic pressures to protect policyholders from high premiums. In the United States, coastal areas are characterized by a large and growing population and an exposure to natural disasters, and therefore, the demand for consumer rate protection is likely to be high in coastal insurance jurisdictions. Coastal Regulators capture domiciliary states’ geographic location in the U.S. sea-water coastal areas. Finally, insurance regulators’ enforcement actions towards captive reinsurance can depend on local citizens’ risk or regulation preferences. States can be characterized as Republican-leaning or Democrat-leaning. A 2016 survey by Gallup on Americans’ views found that Democrats worry about climate change and regulation of Wall Street and banks, while Republicans worry about economic growth, government intervention, and inefficiency (Gallup 2016). Republicanlearning and Democrat-leaning citizens can also differ in their risk preferences. Republican Regulators measures the number of domiciliary insurance regulators in Republican-learning states based on the citizens’ votes in the eight U.S. presidential elections between 1984 and 2012 (U.S. Electoral College 2016). I use two specifications of the enforcement proxies: unique values and total values. In the main tests, the regulatory enforcement factor is based on the total values of enforcement proxies. I use total values as insurance subsidiaries are individually regulated and thus each insurance affiliate has an opportunity to interact with its domiciliary regulator. Similarly, since credit ratings are based on the entire insurance group’s structure, I use total regulatory values in the credit rating analyses. In sensitivity tests, I use unique regulatory enforcement values to 36 reduce potential double-counting. An example of the regulatory variable calculation is presented in Appendix B. 3.3. Captive Reinsurance Determinants The first two hypotheses relate to the effect of regulatory enforcement on the use of captive reinsurers. The main tests use a panel Tobit estimation. In all tests, I regress the dependent variable on the firm’s incentives to use captive reinsurance and control variables. I use the following model to test H1 and H2: 𝐶𝑎𝑝𝑡𝑖𝑣𝑒 𝑅𝑒𝑖𝑛𝑠𝑢𝑟𝑎𝑛𝑐𝑒 = 𝛼 + 𝛽1 𝐸𝑛𝑓𝑜𝑟𝑐𝑒𝑚𝑒𝑛𝑡 + 𝛽2 𝑆𝑢𝑟𝑝𝑙𝑢𝑠 𝐶𝑜𝑛𝑠𝑡𝑟𝑎𝑖𝑛𝑡 + + 𝛽3 𝐸𝑛𝑓𝑜𝑟𝑐𝑒𝑚𝑒𝑛𝑡 ∗ 𝑆𝑢𝑟𝑝𝑙𝑢𝑠𝐶𝑜𝑛𝑠𝑡𝑟𝑎𝑖𝑛𝑡 + ∑ 𝛽(𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠) + 𝜀 (1) My first hypothesis implies a negative coefficient on β1, and my second hypothesis implies a negative coefficient on β3. I use various measures of captive reinsurance presence in an insurance group’s organizational structure. In the main tests, I use C_Number, which equals the number of captive reinsurance subsidiaries in the group’s structure in a given year. In sensitivity tests, I also examine the probability of the captive status (Captive), an indicator variable that equals one if an insurance group has at least one captive reinsurer in their organizational structure in a given year, and zero otherwise. I also examine insurers’ choice with regards to their captive reinsurer(s) location with C_Foreign, a dummy variable that equals two if an insurance group has captives both in the United States and abroad, one if all captives are licensed in the United States, and zero if there are no captive subsidiaries in a firm’s corporate structure. The reason I examine whether an insurance group has captive reinsurance abroad is the regulatory scrutiny of captive reinsurance via off-shore jurisdictions due to the confidentiality concerns. For example, on August 15, 2015 the NAIC signed a memorandum of understanding with the 37 Bermuda Monetary Authority (i.e., the insurance regulator overseeing insurance companies, including captives, domiciled in Bermuda) to increase cooperation between these two regulators; previously U.S. insurance regulators had limited access to the financial reports of captive reinsurers domiciled in Bermuda (NAIC 2015a). To measure regulatory enforcement, I aggregate seven regulatory enforcement proxies into a single regulatory enforcement factor (Enforcement) using principal component analysis. Regulatory enforcement proxies measure regulatory enforcement capacity, enforcement style, and the broader political environment. I interact the regulatory enforcement factor with the proxies for potential surplus constraint. To measure a potential surplus constraint, I use assets-to-surplus among life insurers and premium-to-surplus among P/C insurers (Federal Insurance Office 2013b). While insurance regulators use assets-to-surplus to assess the leverage of life insurers, I did not find authoritative guidance on the levels of leverage that represent either excessive or low leverage for life insurers. As the result, I use the ad-hoc levels of leverage (i.e., above 90th or 75th quantile and below the 25th quantile) in my study. Specifically, Surplus Constraint is a dummy variable that equals one if a life insurer has assets-to-surplus above the 90th quantile or below the 25th quantile (see Appendix C). I argue that the excessive leverage could attract regulatory attention because it implies that a life insurer has a greater exposure to the reserve estimation errors (i.e., and captive reinsurance affects a parent’s reserve levels and increases information asymmetry between insurance regulators and insurers about the reserve adequacy to due to the confidential reporting). I further argue that low leverage could also attract regulatory attention because regulators could be concerned about the under-reserving risk (i.e., insufficient reserves) since 38 reserves-to-surplus are highly correlated with assets-to-surplus and captive reinsurance affects the parent’s reported reserve levels and hence leverage. In the P/C insurance sector, premiums-to-surplus ratio is important in monitoring leverage as a large ratio implies greater underpricing risk (i.e., a higher risk that the surplus could be insufficient to cover expected losses) (NAIC 2016). Surplus Constraint is a dummy variable that equals one if a P/C insurer has premium-to-surplus above the 90th quantile (see Appendix C). I test the model separately for the sample of P/C insurers (i.e., pure property-casualty) and Life insurers (i.e., pure L/H and diversified insurers) to control for the significant differences in their operations and financial structure. In the model, I control for economic incentives for captive reinsurance, which are represented by proxies for contracting costs (i.e., reinsurance underwriting and investment inefficiency) and taxes. The control variables also include proxies for firm size, reinsurance, profitability, investment yields, internal funds, and leverage. Regulators monitor firms’ profitability and hence investment yields. I control for firm’s size since large firms might have incentives to use captive reinsurance to relieve their surplus constraint for growth purposes. Firm performance could be negatively associated with captive reinsurers because captives can be used to manage a firm’s statutory performance. However, since captive reinsurers are consolidated under US GAAP, insurers cannot ‘hide’ the poor performance of their captive reinsurers. Thus, better performing firms might opt for “shadow insurance.” Cash availability could also explain firms’ expansion into the “shadow insurance” market. There are jurisdictions that allow captive reinsurer formation with letters of credit or naked parental guarantees. Thus, cash-constrained firms could be more likely to use captive reinsurance than non-cash-constrained firms. Alternatively, non-cash-constrained firms could be 39 more likely to use captive reinsurance since the formation and operation of a captive requires a capital investment. Debt could creative incentives for captive reinsurance. While captive reinsurance could reduce a reinsurance counterparty’s credit risk, overall risk can be higher if captive reinsurance is not properly structured. Thus, creditors’ oversight could be either positively or negatively associated with captive reinsurance. I test my first two hypotheses using a Tobit model. I select the Tobit model because my dependent variable is truncated at zero as firms cannot have a negative number of captives, while the latent variable (e.g., financial reporting aggressiveness; risk preferences) could be below zero. As a result, my dependent variable equals zero when the latent variable is negative. In this case, Tobit estimates are consistent and asymptotically normal, while OLS estimates are inconsistent and downward biased (Amemiya 1973). Also, since most insurers organize captive reinsurers at the end of the year, my main tests use contemporaneous measures of regulatory enforcement and surplus constraint. I use lagged control variables. In untabulated results, I test other specifications of the independent variables to see whether my results are sensitive to model specification. I include year fixed effects to control for common macroeconomic effects and trends. I include main-regulator fixed effects to control for a common regulatory environment, as well as firms’ self-selection into a regulatory regime. I cluster errors by firm in the main tests and by main (lead) regulator in the sensitivity tests. 3.4. Captive Reinsurance, Regulatory Enforcement, and Credit Ratings To test H3 I examine the association between regulatory enforcement and credit ratings. I also test whether captive reinsurance is associated with S&P credit rating. To test H3, I use the following cross-sectional panel OLS regression model: 40 𝑆&𝑃 𝐶𝑟𝑒𝑑𝑖𝑡 𝑅𝑎𝑡𝑖𝑛𝑔 = 𝛼 + 𝛽1 𝐶_𝑁𝑢𝑚𝑏𝑒𝑟 (𝐶𝑁) + 𝛽2 𝐸𝑛𝑓𝑜𝑟𝑐𝑒𝑚𝑒𝑛𝑡 + + ∑ 𝛽 (𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠) + 𝜀 (2) In equation (2), the dependent variable equals the S&P long-term issuer credit rating assigned to an insurance group (CR_All). I translate twenty two categories into numerical values with one assigned to the highest credit rating score (AAA) and twenty two to the lowest credit rating (D). Non-rated insurers receive the numerical score of twenty-three as non-rated securities are often considered to be speculative grade. In my sample, non-rated insurance groups do not have captive reinsurers in their structure, with the exception of one firm. Nevertheless, there can be multiple reasons why an insurance group is non-rated by a credit rating agency. Non-rated issuers either did not request a rating or the credit rating agency did not have enough information to assign a credit rating. As the result, I exclude non-rated issuers in sensitivity tests, in order to examine the association between captive reinsurance and credit ratings among rated firms (CR_Rated). I multiply credit rating scores by negative one so that the associations are increasing in credit ratings. The regulatory enforcement factor (Enforcement) is my proxy for regulatory enforcement. My third hypothesis implies a positive coefficient on β2. Captive reinsurance is measured as the number of captive reinsurers (C_Number). Since captive reinsurance could result either in risk reduction or risk retention, the coefficient on β1 could be either positive or negative. I control for firm size, performance, contracting inefficiency, reinsurance, internal funds, leverage, and retained earnings. I include year fixed effects to control for common macroeconomic effects and trends. I include main-regulator fixed effects to control for a common regulatory environment as well as firms’ self-selection into a regulatory regime. All 41 variables are contemporaneous, except for control variables, which are lagged. I cluster errors by firm in the main tests and by main regulator in the sensitivity tests. 3.5. Captive Reinsurance, Regulatory Enforcement, and Information Asymmetry In the second set of consequence tests (H4 & H5), I examine whether captive reinsurance is associated with information asymmetry proxies. I also test the association between regulatory enforcement and market illiquidity. To test H5 and H6, I use a cross-sectional panel OLS regression and the following first-differences model: 𝛥 𝑀𝑎𝑟𝑘𝑒𝑡 𝐼𝑙𝑙𝑖𝑞𝑢𝑖𝑑𝑖𝑡𝑦 = 𝛼 + 𝛽1 𝛥 𝐸𝑛𝑓𝑜𝑟𝑐𝑒𝑚𝑒𝑛𝑡 + 𝛽2 𝛥 𝐶_𝑁𝑢𝑚𝑏𝑒𝑟 + + ∑ 𝛽 (𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠) + 𝑌𝑒𝑎𝑟 𝐹𝐸 + 𝑅𝑒𝑔𝑢𝑙𝑎𝑡𝑜𝑟 𝐹𝐸 + 𝜀 (3) In equation (3), the dependent variable is the change in market illiquidity, which I measure with three information asymmetry proxies. First, I use bid-ask spreads (Δ Log (Spread)) that are defined as the yearly median of daily quoted spreads divided by the midpoint. Bid-ask spreads increase with the level of information asymmetry and illiquidity because the spreads are used to address adverse selection in the presence of asymmetrically informed investors (Callahan et al. 1997). Second, I use zero-return days (Δ ZeroReturn), defined as the proportion of trading days with zero daily stock returns out of all possible trading days in a year. The frequency of zero return days increases with market illiquidity (i.e., transaction costs deter marginal investors from trading) (Chen et al. 2007). Third, I use the yearly median of the Amihud’s (2002) illiquidity measure (Δ PriceImpact), which is measured as the firm’s daily absolute stock return divided by US$ trading volume and multiplied by 1,000,000. High PriceImpact implies higher illiquidity because it indicates a low ability of investors to trade in a stock without moving its price (i.e., hence, high transaction costs). Finally, I use principal component analysis to aggregate the three market illiquidity proxies into a single illiquidity factor (Δ Illiquidity). 42 Captive reinsurance is measured as the number of captive reinsurers (C_Number). I use the regulatory enforcement factor (Enforcement) as a proxy for regulatory enforcement. The control variables include firm characteristics that likely affect the information environment and thus market liquidity (Kim et al. 2015): book-to-market ratio (BM), debt- to-assets ratio (Leverage), and return on assets (ROA). I control for market capitalization (MV), share turnover (Turnover), return variability (SD_Ret), stock momentum (ABN_Ret), and the proportion of informed traders (InstOwn%), all of which have been shown in prior research to affect market liquidity (Leuz and Verrecchia 2000; Sadka 2006; Daske et al. 2008; Christensen et al. 2013). All control variables are lagged by one year. All variables are defined in Appendix C. In addition to market liquidity tests, I use a similar model to examine changes in analyst following subsequent to changes in the number of captive reinsurance subsidiaries. I control for changes in firm size, book-to-market ratio, leverage, return on assets, share turnover, return variability, stock momentum, and institutional ownership. I test the following model: 𝛥 𝐴𝑛𝑎𝑙𝑦𝑠𝑡𝑠 = 𝛼 + 𝛽1 𝛥 𝐶𝑁𝑢𝑚𝑏𝑒𝑟 + 𝛽2 𝛥𝐶𝑎𝑝𝑡𝑖𝑣𝑒 + 𝛽3 𝛥 𝐸𝑛𝑓𝑜𝑟𝑐𝑒𝑚𝑒𝑛𝑡 + + ∑ 𝛽 (𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠) + 𝑌𝑒𝑎𝑟 𝐹𝐸 + 𝑅𝑒𝑔𝑢𝑙𝑎𝑡𝑜𝑟 𝐹𝐸 + 𝜀 43 (4) CHAPTER 4: MAIN RESULTS In Chapter 4, I provide the main results. I begin with descriptive statistics and univariate tests. Then, I present the main multivariate tests of H1 through H5. 4.1. Descriptive Statistics Panel A of Table E1 provides descriptive statistics for the two sub-samples: P/C insurers and Life insurers. On average, specialized property-casualty insurers have 0.5 captives whereas specialized life insurers and diversified insurers have 1.79 captive reinsurers in their insurance group’s structure. An average P/C and life insurance group has approximately ten insurance affiliates in the structure and five overlapping insurance regulators. There is considerable variability in regulatory enforcement and illiquidity. Life insurers, on average, have better CR_All than P/C insurers, but both firm types have similar CR_Rated (i.e., -8 or BBB+ rating). On average, CR_All is -10.8 (i.e., BB+ rating) for life insurers and – 13.2 (i.e., BB- rating) for P/C insurers. There are a few P/C insurers in my sample that were assigned the Non-Rated (NR) “rating” by the S&P credit rating agency. A median P/C insurer has $5.6 billion in assets with ROA of three percent and investment yield of eight percent. A median life insurer has $22.8 billion in assets, $14.7 billion in reserves, and $2.4 billion in surplus. Life insurers have larger investment yields and higher investment efficiency but lower return on assets than P/C insurers. A median P/C (life) firm reinsures fourteen (ten) percent of its premiums. Life insurers have larger values for regulatory enforcement than P/C insurers. Overall, the variables have considerable variation (spread) around their mean based on the between group standard deviation. Panel B of Table E1 reports pair-wise correlations among some variables. Among life insurers captive reinsurance is positively correlated with regulatory enforcement, analyst 44 following, size, and reinsurance activity and is negatively correlated with illiquidity and surplus constraint. Captive reinsurance is also positively correlated with leverage (i.e., assets-to-surplus and reserves-to-surplus, which have a correlation of 0.9754). Among P/C insurers captive reinsurance is positively correlated with size and credit ratings and is negatively correlated with premiums-to-surplus and assets-to-surplus. Credit ratings are positively correlated with enforcement among P/C insurers. Furthermore, I find that among life insurers C_Number is positively correlated with investment inefficiency and is negatively correlated with investment yields and return on assets. Among P/C insurers, captive reinsurance is positively correlated with reinsurance inefficiency and return on assets and is negatively correlated with the tax rate. Interestingly, enforcement is negatively correlated with reserves-to-surplus among life insurers. This result suggests that insurance regulators could be more concerned (i.e., more enforcement) about under-reserving than over-reserving in the life insurance sector. Table E2 reports the differences in means between insurers with captives and those without captives. Panel A of Table E2 reports the result for life insurers. Life insurers with captives are larger and have higher market liquidity. Also, they have more debt, less cash (as a percentage of assets), and lower investment yields. They are subject to greater regulatory enforcement and have higher S&P credit ratings. Panel B of Table E2 reports the results for P/C insurers. P/C insurers with captives are larger but have higher market illiquidity. They have higher credit ratings (CR_All), more cash, less debt, greater return on assets, and higher reinsurance inefficiency. There is no difference in regulatory enforcement between P/C insurers with and without captives. P/C insurers with captives have fewer subsidiaries domiciled in states coded as Strict Regulators (i.e., CA, FL, NY, and TX). 