This paper investigates the insurance market where policyholders have private information on their risk type and some of them, opportunistic policyholders, may file fraudulent claims. This study assumes that insurance companies cannot know all types of policyholders, and that they uncover fraudulent claims by costly audit technology. Under the Rothschild and Stiglitz (1976) conjecture, policyholders select different contracts not based on their honesty type, but based on their risk type if market equilibrium exists. As high-risk type policyholders are audited more frequently compared to low-risk type policyholders, they leave the insurance market first as the monitoring cost increases. In this case, the insurance market shuts down, even when low-risk type policyholders can Pareto-improve their expected utility by fair insurance. Consequently, imperfect information regarding risk type makes the degree of market inefficiency from insurance fraud worse. The market equilibrium also depends on whether insurance companies can commit to their audit strategy. The commitment can improve social welfare by completely preventing fraudulent claims.