Determinants and Impact of Credit Ratings: Australian Evidence. Emawtee Bissoondoyal-Bheenick a. Abstract

Transcription

1 Determinants and Impact of Credit Ratings: Australian Evidence Emawtee Bissoondoyal-Bheenick a Abstract This paper examines the credit ratings assigned to Australian firms by Standard and Poor s and Moody s. The study focuses on the quantitative determinants of credit ratings with reference to the financial ratios for the companies. The model also provides forecast ratings to some firms, which are not rated. The analysis is further extended to assess the impact of a rating change on the firms stock returns. The main finding of this paper is that profitability, size and leverage ratios carry a significant weight in the determination of credit ratings. The results also indicate that previous evidence pertaining to the impact of rating changes whereby only rating downgrades impact on the market, cannot be applied to all the rating agencies. Keywords: Credit Ratings, Ordered Response Model, Event Study a C/O School of Economics and Finance, RMIT University, GPO Box 2476V, Melbourne, Victoria 3001, Australia. address: banita.bissoondoyal@rmit.edu.au. The author wishes to thank Robert Brooks for helpful comments on earlier versions of this paper.

2 1. Introduction The purpose of credit ratings is to provide investors with a simple gradation by which the relative creditworthiness of securities may be noted. Companies in Australia, and elsewhere, generally seek credit ratings in order to facilitate their issuance of bonds and commercial papers. Investors prefer rated securities to unrated securities and in general the higher the rating, the higher is the marketability of the financial securities. Rating agencies, therefore provide a good indication of the creditworthiness and default probability of individual firms. The changes in ratings occur so as to reflect variations in the intrinsic relative position of issuers and their obligations. The Australian firms are mostly rated by the two leading agencies, namely Standard and Poor s and Moody s. Each of these agencies rate individual companies for both longterm obligations including medium term notes, and short-term obligations, that is commercial papers. Standard and Poor s further expands the rating classification in terms of local currency and foreign currency ratings. Credit ratings are important not only because they help to assess risk in the Australian capital markets, but plus they also affect the ability of the Australian firms to borrow in overseas markets. There exists a large body of literature analysing the impact of bond rating changes on individual stocks. Early studies on bond ratings suggest that rating changes convey no information to the market. The studies include Pinches and Singleton (1978), Wakeman (1978) and Weinstein (1977). However, recent studies indicate that rating announcement do have an impact on the market. Some examples include Barron, Clare and Thomas (1997); Cornell, Landsman and Shapiro (1989); Dichev and Piotroski (2001); Ederington and Goh (1993), (1998), (1999); Griffin and Savincente (1982); Holthausen and Leftwich 1

3 (1986);Impson, Karafiath and Glascock (1992); Liu, Seyyed and Smith (1999); Matolcsy and Lianto (1995); Wansley, Glascock and Clauretie (1992); Zaima and McCarthy (1988). Another area of research for credit ratings has been to assess what goes behind the ratings, that is the determinants of credit ratings. Packer (2000) examines the credit ratings assigned to Japanese non-financial corporations by Japanese and foreign rating agencies. The study assesses the determinants of the credit ratings, based on quantitative financial ratios and the model used is the ordinary least square model. The main finding of the analysis is that out of eight variables used, a company s rating in Japan is largely determined by its size, retained earnings, profitability and leverage. Both Moody s and Standard and Poor s appraisal for each firm is both quantitative and qualitative that captures the unique characteristics of each issuer. Other factors that could affect the rating include the industry regulatory framework, its competitive and operating activities, management strategy and risk profile, and the quality and stability of the company s shareholding structure. Identifying the relationship between the qualitative factors and the ratings is difficult because they are not quantifiable. However, both Moody s and Standard and Poor s, in their publications, indicate that financial ratios play a key role in assigning ratings (see for example, Packer (2000), Moody s RiskCalc Model for Australian Private Firms: version 1.5 (2002)). Accordingly, the aim of this paper is to assess the relative significance of financial ratios in the determinants of the ratings for each company. The main contribution of this paper to the existing literature is that (1) it uses an ordered response model which is argued to be more appropriate technique than the ordinary least squares technique for rating studies; (2) it utilises the model with the 2

