Organizational Structure and Insurers Risk Taking: Evidence from the Life Insurance Industry in Japan

Transcription

1 Organizational Structure and Insurers Risk Taking: Evidence from the Life Insurance Industry in Japan Noriyoshi Yanase, Ph.D (Tokyo Keizai University, Japan) 2013 ARIA Annual Meeting 1 1. Introduction 2013 ARIA Annual Meeting 2 1

2 Motivation Insurance company s risk taking behavior has an impact on Stockholders Policyholders Employees Regulators Many papers: Organizational Structures and Managerial Activities Alchian/Demsetz (1972), Jensen/Meckling (1976), Mayers/Smith (1981, 1986), Fama/Jensen (1983), Huang et.al, (2011).. Few papers have dealt with the relation between organizational structures and the insurer s risk taking. Lamm-Tennant/Starks (1993), Lee/Mayers/Smith (1997), Ho/Lai/Lee (2013) 2013 ARIA Annual Meeting 3 Purpose This study aims to examine the effects of organizational structures on an insurer s risk taking, Using data from Japanese life insurers for fiscal years between 1983 and 2007, Including ex -deregulation period ( ) and post - deregulation period ( ) Examining if the deregulation has an effect on the relation between organizational structures and an insurer s risk taking behavior ARIA Annual Meeting 4 2

3 Why Japan? Why Insurance? 2 streams of research To examine the relation between organizational structures and a company risk taking, Insurance companies in Japan have two interesting features: 1. Keiretsu vs. Independent <--- (Japanese Firms) Hoshi/Kashyap/Scharfstein (1991), Prowse (1992), Weinstein/Yafeh (1998) 2. Mutual vs. Stock<--- (Insurance Firms) Mayers/Smith (1981, 1986), Lamm-Tennant/Starks (1993) We integrate two streams of research related to the organizational structures ARIA Annual Meeting 5 Related Studies Risk Taking (Insurance) Lamm-Tennant/Starks (1993) --- J. of Business First article that directly investigates the risk taking behavior of the U.S. P/L insurance industry in terms of Underwriting Risks They use the standard deviation of loss ratio as risk measures which is a book-value measure because mutual companies do not have market value risk measure. Lee/Mayers/Smith (1997) --- JFE Examining P/L insurance company s risk-taking in the U.S market in terms of changing of an asset portfolio, or Investment Risks They find that the shift to riskier assets, mainly stocks, following fund enactment occurs only for stock insurers. Ho/Lai/Lee (2012) --- JRI Examining the impact of organizational structure and board composition on risk taking in the U.S. property casualty insurance industry. They find that mutual insurers have lower total risk, underwriting risk, and investment risk than stock insurers ARIA Annual Meeting 6 3

4 Related Studies Keiretsu (Business Group) Recently, diversified business groups around the world is receiving increasing attention in the financial economics literatures. Khanna/Yafeh (2005, 2007) --- J. of Business He et.al., (2013) --- J. of Corporate Finance The prominent role played by business groups has been discussed in a string of literatures on the Japanese bankcentered corporate groups, known as the keiretsu. The keiretsu is a good testing ground for discussing the relation between organizational structures and company s risk taking behavior ARIA Annual Meeting 7 Findings 1. First, keiretsu life insurers have lower investment (underwriting) risk than independent life insurers in the post-deregulation period. 2. Second, keiretsu life insurers hold more stocks in their asset portfolios than independent life insurers. 3. Third, mutual life insurers take a lower investment risk than stock life insurers in the post-deregulation period. Overall, the evidence of ex-deregulation period versus postderegulation period implies that regulation has an impact on risk taking behavior in terms of organizational structures ARIA Annual Meeting 8 4

5 2. Background 2013 ARIA Annual Meeting 9 Japanese Insurance Market 1 Second Largest Market 2 Coexistence of Mutual and Stock 3 Keiretsu-affiliated Insurers 4 Single Regulation 5 Deregulation 2013 ARIA Annual Meeting 10 5

6 (1) Second largest market Share of the world market was 20% in 2011 (Source) WORLD INSURANCE IN 2011 SHOWS NON-LIFE INSURANCE PREMIUMS CONTINUED TO GROW IN 2011 DESPITE AN OVERALL DECLINE IN PREMIUMS, Sigma 3/2012, Swiss Re. (Note) Life premium volume in USD in ARIA Annual Meeting 11 (2) Coexistence of Mutual and Stock 2013 ARIA Annual Meeting 12 6