45 4.2. Main Results 4.2.1. Regulatory Enforcement and Captive Reinsurance Table E3 reports results of the principal component analysis (PCA) of seven regulatory enforcement proxies (i.e., based on total regulatory scores). The first principal component corresponds to the linear combination of regulatory enforcement proxies with maximum variance. The first principal component, PC1, has an eigenvalue of 5.0978 and explains 72.8% of the variation. I use this component (scoring factor) as a proxy for regulatory enforcement. There is a moderate correlation between the regulatory enforcement proxies and PC1, with correlation coefficients ranging from 0.14 to 0.18. Table E4, Panel A presents the results of estimating equation (1) with the dependent variable that measures the number of captive reinsurance subsidiaries in a firm’s corporate structure, C_Number. The results are based on the Tobit estimation. Consistent with H1, I find that regulatory enforcement is negatively associated with captive reinsurance among life insurers. However, regulatory enforcement is positively associated with captive reinsurance among P/C insurers (i.e., p-value < 0.10). This result suggests that regulatory enforcement can vary across firm types. It is consistent with spill-over effects: regulatory attention and monitoring of captive reinsurance among life insurers could decrease expected regulatory non-compliance costs among P/C insurers because insurance regulators have limited resources for monitoring and inspection. Furthermore, P/C insurers are on average smaller than life insurers, and the P/C insurance sector in general consists of a larger number of companies (i.e., 2,544 P/C insurers in 2015) than the life insurance sector (i.e., 872 life insurers in 2015) (Insurance Information Institute 2017). Industries composed of predominantly small firms receive less regulatory monitoring because it is more costly to monitor these firms (Basu and Dixit 2014). Small firms 46 are subject to less regulatory monitoring by the IRS (Hoopes et al. 2012). Resource-constrained regulators cannot perfectly monitor all firms, and thus, P/C insurers may expect smaller noncompliance costs associated with reinsurance standards and the use of captives than life insurers. Similarly, regulatory attention to the use of captives in the life insurance sector could have a spill-over effect in the insurance sector: the increase in information on captive reinsurance could deter potential non-compliance among firms with life operations (i.e., pure life and diversified insurers). Regulatory attention and monitoring can deter potential non-compliance of peers within the same industry (Block and Feinstein 1986; Schenck 2012). Furthermore, I find preliminary evidence that the effect of regulatory enforcement on the use of captive reinsurance is state-dependent. However, it is important to note that I cannot draw any conclusions because the test of H2 is not well specified and uses ad-hoc levels of leverage. The results discussed below are very preliminary and should be interpreted with caution. I find that life insurers with assets-to-surplus ratio above the 90th quantile or below the 25th quantile are less likely to use captive reinsurance, and this association is stronger in the presence of greater regulatory enforcement. Consistent with some anecdotal evidence, insurance regulators did not perceive life insurers’ reserves as redundant under the existing reserving standards (NAIC 2017). Thus, low leverage could attract regulatory attention to the use of captive reinsurance (i.e., increased expected compliance and non-compliance costs could reduce incentives to use captive reinsurance). However, excessive leverage could imply a higher insolvency risk due to the surplus inadequacy concerns, and thus could attract regulatory attention to the use of captives. In untabulated results, I also use reserves-to-surplus ratio because it is positively correlated with assets-to-surplus (i.e., reserves are backed up by assets) and there is anecdotal evidence of reserve reporting as an incentive for captive reinsurance among life insurers (NAIC 2012b, 47 2017). I find qualitatively similar results. However, the results are based on the model that is not well specified, and thus, I cannot draw any conclusions regarding H2. Similarly, I find some preliminary evidence that the effect of regulatory enforcement on the use of captive reinsurance may be state-dependent among P/C insurers. Again, all results discussed below should be interpreted with caution. I find that among P/C insurers premiums-tosurplus is negatively correlated with captive reinsurance. Insurance regulators use premiums-tosurplus ratio to monitor P/C insurers’ leverage. Once again, I use an ad-hoc level of “surplus constraint” to proxy for excessive growth among P/C insurers. P/C insurers need “free” capital to grow and could use captive reinsurance to reduce their capacity constraints. Anecdotal evidence suggests that some insurers use captive reinsurers to relieve their capacity constraints (NAIC 2012b). I assume that when premiums-to-surplus are above the 90th quantile, the use of captive reinsurance by P/C insurers could attract regulatory attention due to the rapid growth concerns (i.e., underpricing risk). I find that P/C insurers with premium-to-surplus ratio above the 90th quantile (in my sample) are more likely to have captive reinsurers, but this association is weaker in the presence of greater regulatory enforcement. This result is consistent with H2. I also find that captive reinsurance is positively associated with firm size and reinsurance among both life and P/C insurers. There is some evidence that life insurers with lower investment inefficiency and P/C insurers with higher reinsurance inefficiency are more likely to use captive reinsurance. Note, however, due to problems with my model that tests the interaction between enforcement and capital constraints, it is not appropriate to draw conclusive inferences. Next, I examine the use of captive reinsurance across two time periods. In 2011, the use of captive reinsurance by life insurers received considerable attention from media and regulators. In October of 2011, the NAIC formed a committee to study the use of captive reinsurers in the 48 life insurance sector. In 2013, the New York Department of Financial Services published a report on “shadow insurance” in which the NY insurance regulator advocated a national moratorium on captive reinsurance until regulators could better assess captive reinsurance risk. Since then, there have been several class action law suits filed against some insurers for their use of captive reinsurance. Insurance regulators also have been criticized for lax enforcement and a regulatory race-to-the-bottom as more jurisdictions have passed captive insurance laws since 2012. Panel B of Table E4 reports the cross-sectional results for two time periods. Column (2) reports the results for the 2006 – 2011 time period (“low public scrutiny”) while Column (3) reports the results for the 2012 – 2015 time period (“high public scrutiny”). Among life insurers, regulatory enforcement is negatively associated with captive reinsurance in both time periods. The coefficient on the interaction term between surplus constraint and regulatory enforcement is negative in both time periods, but it is statistically significant only between 2006 and 2011. The results suggest that public awareness and scrutiny of accounting standards and regulators (who set and enforce those standards) can be important for regulatory enforcement and its credibility. For example, public scrutiny of insurance regulation and insurance regulators’ inquiries into captive reinsurance in the life insurance sector could increase the information uncertainty among life insurers about regulatory enforcement and its credibility. Similarly, I find that among P/C insurers the coefficient on the interaction term between surplus constraint and regulatory enforcement is negative (i.e., p-value < 0.01) only between 2006 and 2011, and it is positive (i.e., p-value < 0.01) between 2012 and 2015. This result suggests that regulatory attention to the use of captives by life insurers could have a spill-over effect in the P/C insurance sector: regulatory attention to life insurers could decrease expected non-compliance costs among P/C insurers (e.g., lower probability of regulatory monitoring and 49 enforcement when regulators are resource-constrained). However, it is important to note, these results are preliminary, and I cannot draw any conclusive inferences on H2 due to the concerns over the specification of the model. 4.2.2. Regulatory Enforcement, Captive Reinsurance, and Credit Ratings The third hypothesis (H3) examines the role of regulatory enforcement in the credit rating process. Table 5 reports the results of estimating equation (2) with a panel OLS regression. The dependent variable is CR_All. Panel A of Table E5 reports the sub-sample results by firm type. Column (1) reports the results for life insurers. Consistent with H3, regulatory enforcement is positively associated (p-value < 0.01) with credit ratings (CR_All). Column (2) reports the result for P/C insurers. I do not find support for H3 among P/C insurers. This result suggests that public awareness of accounting issues and regulatory actions may be important for regulatory enforcement credibility. Panel B of Table E5 reports the cross-sectional results by firm type and time period. The results indicate that the role of regulatory enforcement can vary across firm types and type periods. I find that regulatory enforcement is positively associated with credit ratings among life insurers in both time periods, but the association is statistically significant (p-value < 0.01) only between 2012 and 2015. Once again, the result indicates that public awareness of regulatory attention or actions could be important for the credibility of regulatory enforcement. In contrast, I find that regulatory enforcement is negatively associated (p-value < 0.10) with credit ratings among P/C insurers between 2006 and 2011. In prior tests, I found a positive association between regulatory enforcement and the use of captive reinsurance among P/C insurers between 2006 and 2011. These results indicate that regulatory enforcement may have different implications for different firms. 50 4.3.2. Regulatory Enforcement, Captive Reinsurance, and Information Environment The fourth and fifth hypotheses (H4 & H5) examine the role of regulatory enforcement and captive reinsurance in equity markets. I use three market illiquidity proxies, which are combined in one illiquidity factor (see Table E6 for the Principal Component Analysis results). Table E7 presents primary regression results from tests of H4 and H5 using the three market illiquidity proxies and illiquidity factor. Panel A of Table E7 presents the results for life insurers. Consistent with H4, I find that the change in regulatory enforcement is negatively associated (pvalue < 0.10) with the change in market illiquidity. While I do not examine the channel, this result is consistent with regulatory enforcement reducing the probability of misreporting and increasing the credibility of financial reports. Contrary to my hypothesis H5, I find that the change in the number of captive reinsurance subsidiaries is also negatively associated with the change in market illiquidity, however this association is not statistically significant among life insurers. In terms of control variables, share turnover, as expected, is significantly and negatively associated with market illiquidity. As expected, an increase in return variability is positively associated with market illiquidity. Larger firms have higher market liquidity. Interestingly, I find that the increase in the percentage of sophisticated investors (Δ InstOwn%) is positively associated with the increase in illiquidity (i.e., more zero-return days). This result is consistent with prior studies which show that liquidity risk varies across different types of institutional owners (Sias 2004; Gatev and Strahan 2006, Brunnermeier and Pederson 2009). Panel B of Table E7 presents the results for P/C insurers. The coefficient on regulatory enforcement is negative but not statistically significant. In contrast to H5, I find that the increase in the number of captive reinsurers is negatively associated (p-value < 0.01) with the changes in market illiquidity. A possible explanation of this result is that the use of captive reinsurance by 51 P/C insurers attracts the attention of analysts and investors, who increase their private data collection efforts and hence the firm’s liquidity improves. In contrast to life insurers, there is less public awareness and information on the incentives for captive reinsurance among P/C insurers. Thus, an increase in captive reinsurers among P/C insurers could attract analysts or investors. Also, in prior tests I find that P/C insurers facing greater regulatory enforcement are more likely to use captive reinsurance and have lower credit ratings between 2006 and 2011. Other external monitors can act as substitutes to formal regulatory monitoring (Miller 2006; Dyck et al. 2008). Table E7, Panel C reports the results of estimating equation (4). The dependent variable is the change in analyst following (Δ Analysts). Column (1) and (2) report the results for life and P/C insurers, accordingly. The coefficient on captive reinsurance is negative but not significant. However, I find some evidence that analyst following is positively and statistically associated with captive reinsurance when a life insurer initiates captive reinsurance transactions (Δ Captive). 52 CHAPTER 5: SUPPLEMENTAL ANALYSES In Chapter 5, I perform additional robustness tests. First, I analyze the sensitivity of my main results to the specification of the dependent and independent variables. Then, I examine the sensitivity of my main results to alternative functional forms. Finally, I discuss the results of various robustness tests addressing endogeneity. 5.1. Alternative Specifications of Captive Reinsurance To check the sensitivity of my results to the dependent variable specification, I use two additional measures of captive reinsurance. Since there are not many firm-year observations in my sample where a new captive reinsurer is organized (C_Form), I report the results only with the dependent variables Captive and C_Foreign. Table E8, Panel A reports the results of estimating equation (1) where captive reinsurance is measured as Captive and C_Foreign. I use a probit model in Columns (1) and (3) and an ordered logit in Columns (2) and (4). Consistent with H1, I find that among life insurers regulatory enforcement is negatively associated with captive reinsurance. I do not find results supporting H2. Among P/C insurers, I find that firms with a surplus constraint and facing greater regulator enforcement are less likely to have captives. I do not find support for H1 and H2 when C_Foreign is the dependent variable. To check the sensitivity of my results to the independent variable specification, I use unique regulatory values (see Appendix B for an example) to measure regulatory enforcement. I aggregate these unique values of regulatory enforcement proxies into one regulatory enforcement factor (Enforcement 2) using the Principal Component Analysis. Table E8, Panel C reports the results of estimating equation (1) using he alternative specification of regulatory enforcement. The results are consistent with my prior findings. I find support for H1 and H2 among life 53 insurers and for H2 among P/C insurers. Important to note, nevertheless, that the model is probably misspecified, and it is inappropriate to draw any conclusive inferences. 5.2. Alternative Functional Form In Table E9, I present the results of estimating equation (1) with a quantile estimation. The quantile estimation allows a description of the entire conditional distribution. Also, the parameters in the quantile regression are relatively robust to outliers and can be more efficient than the OLS estimators when the error terms are not normally distributed (Buchinsky 1998). Table E9, Panel A reports the results with a quantile Tobit model in the sub-sample of life insurers. Among life insurers, I find support for H1 at all quantiles between 2006 and 2011 and at the median and the 75th quantile between 2012 and 2015. There is some evidence that regulatory enforcement is state-dependent. Once again, I find support for H2 but only at the 25th quantile and the median (i.e., the coefficient is negative but not statistically significant at the 75th quantile) between 2006 and 2011 and only at the 75th quantile between 2012 and 2015. Interestingly, the coefficient on the interaction term between surplus constraint and regulatory enforcement is positive and statistically significant at the 25th quantile between 2012 and 2015. The results suggest that regulatory attention to the use of captive reinsurance by life insurers with multiple captive subsidiaries could have a spill-over effect in the life insurance sector. Once again, this result is very preliminary, and I cannot draw any conclusions regarding H2 and spillover effects. I also find results consistent with a spill-over effect in the P/C insurance sector. Table E9, Panel B reports the results of estimating equation (1) with a quantile Tobit regression among P/C insurers between 2006 and 2011. The results are statistically insignificant in the 2012 – 2015 time period and are untabulated. I find that among P/C insurers between 2006 and 2011, the 54 coefficient on enforcement is positive at the 25th quantile. Consistent with H2, the coefficient on the interaction term between surplus constraint and regulatory enforcement is negative and statistically significant at all quantiles only between 2006 and 2011. Regulatory attention to the use of captives by life insurers between 2012 and 2015 could have a spill-over effect in the P/C insurance sector. Once again, this result is very preliminary, and I cannot draw any conclusions regarding H2 and spill-over effects. 