4 significant ratios to provide a forecast of the ratings of companies which are rated by Standard and Poor s, but unrated by Moody s, given the ratings of these agencies are comparable (see for example Cantor and Packer (1996a)); (3) it undertakes an event study to assess whether the two rating agencies rating changes have the same impact on stock returns. The plan of the paper is as follows. In section 2, the determinants of the ratings including the data and method used as well as the forecast of the rating are presented. Section 3 introduces the event study to assess the impact of credit rating changes on stock returns and the final section includes some concluding remarks. 2. The Determinants of Australian Corporations Credit Analysis 2.1 Data In the rating materials on their websites and publication, the rating agencies indicate a number of criteria that can be used to assign credit ratings. Nonetheless, it is difficult to use the same criteria for the following reasons. First, the rating agencies provide little guidance as to the relative weight assigned to each of the variables. Second, the choice of the financial ratios reflects data availability for each of the companies. Hence the choice of the seven variables used, has been based on the reasons outlined above as well as following the study of Packer (2000) and the Moody s publication: RiskCalc Model for Australian Private Firms; version 1.5 (2002). Accordingly, the variables used in the model, obtained from the Osiris database are as follows: 3

5 Size: Size is a variable that is correlated with many financial statements inputs and the quality of financial statements. The proxy used for the firm size is the total assets of the firm. Profitability: The focus of most shareholders is on the profit of the company and ultimately the returns available to them. The profitability measure used in this study is the net profit margin. Leverage: The capital structure of the firm is considered to be a key indicator of the fund raising capacity of the firm. It measures the ability to withstand unforeseen circumstances. The gearing ratio has been used in the study. Interest Cover: The ability of the firm to service its debt on a timely basis out of the profit of the firm is very important in assessing its creditworthiness. Growth: It is prudent to examine statistically whether the dynamics of the firm behavior is related to future default behavior. The proxy used for growth is the standard deviation of the return on assets per annum over a 5-year period Activity: The firm s activity is measured by considering the stock turnover of the firm. Liquidity: There may be many liquidity ratios, differing simply on the weight assigned to the different types of current assets. However, the ratio used in this study is the current ratio, which is the most common measure of liquidity. The use of a large number of economic variables in the model introduces the possibility of multicollinearity. Table 1 reports the correlation matrix for the financial ratios used in this study. On average, the correlation between the financial ratios is not substantially high. Both Moody s and Standard and Poor s provide ratings to a large number of firms. However, due to data availability with regard to the financial ratios, the 4

6 study covers the long-term ratings for 30 companies rated by Moody s and the foreign currency rating for 55 companies rated by Standard and Poor s as at December Modeling Framework The modeling framework employs an ordered response model. The use of this model has been justified in the previous literature (see for example, Brooks, Faff and Sokulsky (2002), Berman and Fry (2001), Greene (2000), McKelvey and Zaviona (1975)). In order to run the model the ratings by Standard and Poor s and Moody s for 2002 are replaced by numerical equivalent grades. The study has been repeated for two categories of rating, that is rating 1-21study and rating 1-9 study. This implies that in the rating 1-21 study each of the individual alphabetical ratings are replaced by individual numerical grades and in the rating 1-9 study, the alphabetical grades are grouped by broad letter categories and then each group is assigned a numerical grade. Hence, the mapping of the ratings is as follows: Rating Grade 1-21 Rating Grade 1-9 Moody s Standard & Poors 1 1 Aaa AAA 2 2 Aa1 AA+ 3 Aa2 AA 4 Aa3 AA- 5 3 A1 A+ 6 A2 A 7 A3 A- 8 4 Baa1 BBB+ 9 Baa2 BBB 10 Baa3 BBB Ba1 BB+ 12 Ba2 BB 13 Ba3 BB B1 B+ 15 B2 B 16 B3 B Caa1 CCC+ 18 Caa2 CCC 5

7 19 Caa3 CCC Ca CC 21 9 C SD To motivate the ordered response model, consider the latent variable model, y i * = x i β + ε i (1) where y i * is an unobservable latent variable that measures the risk level, x i is a vector of explanatory variables, in this case, the financial ratios, β is a vector of unknown parameters and ε i is a random disturbance term. If the distribution of ε i is chosen to be normal, then ultimately this produces an ordered probit model. As usual, y i * is unobserved. What we assume is that y i * is related to the observed variable y i, in this case, Moody s long term ratings and Standard and Poor s foreign currency rating in the following way: y i = 0 if y i * < ε 0 1 if ε 0 < y i * ε 1 2 if ε 1 < y i * ε 2 3 if ε 2 < y i * ε 3 21 if ε 20 < y i * where the εs ( ε 0 < ε 1 < ε 2 < ε 3 <.<ε 20 ) are unknown (threshold) parameters to be estimated. The above is an explanation for the rating 1-21 study. The same methodology applies when the numerical grades are reduced to 9 categories only. 6

