6. Conclusion
3.2. Proxy for Bank Quality
4.2.2. Multivariate regression 1
Multivariate regressions are used to determine whether the control variables offer any ability to explain the CAAR for the bank loan rating announcements. In these regressions the dependent variable is the three-day (day t- 1 to day t+ 1 ) CAAR. The regression models for the groups of placement with negative implication and downgrade are given in Equations ( 1 ) and (2), respectively.
CAAR
=
aO + a 1 BKSIZE + a2BKTAG + a3PROF + a4COL + a5�RATP + E. (2)where:
BKSIZE BKTAG
PROF
natural log of lead lender's total assets;
growth rate in total assets of the lead lender, calculated by BKSIZEtO BKSIZEt- 1 IBKSIZEt- 1 , where to indicates the financial year prior to the announcement date;
earnings before interest, tax, depreciation and amortisation divided by total assets;
COL sum of property, plant and equipment and inventory divided by total
INVG
�RATP
assets;
a dummy variable, equal to one if the rating is below investment grade, and zero otherwise; and
number of changes in rating points.
A bank's size, calculated as the natural log of the total assets, is a common variable used as the proxy of a bank's reputation (Stover, 1 996; Johnson, 1 997b; Byers et al. , 1 998; Hubbard et al., 2002; Brewer et al., 2003). In addition; unlike the return on assets or the capital ratio, which are also used as proxies for reputation; this variable is much more readily available, hence reducing concern of a reduced sample size. It is hypothesized that bank size should have a positive effect on the borrower's equity returns, therefore, in the regression it should be significant and have a positive sign.
A bank's growth rate in total assets is used as the proxy for creditworthiness (Huffrnan and Ward, 1 996).9 It is hypothesized that the higher the variation, the riskier the banks, and the more negative the impact on the borrower's equity returns. Therefore, this variable should have a significant and negative sign in the regression.
Other variables are included in the models to control the characteristics of firms and ratings. Profitability ratios (PROF) show the combined effects of liquidity, asset and debt management. Investors are theorised to be more sensitive to those companies with lower profitability. The profitability measurement used in this essay is the ratio of earnings before interest, tax, depreciation and amortisation (EBITDA) to total assets (Titman and Wessels, 1 988). This is the key accounting measurement used by the rating agencies to estimate enterprise value. As suggested by Moody's Staff ( 1 996), EBITDA is the core level of cash flow, regardless of industry or macroeconomic conditions. Therefore, the ratio of EBITDA to total assets is used to measure profitability in this essay. 1 0 It is expected that bad news will have a stronger impact on low profitability firms, so this variable should have a positive sign.
The ratio of tangible assets to total assets is used to proxy information asymmetries or growth opportunity. The collateralised assets (COL); calculated by the sum of property,
9 The previous studies usually use a bank's debt ratings as the measurement of risk. This rating variable is
not used in this essay because of three reasons, besides the concern of sample size. First, as suggested by Gibson ( 1 995), this variable may force all banks with a given rating to have the same effect. Second, because there is little fluctuation in credit ratings within any particular bank, it is more difficult to catch the variance through the sample period. Third, it was more appropriate to use credit rating in previous studies, but not in the current one, as former studies usually focused on events of only one point oftime; for example, the initial public offering of shares, or a bank l oan issuance. In this essay, rating announcements
are more like continuous events, which highlights the problem of the second reason just stated. Therefore, the bank's growth rate in total assets is used here, instead of its credit rating.
\ 0 On the other hand, Fridson ( 1998) and King (200 1 ) argued that the sole use of EBITDA as the measure of value is doubtful . Therefore, the traditional basic earning power (BEP); calculated as EBIT divided by total assets; is also employed as the alternative measurement for profitability. For the group of negative placement, neither BEP nor the alternative model is significant. For the group of downgrade, BEP is still significant at the 5% level, and the alternative model is significant at the 1 0% level.
plant and equipment and inventory; is used as a proxy for intangible assets (Dittmar, 2004). Comell, Landsman and Shapiro ( 1 992) found that the informational value provided by rating actions is correlated to a firm's intangible assets. They included the product of magnitude of the change in ratings and the intangible assets ratio (measured from the tangible assets at current costs) and found that it was significant at the 5% level, making the F-value significant also. ! ! Hsueh and Liu ( 1 9 92) found that rating change announcements had a stronger impact on firms which had less information available to the market. Krishnaswami, Spindt and Subramaniam (1 999) found that the benefits from monitoring accrued more to firms with greater information asymmetry. If the COL ratio
(as the proxy of tangible assets) is used as the proxy of information asymmetries1 2, any results higher the COL are taken to indicate lesser information asymmetry and, hence, are expected to have a positive sign in the regression.