5.3. Endogeneity Robustness Tests 5.3.1. Instrumental Variable, SURE, and Heckman Estimation To address endogeneity concerns, I include main-regulator fixed effects to control for a common regulatory environment and firms’ self-selection into a regulatory regime. However, there still can be unobservable characteristics that are associated with the firm’s regulatory enforcement and captive reinsurance choices. In the robustness testes, I use instrumental variable (IV) regression, simultaneous equation modeling, and Heckman estimation. In the instrumental variable test, I use a firm’s market power and Strict Regulators as instruments for regulatory enforcement. I measure market power using the Lerner index, a negative inverse of the price elasticity of demand.9 The price-cost margin (PCM) is frequently used in the empirical industrial organization literature as a proxy for market power (Bikker and van Leuvensteijn 2008). I find that my measure of market power is correlated with the regulatory enforcement factor, but it is not associated with the number of captive reinsurers. Similarly, Strict Regulators are correlated with the regulatory enforcement factor but are not directly associated with captive reinsurance. 9 The lack of detailed data in my sample limits the opportunity to use other proxies of market power (e.g., concentration) in sensitivity tests. I plan to hand-collect the data in the future work on this topic. 55 Table E10, Panel A reports the result for life insurers. Column (1) reports the results of estimating equation (1) with an instrumental variable model where market power and Strict Regulators are used as instruments for regulatory enforcement. The F-statistic for the weak identification test is 30.5090. The rule of thumb is that the F-statistic should be greater than ten to make sure that the maximum bias in IV estimators is less than 10% (Staiger and Stock 1997). Thus, market power and Strict Regulators are likely not a weak instrument for regulatory enforcement in this test. I also find some evidence that the model is neither underspecified (KP test p-value = 0.0040) nor over-specified. In the instrumental variable test, I find results that are consistent with my main model. Column (2) of Table E10, Panel A reports the results of estimating equation (1) with simultaneous equation modelling (i.e., seemingly unrelated regression). Here, regulatory enforcement is modeled as a function of firm size, reserve and surplus adequacy, tax rates, performance, investment yields, internal funds, and leverage. I include the main regulator indicators. I find preliminary support for H1 and H2. Column (3) of Table E10, Panel A reports the results of estimating equation (1) with a Heckman two-step procedure. Similar to the SURE estimation, the probability of regulatory enforcement is modeled as a function of firm size, reserve and surplus adequacy, tax rates, performance, investment yields, internal funds, and leverage. Based on the Inverse-Mills ratio, I do not find support for self-selection. I find preliminary support for both H1 and H2. Regulatory enforcement is negatively associated with the use of captive reinsurers among life insurers. There is preliminary evidence that the strength of this association may vary across firms’ characteristics. 56 Table E10, Panel B reports the results of estimating equation (1) with IV, SURE, and Heckman regression among P/C insurers. Similar to my previous findings, I do not find support for H1 but there is limited preliminary evidence supporting H2. As predicted, the coefficient on the interaction term between surplus constraint and regulatory enforcement is negative, but it is not statistically significant. Nevertheless, the coefficient on the interaction term between premium-to-surplus and enforcement is negative and statistically significant (i.e., p-value < 0.10). I cannot draw any conclusive inferences due to the model misspecification concerns. 5.3.2. Falsification Test Table E11 reports the results of a falsification test. If insurers who select specific regulatory enforcement are inherently different, I expect to see the effects of regulatory enforcement on other decisions. Here, I use the same approach as in Nicoletti (2015), with variables appropriate to the insurance industry. Security gains and losses are subject to managerial discretion, but are unlikely to be influenced by regulators (Beatty and Harris 1998). This test is designed to check whether the association between realized gains on securities and income before realized gains is moderated by the regulatory enforcement. An insignificant interaction term implies that regulatory enforcement characteristics do not represent firm characteristics (i.e., self-selection) and, instead, capture regulatory enforcement. I estimate the following model which captures the determinants of realized capital gains and losses (RealGAIN): 𝑅𝑒𝑎𝑙𝐺𝐴𝐼𝑁𝑖,𝑡 = 𝛼 + 𝛽1 𝑅𝑒𝑔𝑢𝑙𝑎𝑡𝑜𝑟𝑦 𝐸𝑛𝑓𝑜𝑟𝑐𝑒𝑚𝑒𝑛𝑡𝑖,𝑡 (𝑅𝐸) + 𝛽2 𝐸𝐵𝐺𝐴𝐼𝑁𝑆𝑖,𝑡 +𝛽3 𝐸𝐵𝐺𝐴𝐼𝑁𝑆𝑖,𝑡 ∗ 𝑅𝐸𝑖,𝑡 + 𝛽4 𝐶𝑎𝑝𝑖𝑡𝑎𝑙𝑖,𝑡 + 𝛽5 𝑈𝑛𝑟𝑒𝑎𝑙𝐺𝐴𝐼𝑁𝑖,𝑡−1 + 𝛽6 𝐼𝐴𝑖,𝑡−1 + 𝛽7 𝐼𝑛𝑣𝐼𝑛𝑐𝑖,𝑡−1 + 𝛽8 𝐿𝑖𝑞𝑖,𝑡−1 + 𝛽9 𝑆𝑖𝑧𝑒𝑖,𝑡−1 + ∑ 𝛽 (𝐼𝑛𝑡𝑒𝑟𝑎𝑐𝑡𝑖𝑜𝑛𝑠 & 𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠) + 𝜀 (5) where variables are defined as follows: 57 RealGAIN = realized capital gains (losses) scaled by total assets EBGAINS = earnings before capital gains (losses) scaled by total assets Capital = surplus before capital gains (losses) scaled by total assets UnrealGAIN = unrealized capital gains (losses) scaled by total assets IA = invested assets scaled by total assets InvInc = investment income scaled by total assets Liq = reserves scaled by total assets (or surplus) Unfortunately, the Compustat data on realized security gains and losses for insurance companies is limited and, thus, I have to rely on all realized capital gains and losses (RealGAIN), which could be affected by an insurer’s investments in a captive. I include EBGAIN and Capital to control for the incentives to manage earnings and capital, respectively. I include controls that capture the opportunity to manage earnings through asset sales. Insurers with larger unrealized capital gains or losses and a larger proportion of invested assets have a greater ability to manage earnings through capital asset sales. I control for investment income generated by the invested assets, firm size, and liquidity position. I include year and main regulator fixed effects. Robust standard errors are clustered by firm. Column (1) of Table E11 reports the results of estimating equation (5). The coefficient on the interaction between the income reporting incentive and regulatory enforcement is statistically insignificant. In Column (2), I report the results of estimating equation (5) with an instrumental variable regression where regulatory enforcement is instrumented with market power. The inferences do not change. The result suggests that regulatory enforcement is unlikely to represents firm characteristics. 5.3.3. Propensity Score Matching (PSM) I use propensity score matching in the credit rating and market liquidity tests. I match firms based on their size under the assumption that firm size is the predictor of captive 58 reinsurance (Captive). The differences in size are insignificant between the treatment and control sample (two-tailed p-value > 0.1). I use nearest neighbor matching. Table E12 reports the results of estimating equation (2) with panel OLS using a matched control sample. In Column (1) and (3), the regression is adjusted for the propensity score, while in Column (2) and (4) I use propensity score weighting (i.e., inverse-probability weightings). The matched-pairs’ credit rating tests result in inferences that are qualitatively similar to those from the cross-sectional tests. I find support for H3, but only among life insurers. In addition to propensity score matching, I use the instrumental variable estimation. Table E13 reports the results of estimating equation (2) using an instrumental variable regression where Enforcement is instrumented with market power. In Column (1) and (3) the dependent variable is CR_All, and in Column (2) and (4) the dependent variable is CR_Rated. I exclude insurance groups with a “non-rated” credit rating and use CR_Rated as the dependent variable. The results are based on the sample of forty eight insurance groups (i.e., I lose fourteen insurance groups that have been assigned a “Non-Rated” rating by the S&P credit rating agency). I find support for H3 among life insurers with both dependent variables. Market power is probably not a weak instrument for enforcement in these tests (i.e., F-statistic in the weak identification test is above ten). However, market power could be a weak instrumental variable in the tests among P/C insurers. I find support for H3 among P/C insurers when the dependent variable is CR_All, but not CR_Rated (i.e., the coefficient is positive but not significant). Finally, I use propensity score matching in the market liquidity tests (i.e., H4 and H5). I use a similar approach as in the PSM tests of H3. Table E14 reports the results of estimating equation (3) where the regression is adjusted for the propensity score. I find limited support for H4 among life insurers: the change in regulatory enforcement is negatively associated with the 59 change in market illiquidity. Contrary to H5, I find that changes in captive reinsurers are negatively associated with market illiquidity among P/C insurers (i.e., the coefficient is negative but not significant among life insurers). This result indicates that the use of captive reinsurance (so-called shadow insurance) does not necessarily result in greater opacity and lower market liquidity. Investors and analysts can substitute regulatory monitoring efforts, especially for firms potentially receiving less regulatory attention (i.e., NAIC has been examining the use of captive reinsurers by life insurers but not P/C insurers). However, matching on observables does not rule out an omitted variable bias, and the propensity score results are sensitive to the test specification. Also, propensity score methods work better in large samples which allows the model to achieve distribution balance of observed covariates. Nevertheless, propensity scores can be used for regression adjustment, stratification, and weighting (King and Nielsen 2016). 60 CHAPTER 6: CONCLUSIONS This dissertation examines the association between regulatory enforcement and the firm’s use of off-balance sheet entities. I use regulatory enforcement capacity, enforcement style, and the broader political environment to measure regulatory enforcement of “shadow insurance,” a non-traditional reinsurance that is associated with opaque statutory reporting in the insurance industry. In general, I find that the effect of regulatory enforcement varies across firm types and time periods, and public awareness of accounting issues could be important for regulatory enforcement. Among life insurers, I find that regulatory enforcement is negatively associated with captive reinsurance at all quantiles between 2006 and 2011 and at the median and the 75th quantile between 2012 and 2015. The negative association between regulatory enforcement and captive reinsurance is stronger among life insurers with very high or low leverage (i.e., assets-tosurplus above the 90th quantile or below the 25th quantile). This association is statistically significant at the 25th quantile and the median between 2006 and 2011 and only at the 75th quantile between 2012 and 2015. The result is consistent with a spill-over effect. Important to note, nevertheless, that my findings on H2 should be interpreted with caution due to the model misspecification and endogeneity concerns. P/C insurers, on the other hand, are more likely to use captive reinsurance in the presence of greater regulatory enforcement between 2006 and 2011, especially at the 25th quantile. This result is consistent with a spill-over effect: regulatory attention to the use of captives by life insurers could decrease (anticipated) regulatory attention to the use of captive reinsurance by P/C insurers because insurance regulators are resource-constrained. Also, there are more P/C insurance companies than life insurers, and P/C insurers are usually smaller than life insurers. In 61 general, monitoring smaller firms can be more costly. However, in robustness tests I do not find conclusive evidence of the positive association between regulatory enforcement and captive reinsurance among P/C insurers. Nevertheless, there is some preliminary evidence that regulatory enforcement could be state-dependent (H2) among P/C insurers. I find that P/C insurers with the high premium-to-surplus ratio (i.e., surplus constraint) are more likely to use captive reinsurance, but this association is weaker among P/C insurers facing stronger regulatory enforcement (i.e., at all quantiles between 2006 and 2011). Furthermore, I find some evidence that credit rating agencies rely on regulatory enforcement to infer information on a firm’s default risk. I find support for H3 (i.e., a positive association between regulatory enforcement and credit ratings) among life insurers between 2012 and 2015. This finding is consistent with public scrutiny of captive reinsurance in the life insurance sector since 2012. This result also implies that public scrutiny or awareness could be important for regulatory enforcement of accounting standards and hence regulatory enforcement credibility. In contrast, among P/C insurers, regulatory enforcement is negatively associated with credit ratings between 2006 and 2011 (i.e., the time period where I also find a positive association between regulatory enforcement and captive reinsurance among P/C insurers). However, in the robustness tests I do not find support for the negative association between regulatory enforcement and credit ratings among P/C insurers. Overall, results indicate that public attention or awareness of regulatory efforts could be important for regulatory enforcement credibility. Finally, I find some evidence that the changes in regulatory enforcement are negatively associated with the changes in market illiquidity among life insurers. In contrast to my hypothesis (H5), I find that captive reinsurance is negatively associated with market illiquidity, 62 but this result is statistically significant only among P/C insurers. Thus, there is some evidence that shadow insurance is not as opaque as I predicted. I find that among life insurers the use of captives (i.e., Δ Captive; when a firm licenses the first affiliated captive reinsurer) is positively associated with the changes in analyst following. However, my analysis is subject to a few important caveats. My results may not be generalizable to other settings. I perform various robustness tests including instrumental variable estimation, falsification tests, and propensity score matching and include main regulator fixed effects. But I cannot rule out endogeneity concerns pertaining to firms’ self-selection into a regulatory enforcement environment. Furthermore, some of the models used to test the hypotheses have multicollinearity and could be misspecified. As such, all results should be interpreted with caution. In future work, I plan to reassess my models. For example, to reduce multicollinearity I can use a centered leverage. Also, I will change the model structure to test for a non-linear effect of leverage (e.g., interact leverage with the surplus constraint indicator variable). I will improve the specification of my models (e.g., include additional controls that could be correlated with the firms’ financial position or enforcement and the use of captive reinsurance). I will also test additional specifications of the independent variables of interest (i.e., insurers’ financial position and regulatory enforcement). There are a few lines of research that could be pursued from this work. First, in this setting it is possible to examine the determinants of regulatory enforcement (i.e., regulatory enforcement is heterogeneous across states) and the effects of various dimensions of regulatory enforcement on the use of off-balance sheet entities. Second, future research could examine whether the use of captive reinsurance entities – that are off-balance sheet under SAP but are on 63 the balance sheet under US GAAP – is associated with the disagreement among credit rating agencies and among analysts. Finally, future work could study whether the use of captive reinsurers is associated with the firms’ disclosure choice and financial reporting quality. 64 APPENDICES 65 APPENDIX A Reinsurance Example Company A sold $20 in premiums and incurred $8 in policy acquisition expenses. Without loss of generality, assume that the company received cash when the policy was sold and paid cash for its acquisition expenses. Also, assume that by the end of the year, the entire premium of $20 was earned and recognized as revenue. The company reinsures 100% of its policies with reinsurer B. The company paid $10 for the 100% coinsurance coverage and estimated that this coverage would reduce its expected losses by $15. That is, initially the company expected a net profit of $12 (i.e., $20 - $8), while at the end of the year the expected losses on the policy are $15 (i.e., $12 - $15 = ($3) loss). The purpose of the reinsurance contract is to share the insurance risk (i.e., a probable $3 loss) with another party for a fee. In this case, Company A paid $10 for reinsurance; thus, it can book a $2 gain (i.e., $12 $10) on the reinsurance transaction. The $2 gain and the reduction in expected losses of $3 will result in a $5 increase in Surplus. The effect of this reinsurance transaction on the company’s Balance Sheet is presented below. TABLE A1 Reinsurance Effect on the Balance Sheet Accounts Cash (A) Gross Loss Reserves (L) Loss Reserve Ceded (CL) Net Loss Reserves (L) Surplus (SE) BB $100 Cash/ Premiums Earned $20 Estimated Expense/ Losses/Loss EB EB Cash Reserves (without) Reinsurance (with) ($8) $112 ($10) $102 $40 $15 0 0 $40 $60 $55 $55 $15 $55 $20 ($8) ($15) Gross Reserves / Surplus: Net Loss Reserves / Surplus: Assets / Liabilities: $57 0.9649 0.9649 2.0364 66 $15 $40 $5 $62 0.8871 0.6452 2.5500 The reinsurance does not change Gross Loss Reserves, but it reduces Net Loss Reserves and increases Surplus. The increase in Surplus will improve both the Gross Reserves-to-Surplus and Net Loss Reserves-to-Surplus ratios. Also, reinsurance will reduce Net Premiums; and therefore, Gross Premiums-to-Surplus and Net Premiums-to-Surplus ratios will also improve. TABLE A2 Reinsurance Journal Entries Account Premiums Ceded Expenses Ceded Cash Gain Cr Dr $20 $8 $10 $2 67 Account Loss Reserve Ceded Surplus Dr $15 Cr $15 APPENDIX B Calculation of Regulatory Variables Assume that a fictitious firm A has seven insurance subsidiaries in its group structure and NY is its designated main regulator. These seven subsidiaries are under the supervision of the following regulators: 1 in Michigan, 1 in California, 2 in New York, 1 in Indiana, 1 in Georgia, and 1 captive reinsurer in South Carolina. I do not include the South Carolina captive reinsurer in the regulatory score.10 The number of insurance affiliates in the groups’ structure that could potentially use captive reinsurance equals six. The number of overlapping regulators equals four. Total scores count the number of insurance subsidiaries facing certain regulatory enforcement environment characteristics. Unique regulatory variable scores measure the number of unique insurance regulators’ with a given set of regulatory enforcement characteristics. TABLE B1 Regulatory Score Calculation Unique Total 4 4 Average regulatory budget Total regulatory budget Captive Law Regulators 2 (MI, NY) 3 ( 1 MI + 2 NY) Strict Regulators 2 (CA, NY) 3 (1 CA + 2 NY) Elected Regulators 2 (CA, GA) 2 (1 CA + 1 GA) Coastal Regulators 3 (CA, NY, GA) 4 (1 CA + 2 NY + 1 GA) 2 (IN, GA) 2 (1 IN + 1 GA) Overlapping Regulators Regulatory Resources Republican Regulators 10 While insurers can choose among captive regulators and that choice could reflect firm characteristics or incentives, I do not differentiate among captive regulators in this paper. I do acknowledge that it might be interesting to examine a firm’s choice of captive regulator and whether that choice matters for the firm’s reputation (and, hence, interaction with other regulators and credit rating agencies’ assessments). 