How Low Are PE Ratios? A Comment on Mark Hulbert’s Take

As the market zigzags lower, many experts say we’re entering a recession as bad as 1982. As Nobelist Joseph Stiglitz pointed out recently (minute 19 and beyond here), we have little reason to have confidence in our policy makers.  So I continue to ponder the question of how low PE ratios might go (see here).  After adding more short positions on the S&P 500 SPDR ETF (SPY) Thursday afternoon, I read Mark Hulbert’s MarketWatch article on PEs, which surprisingly contains a major error.   

While I agree with Mark’s conclusion that PE ratios are still historically high, he grossly overstates how high they are.  He says, in referring to Robert Shiller’s (Yale) historical PE ratio studies,

…the current PE ratio for the S&P 500 index is 18.1…Even after a year-long bear market, and especially even after the market's rout over the last month, the stock market's current valuation is still more expensive than 79% of the months over the last 138 years.

Mark uses the PE for trailing twelve month earnings, but he should be using the current SPY PE10, with average earnings over that last 10 years, as Shiller does in his data base.  Today’s SPY PE10 is 14.  Using this, the correct percentage is actually, 40%, or ½ that cited by Hulbert, as I verified by pulling out my cumulative probability plots of Shiller’s data (raw data available on his web site, but I’ve updated to today). 

Below is a cumulative probability plot of month end data for various time frames.   (Note: this is a standard statistical data plot that shows for each value of PE10 along the horizontal axis the corresponding percentage of data points BELOW that PE10  value.)

click to enlarge images

For example, if one looks at the yellow “all time” data line, at today’s PE10 ratio of 14 for SPY (marked by the dotted vertical red line), 40% of the monthly data is below 14.  Moving up along the yellow line we see that about 80% of the monthly PE10 data is below 20.  

Back to Mark’s claim that 79% of the all time months are below today’s value. As you can see, the “all time” line intersects today’s PE10 at 40% probability, meaning that 40% of “all time” months had PE10 below today’s value, not the 79% noted by Mark Hulbert. 

Shiller usually discusses his all time data, however many argue that times are different now -- the SP500 didn’t exist prior to 1950 and has changed makeup since, interest rates are low today, etc., so this all time historical plot is “irrelevant”.  So I’ve looked at 4 different time frames, “all time” (1871 – today), “post 1950” (to today), “all time to 1996”, which excludes the bubble of the late ‘90s, and “1950 to 1996”.  We see that the highest cumulative probability is “all time to 1996” with 43% and the lowest is “post 1950” with 33% of months below today’s PE10.  From the footnote below, if times of high interest rates (>4.5%) are excluded, there is little impact on the conclusion. 

For historical reference, I’ve plotted the PE ratio for several other dates at the top of the plot, ranging from the post 1950 back in 1982 to the high in 1999.  Note that I’m using Shiller’s convention (month-end data), so there may have been higher or lower intra month values not shown here.   

So, what do I conclude from this?   

As many have pointed out (Benjamin Graham, Warren Buffett, etc), markets are driven in the short term by emotions and greed.  In the long term they are driven by fundamentals.  The time to buy is when the fundamentals are at a “value” level and there are today a lot of good stocks out there with their PE10s below 10.  Is this the time to buy?  While I’m still playing the short side of SPY, I’m beginning to make systematic 5% moves from cash into selected stocks.   

At some point I’ll stop playing the short side.  I’ll be watching sentiment, policy moves, and PE10.

If we are heading into a 1980s style recession, or worse, then it seems reasonable to me that the SPY PE10 could reach a level near or below 15-20% cum probability.  If it’s 20%, the PE10 range from these plots is 10-12.  If it’s 15%, the range is 9-10.  If we set an all time low, then it’s below 6.  

Footnote:

Disclosure: Author is increasing his shorts on SPY

Comments

[onzilla:]

not really sure about your data... did you take out the $X Billion of XO losses related to the financials? You would need to unless you think that those numbers will recur on a annual basis indefinitely. Estimated P/E for next year earnings is 11... which is what you should really be looking at... and lets just say that those 'estimates' are wrong... and we are really looking at next years earnings being a 15 P/E , which gives you a earnings yield of ~7%. Compare that with long dated US treasuries. Anyway you cut it, things at these levels are cheap historially v. treasuries. Look at European equities and the valuations are even more pronounced.

2008 Oct 26 09:12 AM

[datacheck:]

onzilla,

I didn't take the huge (expected) reduction in financial earnings out of the data, but you're right, I should have. That of course makes the picture even bleaker, as I pointed out in my previous article.

Ben Graham (in his classic book, "The Intelligent Investor") flatly states that one should use average (as reported) earnings (7 or 10 years) to analyze a company's stock. I'm still playing around with different approaches -- using various combinations of multi year past earnings and forecasts.

By the way, do you know of a source of7 or 10 yr historical earnings for companies? S&P makes historic index earnings readily available, but it's not so easy to find data for individual companies.

2008 Oct 26 11:10 AM

[alajac:]

It doesn't matter how low the PEs go. The herd will keep running until the herd stops running.

2008 Oct 27 09:04 AM

[ERCaptain:]

You can get the 10 historical information off of the S&P Stock Reports. All financials are avaiable for the last 10 years. Per Share Data, Tangible Book Value, Earnings, S&P Core Earnings, Dividends, Payout Ratio, Prices:High, Prices:Low, P/E Ratio:High, P/E Ratio:Low, etc -- and much much more!

2008 Oct 27 11:15 AM

[granger:]

thanks for the probability plot

2008 Oct 27 05:08 PM