Analysts have argued for years about the merits of price/earnings (P/E) ratios. When P/Es are high, as they were in the late 1920s and 1990s, raging bulls would proclaim that the ratios are irrelevant. When P/Es are low, as they were in the 1930s and 1980s, marauding bears would argue that the worst is still ahead. Each time, both were wrong. Here we test a newly designed indicator to determine if P/Es can be effectively used to generate buy and sell signals. To get a complete picture of its effectiveness, we'll look at whether this indicator would have helped the trader beat the returns rendered by a buy-and-hold strategy over the period from 1920 through to 2003.
Trading Tools - Building the P/E SMA Indicator
Simple moving averages (SMAs) are one of the most basic tools for building a trading system but they have remained popular among technicians for one simple reason: they work. A moving average (MA) reduces the noise by smoothing the data, allowing the trader to see the bigger picture more clearly.
Another useful charting metric for analyzing data is a linear regression line. It is very useful in showing a trend and providing insight into potential future price movement. A number of popular charting programs include a function for the linear regression line.
Using annual historic S&P P/E ratio data by Robert Shiller, Yale Professor and author of the best-selling book "Irrational Exuberance" (2000), we constructed charts and a simple moving average crossover system using a shorter-term MA trigger, or fast line, and the long-term MA base, or slow line. The signals generated by changes in S&P Index P/Es, which are charted in figure 1, were used to buy and sell the market as represented by the Dow Jones Industrial Average, which is charted in figure 2.
The best combination of moving averages is somewhat of a juggling act. Longer-term MA periods reduce the number of signals and add delays, which often result in lower returns. Shorter MA periods often increase some individual trade returns at the expense of adding more losing trades, thanks to whipsaws.
Figure 1, as already mentioned, is a chart showing annual S&P 500 Index (and earlier precursors) price/earnings ratios from 1920 through 2003. The chart displays also the two-year (blue line) and five-year (magenta line) simple moving averages. Buy signals occur when the two-year SMA crosses above the five-year, and a sell signal when the two-year crosses below the five-year. The median P/E over the period was 15, but note that the linear regression channel midline (dashed diagonal line) shows that the trend moved from a PE of 12 at the left hand side of the chart to 21 on the right side.
In figure 2, showing the monthly chart of the Dow Jones Industrial Average (DJI) from 1920 through 2003, the green arrows indicate buy signals generated by the two-year S&P P/E SMA crossing above the five-year SMA, and the red arrows show sell signals when the reverse occurs in figure 1. A total of six buy and six sell signals were generated for a total gain of 9439.25 DJIA points.
Figure 2 - Chart provided by MetaStock.com
For the sake of our test, a two-year moving average signal line was found to remove much of the noise without adding excessive delay. The baseline consisting of a five-year moving average was determined to be a good fit. A one-year signal line instead of a two-year was tested and found to provide the same number of trades but with slightly lower returns.
How Did the P/E SMA Indicator Do?
In a total of 12 trades (six buys and six sells), the system returned 9,440 points (see figure 2). A buy-and-hold over the same period would have returned 10,382, so our indicator lead the trader to capture nearly 91% of the gains the Dow made during the 83-year period.
But the real benefit of using the P/E SMA indicator is that it told the investor when to leave the market, thereby protecting investments from losses. Using the P/E indicator, our trader would have been in the market a total of 48 of the 83 years, or 58% of the time, which means he or she would have had money invested elsewhere 42% of the time (25 years) where returns were better.
A buy-and-hold investment in the market for the whole 83-year period earned 10,382 by the end of 2003, which works out to 125 points/year. The trader using our P/E indicator, gaining 9,440 points in 48 investing years, would have made 197 points per year. That is a 58% better return than the buy-and-hold investor!
Given these results using annual data, we can ask whether the use of monthly data would have improved overall results. The best-fitting set of moving averages for our monthly indicator was found to be five- and 21-month SMAs. This system (not shown in a chart here) generated a total of 22 buy and 21 sell signals. The last buy signal was given in Nov 2003 and the system was still long when our test was concluded at the end of Jan 2004.
Trades using the monthly system would have earned 90% of the total Dow gains in 57% of the time (47 years). So, even though this test generated more than three times the number of trades, results were quite similar. The difference was that although trades were entered more quickly, often resulting in bigger gains, the increased number of signals resulted in more exposure to volatility and a greater percentage of losing trades.
Using the P/E SMA Indicator to Short
The next question we can tackle is whether the indicator would have performed if both long and short trades were taken. Entering a short trade of equal size each time a long position was sold would have given losses of 510 points in five short trades for an average loss of 102 points per trade. Based on the chart in figure 1, this makes sense: the linear regression channel shows that the market was in an overall uptrend from 1920, and as all good traders know, it is a bad idea to trade against the trend.
The P/E SMA indicator benefited the trader not so much by offering a straight trading system to generate both long and short trades, but by guiding the trader out of the market during periods of low or negative returns. As a simple long trade timing tool, it worked extremely well.
Taming the Market Beast
It is far easier to make money in a secular than cyclical bull market. A buy-and-hold strategy works well in the former but does not in the latter. It takes more skill and effort to make money when stocks are stuck in a trading range in which the price at the end and the beginning of the investment period are roughly equal.
For example, those who bought Dow stocks at the peak of the secular bull market in 1929 (and held them) did not begin to see profits until nearly 25 years later, in the latter part of 1954. Those buying at the secular bull peak in 1966 had to wait until 1983. It is imperative in these trading range markets that you avoid them altogether, unless you have developed an effective short-term trading system.
You know the well-known market adage: timing is everything. The P/E SMA indicator proves this point. It also demonstrates that the real worth of P/E ratios from a trading perspective is not so much in their absolute values. Those who exited the market in 1996 (at Dow 6448), when the P/E surpassed the prior 1966 bull market peak of 24, would have missed more than 5,000 points in profit in the subsequent three and a half years. Other than at rare extremes, absolute P/E values do not provide accurate entry and exit signals. A relative system combined with a method of detecting rapid change in direction does. But the major benefit of the P/E SMA indicator was found to be in keeping the trader out of the market when it was less profitable.
The next time someone tells you that P/E ratios don't matter, you'll have your answer ready. From a historical standpoint, they certainly do matter, especially if you know how to use them.