Income Hedging, Dynamic Style Preferences, and Return Predictability

• Author: ALOK KUMAR,STEFANOS DELIKOURAS,JAWAD M. ADDOUM,GEORGE M. KORNIOTIS
• Date: 01 August 2019
• Published date: 01 August 2019
• DOI: http://doi.org/10.1111/jofi.12775

THE JOURNAL OF FINANCE •VOL. LXXIV, NO. 4 •AUGUST 2019

Income Hedging, Dynamic Style Preferences, and

Return Predictability

JAWAD M. ADDOUM, STEFANOS DELIKOURAS, GEORGE M. KORNIOTIS,

and ALOK KUMAR∗

ABSTRACT

We propose a theoretical measure of income hedging demand and show that it affects

asset prices. We focus on the value factor and first demonstrate that our demand esti-

mates are correlated with the actual demands of retail and mutual fund investors. We

then show that the aggregate high-minus-low (HML) demand predicts HML returns.

Exploiting the state-level variation in income risk, we demonstrate that state-level

hedging demands predict state-level HML returns. A long-short portfolio that exploits

this hedging-induced predictability earns an annualized risk-adjusted return of 6%.

INCOME RISK IS A FUNDAMENTAL source of uncertainty that households face, and

hence it should affect their financial decisions. Several recent studies demon-

strate that income risk influences the portfolio decisions of U.S. and Euro-

pean households. For example, Angerer and Lam (2009) demonstrate that U.S.

households with higher permanent income risk invest less in risky assets.

Betermier et al. (2012) show that the financial decisions of Swedish households

are sensitive to wage volatility.

A related literature provides household-level evidence that portfolio deci-

sions are affected by income hedging considerations. Bonaparte, Korniotis, and

Kumar (2014) find evidence of income hedging in the decisions of Dutch and

U.S. households. They show that households whose income growth is highly

correlated with market returns participate less in the market and allocate

less of their wealth to risky assets. Using Swedish data, Betermier, Calvet,

and Sodini (2017) argue that the tilt of some investors toward value versus

growth stocks is motivated by income hedging concerns. They show that income

∗Jawad M. Addoum is with S.C. Johnson College of Business, Cornell University. Stefanos De-

likouras, George M. Korniotis, and Alok Kumar are with University of Miami Business School. We

thank three anonymous referees; an anonymous Associate Editor; Sandro Andrade; Tim Burch;

Sean Campbell; Vidhi Chhaochharia; Jason Delaney; Andrew Karolyi; David Ng; Tarun Ramado-

rai; Kenneth Singleton (Editor); Luis Viceira; Deniz Yavuz; and seminar participants at Cornell

University,the Federal Reserve Board in Washington DC, Florida International University,Goethe

University Frankfurt, University of Mannheim, University of Miami, University of Warwick, the

2016 EFA Annual Meeting, NTA 106th Annual Conference on Taxation, and 2016 SFS Finance

Cavalcade for helpful comments and valuable suggestions. We are responsible for all remaining

errors and omissions. The authors have no conflicts of interest as identified by the Journal of

Finance Disclosure Policy.

DOI: 10.1111/jofi.12775

2055

2056 The Journal of Finance R

hedging considerations are strongest for those households with high human

capital and high exposure to aggregate risk.1Korniotis and Kumar (2011)

further demonstrate that certain types of financial assets facilitate income

smoothing and improve risk sharing across U.S. states.

Motivated by the household-level evidence on income hedging, in this pa-

per, we examine its potential asset pricing implications. In particular, we in-

vestigate the potential link between the aggregate income hedging behavior

of U.S. households and the predictability in stock returns. Our primary con-

jecture is that the demand for financial assets that facilitate income hedging

varies as their hedging potential changes over time. If such hedging-induced de-

mand shifts are systematic, they could affect prices. This conjecture is partially

motivated by dynamic portfolio choice models, which demonstrate that when

investment opportunities are time-varying, investors should rebalance their

portfolios as they learn about future investment opportunities (e.g., Campbell

and Viceira (1999), Campbell and Vuolteenaho (2004), and Jurek and Viceira

(2011)).