8 To estimate the model, the variables that constitute x i must be selected, which in this case are the financial ratios detailed in the previous section. Therefore the model to be estimated is: y i = β 1 (Size) +β 2 (Profit Margin) +β 3 (Leverage) +β 4 (Interest Cover) +β 5 (Growth) +β 6 (Stock Turover) +β 7 (Current Ratio) + ε i (2) where y i, represents the observed grades for Moody s long term rating for the 30 companies and Standard and Poor s foreign currency ratings for the 55 companies. 2.3 Results using Rating 1-21 Classification The ordered response model (equation 2) is run with both Moody s long term rating and Standard and Poor s foreign currency ratings. The second column of table 2 reports the regression of Standard and Poor s foreign currency ratings against the seven financial ratios for the 55 companies, while the third column of the table reports the results of Moody s long term ratings. Among the variables used, consistent with the study by Packer (2000), size, profitability and leverage ratios are statistically significant for both Moody s and Standard and Poor s. However, contrary to expectations, the size variable does not have to the anticipated sign. In general, the relationship between the numerical scores and size should have been negative that is as the size of the firm increases, the ratings assigned (alphabetical ratings) are upgraded. Both of the regressions indicate a positive relationship between the numerical scores and size. This indicates that as the size of the firm grows, the ratings are downgraded. However, for the profitability 7

9 and leverage measures, the results have the anticipated signs. For Moody s, in addition to these variables, interest cover and current ratio are statistically significant. Table 3 reports the predictive inaccuracy for the model for both Standard and Poor s and Moody s for the rating 1-21 classification. The expectation prediction table classifies the observation on the basis of the predicted response. The predictive inaccuracy is calculated by dividing the difference between the actual number of ratings and the predicted number of ratings by the actual number of ratings. A zero indicates that the model is predicting a perfect outcome, that is the predicted and the actual number of ratings are the same. An analysis of this table reveals that the percentage of errors is zero in only a few cases. However, the magnitude between the actual and predicted response is not too high Results using Rating 1-9 Classification Variations across a large number of categories (rating 1-21) can be at times very difficult to detect. Accordingly, the rating classifications are reduced to only nine categories and the previous analysis repeated. The results are reported in table 4. The results, on average, do not change much. The same variables with the same signs are statistically significant, that is the size, profitability and the leverage measures. In addition, for Moody s long term rating the current ratio is significant. Table 5 reports the predictive inaccuracy for the rating classification 1-9. Similar to the previous analysis, the predictive inaccuracy does not improve much. The percentage of errors is zero in a few cases only. 8

10 2.4 Forecast of Ratings The study includes 55 companies, which are rated, by Standard and Poor s and 30 companies rated by Moody s. This section presents a forecast of the 25 companies unrated by Moody s in the sample for the year 2002 and this is compared to the actual rating provided to these companies given by Standard and Poor s in It should be highlighted that this comparison is undertaken on the basis that even though the rating symbols used by the two agencies are different, they are comparable (see Cantor and Packer (1996a)). Following the results, which have been outlined in the previous section, the model is adjusted to consider the most significant variables, which in this case include size, profit margin and leverage. The forecast of the ratings is undertaken using the rating 1-9 classification and the model utilized is as follows: y i = β 1 Size +β 2 Profit Margin +β 3 Leverage + ε I (3) where y i represents the observed grades for the 25 companies rated by Standard and Poor s. The forecast of the ratings is therefore based on the size, profit margin and leverage variables for 2002 and the parameter estimates are obtained from the above equation. The score then determines the alphabetical grade to be assigned through the assumed mappingin the rating 1-9 classification. An illustration of the model is as follows. The number of rating categories obtained from the model is five that is there are no observations for the rating categories 1,7, 8 and 9. The numerical score to be assigned to each company is based on the model s limit points obtained from the model as follows: y i = 1 No Observations 2 if y i * <

