The Price/Earnings (p/e) Debate By IndexFunds.com Staff January 15, 2001

Threads Taken from IndexFunds.com Discussion Boards

A war has been raging lately on our discussion boards. The debate centers around p/e ratios, and how they are calculated by Morningstar. The issue is important because many investors and financial professionals use the numbers when making decisions. Generally, Morningstar uses a method of calculating p/e which scraps all the negative-earning companies, takes any company with a p/e of over 60 and converts it to 60, and averages out the group on a cap-weighted basis. Originally sparked by Site Editor Jim Wiandt, Morningstar Research

Director John Rekenthaler joined the fray to defend a system that was created by him. As usual, a wide array of discussion board posters put in their two cents. What follows are edited versions of selected posts in the thread.

Post by John Rekenthaler

Jim:

A brief synopsis of my view (since I determined Morningstar's policy). What are we trying to capture with a "market p/e" ratio? My answer, "the market p/e of the typical stock" ... something that captures what investors are paying for the current earnings of the typical stock. Which, I admit, is quite a different question than "What you would pay per earnings if you were to purchase the entire market?"

A concern with including stocks with negative p/e's is that said stocks don't trade on their earnings. If they did, you would have to be paid to own the stock!

Simple example -

1/3 of all U.S. stocks have p/e of 40. They're glamour companies. 1/3 of all U.S. stocks have a p/e of 20. They're OK companies. 1/3 of all U.S. stocks have p/e of -20. They might be dog companies, they might be young companies, we don't know. All we know is that they now don't turn a profit.

My method - p/e is 30. Your method - p/e is 120.

If you want to answer the question, what are investors willing to pay for corporate earnings?,

then 30 would seem to be the preferable figure. On the other hand, if you had to buy the entire U.S. market, it is certainly true that it would cost you 120 times the market's current level of earnings.

Lots of ways of skinning this cat.

JR

Post by Jim Wiandt

This is great - with you as the calculation's very ORIGINATOR, I have the perfect chance to set things straight. The problem with the calculation is that it just abandons all efforts to calculate valuations at the high end of the market, and leads itself open to unrepresentative samples.

For example, lets say we have the IRekencom 80 index, which consists 5 core telecoms and a bunch of money-losing dot coms. The 75 money losers are scrapped outright, but the 5 telecoms have an average p/e of 20. Fund p/e under M*=20.
With the p/e's capped out at 60, the tech-heavy funds all show up in the high 50s, the average being dragged down kicking and screaming by the handful that aren't 60 or omitted.

Of course, this is where the system most obviously fails - at the richly valued end of the
market. I DO understand where you're coming from...under your system, 60 is just off the charts, beyond consideration. But you get a lot of clump up there with no possibility of differentiation. It's the same sort of problem Harvard has with SATs, only in reverse.
I would also take issue with just averaging out all the p/e's after turning all the inconveniently inflated ones into 60s and scrapping the money losers. I assume (correctly? - I may be wrong on this) that that's what you do with no consideration for market cap. I would prefer that the p/e that shows up for my index reflect the relative heft and/or thinness of the underlying stocks/earnings.

I guess the biggest problem I have with the system is that it just doesn't correspond with what is standard. People are used to seeing tech high-flyers (even big, stable companies) w/ p/e's of over 100, over 300...but Morningstar runs another scale altogether.

Again it seems to me that with a fund's p/e, you are trying to get a gauge of what its price divided by its earnings is...just that "p/e" indicates a formula to me...what you're coming up with seems more like diving scores. '8.9, 8.5, 9.3, 5.2 (East German judge).' Numbers with associations, not meanings.

Also, John, it seems like the negative earnings don't impact p/e's to the degree you point out in your example - much more important is bringing everything down to 60 at least it would seem so based on the Barra returns. To get an idea of how much including negative earnings can change p/e check out the fundamentals on Barra's site:

The S&P 500 Barra growth index, for example has a p/e of 41.52 EXCLUDING negative earnings. Adding in the losses though, brings up the p/e to 46.78. Your guess is as good as mine however, as to how they do THEIR calculations.

Which brings us all the way back around to HOW this came up - what drove the whole debate was the discrepancy between the p/e's reported by M* & those of DFA for identical funds. DFA's (Dimension Fund Advisors) come out far LOWER, interestingly. According to Larry Swedroe,

"M* I believe uses some crazy system to calculate data - they break data into five quintiles I think and then only take the numbers from the mid quintile. I think. Why they do that is beyond me."

Do you have any idea what he's talking about?

Post by John Rekenthaler

Jim -

You write -

"For example, let's say we have the IRekencom 80 index, which consists 5 core telecoms and a bunch of money losing dot coms. The 75 money losers are scrapped outright, but the 5 telecoms have and average p/e of 20. Fund p/e under m*=20 With the p/e's capped out at 60.0, the tech.-heavy funds all show up in the high 50s, the average being dragged down kicking and screaming by the handful that aren't 60 or omitted."

OK, I'll bite. What *should* the p/e be for this group? Let's face it, p/e is a meaningless statistic for describing IRekencom 80. No matter how you massage it.

I'm happy to put the cap at 100, whatever. But under Morningstar's approach, you need some kind of cap. Otherwise you get the company with a penny of earnings and a $100 stock having a p/e of 10,000, and totally screwing up your calculations.

Our results *are* market-cap weighted. Microsoft counts for a lot more than Ben & Jerry's.

