Abstract
This article analyses the effect of rating agencies’ decisions on stock risks for European issuers concerning five kinds of events. Our approach is an extension of dummy variable regression event study methodology, using a GARCH(1,1) estimation to capture simultaneously the impact on both systematic and specific stock risks. This new methodology allows us to obtain both global results by categories of rating decisions and individual results, event by event. We document, globally, a positive impact of upgrading on systematic risk, a negative impact of rating confirmation on specific risk, and no significant impact in all other cases. Regarding event-by-event results, the proportion of rating actions exhibiting a significant effect on risk is almost always observed between 20% and 30%. The weak evidence of a global effect on systematic risk may be due to the lack of informational content of the rating decisions on the stocks’ risk, or the existence of rebalancing effects between systematic and idiosyncratic risks. Furthermore, it should be noticed that the decline in volatility in case of a rating affirmed is an insight of the certification role played by the agencies.
Notes
1. That is, a decrease in the prices on the bond market in case of ‘negative event’ such as a downgrade or an inclusion into watch for possible downgrade and vice versa in case of a ‘positive’ event.
2. To date, most of the authors have looked at systematic risk only.
3. Previously, among the few contributions to the topic of ‘rating and stock risk(s)’ only Abad-Romero and Robles-Fernandez Citation(2006) have presented both global and individual conclusions (cf. Section 4).
4. The relationship built by Carpenter and Chew Citation(1983) is the following: β1=(δ S/δ V)·βu+βu·[(1−T)·(D1/S1)·(δ S/δ V)], where V is the value of the firm, S1 the value of the levered firm's stocks, D1 the value of the risky debt, T the tax rate, and βu the systematic risk of the unleveraged firm's stocks.
5. In a previous version, the model was run including a dummy variable to capture a possible shift in the α parameter. As the estimated coefficient was not significant, we decided to withdraw this dummy from the present version (the results are available upon request).
6. As mentioned in Section 2, information about the motivations behind the rating action (variation in leverage, in economic risk, in the variance of cash flows) would have been helpful to explain the occurrence of a change and its direction. Nevertheless, as most often mentioned in previous literature, it was impossible to obtain such details. The same reason also prevails for the size of the issuer at different dates.