68 APPENDIX C Variables TABLE C1 Variable List Variable Definition P/C Insurers pure property-casualty insurers (SIC code 6331) Life Insurers pure life insurers (SIC code 6311) and diversified insurers (SIC codes 6311 and 6331). C_Number (CN) the number of captive reinsurance subsidiaries in the insurance group's structure Captive an indicator variable that equals one if an insurance group has at least one captive reinsurance subsidiary in their structure, and is zero otherwise C_Form an indicator variable that equals one if an insurance group licenses a captive reinsurance subsidiary, and is zero otherwise C_Foreign an indicator variable that equals zero when there are no captive reinsurance subsidiaries in the group's structure, equals one if all captive reinsurance subsidiaries are licensed in the U.S., and two if captive reinsurance subsidiaries are all located abroad or both in the U.S. and abroad Enforcement a factor score derived from a principal component factor analysis of the standardized measures of the regulatory enforcement variables: Overlapping Regulators, Regulatory Resources, Captive Law Regulators, Strict Regulators, Elected Regulators, Coastal Regulators, Republican Regulators Affiliates the number of insurance affiliates in the group's structure, excluding captive reinsurers Overlapping Regulators the number of unique overlapping insurance regulators who monitor insurance affiliates in the group’s structure, excluding captive reinsurers' regulators Regulatory Resources mean regulatory budget per $1,000 of premiums for regulators that monitor the insurance group Captive Law Regulators the number of unique regulators in the group's structure that have captive laws in their state Strict Regulators the number of unique regulators in the group's structure that are strict in their enforcement Elected Regulators the number of unique regulators in the group's structure that are appointed through an election Coastal Regulators the number of unique regulators in the group's structure whose state is located in a coastal area Republican Regulators the number of unique regulators in the group's structure whose state’s citizens has been Republican-leaning based on their voting in the past eight U.S. presidential elections Surplus Constraint an indicator variable that equals one if the premiums-to-surplus ratio is above the 90th quantile among P/C insures, and zero otherwise; an indicator variable that equals one if the assets-to-surplus ratio is above the 90th quantile (ASR90) or below the 25th quantile (ASR25) among life insurers, and zero otherwise. 69 TABLE C1 (cont’d) Variable Definition Assets / Surplus Ratio (ASR) an indicator variable that equals one if the assets-to-surplus ratio is: above the 90th quantile (i.e., ASR90); above the 75th quantile (ASR75); or below the 25th quantile (ASR 25). RSR / Surplus Ratio (RSR) an indicator variable that equals one if the reserves-to-surplus ratio is greater than three (i.e., RSR2: when reserves-to-surplus are greater than five; RSR3: when reserves-tosurplus are greater than ten; RSR4: when reserves-to-surplus are greater than eighteen; and RSR5: when reserves-to-surplus are greater than fourteen or less than four). Premiums / Surplus Ratio (PSR) an indicator variable that equals one if a P/C insurer has premium-to-surplus ratio above the 90th quantile, and zero otherwise. Log (Premiums) natural logarithm of the insurance group's total gross written premiums Surplus the insurance group's statutory surplus (SRT in WRDS Compustat) Reinsurance Inefficiency reinsurance underwriting expenses scaled by total ceded premiums Investment Inefficiency investment expenses scaled by total investment income Reinsurance a ratio of ceded premiums to total gross written premiums ROA net statutory income (NITS in WRDS Compustat) divided by total assets Investment Yield investment income scaled by invested capital Tax Rate tax expenses (i.e., net income less pre-tax income) scaled by pre-tax income Cash / Total Assets cash scaled by total assets Debt / Total Assets long-term debt scaled by total assets CR_All S&P long-term issuer credit rating assigned to the insurance group; the highest rating (AAA) is encoded 1 while the lowest rating (D) is encoded 22. Non-rated insurers are assigned a credit rating score 23. Then, the credit rating scores are multiplied by – 1 CR_Ranked S&P long-term issuer credit rating assigned to the insurance group if there is a rating. Non-rated insurers are not included. The highest rating (AAA) is encoded 1 while the lowest rating (D) is encoded 22. Then, the credit rating scores are multiplied by – 1 MV market value of equity B/M book-to-market ratio Turnover annual US$ trading volume divided by market cap SD_Return standard deviation of daily stock returns ABD_Return cumulative size and book-to-market adjusted stock return 70 TABLE C1 (cont’d) Variable Definition InstOwn% outstanding shares owned by institutional investors as a percentage of total shares Spread yearly median of daily quoted spreads (i.e., the difference between the bid and ask price divided by the midpoint) ZeroReturn proportion of trading days with zero daily stock returns out of all potential trading days PriceImpact yearly median of Amihud’s (2002) illiquidity measure (i.e., daily absolute stock return divided by US$ trading volume), multiplied by 1,000,000 Illiquidity a factor score derived from a principal component factor analysis of three standardized measures of market illiquidity: Spead, ZeroReturn, PriceImpact Analysts number of analysts following the firm each month, averaged over the year RealGAINS realized capital gains (losses) scaled by total assets EBGAINS earnings before capital gains (losses) scaled by total assets Power product pricing power, i.e, Lerner index, measured as the difference between premiums and reserves divided by total premiums 71 APPENDIX D Sample Selection The table below details the sample selection process. I use CorporateAffiliations database to collect data on the firms’ organizational structure. I start with 120 largest by gross premiums (as reported in the CorporateAffiliations database) public insurance groups between 2006 and 2015. I use multiple sources to verify companies in the firms’ corporate structure: Exhibit 21 in the 10K filings, the NAIC listing of insurance groups, A.M. Best Corporate Structure file, and the YSchedule from statutory filings (e.g., collected from firms’ official websites and regulatory examination reports). I verify the subsidiaries’ licenses (i.e., type, effective date, and parent) from regulatory websites. Then, I remove insurance companies that were acquired by another company. Also, I remove insurers that went through reorganization or liquidation since financial distress would also affect regulatory supervision and enforcement incentives. I lose twenty one insurers due to data limitations (i.e., Compustat has missing statutory data for some public insurers). As the result, I have sixty two insurance groups in the final sample. TABLE D1 Sample Selection Steps # Unique Firms 120 Largest (by gross premiums) public insurance groups in 2006 Less: Mergers & Acquisitions (M&A) between 2006 and 2015 Financial distress (liquidation, bankruptcy) between 2006 and 2015 Missing required financial statement data Final Sample (33) (4) (21) 62 72 APPENDIX E Main Results TABLE E1 Pooled Sample Descriptive Statistics and Correlations Panel A: Descriptive Statistics Panel A of Table E1 provides descriptive statistics for the pooled sample of large public insurers between 2006 and 2015. N is the number of firm-year observations. All variables are defined in Appendix C. Variable Captive (P/C) Captive (Life) C_Number (P/C) C_Number (Life) C_Foreign (P/C) C_Foreign (Life) C_Form (P/C) C_Form (Life) Enforcement (P/C) Enforcement (Life) Illiquidity (P/C) Illiquidity (Life) Log (Spread) (P/C) Log (Spread) (Life) ZeroReturn (P/C) ZeroReturn (Life) PriceImpact (P/C) PriceImpact (Life) Analysts (P/C) Analysts (Life) CR_All (P/C) CR_All (Life) CR_Rated (P/C) CR_Rated (Life) Premiums / Surplus (P/C) PSR (P/C) Assets / Surplus (Life) ASR (> 90%) ASR (> 75%) ASR (< 25%) ASR (25% < > 90%) Reserves / Surplus (Life) RSR2 (Life) N 260 264 260 264 260 264 260 264 260 264 192 190 192 190 192 190 192 190 165 201 247 255 167 212 260 260 264 264 264 264 264 264 264 Mean 0.3962 0.4318 0.4769 1.7879 0.7000 0.6932 0.0192 0.1098 -0.2420 -0.0358 0.2331 -0.1098 -9.0582 -10.725 0.0562 0.0349 32.7902 0.9602 9.2217 9.6269 -13.1903 -10.8118 -8.4910 -8.3396 1.1653 0.0577 15.7315 0.1023 0.2500 0.0720 0.1742 9.8078 0.6667 73 SD 0.4900 0.4963 0.6542 3.4233 0.9060 0.8592 0.1376 0.3133 0.5881 0.8296 1.3016 0.1900 2.4538 1.6853 0.0618 0.0373 186.3751 3.7199 4.5794 3.8157 7.1568 5.9568 2.7040 2.5102 0.6643 0.2336 13.9224 0.3036 0.4338 0.2589 0.3800 7.5413 0.4723 P25 P50 P75 0 0 0 0 0 0 0 0 -0.6042 -0.6564 -0.2011 -0.2024 -10.7501 -11.2963 0.0151 0.0142 0.0170 0.0154 6.5 7 -23 -11 -9 -9 0.6311 0 5.6082 0 0 0 0 3.2861 0 0 0 0 0 0 0 0 0 -0.4626 -0.3616 -0.1214 -0.1690 -9.0756 -10.4961 0.0361 0.0234 0.2472 0.0346 9 10.25 -9 -9 -9 -8 1.0651 0 12.5257 0 0 0 0 8.0028 1 1 1 1 2 2 2 0 0 -0.0315 0.5368 0.0544 -0.0950 -7.9671 -9.2048 0.0758 0.0400 0.9909 0.1649 12 12 -8 -7 -7 -7 1.5085 0 20.0628 0 0.5 0 0 14.0022 1 TABLE E1 (cont’d) Variable RSR3 (Life) RSR4 (Life) RSR5 (Life) Total Assets ($ millions) (P/C) Total Assets ($ millions) (Life) Gross Premiums ($ millions) (P/C) Gross Premiums ($ millions) (Life) Total Assets ($ millions) (P/C) Total Assets ($ millions) (Life) Surplus ($ millions) (P/C) Surplus ($ millions) (Life) Reserves ($ millions) (P/C) Reserves ($ millions) (Life) Reinsurance (P/C) Reinsurance (Life) ROA (P/C) ROA (Life) Investment Yield (P/C) Investment Yield (Life) Reinsurance Inefficiency (P/C) Reinsurance Inefficiency (Life) Investment Inefficiency (P/C) Investment Inefficiency (Life) Cash / Total Assets (P/C) Cash / Total Assets (Life) Debt / Total Assets (P/C) Debt / Total Assets (Life) Retained Earnings / Total Assets (P/C) Retained Earnings / Total Assets (Life) Tax Rate (P/C) Tax Rate (Life) N 264 264 264 260 264 260 264 260 264 260 264 260 264 260 264 260 264 260 264 260 264 260 264 260 264 260 264 258 264 260 264 Mean 0.3939 0.1250 0.5378 13727 77895 3238 6252 13727 77895 3185 4527 7902 42464 0.2332 0.2128 0.0249 0.0136 0.0762 0.2069 0.1426 0.6837 0.0762 0.0496 0.0414 0.0252 0.0741 0.0415 0.1617 0.0982 0.1910 -0.2067 74 SD 0.4895 0.3313 0.4995 22936 140353 5326 8904 22937 140353 4999 6132 14294 67372 0.2557 0.2700 0.0396 0.0153 0.0511 0.1891 0.5577 3.3180 0.0511 0.2920 0.0660 0.0281 0.0825 0.0266 0.2447 0.0912 1.7566 8.4338 P25 0 0 0 1809 9748 532.7 656.9 1809 9748 510.4 828.7 970 6255 0.0677 0.0565 0.0149 0.0037 0.0575 0.1102 0 0 0.0575 0.0134 0.0057 0.0062 0.0353 0.0231 0.1214 0.0260 0.0577 0.0761 P50 0 0 1 5632 22793 1049 3068 5632 22792 1261 2437 2857 14720 0.1410 0.1036 0.0287 0.0101 0.0792 0.1717 0 0 0.0792 0.0240 0.0238 0.0189 0.0549 0.0371 0.1909 0.0918 0.2507 0.2817 P75 1 0 1 16108 64151 2963 7369 16108 64151 3507 5469 9154 48882 0.3430 0.2514 0.0447 0.0227 0.0973 0.2453 0 0 0.0973 0.0378 0.0520 0.0362 0.0767 0.0578 0.2702 0.1616 0.4110 0.4677 TABLE E1 (cont’d) Variable Enforcement2 (P/C) Enforcement2 (Life) Affiliates (P/C) Affiliates (Life) Overlapping Regulators (P/C) Overlapping Regulators (Life) Regulatory Resources (Total) (P/C) Regulatory Resources (Total) (Life) Captive Law Regulators (Total) (P/C) Captive Law Regulators (Total) (Life) Strict Regulators (Total) (P/C) Strict Regulators (Total) (Life) Coastal Regulators (Total) (P/C) Coastal Regulators (Total) (Life) Elected Regulators (Total) (P/C) Elected Regulators (Total) (Life) Republican Regulators (Total) (P/C) Republican Regulators (Total) (Life) Regulatory Resources (Aver.) (P/C) Regulatory Resources (Aver.) (Life) Captive Law Regulators (Uniq.) (P/C) Captive Law Regulators (Uniq.) (Life) Strict Regulators (Uniq.) (P/C) Strict Regulators (Uniq.) (Life) Coastal Regulators (Uniq.) (P/C) Coastal Regulators (Uniq.) (Life) Elected Regulators (Uniq.) (P/C) Elected Regulators (Uniq.) (Life) Republican Regulators (Uniq.) (P/C) Republican Regulators (Uniq.) (Life) N 260 264 260 264 260 264 260 264 260 264 260 264 260 264 260 264 260 264 260 264 260 264 260 264 260 264 260 264 260 264 Mean -0.1337 0.1317 10.4692 10.5151 4.6346 5.1591 11.9406 12.5508 2.1154 3.1894 1.7808 2.0189 1.8077 2.4470 0.8000 1.1402 3.7533 5.2161 1.5408 1.6818 1.3115 2.2159 1.0269 1.3447 2.3731 2.9129 0.8000 1.0379 1.9077 2.6061 75 SD 0.7961 1.1530 10.4110 9.4140 3.2121 4.1536 7.4595 7.0382 1.5848 3.2561 2.4277 2.3617 1.7089 2.3032 1.0427 1.7461 2.5004 4.3620 0.7034 0.8397 1.4727 2.2341 1.1129 1.1093 1.9301 2.3195 1.0427 1.3920 1.6057 2.5492 P25 -0.6147 -0.7099 4 4 2 2 7.8496 7.2459 1 1 0 1 0 1 0 0 1.92 1.88 0.9947 1.1491 0 1 0 1 1 1.5 0 0 1 1 P50 -0.3824 -0.1580 7 7 4 4 9.7016 9.3441 2 2 1 1 2 2 1 0 3.03 3.96 1.3440 1.4704 1 2 1 1 2 2 1 0 2 2 P75 0.3553 0.7504 13 15.5 6 7 12.7117 17.1354 3 4 2 2 3 4 1 2 5.92 6.12 1.8567 2.0200 2 3 2 2 3 4 1 2 2 4 TABLE E1 (cont’d) Panel B: Correlation Matrix Panel B of Table E1 provides correlations for the pooled sample of large public insurers between 2006 and 2015. Pearson (Spearman) correlations are presented above (below) the diagonal. Correlations in bold are significant at the 5% level or better. All variables are defined in Appendix C. Continuous variables are winsorized at the 1st and 99th percentiles. 1 2 3 4 5 6 7 8 9 10 11 Variable Captive C_Number Enforcement Illiquidity Analysts CR_All Assets / Surplus Surplus Constraint Reserves / Surplus Premiums Reinsurance 1 2 3 4 5 6 7 8 9 10 11 Variable Captive C_Number Enforcement Illiquidity Analysts CR_All Premiums / Surplus Surplus Constraint Assets / Surplus Premiums Reinsurance 1 0.9246 0.5126 -0.2088 0.1837 0.1126 0.0244 -0.2261 0.0007 0.3974 0.3956 1 0.9849 0.0121 -0.1285 -0.1368 0.3950 -0.4084 0.0447 -0.3060 0.3664 -0.1056 2 0.6002 0.4616 -0.1529 0.1798 0.1797 0.1752 -0.1303 0.1421 0.4079 0.3609 2 0.9017 -0.0237 -0.1240 -0.1505 0.4149 -0.4304 0.0334 -0.3445 0.3827 -0.0964 3 0.4980 0.3032 0.0265 0.1687 -0.0849 0.0349 -0.1350 -0.0881 0.4804 0.3536 3 0.0173 -0.0387 0.1356 0.0464 0.3323 0.1068 0.0469 0.0661 0.3908 -0.3515 Life Insurers 4 5 -0.1381 0.0794 -0.0585 0.0586 -0.0989 0.1038 -0.1491 -0.1543 -0.0032 0.4101 -0.0327 0.3128 0.2535 0.1797 -0.0863 0.3349 0.1027 0.4833 -0.0849 0.0740 P/C Insurers 4 5 0.1492 0.0934 0.0824 0.0903 0.0771 -0.0261 0.0528 0.1066 0.0134 -0.1619 -0.2087 -0.0525 0.0089 -0.1073 0.3473 0.1494 -0.1148 -0.2275 -0.1441 -0.0824 76 6 0.1860 0.2022 0.1161 -0.0540 0.3419 0.2316 0.1890 0.2346 0.4855 -0.1691 6 0.1720 0.1847 0.3954 0.0041 -0.0668 -0.2133 0.0429 -0.1682 0.8139 -0.3157 7 -0.0639 0.2069 -0.1122 0.0207 0.4995 0.1894 0.1489 0.9754 0.1030 0.1057 7 -0.0523 -0.0475 -0.0092 -0.0237 -0.2757 0.0376 0.2346 0.1012 -0.2273 -0.1945 8 -0.1585 0.0753 -0.1526 0.1730 0.1579 0.1013 0.4497 0.1560 0.2722 -0.1131 8 -0.1330 -0.1302 0.3208 -0.0182 -0.1376 0.2085 0.2545 0.1229 0.0156 -0.2033 9 -0.1231 0.1998 -0.1718 -0.0508 0.4598 0.1120 0.9326 0.3368 0.0282 0.1007 9 -0.0519 -0.0474 -0.0117 -0.0231 0.1142 0.0365 0.9998 0.2525 -0.1361 0.0019 10 0.2168 0.0728 0.3371 -0.0774 0.4991 0.3660 0.3107 0.1302 0.2003 11 0.3764 0.1532 0.2260 -0.0194 -0.1010 -0.1759 -0.0002 -0.0654 0.0100 -0.1979 -0.1055 10 0.0914 0.0345 0.6849 -0.0777 -0.1013 0.4310 -0.0247 0.3712 -0.0263 -0.1300 11 -0.0612 -0.0708 -0.2090 -0.0209 -0.1005 -0.2251 -0.0558 -0.1731 -0.0561 -0.2409 TABLE E1 (cont’d) 12 13 14 15 16 17 18 19 20 21 22 Variable 12 C_Number Enforcement 0.5526 Premiums 0.4062 Reserves / Surplus 0.0807 Reinsurance Inefficiency 0.0138 Investment Inefficiency 0.1321 Tax Rate 0.1192 ROA -0.1471 Investment Yield -0.1682 Cash / Total Assets -0.0437 Debt / Total Assets 0.2873 12 13 14 15 16 17 18 19 20 21 22 Variable C_Number Enforcement Premiums Premiums / Surplus Reinsurance Inefficiency Investment Inefficiency Tax Rate ROA Investment Yield Cash / Total Assets Debt / Total Assets 12 0.0142 0.1898 -0.4264 0.4437 0.0891 -0.1986 0.2103 -0.0183 0.0423 -0.0297 13 0.3032 0.5020 -0.1367 0.0143 0.2252 -0.0209 0.1860 -0.4744 0.2390 0.5424 13 -0.0387 0.5096 0.1715 -0.2714 -0.2933 -0.0110 0.0218 0.0364 -0.0994 0.2888 14 0.0728 0.3371 -0.0551 0.2382 0.3198 0.0903 0.1280 -0.3062 0.1158 0.4404 14 0.0345 0.6849 0.1540 0.2806 -0.2204 -0.1719 0.1463 0.1456 -0.3093 0.1615 Life Insurers 15 16 0.1998 -0.0271 -0.1718 -0.1327 0.2003 -0.0316 -0.1975 -0.3882 -0.0552 0.1743 0.2134 -0.1851 -0.7235 0.2567 0.7099 -0.2602 -0.2741 0.3674 -0.2641 0.2829 P/C Insurers 15 16 -0.0475 0.2610 -0.0092 -0.1234 -0.0247 0.0175 -0.0159 -0.2078 0.0631 0.1567 0.2974 -0.1993 -0.1984 0.1500 0.0801 0.0845 -0.0020 -0.0011 0.1648 -0.2127 77 17 -0.0320 0.0315 -0.0083 -0.0574 0.0462 -0.0801 -0.0409 -0.1692 0.1050 -0.0144 17 -0.0045 -0.2084 -0.1767 -0.0075 -0.0185 -0.1343 -0.1525 -0.3330 0.1449 -0.0638 18 0.0515 -0.0145 0.0300 0.1079 -0.0011 0.0035 -0.3322 0.0747 -0.0795 -0.0182 18 -0.0799 0.0385 0.0188 0.0045 0.0397 -0.0088 -0.3427 0.0035 0.0038 0.0412 19 -0.3102 0.1922 -0.0080 -0.5986 0.2792 0.0051 -0.0610 -0.4800 0.1617 0.2615 19 0.1506 0.0908 0.1551 -0.0090 0.0240 -0.1553 -0.0345 0.1658 -0.0152 -0.2159 20 -0.1336 -0.3293 -0.1579 0.5145 -0.1184 -0.0826 0.0331 -0.3650 -0.5203 -0.5034 20 0.0017 -0.0094 0.0484 -0.0209 0.0475 -0.1184 -0.1350 0.1250 -0.2944 -0.2760 21 0.0290 0.2371 -0.1069 -0.2241 0.1858 0.0959 -0.0331 0.2193 -0.2793 22 0.2350 0.4027 0.0808 -0.1709 0.0440 0.1554 0.0707 0.2650 -0.3498 0.3814 0.3418 21 0.0753 -0.1329 -0.2310 0.0021 -0.0499 0.1477 -0.0495 0.1250 -0.1670 0.1142 22 -0.1201 -0.0941 -0.0592 0.0204 -0.0828 0.2234 0.0551 -0.4101 -0.2311 0.2137 TABLE E1 (cont’d) 23 24 25 26 27 28 29 30 31 32 33 Variable C_Number C_Foreign Enforcement Assets / Surplus Surplus Constraint Premiums Reinsurance Inefficiency Investment Inefficiency Tax Rate ROA Investment Yield 23 24 25 26 27 28 29 30 31 32 33 Variable C_Number C_Foreign Enforcement Premiums / Surplus Surplus Constraint Premiums Reinsurance Inefficiency Investment Inefficiency Tax Rate ROA Investment Yield 23 0.9217 0.5526 0.1822 -0.0972 0.4062 0.0138 0.1321 0.1192 -0.1471 -0.1682 23 0.9696 0.0142 -0.4264 -0.1353 0.1898 0.4437 0.0891 -0.1986 0.2103 -0.0183 24 0.6151 0.5570 0.0060 -0.1549 0.3603 -0.0012 0.0768 0.0728 0.0120 -0.3214 24 0.8937 0.0011 -0.3548 -0.1370 0.2324 0.4996 0.0892 -0.2124 0.2093 0.0058 25 0.3032 0.5168 -0.0326 -0.1301 0.5020 0.0143 0.2252 -0.0209 0.1860 -0.4744 25 -0.0387 -0.0256 0.1715 0.2625 0.5096 -0.2714 -0.2933 -0.0110 0.0218 0.0364 Life Insurers 26 27 0.2069 0.0753 -0.0862 -0.1501 -0.1122 -0.1526 0.4497 0.1143 0.0276 0.1782 -0.4298 0.0887 0.0082 0.1030 0.2301 0.1309 -0.7073 -0.1018 0.6499 0.0153 P/C Insurers 26 27 -0.0475 -0.1302 -0.0497 -0.1368 -0.0092 0.3208 0.2545 0.4038 0.1540 0.2718 -0.2078 -0.0866 0.0631 -0.0992 0.2974 0.1401 -0.1984 0.1173 0.0801 -0.0595 78 28 0.0728 0.1695 0.3371 0.3107 0.1302 0.2382 0.3198 0.0903 0.1280 -0.3062 28 0.0345 0.1349 0.6849 -0.0247 0.3712 0.2806 -0.2204 -0.1719 0.1463 0.1456 29 -0.0271 -0.0484 -0.1327 -0.1721 0.1479 -0.0316 0.1743 -0.1851 0.2567 -0.2602 29 0.2610 0.3558 -0.1234 -0.0159 -0.0525 0.0175 0.1567 -0.1993 0.1500 0.0845 30 -0.0320 -0.0497 0.0315 -0.0379 -0.0080 -0.0083 0.0462 -0.0801 -0.0409 -0.1692 30 -0.0045 0.0173 -0.2084 -0.0075 -0.0811 -0.1767 -0.0185 -0.1343 -0.1525 -0.3330 31 0.0515 0.0431 -0.0145 0.0653 0.0674 0.0300 -0.0011 0.0035 -0.3322 0.0747 31 -0.0799 -0.1081 0.0385 0.0045 0.0610 0.0188 0.0397 -0.0088 -0.3427 0.0035 32 -0.3102 0.0149 0.1922 -0.5095 -0.0219 -0.0080 0.2792 0.0051 -0.0610 33 -0.1336 -0.3176 -0.3293 0.3419 0.0436 -0.1579 -0.1184 -0.0826 0.0331 -0.3650 -0.4800 32 0.1506 0.1772 0.0908 -0.0090 0.0917 0.1551 0.0240 -0.1553 -0.0345 0.1658 33 0.0017 0.0224 -0.0094 -0.0209 -0.0763 0.0484 0.0475 -0.1184 -0.1350 0.1250 TABLE E1 (cont’d) 34 35 36 37 38 39 40 41 42 43 44 Variable Enforcement Illiquidity Analysts CR_All CR_Rated Assets / Surplus Reinsurance Inefficiency Investment Inefficiency Tax Rate ROA Investment Yield 34 35 36 37 38 39 40 41 42 43 44 Variable Enforcement Illiquidity Analysts CR_All CR_Rated Premiums / Surplus Reinsurance Inefficiency Investment Inefficiency Tax Rate ROA Investment Yield 34 -0.0033 0.1406 0.0639 -0.0662 -0.0662 -0.1978 0.3048 0.0038 0.1826 -0.4568 34 0.1944 -0.0485 0.2727 0.2727 0.2634 -0.4507 -0.5110 0.2309 0.1515 0.0199 35 -0.0989 -0.2586 0.0559 -0.0288 -0.0288 -0.0932 0.0987 0.1117 0.1323 -0.0396 35 0.0771 0.1382 -0.3519 -0.3519 -0.0424 -0.3468 -0.1764 0.1414 -0.2837 -0.2175 36 0.1038 -0.1491 0.2898 0.3544 0.3544 -0.1545 0.1445 0.1342 -0.2298 0.1126 36 -0.0261 0.0528 -0.1091 -0.1091 -0.1176 -0.3250 -0.0254 0.1326 -0.0653 -0.0494 Life Insurers 37 38 -0.1122 0.1161 0.0207 -0.0540 0.4995 0.3419 0.1894 -0.0365 -0.0365 1.0000 -0.5557 0.2624 -0.0746 0.4528 0.2537 -0.1092 -0.6569 0.0039 0.5175 0.0339 P/C Insurers 37 38 0.3954 0.3517 0.0041 -0.0171 -0.0668 -0.4341 1.0000 1.0000 -0.1002 -0.1002 0.3823 0.3823 -0.1599 -0.1599 -0.0526 -0.0526 0.4775 0.4775 0.2018 0.2018 79 39 0.0889 0.0083 0.1670 -0.0877 1.0000 0.2624 0.4528 -0.1092 0.0039 0.0339 39 -0.0092 -0.0237 -0.2757 0.0376 -0.0137 -0.4527 -0.2970 0.4868 -0.2859 0.1038 40 -0.1327 -0.0370 -0.1609 -0.1721 0.1524 0.1800 0.2073 -0.1889 0.2770 -0.2850 40 -0.1234 -0.0094 -0.1484 0.2037 0.1424 -0.0159 0.3408 -0.4212 0.2066 -0.0533 41 0.0315 -0.0390 -0.0669 -0.0379 -0.1138 0.2669 0.0462 -0.1499 0.1460 -0.3294 41 -0.2084 0.0455 -0.0815 -0.2031 -0.0322 -0.0075 -0.0185 -0.3876 -0.2459 -0.3439 42 -0.0145 -0.0482 0.0608 0.0653 0.0469 -0.0004 -0.0011 0.0035 -0.3966 0.0507 42 0.0385 -0.0280 0.0301 0.0461 -0.2114 0.0045 0.0397 -0.0088 -0.3386 0.0920 43 0.1922 0.0779 -0.2507 -0.5095 0.0743 0.2742 0.2792 0.0051 -0.0610 44 -0.3293 -0.0311 0.0697 0.3419 -0.0285 -0.4585 -0.1184 -0.0826 0.0331 -0.3650 -0.3342 43 0.0908 0.0883 -0.2195 0.1496 0.3647 -0.0090 0.0240 -0.1553 -0.0345 0.3680 44 -0.0094 0.0522 -0.0537 -0.0048 0.0993 -0.0209 0.0475 -0.1184 -0.1350 0.1250 TABLE E2 Descriptive Statistics and Univariate Tests of Differences Panel A: Life Insurers Panel A of Table E2 reports the results for the sub-sample of life insurers. This table reports the mean differences across two groups: insurance groups with at least one captive reinsurer in their structure (Captive) and insurance groups with no captive reinsurance subsidiaries. In reported results, I do not control for the main regulator. The differences between variable means differ from the results reported below when I compare firms with the same main regulator. All variables are defined in Appendix C. Significance at the .10, .05 and .01 level for two-sided tests is denoted by *, ** and ***, respectively. Variable Enforcement Illiquidity Log (Spread) PriceImpact ZeroReturns Analysts S&P Credit Ratings (CR_All) S&P Credit Ratings (CR_Rated) Premiums Assets / Surplus Surplus Constraint Reserves / Surplus RSR5 Reinsurance Reinsurance Inefficiency Investment Inefficiency ROA Investment Yield Cash / Total Assets Debt / Total Assets Affiliates Overlapping Regulators Regulatory Resources (Total) Captive Law Regulators (Total) Strict Regulators (Total) Elected Regulators (Total) Coastal Regulators (Total) Republican Regulators (Total) Regulatory Resources (Ave.) Captive Law Regulators (Uniq.) Strict Regulators (Uniq.) Elected Regulators (Uniq.) Coastal Regulators (Uniq.) Republican Regulators (Uniq.) Mean [Captive = 1] 0.4372 -0.1387 -10.4760 0.8767 0.0301 9.9495 -9.5625 -7.6429 8462 14.7113 0.1053 8.7445 0.4825 0.3292 0.5304 0.0337 0.0138 0.1460 0.0221 0.0467 15.7105 7.2017 16.9108 4.7895 2.6228 1.6491 3.7192 8.0082 1.3034 3.3684 1.7281 1.4122 4.1403 3.7456 Mean [Captive = 0] -0.3952 -0.0860 -9.9215 1.0292 0.0388 9.3435 -11.7902 -8.9386 4572 16.5068 0.2267 10.6159 0.5800 0.1244 0.8003 0.0617 0.0135 0.2533 0.0276 0.0375 6.5667 3.6067 9.2372 1.9733 1.5600 0.7533 1.4800 3.0939 1.9694 1.3400 1.0533 0.7533 1.9800 1.7400 80 Difference 0.8324 -0.0527 -0.5545 -0.1525 -0.0085 .6060 2.2277 1.2957 3890 -1.7955 -0.1214 -1.8714 -0.0975 0.2058 -0.2699 -0.0279 0.0003 -0.1073 -0.0055 0.0092 9.1439 3.5950 7.6736 2.8162 1.0628 0.8958 2.2392 4.9143 -0.6660 2.0284 0.6747 0.6589 2.1603 2.0056 t-stat 9.2952 1.9116 2.2822 -0.2807 1.6048 1.1242 3.0105 3.8692 3.5949 1.0366 2.5990 2.0086 1.5761 6.5770 -0.6539 -0.7695 0.1212 -4.7481 -1.6014 2.8027 8.9052 7.6992 10.4143 7.6918 3.7086 4.2616 8.9161 10.9158 6.9306 8.1696 5.1251 3.9120 8.4369 6.8650 *** ** *** * *** *** *** *** ** * *** *** * *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** TABLE E2 (cont’d) Panel B: P/C Insurers Panel B of Table E2 reports the results for the sub-sample of P/C insurers. This table reports the mean differences across two groups: insurance groups with at least one captive reinsurer in their structure (Captive) and insurance groups with no captive reinsurance subsidiaries. In reported results, I do not control for the main regulator. The differences between variable means differ from the results reported below when I compare firms with the same main regulator. All variables are defined in Appendix C. Significance at the .10, .05 and .01 level for two-sided tests is denoted by *, ** and ***, respectively. Variable Enforcement Illiquidity Log (Spread) PriceImpact ZeroReturns Analysts S&P Credit Ratings (CR_All) S&P Credit Ratings (CR_Rated) Premiums Premiums / Surplus Surplus Constraint Reinsurance Reinsurance Inefficiency Investment Inefficiency ROA Investment Yield Cash / Total Assets Debt / Total Assets Affiliates Overlapping Regulators Regulatory Resources (Total) Captive Law Regulators (Total) Strict Regulators (Total) Elected Regulators (Total) Coastal Regulators (Total) Republican Regulators (Total) Regulatory Resources (Ave.) Captive Law Regulators (Uniq.) Strict Regulators (Uniq.) Elected Regulators (Uniq.) Coastal Regulators (Uniq.) Republican Regulators (Uniq.) Mean [Captive = 1] -0.2295 0.4918 -8.7889 85.8335 0.0630 9.7376 -11.7128 -8.5696 38372 0.8783 0.0194 0.2139 0.3520 0.0450 0.0312 0.0778 0.0477 0.0642 11.0388 4.9320 12.8619 2.4660 1.2913 0.8447 1.8641 3.4109 1.5484 2.0777 0.9515 0.8447 2.3204 1.9903 81 Mean [Captive = 0] -0.2502 0.0880 -9.2092 3.0341 0.0525 8.8690 -14.2123 -8.4205 2844 1.3547 0.0828 0.2458 0.0052 0.0479 0.0207 0.0751 0.0371 0.0806 10.0955 4.4395 11.3362 1.8854 2.1019 0.7707 1.7707 3.9780 1.5358 0.8089 1.0764 0.7707 2.4076 1.8535 Difference 0.0207 0.4038 0.4203 82.7994 0.0105 0.8685 2.4995 -0.1492 3238 -0.4764 -0.0634 -0.0319 0.3468 -0.0029 0.0105 0.0027 0.0106 -0.0164 0.9433 0.4925 1.5257 0.5806 -0.8106 0.0740 0.0934 -0.5671 0.0126 1.2688 -0.1249 0.0740 -0.0873 0.1368 t-stat 0.2777 2.0802 1.1399 3.0156 1.1343 1.1980 2.7337 -0.3550 1.4742 6.0224 2.1548 0.9841 5.1388 -0.4015 2.1080 0.4153 1.2561 -1.5656 0.7139 1.2104 1.6181 2.9316 2.6642 0.5586 0.4302 1.7965 0.1418 7.4802 -0.8852 0.5586 -0.3559 0.6711 ** *** *** * *** ** *** ** * * * *** *** ** *** TABLE E3 Regulatory Enforcement Factor Panel A: Eigenvalues in the Principal Component Analysis Table E3 reports the results from the principal component analysis (PCA) of seven regulatory enforcement proxies. Panel A of Table E3 reports the eigenvalues of the factors identified in the principal component analysis. All variables are defined in Appendix C. Factor Eigenvalue Difference Proportion Cumulative Factor 1 5.0978 4.1772 0.7283 0.7283 Factor 2 0.9205 0.5743 0.1315 0.8598 Factor 3 0.3462 0.0909 0.0495 0.9092 Factor 4 0.2553 0.0367 0.0365 0.9457 Factor 5 0.2186 0.1078 0.0312 0.9769 Factor 6 0.1108 0.0600 0.0158 0.9927 Factor 7 0.0508 0.0073 1.0000 Panel B: Coefficients in the Principal Component Analysis Table 3 reports the results from the principal component analysis (PCA) of seven regulatory enforcement proxies. Panel B of Table 3 presents the factor loadings for Factor 1 (Enforcement). All variables are defined in Appendix C. Variable Overlapping Regulators Coefficient 0.14 Regulatory Resources 0.17 Captive Law Regulators 0.18 Strict Regulators 0.17 Elected Regulators 0.16 Coastal Regulators 0.17 Republican Regulators 0.17 82 TABLE E4 Regulatory Enforcement and Captive Reinsurers Panel A: Firm Type and Captive Reinsurance Panel A of Table E4 reports the results from the estimation of equation (1) using a Tobit model. The dependent variable is C_Number. Columns (1) – (3) report the results for pure life-health (L/H) insurers and diversified insurers (P/C + L/H) between 2006 and 2015. Column (4) reports the results for pure propertycasualty (P/C) insurers between 2006 and 2015. Robust standard errors are clustered by firm. All variables are defined in Appendix C, and continuous variables are winsorized at the 1st and 99th percentiles. Significance at the .10, .05 and .01 level for two-sided tests is denoted by *, ** and ***, respectively. DV = C_Number Life Insurers (2) ASR > 75% (1) ASR > 90% Variable Predict. Enforcement H1 (-) Assets-to-Surplus Ratio (ASR) ASR * Enforcement H2 (-) Assets / Surplus (AS) AS * Enforcement Premium-to-Surplus Ratio (PSR) PSR * Enforcement H2 (-) Premiums / Surplus (PS) PS * Enforcement Log (Premiums) Reinsurance Inefficiency Investment Inefficiency Tax Rate Reinsurance ROA Investment Yield Cash / Total Assets Debt / Total Assets Constant Year & Regulator FE Observations Adjusted R-squared Log pseudolikelihood Left / Right - censored Coeff -2.0606 -0.6082 -1.4287 -0.0319 0.1424 1.2807 0.9722 -5.7474 0.0115 2.9708 0.2456 -0.9868 -3.0264 -19.7433 -7.4371 SE 0.8101 0.2714 0.6573 0.0483 0.0521 0.3244 0.4707 3.4891 0.0051 0.9425 8.4064 0.8609 3.3566 8.4615 2.4878 Yes 264 0.4476 -255.8343 150 / 114 *** ** ** *** *** ** * ** *** ** *** Coeff -1.9242 1.6668 6.3089 -0.0134 0.0929 1.6707 0.0017 -13.2160 -0.0019 1.1255 -39.8335 -1.9673 2.9694 -6.7551 -9.8426 SE 1.3307 1.2728 2.8144 0.1082 0.1541 0.5632 0.1301 7.1892 0.0081 1.1906 22.4781 2.4988 11.8126 20.6981 3.8409 Yes 264 0.4604 -249.9175 150 / 114 83 P/C Insurers (4) PSR > 90% (3) ASR < 25% ** *** * * *** Coeff -4.0348 -11.1209 -14.9582 0.1374 0.1730 2.2824 -0.1112 -12.4245 0.0007 2.0777 -26.7995 -2.4844 9.7898 -6.8758 -16.0329 SE 0.8467 4.2026 6.5325 0.0765 0.1094 0.6442 0.1336 5.7281 0.0085 1.2358 29.5120 2.8323 14.0289 21.3924 4.5371 Yes 264 0.4595 -250.3552 150 / 114 *** *** ** * *** ** * *** 0.6488 0.3595 * 0.8313 -1.7503 -1.2702 -0.7211 0.7997 0.1116 -1.1244 0.0827 2.0478 0.4612 1.1586 0.0727 -1.0790 -2.7553 0.3521 0.9547 0.2261 0.2701 0.1132 0.0592 1.3817 0.0273 0.6741 1.1751 1.4758 3.7826 1.8679 0.8882 ** * *** *** *** * Yes 260 0.5708 -119.3915 157 / 103 *** *** *** TABLE E4 (cont’d) Panel B: Captive Reinsurance across Time Periods [Life Insurers] Panel B of Table E4 reports the results from the estimation of equation (1) using a Tobit model. The dependent variable is C_Number. Column (1) reports the results for the entire time period between 2006 and 2015 for life insurers. Column (2) reports the results for the 2006-2011 time period for life insurers. Column (3) report the results for the 2012 – 2015 time period. Robust standard errors are clustered by firm. All variables are defined in Appendix C, and continuous variables are winsorized at the 1st and 99th percentiles. Significance at the .10, .05 and .01 level for two-sided tests is denoted by *, ** and ***, respectively. DV = C_Number (2) [2006-2011] (1) [2006-2015] Variable Predict. Enforcement H1 (-) Surplus Constraint (SC) SC * Enforcement H2 (-) Assets / Surplus (AS) AS * Enforcement Log (Premiums) Reinsurance Inefficiency Investment Inefficiency Tax Rate Reinsurance ROA Investment Yield Cash / Total Assets Debt / Total Assets Constant Year & Regulator FE Observations Adjusted R-squared Log pseudolikelihood Left / Right - censored Coeff -4.2150 -3.2601 -0.6738 0.2118 0.2354 1.8157 -0.0324 -13.2146 0.0012 1.0041 -15.2373 -2.9150 7.4127 7.2795 -13.6579 SE 1.2212 0.9951 1.8237 0.1039 0.1495 0.5043 0.1260 5.8663 0.0075 1.1691 28.6243 3.1145 13.0073 18.1473 4.0634 Coeff -6.1263 -2.5498 -4.3358 0.0405 0.4056 2.2710 0.0394 -3.8002 0.0982 2.1468 -26.1213 0.9122 16.0210 -13.4395 -14.9616 *** *** ** *** ** *** Yes 264 0.4636 -248.4673 150 / 114 SE 0.6950 0.5027 0.9534 0.0484 0.0811 0.4006 0.2982 1.9059 0.0273 0.7240 16.3260 0.7078 5.4072 10.6703 2.5547 Yes 151 0.7979 -50.9031 90 / 61 84 (3) [2012 - 2015] *** *** *** *** *** ** *** *** *** *** Coeff -3.2884 -2.5316 -1.2983 0.4285 0.1938 1.2176 0.3932 -38.5848 0.0005 -0.7747 -7.0481 -22.5374 46.1049 17.4617 -8.3902 SE 1.4344 0.8671 1.8748 0.1377 0.1593 0.5723 0.2301 14.1597 0.0106 1.3327 23.4287 9.7201 23.5189 28.9450 3.7463 Yes 113 0.5216 -100.8728 60 / 53 ** *** *** ** * *** ** ** ** TABLE E4 (cont’d) Panel B: Captive Reinsurance across Time Periods [P/C Insurers] Panel B of Table E4 reports the results from the estimation of equation (1) using a Tobit model among P/C insurers. The dependent variable is C_Number. Column (1) reports the results for the entire time period between 2006 and 2015.Column (2) reports the results for the 2006-2011 time period for P/C insurers. Column (3) report the results for the 2012 – 2015 time period. Robust standard errors are clustered by firm. All variables are defined in Appendix C, and continuous variables are winsorized at the 1st and 99th percentiles. Significance at the .10, .05 and .01 level for two-sided tests is denoted by *, ** and ***, respectively. DV = C_Number (2) [2006-2011] (1) [2006-2015] Variable Predict. Enforcement H1 (-) Surplus Constraint (SC) SC * Enforcement H2 (-) Premium / Surplus (PS) PS * Enforcement Log (Premiums) Reinsurance Inefficiency Investment Inefficiency Tax Rate Reinsurance ROA Investment Yield Cash / Total Assets Debt / Total Assets Constant Year & Regulator FE Observations Adjusted R-squared Log pseudolikelihood Left / Right - censored Coeff 0.6488 0.8313 -1.7503 -1.2702 -0.7211 0.7997 0.1116 -1.1244 0.0827 2.0478 0.4612 1.1586 0.0727 -1.0790 -2.7553 SE 0.3595 0.3521 0.9547 0.2261 0.2701 0.1132 0.0592 1.3817 0.0273 0.6741 1.1751 1.4758 3.7826 1.8679 0.8882 Coeff 0.9360 -0.2300 -1.7558 -1.8959 -0.9990 1.0970 0.0354 -0.6724 0.1654 3.2020 -3.2348 0.1203 -5.7938 -3.6097 -3.9231 * ** * *** *** *** * *** *** *** Yes 260 0.5708 -119.3915 157 / 103 SE 0.0189 0.0145 0.0233 0.0164 0.0174 0.0026 0.0047 0.2906 0.0072 0.0536 0.1286 0.1656 0.3592 0.2376 0.0200 Yes 155 0.7581 -39.5959 95 / 60 85 (3) [2012 - 2015] *** *** *** *** *** *** *** ** *** *** *** *** *** *** Coeff 4.1944 2.6677 9.7839 -2.8758 -6.7813 1.2252 1.3196 0.5652 -0.7362 -0.2788 -0.3813 1.5905 -31.4475 10.1062 -3.3841 Yes 105 0.9427 -6.5515 62 / 43 SE 0.0063 0.0099 0.0191 0.0044 0.0048 0.0010 0.0063 0.0708 0.0029 0.0350 0.1708 0.0893 0.0248 0.0608 0.0083 *** *** *** *** *** *** *** *** *** *** ** *** *** *** *** TABLE E5 Regulatory Enforcement, Captive Reinsurers, and Credit Ratings Panel A: Firm Type Panel A of Table E5 reports the results from the estimation of equation (2) using a panel OLS regression. The dependent variable is CR_All. Column (1) reports the results for life insurers, and Column (2) reports the results for P/C insurers between 2006 and 2015. Robust standard errors are clustered by firm. All variables are defined in Appendix C, and continuous variables are winsorized at the 1st and 99th percentiles. Significance at the .10, .05 and .01 level for two-sided tests is denoted by *, ** and ***, respectively. DV = CR_All (1) "Life" Variable C_Number Enforcement Reserves/Surplus (RS) Reserves/ Surplus Ratio (RSR) Premium / Surplus Ratio Log (Premiums) Reinsurance Inefficiency Investment Inefficiency Reinsurance ROA Cash / Total Assets Debt / Total Assets Retained Earnings / Total Assets Constant Year & Regulator FE Observations Adjusted R-squared Predict. H3 (+) Coeff 0.56910 3.14347 0.40597 1.52640 -3.77849 0.75300 0.38958 -2.93106 -5.38211 17.31507 8.88173 28.56863 16.76629 -28.72266 SE 0.21848 1.04997 0.10186 1.77986 2.00398 