Relative Altman Z-Score

4th Nov '19 by AnonymousUser228041

7 comments 520 reads

Hello everyone.

Too often we find a red warning saying that "Altman Z-Score Screen" was triggered for a security we are interested in. Indeed, around 45% of non-financial shares have a Z-Score lower than 1.8: but a warning that pops up half of the time is not a very useful warning. So, to say that a company with a Z-score of less than 1.8 is very likely to go bankrupt within the year is like to say that the female half of the population (from 0 to 100+ years old) has a good chance of giving birth to a child within the year. In conclusion, to stay in the maternity metaphor, using Z-Score like this you really risk throwing the baby out with bathwater.

I believe that the problem derives from the fact of using fixed values in a theoretically still valid model. Perhaps certain values were perfect in 1968, but today they no longer work so well. That's why I worked out my personal version of the Altman screen: only 10% worse for Z-Score in the same industry. In this way the warning is triggered only for about 3-4% of non financial companies. This seems to me to be more reasonable, given that listed companies that go bankrupt every year are usually a smaller percentage. In 2018 there were only 58 bankruptcies over thousands of american public companies.

What do you think?

Unlock with Email

Unlock with Google

Already have an account?
Login here

Nick Ray

4th Nov '19

Not sure if I've misunderstood you, but looking at Altman Z(2) scores, I get about 20% of non-Financial stocks below 1.8 rather than 45%. Maybe it's different for Z(1) scores.

However I definitely agree that using rank-based comparisons is generally a good idea. For Altman Z(2) I find that being above the median (which currently is roughly a Z(2) score of greater than 5) is a good place to be. So I guess I am still rejecting half the stocks - not because I think they'll go bust but because they tend to have less good performance.

AnonymousUser228041

Yes, "Altman Z-Score Screen" uses Z(1) score.

Remember that Altman developed his score to help avoid the next Enron, not improve performance. It's a filter to select the worst not the best. There is no reason to think that a portfolio with a higher Z-score will perform better than others. It has only a smaller probability to pick a bad apple. Bad apples are usually only in the bottom of the basket.

There is no reason to think that a portfolio with a higher Z-score will perform better than others.

That's true. But in fact, it does have quite reasonable discriminatory powers. The plot below shows the "likelihood ratio" for Altman Z(2) over the "prior" odds. It's fairly decent:

iwright7

Nick,

As I understand it Alt Z 1 is designed for manufacturing companies and Alt Z 2 for non manufacturing companies.

Are you restricting your likelihood ratio Alt Z 2 observation to non-manufactures, and/or only to companies with a calculated Alt 2 metric in Stocko Bankruptcy Risk score?

Ok. Just be careful not to pay too much for quality. Only about 0,5% of firms go bankrupt every year and they have an average Z-Score of -0.15. If you discard half of the business for their Z-Score, it will be very difficult to still find a good deal.

Gromley

There is I think a weakness in the idea of only "screening out" the worst 10% of Z-Scores for each industry.

That discounts the possibility that at a given point in time (or perhaps even permanently) a whole industry could be at elevated risk of bankruptcy - the financial crisis should have taught us that as more latterly should the state of "bricks and mortar retail".

I also would not, personally, just want to screen out the worst 10% of the whole market and by inference assume that the other 90% are "safe". (Although I must give a nod to Nick's figures that show there is a probability curve)

Personally if I use Z-score at all I use it as a flag to have a think and a further dig in the numbers to observe whether in the specific case the numbers really do highlight a risk or whether there is a logical explanation.

I have to confess that I found the Altman papers so turgid and jargon bound that I could not summon up the will to read them properly. However if I understand it correctly, some of the descriptions of the Altman work are misleading. Including Stocko's very own :

the model was found to be 80-90% accurate in predicting bankruptcy one year prior to the event.

To go back to the original analogy, I believe that is equivalent to saying that "being female is 100% accurate in predicting pregnancy one year prior to the event" - On the basis that 100% of those getting pregnant in the next year will be female (probably!)

I don't know how much post 1999 analysis there has been of the effectiveness of the Z-Score (Nick's graph certainly implies that it still has some power).

I note that the calculations are dominated by Assets. 4 out of 5 of the tests have Total Assets as their denominator.

I would hazard two guesses about 21st Century business vs. their 20th Century counterparts.

21st Century businesses are on average less Asset heavy than the former crop.
C.21 companies can be more prone to have substantial intangibles/goodwill on their balance sheet but this varies significantly dependent on (i) How acquisitive the company has been and (ii) specific policies on whether "development spend" is capitalised or expensed.

I wrote a piece that alludes to some of this over on Graham's cube investments site a few weeks ago and am overdue a follow up piece that explores the concept and value of assets more generally.

I haven't got the bandwidth to prove it, but I strongly suspect that Altman would reach somewhat different conclusions if he conducted the experiment based on data for the last two decades.

Although if anyone has information on more recent detailed research (not written in Martian) that disproves my conjecture I would love to read it.

Gromley, you pose some interesting questions.

Bankruptcies are indeed concentrated in small, young manufacturing firms. This is also the reason Altman concentrated his method initially on manufacturing firms. Afterward he developed a second version of Z-Score for non manufacturing firms. Because industry actually matters.

The reason I think that it is not sound to avoid more than a little number of firms is that as a value investor I look for undervalued firms. If I find a company with a big Z-Score, a full Piotroski, a small debt, a stellar ROIC... probably it is very expensive as well. Safe havens and glamour stocks are rarely cheap. So I use Altman to avoid the most dangerous firms, not to select the safest ones.

Some of these questions are discussed in "Quantitative Value" of Wesley R. Gray. A critic of Altman Z-Score is in "Bankruptcy Prediction With Industry Effects" of Sudheer Chava and Robert A. Jarrow.

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=287474

The paper is quite technical, but conclusions are already clear in the abstract. Beside it contains a lot of interesting data in appendix.

Relative Altman Z-Score

Unlock the rest of this article with a 14 day trial

About the Author

AnonymousUser228041

7 comments

Unlock the metrics for all these stocks and more...

Unlock the metrics for all these stocks and more...