The Dog Has Barked For A Long Time: Dividend Growth Is Predictable
University of Denver – Reiman School of Finance
University of Denver – Reiman School of Finance
March 15, 2016
The Campbell-Shiller present value identity implies that (i) the dividend-price ratio should forecast returns or dividend growth, and (ii) any other predictor of returns should also forecast dividend growth, and vice versa. Prior literature investigates (i) and finds that dividend growth appears unpredictable based on forecasting regressions with the dividend-price ratio. We investigate (ii) and combine out-of-sample dividend-growth forecasts from 14 individual predictive regressions based on common return predictors. Encompassing and structural break tests show these individual return predictors contribute disparate and unstable univariate forecasts of dividend growth. Combination forecast methods of 14 common return predictors overcome these econometric problems and generate robust out-of sample predictability of dividend growth over the entire post-war period as well as over most sub-periods. The dividend-growth forecasts coupled with the dividend-price ratio also significantly forecast excess returns out-of-sample.
The Dog Has Barked For A Long Time: Dividend Growth Is Predictable – Introduction
Following the present value identity of Campbell and Shiller (1988), a large literature argues that time variation in the dividend-price ratio of the aggregate stock market necessarily represents news about expected future returns (“discount-rate news”) and/or news about expected future dividend growth (“cash-flow news”). Accordingly, many studies in this literature investigate whether the dividend-price ratio forecasts returns or dividend-growth.1 Regression-based tests frequently fail to find that the dividend-price ratio predicts dividend growth and conclude that time variation in the dividend-price ratio primarily results from time variation in expected expected returns (see, e.g., Cochrane (2008), Cochrane (2011)). However, these regression-based tests suffer from at least two econometric problems that limit their inferences about dividend-growth predictability, which is important as it is a key input to equity and contingent-claim valuation.
First, only using the dividend-price ratio in dividend-growth-forecasting regressions unnecessarily limits the set of candidate predictive variables. The prior literature on return predictability (see, e.g. Rapach, Strauss and Zhou (2010)) finds that multiple variables forecast returns and these predictors should also forecast dividend growth by the Campbell and Shiller (1988) present value identity (see e.g. Cochrane (2011) and Koijen and Van Nieuwerburgh (2011)). Second, forecasting regressions based on the dividend-price ratio exhibit statistical biases due to the persistence of the dividend-price ratio (see, e.g., Stambaugh (1999)) as well as structural breaks (see, e.g. Lettau and Van Nieuwerburgh (2008) and Koijen and Van Nieuwerburgh (2011)) that limit their out-of-sample reliability.
In this paper, we investigate whether dividend-growth is predictable out-of-sample by using combination forecast methods. Our combination forecasts are weighted averages of the out-of-sample univariate dividend-growth forecasts using 14 common return predictors identified by Goyal and Welch (2008). Stock and Watson (2003), Stock and Watson (2004), Elliott, Granger and Timmermann (2006), and Rapach et al. (2010) find that combination forecast methods overcome the two sets of econometric problems cited above. They provide more stable and reliable out-of-sample forecasts of macro variables and returns from the relatively unstable univariate forecasts.
We find that the dividend-price ratio as well as other common return predictors fail to individually predict dividend growth out-of-sample. However, combination forecasts of these predictors generate significant out-of-sample evidence of dividend-growth predictability for horizons of one or two years over the entire post-war sample period. The equal-weighted average of the different combination forecasts predicts dividend growth with an out-of-sample R2 of more than 7% at the one-year horizon. Moreover, we find that different weighting schemes that define each combination forecast all generate forecasts with significantly positive out-of-sample R2 statistics. Goyal and Welch (2008) find return forecasting relationships change across time and the same is likely true for those of dividend growth. Hence, we also show the combination dividend-growth forecasts provide significant out-of-sample predictability over most subsamples. Encompassing tests reject the null that the dividend-growth forecasts based on the 14 different predictive variables subsume each other. These tests highlight the shortcomings of relying solely on forecasts from a single predictor of dividend growth, further motivating and showing the effectiveness of combination forecast methods.
Lettau and Van Nieuwerburgh (2008) argue that it is hard to model structural breaks in return and dividend growth forecasts based on the dividend-price ratio in real time. When structural breaks are not perfectly correlated across univariate forecasts, combination forecast methods can mitigate the instability through a diversification-type effect (see, e.g. Elliott et al. (2006)). For example, Rapach et al. (2010) find that return forecasts based on the 14 Goyal and Welch (2008) predictors exhibit structural breaks, but also that combination forecasts of these predictors overcome this instability resulting in significant out-of-sample return predictability. Using Bai and Perron (1998) tests, we show that dividend-growth forecasts with the 14 Goyal and Welch (2008) predictors also exhibit structural breaks, further explaining the superior performance of combination forecast methods for dividend growth.
See full PDF below.