Tax Sheltering Behavior of Property-Liability Insurers: Some Additional Evidence
利用加州保险公司数据，通过分析股权水平与投资组合选择的关系，检验财产责任保险公司的避税行为，发现股权水平高的公司更多投资于税收优惠证券。
Past tests of the tax sheltering behavior of property-liability insurers have looked at the relation between net income and tax-exempt income. This article proposes an alternate test that uses the data on insurers' portfolio choices. The relation between insurers' equity level (policyholders' surplus to total assets) and their portfolio choice is formally derived and tested using data for California insurers. The prediction that insurers with higher equity levels will invest more in tax-favored securities is supported by the data. The test results also support the view that policyholders bear the corporate tax burden of insurers. The portfolio behavior of property-liability insurers has been of considerable interest to researchers in finance and insurance. Hendershott and Koch (1980) and Cummins and Grace (1988) provide recent examples of such research. These authors base their models of property-liability insurers' portfolio choices between taxable and tax-exempt bonds on profit maximizing behavior. They assume that the yields from tax-exempt bonds are higher than the after-tax yields from taxable bonds for property-liability insurers. Thus an insurer will invest in taxable bonds until all the underwriting losses and other expenses are offset by the investment income from these bonds. Thereafter, the insurer will invest in tax-exempt bonds. Such a model leads to an empirically testable hypothesis that an insurer's reported net income should equal its investment income from tax-exempt bonds. The aforementioned studies, therefore, focus on testing the relation between reported net income and tax-exempt income. In this article an alternative test for the tax sheltering behavior of property-liability insurers is proposed. Unlike the previous studies which examined property-liability insurers' bond portfolio choices by using income CharngYi Chen and Richard PonArul are Professor of Finance and Associate Professr of Finance, California State University at Chico. The authors thank the Associate Editor and two anonymous referees for valuable suggestions. This content downloaded from 157.55.39.163 on Wed, 21 Sep 2016 04:44:02 UTC All use subject to http://about.jstor.org/terms Tax Sheltering Behavior of Property-Liability Insurers 723 data, this test directly employs portfolio data. Property-liability insurers are assumed to select a mix of taxable and tax-favored securities in order to maximize the tax benefits of underwriting losses.' This leads to a hypothesis about the effect of the insurers' equity level (policyholders' surplus to assets ratio) on their tax sheltering behavior. The rest of the article is organized as follows: The next section reviews the theory of property-liability insurer's tax sheltering behavior and develops the testable hypothesis. Then, the data and results of empirical tests of the proposed hypothesis are discussed. Concluding remarks appear at the end of the article. Theory and Hypothesis Property-liability insurers are assumed to have a choice of two classes of securities-taxable and tax-favored. The investment income from a taxable security is fully taxed, while the investment income from a tax-favored security is either partially or fully tax-exempt. As in past studies, the following market equilibrium condition is assumed: for a given level of risk, the pre-tax yield from a taxable security is higher than that from a taxfavored security, but the after-tax yield (for the insurer) from the taxable one is lower than that from the tax-favored one.2 To formally derive the relation between the equity level and the portfolio choice, consider a single-period model of the insurer, like the one found in Smith (1989). At the beginning of the period P dollars of premiums are written. Losses will occur and loss payments will be made at the end of the period. The following notations are used: X = proportion of portfolio invested in tax-favored securities P = total premiums collected s = policyholders' surplus to total premiums ratio at the beginning of the period S = equity level (policyholders' surplus to assets ratio) at the beginning of the period Rs = required rate of return for the insurer's equity Te = effective tax rate for returns on tax-favored securities Re = equivalent pre-tax rate of return for tax-favored securities. If the tax-favored portion of the portfolio is entirely in municipal bonds, then Te is the implicit tax rate defined as one minus the ratio of tax-exempt and taxable yields (see Heaton, 1986, p. 484). If the tax-favored portion is entirely 'Treasury bonds and corporate bonds are considered taxable securities since the interest income is fully taxable. Common and preferred stocks are considered tax-favored since only 20 percent of the dividend income is taxable. Further, in the case of common stock, a substantial portion of the return is typically in the form of capital gains. The effective tax rate on capital gains can be lowered by postponing the realization of gains. Municipal bonds are considered, in this article, as tax-favored rather than tax-exempt since their reduced yields create a defacto tax incidence. 2This assumption implies that the tax rate of marginal investors is lower than those of property-liability insurers. Heaton (1986) reports evidence to support this assumption. This content downloaded from 157.55.39.163 on Wed, 21 Sep 2016 04:44:02 UTC All use subject to http://about.jstor.org/terms 724 The Journal of Risk and Insurance in preferred and common stocks, Te is one minus the ratio of average after-tax expected return to the average pre-tax expected return. For each insurer the value of Te will depend on the composition of the tax-favored portion of its portfolio. The after-tax rate of return for the tax-favored portion of the portfolio will be Re(1-Te). The insurance contracts are assumed to be zero beta securities.3 To simplify the model, all the financial assets in the insurer's portfolio are assumed to be zero beta.4 The insurer's decision rule for choosing between taxable and tax-favored securities will be as described by Hendershott and Koch (1980): invest in taxable securities to the point where income from these securities just offset the underwriting losses and invest the rest in tax-favored securities. In a competitive insurance market, the economic profit for the insurer should be zero. Hence the expected income from tax-favored securities should provide the expected return for the shareholders of the insurer. At the beginning of the period P(1 + s)X is invested in tax-favored securities. The after-tax expected return from the tax-favored securities will be P(1 + s)XE[Re](1 Te). The required return for the insurer's equity, sP, will be sPE[R,]. Thus, P(I1 + s)XE [Re] ( 1-Te) = sPE [R,] ( 1) Since the betas of the policies and the investment portfolio are zero, the insurer's equity will be zero beta. Thus the expected rate of return for the insurer's equity, E[R,], will be equal to E[Re] and (1) will reduce to (1 + s)X(1-Te) = s (2) By definition S = s/(1 + s). Substituting this into (2) and solving for X