<<

. 121
( 208 .)



>>


Table 14-1 Debt Ratings in December 1998: Example Firms and Median
Financial Ratios by Category
Median ratios for overall category in 1998
(industrials only)
...............................................................................
Percentage
of public Pretax Cash ¬‚ow Long-term
S&P industrials return on Pretax from debt to
debt given same long-term interest operations total
rating Example ¬rms in 1998 rating by S&P capital coverage to total debt capital
...........................................................................................................................................................

AAA General Electric
Johnson & Johnson
Merck and Co. 1.9% 35.3% 11.6 times 100.1% 9.7%
AA McDonald's Corp.
J. P Morgan
.
Wal-Mart Stores, Inc. 7.0% 25.0% 7.2 times 59.8% 29.4%
A Ford Motor Company
General Motors
Sears Roebuck & Co. 21.8% 16.6% 4.8 times 34/3% 39.0%
BBB Delta Airlines
MCI Communications 28.2% 12.6% 3.0 times 24.8% 45.0%
BB Northwest Airlines
RJR Nabisco 21.0% 11.1% 1.9 times 11.1% 59.5%
B Apple Computer
Greyhound Lines
Loehmanns 18.3% 7.4% 0.7 times 3.1% 78.4%
CCC Oxford Health Plans
Trans World Airlines 1.7% “5.3% “1.7 times “17.3% 80.8%
...........................................................................................................................................................
Source: Standard and Poor™s Compustat, 1998.




Factors That Drive Debt Ratings
Research demonstrates that some of the variation in debt ratings can be explained as a
function of selected ¬nancial statement ratios, even as used within a quantitative model
that incorporates no subjective human judgment. Some debt rating agencies rely heavily
on quantitative models, and such models are commonly used by insurance companies,
banks, and others to assist in the evaluation of the riskiness of debt issues for which a
public rating is not available.
Table 14-2 lists the factors used by three different ¬rms in their quantitative debt-rat-
ing models. The ¬rms include one insurance company and one bank, which use the mod-
els in their private placement activities, and an investment research ¬rm, which employs
565
Credit Analysis and Distress Prediction




14-13 Part 3 Business Analysis and Valuation Applications




Table 14-2 Factors Used in Quantitative Models of Debt Ratings

Firm 1 Firm 2 Firm 3
...........................................................................................................................................................
Pro¬tability measures Return on long-term Return on long-term Return on long-term
capital capital capital
Leverage measures Long-term debt to Long-term debt to Long-term debt to
capitalization capitalization capitalization
Total debt to total capital
Pro¬tability and leverage Interest coverage Interest coverage Fixed charge coverage
Cash ¬‚ow to long-term Cash ¬‚ow to long-term Coverage of short-term
debt debt debt and ¬xed charges
Firm size Sales Total assets
Other Standard deviation of
return
Subordination status
...........................................................................................................................................................


the model in evaluating its own debt purchases and holdings. In each case, pro¬tability
and leverage play an important role in the rating. One ¬rm also uses ¬rm size as an in-
dicator, with larger size associated with higher ratings.
Several researchers have estimated quantitative models used for debt ratings. Two of
these models, developed by Kaplan and Urwitz and shown in Table 14-3, highlight the
relative importance of the factors.3 Model 1 has the greater ability to explain variation
in bond ratings. However, it includes some factors based on stock market data, which
are not available for all firms. Model 2 is based solely on financial statement data.
The factors in Table 14-3 are listed in the order of their statistical signi¬cance in
Model 1. An interesting feature is that the most important factor explaining debt ratings
is not a ¬nancial ratio at all”it is simply ¬rm size! Large ¬rms tend to get better ratings
than small ¬rms. Whether the debt is subordinated or unsubordinated is next most im-
portant, followed by a leverage indicator. Pro¬tability appears less important, but in part
that re¬‚ects the presence in the model of multiple factors (ROA and interest coverage)
that capture pro¬tability. It is only the explanatory power that is unique to a given vari-
able that is indicated by the ranking in Table 14-3. Explanatory power common to the
two variables is not considered.
When applied to a sample of bonds that were not used in the estimation process, the
Kaplan-Urwitz model (1) predicted the rating category correctly in 44 of 64 cases, or 63
percent of the time. Where it erred, the model was never off by more than one category,
and in about half of those cases its prediction was more consistent with the market yield
on the debt than was the actual debt rating. The discrepancies between actual ratings and
those estimated using the Kaplan-Urwitz model indicate that rating agencies incorporate
566 Credit Analysis and Distress Prediction