11 3 if < y i * if < y i * if < y i * if y i * > No Observations 8 No observations 9 No Observations The numerical score then determines the alphabetical grade to be assigned for each company following the classification under the rating 1-9 study. The results of the forecast rating that should be assigned by Moody s are reported in table 6. The table shows the number of hit and misses by comparing the ratings obtained in this study and the actual rating assigned by Standard and Poor s. From table 6, it is clear that 68 percent of the results indicate a hit. For the remaining number of misses, however, the ratings difference is only up to two notches, which is not considered as being too substantial. 3. The Impact of Company Rating Changes on the Returns The credit quality of most issuers and their obligations is not fixed and steady over a period of time, but tends to undergo changes. The focus of this section is to assess the impact of a rating change on the stock returns. The data used for this study includes the population of rating change announcements for the period January 1996 to December 2002, available from the Osiris database, by the two leading agencies. Over the 7-years period, the number of rating changes announced by Moody s for Australian companies 10

12 comprise of 13, including 5 upgrades and 8 downgrades. In the case of Standard and Poor s, 21 rating changes were observed, that is 10 upgrades and 11 downgrades. To determine the impact of the rating changes by both Moody s and Standard and Poor s, an event study methodology is employed to detect the abnormal returns resulting from an upgrade or downgrade announcement. Daily risk adjusted returns are derived from the conventional market model: AR it = R it (α i + β i R mt ) (4) Where R it is the return on company i at day t, R mt is the corresponding return on the Australian stock market, that is the All Ordinaries Index at day t, and α i and β i are the market model parameters obtained from an ordinary least squares regressions. The market model parameters are based upon approximately six months of daily return observations beginning 120 days before through to 21 days before the sovereign rating change. The event period ranges from 10 days before to 10 days after the rating change. Abnormal return test statistics are taken from Boehmer, Musumeci and Poulsen (1991) and are estimated as follows: risk adjusted abnormal returns are first standardised to give the standardised abnormal return (SAR): 2 ( R mt R m ) 2 ( R mt R m ) 1 SAR it = AR it / σ ˆ i (5) 21 T i E =

13 where σˆ i is company i s standard deviation of the risk adjusted abnormal share price return during the estimation period; T i is the number of trading days in the estimation period of company i; and R m is the average All Ordinaries return during the estimation period. For each day in the event period, the cross-sectional standard deviation of the standardised abnormal returns is then calculated. This can be written as: ( N 1) 2 N N SARit SARit N i= 1 i= 1 σ = SAR t (6) N The standardised cross-sectional test statistic is thus: Z N SAR it N i= = 1 (7) σ SAR t The individual standardised abnormal returns are assumed to be cross-sectionally independent and distributed normally. By the Lindberg-Levy and Lindberg-Feller central limit theorems [Greene (2000)], the distribution of the sample average standardised abnormal returns will converge to normality. Table 7 reports the results of the market reaction to Standard and Poor s foreign currency re-ratings. From the table, it is clear that rating downgrades have a significant wealth effect. Two days prior and on the announcement day, there is a significant impact following the rating downgrade. However, contrary to expectation, the average abnormal 12

14 returns are positive on the announcement day. In contrast, except on the announcement day, rating upgrades do not seem to have an impact on the returns. Table 8 reports the results following Moody s rating change announcements. Contrary to the results obtained for Standard and Poor s rating, both rating upgrades and downgrades have an impact on the stock returns. For the rating downgrades, the average abnormal return are significantly negative at 1.03% (one day return) and for several days over the event window the returns are significant. For the rating upgrades, it is clear that for several days over the entire event window, the average abnormal returns are statistically significant. These results suggest that different rating agencies have different impact on the market. 4. Conclusion The creditworthiness of the firm serves the interests of institutions, borrowers and investors alike. This paper aims at analysing the determinants and impact of credit ratings of Australian firms. An ordered response model using seven financial ratios is used so as to model the determinants of the ratings that Standard and Poor s and Moody s utilise in the ordering of risks of Australian firms. The main finding of this paper is that of the quantitative variables used in the analysis, a company s rating appears to be largely determined by its size, profitability and leverage measures. This obviously suggests that the information publicly available in the financial statements do play a role in the analysis undertaken by the rating agencies. However, they are not the only drivers of the credit ratings assigned. There are other factors which play a significant role in the ratings, which include industry structure, a 13

