Decomposing the accrual premium: The evidence from two markets

• Published date: 01 July 2019
• DOI: http://doi.org/10.1111/jbfa.12394
• Author: Anthony Holder,Doina Chichernea,Alex Petkevich
• Date: 01 July 2019

DOI: 10.1111/jbfa.12394

Decomposing the accrual premium: The evidence

from two markets

Doina Chichernea1Anthony Holder1Alex Petkevich2

1Reiman School of Finance, Daniels College of

Business, University of Denver

2John and Lillian Neff Department of Finance,

College of Business and Innovation,University of

Tol ed o

Correspondence

AlexPetkevich, University of Toledo,Mail Stop

103,2801 W. Bancroft St., Toledo,OH 43606,

USA.

Email:alexey.petkevich@utoledo.edu

Fundinginformation

Universityof Toledo

Abstract

We decompose the accrual premium and study its components

in the debt and equity markets. We show that the importance of

each accrual component depends on the sample and the type of

market considered. The short-term accruals component is primarily

observed in equity markets, among small and young companies,

which is consistent with mispricing arguments. The long-term

accruals premium is consistently positive and significant in different

samples and markets. This component reflects growth in capital

expenditures, and it is counter-cyclical and predictable, which is

in line with investment-based explanations. Finally, the financial

accruals component does not generate predictability.

KEYWORDS

accruals, debt, equity, investment, long-term accruals, short-term

accruals

JEL CLASSIFICATION

G11, G12

1INTRODUCTION

Economists have traditionally relied on financial statement information to make inferences about future expected

returns and about the overall financial health of any given company. In particular, accruals draw the attention of

researchers due to their strong predictive connection with returns (originally documented bySloan, 1996). The return

predictability of accruals is one of the most robust anomalies in the asset pricing literature, which is present in all size

groups, is not related to sentiment proxies, and is not affected by credit risk (Avramov, Chordia, Jostova, & Philipov,

2013; Fama & French, 2008, 2016; Jacobs, 2015).

Recently,several studies have brought the accrual anomaly back under scrutiny. Mashruwala, Rajgopal, and Shevlin

(2006) present evidence that low accruals stocks tend to have relatively higher volatility,trading costs, and lower liq-

uidity, making this strategy difficult to realize. Green, Hand, and Zhang (2017) present a comprehensivestudy of 94

different stock characteristics and show that most fundamental accounting information variables (including accruals)

do not provide independent information about equity returns. On the contrary,Hou, Xue, and Zhang (2018) compile

a replication study for 452 anomalies and report that fundamental information based anomalies replicate better than

J Bus Fin Acc. 2019;46:879–912. wileyonlinelibrary.com/journal/jbfa c

2019 John Wiley & Sons Ltd 879

880 CHICHERNEA ET AL.

tradingfrictions anomalies. 1Giventhis debate, understanding the driving forces behind the predictive power of accru-

als becomes increasingly important. The two most popular explanations of the accrual phenomenon in the literature

either relate to the persistence hypothesis (Richardson, Sloan, Soliman, & Tuna, 2005; Sloan, 1996) or to the invest-

ment Q-theory hypothesis (Wu, Zhang, & Zhang, 2010).2

In this paper,we attempt to identify the information captured by accruals in order to better understand the source

of predictive power that this fundamental characteristic has for returns. In order to address this question, we con-

sider a comprehensive definition of accruals and we study the difference in the accrual anomaly in the debt and equity

markets. We argue that conflicting results in the previous literature are partly generated byinconsistent definitions

of accruals used across various studies which make comparison difficult.3Not all types of accruals predict returns in

a similar fashion. We use the Richardson et al. (2005) methodology to decompose accruals into short-term (working

capital), long-term (non-current), and financial components of accruals. Based on our results, the predictability of the

short-term (working capital) accrual is primarily observed among relatively small and young companies and it is more

likely related to mispricing. The long-term (non-current) component of accruals persistently drives the accrual pre-

mium across different samples and different markets and is consistent with the investment-basedexplanation. Finally,

neither shareholders nor bondholders react significantly to the financial accruals component.

We also argue that using this comprehensive definition to examine the behavior of the accrual anomaly and its

components in both equity and credit markets provides us with a good setting to differentiate between mispricing

and risk-based explanations provided for the accrual phenomenon.4The bond market investors are primarily large,

sophisticated institutional investors, and, therefore, “mispricing” anomalies are less likely to be observed in this

environment. Using TRACE, Bloomberg, and FISD databases, we compile a comprehensive dataset of bond returns

containing on average 2,660 corporate bonds per year (corresponding to about 730 firms) from 1995 to 2016.

Overall, our results strongly support an investment-basedexplanation for the accrual phenomenon and can be briefly

summarized as follows:

1. The long-term component of accruals drives the predictive power of accruals for returnsin both equity and credit markets.

We document that the predictability of bond returns is driven only by the long-term (non-current) component of

accruals. The working capital and financial components of accruals have no significant relation with future bond

returns.5Moreover, when further decomposing the long-term component of accruals, we observe that the asset

part (related to investmentpolicy) primarily drives the results. For equity markets, the working capital accrual com-

ponent has predictive power mainly among small and young companies, which is consistent with mispricing argu-

ments. However, the long-term accrual premium is consistently positive and significant in both debt and equity

markets.

1Specifically, Hou, Xue, and Zhang (2018) confirm that accrual- and investment-basedanomalies provide significant returns, albeit smaller than originally

documentedin the literature.

