A Framework for Financial Statement Analysis Part 5

advertisement
A Framework for Financial Statement Analysis Part 5:
Prediction of Financial Distress.
By GALLINGER, GEORGE W.
Publication: Business Credit
Date: Sept 2000
Subject: Credit ratings (Methods), Credit management (Methods), Financial analysis (Methods),
Business losses (Forecasts and trends)
Product: Credit Management
Location: United States
Sorry I missed getting this segment of the financial analysis series into the June and July/August
editions of Business credit. Given the three-month lapse, hopefully, you've been able to practice
your analytical and interpretative skills. Honestly, the skills you've been learning from the prior
articles are more important than the topic I'm going to examine here--financial distress scores. In
my opinion, too many credit professionals use scoring techniques of one sort or another in place
of meaningful analysis. The common approach is to look at the financial distress (or credit)
score, see what range it is in--good, bad or indifferent--and then decide on an appropriate action.
You need to keep in mind that most scoring models are based on statistical relationships and they
may have a significant level of error associated with them.
Let me give you an analogy. I live in Phoenix, Arizona. When I go to bed, I can predict with a
high degree of accuracy that the sun will shine tomorrow. I could use historical weather data to
derive a statistical model that would tell me the probability of the sun shining on any particular
day in Phoenix. And rest assured that the model would do a very good job of predicting another
glorious day. However, can I use the same model to predict whether the sun will shine in Seattle,
Chicago, New York or Washington? Of course I can't.
The problem I have in taking my sunshine forecasting model to other parts of the country also
applies to scoring trade credit accounts. What are some of the important factors causing a credit
scoring model to lose, at least, some of its predictive credibility? For example, there are industry
differences retailers have different characteristics, than manufacturers and manufacturers of
consumer goods are different from manufacturers of heavy industrial goods Further, the
economic environment can differ by geographic region. Firms located in the Central Plains may
face very different challenges than firms located in Silicon Valley. You can undoubtedly add to
this list. The point is that a statistical model is not able to successfully capture all the, effects that
differentiate firms into different levels of pending financial difficulty. So don't rely on them
excessively.
Don't get me wrong. There is a place for a scoring model in your toolkit. My caution to you is
don't expect it to provide the right answer all the time. I advocate the use of scoring models to
quickly analyze those accounts that your company has relatively low exposure. It is impossible
for you to spend the amount of time necessary to conduct the type of in-depth analysis discussed
in this series of articles on financial statement analysis. So, let me say this. Use a credit scoring
model a couple of ways. First, use the model to quickly review low exposure type accounts. If
the score indicates a problem, then conduct a more detailed analysis. Second, for high exposure
accounts, do a detailed analysis and then run the scoring model to buttress your findings.
Don't get me wrong. There is a place for a scoring model in your toolkit. My caution to you is
don't expect it to provide the right answer all the time. I advocate the use of scoring models to
quickly analyze those accounts that your company has relatively low exposure. It is impossible
for you to spend the amount of time necessary to conduct the type of in-depth analysis discussed
in this series of articles on financial statement analysis. So, let me say this. Use a credit scoring
model a couple of ways. First, use the model to quickly review low exposure type accounts. If
the score indicates a problem, then conduct a more detailed analysis. Second, for high exposure
accounts, do a detailed analysis and then run the scoring model to buttress your findings.
Many credit professionals tell me that they use a scoring model on all accounts--large and small-because they don't have time to analyze them. I don't find this argument very convincing. If one
of your large accounts files bankruptcy, and the scoring model failed to note it, you will wish
that you had taken the time to do the, analysis. At a minimum, the analysis should encompass a
detailed reading of the annual report Such a reading often reveals insights that ratio analysis or
scoring models don't uncover.
Before I get to the discussion of the scoring models, let me reiterate. Financial scorings aren't
perfect. They are based on many simplifying assumptions and may be derived on a sample of
firms that isn't representative of your accounts. With this caution, let's proceed to "test" the
solvency of Salton using the models in Number Cruncher (a model that predicts the bond rating
of the company); two Z-score models by Ed Altman of New York University; and the lambda
model by Gary Emery of the University of Oklahoma.