15 company s role in the sector, business risk, the regulatory environment, capital structure and covenant protection in a company s financing structure. In fact, it is equally argued that the rating agencies often place emphasis on these qualitative factors in the rating process. Nevertheless, financial ratios are also used extensively in the rating process for comparing the financial strengths of different companies. In addition, the paper assesses the impact of rating changes by different rating agencies on stock returns. Consistent with previous research on credit rating, rating downgrades have a significant impact on the stock returns, whereas rating upgrades do not carry the same informative value. Surprisingly, an important conclusion in this study is that this finding cannot be generalised for both the rating agencies. The results suggest that different rating agencies affect the market differently. In the case of Moody s longterm ratings changes, rating upgrades do have an impact on the market. This suggest that before generalising that upgrades do not have an impact while downgrades impact in the market, we need to be cautious. The evidence from this study suggest that this can be applied in the case of Standard and Poor s rating changes only. 14

16 References Barron, M. J., Clare, A. D. and Thomas, S. H., (1997), The Effect of Bond rating Changes and New Ratings on UK Stock Returns Journal of Business Finance and Accounting 24, pp Berman, G. and Fry, T. R. L, (2001), A Charitable Ranking, Economic Papers 20, pp Boehmer, E, Musumeci, J, and Poulsen, A B, (1991), Event-Study Methodology Under Conditions of Event-Induced Variance, Journal of Financial Economics 30, pp Brooks, R. D., Faff, R. W. and Sokulsky, D., (2002), An Ordered Response Model of Test Cricket Performance, Applied Economics 34, pp Cantor, R., and Packer, F., (1996a), Determinants and Impact of Sovereign Credit Rating, Federal Reserve Bank of New York Economic Policy Review, (October), pp Cornell, B., Landsman W. and Shapiro, A. C., (1989), Cross-Sectional Regularities in the Response of Stock Prices to Bond Rating Changes, Journal of Accounting, Auditing, and Finance 4, pp Dichev, I., and Piotroski, J., (2001), The Long-run Stock returns Following Bond Rating Changes, Journal of Finance 55, pp Ederington, H. and Goh, J. C., (1993), Is a Bond Rating Downgrade Bad news. Good News, or No news for Stockholders?, Journal of Finance 48, pp Ederington, H. and Goh, J. C., (1998), Bond Rating Agencies and Stock Analysts: Who Knows What When?, Journal of Financial and Quantitative Analysis 33, pp

17 Ederington, H. and Goh, J. C., (1999), Cross-Sectional Variation in the Stock market Reaction to Bond rating Changes, Quarterly review of Economics and Finance 39, pp Greene, W.H., (2000), Econometric Analysis, Prentice Hall, 4 th Ed. Griffin, P. A. and Sanvincente, A. Z., (1982), Common Stock Returns and rating Changes: A Methodological Comparison, Journal of Finance 47, pp Holthausen, R. W. and Leftwich, R. W., (1986), The Effect of Bond Rating Changes on Common Stock Prices, Journal of Business Research 24, pp Impson C.M., Karafiath, I. and Glascock, J., (1992), Testing Beta Stationarity across Bond Rating Changes, Financial Review 27, pp Liu, P., Seyyed F. J. and Smith, S. D., (1999), The Independent Impact of Credit rating Changes The Case of Moody s rating Refinement on Yields Premiums, Journal of Business Finance and Accounting 26, pp Matclosy, Z. P. and Lianto, T., (1995), The Incremental Information Content of Bond rating Revisions: The Australian Evidence, Journal of Banking and Finance 19, pp McKelvey, R. C. and Zaviona, W., (1975), A Statistical Model for the Analysis of Ordinal Level Dependent Variables, Journal of Mathematical Sociology 4, pp Moody s Publication. (2002)., Moody s RiskCalc TM Firms:version 1.5 Model for Australian Private Packer, F., (2000), Credit Ratings and the Japanese Corporate Bond Market, IMF Discussion paper No.2000-E-9 Pinches, G. E. and Singleton, J.C.,(1978), The Adjustment of Stock Prices to Bond Rating Changes., Journal of Finance, Vol. 33, pp Pinches, G. E. and Singleton, J.C.,(1978), The Adjustment of Stock Prices to Bond Rating Changes., Journal of Finance, Vol. 33, pp Wansley, J. W., Glascock, J. L. and Clauretie, T. M., (1992), Institutional Bond Pricing and Information Arrival: The Case of Bond Rating Changes, Journal of Business Finance and Accounting 19, pp Weinstein, M.I., (1977), The Effect of a Rating Change Announcement on Bond Price, Journal of Financial Economics, Vol 5, oo