We connect this basic prediction of dynamic portfolio choice models with the

literature on style investing, which shows that investors systematically move

in and out of certain investment styles (e.g., Barberis and Shleifer (2003),

Kumar (2009), and Wahaland Yavuz (2013)). We argue that style shifts between

value and growth stocks might be motivated in part by the time-variation in

income hedging opportunities. Such systematic demand shifts could generate

predictable variation in the returns of value and growth portfolios.

We formalize this economic insight in a model that combines the noise trader

specification of De Long et al. (1990), the Bayesian framework of Barberis

(2000), and the income hedging model of Viceira (2001). In the model, there

are two groups of investors. The first group comprises workers, who receive

stochastic income and are concerned about income risk. The second group is

the sophisticated investors, who are risk neutral and face transaction costs.

We solve the model analytically, and obtain explicit solutions for the optimal

portfolios of workers and sophisticated investors. Similar to Viceira (2001),

asset demands by workers include an income hedging term, which depends on

the covariances between asset returns and income, scaled by the covariance

matrix of asset returns. This covariance-based term is our proposed measure

of income hedging demand (IHD).

When the conditional covariance between income and an asset decreases,

the income hedging potential of this asset increases and the IHD measure

increases. Workers then demand more of the asset, creating price pressure,

which in our model cannot be fully absorbed by the sophisticated investors

due to transaction costs. This price pressure today generates lower returns in

the future. Overall, the key theoretical prediction is that there is a neg ati ve

relation between our IHD measure and future asset returns.

1Other papers related to income hedging include Heaton and Lucas (2000), Viceira (2001), and

Vissing-Jorgensen (2002).

Income Hedging, Dynamic Style Preferences, and Return Predictability 2057

To test our predictability hypothesis, we focus on the value factor or the

high-minus-low (HML) style portfolio. We focus on the HML factor because ex-

isting literature suggests that income risk affects the choice between value and

growth stocks. For example, studies using aggregate data find that the value

premium is related to income hedging. In particular, Koijen, Lustig, and

Nieuwerburgh (2017) show that value stocks are especially risky because their

cash flows are low during deep recessions, that is, during periods when aggre-

gate income growth is very negative. Thus, investors could hedge against this

risk by investing in growth stocks.

Betermier, Calvet, and Sodini (2017) provide direct evidence of income hedg-

ing using household-level data. They find that households with high income

risk or high human capital and households employed in highly cyclical sectors

avoid value stocks and prefer growth stocks. Cronqvist, Siegel, and Yu (2015)

also find that value versus growth investing is one of the most predominant

investment styles. Further, they show that households with higher human cap-

ital (i.e., high labor income and high education) tend to prefer growth stocks.

Moreover,investors with more procyclical income (i.e., high correlation between

labor income growth and GDP growth) also tend to prefer growth stocks.

We also focus on the HML portfolio because it satisfies the necessary con-

ditions required to generate predictability in returns. First, value and growth

investment styles have been popular as far back as the 1930s (e.g., Graham and

Dodd (1934)). Further, the average investor can easily gain exposure to value

and growth portfolios (Wahal and Yavuz (2013)). Jurek and Viceira (2011) re-

port that in the universe of mutual funds in the Center for Research in Security

Prices (CRSP) database, about 78% of funds can be categorized as either value

or growth funds. Second, the income hedging potential of the HML portfolio

varies significantly over time, which generates time-varying demand shifts

that can potentially influence stock prices.

Since hedging-induced investments cannot be directly observed, we estimate

the IHD using the relation predicted by our model. Because we want to examine

the aggregate asset pricing implications of income hedging, we construct the

IHD using aggregate U.S. and state-level income data. Specifically, we adopt a

conditional estimation approach where we compute variances and covariances

using a 10-year rolling window. Using these conditional moment estimates we

compute IHD at both the U.S. state level and the national level. As shown in

Figure 1, the IHD estimates exhibit substantial variation in the cross section

and over time. For instance, in every period some U.S. states have negative

while others have positive IHD estimates.

Before using our IHD measure to predict asset returns, we investigate

whether these theoretical demand estimates are related to actual investor

demands for value and growth stocks. We perform two validation tests. In the

first test, we compute the relative portfolio weights in value and growth stocks

based on the actual stock holdings of a sample of retail investors. We find that

the correlation between the actual relative weight and our IHD measure is

positive and statistically significant.