Discussion of Results

The descriptive statistics of the sample, the comparison between the solicited and unsolicited groups, and the results from the treatment effects model using Wooldridge's instrumental variable method are discussed in this section.

4.1 Descriptive Statistics

Table 2 [ PDF 19.1KB | 1 page ] provides a brief description of LTRs ranging from AAA to D. The table also lists the sample frequency and sample percentage of 3,392 observations across the nine rating levels and by solicited and unsolicited rating subgroups. Only 2.2% of the sample ratings (all from the solicited group) received AAA ratings, which indicates that these firms are “extremely strong” according to S&P's definition. About 70% of the sample firms obtained a rating of BBB or above (labeled as investment-grade ratings hereafter), while about 30% of the sample obtained a rating of BB or below (labeled as speculative-grade ratings hereafter). In the overall sample, the majority of the issuers (82.2%) had solicited ratings from S&P, and 17.8% had unsolicited ratings. Interestingly, while most of the solicited ratings (2,049 out of 2,789 or 73%) are investment grade, only about half of the unsolicited ratings (318 out of 603) are investment grade.

Table 3 [ PDF 40KB | 1 page ] and Table 4 [ PDF 16KB | 1 page ] illustrate the distribution of 3,392 sample ratings from 53 countries during the study period from 1998 to 2003 sorted by country and by year, respectively. Japan had the highest number of ratings in the overall sample (23.5% of the overall sample) and all the unsolicited ratings of the sample. Canada had the highest number of solicited ratings (18.3% of the solicited subsample and 15.1% of the overall sample). Table 4 [ PDF 16KB | 1 page ] shows that there were increasing numbers of both solicited and unsolicited ratings during the study period. About 22% of the sample ratings are from 2003.

4.2 Comparison of Solicited and Unsolicited Ratings

The Mann-Whitney U-test and t-test results are reported in Table 5 [ PDF 16.9KB | 1 page ]. Panels A and B in Table 5 present the results of the overall sample with 3,392 ratings (Panel A) and the Japanese subsample with 798 ratings (Panel B). Japan was the only country with both solicited and unsolicited ratings during the study period.

The null hypothesis that unsolicited and solicited bank ratings have identical distributions (H1) can be rejected at the 1% level for both panels in Table 5. The mean rankings in both tables indicate that unsolicited ratings, on average, are lower than solicited ratings for the overall sample and for the Japanese subsample.

4.3 Profiles of Firms

Panels A and B of Table 6 [ PDF 52.2KB | 2 pages ] present the descriptive statistics and the t-test and Z-test results of the two groups.⁵ Panel A shows the results of the overall sample while Panel B shows the results of the Japanese subsample. First, referring to the t-test and U-test results of the profitability ratios (see ratios ROC, OM, and ROA), the solicited group had significantly higher profitability and earning power than the unsolicited group. Second, all debt ratios in the capital/debt structure group were significantly higher for the firms with unsolicited ratings. This suggests that unsolicited issuers had significantly higher leverage in terms of debt to

Third, the cash flow protection ratios (FFOTD and NCFCAPEX) show that the firms in the solicited group had more funds from operation (FFO) and net cash flow (NCF) compared to their total debt and capital expenditure than those in the unsolicited group in the overall sample. On the other hand, the results of the Japanese subsample indicate that the Japanese firms with unsolicited ratings had more NCF to payments to fixed assets than the Japanese firms with solicited ratings.

Lastly, the comparisons of the three liquidity ratios (CR, QUICK, and CASH) illustrate that the issuers with solicited ratings were more liquid than the issuers with unsolicited ratings. In addition, the SDTD ratio in the financial flexibility group indicates that the firms in the unsolicited group were tied up with more short-term debts compared to their total debts in both the overall sample and the Japanese subsample. However, the results of CASHEQ, TA, and SALE were mixed. The unsolicited group in the overall sample was significantly larger in asset size and sales revenues and had more cash than the solicited group, while the solicited group in the Japanese subsample was larger and had more cash than the unsolicited group.

In sum, the results indicate that those issuers with solicited ratings were more profitable and more liquid, and had lower leverage, than the issuers with unsolicited ratings. These results suggest that the solicited group had stronger financial profiles in terms of profitability, liquidity, and debt structure than the unsolicited group. Therefore, the null hypothesis of H2, that there is no difference in the financial profiles between the issuers with solicited and unsolicited ratings, can be rejected for the sample in this study.

4.4 Results of Two-Step Treatment Effects Model

We present the rating decision equation in Panel A of Table 7 [ PDF 52.6KB | 2 pages ]. Because there are disproportionately more Japanese companies in the sample, we used a dummy variable to control for the dominance of Japanese companies. We also separately estimated a Japan-only subsample. The findings suggest that firm size (natural logarithm of total assets) and profitability (ROA) are positively related to the probability of seeking a credit rating while market-to-book ratio (MTB) and debt ratio (DTA) are negatively related to the probability. The signs of these variables are consistent with the literature. Both the full sample and the Japanese subsample offer similar results.

The findings of the determinants of credit ratings are presented in Panel B of Table 7. Both the full sample and the Japan subsample offer similar findings. The firm size (TA), sovereign risk (SOV), and profitability (ROA) are positively significant at the 1% level. The results suggest that a larger and more profitable firm located in a country with a higher sovereign rating would get a higher LTR. The debt ratio is negatively significant at the 1% level. Other things being equal, a firm with higher leverage would have a lower LTR. The fit probability variable is positive and significant at the 1% level. Hence, there is indeed a sample-selection bias (i.e., a statistically significant Y_HAT variable). Therefore, we were able to reject H3 (that corporate credit ratings do not reflect a sample-selection bias). In addition, a positively significant Y_HAT variable suggests that unsolicited firms exhibit a lower rating after controlling for other financial characteristics and sample-selection bias. There are a total of nine industries in our sample and we used the technology industry as the basis for comparison. Many industry dummy variables are significant, which suggests that there are industry effects on credit ratings.

Download this Paper [ PDF 293.6KB| 26 pages ].

Comment(s)

There are [0] comment(s) for this entry. Post a comment.