What Does Salton's Bond Rating Look Like?
The model for predicting Salton's bond rating is based on the academic research of two financial
economists by the names of Kaplan and Urwitz. Their study appeared in academic literature
several years ago, but I feel it does a reasonably good job of forecasting a company's bond rating.
If you look at Exhibit 1, you will notice that the bond rating is dependent on the firm's asset size
(the bigger the better), whether it has debt that is subordinated (bad) or unsubordinated (good),
the long-term debt-to-total asset ratio (the smaller the better), the profitability of the company as
measured by net income divided by sales (the higher the better), the variability in net income as
measured by a statistic called the coefficient of variation (the lower the better) and, finally, the
interest coverage ratio (the higher the better). I think if you stop and ponder each of these
measures, you will agree that their impact on the bond rating is in the direction you would
expect. For example, if the firm has a high level o f debt relative to total assets, it is riskier than a
firm with a relatively low debt level, all else the same.
We can see that Salton's bond rating was in the high-risk area for years 1995 through 1998. Any
rating of BB or less is classified as speculative debt--commonly referred to as junk bonds.
Salton's credit risk deteriorated in 1996 as compared to 1995, and eroded again in 1997
compared to 1996. This increasing risk level is noticeable by the more negative score each year
from 1995 through 1997. In 1998, there was a significant improvement in credit risk, as the score
improved to a positive number of 0.3072. However, the debt is still considered junk and the
improvement was short-lived. In 1999, the score fell back and the bond rating is again a single B.
What caused the improvement in 1998? There are several contributing factors. First, the debt-tototal asset ratio declined. Second, and associated with the first item, the interest-coverage ratio
improved. Third, profitability as measured by net income over assets got better. Last, the threeyear variability in net income became less. However the improvement was fleeting. Although
Salton was a larger firm in 1999, as measured by total assets, it also had considerably more debt
and a lower interest-coverage ratio.
What do you conclude about Salton's financial risk after looking at the bond rating predictions?
You may feel that the firm's debt has a relatively high credit risk, but compared to many of your
accounts that pay you may not be overly concerned. My perspective is that Salton, indeed, has
high credit risk. Ratings don't go too much lower than single B. According to Moody's bond
rating information for Salton, as of April 2000, its actual rating is B3. Moody's definition of a Brating states the bond lacks desirable investment characteristics, and assurance of interest and
principal payments or maintenance of other terms of the contract over any long period of time is
small. The number 3 in the rating scheme indicates that the bond is in the lower third of firms in
this rating category. Thus, the model has done a good job in agreeing with Moody's credit risk
rating of Salton.
Altman Z-Score Model Predictions
Many credit professionals are familiar with credit scoring models developed by Altman. These
models are frequently referred to as Z-score models.
Exhibit 2 shows two of his models, One model is for private and nonmanufacturing firms, which
Salton isn't. The other model is for public manufacturing firms, which Salton is. If Salton was a
private company, you should be able to easily translate the follow discussion to fit that model.
Exhibit 2 summarizes the Z-score models.
Altman's public manufacturing model was derived in the early 197Os on a group of
manufacturing firms. The significant predictor variables measure profitability (EBIT/total
assets), liquidity (net working capital/total assets), financial leverage (equity/debt) and earning
power (retained earnings/total assets). In the model, each of these ratios is multiplied by a
coefficient that is derived from a statistical procedure called multiple discriminant analysis. Trust
me, you probably don't want to know all the details of this procedure. The sum of the products of
the coefficients times their respective ratios provides the Z-score.
Which model do we use with Salton? Although Salton does some manufacturing, it tries to
outsource most of the products it sells and simply be a marketing and distribution company. The
"public manufacturing firms" model is probably more appropriate to use in analyzing Salton,
although an argument can be made to use the other model The alternative model pertains to
either private companies or public businesses classified as non-manufacturing entities.