0.59677 0.16101 1.18028 2.80569 28.30302 18.49395 26.45679 10.56653 6.21876 Yes 255 0.7004 86 (2) "P/C" *** *** *** * ** ** * *** Coeff 0.3083 -1.0189 0.0013 -2.1348 -0.5078 3.6772 0.2768 -11.7287 3.1897 -6.7887 55.0608 14.2136 12.3022 -35.9458 SE 1.1037 0.7618 0.0014 1.7183 1.5204 0.9092 0.4078 6.9006 2.6793 6.4882 20.9066 10.0524 3.9608 5.7743 Yes 247 0.7888 *** * *** *** *** TABLE E5 (cont’d) Panel B: Firm Type and Public Scrutiny of Captive Reinsurance Panel B of Table E5 reports the results from the estimation of equation (2) using a panel OLS regression. The dependent variable is CR_All. Column (1) and (3) report the results for the 2006-2011 time period for life and P/C insurers, accordingly. Column (2) and (4) report the results for the 2012 – 2015 time period for life and P/C insurers, accordingly. Robust standard errors are clustered by firm. All variables are defined in Appendix C, and continuous variables are winsorized at the 1st and 99th percentiles. Significance at the .10, .05 and .01 level for two-sided tests is denoted by *, ** and ***, respectively. DV = CR_All Variable Predict. C_Number Enforcement H3 (+) Reserves/Surplus (RS) Reserves/ Surplus Ratio (RSR) Premium / Surplus Ratio Log (Premiums) Reinsurance Inefficiency Investment Inefficiency Reinsurance ROA Cash / Total Assets Debt / Total Assets Retained Earnings / Total Assets Constant Year & Regulator FE Observations Adjusted R-squared (1) "Life" [2006 - 2011] Coeff SE 0.8484 0.3887 1.7281 1.2680 0.3796 0.0951 -0.2815 2.8282 -2.5479 1.6778 0.9427 0.7690 0.2135 0.1535 -1.7155 1.1071 -4.6304 2.7908 23.2079 39.5089 14.1471 11.9148 3.1336 34.8702 18.8038 14.8289 -19.3388 9.4719 Yes 149 0.7063 ** *** ** (2) "Life" [2012 - 2015] Coeff SE 0.4941 0.2973 5.5380 1.7199 0.4589 0.1866 1.2869 2.4593 -5.9456 4.3731 0.6158 0.6080 1.8583 0.6219 -26.3426 27.4792 -4.8122 4.1433 -41.1119 48.4133 -21.6989 43.4989 38.8464 41.4884 6.0777 10.7232 -36.7751 9.0852 Yes 106 0.8290 87 *** ** *** *** (3) "P/C" [2006 - 2011] Coeff SE -0.4828 1.2329 -1.4213 0.8042 0.0023 0.0011 -3.2135 1.9449 -1.6013 1.2899 4.1836 0.7613 -0.0351 0.2644 -4.9086 9.6810 -1.7172 3.1750 -13.8270 8.7492 53.7631 18.8957 4.0893 9.5908 15.1464 6.1493 -37.4780 5.1317 Yes 150 0.8926 * * *** *** ** *** (4) "P/C" [2012 - 2015] Coeff SE -0.4401 2.5451 0.2918 3.6167 -0.6528 1.0102 2.6764 3.4295 4.9871 10.2810 1.8075 2.7633 9.1560 7.6430 -22.0090 11.1051 1.5836 3.5729 3.5817 17.6741 73.2519 46.4554 23.5273 24.1607 11.4589 14.8093 -24.1410 16.6836 Yes 97 0.7871 * TABLE E6 Illiquidity Factor Panel A: Eigenvalues in the Principal Component Analysis Table E6 reports the results from the principal component analysis (PCA) of three information asymmetry proxies. Panel A of Table E6 reports the eigenvalues of the factors identified in the principal component analysis. All variables are defined in Appendix C. Factor Eigenvalue Difference Proportion Cumulative Factor 1 1.9092 1.1310 0.6364 0.6364 Factor 2 0.7783 0.4657 0.2594 0.8958 Factor 3 0.3125 . 0.1042 1.0000 Panel B: Factor Loadings and Scoring Coefficients Table E6 reports the results from the principal component analysis (PCA) of three information asymmetry proxies. Panel B of Table E6 presents the scoring coefficients for Factor 1 (i.e., Illiquidity). All variables are defined in Appendix C. Variable Log (Spread) Coefficient 0.4520 ZeroReturn 0.4642 PriceImpact 0.3226 88 TABLE E7 Regulatory Enforcement, Captive Reinsurers, and Information Asymmetry Panel A: Life Insurers Panel A of Table E7 reports the results from the estimation of equation (3) using a first-differences regression among life insurers between 2006 and 2015. Column (1) reports the results for the dependent variable Spread, Column (2) reports the results for the dependent variable ZeroReturns, Column (3) reports the results for the dependent variable PriceImpact, and Column (4) reports the results for the dependent variable Illiquidity. Robust standard errors are clustered by firm. All variables are defined in Appendix C, and continuous variables are winsorized at the 1 st and 99th percentiles. Significance at the .10, .05 and .01 level for two-sided tests is denoted by *, ** and ***, respectively. DV = Δ Market Illiquidity Variable Predict. Δ Enforcement H4 (-) Δ C_Number H5 (+) Δ Log (MV) Δ B/M Δ Leverage Δ ROA Δ Log (Turnover) Δ Log (SD_Return) Δ Log (ABN_Return) Δ InstOwn% Constant Year & Regulator FE Observations Within R-squared Between R-squared (1) Δ Log(Spread) Coeff SE -0.3296 0.1418 -0.2975 0.1822 -1.7145 0.3649 0.0000 0.0000 -2.0509 7.2294 -2.5547 3.2068 -0.1599 0.1285 0.4366 0.2110 -11.5103 58.7851 -0.0018 0.0008 17.2816 5.4077 Yes 130 0.7752 0.7542 ** *** ** ** *** (2) Δ ZeroReturn Coeff SE -0.0113 0.0036 -0.0009 0.0074 -0.0336 0.0128 0.0000 0.0000 -0.4515 0.1884 0.1223 0.1181 -0.0098 0.0048 0.0065 0.0110 2.5559 2.2720 0.0005 0.0001 0.5633 0.1992 Yes 130 0.5970 0.5815 *** *** ** ** *** *** (3) Δ PriceImpact Coeff SE -0.8519 0.5398 -1.4469 0.7099 -9.5082 3.7510 0.0000 0.0001 -82.8914 42.2584 -12.4157 16.9640 -2.3208 1.1393 2.9401 2.3258 120.2460 218.5353 -0.0002 0.0063 160.6998 58.5080 Yes 130 0.5094 0.3937 89 ** ** * ** *** (4) Δ Illiquidity Coeff SE -0.0392 0.0203 -0.0389 0.0401 -0.2883 0.1097 0.0000 0.0000 -2.5119 0.9332 0.5659 0.5621 -0.0931 0.0468 0.1433 0.0941 6.7568 16.7823 0.0019 0.0003 4.9418 1.9753 Yes 130 0.4958 0.5217 * *** *** * *** ** TABLE E7 (cont’d) Panel B: P/C Insurers Panel B of Table E7 reports the results from the estimation of equation (3) using a first-differences regression among P/C insurers between 2006 and 2015. Column (1) reports the results for the dependent variable Spread, Column (2) reports the results for the dependent variable ZeroReturns, Column (3) reports the results for the dependent variable PriceImpact, and Column (4) reports the results for the dependent variable Illiquidity. Robust standard errors are clustered by firm. All variables are defined in Appendix C, and continuous variables are winsorized at the 1 st and 99th percentiles. Significance at the .10, .05 and .01 level for two-sided tests is denoted by *, ** and ***, respectively. Variable Predict. Δ Enforcement H4 (-) Δ C_Number H5 (+) Δ Log (MV) Δ B/M Δ Leverage Δ ROA Δ Log (Turnover) Δ Log (SD_Return) Δ Log (ABN_Return) Δ InstOwn% Constant Year & Regulator FE Observations Within R-squared Between R-squared (1) Δ Log(Spread) Coeff SE -0.3198 0.1680 -0.5979 0.0928 -1.9387 0.0754 0.0000 0.0000 -0.3719 1.0069 -1.8341 0.8547 -0.2551 0.1187 0.4477 0.0973 -12.4000 29.5044 -0.0092 0.0024 20.7428 1.3503 Yes 142 0.9222 0.6477 * *** *** ** ** *** *** *** (2) Δ ZeroReturn Coeff SE -0.0135 0.0101 -0.0252 0.0080 -0.0764 0.0151 0.0000 0.0000 -0.1358 0.0345 -0.1631 0.0868 -0.0248 0.0226 -0.0309 0.0127 -6.0026 3.6522 -0.0006 0.0004 1.1240 0.2753 Yes 142 0.4291 0.2498 DV = Δ Market Illiquidity (3) Δ PriceImpact Coeff SE 13.0016 16.8803 *** 0.1955 13.7757 *** -16.5566 31.3996 0.0001 0.0001 *** -150.9025 95.9216 * -95.2662 121.1160 -16.6236 18.7399 ** 36.5550 33.3026 -12168.9100 4848.5070 -1.8243 1.2550 *** 602.1364 530.0397 Yes 142 0.2695 0.2641 90 ** (4) Δ Illiquidity Coeff SE -0.2697 0.4795 -0.5499 0.1841 -0.9737 0.3802 0.0000 0.0000 -1.7713 1.0062 -4.2323 1.8967 0.0671 0.2039 -0.0462 0.2790 -389.7654 139.4336 0.0024 0.0060 13.8652 4.6420 Yes 142 0.5429 0.3701 *** ** * * ** *** *** TABLE E7 (cont’d) Panel C: Captive Reinsurance and Analyst Environment Panel C of Table E7 reports the results from the estimation of equation (3) using a first-differences regression. The dependent variable is Analysts. Column (1) reports the results for life insurers, and Column (2) reports the results for P/C insurers. Robust standard errors are clustered by firm. All variables are defined in Appendix C, and continuous variables are winsorized at the 1 st and 99th percentiles. Significance at the .10, .05 and .01 level for two-sided tests is denoted by *, ** and ***, respectively. DV = Δ Analysts Variable Predict. Δ C_Number Δ Captive Δ Enforcement Δ Log (MV) Δ B/M Δ Leverage Δ ROA Δ Log (Turnover) Δ Log (SD_Return) Δ Log (ABN_Return) Δ InstOwn% Constant Year & Regulator FE Observations Within R-squared Between R-squared (1) "Life" (2) "P/C" Coeff SE -0.8925 0.8378 3.0373 0.5607 0.8129 0.7525 -2.9094 2.5776 -0.0001 0.0001 18.0534 10.9150 -3.6467 15.6578 0.2974 0.4297 0.6895 0.8590 -457.7250 124.6623 0.0020 0.0032 51.0619 34.5642 Coeff SE omitted omitted -1.0306 1.6010 -0.0782 1.5251 0.0000 0.0000 1.3637 5.7079 -7.7302 9.1039 0.7765 0.8735 -0.8043 0.6978 -340.0047 214.4161 -0.0054 0.0283 1.7952 14.8770 Yes 105 0.5068 0.2349 *** *** Yes 77 0.5921 0.0041 91 TABLE E8 Alternative Specifications Panel A: Alternative Specifications of Captive Reinsurance and Firm Type Panel A of Table E8 reports the results from the estimation of equation (1) using two measures of captive reinsurance. The dependent variable is Captive in Columns (1) and (3) and C_Foreign in Columns (2) and (4). The results in Columns (1) and (3) are based on the probit estimation and in Columns (2) and (4) on ordered logit model. All variables are defined in Appendix C, and continuous variables are winsorized at the 1 st and 99th percentiles. Significance at the .10, .05 and .01 level for two-sided tests is denoted by *, ** and ***, respectively. Life Insurers (1) DV = Captive Variable Enforcement Surplus Constraint (SC) SC * Enforcement Assets / Surplus (AS) AS * Enforcement Premium / Surplus (PS) PS * Enforcement Log (Premiums) Reinsurance Inefficiency Investment Inefficiency Tax Rate Reinsurance ROA Investment Yield Cash / Total Assets Debt / Total Assets Constant / Cut 1 Year / Regulator FE Observations Adjusted R-squared Log pseudolikelihood Predict. H1 (-) H2 (-) Coeff -3.4359 -0.0685 6.3104 0.0482 0.4447 0.9409 0.1580 -1.6422 0.0115 8.7310 9.0657 -2.8133 -54.7024 -20.7379 -8.8447 SE 0.9713 1.6766 3.1167 0.0527 0.1163 0.2971 0.0823 0.4601 0.0133 1.9124 19.6510 3.3106 18.3284 11.4395 2.3092 Yes / No 264 0.7763 -40.3855 P/C Insurers (2) DV = C_Foreign *** ** *** *** * *** *** *** * *** Coeff -4.0365 1.9846 -2.2759 -0.2362 0.5535 0.5042 -0.0276 -13.7305 0.0096 3.0960 -9.2051 -18.0431 -15.5656 -37.9844 -9.7394 SE 3.0161 0.9473 4.3946 0.1686 0.3030 0.9838 0.2109 10.9989 0.0113 2.2248 33.5872 5.7293 20.3303 26.2416 13.5607 Yes / Yes 264 0.6417 -92.0885 92 (3) DV = Captive ** Coeff 0.1388 1.4647 -11.6679 SE 0.5652 1.2298 1.9925 -1.0954 0.2092 0.1700 5.9649 0.9114 -0.0606 0.0068 1.2369 1.5035 4.9073 0.2737 -1.0206 0.4830 0.5712 0.2819 2.3411 3.1502 0.0416 0.7857 3.2266 2.1760 2.6396 1.9579 2.2288 (4) DV = C_Foreign *** Coeff 0.5812 -8.4089 1.0411 SE 1.6522 7.8921 6.8576 -1.1591 0.2678 1.2489 156.6384 -8.1286 -0.1691 2.5801 1.2885 0.2787 -40.3431 -4.7036 4.0139 1.6895 2.6649 0.9274 7.4317 4.2121 0.1717 2.6602 5.3885 10.1664 52.5041 6.6365 6.3127 * *** Yes / No 260 0.4070 -103.5254 ** *** * Yes / Yes 260 0.7210 -64.2998 *** ** TABLE E8 (cont’d) Panel B: Alternative Regulatory Enforcement Specification and Captive Reinsurers Panel B of Table E8 reports the results from the estimation of equation (1) using a Tobit model. The dependent variable is C_Number. Columns (1) and (2) report the results for pure life insurers and diversified insurers (P/C + L/H). Columns (3) and (4) report the results for pure property-casualty (P/C) insurers. Robust standard errors are clustered by firm. All variables are defined in Appendix C, and continuous variables are winsorized at the 1st and 99th percentiles. Significance at the .10, .05 and .01 level for two-sided tests is denoted by *, ** and ***, respectively. DV = C_Number (1) "Life" [2006-2011] Variable Predict. Enforcement2 H1 (-) Surplus Constraint (SC) SC * Enforcement2 H2 (-) Assets / Surplus (AS) AS * Enforcement2 Premium / Surplus (PS) PS * Enforcement2 Log (Premiums) Reinsurance Inefficiency Investment Inefficiency Tax Rate Reinsurance ROA Investment Yield Cash / Total Assets Debt / Total Assets Constant Year & Regulator FE Observations Adjusted R-squared Log pseudolikelihood Left / Right - censored Coeff -4.8198 -2.2916 -2.1059 0.0624 0.2332 3.6233 -0.4706 -2.3544 0.0971 3.2201 -11.4019 0.4507 15.4148 -11.4138 -24.1688 SE 0.6238 0.7564 0.5713 0.0564 0.0609 0.5357 0.4553 0.8948 0.0323 1.1024 15.3859 0.4856 5.4350 12.4097 3.8420 Yes 151 0.8070 -48.6086 90 / 61 (2) "Life" [2012-2015] *** *** *** *** *** *** *** *** *** *** Coeff -2.7245 -2.6533 -1.6016 0.4979 0.1668 0.9379 0.3951 -39.2007 -0.0049 -1.3729 20.6999 -27.0875 70.3334 34.8813 -7.2958 SE 1.1006 0.8450 1.5218 0.1421 0.1113 0.5302 0.2233 14.8645 0.0124 1.1559 20.7852 11.9162 23.6617 25.6321 3.7547 Yes 113 0.5324 -98.6094 60 / 53 93 (3) "P/C" [2006-2011] ** *** Coeff 1.2040 1.6876 -3.7207 SE 0.0111 0.0207 0.0318 -2.0390 -1.4262 1.2727 0.0221 0.5217 0.2241 4.0860 -4.2287 0.1079 -4.2580 -2.6206 -5.1305 0.0209 0.0121 0.0035 0.0079 0.4179 0.0087 0.0711 0.1734 0.2142 0.4231 0.3249 0.0270 (4) "P/C" [2012-2015] *** *** *** Coeff 1.0178 6.4718 -1.4349 SE 0.0093 0.0115 0.0265 *** *** *** -2.9299 -2.0845 0.0390 1.2221 1.6593 -0.1120 0.3187 2.8708 -5.4674 3.2041 -3.8477 3.2545 0.0119 0.0115 0.0029 0.0173 0.1302 0.0004 0.0576 0.2760 0.2280 0.0519 0.1547 0.0225 *** *** *** *** *** *** *** *** *** *** *** *** *** * * *** ** *** * Yes 155 0.7705 -37.5601 95 / 60 *** *** *** *** *** *** *** *** *** *** Yes 105 0.9305 -7.9534 62 / 43 TABLE E9 Quantile Tobit Model: Regulatory Enforcement and Captive Reinsurers Panel A: Life Insurers Panel A of Table E9 reports the results from the estimation of equation (1) using a quantile Tobit estimation in the sub-sample of life insurers between 2006 and 2011. The dependent variable is C_Number. Column (1) reports the results at the 25th quantile for life insurers. Column (2) reports the results at the median, and Column (3) reports the results at the 75th quantile. All variables are defined in Appendix C, and continuous variables are winsorized at the 1 st and 99th percentiles. Significance at the .10, .05 and .01 level for two-sided tests is denoted by *, ** and ***, respectively. DV = C_Number (2) 50% [2006 - 2011] (1) 25% [2006 - 2011] Variable Predict. Enforcement H1 (-) Surplus Constraint (SC) SC * Enforcement H2 (-) Assets / Surplus (AS) AS * Enforcement Log (Premiums) Reinsurance Inefficiency Investment Inefficiency Tax Rate Reinsurance ROA Investment Yield Cash / Total Assets Debt / Total Assets Constant Year & Regulator FE Observations Adjusted R-squared Objective function Coeff -3.8347 -2.2811 -3.8808 0.1472 0.2897 0.6582 -0.0790 -0.2500 0.0006 -0.0015 -1.6374 0.0301 4.1127 13.8701 -10.3615 SE 1.9560 1.1640 1.8935 0.0950 0.1761 0.3819 0.1125 0.0944 0.0236 1.5428 12.5327 0.2397 2.7581 15.9685 5.1408 Coeff -2.7498 -2.3147 -3.1696 0.1184 0.2141 0.5936 -0.0596 -0.2809 -0.0098 0.4492 1.6597 0.0069 -1.8022 11.5241 -7.0076 ** ** ** * * *** ** Yes 151 0.9364 0.1576 SE 0.8150 0.2309 0.5856 0.0273 0.0524 0.1526 0.0227 0.0437 0.0206 0.7716 11.0331 0.6330 1.2527 8.7377 2.0465 Yes 151 0.9195 0.2099 94 (3) 75% [2006 - 2011] *** *** *** *** *** *** *** *** *** Coeff -2.6344 -1.6208 -2.3921 0.1012 0.1851 0.6312 -0.0677 -0.1055 -0.0221 -0.7831 -6.7699 0.1597 -0.2108 13.2491 -7.2116 SE 1.4557 1.3793 1.7252 0.0692 0.1390 0.1196 0.0954 0.6809 0.1765 0.4427 8.2089 0.9480 2.5124 31.6579 4.1342 Yes 151 0.9266 0.1645 * *** * * TABLE E9 (cont’d) DV = C_Number (2) 50% [2012 - 2015] (1) 25% [2012 - 2015] Variable Predict. Enforcement H1 (-) Surplus Constraint (SC) SC * Enforcement H2 (-) Assets / Surplus (AS) AS * Enforcement Log (Premiums) Reinsurance Inefficiency Investment Inefficiency Tax Rate Reinsurance ROA Investment Yield Cash / Total Assets Debt / Total Assets Constant Year & Regulator FE Observations Adjusted R-squared Objective function Coeff -0.9095 2.7870 7.2278 0.0846 0.1330 0.4410 0.3131 -15.2705 -0.0053 0.4908 -16.9798 -0.7463 4.0484 13.7834 -2.1207 SE 0.8497 1.3607 2.9542 0.0194 0.0662 0.3562 0.1998 8.5286 0.0050 0.9068 20.7359 2.1772 12.3232 10.7662 3.8007 Coeff SE -4.8359 1.6117 -1.5561 1.6294 -2.7153 2.4982 0.2116 0.0656 0.3609 0.2417 0.6870 2.5513 0.0608 1.9231 -15.7173 183.2298 -0.0042 0.0158 -0.5371 2.0024 -7.7942 72.0825 -0.2820 4.0575 12.0806 41.3604 38.3405 82.2298 -10.9878 22.1660 ** ** *** ** * Yes 113 0.7135 0.2196 Yes 113 0.8898 0.3084 95 (3) 75% [2012 - 2015] *** *** Coeff -4.5410 -1.3947 -3.1564 0.2199 0.3647 0.5194 0.3274 -17.9916 0.0004 -0.6976 -10.5253 -0.6451 6.0700 41.2770 -9.5967 SE 0.2959 0.2098 0.3352 0.0130 0.0245 0.1055 0.0883 4.3700 0.0035 0.4755 11.2610 0.6334 13.4801 9.9198 0.9082 Yes 113 0.8807 0.2265 *** *** *** *** *** *** *** *** *** *** TABLE E9 (cont’d) Panel B: P/C Insurers Panel B of Table E9 reports the results from the estimation of equation (1) using a quantile Tobit estimation in the sub-sample of P/C insurers between 2006 and 2011. The dependent variable is C_Number. Column (1) reports the results at the 25th quantile for P/C insurers. Column (2) reports the results at the median, and Column (3) reports the results at the 75th quantile. All variables are defined in Appendix C, and continuous variables are winsorized at the 1 st and 99th percentiles. Significance at the .10, .05 and .01 level for two-sided tests is denoted by *, ** and ***, respectively. DV = C_Number (2) 50% [2006 - 2011] (1) 25% [2006 - 2011] Variable Predict. Enforcement H1 (-) Surplus Constraint (SC) SC * Enforcement H2 (-) Premium / Surplus (PS) PS * Enforcement Log (Premiums) Reinsurance Inefficiency Investment Inefficiency Tax Rate Reinsurance ROA Investment Yield Cash / Total Assets Debt / Total Assets Constant Year & Regulator FE Observations Adjusted R-squared Objective function Coeff 0.3334 0.4013 -1.0063 -0.0064 -0.0177 0.0566 0.0069 0.0954 -0.0188 0.4099 0.0867 -0.0393 -0.0316 -0.2959 0.7756 SE 0.0931 0.1013 0.3647 0.0673 0.1973 0.1226 0.0163 0.3788 0.0350 0.5956 0.4058 0.1280 0.5477 0.4862 0.8691 Coeff 0.3392 0.4472 -0.7333 -0.1129 -0.3300 0.3476 0.0173 -0.6676 0.0159 1.4159 -0.3566 -0.3419 1.7099 -0.2776 -1.0693 *** *** *** Yes 155 0.3326 0.0721 Yes 155 0.5488 0.1147 96 SE 0.2606 0.1437 0.2933 0.0952 0.2792 0.4301 0.0244 1.0878 0.0476 1.1057 4.0920 0.6744 5.9761 1.1961 2.4515 (3) 75% [2006 - 2011] *** *** Coeff 0.1200 0.3895 -0.8489 -0.0404 -0.1173 0.2717 0.0411 -1.1087 -0.0465 0.8370 0.5974 -0.7662 5.2635 -0.9815 0.3013 Yes 155 0.5297 0.0934 SE 0.2596 0.1249 0.3304 0.1037 0.3043 0.2321 0.1076 1.7503 0.0472 1.0314 1.0773 0.4927 1.3814 0.9590 1.4069 *** *** *** TABLE E10 IV, SURE, and Heckman Estimation: Regulatory Enforcement and Captive Reinsurers Panel A: Life Insurers Panel A of Table E10 reports the results from the estimation of equation (1) using an instrumental variable (IV), seemingly unrelated regression (SURE), and Heckman estimation. The