I’m going to test for the effect corporate credit rating announcements have on stock prices through different economic climates (good times vs. bad times). I want to research whether or not the stock price reaction to a corporate credit rating upgrade/downgrade is more extreme in a tough economic climate (financial crisis).

A lot of research has been done on corporate credit rating announcements (upgrade, downgrade, positive and negative watch) effect on stock prices but not a lot of research has compared the results from different time periods (before/after vs. during financial crisis, before vs. after regulation). Joo & Pruitt (2006) studied the Korean financial crisis, and Jorion et al. (2006) studied the effects before and after Regulation Fair Disclosure. However, these papers are fairly short and don’t explain the statistical framework in detail (At least to me it’s not clear what they are doing).

I have done some previous research using the event study methodology presented by Brown & Warner (1985) and MacKinlay (1997) but I’m not sure if it’s applicable to this problem. And if it is, I’m not completely sure how to compare the results from different periods.

I'd be forever grateful if anyone could point me in the right direction.

  • Brown, S. J., & Warner, J. B. (1985). Using daily stock returns: The case of event studies. Journal of financial economics, 14(1), 3-31.
  • Joo, S. L., & Pruitt, S. W. (2006). Corporate bond ratings changes and economic instability: Evidence from the Korean financial crisis. Economics Letters, 90(1), 12-20.
  • Jorion, P., Liu, Z., & Shi, C. (2005). Informational effects of regulation FD: evidence from rating agencies. Journal of financial economics, 76(2), 309-330.
  • MacKinlay, A. C. (1997). Event studies in economics and finance. Journal of economic literature, 13-39.
  • $\begingroup$ Hi @GaryUpper and welcome to quant.SE! Could you specify which effects you are going to measure? If I understood, you want to measure the effects of an event during financial crises and during normal times, seeing if it is greater during economic distress time period, right? If it is, I suggest you to edit the question specifying which event you want to model, because, in this way, it should be easier helping you; otherwise, it would be a too broad question, IMHO! $\endgroup$
    – Quantopik
    Jul 10, 2015 at 18:04
  • $\begingroup$ Thanks! Appreciate the feedback @Quantopic. Better now? Changed it from a very general question to more specific. $\endgroup$
    – Gary Upper
    Jul 10, 2015 at 20:10
  • $\begingroup$ Hi @GaryUpper! Please, remember to mark the answer with a check, in the case your question has been fulfilled! $\endgroup$
    – Quantopik
    Jul 11, 2015 at 16:05

1 Answer 1


The rating downgrade/upgrade effect is definitely more extreme during financial crisis, because of several effects (among all, flight-to quality, flight-to-liquidity and news effects itself), as shown by:

Arezki, Rabah, Bertrand Candelon, and Amadou Nicolas Racine Sy. "Sovereign rating news and financial markets spillovers: Evidence from the European debt crisis." IMF working papers (2011): 1-27.

The paper analyzes the news effects on financial markets by using simply linear regression model, as follows:

$r_{i,t}$ = $\alpha$ + $\sum\limits_{i=1}^n \beta_i*D_{i,t}$ + $\epsilon_t$

where $r_{i,t}$ is the market returns relative to the country you're going to analyze and $D_{i,t}$ is the variable relative to the presence of news; if $D_{i,t}$ results to be statistically significant and different from $0$, then news will have been affected the market.

This could be a solution to your question, but, alternatively, you could take into consideration:

Brown, S. J., & Warner, J. B. (1985). Using daily stock returns: The case of event studies. Journal of financial economics, 14(1), 3-31.

that you cited in the question, or, better:

Kothari, S. P., and Jerold B. Warner. "The econometrics of event studies." Available at SSRN 608601 (2004).

that is more recent and used in event study methodology.

If you want to carry with linear regression model on, I suggest you to implement the following model:

$r_{i,t}$ = $\alpha$ + $\sum\limits_{i=1}^n \beta_i*D^1_{i,t}$ + $\sum\limits_{i=1}^n \delta_i*D^2_{i,t}$ + $\epsilon_t$

where, in this case, $D^1_{i,t}$ is the news presence during normal times and $D^2_{i,t}$ is the news presence during financial crisis time periods. If $\delta$ $>$ $\beta$, you showed that, on average, news affects financial markets much more during financial crisis than during normal times.

Hope this helps.

  • $\begingroup$ I have to apologize for not specifying that I'm researching -corporate- credit rating announcements. I'm not sure if those regression models can be applied in that case? I'm thinking CAPM, FF, and I'm not sure i completely understand the regression models in the above papers. I hope you can take some of your time to elaborate a little bit further. Again, I apologize for not specifying exactly what i was after, @Quantopic. Thank you! $\endgroup$
    – Gary Upper
    Jul 12, 2015 at 13:44
  • $\begingroup$ Generally, the event-study methodology is equal for all assets/mkts you're going to analyze; moreover, I think that the models I proposed should be better too for the corporate credit market, because you will have more rating news observations with respect to the sovereign one. Said that, can you specify what do you not understand in the regression model explained above, @GaryUpper? It would be easier helping you! $\endgroup$
    – Quantopik
    Jul 12, 2015 at 13:57
  • $\begingroup$ Thanks for coming back to me so quick, @Quantopic. I mean, is it just regressing returns with a dummy variable where news are present? Just the day of the announcement or can i have different size event windows? I might be too caught up in my previous event study research, using the estimation period to obtain the parameters (CAPM Beta) and then projecting the expected returns in the event window based on these. $\endgroup$
    – Gary Upper
    Jul 12, 2015 at 14:25
  • $\begingroup$ If you take a look to the Kothari's paper, you'll see that the windows size usually used is 60 days, but it is a rule of thumb only; you can change that. As regards the methodology, it is the simpler case with one dummy as regressor; you should after calibrate and adjust for the risk the model. I just give you the hint, now, it depends on you and on what you need for! :) Anyway, I suggest you starting from the simpler model case and after adjust calibrate and add other variables! $\endgroup$
    – Quantopik
    Jul 12, 2015 at 14:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.