Does the Trend Matter?

December 24, 1969

1273

I am fascinated by the growing body of research that revolves around the P/E10 ratio (and its siblings, the P/E5, P/E15 and so on), also known as the cyclically adjusted or normalized P/E ratio. In the past 5 years there have been numerous articles published at Advisor Perspectives as well as independently and in professional journals by numerous researchers: Robert Shiller, Doug Short, Wade Pfau, Michael Kitces, John Hussman, Crestmont Research, Jim Otar, the team of Mike Philbrick and Adam Butler, Rob Bennett and others.

Kitces in particular, in the April 2009 issue of The Kitces Report, proposes a tactical asset allocation investment strategy based on the current value of the P/E10 ratio. In the December 2011 issue of the Journal of Financial Planning he and colleagues Kenneth Solow and Sauro Locatelli propose a variation of the theme and subject the proposed strategy to a rigorous analysis to conclude that the approach is highly likely to lead to better risk adjusted returns.

To summarize my understanding of these strategies, the investor lowers or increases her allocation to stocks when the normalized P/E ratio hits certain value milestones. In essence, when stocks are “cheap” as per the normalized P/E ratio the investor holds more stocks and when the market it expensive as per the P/E ratio the investor pairs back her allocation to stocks. Kitces proposed strategies don’t take into account the long term trend of the P/E ratio. That is, the proposed strategy doesn’t call for investors to invest differently during uptrends versus downtrends. I am not criticizing here, I’m simply making observations (I, in fact, think very highly of Michael Kitces and, via email, commended him on his recent JFP article).

Doug Short’s monthly updated P/E10 chart, which so clearly shows the secular trends of the P/E10 ratio, really stimulates my imagination. Does the trend matter? Should investors be investing differently during P/E10 ratio downtrends versus uptrends? For example, once investors have increased/lowered their allocation to stocks, should they then ride the secular trend up/down to some degree that is suitable based on their risk profile? Others mentioned above besides Kitces do advocate for this though I have not seen an analysis of available stock market return data that supports trend informed investment strategies (it’s very possible I’ve simply missed such analyses and would appreciate receiving pointers to them).

To begin to answer this question it seems the first order of business is to determine if historical return data lends any support to the idea that the trend matters.

Professor’s Shiller’s database of monthly values affords an excellent starting point for delving into this question.

The first step in the analysis is to add a column to Shiller’s data set that flags each row of data as occurring in either an uptrend or a downtrend. I marked the uptrend rows with a value of 1 in the new column and the downtrend values with a value of 0. Trends are demarcated by a highest or lowest value for that particular trend (as seen through a rear view mirror).

It is worth noting that the P/E10 ratio didn’t start to exhibit widely swinging, long term trends until July of 1901. Thus, the July 1901 monthly data point serves as the start of the usable data for trend analysis. The first trend is a downtrend that starts with a P/E10 value of 25.24 (June 1901) and lasts until November of 1920 when a durable low P/E10 value of 5.13 was established.

The following table shows how I categorized each monthly data point in the Shiller Data base:

All month end data points from July of 1901 through June of 2011 are included in the analysis We have experienced 4 completed uptrends and 4 completed downtrends. We are currently 12 years into the 5th downtrend that follows a new, all-time high P/E10 value of 44.20. It seems reasonable to assume that we have years to go before we hit the bottom of the current trend.

The next step in the analysis is to calculate 3, 5 and 10 year real, future, total (price change with dividends reinvested) returns of the stock market for every monthly data point in the Shiller data set from July 1901 onward.

At this point in my analysis I observed that for each N-year window (3, 5, 10 years), some of the N-year periods fell totally in an uptrend or totally in a downtrend and that others started (finished) in a different trend than they ended (began). Truthfully, the Solow/Kitces/Locatelli 12/2011 JFP article alerted me to the fact that I should probably analyze the “pure trend” returns and the “hybrid trend” returns distinctly from each other (the JFP article focuses on the top and bottom deciles of historical P/E10 values as the triggers for asset allocation changes). So I re-categorized each monthly N-year return data point as occurring in either an uptrend, downtrend or hybrid trend.