For rating characteristics, investment grade (INVG) is used for the sample of negative placement. The lowest rating to qualify as investment grade (INVG) is BBB- (Standard and Poor's; Baa3 by Moody's). Many institutional investors; pension funds for example; are restricted to investing in above-investment-grade securities. If investment grade is viewed as a benchmark of bank loan credit quality, it can be expected that any
creditworthiness announcement effects will be stronger for firms with below-investment- grade, or speCUlative, ratings (see, for example Hand, Holthausen and Laftwich, 1 992; Dichev and Piotroski, 200 1 ). In the regressions, a dummy variable is set to one if the rating is below-investment, and is zero otherwise. The below-INVG group is expected to
1 1
The F-value of regression models in most of the rating literature is insignificant. Although the product of the number of grades changed (""RA TP) and the intangible asset ratio in the regression model of Cornell et al. (1992) increase the significance of the F-value, the variable of the product is excluded in this essay because it is highly correlated with ""RA TP.
1 2 Tobin's Q; calculated by the sum of the total liabilities and the market value of equity divided by the total
assets; and size are also employed as the alternative proxy for information asymmetry, but the results are not qualitatively different.
be associated with a stronger market reaction and, therefore, should have a negative sign in the regression model for the sample of negative placement.
The number of grade changes in ratings (llRATP) is used for the sample of downgrade. Based on Holthausen and Leftwich ( 1 986), a number is assigned to each rating; from 2 8 assigned to the highest rating of AAA, to I assigned to the lowest rating of D. The midpoint of these numbers is assigned to the class without the plus or minus modification (AAA, CC, C and D). The llRATP is calculated by subtracting the new rating from the old one. This is, therefore, positive, and measures the magnitude of the bad news. This is the most common variable used in the rating literature for the multivariate regression methodology. Nevertheless, while the expected sign is almost always correct in all studies; which indicates that llRA TP has a positive relationship to the market reaction; its significance is not consistent. Therefore, this variable is expected to have a negative sign in the regression for the sample of downgrade, but the significance may be uncertain.
5. EMPIRICAL RESULTS
This section covers two parts of the empirical results for the sample of bank loan rating placement announcements on CreditWatch with negative implication, and downgrade
announcements. The first part presents the results of the event study of the univariate tests based on different hypothesized variables. The second part is the results of the cross-sectional regression.
5.1 Univariate Tests of Returns based on Hypothesized Variables
The event study results of the univariate tests for negative placement are shown in Table 4, based on different variables. The three-day and two-day CAAR, CMAR, the number of positive versus negative abnormal returns, SCS Z and GSIGN Z are reported.
For the variable of bank size, the three-day CAAR for the group with smaller (larger) lending banks is -5.78% (-0. 00%), and the difference between these two groups is at 0. 1 % level. These results support Hypothesis l a, that larger lenders are more effective in monitoring borrowers, and mitigate the negative impact on borrowers. This finding is consistent with DeAngelo ( 1 98 1 ) and Johnson ( l 997b).
For the variable of the bank's growth rate in total assets, the group with high-growth lenders has unexpectedly less negative returns (three-day CAAR of -2.28%) than the group with low-growth lenders (three-day CAAR of -4.02%). The difference is, however, not significant.
Table 4
Equity Market Reaction of Hypothesized Variables to Bank Loan Rating Announcements of CreditWalcll Placement with Negative Implication
N is the number of returns for a given category. CAAR3 and CAAR2 are the three-day (t- 1 , t+ I ) and two-day (I- I , to) cumulative average abnormal return, with mean and median (MED) provided. P:N shows how many of the returns are positive or negative on a given day. SCS Z is the statistical test for a significant difference in the average abnormal return from zero. GSIGN Z is the generalised sign Z, which is the non-parametric test statistic for a significant difference from zero, considering the ratio of positive to negative returns. Bank ' s size is the log of the lead lending bank's total assets, bank's growth rate is the percentage change in total assets, profitability is the return on assets calculated as net income divided by total assets and capital ratio is the risk-adjusted capital ratio extracted from Compustat. The bank's credit rating (according to Moody's) is high (Iow) if it is Aa3 and above (below Aa3). Difference is the above median CAAR minus the below median CAAR, unless specified otherwise.