14-14
Credit Analysis and Distress Prediction




Table 14-3 Kaplan-Urwitz Models of Debt Ratings

Coef¬cients
...............................
Firm or debt characteristic Variable re¬‚ecting characteristic Model 1 Model 2
..........................................................................................................................................................
Model intercept 5.67 4.41
Total assetsa
Firm size .0010 .0012
Subordination status of debt 1 = subordinated; 0 = unsubordinated “2.36 “2.56
Leverage Long-term debt to total assets “2.85 “2.72
Systematic risk Market model beta, indicating sensitivity of stock
price to market-wide movements (1 = average)b “.87 NA
Pro¬tability Net income to total assets 5.13 6.40
Unsystematic risk Standard deviation of residual from market model
(average = .10)b “2.90 NA
Riskiness of pro¬t stream Coef¬cient of variation in net income over 5 years
(standard deviation/mean) NA “.53
Interest coverage Pretax funds ¬‚ow before interest to interest expense .007 .006
The score from the model is converted to a bond rating as follows:
If score > 6.76, predict AAA
If score > 5.19, predict AA
If score > 3.28, predict A
If score > 1.57, predict BBB
If score < 0.00, predict BB
..........................................................................................................................................................
a. The coefficient in the Kaplan-Urwitz model was estimated at .005 (Model 1) and .006 (Model 2). Its scale has been
adjusted to reflect that the estimates were based on assets measured in dollars from the 1960s and 1970s. Given that $1
from 1970 is approximately equivalent to $5 in 1995, the original coefficient estimate has been divided by 5.
b. Market model is estimated by regressing stock returns against the return on the market index, using monthly data for
prior five years.




factors other than ¬nancial ratios in their analysis. These are likely to include the types
of strategic, accounting, and prospective analyses discussed throughout this book.
Given that debt ratings can be explained reasonably well in terms of a handful of ¬-
nancial ratios, one might question whether ratings convey any news to investors”any-
thing that could not already have been garnered from publicly available ¬nancial data.
The answer to the question is yes, at least in the case of debt rating downgrades. That is,
downgrades are greeted with drops in both bond and stock prices.4 To be sure, the capital
markets anticipate much of the information re¬‚ected in rating changes. However, that
is not surprising, given that the changes often represent reactions to recent known
events, and that the rating agencies typically indicate in advance that a change is being
considered.
567
Credit Analysis and Distress Prediction




14-15 Part 3 Business Analysis and Valuation Applications




PREDICTION OF DISTRESS AND TURNAROUND
The key task in credit analysis is assessing the probability that a ¬rm will face ¬nancial
distress and fail to repay a loan. A related analysis, relevant once a ¬rm begins to face
distress, involves considering whether it can be turned around. In this section, we con-
sider evidence on the predictability of these states.
The prediction of either distress or turnaround is a complex, dif¬cult, and subjective
task that involves all of the steps of analysis discussed throughout this book: business
strategy analysis, accounting analysis, ¬nancial analysis, and prospective analysis.
Purely quantitative models of the process can rarely serve as substitutes for the hard
work the analysis involves. However, research on such models does offer some insight
into which ¬nancial indicators are most useful in the task. Moreover, there are some set-
tings where extensive credit checks are too costly to justify, and where quantitative dis-
tress prediction models are useful. For example, the commercially available “Zeta”
model is used by some manufacturers and other ¬rms to assess the credit-worthiness of
their customers.5
Several distress prediction models have been developed over the years.6 They are
similar to the debt rating models, but instead of predicting ratings, they predict whether
a ¬rm will face some state of distress within one year, typically de¬ned as bankruptcy.
One study suggests that the factors most useful (on a stand-alone basis) in predicting
bankruptcy one year in advance are7:
Net income
---------------------------
1. Profitability =
Net worth

