Monday, December 22, 2008

SPX and P/E

In an earlier post, I had looked at some of the earnings forecast revisions that S&P had made in the past months. This time, I'm going to confine my discussion to historic PE levels and implications for the market in the future, considering the current earnings expectations.

For this, I downloaded Robert Shiller's monthly S&P data set (see sources at bottom) that goes back to the dark ages of the market. But since I just couldn't see how the S&P's valuation in the 19th century was relevant, I decided to restrict the set to 1950 until the present time. There are some caveats with the set that he explains and you should read but that I will omit here. Shiller's data set uses the "as reported" earnings as opposed to "operating" earnings. The former includes all the write-downs, write-offs, one-time expenses, etc. whereas the latter omits all of these things and is typically the number that you will read has been forecasted by analysts. Here is the chart of P/E & SPX. SPX on this chart is on a log scale to allow the movements in the 50s to be seen.

The last 3 months in the data set are estimated using the actual SPX levels but with the most recent S&P Q4/2008 EPS estimates. The average P/E from 1950 to present is 16.59 with a standard deviation of 6.76. The majority of the points (~73%) fall withing +/- 1 standard deviation. The only points that are above +2 standard deviations (PE>= 30.06) are unsurprisingly on the high end of the distribution. Approximately 4.2% of the months fall into this range and perhaps also unsurprisingly, essentially all of these points originate in the 2001/2002 bust period.

The next step was attempting to derive something useful out of this historic information. If you have had the displeasure of viewing any financial TV since the market began crashing in earnest, you've likely heard that this is a "great time to buy stocks because they are undervalued historically," or something of that nature. What seemed like the best thing to do with this data set was to bin the monthly PE values against 1 year SPX % returns and compare the data sets to see if there are any statistical differences.

In the chart, you'll see some circles that overlap and some that don't. The degree of overlap indicates how distinct these sets are. Similarly, the positive values in the comparison's table (just below the mean chart) show which groups are (and are not) connected. The deviations within each group are rather large, however the bins (>+2 std. devs, +1 to +2 std. devs, -1 to +1 std. devs and <-1 std. devs) are - at least on the extremes - relatively distinct. The +1 to +2 std. dev and -1 to +1 std. dev groups are not distinct and the very small circle is the representation of the former category. The summary for average 1 year SPX percent returns: <-1 = 13.91% -1 to +1= 8.42% +1 to +2 = 6.41% >+2 = -2.14%
Entire data set: 8.66%

So... where is the PE of SPX currently? Well, if you believe the S&P EPS estimates for Q4, the trailing 12 month PE should be a bit over 18. Not exactly in the "cheap" range to say the least but not as wildly overvalued as it had been in the 2nd and parts of the 3rd quarter. Most optimistically, I'd say SPX is averagely valued and an average return could be expected rather than a sharp rebound.

One other topic that I've not heard addressed in a very satisfying way: multiple contraction and multiple expectations if appetite for securities disappears or diminishes in the retail arena. After all, a lot of people would have been served quite well by simply putting their 401k contributions into US treasuries for the last 10 years. (Yes, I'm aware this is simplistic and overlooks dividends. Maybe I'll look at that in the future.)

Sources: (download ie_data.xls)
S&P 500 EPS estimates (see link on right)