For either model, Salton's Z-score, with the exception of the latest year, has continually
improved. In the case of the "private and non-manufacturing" model, Salton is classified in the
"no threat region" each year. The "public manufacturing" model changes the risk classification
for Salton from the "gray" category to the "no threat" category. Hence, as of the end of Salton's
1999 fiscal year, the Altman Z-score model predicts that the company is financially sound. Yes,
its score has fallen from the lofty levels the two models report for 1998. However, Salton is still
safely in Altman's no-bankrupt" area.
Let me make a slight digression to explain what "gray" means. In Altman's model, he derived the
decision rule using firms that filed bankruptcy and a matching sample of firms that didn't and
were financially strong. Thus, he had a sample of credit risky firms and a sample on non-risky
firms. He used multiple discriminant analysis to find a weighting (the coefficients) and
combination of financial variables that would correctly classify the firms into their respective
credit risk categories. Altman wasn't able to derive a prediction model that perfectly classified
firms into their respective bankrupt and non-bankrupt categories. However, he did find that firms
with scores of 1.81 or lower have a high probability of becoming bankrupt within the next year.
Thus, they belong in The bankrupt category. He concluded that firms with scores above 2.99
have virtually no probability of becoming bankrupt. Firms with scores above 1.81 but, below 3
are in the "gray" area. These firms could be financially distressed compan ies. on the other hand,
maybe they are financially sound. The model just doesn't know.
What is your conclusion about Salton? How do you feel about its credit risk? If you rely on the
Z-score model to tell you whether Salton is a risky customer or not, you have to conclude that it
is a good credit risk as of the end of its fiscal 1999 year. Is this result consistent with the opinion
you've been forming about Salton as you've read the prior articles in this series?
Emery's Lambda Index
I've left discussion of the best model for last. The lambda model is a big improvement over the
Z-score model, at least in my mind, and I suspect in Gary Emery's as well. Why? Because it
directly addresses the issue of cash flows After all, it takes cash to pay the bills. The Altman Zscore models don't directly address cash flow. They hint at the issue through the use of the net
working capital-to-asset ratio. However, if you've read my articles in Business Credit (volumes
98 and 99) about the shortcomings of traditional liquidity ratios, you will undoubtedly agree that
net working capital is an inadequate measure of liquidity.
Exhibit 3 shows Number Cruncher's results for the lambda model. Let me briefly describe the
model. I've taken the liberty to change the model somewhat from Emery's actual model.
However, I will get to that issue later. The model's outcome, called lambda, is simply a ratio. The
numerator of the ratio is the firm's cash and marketable securities, as shown on the balance sheet.
To this number is added any unused lines of credit that are available to the firm. You need to
look in the notes to the financial statements to find this number. When added together, these two
numbers represent the "liquid reserve."
Next, you add or subtract the firm's average free cash flow to the "liquid reserve." Yes, I know,
you are wondering what is free cash flow. Think of it as operating cash flow less capital
expenditures. Both of these items are found in the statement of cash flows. They are normally
called cash flow from operations and cash flow in investing activities, respectively. Look back to
the fourth article in this series and refer to Exhibit 1, page 33. For 1999, the average free cash
flow number is based on the free cash flows for that and preceding years. For 1998, the average
uses free cash flows for 1998 and earlier. For other years, the calculation is similar. By adding
this average free cash flow number to the liquid reserve number, the result is called the "net
liquid reserve." This number represents the numerator of the lambda ratio.
The denominator of the ratio is the standard deviation of the free cash flow numbers. Standard
deviation represents the variability of free cash flow during the historical period. This article is
not the place to provide you with a learning experience for calculating standard deviation. If you
have forgotten the calculation, I recommend a couple of sources. First, look in the manual for
your calculator. If this exercise proves fruitless, then move to phase two. Ask your daughter or
son, assuming they have been to college. If this doesn't help, ask the neighbor's daughter or son,
or your nieces or nephews. Someone has to know!
