Productivity Risk and Industry Momentum

• Published date: 01 September 2018
• Date: 01 September 2018
• DOI: http://doi.org/10.1111/fima.12206
• Author: Efdal Ulas Misirli

Productivity Risk and Industry

Momentum

Efdal Ulas Misirli∗

A productivity shock identified through a vector autoregression model is a priced risk factor

for one-month industry momentum portfolios and commands a positive risk premium. Stocks in

winning industries have greater sensitivity to productivity news, thereby earning higher average

returns than stocks in losing industries. This evidence lends support to an Intertemporal Capital

Asset Pricing Model (ICAPM) with human wealth. In many specifications, exposure to productivity

risk captures more than half of the observed industry momentum profits. This paper studies the

sources of profits and attributes the risks of industry momentum portfolios to the behavior of their

underlying cash flows.

The industry momentum effect, or the tendency of industry portfolio investments to display

persistence in their relative performance, is an important anomaly in cross-sectional asset pricing.

Moskowitz and Grinblatt (1999) find that this anomaly is strongest at the one-month horizon.

Moreover, it is not subsumed by the individual stock momentum anomaly.

The rewards to a one-month industry momentum strategy are also sizeable. The long/short

winner-minus-loser portfolio earns an average return of 0.93% per month that is twice as high

as the returns associated with several other important anomalies, such as size, book-to-market,

asset growth, long run reversal, short run reversal, accrual, and gross profitability. In addition,

the industry momentum strategy has an annualized Sharpe ratio of 0.56, which is higher than the

Sharpe ratios of the other anomaly-based strategies listed.

Despite its large economic benefits, the existence of industry momentum poses a challenge to

rational asset pricing. In fact, many asset pricing models, such as the capital asset pricing model

(CAPM), the Fama and French (1993, 1996) three-factor model, the Carhart (1997) four-factor

model, and the Fama and French (2015) five-factor model, fail to explain this effect. Uncovering

the risk behind one-month industry momentum is still an interesting open question. This paper

provides a solution to it by suggesting a plausiblerisk factor and explaining, in detail, why winner

and loser industries load differently on it.

The key risk factor underlying one-month industry momentum profits is the innovation in

total factor productivity. I motivate the productivity innovation within an intertemporal capital

asset pricing model (ICAPM) and generate its proxy using a vector autoregression model. Cross-

sectional regressions reveal that the productivity shock becomes a priced risk factor and carries

The author thanks John Long, Bill Schwert, Jerry Warner, Fabian Hollstein (MFA discussant), Mihail Velikov (IBEFA

discussant), Yufeng Han (SFA discussant), and seminar participants at the University of Rochester, the 2016 Midwest

Finance Association Meetings in Atlanta, 2017, the IBEFA Summer Meetings in San Diego, and the 2017 Southern

Finance Association Meetings in Key West for their valuable feedback and suggestions.I am also grateful to Bing Han

(Editor) and an anonymous reviewer for their thoughtful comments and suggestions. The viewsexpressed in this paper

are those of the author and do not necessarily reflectthe position of the Federal Reserve Bank of Richmond or the Federal

Reserve System.

∗Efdal Ulas Misirli is a Financial Economist in the Quantitative Supervision and Research Department of the Federal

Reserve Bank of Richmond, Baltimore, MD.

Financial Management •Fall 2018 •pages 739 – 774

740 Financial Management rFall 2018

a positive price of risk. Winning industries load positively on this innovation, thereby earning

higher average returns than losing industries.

Merton’s (1973) ICAPM is an important model to identify reasonable risk factors and to

study the cross-sectional variation of expected returns. Campbell (1996) extends Merton’s (1973)

analysis by developing an empirical method to test the ICAPM and proposing a multifactor

model where innovations in state variables that forecast future stock returns and labor income

growth become plausible risk factors. Campbell (1996) emphasizes the latter group of forecasting

variables as changes in labor income affect consumption-investment behavior.

The cyclical dynamics of employment are also an important research area in macroeconomics.

Since the introduction of seminal articles by Kydland and Prescott (1982) and Long and Plosser

(1983), business cycle research has focused on the impact that productivity shocks have on

employment and other macroeconomic variables. A well-established result in this area is that

a positive productivity shock leads to improvement in output, employment, wages, and labor

income (King, Plosser, and Rebelo, 1988; King, Plosser, Stock, and Watson, 1991; Chang and

Hong, 2006).

Linking these two lines of research, this paper suggests total factor productivity growth as

an ICAPM state variable and further indicates that productivity risk exposure explains more

than half of the one-month industry momentum profits. This result is encouraging as several

ICAPM factors studied in the literature fail to account for these profits. As a result, sensitivity

to productivity news seems to be the major risk affecting industry momentum portfolios. Event

time regressions also reveal that productivity risk exposure of industry momentum portfolios are

consistent with the short life of profits.

Motivated by the theoretical work of Johnson (2002), this paper also examines why the pro-

ductivity risk exposure of winner industries differ from those of loser industries. Johnson (2002)

provides a rational economic mechanism in which momentum sorts (both individual stock mo-

mentum and industry momentum) become equivalent to sorts on expected growth rate levels.

