بازار بیمه عمر: اطلاعات نامتقارن بازبینی شده

This paper finds evidence for the presence of asymmetric information in the life insurance market, a conclusion contrasting with the existing literature. In particular, we find a significant and positive correlation between the decision to purchase life insurance and subsequent mortality, conditional on risk classification. Individuals who died within a 12-year time window after a base year were 19% more likely to have taken up life insurance in that base year than were those who survived the time window. Moreover, as might be expected when individuals have residual private information, we find that the earlier an individual died, the more likely she was to have initially bought insurance. The primary factor driving the difference between our and the prior literature's findings is that we focus on a sample of potential new buyers, rather than on the entire cross section, to address the sample selection problem induced by potential mortality differences between those with and those without coverage.

Empirical testing of contract theory comprises a burgeoning area of economic research (see Chiappori and Salanié, 2003 for a review). Especially important in this literature have been inquiries into whether asymmetric information prevails in particular insurance markets. Much of the literature has adopted the “conditional correlation” approach illustrated in Chiappori et al. (2006), in which the presence of information asymmetry implies that, conditional on risk classification, the risk outcome is positively correlated with insurance coverage. Evidence has been mixed.1 The life insurance market is of particular interest for asymmetric information tests. It is an important market on account of size alone. In 2004, 77% of American households held life insurance. The industry had overall assets of $4.5 trillion and invested $4 trillion in the economy, making it one of the most important sources of investment capital in the United States (American Council of Life insurers (ACLI), 2007a and American Council of Life insurers (ACLI), 2007b). Life insurance contracts also are relatively explicit and simple, and the risk outcomes–policyholders' deaths–are in principle easy to verify and measure. In an important contribution, Cawley and Philipson (1999) use the Health and Retirement Study (HRS) data to examine cross-sections of individual term life insurance contracts and find a negative or neutral correlation between mortality risk and coverage.2 This negative-or-neutral-correlation result, together with their evidence for bulk discounts, has been widely cited as evidence that life insurance markets are free of asymmetric information.3 We find, in contrast, evidence of asymmetric information in these very markets.4 With the same HRS dataset, we recover a significant positive correlation between the mortality outcome and the decision to purchase individual term life insurance, conditional on risk classification. In particular, individuals with higher risk (those who died within a 12-year time window after a base year) were 19% more likely to have purchased individual term insurance in that base year than were individuals with lower risk (those who survived beyond the window). Indeed, decomposing the mortality outcome into time-until-death categories, we find that the earlier an individual died, the more likely she was to have initially taken up insurance. Such monotonicity further suggests the prevalence of asymmetric information.5 The primary factor driving the difference between our and Cawley and Philipsons' earlier findings is that we focus on a sample of potential new life insurance buyers rather than on the entire cross sectional sample. Potential new buyers are the subset of the total sample who did not own life insurance at the beginning of the sample period. They are not subject to the sample selection problem inherent in cross-sectional samples for asymmetric information tests in life insurance markets. This sample selection problem is as follows. Suppose individuals do have residual private information about their mortality risk. Those for whom the information is unfavorable, and who thus decide to buy life insurance, then are more likely to die early and thus less likely to be found in a cross-sectional sample than are those for whom the information is favorable. High-risk individuals with coverage therefore are under-represented in cross-sectional samples. Sample selection induced by potential mortality differences between the covered and uncovered may bias estimates of the conditional correlation between insurance coverage and mortality risk.6 To illustrate, consider the following thought experiment. Four individuals are alive, with the same appearance of good health, at time t − 5. Individuals 1 and 2 choose not to obtain coverage because they know they are in good health. Individuals 3 and 4 do choose coverage because, despite their healthy appearance, they know they are in poor health. At year t − 1, individual 4 dies. The remaining three are randomly drawn into a sample and survive the entire sample period from year t to t + 5. A researcher examining this sample will conclude that asymmetric information is absent: observed mortality in the t through t + 5 window does not differ between the two individuals without insurance and the one with insurance, inasmuch as all three have survived the five-year sample period. The real story, however, is that half of the covered, and neither of the uncovered, have died within the full ten-year (t − 5 to t + 5) horizon. The remainder of the paper is organized as follows. Section 2 describes the dataset. Section 3 discusses the empirical strategy, in which we focus on the sample of potential new buyers together with proper risk classification controls and a 12-year-window ex-post mortality risk measure. Section 4 presents the results. Section 5 concludes.

We find, contrary to the earlier literature, evidence for asymmetric information in the life insurance market. After risk classification is carefully taken into account, individuals with higher mortality risk are 19%–49% more likely to buy individual term life insurance than are those with lower risk, depending on the length of the time window within which the mortality risk is defined. Moreover, buyers appear to employ this informational advantage by seeking to take up insurance 4–6 years before death. Such results provide an alternative view of the informational content of life insurance markets, calling into question the widely held notion that life insurance is free of asymmetric information. Furthermore, the failure of the life insurance industry's comparatively stringent underwriting practices to eliminate strategic purchasing suggests that informational asymmetry might be even more prevalent in other insurance markets. Our focus on potential new buyers is what drives the difference between our and the earlier literature's findings. When individuals do hold private information, analysis based on the entire cross-sectional data produces downward-biased estimates of the coverage-risk correlation. This bias arises because high-risk individuals with coverage are under-represented in cross-sectional samples, so that observed mortality differences between those with and without coverage are dampened in such samples. Our main contribution has been to show that restricting the sample to potential new buyers solves this sample selection problem, providing unbiased asymmetric information tests.