When Number Cruncher divides the numerator (the expected cash resources) by the denominator
(the variability of cash flows), the resulting ratio is used by Excel to provide an estimate of the
probability that the firm will be out of cash within the next year. For Salton, this probability has
pretty much become larger each year since 1996. You can tell this is the case by looking at the
lambda values. As they become smaller, the probability of being out of cash becomes larger. In
1999, the free cash flow version of the lambda model suggests that there is better than an 11
percent probability that Salton will be out of cash during the next year. Although not shown in
Exhibit 3, if the "unused line of credit" is zero each year, Salton has about a 50 percent
probability of being out of cash in year 2000. Clearly the line of credit is an important source of
liquidity.
Let me return to my version of Emery's model. He uses operating cash flow in the numerator, not
free cash flow. In other words, the lambda model derived by Emery ignores capital expenditures.
So why have I included the capex effect? The reason is simple. Capital expenditures usually
consume cash--cash that once it is consumed isn't available to meet other obligations. Many
managers spend money on questionable assets under the guise that the investment is needed to
grow the business. And many firms that file bankruptcy often do so because they grew too
quickly. So, my feeling is let's calculate the probability of the firm being out of cash under the
assumption that management is going to continue to invest in capital expenditures. The true
probability of a firm becoming bankrupt is most likely somewhere between these two probability
measures.
Thus, Salton's lambda score is probably in the range of 11 to 50 percent, depending on its
availability of credit. The 10K statements for the past four years indicate that Salton is constantly
changing the credit facility and incurring significant penalty costs to cancel it and establish a new
line. You need to interpret the score in terms of your situation. There is no magic cutoff score
similar to the Altman model. If your firm operates with high profit margins, you can take more
risk in extending a credit line than can companies with low margins.
Let's Summarize the Results
As of the end of fiscal year 1999, the financial distress models attempting to capture Salton's
credit risk are sending mixed signals. On the negative side, the predicted bond rating has dropped
from double-B to single-B. The Altman Z-scores, although less positive than in 1998, are still in
the "no threat" bankruptcy region. The lambda indices suggest that liquidity risk increased in
1999. What's the overall conclusion after looking at this data? Salton's financial distress risk is
increasing, but it does appear to have adequate access to lines of credit, particularly if it cuts
back on capital expenditures.
Is this result consistent with your analysis, opinions and conclusions discussed in the previous
four articles of this series? Given that you must have an opinion of some sort, I'm throwing out a
challenge to you, and it is supported by Business Credit magazine.
Here's The Challenge
Go back and reread the five articles that have appeared in this series to gain a true understanding
of financial analysis of Salton. Once you've satisfied yourself that you know what is going on,
submit a 2500-word, typed, double-spaced paper of your analysis and interpretation to Business
Credit. On the first page of the paper, type your name and the company you work for, including
your telephone and fax numbers, and your e-mail address. Start your report on the second page.
Business Credit will code all papers received so your name is tied to the code. Before they send
them to me, they will delete the first page. The reason for this approach is that I don't want to
know whose paper I'm reading.
There is no cash prize involved in this challenge. I'm simply appealing to your ego and
competitiveness. Business Credit will publish the best paper in a future edition--assuming at least
half a dozen people take me up on the challenge. Honorable mention will be made of the authors
of the second and third place papers. I will provide some written commentary about the best
paper in the same edition of Business credit. The deadline for submitting your paper is
November 1, 2000. And to state the obvious, you don't need to push the numbers, that has
already been done for you. The challenge is to interpret the results and write a report that makes
you proud of your effort. The reason for offering this challenge is to stimulate your critical
thinking process. Most people can generate the numbers--that is the easy part. However, writing
a report that provides the basis for a knowledgeable decision is another matter. Take the
challenge!
George W. Gallinger, Ph.D., is associate professor of finance at Arizona State University in
Tempe, AZ.
Download