Johnson (2002) argues that empirical studies that seek to rationalize momentum using his

framework should be able to provide evidence for two conditions. First, the empiricist should

find that expected growth rates rise in the momentum cross-section from the loser portfolio to

the winner portfolio. This evidence will be sufficient to argue that growth rate risk rises with

growth rates. In addition, the empiricist should demonstrate that the growth rate risk is priced.

The empiricist can satisfy the latter condition by indicating that the estimated cash flow betas

with respect to the candidate systematic risk factor also increase from the loser portfolio to the

winner portfolio. If the empiricist produces supporting evidence, these two conditions then imply

that winners should have higher expected returns than losers as winners have greater cash flow

risk than losers and winners’ cash flow risk is amplified by their high expected growth.

To satisfy the first condition within the context of the industry momentum anomaly,this paper

confirms that winner industries have temporarily higher average cash flow growth, sales growth,

and investment growth rates than loser industries. In addition, the paper finds that the duration

of these expected growth spreads matches roughly that of industry momentum profits.

Tosatisfy the second condition, this paper estimates a novel cash flow risk measure for industry

momentum portfolios. Following Campbell, Polk,and Vuolteenaho (2010), I form portfolio-level

cash flow news variables based on return-on-equity and regress them on ICAPM factors. I find

that winner industries’ cash flows have higher productivity risk exposure than loser industries’

cash flows. As a result, the productivity risks of industry momentum portfolios are derived from

the behavior of their underlying cash flows.

An alternative explanation for the one-month industry momentum effect concerns lead-lag

relations within industries. Moskowitz and Grinblatt (1999) rule out the possibility that industry

Misirli rProductivity Risk and Industry Momentum 741

momentum is driven bysize-, liquidity-, or microstr ucture-related lead-lag effects.However, Boni

and Womack(2006) f ind some evidence that the returns of firms with more analyst coverage lead

those with less coverage in the same industry.Similarly, Hou (2007) confirms that firms with high

market shares lead firms with low market shares in the same industry. Moskowitz and Grinblatt

(1999) argue that, in a value-weighted portfolio, lead-lag effects playa minor role. They also f ind

a sizable profit for their one-month industry momentum strategy among large and liquid stocks.

Unlike these studies, this paper seeks a risk-based explanation and attributes industry momentum

profits to compensation for aggregate productivity risk.

Several papers illustrate the importance of productivity risk in cross-sectional asset pricing.

Balvers and Huang (2007) study productivity risk in a conditional asset pricing model and mea-

sure aggregate productivity with total factor productivity. They attribute the size premium to

differences in unconditional sensitivities. Small firms are more sensitive to aggregate produc-

tivity shocks than large firms. They determine that a value premium emerges from differences

in conditional sensitivities to productivity innovations. In addition, Belo (2010) formulates a

productivity-based pricing kernel and studies the cross-section of size and book-to-market port-

folios.1

Since productivity news is an important source of business cycle fluctuations, it becomes

a part of the pricing kernel in several other macro-based asset pricing models. Studies that

empirically test these models using unconditional sensitivities, however, tend to find a role for

macroeconomic shocks other than the productivity shock (i.e., investment-specific technology

shocks and financial shocks) in the cross-section of popular anomalies, such as book-to-market

and investment (Koganand Papanikolaou, 2013; Belo, Lin, and Yang,2014). Unlike these studies,

this paper focuses on a less studied, yet economically more important, anomaly and uncovers the

importance of productivity risk in its cross-section.

This paper also contributes to the broader literature that identifies several risk factors for

the cross-section of stock returns using the ICAPM, including Campbell (1996), Campbell

and Vuolteenaho (2004), Ang, Hodrick, Xing, and Zhang (2006), Petkova (2006), Adrian and

Rosenberg (2008), Wang (2013) and Campbell, Giglio, Polk, and Turley (2016). Similar to these

papers, this work reveals the pricing ability of a plausible ICAPM innovation in the cross-section

of an important stock market anomaly.

The remainder of the paper is organized as follows. Section I provides the descriptive statistics

of the one-month industry momentum anomaly and compares them with the statistics of other

well-known stock marketanomalies. This section also presents the f ailure of Fama-French (1996,

2015) multifactor models in explaining the industry momentum effect. Section II motivates

total factor productivity (TFP) growth as an ICAPM state variable and presents the relation

between productivity shock and other macroeconomic variables.Section III repor ts the results of

the asset pricing tests that employ the TFP shock and demonstrates the economic significance of

productivity risk for the industry momentum anomaly.Section IV reports the short life of industr y

momentum profits and shows its consistency with the evolution of TFP shock loadings in event

time. Section V gives economic meaning to TFP shock loadings and rationalizes the industry

momentum anomaly using Johnson’s (2002) theoretical framework. Section VI runs horse races

with alternative systematic risk factors and provides additional robustness checks. This section

1While the innovation in total factor productivity (TFP) is a common proxy for technologyshock in macro-economics,

recent studies in finance construct alternative proxies from patent data, and explore their asset pricing implications. For

example, Hsu (2009) finds that aggregate patent and research and development (R&D) shocks predict future market

returns. Motivated by this evidence, Hsu and Huang (2010) form a technologyrisk factor, and examine its pricing ability

for size, book-to-market, individual stock momentum, and R&D intensity portfolios.