18 Zaima, J. K. and McCarthy, J., (1988), The Impact of Bond Rating Changes on Common Stocks and Bonds: Tests of the Wealth Redistribution Hypothesis, Financial Review 23. pp Table 1: Correlation matrix of the Independent Financial Ratios 2002 Panel A: Moody s -30 Companies Current ratio Gearing Int. Cover Profit Margin Growth S.Turnover Size Current ratio 1 Gearing Int. Cover Profit Margin Growth S.Turnover Size Panel B: Standard and Poor s - 55 companies Current ratio Gearing Int. Cover Profit Margin Growth S.Turnover Size Current ratio 1 Gearing Int. Cover Profit Margin Growth S.Turnover Size

19 Table 2: Rating 1-21 Determinants Results This table shows the relative significance of the financial ratio applied to both Standard and Poor s and Moody s in the rating classification Values in parenthesis are p- values. S&P Foreign Currency Rating- 55 companies Moody s Long Term Rating- 30 companies Size(Total Assets) (0.0111) (0.0219) Profit margin (0.0021) (0.0004) Leverage(Gearing) (0.0306) (0.0347) Interest Cover (0.4455) (0.0835) Growth (Stdev ROA) (0.4584) (0.1337) Stock Turnover (0.4591) (0.2532) Current Ratio (0.2410) (0.0170) Pseudo R 2 13% 32% 18

20 Table 3: Predictive Inaccuracy Table under Rating 1-21 Classification This table shows the predictive inaccuracy for Standard and Poor s and Moody s in the rating 1-21 study. Standard and Poor s Foreign Currency Rating Moody s Long Term Rating Grades Predicted No. of rating Error % Predicted No. of rating Error % Actual No of Rating Actual No of Rating

21 Table 4: Rating 1-9 Determinants Results This table shows the relative significance of the financial ratio applied to both Standard and Poor s and Moody s in the rating classification 1-9. Values in parenthesis are p- values. S&P Foreign Currency Rating- 55 companies Moody s Long Term Rating- 30 companies Size(Total Assets) (0.0104) (0.0822) Profit margin (0.0112) (0.0037) Leverage(Gearing) (0.0262) (0.0720) Interest Cover (0.7308) (0.1729) Growth (Stdev ROA) (0.3715) (0.1499) Stock Turnover (0.3743) (0.2338) Current Ratio (0.2313) (0.0449) Pseudo R 2 19% 41% 20

22 Table 5: Predictive Inaccuracy Table under Rating 1-9 Classification This table shows the predictive inaccuracy for Standard and Poor s and Moody s in the rating 1-9 study. Standard and Poor s ForeignCurrency Rating Moody s Long Term Rating Grades Actual No of Rating Predicted No. of rating Error % Actual No of Rating Predicted No. of rating Error % Table 6: Hit and Miss Table This table shows the forecast rating of the 25 companies not rated by Moody s using equation 3 for the year A rating notch is a one-level difference between the forecast rating and the actual rating. No of Companies 21

23 Hit 17 Miss: 1 Notch 6 2 Notches 2 Total 25 Table 7: Standard and Poor s Event Study This table reports average abnormal returns (AAR) and cumulative abnormal returns (CAR) as measures of the stock returns reaction to Standard and Poor s ratings. AAR and CAR are generated using a standard mean adjusted event study methodology. Note that standard errors are estimated using standardized abnormal returns (SARs) but only average abnormal returns are reported. S&P Upgrade n=10 S&P Downgrade n=11 Event Day AAR t- stat CAR t-stat AAR t- stat CAR t-stat ** ** ** * *

24 ** * Denotes statistical significance at 10% level ** Denotes statistical significance at 5% level Table 8: Moody s Event Study This table reports average abnormal returns (AAR) and cumulative abnormal returns (CAR) as measures of the stock returns reaction to Moody s ratings. AAR and CAR are generated using a standard mean adjusted event study methodology. Note that standard errors are estimated using standardized abnormal returns (SARs) but only average abnormal returns are reported. Moody s Upgrade n=5 Moody s Downgrade n=8 Event Day AAR t- stat CAR t-stat AAR t- stat CAR t-stat ** ** * ** * * ** * * * *

25 ** * * * ** ** * Denotes statistical significance at 10% level ** Denotes statistical significance at 5% level 24