[vc_column_text disable_pattern=”true” align=”left” margin_bottom=”0″ width=”1/1″ el_position=”first last”]
In “Investment Performance and Price-Earnings Ratios: Basu 1977 Revisited” I reexamined Basu’s 1977 work “Investment performance of common stocks in relation to their price-earnings ratios: a test of the efficient market hypothesis.” Basu examined data on common stock returns during the period September 1956 – August 1971, included an average of 500 assets for each of the 14 years, and found that the returns on common stocks were inversely, linearly, and statistically significantly related to their price-earnings (P/E) ratios. To put it another way, low P/E ratios predict future higher than normal returns; this suggests that an investor could follow Basu’s methodology to get high returns. To do so, the investor would need to group stocks into five portfolios based on the P/E ratios of those firms, then purchase the portfolio with the lowest P/E. By rebalancing this strategy at given intervals (monthly, quarterly, or annually as in Basu’s work) an investor could consistently out-perform the market average rate of return.
In my paper I used data on assets selected from the 2012 Russell 3000 list covering the period November 2002 – December 2012; the selection process, while introducing some degree of across-the-board upward bias on returns, ensured that assets included in the sample would appeal to a rational investor, and excluded those which would not. Ultimately, once firms which did not exist for the entire period, or failed to report financial data were removed, the assets of 1671 firms were included in the final dataset with data for every month in the dataset. While monthly returns in the dataset closely followed a normal distribution, typical in stock returns this dataset exhibited a slight degree of negative skewness (-0.3623) and some excess kurtosis (1.4496). These deviations from a standard normal distribution are assumed to be minor enough in degree to treat the data as normal, and therefore, standard errors are used throughout the study. Characteristics of the data also follow previous studies including Fama and French’s 1993 work, “Common risk factors in the returns on stocks and bonds.”
I examined the data in a similar fashion to Basu, adhering as closely as possible to his methodology albeit with a more frequent (monthly) portfolio rebalance. This required ranking assets by P/E ratio and creating 5 portfolios, or quintiles, for each month of the sample. My work revealed a statistically significant non-linear relationship between P/E ratio portfolio and asset returns, with return generally increasing as P/E ratio increases. To put it another way, in the period 2002 – 2012, higher P/E ratios can be seen as predicting higher returns to a degree. The non-linearity of the relationship suggests that there is some optimal P/E ratio, and that to obtain the highest predicted return an investor should invest in the second highest P/E portfolio.
My work contradicted Basu, finding not only an essentially opposite relationship between price-earnings ratios and common stock returns, but a non-linear relationship as well. Assuming that the data from these two periods are not, in some way, vitally abnormal, the conclusions which can be reached from these results are:
1) Stock returns during the period 2002-2012 can be predicted by price-earnings ratios
2) In general, predicted stock returns are higher for higher price-earnings ratios, however, that relationship is non-linear in nature
3) The highest asset returns are displayed in 2nd highest quintile (by P/E) ratio firms
In short, were an investor during the period of 2002-2012 to have followed the methodology outlined in this study, following Basu’s 1977 work, using price-earnings ratio quintiles to determine investments, that investor would have earned a monthly average of 0.505% (1.345% excess of risk-free rate) more than the market monthly excess rate of return (0.839% in excess of the risk-free rate).
[/vc_column_text] [vc_column_text disable_pattern=”true” align=”left” margin_bottom=”0″ width=”1/1″ el_position=”first last”]
This represents the thoughts and assertions of the author. It does not represent any position taken by Strata or it’s affiliates.