Re; Bankruptcy Predictors Revisted

Hi All,

Been away on vacation so I missed the discussion. The expert on this subject is Edward Altman. All your questions about bankruptcy prediction, including its methods, statistics and accuracy can be found here:

formatting link
Note that Dr. Altman did not start out with a theory about why companies go out of business, he just is an expert in the stats. Companies with certain types of ratios tended to be out of business within a year. That is where it all started. The accuracy of these predictors is much higher that you think they would be.

Cliff

Reply to
___cliff rayman___
Loading thread data ...

...

Yes, that they are empirical models was observed in the summary papers I saw referenced. How accurate such models are for future predictions is dependent on how well the conditions of the model are reproduced by the company and external events comprising the prediction. Whether the underlying sample space can actually be considered a random process is often a key. IOW, "past performance isn't indicative of future results..." :)

Reply to
Duane Bozarth

BTW, thanks for the link--I intended to add that on other note re: prediction accuracy/validity.

Reply to
Duane Bozarth

You are right about "past performance isn't indicative of future results", but if a company is in trouble and does not do something about it, things will not get better. These bankruptcy predictors can show that trouble exists and give a business-owner reason to be cautious in the future.

Wayne Brasch

Reply to
Wayne Brasch

Surely. I've just started reading "CORPORATE DISTRESS PREDICTION MODELS IN A TURBULENT ECONOMIC AND BASEL II ENVIRONMENT". I've skimmed most of it quickly to get the overall drift. While not particularly long on detail, one interesting note wrt to prediction accuracy is that he points out that at the optimum Z-criterion level for model fitting, the Type II prediction error for a non-static sample not in the model base approached 20%. That's fairly high altho Type II error isn't a major cost here as it might be for other fields. Well, actually, I can think of one potential cost--if an analyst were making recommendations based partly on such factors and subsequently the recommendation/rating were non-positive, it could have a fairly strong impact on stock prices, say.

Anyway, is interesting...I'll do some more reading in depth and may post further thoughts/questions...

(I can now

Reply to
Duane Bozarth

I'm not so much concerned with stock prices as I am about the continuing existence of the company.

Wayne Brasch

Reply to
Wayne Brasch

Wayne Brasch wrote: ...

...

Granted, I was commenting on a result that I found interesting although w/ some publicly traded companies, a precipitous drop in stock price could hasten the end.

I note you seem somewhat reluctant to discuss the models themselves...

Reply to
Duane Bozarth

OK, I've read enough to satisfy my former curiousity! :)

Actually, in one line the stability of the Altman Z-factor is much greater than I would have thought. Actually appears quite remarkable that the original model coefficients from 1968 are the same as those of the base model for publicly traded manufacturing firms analyzed in "CORPORATE DISTRESS PREDICTION MODELS IN A TURBULENT ECONOMIC AND BASEL II ENVIRONMENT" (2002, available at

formatting link
A comparison there of the original model for that class of companies shows a mid-90% rate in the original test set, dropping to low- to mid-80% range for a group from 1969-1975 and 1976-1995, respectively, then rebounding back to mid-90% for those bretween 1997-1999. These are the one-year lead time predictions for the 2.67 cutoff factor value. However, at two years, these success rate drops significantly, being only roughly 75% for the last two time periods above. This still is undoubtedly valuable no doubt as a screening tool as Wayne (I presume) recommends.

Regarding the analysis of the Enron and WorldCom cases, in the former both the modified Z-factor (Z'', accounting for the non-manufacturing market) and the KMV EDF predictor showed definite signs of distress in time for any observer watching to have escaped the debacle (assuming, of course, one weren't a captive member of a plan which wouldn't allow asset transfer). Otoh, neither model, while showing signs of degradation predicted the actual failure of WorldCom --

"So, while both models were indicating a non-investment grade company as much as 18 months before the actual downgrade to non-investment grade and its eventual bankruptcy, we would not have predicted its total demise based on the available financials. But it did go under, primarily because of the fraud revelations and its attendant costs due to the loss of credit availability."

So, I concur there's value in the models as Wayne has been touting and both appear more robust than I had initially expected (but I'm a physicst/engineer, not an accountant, so what do I know? :) ). It is certainly clear the selection of which Z-factor model based on the appropriate sector is significant in using the scoring although since all are positive coefficient models, trends in negative directions could still be meaningful in raising concern levels.