CAAR3 CAAR2
N MEAN MED P:N SCS Z GSIGN Z MEAN MED P:N SCS Z GSIGN Z
All sample 32 -2.79 -2.40 8:24 -2.974** -2. 7 1 4** -3.41 -3.29 7:25 -2.976** -3.067*'"
Bank's Size
Above median 1 3 0.00 -0.5 1 6:7 -0.009 -0. 1 59 -0.40 -0. 9 1 5:8 -0.369 -0. 7 1 4
Below median 1 6 -5.78 -3.87 1 : 1 5 -3.726*** -3.520*** -4.21 -2.86 3 : 1 3 -3.562*** -2.520*
Difference -3 5.78 4.494*** 3.73 2.885**
Bank's Growth Rate
Above median 1 4 -2.28 -2. 1 5 3 : 1 1 - 1 .654$ - 1 .924$ - 1 .50 -0.99 6:8 -0.722 -0.3 1 8 Below median 1 5 -4.02 -3.76 4: 1 1 - 1 .966* - 1 .929$ -3.50 -3.5 1 2 : 1 3 -2.93 3 * * -2.962** Difference -1 1 .74 1 .349 2.00 1 .55 1 Bank's Above median 1 3 -0.66 - 1 .93 5:8 -0.799 -0.773 - 1 .36 -2.29 2 : 1 1 - 1 .322 -2.437* Below median 1 3 -4. 1 2 -3.97 2: 1 1 -2.630** -2.361 * -3.03 -3.20 6:7 - 1 .802$ -0. 1 4 1 Difference -- 3 .46 2.565* 1 .67 1 .238 Bank's Ratio Above median 1 1 - 1 .55 - 1 .25 5:6 -0.554 -0.249 -2.80 -2.29 2:9 -2.874** -2.059* Below median 1 2 -2.88 -4.97 2 : 1 0 - 1 .996* -2.2 1 8* -0.71 - 1 .80 5 : 7 -0.559 -0.485 Difference - I 1 .33 0.968 -2.09 - 1 .52 1 Bank's Credit High 1 3 -2.32 - 1 .25 4:9 - 1 .208 - 1 .276 -2.00 - 1 .69 4:9 - 1 .600 - 1 .276 Low 1 7 -3.90 -3.76 3 : 14 -3.292*** -2.652** -2.82 -2.52 4 : 1 3 -2.005* -2. 1 67* Difference -4 1 .58 1 .205 0.82 0.626 US vs. Banks US 23 -2.24 -2.37 7 : 1 6 - 1 . 9 1 0$ - 1 . 774$ - 1 .7 1 -2.29 7: 1 6 - I . 740$ - 1 .774$ Foreign 7 -6.4 1 -3.76 0:7 - 1 .864$ -2.656** -4.96 - 1 .27 1 :6 - 1 .976* - 1 .900$ Difference 1 6 4. 1 7 2.735 * 3.25 2. 1 32*
For the variable of bank profitability, the group with highly-profitable banks are
associated with much less negative abnormal returns, as expected (three-day CAAR of - 0.66%). Meanwhile the group with less-profitable banks has a three-day CAAR of - 4. 1 2%. The difference between these two groups is significant at the 5% level, which supports Hypothesis 3a, that borrowing from highly profitable lenders mitigates the negative reaction on negative placements of bank loan ratings. This finding is consistent with Stover ( 1 996) and Brewer et al. (2003).
For the variable of the bank risk-adjusted capital ratio, the three-day CAAR for the group with highly- (less-) capitalised lenders is - 1 .55% (-2. 8 8%), as expected. The difference is, however, not significant. Similarly, the group with higher-rated lenders has less negative abnormal returns than the group with lower-rated banks (-2.32% versus -3 .90%), however this difference is not significant either.
Finally, the results in the bottom panel of Table 4 show that the group with a foreign bank as the lead lender has much more negative abnormal returns (three-day CAAR of -6.4 1 %) than does the group with a US lead lending bank (-2.24%). The difference is significant at the 5% level. This supports Hypothesis 6a, that investors find the monitoring done by foreign banks is not as effective/efficient as that of US banks.
The event study results of the univariate tests for downgrade are shown in Table 5, based on the different variables. For the variable of bank size, the group with the larger (smaller) lead lending bank has -2. 3 8% (-4.29%) three-day CAAR. The difference is significant at the 5% level, which supports Hypothesis I b, that larger banks have a positive effect on borrowers.
For the variable of the bank's growth rate in total assets, the group with higher- (lower-) growth lenders has a three-day CAAR of -4. 1 3% (-2. 57%). The difference is significant at the 1 0% level, supporting Hypothesis 2b, that safer banks mitigate the negative reaction of deteriorated borrowers. This finding is consistent with Slovin et al. ( 1 993).
For the variable of bank profitability, the group with highly- (less-) profitable banks has a -2. 1 4% (-4. 1 5%) three-day CAAR. The difference is significant at the 5% level,
supporting Hypothesis 3b that investors consider the monitoring performed by profitable banks is more reliable.
