Academic journal article Journal of Risk and Insurance

Decomposing Asymmetric Information in China's Automobile Insurance Market

Academic journal article Journal of Risk and Insurance

Decomposing Asymmetric Information in China's Automobile Insurance Market

Article excerpt



In theory, insurance markets are highly susceptible to problems of asymmetric information. However, empirical studies have not provided consistent support for this premise. Moreover, even in cases of empirically demonstrable asymmetry, it rarely is clear to what extent problems can be assigned to the distinctly different categories of adverse selection, ex ante moral hazard (i.e., morale hazard and/or planned fraud), and ex post moral hazard (i.e., claim build-up and/or opportunistic fraud).

This state of affairs is attributable to two factors. On the one hand, asymmetric information is likely to manifest itself differently in different markets, depending on both market structure and development. On the other hand, empirical analyses have been hampered by both data shortcomings and methodological limitations. In the present article, we investigate the impact of asymmetric information on China's automobile insurance market using dynamic data from a large domestic insurance company. Our approach differs from previous research both by distinguishing between the often-confounded effects of adverse selection and ex ante moral hazard on the probability of a claim, and by testing for the separate presence of ex post moral hazard.

Previous Literature

Rothschild and Stiglitz (1976) and Shavell (1979) provided the first simple theoretical models of adverse selection and moral hazard in insurance markets. Dionne (1983), Cooper and Hayes (1987), and Dionne and Lasserre (1985, 1987) introduced multiple-period models to demonstrate the difficulty of concealing a high risk level over time because of the insurance company's ability to learn from policyholder claim histories. In particular, Dionne and Lasserre (1988) noted the importance of information disclosure in simultaneously reducing problems of adverse selection and ex ante moral hazard. Boyer and Dionne (1989) and Chassagnon and Chiappori (1997) noted further that the claim probability is not simply an exogenous variable, but derives from the policyholder's efforts to enhance personal safety and reduce collision-related costs. This observation became the foundation for subsequent empirical studies attempting to distinguish between adverse selection and ex ante moral hazard.

Over the past few decades, researchers occasionally have gained access to policyholder-level insurance company claim data, enabling empirical tests for the presence of asymmetric information. Unfortunately, the results of these studies have been inconsistent. Puelz and Snow (1994), Fong (2002), Finkelstein and Poterba (2004), Cohen (2005), He (2009), and Gao and Wang (2011) found support for the existence of asymmetric information in various insurance markets using a test of positive correlation between risk and insurance coverage. Moreover, Finkelstein and McGarry (2006), Fang, Keane, and Silverman (2008), and Gan, Huang, and Mayer (2011) detect asymmetric information after accounting for heterogeneity of insurance demand. However, Cawley and Phillipson (1999), Chiappori and Salanie (2000), and Dionne, Gourieroux, and Vanasse (2001) did not find evidence of asymmetric information.

Even within the automobile insurance literature itself, results are mixed. In the seminal work of Puelz and Snow (1994), the authors evaluate 1986 data from commercial insurance companies in the U.S. state of Georgia. Using a two-equation regression model, they were able to verify the prediction of Rothschild and Stiglitz (1976) that higher-risk policyholders tend to purchase greater insurance coverage. However, when Dionne, Gourieroux, and Vanasse (2001) introduced a new variable (i.e., an estimate of the claim probability based upon observable covariates) into Puelz and Snow's second equation, they reached the opposite conclusion. At about the same time, Chiappori and Salanie (2000) applied a revised version of Puelz and Snow's model to French automobile insurance data, and also found no evidence of asymmetric information. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.