sample includes life insurers between 2006 and 2015. The dependent variable is C_Number. Column (1) reports the results using IV regression, where Enforcement is instrumented with market power (Power) and Strict Regulators. All variables are defined in Appendix C, and continuous variables are winsorized at the 1st and 99th percentiles. Significance at the .10, .05 and .01 level for two-sided tests is denoted by *, ** and ***, respectively. (1) IV Predict. Variable Enforcement H1 (-) Surplus Constraint (SC) SC * Enforcement H2 (-) Assets / Surplus (AS) AS * Enforcement Log (Premiums) Reinsurance Inefficiency Investment Inefficiency Tax Rate Reinsurance ROA Investment Yield Cash / Total Assets Debt / Total Assets Constant Year & Regulator FE Observations Adjusted R-squared Chi-squared Underidentification test (KP) Weak identification (F stat) Coeff -3.7155 -1.9046 -2.1470 0.1544 0.2626 0.7872 -0.0683 -0.2318 -0.0056 -0.8061 -16.6004 0.3979 1.2807 28.2645 -5.8780 SE 0.8332 0.6353 0.5281 0.0331 0.0328 0.1903 0.0280 0.0886 0.0046 0.5009 12.5062 0.5030 8.7333 10.5686 1.6350 *** *** *** *** *** *** ** *** *** *** DV = C_Number (2) SURE SE Coeff 0.2595 -3.7233 0.3349 -1.9245 0.5746 -2.2263 0.0145 0.1545 0.0211 0.2670 0.1040 0.7811 0.0319 -0.0718 0.2695 -0.2345 0.0065 -0.0053 0.3812 -0.7800 8.2953 -16.2970 0.5980 0.4157 3.8725 1.4434 4.5229 27.8118 0.9537 -7.3717 Yes 264 0.8863 2048.96 Yes 264 0.8863 0.0040 30.5090 97 *** *** *** *** *** *** ** ** ** *** *** (3) Heckman SE Coeff 0.2597 -3.8068 0.3351 -1.9443 0.5750 -2.2326 0.0145 0.1562 0.0211 0.2682 0.1040 0.7980 0.0319 -0.0728 0.2697 -0.2286 0.0065 -0.0055 0.3814 -0.7891 8.2967 -15.7086 0.5980 0.4285 3.8727 1.4602 4.5235 28.2390 0.9539 -7.5673 Yes 264 2059.63 *** *** *** *** *** *** ** ** * *** *** TABLE E10 (cont’d) Panel B: P/C Insurers Panel B of Table E10 reports the results from the estimation of equation (1) using an instrumental variable (IV), seemingly unrelated regression (SURE), and Heckman estimation. The sample includes pure P/C insurers between 2006 and 2015. The dependent variable is C_Number. Column (1) reports the results using IV regression, where Enforcement is instrumented with market power (Power) and Strict Regulators. All variables are defined in Appendix C, and continuous variables are winsorized at the 1 st and 99th percentiles. Significance at the .10, .05 and .01 level for two-sided tests is denoted by *, ** and ***, respectively. DV = C_Number (2) SURE Coeff SE 0.1517 0.1281 0.0075 0.2085 -0.1473 0.3879 -0.0800 0.0441 -0.2346 0.1295 0.2536 0.0564 0.1944 0.0677 -0.6643 0.6363 -0.0045 0.0159 0.4851 0.1777 1.7845 0.8716 -0.0510 0.6303 1.3335 1.0425 -0.7929 0.4933 0.0832 0.4057 (1) IV Variable Predict. Enforcement H1 (-) Surplus Constraint (SC) SC * Enforcement H2 (-) Premium / Surplus (PS) PS * Enforcement Log (Premiums) Reinsurance Inefficiency Investment Inefficiency Tax Rate Reinsurance ROA Investment Yield Cash / Total Assets Debt / Total Assets Constant Year & Regulator FE Observations Adjusted R-squared Chi-squared Underidentification test (KP) Weak identification test (F stat) Coeff 0.7016 0.1597 -0.3788 -0.1356 -0.3973 0.3226 -0.2714 -1.3467 0.0134 0.2519 0.0463 -0.2635 2.0611 0.8721 0.1157 SE 0.2647 0.0977 0.2940 0.0795 0.2332 0.1624 2.7409 0.8100 0.0070 0.2198 0.6314 0.6540 1.1746 0.9490 1.1429 *** * * * ** * * * Yes 260 0.7613 Yes 260 0.6079 402.78 0.0308 63.2730 98 * * *** *** *** ** (3) Heckman Coeff SE 0.1254 0.1281 0.0045 0.2085 -0.1518 0.3879 -0.0790 0.0441 -0.2316 0.1295 0.2568 0.0564 0.1945 0.0678 -0.6561 0.6364 -0.0043 0.0159 0.4845 0.1778 1.7578 0.8717 -0.0409 0.6303 1.3326 1.0425 -0.8278 0.4933 0.0531 0.4057 Yes 260 403.29 * * *** *** *** ** * TABLE E11 Falsification Test: Regulatory Enforcement and Capital Gains Table E11 reports the results from the estimation of equation (4) which examines the effect of regulatory enforcement on the realization of gains or losses (RealGAIN). EBGAINS masures earnings before capital gains (losses) scaled by total assets. Controls include Capital (i.e., surplus before capital gains (losses) scaled by total assets), UnrealGAIN (i.e., unrealized capital gains (losses) scaled by total assets), IA (i.e., invested assets scaled by total assets), InvInc (i.e., investment income scaled by total assets), and Liq (i.e., reserves scaled by total assets (or surplus)). Column (1) reports the results using panel OLS estimation while Column (2) reports the results using instrumental variable (IV) regression. Enforcement is instrumented with market power (Power). Robust standard errors are clustered by firm. All variables are defined in Appendix C, and continuous variables are winsorized at the 1 st and 99th percentiles. Significance at the .10, .05 and .01 level for two-sided tests is denoted by *, ** and ***, respectively. DV = RealGAIN (1) OLS Variables Enforcement EBGAINS EBGAINS* Enforcement Controls & Interactions Year & Regulator FE Observations Adjusted R-squared Coeff. 0.0001 0.0022 -0.0011 Yes Yes 461 0.9925 Underidentification (KP LM) Weak identification (KP Wald F) 99 (2) IV SE 0.0009 0.0027 0.0038 Coeff. 0.0007 0.0017 -0.0021 SE 0.0012 0.0021 0.0036 Yes Yes 461 0.9925 23.227 (p-value = 0.0000) 29.564 TABLE E12 PSM: Regulatory Enforcement, Captive Reinsurers, and Credit Ratings Table E12 reports the results from the estimation of equation (2) using a panel OLS regression. In Column (1) and (3), the regression is adjusted for propensity scores for Captive. In Column (2) and (4), I use propensity score weighting. The dependent variable is CR_All. Control sample is matched based on size. Robust standard errors are clustered by firm. All variables are defined in Appendix C, and continuous variables are winsorized at the 1st and 99th percentiles. Significance at the .10, .05 and .01 level for two-sided tests is denoted by *, ** and ***, respectively. DV = CR_All Variable Predict. C_Number Enforcement H3 (+) Reserves/Surplus (RS) Reserves/ Surplus Ratio (RSR) Premium / Surplus Ratio Log (Premiums) Reinsurance Inefficiency Investment Inefficiency Reinsurance ROA Cash / Total Assets Debt / Total Assets Retained Earnings / Total Assets Constant Year & Regulator FE Observations Adjusted R-squared (1) "Life" [Adjusted] Coeff SE 0.6478 0.1868 2.7517 0.9352 0.4015 0.0652 3.1377 1.8795 2.1064 1.6505 -3.6427 0.6372 0.2415 0.1219 -2.9749 0.9185 -5.1411 1.9806 33.9343 27.7527 3.2134 14.7452 50.8120 19.2644 28.4362 10.9813 -12.1178 5.7375 Yes 255 0.7829 *** *** *** *** * *** *** *** ** ** (2) "Life" [Weighting] Coeff SE 0.5324 0.1901 3.4218 0.8496 0.4478 0.0936 1.2288 1.7077 -3.3802 1.7190 0.8033 0.6180 0.3503 0.1588 -2.3202 1.3058 -6.0418 2.5414 16.9638 32.9194 -0.3010 19.6143 25.0718 29.6709 21.9404 10.1358 -26.9776 4.1114 Yes 255 0.7795 100 *** *** *** * ** * ** ** *** (3) "P/C" [Adjusted] Coeff SE 0.0969 1.0456 -0.7959 0.6717 0.0016 0.0018 -1.5799 1.7257 -0.0189 1.7912 2.7445 0.9950 *** 0.2395 0.3983 -10.2996 7.3634 3.0731 2.5168 -8.8666 6.7799 52.9629 21.5246 ** 13.3737 10.5202 12.5827 3.8595 *** -31.9421 5.7574 *** Yes 247 0.7909 (4) "P/C" [Weighting] Coeff SE 0.1504 0.9096 -0.9153 0.8075 0.0011 0.0013 -2.7788 1.3950 0.0721 1.8187 3.2384 0.9664 0.2972 0.4725 -13.9150 7.4249 0.2079 4.7327 -0.1880 6.2169 53.0277 17.9105 13.7437 11.0180 10.8295 4.9601 -32.2607 6.3616 Yes 247 0.7926 * *** * *** ** *** TABLE E13 Instrumental Variable Regression: Regulatory Enforcement, Captive Reinsurers, and Credit Ratings Table E13 reports the results from the estimation of equation (2) using instrumental variable (IV) regression, where Enforcement is instrumented with market power (Power) and Strict Regulators. The dependent variable is CR_All in Column (1) and (3) and CR_Rated in Column (2) and (4). Robust standard errors are clustered by firm. All variables are defined in Appendix C, and continuous variables are winsorized at the 1st and 99th percentiles. Significance at the .10, .05 and .01 level for two-sided tests is denoted by *, ** and ***, respectively. DV = Credit Ratings Variable Predict. C_Number Enforcement H3 (+) Reserves/Surplus (RS) Reserves/ Surplus Ratio (RSR) Premium / Surplus Ratio Log (Premiums) Reinsurance Inefficiency Investment Inefficiency Reinsurance ROA Cash / Total Assets Debt / Total Assets Retained Earnings / Total Assets Constant Year & Regulator FE Observations Adjusted R-squared Underidentification test (KP) Weak identification (F stat) (1) DV = CR_All ["Life"] Coeff SE 1.7019 0.3793 16.4203 3.5108 0.4200 0.1135 -0.4122 2.3655 -18.7032 4.8936 -2.9271 1.0942 0.6453 0.1923 -2.2883 0.8438 -12.6133 2.7392 14.8848 45.3207 59.6843 25.6620 -54.8030 34.3091 4.5932 8.8159 13.7892 9.0664 Yes 255 0.2483 p-val = 0.0000 60.0190 *** *** *** *** *** *** *** *** ** (2) DV = CR_Rated ["Life"] Coeff SE 0.0405 0.1291 1.8327 0.9321 0.1097 0.0427 -0.4570 0.5433 -2.3161 0.8855 0.6328 0.2864 0.1222 0.0482 -4.5265 2.8671 -2.3230 0.9114 -14.2759 15.8298 2.0285 7.1458 -34.5116 9.4029 11.0854 3.5173 -9.7590 2.2281 Yes 212 0.6740 p-val = 0.0131 17.0280 101 ** *** *** ** *** *** *** *** *** (3) DV = CR_All ["P/C"] Coeff SE 1.5488 0.8987 11.6978 5.4358 0.0033 0.0019 -0.2052 1.2320 -2.8084 2.7192 0.4282 1.5879 0.9442 0.8403 -18.4376 7.0603 1.0580 2.9796 -6.5919 15.3843 61.9057 14.0496 38.7784 15.5235 21.4645 6.2581 -15.4665 11.0793 Yes 247 0.5774 p-val = 0.0027 6.6250 * ** * *** *** *** *** (4) DV = CR_Rated ["P/C"] Coeff SE 0.4533 0.7740 1.1771 3.0506 0.0000 0.0006 -0.7089 0.4582 -0.1143 0.7871 1.5096 1.0842 0.1252 0.1040 -9.1092 4.2602 0.3163 1.8260 -1.8624 5.2712 8.8084 8.8517 11.1915 10.9955 13.8107 2.8946 -21.9449 6.9675 Yes 167 0.7505 p-val = 0.0043 8.5440 ** *** *** TABLE E14 PSM: Regulatory Enforcement, Captive Reinsurers, and Information Asymmetry Table E14 reports the results from the estimation of equation (3) using a first-differences regression, which is adjusted for propensity scores for Captive. The dependent variable is the change in Illiquidity. Column (1) reports the results for life insurers and Column (2) reports the results for P/C insurers between 2006 and 2015. Robust standard errors are clustered by firm. All variables are defined in Appendix C, and continuous variables are winsorized at the 1st and 99th percentiles. Significance at the .10, .05 and .01 level for two-sided tests is denoted by *, ** and ***, respectively. DV = Δ Illiquidity (1) "Life" Variable Predict. Δ Enforcement H4 (-) Δ C_Number H5 (+) Δ Log (MV) Δ B/M Δ Leverage Δ ROA Δ Log (Turnover) Δ Log (SD_Return) Δ Log (ABN_Return) Δ InstOwn% Constant Year & Regulator FE Observations Within R-squared Between R-squared Coeff -0.0561 -0.0332 -0.2870 0.0000 -2.2848 0.4206 -0.0945 0.1410 6.4384 0.0019 4.7757 (2) "P/C" SE 0.0281 0.0413 0.1104 0.0000 0.9698 0.5693 0.0470 0.0941 16.8204 0.0003 1.9810 Yes 130 0.4989 0.5491 * ** ** * *** ** Coeff SE -0.1746 0.4442 -0.6596 0.2079 -1.0118 0.3437 0.0000 0.0000 -1.3481 1.1678 -4.6626 2.4442 0.1543 0.2215 -0.1710 0.2613 -384.1571 146.3586 0.0086 0.0065 13.2268 3.9571 Yes 142 0.5631 0.3568 102 *** *** * * *** *** REFERENCES 103 REFERENCES Acemoglu, D., S. Johnson, and J. Robinson. 2005. Institutions as a Fundamental Cause of Longrun Growth. Handbook of Economic Growth, Volume A. Elsevier B.V. Acharya, V.V, P. Schnabl, and G. Suarez. 2013. Securitization without Risk Transfer. Journal of Financial Economics, 107(3): 515-536. Adiel, R.1996. Reinsurance and the Management of Regulatory Ratios and Taxes in the Property-Casualty Insurance Industry. Journal of Accounting and Economics, 22: 207–240. Agarwal, S., D. Lucca, A. Seru and F. Trebbi. 2014. Inconsistent Regulators: Evidence from Banking. The Quarterly Journal of Economics, 129(2): 889–938. Alt, J., E. Bueno de Mesquite, and S. Rose. 2011. Disentangling Accountability and Competence in Elections: Evidence from US Term Limits. Journal of Politics, 73: 171 – 186. Altamuro, J. 2006. The Determinants of Synthetic Lease Financing and the Impact on the Cost of Future Debt. Working paper. The Ohio State University. Amihud, Y. 2002. Illiquidity and Stock Returns: Cross-Section and Time-Series Effects. Journal of Financial Markets, 5: 31-56. Amemiya, T. 1973. Regression Analysis when the Dependent Variable is Truncated Normal. Econometrica, 41 (6): 997-1016. American Academy of Actuaries. 2014. Regulatory Capital Requirements for U.S. Life Insurers. Indusrance Industry Work Group. Available from https://www.actuary.org/files/Regulatory_Capital_Requirements_US_Life_Insurers_6-17-14.pdf (last accessed May 9, 2017). Ashbaugh, H. and M. Pincus. 2001. Domestic Accounting Standards, International Accounting Standards, and the Predictability of Earnings. Journal of Accounting Reseach, 39: 417 – 434. Baldwin R, and J. Black. 2008. Really Responsive Regulation. The Modern Law Review, 71: 5994. Ball, R. 2001. Infrastructure Requirements for an Economically Efficient System of Public Financial Reporting and Disclosure. Brookings – Wharton Papers on Financial Services: 127 – 169. Barth, M. E., W. H. Beaver, and W. H. Landsman. 2001. The Relevance of the Value Relevance Literature for Financial Accounting Standard Setting: Another View. Journal of Accounting and Economics, 31(13):77 – 104. 104 Basu, K. and A. Dixit. 2014. Too Small to Regulate. World Bank Research Working Paper No. 6860. Beatty, A., P.G. Berger, and J. Magliolo. 1995. Motives for Forming Research & Development Financing Organizations. Journal of Accounting and Economics, 19: 411-442. Beatty, R.P., Bunsis, H., and J.R. Hand, 1998. The indirect economic penalties in SEC investigations of underwriters. Journal of Financial Economics, 50: 151-186. Beatty, A. and D. Harris. 1998. The Effects of Taxes, Agency Costs and Information Asymmetry on Earnings Management: A Comparison of Public and Private Firms. Review of Accounting Studies, 4(3–4): 299–326. Beatty, A and J. Weber. 2006. Accounting discretion in fair value estimates: An examination of SFAS 142 goodwill impairments. Journal of Accounting Research, 44(2): 257 - 288. Beck, P. J., J. S. Davis, and W. Jung. 2000. Taxpayer Disclosure and Penalty Laws. Journal of Public Economic Theory, 2 (2): 243–272. Beland, L. 2015. Political Parties and Labor-Market Outcomes: Evidence from US States. American Economic Journal: Applied Economics, 7: 198 – 220. Beneish, M., 1999. Incentives and Penalties Related to Earnings Overstatement That Violate GAAP. The Accounting Review, 74 (4): 425-457 Bens, D. A., M. Cheng, and M. Neamtiu. 2016. The Impact of SEC Disclosure Monitoring on the Uncertainty of Fair Value Estimates. The Accounting Review, 91 (2): 349–375. Bertomeu, J. and I. Marinovic. 2016. A Theory of Hard and Soft Information. The Accounting Review, 91(1): 1-20. Besley, T. and A. Case. 1995. Does Electroral Accountability Affect Economic Policy Choices? Evidence from Gubernatorial Term Limits. Quarterly Journal of Economics, 110: 769 – 798. Beyer, A., D.A. Cohen, T.Z. Lys, and B.R. Walther. 2010. The Financial Reporting Environment: Review of Recent Literature. Journal of Accounting and Economics, 50: 296 – 343. Bikker, J.A. and M. van Leuvensteijn. 2008. Competition and Efficiency in the Dutch Life Insurance Industry. Applied Economics, 40 (16): 2063 – 2084. Blacconiere, J. R. Frederickson, M. F. Johnson, and M. F. Lewis. 2011. Are voluntary disclosures that disavow the reliability of mandated fair value information informative or opportunistic? Journal of Accounting and Economics, 52(2 3): 235 - 251. Block, M.K. and J.S. Feinstein. 1986. The Spillover Effect of Antitrust Enforcement. The Review of Economics and Statistics, 68: 122 – 131. 105 Bonsall, S., K. Koharki, and M. Neamtiu. 2015 The Effectiveness of Credit Rating Agency Monitoring: Evidence from Asset Securitizations. The Accounting Review, 90(5): 1779-1810. Bozanic, Z. and P. Kraft. 2014. Qualitative Corporate Disclosure and Credit Analysts’ Soft Rating Adjustments. Working paper. Bozanic, Z., J. Hoopes, J. Thornock, and B. Williams. 2016. IRS Attention. Working paper. Bratten, B., L. M. Gaynor, L. McDaniel, N. R. Montague, and G. E. Sierra. 2013. The audit of fair values and other estimates: The effects of underlying environmental, task, and auditorspecific factors. Auditing: A Journal of Practice & Theory, 31: 127–146. Brochet, F., A. D. Jagolinzer, and E. J. Riedl. 2013. Mandatory IFRS Adoption and Financial Statement Comparability. Contemporary Accounting Research, 30 (4): 1373-1400. Brunnermeier, M.K., and L.H. Pedersen. 2009. Market Liquidity and Funding Liquidity. Review of Financial Studies, 22: 2201–2238. Buchinsky, M. 1998. Recent Advances in Quantile Regression Models: A Practical Guideline for Empirical Research. Journal of Human Resources, 33(1): 88 – 126. Bushee, B., I.D. Gow, and D. J. Taylor. 2016. Linguistic Complexity in Firm Disclosures: Obfuscation or Information? Working Paper. Bushman, R. and C. Williams. 2012. Accounting Discretion, Loan Loss Provisioning, and Discipline of Banks Risk-taking. Journal of Accounting and Economics, 54(1): 1-18. Callahan, C., C.M.C. Lee, and T. Yohn. 1997. Accounting Information and Bid-Ask Spreads. Accounting Horizons, 11: 50-60. Campbell, J., and P. E. Yeung. 2016. Earnings Comparability, Accounting Similarities, and Stock Returns: Evidence from Peer Firms’ Earnings Restatements. Working paper, University of Georgia and Cornell University. Carrigan and Harrington. 2015. Choice in Regulatory Program Design and Enforcement. Research paper for the Penn Program on Regulation’s Best-in-Class Regulator Initiative. CAS Institute. 2006. Considerations Regarding Materiality and Range of Reserves in Connection with Actuarial Standard of Practice #36. Available from https://www.casact.org/pubs/forum/02fforum/02ff259.pdf (last accessed May 9, 2017). Cassell, C. A., L. M. Dreher, and L. A. Myers. 2013. Reviewing the SEC’s Review Process: 10K Comment Letters and the Cost of Remediation. The Accounting Review, 88 (6): 1875–1908 Chen, L., D.A. Lesmond, and J. Wei. 2011. Corporate Yield Spreads and Bond Liquidity. Journal of Finance, 62 (1): 119 – 149. 106 Cheng, M. and K.R. Subramanyam. 2008. Analyst Following and Credit Ratings. Contemporary Accounting Research, 25 (4) (Winter 2008): 1007 – 1043. Coglianese, C. and R. A. Kagan. 2007. Introduction. Regulation and Regulatory Processes. Ashgate Publishing, Burlington, VT. Cornaggia J., K. J. Cornaggia and H. Xia. 2016. Revolving Doors on Wall Street. Journal of Financial Economics, 120(2): 400 – 419. Costello, A., J. Granja and J. Weber. 2016. Do Strict Regulators Increase the Transparency of the Banking System? Working paper. Massachusetts Institute of Technology. Daske, H., L. Hail, C. Leuz, and R. Verdi. 