For the variable of bank risk-adjusted capital ratio, as expected, the three-day CAAR for the group with highly-capitalised lenders is less negative than that with less-capitalised lenders (-2.77% versus -3 .56%), however, the difference is not significant. Similarly, the group with higher-rated lenders has less negative abnormal returns than that with lower rated lenders (-3.06% versus -3.20%), however, the difference is, again, not significant. The last panel of Table 5 shows that, whether bank loans are granted by a US or a foreign lead lending bank, there is no significant difference in returns.
In summary, these results support the hypotheses that larger, less-risky and more profitable banks are regarded by the equity market as reputable monitors and, therefore, impart positive effect to borrowers, mitigating the negative effect caused by downside rating news. This indicates investors perceive informational value from banks is different than from rating agencies, specifically from high quality banks. As hypothesized, the
Table S
Equity Market Reaction of Hypothesized Variables to Bank Loan Rating Announcements of Downgrade
N is the number of returns for a given category. CAAR3 and CAAR2 are the three-day (t- I , t+ l ) and two-day (t- I , to) cumulative average abnormal return, with mean and median (MED) provided. P:N shows how many of the returns are positive or negative on a given day. SCS Z is the statistical test for a significant difference in average abnormal return from zero. GSIGN Z is the generalised sign Z, which is the non-parametric lest statistic for a significant difference from zero, considering the ratio of positive to negative returns. Bank's size is the log of the lead lending bank's total assets, bank's growth rate is the percentage change in total assets, profitability is the return on assets calculated as net income divided by total assets and capital ratio is the risk-adjusted capital ratio extracted from Compustat. The bank's credit rating (according to Moody's) is high (Iow) if it is Aa3 and above (below Aa3). Difference is the above median CAAR minus the below median CAAR, unless specified otherwise.
CAAR3 CAARl
N MEAN MED P:N SCS Z GSIGN Z MEAN MED P:N SCS Z GSIGN Z
All sample 1 2 1 -3.07 -2.4 1 43:78 -5.224*** -3.022** -2.33 - 1 .57 43:78 -4.291 *** -3.022**
Bank's Size
Above median 50 -2.38 -3. 1 8 1 7:33 -2.895** -2.350* - 1 .98 -2. 1 9 1 8 :3 2 -2.58 1 ** -2.067* Below median 5 1 -4.29 - 1 .98 1 7:34 -4. 1 63*** -2. 1 6 1 * -3.41 - 1 .58 1 7:34 -3.400*** -2. 1 6 1 *
Difference - I 1 .9 1 2.222* 1 .43 1 .664$
Bank's Growth Rate
Above median 50 -4. 1 3 -3. 1 8 1 2:38 -5. 1 36*** -3.653*** -3.65 -3.06 1 0 :40 -4. 9 1 4*** -4.2 1 8*** Below median 5 1 -2.57 - 1 .79 22:29 -2.289* -0.870 - 1 .77 -0. 1 2 25:26 - 1 .528 -0.030 Difference - I -1 .56 - 1 .805$ - 1 .88 -2. 1 76* Bank's Above median 42 -2. 1 4 -0.70 1 8:24 -2.492* - 1 .0 1 0 -2.50 - 1 .48 1 7:25 -2.5 1 2* - 1 . 3 1 9 Below median 44 -4. 1 5 -3. 9 1 1 2:32 -3.985*** -2.907** -3.46 -2.5 1 1 2:32 -3.774*** -2.907** Difference -2 2.01 2.245* 0.96 1 .072 Bank's Ratio Above median 43 -2.77 -3.46 1 7:26 -2.436* - 1 .263 -2.88 - 1 .78 1 7:26 -2.698** - 1 .263 Below median 43 -3.56 -2.45 1 3:30 -4. 1 69*** -2.676** -3. 1 0 -2.62 1 2:3 1 -3.368*** -2.98 1 ** Difference -- 0.79 0.88 1 0.22 0.245 Bank's Credit High 70 -3.06 -2.93 22:48 -4.640*** -3. 1 0 1 ** -2.08 - 1 .89 24:46 -3.446*** -2.623** Low 3 7 -3.20 - 1 .59 1 6:2 1 -2. 1 79* -0.629 -3. 1 8 - 1 .57 1 5 :22 -2.204* -0.958 Difference 33 0. 14 0 . 1 59 1 . 1 0 1 .249 US vs. Banks US 89 -3 . 1 7 -3 . 1 3 3 1 :58 -4.646*** -2.828** -2.95 -2.39 30:59 -4.353*** -3.040** Foreign 20 -3. 1 3 -2.05 7: 1 3 - 1 .805$ - 1 . 1 02 -0.41 -0.38 9 : 1 1 -0.34 1 -0.206 Difference 69 -0.04 -0.037 -2.54 -2.362*
high capital ratio, high ratings and the identity of US banks also have positive effects, but these are not significant.