Standard deviation of « --------------------------- 
Net income
2. Volatility =
 Net worth 

Market value of equity
-------------------------------------------------------------------------------------------------------------------
-
3. Financial leverage =
( Market value of equity + Book value of debt )
The evidence indicates that the key to whether a ¬rm will face distress is its level of
pro¬tability, the volatility of that pro¬tability, and how much leverage it faces. Interest-
ingly, liquidity measures turn out to be much less important. Current liquidity won™t save
an unhealthy ¬rm if it is losing money at a fast pace.
Of course, if one were interested in predicting distress, there would be no need to re-
strict attention to one variable at a time. A number of multi-factor models have been de-
signed to predict ¬nancial distress. One such model is the Altman Z-score model8:
.717 ( X 1 ) + .847 ( X 2 ) + 3.11 ( X 3 ) + .420 ( X 4 ) + .998 ( X 5 )
Z =
where X1 = net working capital/total assets
X2 = retained earnings/total assets
X3 = EBIT/total assets
X4 = shareholders™ equity/total liabilities
X5 = sales/total assets
568 Credit Analysis and Distress Prediction




14-16
Credit Analysis and Distress Prediction




The model predicts bankruptcy when Z < 1.20. The range between 1.20 and 2.90 is la-
beled the “gray area.”
The following table presents calculations for two companies, Northwest Airlines and
Merck:
Northwest Airlines Merck
Model
Coef¬cient Ratios Score Ratios Score
Net working capital/assets 0.717 “0.15 “0.108 0.13 0.093
Retained earnings/Total assets 0.847 “0.06 “0.051 0.63 0.534
EBIT/Total assets 3.11 “0.01 “0.031 0.26 0.809
Shareholders™ equity/Total
liabilities 0.42 “0.02 “0.008 0.67 0.281
Sales/Total assets 0.998 0.88 “0.878 0.84 0.838
“0.680 2.555


As noted earlier, in 1998 Northwest Airlines experienced a signi¬cant decline in reve-
nues and pro¬ts as a result of a costly pilot strike and a downturn in demand from Asia.
Consequently, it is not surprising to see that the model rates Northwest™s likelihood of
failure as quite high. Merck™s ¬nancial performance ratios are much healthier than for
Northwest Airlines. However, it is interesting to note that the model rates Merck as in
the “gray area.” Of course, Merck is a highly successful company. Its relatively low
model score re¬‚ects limitations of the model and the method of accounting for its most
signi¬cant asset, R&D, rather than performance.
Such models have some ability to predict failing and surviving ¬rms. Altman reports
that when the model was applied to a holdout sample containing 33 failed and 33 non-
failed ¬rms (the same proportion used to estimate the model), it correctly predicted the
outcome in 63 of 66 cases. However, the performance of the model would degrade sub-
stantially if applied to a holdout sample where the proportion of failed and nonfailed
¬rms was not forced to be the same as that used to estimate the model.
As re¬‚ected in the Merck analysis, simple distress prediction models like the Altman
model cannot serve effectively as a replacement for in-depth analysis of the kind dis-
cussed throughout this book. But they provide a useful reminder of the power of ¬nan-
cial statement data to summarize important dimensions of the ¬rm™s performance. Even
in the absence of direct information about management expertise, corporate strategy, en-
gineering know-how, and market position, ¬nancial ratios can reveal much about who
will make it and who will not.


SUMMARY
Credit analysis is the evaluation of a ¬rm from the perspective of a holder or potential

<<

. 121
( 208 .)



>>