Equity yields

Abstract

We study a new data set of dividend futures with maturities up to ten years across three world regions: the US, Europe, and Japan. We use these asset prices to construct equity yields, analogous to bond yields. We decompose the equity yields to obtain a term structure of expected dividend growth rates and a term structure of risk premia, which decomposes the equity risk premium by maturity. We find that the slope of the term structure of risk premia is pro-cyclical, whereas the slope of the term structure of expected dividend growth rates is counter-cyclical. The comovement of yields across regions is, on average, higher for long-maturity yields than for short-maturity yields, whereas the variation in this comovement is much higher for short-maturity yields.

Introduction

There exists a large literature studying fluctuations of, and the information contained in, the term structures of nominal and real interest rates.² At each point in time, these term structures summarize pricing information of either nominal or real claims with different maturities. In this paper, we study a novel term structure of assets that are direct claims to future dividends paid by firms to shareholders. A long position in these so-called dividend futures implies that in exchange for a known payment due in n years from now, one receives the dividends paid on the underlying index over the year leading up to the settlement. Our data set is available at a daily frequency with maturities up to ten years, with 1-year increments. Dividend futures prices allow us to construct a term structure of equity yields that are analogous to real and nominal bond yields. The key difference between dividend futures and either nominal or real bonds is that the final payoff of dividend assets is variable whereas the payoff of nominal and real bonds is fixed in nominal and real terms, respectively. In this paper, we explore the information contained in equity yields across three major equity markets: the US, Europe, and Japan.

The equity yield at time t with maturity n can be written as the sum of three components. It consists of the nominal bond yield with maturity n, plus a maturity-specific risk premium that investors require for holding dividend risk, minus the expected dividend growth rate, which represents the average expected dividend growth over the next n periods. Higher discounting increases the yield, whereas higher expected dividend growth lowers the yield.³

Dividend assets, also called dividend strips, are generally traded in futures or swap markets, not in spot markets. Spot prices and futures prices are linked through bond prices. Assuming no-arbitrage, we can replace spot prices with futures prices in our computations to obtain forward equity yields, denoted by $e_{t,n}^f$, which do not depend on the n-year bond yield. The forward equity yield is simply equal to the difference between the maturity-specific dividend risk premium, which we denote by ${\theta }_{t,n}$, and the average n-year expected dividend growth rate $g_{t,n}$:

$$\underset{n-\mathrm{year}\mathrm{forward}\mathrm{equity}\mathrm{yield}}{\underbrace{e_{t,n}^f}}=\underset{\mathrm{risk}\mathrm{premium}}{\underbrace{{\theta }_{t,n}}}-\underset{\mathrm{expected}\mathrm{dividend}\mathrm{growth}}{\underbrace{g_{t,n}}}.$$

This implies that, by definition, forward equity yields must either predict dividend growth rates or excess returns (in excess of bonds) on dividend assets, or both. A high (low) value of the forward equity yield implies that the risk premium is high (low) or that the expected dividend growth rate is low (high). This makes forward equity yields natural candidates to forecast dividend growth across various maturities. We find that forward equity yields fluctuate strongly over time, for all maturities, and for all geographic regions. These fluctuations are due to both expected dividend growth variation and risk premium variation. Particularly during the great recession, 1-year forward equity yields turn strongly positive with values above 30% for the US, and values above 50% for Europe and Japan. We find that for all regions, expected dividend growth rates were low (negative) and risk premia were high during this period.

This paper is the first to compute and analyze the behavior of the full term structure of equity yields. Our new data set allows us to make two important additional contributions to the asset pricing literature. First, risk pricing across maturities has recently received a lot of attention. Important contributions in this literature are Lettau and Wachter (2007) and Hansen, Heaton, and Li (2008). In a recent paper, Binsbergen, Brandt, and Koijen (2012) show that, unconditionally, the average excess return over their sample (the estimated risk premium) is high for short-maturity dividend strips, which seems inconsistent with several leading asset pricing models. In this paper, we study the time variation in risk pricing (risk premia) across maturities. We find that the slope of the term structure of dividend risk premia moves in a strongly pro-cyclical fashion. During recessions, the slope of the risk premia seems particularly downward-sloping. The opposite holds for the slope of the term structure of expected dividend growth rates, which moves in a strongly counter-cyclical fashion. Hence, our paper contributes to a large literature documenting that the equity risk premium fluctuates over time.⁴ We use equity yields to study whether variation in the equity risk premium is largely driven by short- or long-maturity variation in risk premia and our results suggest it is the former.⁵

Second, we study the degree of comovement of dividend future returns across regions and compare it to the comovement in index returns across regions. On average, short-maturity dividend strip returns comove less compared to index returns, but the time-variation in this comovement is much higher. The average correlation of the 2-year dividend future returns across regions is only 0.4, but increases to 0.8 in 2008. The correlation between the index returns across the three regions is on average higher than the dividend futures returns, but does not change as much over time. We also study the time-variation in the Capital Asset Pricing Model (CAPM) betas of the 2-year and the 5-year dividend futures returns. We find that these betas are strongly time varying, and this time variation is decreasing with maturity. The CAPM betas increase substantially during the great recession. In summary, dividend futures seem to have large time-variation in their co-movement with each other as well as with asset markets in general, providing an interesting avenue for future research as to why these assets have such high risk premia. In this way, we extend the literature on comovement across regions⁶ by studying whether claims to short-maturity or long-maturity cash flows tend to comove more strongly.