The current paper takes an alternative view to the observed “triple-A failure”. It considers truth-telling credit rating agencies that chose endogenously their rating efforts, i.e. the accuracy of ratings. The possibility to lie to the market is therefore explicitly excluded. The motivation is rating agencies are subject to regulation and have to keep the documents and rating procedures upon which the published ratings are based. Since regulators can verify the rating results by repetition, the rating agencies may risk considerable penalties by lying. On the other hand, unlike corporate bonds, rating structured finance products is much like a statistical inference: The default rate is estimated by a statistical model. Naturally, the quality of such estimation can be improved if more data are available or more meticulous estimation methods are used. Given the rating failure is most prominent for such products, it is important to know whether the credit rating industry is well designed so that rating agencies are induced to put high effort on their job.

From a social point of view, credit ratings can be seen as a costly investigation which yields a better understanding of the risk of certain assets such that some inefficient investment can be avoided. Specifically, two types of inefficiency can emerge in the allocation of risky assets. One is investors underestimate the risk of the asset and make a loss due to their purchase. The other is the risk level is overestimated such that investors refrain from investments that are potentially profitable. As an accurate rating helps to avoid both types of inefficiency, when deciding the rating efforts, a social planner will take into account both the potential profits and the avoided loss. For private rating agencies, however, profits are derived only from the investors who buy the risky asset. Warning of the underestimated risk will discourage the purchase and reduce the revenue of the private agencies’. Since the avoided loss is not taken into account as private benefit, rating agencies may pay more attention to avoid the overestimation of risk than to the underestimation, leaving the rating efforts distorted below the social optimal level in certain circumstances. The inaccurate ratings in turn will increase the amount of risky assets sold in the markets, a pattern observed in the current subprime turmoil. The rest of the paper is organized as follows. A brief literature review is presented in section 2. Section 3 sets up the formal model. To provide a benchmark, section 4 analyzes the social optimal effort provision and some comparative statics. In section 5, the agency problem is examined. The analysis mainly concentrates on possible reasons for rating effort underprovision and its impact on the amount of risky asset sold in the economy. Section 6 discusses the future extension of the model by allowing competition. Section 7 concludes and suggests some caveats.