7 (3) Keiretsu 10% 20% Firm A (business) 15% Firm B (Insurer) Cross- share holdings 10% (Main) Bank Horizontal conglomerates Firm D (business) Firm C (business) 2013 ARIA Annual Meeting 13 (3) Keiretsu (cont.) Keiretsu is a special type of organizational structure in Japanese industries (incl. insurance). Past researches have shown that the unique organization structure (corporate groupings) in Japan known as the keiretsu have relatively Low bankruptcy costs (e.g., Nakatani, 1984) Low agency conflicts (e.g., Hoshi/Kashyap/Scharfstein, 1990, JFE) Low information asymmetry (e.g., Hoshi/Kashyap/Scharfstein, 1991, QJE) Low effective taxes (e.g., Gramlich/Limpaphayom/Rhee, 2004, JAE) Some of JP life Insurance companies belong to the keirestu ARIA Annual Meeting 14 7

8 (4) Single regulation Japanese life insurance industry is regulated by single insurance business law, the authorities, and tax code, while in U.S. the insurance is sold in different states and regulated by the law of various states. The results are reliable ARIA Annual Meeting 15 (5) Deregulation The regulation of the post-war life insurance industry was characterized by the so-called convoy system until the mid- 1990s. By utilizing the convoy system, after the war, the government tried to revitalize existing companies by limiting competition and forbidding new entries into the market, thus sheltering the industry. The Japanese government initiated a sweeping relaxation of regulations from the mid-1990s. In 1995, the insurance business law of Japan was drastically amended for the first time in 56 years. One of the primary purposes of the new law, which went into effect in April 1996, is to promote competition among insurers. Differential managerial activities (e.g., risk taking) cannot (can) be observed in the ex (post) deregulation period ARIA Annual Meeting 16 8

9 3. Hypothesis Development 2013 ARIA Annual Meeting 17 Keiretsu vs. Independent Mitsubishi 三 菱 Mitsui 三 井 Yasuda 安 田 Sumitomo 住 友 2013 ARIA Annual Meeting 18 9

10 Keiretsu Effects In general, shareholders can increase their wealth by taking more risk because of the limited liability provision. On the other hand, there is less incentive for shareholders to expropriate wealth from other stakeholders (policyholders, bondholders, and so on) in Keiretsu insurance companies. In other word, shareholders of Keiretsu affiliate insurance companies do not take too much risk at the expense of other stakeholders. Why? 2013 ARIA Annual Meeting 19 Aspect of Policyholder Policyholders Purchasing insurance products Keiretsu Insurance Company Other keiretsu member Company Holding much stocks 2013 ARIA Annual Meeting 20 10

11 Aspect of Owner Owner Underwriting (1) Subordinated bond (2) Mutual insurance fund (Mutual form) (3) Stocks (Stock form) Keiretsu Insurance Company Other keiretsu member Company Holding much stocks 2013 ARIA Annual Meeting 21 Aspect of Owner Keiretsu-affiliate Insurance Company The link between owner and policyholder in the Keiretsu could cause the Keiretsu affiliated insurance companies taking less risks than independent insurance companies. Policyholder = Owner Keiretsu Affiliation Keiretsu-affiliate Company Keiretsu-affiliate Company Keiretsu-affiliate Company 2013 ARIA Annual Meeting 22 11

12 Keiretsu vs. Independent Stockholder Independent Keiretsu Policyholders Take much risk at the expense of policyholders! Managers Not take much risk! Hypothesis #1 Keiretsu life insurers tend to take less risk than independent insurers ARIA Annual Meeting 23 Robustness: Stock holdings (H3) A keiretsu life insurer usually is a large shareholder (royal shareholder) of its keiretsu member firms and provide stability for the member firms. The member firms underwrite equity funds, subordinate bonds of the insurer, and purchase insurance from the insurer in return. Hypothesis #2 Keiretsu life insurers hold more stocks in their asset portfolios than independent life insurers do. Keiretsu Insurer Large shareholder (Royal shareholder) Underwrite equity funds, subordinate bonds of the insurer Purchase insurance from the insurer in return 2013 ARIA Annual Meeting Keiretsu Member Firm 24 12