With these categorized return values at the ready I could then calculate numerous measures that I thought would give us a first inkling of whether or not the trend matters. Table 2 contains these measures.

As a “first blush” take on the question of whether or not trend matters, these numbers are revealing.

I’d like to highlight what I think are the most telling numbers, specifically, those in the last three rows (bolded). Figures 1-3 show the Average Return, Average Standard Deviation, and Average Sharpe ratio data points from the table above.

The average annualized 10 year future return during P/E10 ratio downtrends is less than 1%! Downtrends would seem to pose substantial challenges even for long term investors.

And, while average returns vary quite a bit between the trends (Figure 2), standard deviation doesn’t which leads to quite divergent Sharpe Ratios (Figure 3)!

It seems fair to say that 3, 5 and 10 year risk adjusted future returns tend to be quite a bit higher (as do absolute returns) during uptrends versus down trends!

Kitces 12/2011 JFP article finds that making tactical assets allocation changes when a normalized P/E ratio (P/E5 in this case) is in the top or bottom decile of historical values (i.e. during hybrid periods) can lead to higher risk adjusted returns for the investor. The measures in the Table 1 above lends support to the hypothesis that better risk-adjusted returns through tactical asset allocation should also be easier to obtain in normalized P/E uptrend periods than downtrend periods.

There is yet another way we can analyze the Shiller data to gain perspective about whether the trend of the P/E10 ratio matters. I’ll call the next exploration “recovery analysis”.

For our purposes, recovery analysis involves analyzing how long it takes to recover a previous stock market high after the market falls below that high. Because Shiller’s data forms the starting point of the analysis, the highs will be monthly market highs and the recovery periods will be measured in months. Also, the highs are inflation adjusted and total return (price + dividends) highs. Again, the time period analyzed is June of 1901 through June of 2011 (last period for which final dividend return numbers are available).

The methodology of this analysis is straightforward. For every monthly data point where the market hits a new high we determine if the month falls in a downtrend or an uptrend and then we count the number of months until the market hits another new high. At first it might seem counterintuitive that the market could reach a new high during a P/E10 ratio downtrend. When inflation and dividend returns are factored in to the analysis this can and does happen.

Once again, we want to notice the difference between uptrends and down trends. Table 2 shows the results of this analysis.

The DT Max Recovery period of 12 years (144 months) unfolded from January of 1973 through December of 1984.

These measures also strongly suggest that downtrends can be painful periods in which to be invested in the stock market. The minimum recovery time is the same for each trend but the maximum recovery time is just shy of three times longer in downtrends!

Interestingly we may be on course to establish a new record for number of months to recovery. The most recent inflation adjusted market high occurred in August of 2000 which (counting back from June 2011) was 131 months ago and the market is 23% below that high (on an inflation adjusted basis).

Conclusion

This rudimentary analysis suggests that more research into making asset allocation decisions based on the trend of the P/E10 ratio might prove worthwhile.

Such future research might include statistical significance testing, calculating up and down trend Sortino ratios, measuring the Sharpe and Sortino ratios for tactically allocated stock+bond portfolios and studying how trend-sensitive asset allocation strategies would have fared in the past. Finally, Kitces recent JFP article also suggests that studying the P/E5 ratio may also have some value.

References

Kitces, Michael E. 2009. “Dynamic Asset Allocation and Safe Withdrawal Rates.” The Kitces Report April.

Solow, Kenneth R., Kitces, Michael E. and Sauro Locatelli 2011. “Improving Risk-Adjusted Returns Using Market-Valuation-Based Tactical Asset Allocation Strategies.” The Journal of Financial Planning 24, 12 (December): 48-59.