2008. Mandatory IFRS Reporting Around the World: Early Evidence on the Economic Consequences. Journal of Accounting Research, 46: 10851142. Dechow, P.M., R. Sloan and A. Sweeney, 1996. Causes and Consequences of Earnings Manipulation: An Analysis of Firms Subject to Enforcement Actions by the SEC. Contemporary Accounting Research, 13 (1): 1-36. Dechow, P.M. and C. Shakespeare. 2009. Do Managers Time Securitization Transactions to Obtain Accounting Benefits? The Accounting Review, 84 (1): 99-132. Dechow, P.M., L. A. Myers, and C. Shakespeare. 2010. Fair value accounting and gains from asset securitizations: A convenient earnings management tool with compensation side benefits. Journal of Accounting and Economics, 49(1 2): 2 - 25. Dechow, P.M., A. Lawrence, and J. Ryans. 2016. SEC Comment Letters and Insider Sales. The Accounting Review, 91 (2): 401–439. De Franco, G., S. P. Kothari, and R. S. Verdi. 2011. The Benefits of Financial Statement Comparability. Journal of Accounting Research, 49 (4): 895-931. deHaan, E., S. Kedia, K. Koh, and S. Rajgopal. 2015. The Revolving Door and the SEC’s Enforcement Outcomes: Initial Evidence from Civil Litigation. Journal of Accounting and Economics, 60 (2-3): 65-96. Djankov, S., La Porta, R, Lopez-de-Silanes, F., and A. Shleifer. 2008. The Law and Economics of Self-dealing. Journal of Financial Economics, 88: 430-465. Government Publishing Office. 2010. Dodd-Frank Wall Street Reform and Consumer Protection Act. Public Law 111 – 203 – July 21, 2010. Available from https://www.gpo.gov/fdsys/pkg/PLAW-111publ203/pdf/PLAW-111publ203.pdf (last accessed June 19, 2017). Doherty, N. and L. L. Posey. 1997. Availability Crisis in Insurance Market: Optimal Contracts with Asymmetric Information and Capacity Constraints. Journal of Risk and Uncertainty, 15: 55 – 80. 107 Dyck. A., Volchkova, N. and L. Zingales. 2008. The Corporate Governance Role of the Media: Evidence from Russia. Journal of Finance, 63: 1093 – 1135. Edelman, L.B. and M.C. Suchman. 1997. The Legal Environments of Organizations. Annual Review of Sociology, 23: 479 – 515. El Ghoul, S., O. Guedhami and J. Pittman. 2011. The Role of IRS Monitoring in Equity Pricing in Public Firms. Contemporary Accounting Research, 28: 643–674. Ely, K. 1995. Operating Lease Accounting and the Market’s Assessment of Equity Risk. Journal of Accounting Research, 33 (Autumn): 435 – 458. Ettredge, M., K. Johnstone, M. Stone, and Q. Wang. 2011. The effects of Firm Size, Corporate Governance Quality, and Bad News on Disclosure Compliance. Review of Accounting Studies, 16 (4): 866–889. Fang, X., Y. Li, B. Xin, and W. Zhang. 2016. Financial Statement Comparability and Debt Contracting: Evidence from the Syndicated Loan Market. Accounting Horizons, 30 (2): 277˗303. Farber. 2005. Restoring Trust after Fraud: Does Corporate Governance Matter? The Accounting Review, 80 (2): 539–561. Federal Insurance Office. 2013a. How to Modernize and Improve the System of Insurance Regulation in the United States. Federal Insurance Office. 2013b. Annual Report on the Insurance Industry. Available from https://www.treasury.gov/initiatives/fio/reports-andnotices/Documents/FIO%20Annual%20Report%202013.pdf (last accessed June 19, 2017). Feng, M, J.D. Gramlich, and S. Gupta. 2009. Special Purpose Entities: Empirical Evidence on Determinants and Earnings Management. The Accounting Review, 84: 1833 – 76. Financial Stability Oversight Council (Office of Financial Research). 2016. Mind the Gaps: What do New Disclosures Tell Us About Life Insurers’ Use of Off-Balance-Sheet Captives? Fu, L. 2012. Optimal Growth for P&C Insurance Companies. Variance, 6 (1): 102 – 121. Gallemore, J. 2016. Does Bank Opacity Enable Regulatory Forbearance? Working paper, University of Chicago. Galloway. C. and J. Galloway. 1986. Handbook of Accounting for Insurance Companies. McGraw-Hill. Gallup. 2016. Democrats, Republicans Agree on Four Top Issues for Campaign. Available from http://www.gallup.com/poll/188918/democrats-republicans-agree-four-top-issues-campaign.aspx (last accessed November 14, 2016). 108 Gao, P. 2013. A Measurement Approach to Conservatism and Earnings Management. Journal of Accounting and Economics, 55(2-3): 251 – 268. Gatev, E. and P.E. Strahan. 2006. Banks’ Advantage in Hedging Liquidity Risk: Theory and Evidence from the Commercial Paper Market. Journal of Finance, 61: 867 – 892. Greenberg, J. 1984. Avoiding Tax Avoidance: A (Repeated) Game-Theoretic Approach. Journal of Economic Theory, 32(1): 1 – 13. Guedhami, O. and J. Pittman 2008. The Importance of IRS Monitoring to Debt Pricing in Private Firms. Journal of Financial Economics,90: 38-58. Hanlon M., J. L. Hoopes, and N. Shroff. 2014. The Effect of Tax Authority Monitoring and Enforcement on Financial Reporting Quality. The Journal of the American Taxation Association, 36 (2): 137-170. Harrington, W. 1988. Enforcement Leverage when Penalties are Restricted. Journal of Public Eoconomics, 37: 29 -53. Harrington, S. E. 2015. The Economics and Regulation of Captive Reinsurance in Life Insurance. Working paper. Healy, P.M., and K.G. Palepu. 2001. Information Asymmetry, Corporate Disclosure and the Capital Markets: A Review of the Empirical Disclosure Literature. Journal of Accounting and Economics, 31: 405 – 40. Heese, J., R. Krishnan, and F. Moers. 2016. Selective Regulator Decoupling and Organizations' Strategic Responses. Academy of Management Journal, 59(6): 2178 – 2204. Hepfer, B., J. Wilde, and R. Wilson. 2016. Taking Shadow Insurance Out of the Shadows: Regulatory Arbitrage, Taxes, and Capital. Working paper. Hood C. and H. Margetts. 2007. The Tools of Government in the Digital Age. Palgrave Macmillan, 2nd edition. Hoopes J.L., D. Mescall, and J. A. Pittman. 2012. Do IRS Audits Deter Corporate Tax Avoidance? The Accounting Review, 87(5): 1603-1639. Horton, J., G. Serafeim, and I. Serafeim. 2013. Does Mandatory IFRS Adoption Improve the Information Environment? Contemporary Accounting Research, 30 (1): 388-423. Hope, O. 2003. Disclosure Practices, Enforcement of Accounting Standards, and Analysts’ Forecast Accuracy: An International Study. Journal of Accounting Research, 41(2): 235 – 273. Huizinga, H., and L. Laeven. 2012. Bank Valuation and Accounting Discretion during a Financial Crisis. Journal of Financial Economics, 106 (3): 614-634. Hutter, B. M. 1989. Variations in Regulatory Enforcement Styles. Law & Policy, 11: 153–74. 109 Insurance Information Institute. 2015. Captives by State. Available from http://www.iii.org/publications/a-firm-foundation-how-insurance-supports-the-economy/a-50state-commitment/captives-by-state (last accessed November 3, 2016). Insurance Information Institute. 2017. Industry Overview. Available from http://www.iii.org/fact-statistic/industry-overview (last accessed June 19, 2017). Jaffee, D.W., and T. Russell. 1998. The Causes and Consequences of Rate Regulation in the Auto Insurance Industry. The Economics of Property-Casualty Insurance (David F. Bradford ed.). Chicago: University of Chicago Press. Johnston, R. and R. Petacchi. 2016. Regulatory Oversight of Financial Reporting: Securities and Exchange Commission Comment Letters. Contemporary Accounting Research, forthcoming. Kagan, R.A. 1984. Regulatory Enforcement. Handbook of Regulation and Administrative Law. Marcel Dekker, NY. Kagan, R. A. 1989. Editor’s Introduction: Understanding Regulatory Enforcement. Law & Policy, 11: 89–119. Karpoff, J.M, D.S. Lee, and G.S. Martin. 2008. The Consequences to Managers for Financial Misrepresentation. Journal of Financial Economics, 88(2): 193 – 215. Kedia, S. and S. Rajgopal. 2011. Do the SEC's Enforcement Preferences Affect Corporate Misconduct? Journal of Accounting and Economics, 51: 259-278. Kim, S., P. Kraft, and S. G. Ryan. 2013. Financial Statement Comparability and Credit Risk. Review of Accounting Studies, 18 (3): 783-823. King. G. and R. Nielsen. 2016. Why Propensity Scores Should Not Be Used for Matching. Working Paper. Klein, R. W. 2005. A Regulator’s Introduction to the Insurance Industry. Kansas City, MO: National Association of Insurance Commissioners, 2nd Edition. Koijen, R. and M. Yogo. 2015. The Cost of Financial Frictions for Life Insurers. American Economic Review, 105: 445–475. Koijen, R. and M. Yogo. 2016. Shadow Insurance. Econometrica, 84(3): 1265 – 1287. Kolev, K. 2013. Do Investors Perceive Marking-to-Model as Marking-to-Myth? Early Evidence from FAS 157 Disclosure. Working paper, Yale University. Kothari, S.P. 2000. The Role of Financial Reporting in Reducing Financial Risks in the Market. Building an Infrastructure for Financial Stability, Federal Reserve Bank of Boston Series 44: 89 – 102. 110 Kraft, P. 2015. Rating Agency Adjustments to GAAP Financial Statements and Their Effect on Ratings and Credit Spreads. The Accounting Review, 90(2): 641-674. Kubick, T. R., D. P. Lynch, M. A. Mayberry, and T. C. Omer. 2016. The Effects of Regulatory Scrutiny on Tax Avoidance: An Examination of SEC Comment Letters”. The Accounting Review, forthcoming. Kunreuther, H., M. Pauly, and S. McMorrow. 2013. Insurance and Behavioral Economics: Improving Decisions in the Most Misunderstood Industry. Cambridge University Press, NY. Landsberger, M. and I. Meilijson. 1982. Incentive Generating State Dependent Penalty System: The Case of Income Tax Evasion. Journal of Public Economics, 19 (3) (12): 333-52. Lang, M., K. Lins, and M. Maffett. 2012. Transparency, Liquidity and Valuation: International Evidence on When Transparency Matters Most. Journal of Accounting Research, 50: 729-774. La Porta, R. Lopez-de-Silanes, F., and A. Shleifer. 2006. What works in securities laws? Journal of Finance, 61: 1-32 Laux, V. and P. Stocken. 2016. Accounting Standards, Regulatory Enforcement, and Innovation. Working paper. Lehavy, R., F. Li, and K. Merkley. 2011. The Effect of Annual Report Readability on Analyst Following and the Properties of Their Earnings Forecasts. The Accounting Review, 86 (3): 1087– 1115. Leigh, A. 2008. Estimating the Impact of Gubernatorial Partisanship on Policy Setting and Economic Outcomes: A Regression Discontinuity Approach. European Journal of Political Economy, 24: 256 – 268. Leuz, C., and R. Verrecchia. 2000. The Economic Consequences of Increased Disclosure. Journal of Accounting Research, 38: 91-124. Leuz, C., and L. Hail. 2006. International Differences in the Cost of Equity Capital: Do Legal Institutions and Securities Regulation Matter. Journal of Accounting Research, 44 (3): 485 - 531. Leuz, C., and P. Wysocki. 2016. The Economics of Disclosure and Financial Reporting Regulation: Evidence and Suggestions for Future Research. Journal of Accounting Research, 54(2): 525 – 622. Lunawat, R., K. Pronin, J. Stecher, and G. Zhang. 2014. Aggressive Reporting under Imprecise Standards: A Laboratory Investigation. Working Paper. Lys, T., and L.G. Soo. 1995. Analysts' Forecast Precision as a Response to Competition. Journal of Accounting, Auditing, and Finance, 10: 751-765. Malik, A.S. 2014. The Desirability of Forgiveness in Regulatory Enforcement. Journal of Regulatory Economics, 46(1): 1 – 22. 111 Miller, G.S. 2006. The Press as a Watchdog for Accounting Fraud. Journal of Accounting Research, 44: 1001 – 1033. Mills, L., and R. Sansing. 2000. Strategic Tax and Financial Reporting Decisions: Theory and Evidence. Contemporary Accounting Research, 17 (1): 85-106. Mills, L., L. Robinson, and R. Sansing. 2010. FIN 48 and Tax Compliance. The Accounting Review, 85 (5): 1721–1742. Mishkin, F. 2000. Prudential Supervision: Why is it important and what are the issues? Working paper, NBER. Moody’s 2007. Financial Reporting and Credit Ratings (Presentation by Greg Jonas), CARE Conference, Napa California. Available from http://caremendoza.nd.edu/assets/152341/greg_jonas.pdf (last accessed May 27, 2017). NAIC. 2011. State Insurance Regulation. Available from http://www.naic.org/documents/topics_white_paper_hist_ins_reg.pdf (last accessed November 3, 2016). NAIC. 2012a. Insurance Group Supervision. Available from http://www.naic.org/cipr_newsletter_archive/vol3_ins_group_supervision.htm (last accessed November 14, 2016). NAIC. 2012b. Captives & SPV Use (E) Subgroup Call for Comments – Survey Results. Available from http://www.naic.org/documents/committees_e_cspv_sg_related_docs_survey_results.pdf (last accessed June 28, 2017). NAIC. 2013. Captives and Special Purpose Vehicles. Available from http://www.naic.org/store/free/SPV-OP-13-ELS.pdf (last accessed May 9, 2017). NAIC. 2015a. NAIC Signs MOU with Bermuda. Available from http://www.naic.org/Releases/2015_docs/naic_signs_mou_with_bermuda.htm (last accessed June 19, 2017). NAIC. 2015b. State Commissioners 2015. Available from http://www.naic.org/documents/members_state_commissioners_elected_appointed.pdf (last accessed November 3, 2016). NAIC. 2016. Insurance Regulatory Information System (IRIS) Ratios Manual. Available from http://www.naic.org/documents/prod_serv_fin_receivership_uir_zb.pdf (last accessed May 9, 2017). NAIC. 2017. Captive Insurance Companies. Available from http://www.naic.org/cipr_topics/topic_captives.htm (last accessed June 19, 2017). 112 Neel, M. 2017. Accounting comparability and economic outcomes of mandatory IFRS adoption. Contemporary Accounting Research, 34(1): 658 – 690. New York DFS 2013. Shining a Light on Shadow Insurance. Available from http://www.dfs.ny.gov/reportpub/shadow_insurance_report_2013.pdf (last accessed November 7, 2016). New York Times. 2011. Seeking Business, States Loosen Insurance Rules. Available from http://www.nytimes.com/2011/05/09/business/economy/09insure.html?pagewanted=all (last accessed November 3, 2016). North, D. 1990. Institutions, institutional change and economic performance. Cambridge University Press, Cambridge. North, D. 1992. Institutions, Ideology, and Economic Performance. Cato Journal, 11 (3): 477 – 496. Oded, S. 2013. Corporate Compliance: New Approaches to Regulatory Enforcement. Edward Elgar Publishing. Organization for Economic Cooperation and Development (OECD). 2014. OECD Best Practice Principles for Regulatory Policy: Regulatory Enforcement and Inspections. OECD Publishing. Peltzman, S. 1976. Toward a More General Theory of Regulation. Journal of Law and Economics, 19: 211–40. Peterson, K., R. Schmardebeck, and T. J. Wilks. 2015. The Earnings Quality and Information Processing Effects of Accounting Consistency. The Accounting Review, 90 (6): 2483-2514 Petroni, K. 1992. Optimistic Reporting in the Property-Casualty Insurance Industry. Journal of Accounting and Economics, 15: 485-508. Petroni, K. and D. Shackelford. 1995. Taxation, Regulation, and the Organizational Structure of Property-Casualty Insurers. Journal of Accounting and Economics, 20(3): 229-253. Richter, R. 2015. “The Role of Law in the New Institutional Economics in Comparison with the Economic Analysis in law.” Essays on New Institutional Economics, Springer International Publishing: 77 – 96. Robinson, J. R., Y. Xue, and Y. Yu .2011. Determinants of Disclosure Noncompliance and the Effect of the SEC Review: Evidence from the 2006 Mandated Compensation Disclosure Regulations. The Accounting Review, 86 (4): 1415–1444. R Street Institute. 2015. 2015 Insurance Regulation Report Card. R Street Policy Study No. 46, December 2015. Available from http://www.rstreet.org/wpcontent/uploads/2015/11/RSTREET46.pdf (last accessed May 27, 2017). 113 Sadka, R. 2006. Momentum and Post-Earnings-Announcement Drift Anomalies: The Role of Liquidity Risk. Journal of Financial Economics, 80: 309-349. Sage Business Researcher. 2017. Shadow Banking. Available from http://businessresearcher.sagepub.com/sbr-1863-101611-2765611/20170102/shadow-banking (last accessed April 15, 2017). Schenck, K. M. 2012. The Deterrence Effects of SEC Enforcement Actions. Working Paper. Scholz, J. T. 1984. Cooperation, Deterrence, and the Ecology of Regulatory Enforcement. Law and Society Review, 18: 179 – 224. Schwarcz, D. 2015. A Critical Take on Group Regulation of Insurers in the United States. Working Paper at University of Minnesota Law School. Shevlin. T. 1987. Taxes and Off-balance Sheet Financing: Research and Development Limited Partnerships. The Accounting Review, 62 (July): 480 – 509. Sias, R.W. 2004. Institutional Herding. Review of Financial Studies, 17: 165 – 206. Smith, C. W., and R. Stulz. 1985. The Determinants of Firms' Hedging Policies. Journal of Financial and Quantitative Analysis, 20: 391-405. Society of Actuaries. 2000. Managing the “Free Surplus” Portfolio of an Insurance Company. Presented at Chicago Annual Meeting, Session 31PD (Oct 15 – 18, 2000). Society of Actuaries. 2014. Captive Reinsurance and Other Innovative Structures. Available from https://www.soa.org/Professional-Development/Event-Calendar/2014/Adv-Current-TopicReinsurance/Day-1/Session-3-B.aspx (last accessed November 14, 2016). Sparrow, M. K. 2000. The Regulatory Craft: Controlling Risks, Solving Problems, and Managing Compliance. Brookings Institution Press, Washington, DC. Staiger, D., and J. H. Stock. 1997. Instrumental Variables Regression with Weak Instruments. Econometrica, 65(3): 557 – 586. Stigler, G. J. 1971. The Theory of Economic Regulation. Bell Journal of Economics and Management Science, 2: 3–21. Sunder, S. 1997. Theory of Accounting and Control. South-Western Publishing, Cincinnati, OH. Tan, H., S. Wang, and M. Welker. 2011. Foreign analysts following and forecast accuracy around mandatory IFRS adoptions. Journal of Accounting Research, 49(5): 1307-1357. U.S. Electoral College. 2016. Historical Election Results. Available from https://www.archives.gov/federal-register/electoral-college/historical.html (last accessed November 3, 2016). 114 Verisk. 2015. Property/ Casualty Insurance Results 2015. Available from http://www.verisk.com/downloads/InsuranceResultsReport2015Q4.pdf (last accessed June 19, 2017). Wall Street Journal. 2002. “Systematic Failure by SEC is Seen in Enron Debacle.” Jonathan Weil. Available from https://www.wsj.com/articles/SB1033944629262271233 (last accessed May 9, 2017). Wang, X., G. Young and Z. Zhuang. 2008. The effects of mandatory adoption of International Financial Reporting Standards on information environments. Working paper, presented at the American Accounting Association Annual meeting, Chicago, IL. Weinberg, J. 2002. Competition among Bank Regulators. Federal Reserve Bank of Richmond Economic Quarterly, 88(4): 19–36. Williamson, O. 2000. The New Institutional Economics: Taking Stock, Looking Ahead. Journal of Economic Literature, 38: 595 – 613. Zechman, S. 2010. The Relation Between Voluntary Disclosure and Financial Reporting: Evidence from Synthetic Leases. Journal of Accounting Research, 48: 725-765. 115