13 Stock vs. Mutual 2013 ARIA Annual Meeting 25 Stock Insurers Too much risk taking Managers Shareholders Shareholders (and Managers) could take too much risk at the expense of policyholders (debt holders). Potential conflict Principal of Limited Liability Policyholders (Debt holders) 2013 ARIA Annual Meeting 26 13

14 Mutual Insurers Less risk taking (than in the case of stock insurers) Managers Owners = Policyholders Mutual life insurance companies take less risk because potential conflict between owners and policyholders are eliminated in mutual company. Agency cost is reduced because No potential conflict there is no incentive for shareholders to expropriate wealth from policyholders (Lamm-Tennant and Starks,1993) ARIA Annual Meeting 27 Mutual vs. Stock Stock Mutual Stockholder Policyholders Take much risk at the expense of policyholders! Managers Not take much risk! Hypothesis #3 Mutual life insurers take less risk than stock insurers ARIA Annual Meeting 28 14

15 4. Data and Methodology 2013 ARIA Annual Meeting 29 Data Source: Statistics of Life Insurance Business in Japan [SLBJ] (FY1983 edition-fy2007 edition), edited by Insurance research Institute (Tokyo, Japan). All data is set up by manual entry. Sample: The sample consists of all existing licensed life insurance companies with complete records during FY1983- FY2007 (for 25years). Including both public and private companies Insurers which do business as a branch operation by a foreign insurers (e.g., ALICO Japan, AFLAC, et.al) are excluded from the sample. When insurers were involved in M&A deals, we have to identify the new insurers after the deals for an analysis in panel data set. In this study, we identify (legal) surviving insurers as the new insurers ARIA Annual Meeting 30 15

16 Scope of the sample There are mutual-type organizations similar to a mutual-type insurance company, called cooperative insurers. Since there are many differences between insurance companies and the cooperative insurers in terms of disclosure requirements, capital requirements, and so on, the cooperative insurers are not generally classified as part of the insurance industry. This study uses just the data of mutual-type licensed insurance companies in Japan and does not include the data of the cooperative insurers because consistent data are not available for both organizations, which is a limitation of our analysis ARIA Annual Meeting 31 The number of Life Insurers in Japan Fiscal Year Ownership Keiretsu (A) Keiretsu affiliation Mutual Stock Keiretsu Independent ARIA Annual Meeting

17 Regression (1) S.D. of loss ratio as the underwriting risk to test (H1,H3) (2) S.D. of the rate of return on investment to measure investment risk to test (H1, H3) (3) A ratio of stock holding to total marketable securities as a dependent variable to test (H2) We treat ownership structures as exogenous variables (e.g., Saunders/Strock/Travlos (1990)) 2013 ARIA Annual Meeting 33 Variables Variable Definition Expected Sign Dependent Variables: Underwriting risk Investment risk Stock holdings Independent Variables: Standard deviation of of loss ratio [t+1,t] Standard deviation of the rate of return on investment [t+1,t] A ratio of stock holding to total marketable securities at the end of each fiscal year ASSET Logarithm natural of the amount of total asset that is deflated to real 2005 values using the Consumer Price Index. (+) / (-) BOARD Total number of directors in the board (+) / (-) GROWTH The average growth ratio of premium income each year (+) HHI Herfindahl-hirschman index across the line of business in premium income? ID A proxy variable for the proportion of individual insurance to total premium income? IA A proxy variable for the proportion of individual annuity to total premium income? GD A proxy variable for the proportion of group insurance to total premium income? GA A proxy variable for the proportion of group annuity to total premium income? LEV Debt to Asset Ratio (+) SOLVEN A dummy variable = 1 (if a fiscal year is after 1996) (-) ORG: H1: Negative MUTUAL A dummy variable = 1 (if a insurer is a mutual company) H2: Negative (-) KEIRETSU A dummy variable = 1 (if a insurer is a keiretsu-affiliated company) H3: Positive (-) 2013 ARIA Annual Meeting 34 17

