Do Rating Announcements convey new iInformation?

Diplomarbeit von Jan Klobucnik

Introduction:

Since the beginning of the last century, investors in capital markets have strongly relied on rating agencies’ assessments of credit quality to decide on investments. Due to their important role in debt markets, they are supposed to provide accurate ratings without delay. However, cases like the defaults of WorldCom or Enron have damaged their reputation. In particular, credit rating agencies have been heavily criticized for their role during the financial crisis of 2007-2009. Many economists blame the rating agencies for having played a major part in the securitization process of mortgage loans by providing too high rating grades; and thus sowing the seeds of the crisis. Having rated credit derivatives like collateralized debt obligations with best grades, the rating agencies encouraged banks and other financial institutions to keep these assets in their portfolios.

As a result, it caused severe problems for the banking sector when these products heavily lost in value. Along with imprecise assessments of creditworthiness, the slow reaction of rating agencies has been critizised over the last few years. Therefore, the question of how well the agencies assess credit quality arises. This question is of great importance because of their dominant role on capital markets and the fact that decisions are made upon their ratings. To put it more precisely, this study asks whether the agencies process and convey new information to the market. On the other hand, it might be the case that market participants anticipate any change in the credit quality of a company before these institutions publish their assessments. Answering this question is of particular importance: if the rating announcements convey unknown information and the market reacts, then rating agencies are a systemic part of capital markets and policy should consider stricter regulation to prevent manipulation and failures like those described above. Conversely, if their announcements do not contain any new information – or to put it differently, if markets react faster – then we could think about using market based indicators instead in order to assess credit risk. In this case, the economic task of signaling creditworthiness could be handed over, among others, to Credit Default Swaps (see Chapter 2), which is also suggested by Hart & Zingales. This thesis contributes to the field of rating agencies’ performance measurement.

Evaluating their announcements with the help of the Credit Default Swap (CDS) market, I examine the information content of their ratings. At the same time, this study can be seen as part of the market efficiency research on the CDS market. This empirical analysis shows that there are only weak effects of rating announcements, which might indicate that rating changes is no news to the markets. By observing the CDS market, the empirical evidence displays some anticipation of rating changes, especially for downgrades, such that markets seem to react faster. Moreover, there seems to be some counter-movement to the prior adjustment after the rating was announced. Finally, I can detect both an asymmetric reaction between upgrades and downgrades, which means that latter ones have a stronger impact on the market, and a higher sensitivity in poorer rating classes. In order to give an answer to the question in the title, the thesis proceeds in the following way: The next section provides background information on rating agencies and CDS to build up the theoretical foundation of this study. The previous research and the working hypotheses are presented thereafter in section 3. The subsequent section introduces the data and its transformations. In order to perform the analyses, section 5 addresses the methodology of the event study. The empirical results are then stated in section 6 and finally, section 7 concludes.

Table of Contents:

Text Sample:

Chapter 5.1, Framework:

This study applies the well developed Event Study Methodology, which is described, for example, by Campbell et al.. Generally, it is adequate to measure the impact of a specific event on the value of a firm. Assuming efficient CDS markets with rational agents, new information should lead to a significant reaction in the spreads because only expected information is already reflected in the prices. The idea of an event study is to measure the abnormal change in order to investigate whether an event conveys new and unexpected information to the market. In this application, the event is the rating change while the event window is the day of the announcement (event day) and the following day. As argued above, this two day event date is chosen to capture all event day effects due to the uncertainty about the release time. The pre-announcement period, which this study considers, is from 50 days to one day prior to the event, hence, whereas the post-announcement period is from two days to 30 days after which is. The resulting event period, which looks for abnormal changes in the spreads, is.

The estimation period, which is used to estimate the parameters below, reaches from 300 to 51 days before the event, i.e.. These windows and all following dates refer to working days only. To conduct the event study properly one has to assume the event to be exogenous. In this context it implies that a rating announcement can lead to a reaction in the spreads, but not the other way around. This might be questionable if rating agencies take the information from CDS markets into account to decide upon a rating change: A significant movement in the spreads could then trigger a rating change. One reason for this could be that rating agencies put more effort in designing initial ratings than in the maintenance of these ratings, resulting in incentives to leave this task to the markets and adapt the ratings accordingly. Consequently, such effects would make it hard to detect any reactions that are due to the event or argue in favor of the information content of rating changes. In order to examine the abnormal reactions, one needs a proxy for the „normal” spread changes that would have occurred if there had not been any rating event. There are several different models; two of which are the constant mean return model and the market model. The first one assumes that the normal spread changes are constant. However, for this study we will work with the market model, which proposes the market portfolio as a good choice to simulate the normal changes. The reason is that it is the factor with the single most explaining power for an individual CDS spread change. In this study, I use an index of CDS spreads to approximate the market portfolio.

The market model allows us to remove the portion of the spread change and return that is related to variation in the market’s change. Thereby reducing the variance of the abnormal changes we calculate below, it increases the ability to detect event effects. The index is created as an equally weighted average from the 80 most liquid CDS time series in the data. The series that are included have transactions on at least 90% of all working days between 01/2002 and 09/2009. In addition, the missing values in the series are linearily interpolated. This index describes the trends on the CDS market quite well. The spreads of the index are then transformed into spread changes data SCm,t and return data Rm,t as it was done in equations (4). The market model assumes a linear relation between market movements and CDS price changes. It is a single factor model with the index as explanatory variable to forecast the „normal” spread changes NSCi,t and the „normal” returns NRi,t, which would have been expected if no rating change had taken place.