One of the primary reasons for doing the original analysis was to get a feel for the range of multiples that have been applied to stock market earnings in the past 100 years or so. As the table below shows, the range of earnings multiples is very wide; the normalized price earnings ratios, NPEs, range from around 33 to around 7. (Note: NPE is price divided by normalized earnings.) What this means is that there have been years when the market charged ($100 of earnings x an NPE of 33=) $3300 for $100 of stock market earnings; there have been other years when you could buy the same level of earnings for about $700. The range between expensive and cheap is very wide indeed.
Pricing the Dow Using Historic Earnings MultiplesAn important question is, based on our history of earnings multiples since 1900, what price might the stock market have assigned to the Dow given recent earnings. Answering that question will give us some perspective. First, we will look at 2002, and then extrapolate to 2008/2009.
Implications of Price to Earnings Ratios on Expected Dow Price Range for 2002
Note: The average NPE between 1898 and 2007 was 14.9.
The table above approximates the range of prices that the market might have paid for 2002 normalized earnings (NE) of $455. (Because of the way I normalize earnings, 2002 is the most recent year for which I have calculated normalized earnings.) The table shows:
- Normalized earnings: For 2002, $455 is the average of actual Dow earnings from 1997 through 2007.
- NPE Percentile: 10% of the NPE readings are at or below the 10th percentile; 20% are at or below the 20th percentile; 100% are at or below the 100th percentile.
- NPE: The normalized price earnings ratio for that percentile. For example, 10% of the years had NPEs of 9.4 or below.
- Dow Estimated Price: For 2002, this is normalized earnings (NE) multiplied by the NPE for each percentile. In other words, price is equal to earnings times the earnings multiple.
Extrapolating Normalized Earnings to 2008
I can't yet calculate the normalized earnings for 2008 since we don't know the earnings for 2009-2013. Therefore, I have had to estimate. That's ok; we're just trying to get some ballpark numbers anyway. I have estimated 2008 normalized earnings three ways -- that is, by extrapolating from three different "base years":
- Based on 2002 -- the most recent year for which we have normalized earnings. If you increase the 2002 normalized earnings of $455 by the long-term earnings growh rate of 5.72%/year, you arrive at an estimate of $635 for 2008 normalized earnings. For various reasons, I think the 2002 NE are above trend; they probably generate an optimistic estimate for 2008. (See this post for how I arrived at 5.72%)
- Based on 1974 -- the most recent secular trough. If you increase the 1974 normalized earnings of $83.11 by 5.72%/year, the estimate for 2008 normalized earnings is $551. Since 1974 and 1932 are secular bottoms, they probably (hopefully?) generate somewhat pessimistic estimates.
- Based on 1932 -- the trough associated with the Great Depression. The 1932 NE of $8.32 project to $570 in 2008.
2008-2009 Worst-Case Scenarios
Based on 100 years of stock market history, it appears that the February 23, 2009 Dow close of 7115 is somewhere between the 30th and 50th percentile. For example, using the $635 estimate for 2008 normalized earnings, the 30th percentile is 7091 -- just below the actual close of 7115. Apparently, even after all this pain, the market is still not extraordinarily cheap. Some additional observations:
- No matter which of the three estimates for 2008 normalized earnings we use, in at least 30% of the years in our database those earnings would have resulted in a Dow price lower than 7115.
- Regardless of which estimate you use, in at least 10% of the years the Dow would be valued at less than 6000.
- If "Mr. Market" decided to apply valuations comparable to those applied in 1932 and 1974, we might see the Dow in the neighborhood of 4100. In a way, this post clarifies the worst-case estimate in my previous worst-case scenario post. The years 1932 and 1974, are basically "bottom of barrel" valuations -- earnings multiples that we have seen only about twice in 100 years. So, another way to think about the Dow 4084 figure quoted in the previous Worst-Case post is that it assumes a) 1974 (i.e., near absolute worst-case) normalized earning -- projected forward, and b) 1974 (again, near worst-case) normalized price earnings ratios.
Obviously these numbers are not predictions; I am simply trying to "ballpark" POSSIBILITIES -- based on estimates and extrapolations. Primarily, I'm trying to get a feel for what MIGHT happen so that I can prepare for it. What I conclude from the analysis above is that prices could go lower and still be consistent with stock market history since 1900. (Note that though this analysis is based on Dow Jones data the results would have been similar if I had used S&P 500 data.) A future post will describe how I go about preparing for these possibilities. Bottom line: I'm not betting the farm quite yet.
Related MaterialsDow Price to Earnings Ratios Since 1900 - A Summary: The precursor to this post.
Stock Market Earnings History, Average Returns, and a Worst Case Scenario: My previous approach to developing worst-case scenarios.
The 1929-1932 Stock Market Crash Revisited: the ultimate worst case?
Managers Say Stocks Have Been Cheaper: Morningstar article quoting John Hussman and Jeremy Grantham who reach similar conclusions.
Why Stocks Still Aren't Cheap: New York Times article: same conclusion.
Three Scenarios for the Economy (and the Stock Market): A qualitative instead of quantitative description of scenarios.
The Extraordinary Impact of Price to Earnings Ratios
Al's Stock Market Analysis Model: The spreadsheet containing my source data. Note that that version contains closing prices beginning in 1898, but earnings only from 1929. However, for this post, I estimated earnings prior to 1929 using data from another stock market index. If you do the calculations using that spreadsheet you will get slightly different NPEs. From the 30th percentile down the results are the same as presented above plus or minus 0.1.
Last updated 12/13/2009