18 Empirical Results 2013 ARIA Annual Meeting 35 Variables (1) (2) (3) (4) Keiretsu Whole Period Ex-Deregulation (A) vs. Post-Degulation Independent Whole Period ( ) ( ) Underwriting Risk Investment Risk ( ) ( ) (5) (6) Ex-Deregulation Post-Degulation ( ) ( ) ASSET BOARD GROWTH The coefficient on KEIRETSU in SOLVEN KEIRETSU Number of observation [0.246] [1.918] * [-0.372] [-2.678] *** [-2.723] *** [-0.020] [-0.365] [-2.865] *** [-0.135] [-1.099] [1.100] [-2.958] *** [2.070] ** [-1.716] * [3.543] *** [1.657] * [0.478] [1.632] column HHI (3) is negative (-2.354) and [-1.712] * [-1.361] [-1.530] [1.448] [2.108] ** [0.412] significant at the 1% level, while that in column (2) is not strong significant. ID (Individual Death) IA (Individual Annuity) These results suggest that the keiretsu GD (Group Death) effect on underwriting risk is strongly driven by the results of the GA (Group Annuity) post-deregulation period. LEV (Debt to Equity) [-8.170] *** [0.193] [-6.079] *** [-6.934] *** [-2.087] ** [-6.667] *** [-8.462] *** [0.324] [-6.304] *** [-6.911] *** [-1.820] * [-6.710] *** [-6.646] *** [-0.515] [-4.071] *** [-6.448] *** [-2.194] ** [-5.011] *** [-8.117] *** [0.438] [-5.949] *** [-6.405] *** [-1.629] [-6.252] *** [-1.129] [-3.357] *** [0.662] [2.406] ** [0.427] [1.752] * [0.312] [-2.501] ** [-2.723] *** [-1.762] * [-2.295] ** [-1.169] [0.216] [-1.982] ** 738 The coefficient on KEIRETSU in column (6) is negative (-0.054) and significant at the 5% level, while that in column (4) is not significant. These results suggest that the keiretsu effect on investment risk is also strongly driven by the results of the post-deregulation period ARIA Annual Meeting R-squared

19 Variables ASSET BOARD GROWTH LEV (Debt to Asset) KEIRETSU (1) Stock holdings Whole Period Ex-Deregulation ( ) ( ) The coefficient on KEIRETSU in [1.805] * [-1.060] [-0.100] [1.063] [2.885] *** [0.666] [1.425] [2.078] ** [-2.303] ** [0.815] [-0.627] [1.421] [0.661] [-0.600] [1.613] [2.385] ** [-0.611] [4.711] *** 0.408* [1.928] * [-0.307] [3.359] *** [1.253] [4.438] *** [-1.267] [-6.535] *** Keiretsu (A) column HHI (3) is positive (14.546) and [1.047] [-0.031] [0.419] significant at the 1% level, while that in column (2) is not strong significant. ID (Individual Death) IA (Individual Annuity) These results implies that keiretsuaffiliated GD (Group Death) companies tend to hold more stocks especially in the postderegulation GA (Group Annuity) period. SOLVEN (2) (3) Post-Degulation ( ) [5.323] *** [1.907] * [6.710] *** 2013 ARIA Annual Meeting 37 Number of observation R-squared Variables (1) Underwriting Risk (2) (3) (4) Ex-Deregulation Post-Degulation Mutual vs. Stock Whole Period ( ) ( ) Whole Period ( ) ( ) Investment Risk (5) (6) Ex-Deregulation Post-Degulation ( ) ( ) ASSET BOARD GROWTH The coefficient on MUTUAL in column HHI (4) is negative (-0.051) and significant at the ID (Individual 1% level, Death) implying that mutual insurers take less investment risk than IA (Individual Annuity) stock insurers do as a whole. GD (Group Death) GA (Group Annuity) LEV (Debt to Asset) SOLVEN MUTUAL Number of observation [-0.393] [1.627] [-0.778] [-2.007] ** [-2.831] *** [0.150] [-0.301] [-2.757] *** [-0.224] [-1.145] [1.208] [-2.733] *** [2.090] ** [-1.691] * [3.498] *** [1.511] [0.525] [1.390] [-1.591] [-1.480] [-1.424] [1.167] [2.224] ** [0.341] [-8.181] *** [0.319] [-6.114] *** [-6.910] *** [-2.200] ** [-6.867] *** [-8.422] *** [0.455] [-6.314] *** [-6.905] *** [-1.932] * [-6.909] *** [-6.699] *** [-0.441] [-4.251] *** [-6.666] *** [-2.259] ** [-5.488] *** [-8.237] *** [0.534] [-6.025] ** [-6.257] *** [-1.749] * [-6.165] *** [-0.795] [-3.194] *** [0.966] [2.085] ** [0.559] [1.796] * [0.418] [-2.810] *** [-0.395] [-0.691] [-0.531] [-2.576] *** [0.810] [-3.533] *** 738 The coefficient on MUTUAL in column (6) is negative (-0.096) and significant at the 1% level, while that in column (4) is insignificant. This suggests that the negative effect of MUTUAL on insurers investment risk is strongly contingent upon the results of the post-deregulation period ARIA Annual 413 Meeting R-squared