The one nagging question is whether if the trend to less precision as the time period is expanded were to grow even more whether there would, in essence, be a risk of almost everybody being "at risk", thus losing the benefit. More data on specific market sectors and companies tracked for a longer time would be required to answer that question. It is significant to note that --

"The Type II error (classifying the firm as distressed when it does not go bankrupt or defaults), however, has increased substantially in recent years with as much as 25% of all firms having Z-Scores below 1.81. Using the lower bound of the zone-of-ignorance (1.81) gives a more realistic cutoff Z-Score than the 2.675, although the latter resulted in the lowest overall error in the original tests. The model was 100% accurate when scores were below 1.81 or above 2.99."

If the factors are used as an indicator routinely, however, and trended for a particular corporate entity, changes in the indicators could well be more significant than absolute numbers. This hypothesis mirrors experience that shows that some physical models, while not good predictors of absolute values, be very useful for predicting in the same properties.

HTH some, interests others...

Reply to
Duane Bozarth

If the accountants of a particular client would routinely enter that client's financial data into these bankruptcy predictor models, that accountant should spot a trend whether good or bad. I have no reluctance in discussing any of these predictors-I use them all the time with my clients.

What I seem to have found, though, is that most accountants don't know these predictors even exist. Either that, or they don't care.

Wayne Brasch

Reply to
Wayne Brasch

...

I can believe both...

Have you had a case in which the indicators went south in time for the client to subsequently fix the problem successfully? If so, I would assume there are applied journals in your field which would accept case studies for publication which would be one way to promote their usage more widely.

Of course, as noted, the actual computation of the Z-factor for trending is almost secondary--simply the direction of change of the ratios is indicative of the direction of change in Z.

Anyway, I learned something that could be useful for monitoring some portfolio holdings. Thanks for that even though I'm not your target audience.

Reply to
Duane Bozarth

I'm truly glad your broken arm seems better-at least, you can type with it. I have had occasion to catch a company in a situation in time to be able to turn them around as has Cliff Rayman. More accountants and business consultants could use these indicators, if they would. I wish I could get a better picture of how many accountants actually don't know about these predictors. I know I never had it mentioned in any accounting class I ever took in college. The various editions of the accounting book I taught from for 17 years at a local vocational-technical school didn't mention them at all. In the local university continuing education classes I teach periodically now, none of the students have ever heard of them and are amazed as to the accuracy of them when I present this information to them. Maybe it is that nobody teaches about them.

Wayne Brasch

Reply to
Wayne Brasch

...

So am I...thanks! :)

As noted, I'm not an accountant but I roomed w/ an accounting major for a while in school. Observing the classwork, such theoretic econometrics wasn't part of the curriculum at all--simply the "nuts 'n bolts" of the balance sheet is how I'd characterize it. (As an engineer, I actually graduated w/ more econometrics than he did as an accountant. As I recall, it was mostly Econ majors and similar, not Accounting majors who were the primary participants. I'd also note that except for a small fraction, almost all were woefully inadequately prepared mathematically for the topics--it was that which allowed me to go find out what I didn't know in Econ 202 and keep up or stay ahead of them as that information was much easier to assimilate quickly than the differential calculus.) It would take a wider view than "just accounting" to have brought this type of material (or even the objective) into the mix.

I think underlying the issue is what the objective of the accountant in question is--whether they're interested only in the balance/numbers or can step back and view "the big picture" as you're obviously interested in doing. I've known engineers who had similar shortsightedness--I'm sure the syndrome is present in any field of endeavor.

I'd also reemphasize that the absolute number in these predictors isn't really even required--if the accountant isn't using them (the correlations), are the predictive factors something that would routinely be generated anyway? I gather probably not as if they were I would presume a change in the negative direction would show up as a flag. Of course, if the individual isn't looking for the assessment of condition and future prospects, it wouldn't matter even if the Z-factor itself were spit out.

What I'm driving at is that to accomplish your stated goal I think your objective needs to be larger--that is, changing the objective of the average accountant to be more attuned to the larger question than the detailed balance.

The way to do that would be through whatever methods your profession has for continuing education and trade-related journals, I'd think. Working to influence those would seem one possible course of action.

Reply to
Duane Bozarth

Duane, you are right. I'm afraid that most accountants view themselves as company historians-just recording the daily activities of the company in the accounting records. They print out the financial statements from their software, hand them to the client, collect their fat check, and run to the bank with it. That is not what should be done, of course.

That is and was my reason for starting this thread in the first place. I wanted to make people aware of the fact that accounting is much, much more than that. I will try to get my message out in other ways. Thank you for your comments.

Wayne Brasch

Reply to
Wayne Brasch

Wayne Brasch wrote: ...