Regarding Larry Swedroe, he's off base here. The quintile thing is a red herring, means nothing at all. We find the average weighted mean of the quintile ...it's called a "trimmed mean." Our Ph.D. stat boys got excited, did this to "smooth out" lumpy medians. Basically, it gives you 99.9% of the same answer as the median.

Post by Jonathan Kandell

John, the purpose of a trimmed mean is to weed out small aberrations in a larger picture. But with p/e this method is quite dangerous, since those exceptions can wield much weight in the cap-weighted return! It might well be that a few stocks with stratospheric p/e's don't "fit in" with the p/e's of the rest of the fund - but those same few stocks can swing down the whole fund when they crash. I can see, perhaps, trimming the bottom end of cap-weight, but not the top!

Post by Larry Swedroe

Median is quite bit different from mean. And since you buy a market cap weighted index you want to know the mean not the median p/e and then you now have the issue of the 60 p/e cut off. It seems to me that what I want to know if I am buying an index what the weighted average p/e and book to market are. This is different than what JR calls what the market is willing to pay for earnings.

Post by Jim Wiandt

John:

I still maintain that calculated cumulatively, p/e can be a good gauge of where a richly valued fund stands in terms of earnings. I'm the first to admit that the IRekencom 80 is a stretch - and what exactly p/e would be telling you is open to question (maybe 2 p/e's would do the trick...). Much more important is that a cap of 60 makes it impossible to differentiate funds at the richer end of the scale.

On point 2: this is more a problem of your calculation method than a real issue. I see no benefit in doing any sort of averaging of the p/e numbers themselves, even if they ARE weighted. Under the M* system of averaging out the p/e numbers themselves, of course you've got to have a cap. But if you took the raw underlying numbers and used those for cumulative calculation, the point would be moot.

Sorry I failed to state my favorite point clearly enough. Equating your calculation to diving scores has to do with the ballpark nature of your end results. What you are aiming for is a sort of "p/e of 20 is a cheap company, "p/e of 30 is a normal company" and "p/e of 40 means richly valued." "3 is a horrendous dive, 6 is pretty good, 10 is perfect" By choosing a cutoff you are converting your p/e to a number that is meant to fit into your scale but is not what it portends to be, that is price divided by earnings. It is a fudged calculation, W Bush's fuzzy math, if you will. Does that better explain it?

On the last point, still not sure I understand it. Larry's point, let me see if I understand you, is that a mean is calculated using just the middle quintile rather than the entire universe. What mean are you finding, exactly, and how is it used in the calculation? Are your removing weighted p/e's themselves off the ends and then just averaging out the middle? I'm interested to see exactly how you come up with your calculations (can't be THAT complicated), and see the same for DFA and run some comparisons. Maybe they cap at 30 - that would do a fine job of keeping their p/e's down.

In conclusion, I would recommend that M* implement the IndexFunds.com p/e calculation method, which is blindingly simple and unfailingly accurate, because it calculates for each index or fund, price divided by earnings, period. How does it work? Total market cap divided by total 1-year trailing earnings.
In lieu of the above, which I feel very strongly is the most accurate way to do things, I would suggest that M* move up its cap to more realistic levels. Even after their prices fell off a cliff, there is a whole bevy of stocks, including many of the most prominent, richly capitalized stocks in the country with p/e's of over 100. Even after dropping 80% of it's value, Yahoo! still has a p/e of over 150 [Editor's Note: Times, of course, have changed in the two months since this discussion.]. Cisco is about 125. So I guess if you've got to have one, a cap of 150 would capture most of the significant ones. I just can't think of any good reason to eliminate the negative ones outright. Why does a company with 10 dollars of earnings play a factor in the calculation and one with no earnings doesn't? At the least, the non-earners should come in at the top of the scale in the way that the extreme p/e stocks do.

Post by John Rekenthaler

OK, we're in the home stretch. You are right, our method is indeed fuzzy math. In contrast, your method, as you say, is "blindingly simple and unfailingly accurate." That is a benefit for your approach. However, I find your output, in many cases, to be uninterpretable. With rare cases, our output *does* function as a diving score. It is intended to be a "judgment," and it functions as such. That, to my belief, is a benefit for our approach.

Trimmed mean ... let's take price to book ration (p/b) as an example. Line up all stocks in the fund from highest p/b to lowest p/b. Count down by % of fund assets in each position. When you hit the middle dollar, you've have just found the "dollar-weighted median." If you find the region between the 40th percentile and 60th percentile, and then find the "dollar-weighted mean" of this middle quintile, you have something called the "trimmed mean." It is essentially the same number as the median, only in those rare cases when the fund has a "lumpy" portfolio that consists of only a few holdings it improves upon a median.

In summary, the trimmed mean vs. median discussion means almost nothing, except in a few cases wherein the trimmed mean provides the better result. But I'm tired of talking about it.

I'm game for raising the cap to 100 or even 150.
I'm not game for calling companies with losses a high P/E stock. A high P/E, by our system, means "priced for high future growth, that is a high expectation stock." A start-up network equipment company has losses because investors are financing a potentially great future. If Bethlehem Steel has losses, investors will pay *something* for the company because they view this situation as a temporary anomaly. But they're not really expecting a great future. It's
more like they own a bond that didn't pay out a coupon this quarter, but might next quarter.

I'm happy with a system that calls the Network Equipment company a high P/E stock but does not apply the same label to Bethlehem Steel. Even if the system is fuzzy. And I'm unhappy with one that treats the two companies similarly. Even if it's pristine.