20 Conclusion Organizational Structures Mutual Keiretsu Less Risk taking In the postderegulation period More stocks in the asset portfolio 2013 ARIA Annual Meeting 39 Further issue We need to address the following explanation in the future study. We find that (1) Keiretsu insurance companies take relatively low investment risk. (2) Keiretsu insurance companies hold relatively high stock holdings. Generally speaking, An investment for stocks relatively correlates with a high investment risk of an investor (an insurance company in this study). So, Can we explain the above findings (1) & (2)? It s a our one of the future tasks ARIA Annual Meeting 40 20

21 Thank you! Welcome your comments! Noriyoshi YANASE, Ph.D. (Tokyo Keizai Univ., JAPAN) 2013 ARIA Annual Meeting 41 RESPONSE TO THE COMMENTS 2013 ARIA Annual Meeting 42 21

22 Suggestions Motivation / Contribution What does the analysis of stock vs. mutual risk taking add to the literature? Lamm-Tennant and Starks (1993) Lee, Mayers, & Smith (1997) The keiretsu insurer structure is similar to the alliance governance. This may provide additional and possibly stronger motivation for your hypothesis 2 and provide you a broader audience Yes, I know that the analysis of stock vs. mutual risk taking has no additional contribution except using out of the sample Japanese data. Absolutely, I agree with you. So, I try to more strengthen the analysis of the keiretsu effect rather than mutual effect. Suggestions Organization Shorten the introduction by 1-2 pages Move the description of the keiretsu organization into one subsection and provide details and cites Greater description of the empirical methods in the paper Type of regression analysis Diagnostics Yes, I ll make clear the description of the keiretsu organization, and strengthen description of the empirical methods in the paper for readers. 22

23 Suggestions Empirical (1) Inclusion of the four life insurance segments creates a linear dependence Actually, there are other segments but the four segments in the data book, including health, disability, and so on. Also, I checked the correlation matrix among the 4 variables, but I am little concern about a perfect linear dependence. However, for robustness, I will check the results dropping one or two insurance segment variables in the regression. Suggestions Empirical (2) Hausman test You state that individual effects are correlated with explanatory variables. Shouldn t this indicate a need for a fixed effects model? Sorry about my sentence in the footnote(22) is confusing. Yes, your mention is absolutely right. But, actually, my regression results come from the random effects model. Thank you, I will make clear in the sentence for the methodology. 23

24 Suggestions Empirical (3) Define the leverage variable. What are debt and equity in this case? Definition of the leverage variable is very simple in this study. Due to the disclosure problem, I can consistently get the amount of total liabilities and the total asset in book value. So, I just calculated the leverage as (total liabilities) over (total asset minus total liabilities). Suggestions Empirical (4) Are the stock holdings of life insurers regulatory limited? Is there a risk based capital requirement? How does this influence hyp. 3? Before the deregulation age, the Insurance Business Law (IBL) had regulations for insurers asset allocation, therefore, the stock holdings were in effective limited. In the deregulation period, on the one hand, the regulations of insurers asset allocation have been loosened. On the other hand, the solvency margin regulation, a kind of a risk based capital requirement, introduced to the industry. That s why I need to control the solvency margin regulation, and so this study conducts a dummy variable for implementation of the solvency regulation. Of course, for more robustness, I try to conduct an additional regression using a control variable for a ratio of solvency margin of each firm. 24