You're welcome...it was an interesting exercise. I'll note in closing that if the above is the purpose, the manner you responded in early on in the previous thread wasn't very conducive to achieving the goal. It took another respondent pointing to a url to provide me a convenient way to find more current information and your "discussion" consisted almost exclusively of trying to e-mail what turned out to be a spreadsheet containing the correlations discussed above, and it took several tries to get you to even state that is what you had. It wasn't until I had the opportunity to read the last paper and posted a brief synopsis of it with some comments on my take on it that this last discussion occurred. Would seem to be much more likely to be fruitful "evangelism" if you were to prepare some more discussion, examples, etc., to use in the fight.

Reply to
Duane Bozarth

The papers are a little complex for the average accountant. There has to be some way of crafting a message that convinces accountant's of the importance of performing bankruptcy prediction and doing ratio analyses and delivering that information to the client in a way that makes sense to them also. C

Reply to
___cliff rayman___

Cliff, I will work on something that may make it a little easier to understand. I don't understand why some people write and talk in such complex ways. Maybe it's just to impress others!

Wayne Brasch

Reply to
Wayne Brasch

Wayne Brasch wrote: ...

Some, undoubtedly. The Altman papers I've read certainly aren't what I'd call "complex" analytically--if so, that's perhaps an indictment of the current Accounting curriculum? As noted, my experience has been that mathematics beyond perhaps algebra was sadly lacking based on my elective coursework, but that's 40 years hence, now. I'd have thought the situation might have changed at least some.

That said, they however, written by a PhD econometrics guy, an accountant, and therefore can't be expected to be completely devoid of some basic theory. I do think to achieve any widespread accepetance will require simplified presentation, however, probably as I suggested concentrating on case studies as a teaching mechanism. I'd caution, however, against using the spectacular cases as a prime motivator--recall, particularly, that WorldCom wasn't predicted to fail, and in fact, both models considered by Altman showed improvement from previous estimates at the last reporting period before the actual collapse. Parenthetically, also note that Altman points out specfically that the sinking bond ratings if actual public ratings had been correlated to ratio analysis predictions could have well accelerated the WorldCom collapse so the comment I posed previously regarding Type II error is not one to be treated too lightly, especially for publicly traded companies.

Anyway, good luck w/ your mission. I agree w/ you and Dr. Altman these methods should undoubtedly get more broad exposure, but that usage needs to be only an additional tool in the kit of the accountant, not a "magic bullet".

Reply to
Duane Bozarth

Generally, accountants do not need to understand this level of analytics. Its not part of what they do on a regular basis.

It is all taught, but then never used, so hardly remembered.

Of course, this can be argued two ways. When a company faces a crisis publicly, they are more likely to get the help they need. Better to get thrown into hot water quickly, rather than in a pot brought slowly to boil. Also, even though these are statistically called Type II errors, they are probably not as statistially pure as in other usages. Perhaps the companies that did not become insolvent, realized their troubles and took action to turn their companies around. I am in the turnaround business, so my belief is that many companies that are heading towards insolvency can be turned around as long as correct and decisive action are taken early enough in the decline.

The worst that can happen when you predict a high possibility of bankruptcy is a more thorough look into the company operations and it's financing. If deeper analysis of the company's financial and operating ratios, and its practices reveal a healthy company, then the predictor rating can be ignored as an anomoly. Generally, when I see a high prediction of failure, I can tell why the predictor is not happy with a glance at the company's ratios. The company is usually truly in danger, and although it might get lucky or smart and avoid insolvency, it does not mean that the predictor returned a truly false result. A highly leveraged company that is losing lots of money, will show a high prediction of failure. The company may not fail, which is a Type II error, but they could have gotten a lucky break in the market, or some help (whether voluntarily or forced), that allowed them to avoid insolvency and bankruptcy. Cliff

Reply to
___cliff rayman___

Many valid points...

On education of Accountants, no argument at all--simply a rejoinder comment to Wayne's apparent wondering why not more are aware after they have completed course work--it seemed to me that that sort of analysis wasn't the target job for those being trained as accountants--whether that's shortsighted or not I leave for the accountants to decide amongst themselves... :)

I agree "Type II" errors in this area are not as risky as in some and not be particularly costly. I was just wanting to point out it isn't zero cost, however. One cost is simply in lack of discernment which requires other effort to uncover real problem--as long as the proportion is only 25% or so, that most likely is minor. And, of course, any methodology is only as good as the data. In a public situation such as WorldCom where there were conscious efforts to obfuscate, the point isn't to uncover but to hide. One would presume in your and Wayne's situations the client is actively trying to find out and is open and data are accurate. That undoubtedly helps cut down the likelihood of false positives.

Reply to
Duane Bozarth

BeanSmart website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.