Stock return can be divided into two components: payout yield and capital gains yield. Dividend yield (DY), as the major component of the payout yield, is one of the most common predictors of future stock return. It seems reasonable that variation in DY should be related to future dividend growth or future returns.

In recent years, DY has become more persistent and does not closely track real economic activity. The authors attempt to use net payout yield (NPY), which equals DY plus repurchase yield minus issuance yield, to predict future returns and find that it outperforms DY and other alternative payout measures.

Although DY and NPY often co-move, their correlations with contemporaneous shocks are different. DY is strongly negatively correlated with the shocks, but NPY shows only weak negative correlation. As a result, a long-horizon portfolio calibrated by DY requires larger hedging demand than one calibrated by NPY during the shocks.

Because the return predictability of NPY outweighs that of DY, an investor in a DY-based portfolio may be willing to pay up to a 4% annual management fee to switch to an NPY-based portfolio. Investors with lower risk aversion and longer horizons are willing to pay higher fees.

Both univariate and multivariate predictive regressions of excess stock returns on alternative yields are used to measure the predictabilities of payout yields. Five alternative yields are considered, including DY, NPY, issuance yield, and net issuance yield. The sample period covers 1927–2014. After bias correction, the predictability of NPY (measured by R²) is around 20%, which is more than double that of total payout yield (TPY) and more than triple DY.

Out-of-sample analysis is also conducted to avoid potential data mining and check the robustness against structural change. Both rolling and recursive estimations are considered, and the results show that NPY achieves the highest out-of-sample R² among alternative measures.

Rather than testing the predictability of payout yields on returns and cash-flow growth separately, the authors adopt a joint test that exploits information in slope estimates for returns and dividend growth together. The approach links two dependent variables by the vector autoregression (VAR) system. By assuming no predictability of the payout measures (null hypothesis), the authors estimate slope coefficients under 10,000 simulated samples. There is strong evidence against the no-predictability null hypothesis. Evidence shows that the “issuance growth” component of NPY has strong predictability in the joint test.

Lastly, by taking risk aversion, investment horizons, and utility into consideration, the authors conduct a long-horizon portfolio choice analysis. By estimating the VAR system of equations and using historical data to calibrate VAR parameters, they compare the hedging demand of portfolios under various payout measures. They estimate the annual management fee of switching from a dividend-based portfolio to an NPY-based portfolio considering the superior performance of the NPY-based portfolio.

The authors show that NPY has better predictability than TPY. In fact, some previous researchers have observed that negative stock returns are associated with share issuance since 1970. Signaling can explain the negative relationship. Because firm management has more information about the true value than the general public, it may take this advantage to issue new shares when the stock is overpriced. Share issuance may also imply debt overload. Because the cost of debt is normally cheaper than the cost of shares, raising capital by issuing shares may imply difficulty in issuing low-cost debt, which leads to a stock price decrease.

In short, NPY seems to be a more accurate predictor of future returns when compared with DY and TPY. The reasons behind their difference need to be carefully examined.