A measure of cash flow duration derived directly from balance sheet data supports evidence for a downward-sloping term structure, which is inconsistent with what is implied by commonly used asset pricing models. The author finds that low cash flow duration stocks earn 1.10% per month higher returns compared with high cash flow duration stocks and that standard factor models can explain only half of this return differential.
How Is This Research Useful for Practitioners?
Cash flow duration for stocks in this research resembles the Macaulay duration for bonds. That is, cash flow duration is the weighted-average time to maturity of the cash flows generated by the stock. For a more concrete example, growth and value stocks can be compared. Growth stocks, which pay more of their cash flows in the long-term future, have higher cash flow duration. By contrast, value stocks, which pay more of their cash flows in the near term, have lower cash flow duration.
From a relative valuation perspective, practitioners would be interested in the apparent mispricing between low and high cash flow duration stocks and the fact that this mispricing can be identified using a cash flow duration metric derived directly from balance sheet data. The magnitude of the mispricing is also quite large; the author finds that low cash flow duration stocks earn 1.10% per month higher returns than high cash flow duration stocks. This return differential is three times higher after periods of high investor sentiment. Further analysis reveals that this mispricing is limited to stocks that are short sale constrained, which implies that investors’ ability to take advantage of this mispricing may be limited.
Another useful finding for practitioners is the relationship between cash flow duration and optimistic analyst behavior. In particular, the author finds that analysts have overly optimistic growth forecasts for high-duration stocks, analysts extrapolate from past earnings growth into the future, and high-duration stocks have negative earnings surprises and seem to engage in earnings management. The author also finds that analyst forecasts for expected returns are upward sloping after periods of high investor sentiment and downward sloping after periods of low investor sentiment, which is at odds with what is observed ex post.
How Did the Author Conduct This Research?
The author uses standard financial databases for his data, including stock price data from the CRSP monthly stock file and balance sheet data from S&P’s Compustat database. He uses the Thomson Reuters 13F database for institutional ownership data. Analyst forecasts for EPS and earnings growth are retrieved from I/B/E/S, and the Fama–French factors are obtained from Ken French’s website.
The full sample period is July 1963 to June 2014, but the sample is then restricted to various subperiods. For example, the most restrictive subset is that of I/B/E/S forecasts on price targets—July 2001 to June 2014. To minimize the impact of outliers, the author winsorizes all variables at the 1% and 99% levels.
The author develops a measure of cash flow duration that resembles the traditional Macaulay duration for bonds, which reflects the weighted average time to maturity of the cash flows. The weight the author uses is the ratio of discounted cash flows to price. Unlike bonds, stocks do not have finite maturities, so the author splits his duration formula into a finite forecast period of 15 years and an infinite terminal period. The terminal value is calculated as a level perpetuity payout. Also unlike bonds, stocks’ cash flows are not known in advance, so the author assumes clean surplus accounting, which implies that all changes in the book value of equity aside from ownership transactions are reflected in income. This approach also forecasts cash flows via forecasting return on equity (ROE) and growth in book value of equity. Both these variables are calculated directly using financial statement data and are modeled as autoregressive processes with parameters using the combined CRSP–Compustat universe. ROE is assumed to mean revert to the average cost of equity. The growth in book equity is assumed to mean revert to the average growth rate of the economy.
The author also carries out a number of robustness tests to confirm his findings by using firm size and book to market value as conditional variables. His conclusion remains the same.
The Fama–French three-factor model has unseated the CAPM for routine risk adjustment in academic empirical research. The CAPM includes only the market portfolio, whereas the Fama–French three-factor model adds size and value premiums as explanatory variables. Much research has shown that since the mid-1980s, a size premium no longer exists. It appears that one implication of this research is that value stocks (i.e., those with low cash flow duration) outperform growth stocks (i.e., those with high cash flow duration) but only for stocks that are short sale constrained. Put differently, the author suggests that the value premium may be limited to short-sale-constrained stocks. What implications do these findings—that the two additional factors (i.e., size and value) do not appear to matter as much as originally thought—have for the Fama–French three-factor model?