私有信息、评级偏差和中国评级机构的市场声誉

考虑到评级机构拥有市场声誉的本质在于其可以通过扭曲评级从而对市场产生影响,本文分两步研究中国发行人付费评级机构的市场声誉:第一步,将信用评级对各种基本面因素进行回归,并以实际评级与回归预测值的差值作为评级偏差的量度。与既有文献相比,本文的重要改进是在基本面因素中引入了发债企业与各评级机构(分支机构)最短距离的均值和方差,这能较好地控制因发债企业私有信息可能造成的选择偏误。第二步,考察评级偏差和机构特征如何影响企业的发债成本。研究表明,中国评级机构作为一个整体具有显著的市场声誉,但各评级机构之间存在很大的差异性。最后,考虑到评级机构与发债企业在选址上可能会有集聚效应,我们基于高铁开通事件进行双重差分检验,研究表明本文结论是稳健的。

Private Information,Rating Distortion,and Market Reputation of Credit Rating Agencies in China

Credit rating agencies (CRAs) play an important role in the capital market. In theory, they provide reliable decision-making information for investors and mitigate information asymmetry in the capital market. However, the increasing number of debt defaults has led some investors to have serious doubts about CRAs. They suspect that CRAs do not play the role of gatekeeper in the capital market and may even collude with issuers to transfer risk to innocent investors. This paper studies whether Chinese CRAs have lost their market reputation. Measuring their market reputation is difficult because it involves a paradox: on the one hand, to build their market reputation, CRAs must provide reliable information to investors; on the other hand, CRAs with a good market reputation can milk investors by providing misleading information. Indeed, if a CRA manipulates a credit rating but has no significant impact on investors or on the debt issuance cost, the rating will be a “rubber stamp”; that is, the CRA actually has no market reputation. This paper studies the market reputation of Chinese CRAs in two steps: (1) measuring the rating distortion of CRAs; and (2) examining whether and how rating distortion and CRA characteristics affect the debt issuance cost. To measure rating distortion, this paper makes two salient contributions to the literature. (1) We separate rating distortion from CRAs' private information. In the literature, it is common practice to regress a credit rating with respect to public information and use the residual to measure rating distortion. However, considering that the de facto observable ratings are only given by those CRAs that ultimately win the rating competition, this approach may pool the rating distortion and favorable private information observed by the winning CRA. To extract real rating distortion from the residual, we follow Tian (2011) and introduce the distances between CRAs and issuers to control private information. However, to address selection bias, we not only introduce the average distance between the issuer and all CRAs to capture the monitoring effect but also the variance of the distance to capture the effect of private information on rating competition. Intuitively, credit ratings should decrease in average distance because monitoring becomes more difficult over a greater distance; however, the “observed” credit rating should increase in distance variance, because with a mean-preserving transformation, the debt issuer will choose a CRA’s rating only when the CRA moves closer to the issuer, finds favorable information, and gives a higher rating. (2) We use the propensity score matching (PSM) method to mitigate possible endogeneity. Because we can only observe the actual rating for each bond, the regression prediction value based on the full sample will result in serious measurement errors if there is great heterogeneity among different bonds. Therefore, correctly measuring rating distortion entails constructing a reasonable benchmark rating for each bond. Our solution is as follows: for each bond, we use PSM based on our public information and distance variables to find similar bonds as a control group. The bonds in the control group should be rated by other CRAs; thus the treatment effect between two groups is a more accurate measurement of rating distortion. The main data, taken from the Wind database, cover information on corporate bonds and enterprise bonds from January 2009 to October 2017. After excluding bonds without ratings, there are 6,073 observations. Our findings are as follows. First, credit ratings decrease in average distance between debt issuers and the CRAs, in accordance with the monitoring mechanism. Second, credit ratings increase in distance variance, in accordance with the private information channel. Third, on average, CRAs in China still have a good market reputation because upward rating distortion significantly reduces the cost of issuing bonds. Fourth, there is significant heterogeneity among CRAs. Fifth, we perform a DID analysis using high-speed railway openings as a quasi-natural experiment to tackle possible endogeneity problems due to the agglomeration effect in the locations of CRAs and issuers. We find that all of the main results are robust.