2According to the reliability argument, the least reliable components of accruals – working capital accruals (ΔWC) and non-current accruals (ΔNCO)–are

mostlikely to be mispriced by investors and therefore drive the relation with returns. Based on the Q-theory proposed by Wu et al. (2010), the working capital

and non-current accruals are highly related to growth and, in turn, to the investment risk of the firm. In this context, accruals predict returns because they

containinformation about the investment decisions of the firm.

3Forexample, the original working capital accruals measure proposed by Sloan (1996) and used by Green et al. (2017) and Chordia, Goyal, Nozawa, Subrah-

manyam, and Tong(2017) incorporates the effect of depreciation expense, which is more closely related to growth in long-term operating assets. However,

this measure misses other important elements related to the long-term component of accruals, which we argue is the main determinant of the predictive

powerof accruals and the most robust across markets. Given the different type of information captured by each element and the implications that can lead to

erroneous/conflictingconclusions, it is important to clarify the differences in alternative definitions. See further discussion on this topic in Section 3.2.

4We are not the only paper to study the accrual phenomenon simultaneously in bond and equity markets.For example, Chordia et al. (2017) study several

equitycharacteristics (including accruals) in bond and equity markets. Since they use the reduced Sloan definition of accruals, their conclusion is that accruals

do not impact bond returns. Our use of a more comprehensive definition of accruals leads to a different conclusion related to the long-term component of

accruals,which is consistent with investment-related variables found significant in the aforementioned study.

5Thisresult confirms and extends Bhojraj and Swaminathan (2009), whose sample period ends in 1997.

CHICHERNEA ET AL.881

2. The long-term component of accruals is related to growth/investment risk. Consistent with the implications of Q-theory,

we show that expected returns to accrual-based trading strategies are counter-cyclical and time-varying in the

bond markets.6The predictability of accruals in the credit marketis significantly higher during bad states and low-

accrual portfolios have significantly higher exposureto growth risk.7These results are particularly pronounced for

the long-term component of accruals, which is more likely related to investments.

3. Apparent differences in accrual predictability between equity and bond markets aredriven by sample characteristics.We

investigate what drives the apparent difference in the accrual premium between the bond and equity markets.

Once we control for differences in sample characteristics by using a matched equity to bond sample, we show

that both bond and equity markets produce similar return predictability. Specifically, in the matched sample, the

equity accrual premium is also only driven by the non-current component of accruals, and current working capital

does not affect equity returns.8Similarly,our results show that the non-current component of accruals is related to

growth-related risk in the equity market. On the contrary,the working capital accruals are less likely to be driven

by investments risk. However,this component is important and significant, especially for smaller, less established

firms.9

Our paper makes a number of important contributions to the literature. First, our results add to the debate about

the sources of predictability of accruals. While we confirm the proposition by Richardson et al. (2005) that working

capital accruals are more likely to be driven by mispricing, we document that the most robust predictor of returns is

the long-term accrual component. Further decomposing long-term accruals, we observe that the asset part, related

to investment policy,primarily drives the results. We provide additional support for the Q-theory explanation, show-

ing that non-current accruals are highly related to investment risk and their premium is related to growth risk. These

findings are consistent across marketsand sub-samples. We also document that the short-term accrual premium is pri-

marily observed among small and young companies and nonexistent among large and mature firms, which challenges

the reliability theory. Therefore, we contribute to the extensiveline of literature debating the potential explanations

for the accrual anomaly.

Second, our results shed light on the role of accrual information as a driver of future bond returns. Wecontribute to

this stream of literature by showing that the accrual premium in bonds is significant and primarily driven by the long-

term component of accruals. We further show that this premium is time-varying and counter-cyclical. Compared to

other studies looking at accruals and bond returns (e.g. Bhojraj & Swaminathan, 2009; Chordia, Goyal, Nozawa, Sub-

rahmanyam, & Tong,2017), the main advantage of our approach is using a more comprehensive measure of accruals,

which allows us to clarify apparently conflicting results in previous literature. The distinction is important because dif-

ferent types of accruals provide different information and thus lead to different conclusions in terms of returns. For

example, we show that the main reason that the working capital accrual measure advocated by Sloan (1996) drives a

premium in both equity and bond markets is because this measure incorporates the effect of depreciation (which cap-

tures information related to long-term assets and investments). By separatingshort-term and long-term accrual infor-

mation, we reconcile seemingly conflicting results. Forinstance, while Green et al. (2017) and Chordia et al. (2017) use

the Sloan definition of accruals and conclude that accruals do not provide useful information about equity and bond

returns (respectively), both papers agree that investment-relatedcharacteristics do matter for predicting returns. This

6The intuition for this hypothesis is similar to the argument put forth by Zhang (2005) related to the value premium, which builds on the notions of costly

reversibilityand countercyclical price of risk. Chichernea, Holder, and Petkevich (2015) show similar results for equity markets.

7Weuse cross-sectional return dispersion (RD) to capture growth risk. See Section 3.3 for a more detailed discussion of the return dispersion variable.

8Forexample, Richardson et al. (2005) show that both ΔWC and ΔNCO generate significant hedge portfolio returns when it comes to equity. On the contrary,

weshow that, for our matched sample, ΔNCO is the only significant source of positive returns for the total accrual hedge portfolio.

9Firms that haveaccess to credit markets are larger, more mature firms. Our results show that the short-term (working capital) accrual premium monotoni-

callydecreases in magnitude as one moves from the smallest to the largest firms. This would explain why in the matched sample (which contains mostly large

firms) there is no working capital accrual premium present. Although the long-term accrual premium also decreases as we move towards larger firms, it is

consistentlysignificant even among the largest companies.