INDUSTRY MOMENTUM IN AN EARLIER TIME: EVIDENCE FROM THE COWLES DATA

• Date: 01 September 2015
• DOI: http://doi.org/10.1111/jfir.12062
• Published date: 01 September 2015
• Author: Xiwen Zhou,Andrew C. Szakmary

INDUSTRY MOMENTUM IN AN EARLIER TIME:

EVIDENCE FROM THE COWLES DATA

Andrew C. Szakmary and Xiwen Zhou

University of Richmond

Abstract

Virtually all evidence on the efficacy of momentum strategies arises from the post-1962

era, and momentum returns across different markets and asset classes are highly

positively correlated. We examine industry momentum in an earlier time and find that

these strategies would have earned gross returns over the 1900–1925 period that are at

least equal to those in the modern era; however, there is little evidence of profitability

before 1900. The paucity of industry coverage in the earlier years of the Cowles database

likely does not explain this differential profitability. We also show that the market

state dependence of momentum strategies in the 1900–1925 period is similar to the

modern era.

JEL Classification: G11, G12, G14

I. Introduction

Beginning with Jegadeesh and Titman (1993), a voluminous literature documents the

profitability of intermediate-horizon momentum strategies, whereby securities that have

experienced relatively high (low) returns over prior formation periods up to 12 months

continue to earn relatively high (low) returns over subsequent holding periods of up to

12 months. In addition to individual U.S. stocks, the profitability of momentum strategies

has been established for foreign stocks in all but a handful of countries (Rouwenhorst

1998; Griffin, Ji, and Martin 2003; Chui, Titman, and Wei 2010), and several studies

beginning with Asness, Liew, and Stevens (1997) and Chan, Hameed, and Tong (2000)

demonstrate the efficacy of momentum strategies implemented with country stock

indices. Albeit sometimes with shorter formation and/or holding periods, classic

momentum strategies sorting on past cross-sectional returns have also been found to be

profitable in industry stock indices (Moskowitz and Grinblatt 1999 and numerous other

studies discussed below), commodity futures (Shen, Szakmary, and Sharma 2007; Miffre

and Rallis 2007), and exchange rates (Harris and Yilmaz 2009; Serban 2010). Recently,

Asness, Moskowitz, and Pedersen (2013) confirm the profitability of both value and

momentum strategies across many different markets and asset classes along with a strong

common factor structure among their returns: both momentum and value strategy returns

We would like to thank an anonymous referee; Steve Jones, the associate editor; and seminar participants at

the 2014 European Financial Management Association meetings (Rome) for valuable comments and suggestions.

The Journal of Financial Research Vol. XXXVIII, No. 3 Pages 319–347 Fall 2015

319

© 2015 The Southern Finance Association and the Southwestern Finance Association

RAWLS COLLEGE OF BUSINESS, TEXAS TECH UNIVERSITY

PUBLISHED FOR THE SOUTHERN AND SOUTHWESTERN

FINANCE ASSOCIATIONS BY WILEY-BLACKWELL PUBLISHING

exhibit high positive correlation across markets/asset classes but negative correlation

with each other.

Overall, it is clear from the literature that the profitability of momentum

strategies is pervasive across many countries, markets, and asset classes, including equity

indices grouped by industry. Given these extensive findings, it appears on the surface that

yet another study of the momentum phenomenon is hardly necessary. It is noteworthy,

however, that virtually all of the existing momentum evidence stems from post-1962 data

(which we will hereafter refer to as the modern era), and that as Asness, Moskowitz, and

Pedersen (2013) show, returns resulting from momentum (and value) strategies

implemented across countries and asset classes are more highly positively correlated

than those resulting from passive investments in the same countries/asset classes. Thus,

the simultaneous global existence of individual stock momentum, country index

momentum, industry momentum, exchange rate momentum, commodity futures

momentum, and so on in the modern era does not necessarily prove that data mining

is not a factor underlying the profitability of momentum strategies. To completely rule

out data mining, it would be helpful to determine whether these strategies would have

worked during a period that has not previously been examined. This is the main

motivation for our study, which examines industry momentum in a relatively high-

quality data set covering an earlier period (1871–1925) that has received little attention.

The literature that is most closely related to our article consists of studies that

examine industry momentum. Moskowitz and Grinblatt (1999) form 20 value-weighted

industry portfolios over the 1963–1995 period and apply momentum strategies to these

industry portfolios. Although they find profitability to be highest for 1-month formation

and holding periods with no gap between the two (which, for reasons explained below,

we cannot replicate with the Cowles data), Moskowitz and Grinblatt do report significant

profitability for 6- to 12-month formation and holding periods with a 1-month gap

between the formation and holding periods. Among other significant findings, they report

that unlike when the strategies are implemented at the level of individual stocks, the

profitability of industry momentum strategies is primarily driven by long positions, and

(more controversially) they claim that once individual stock returns are adjusted for

industry effects, momentum profits from individual equities are weaker and generally

statistically insignificant. This latter claim is not supported by Grundy and Martin (2001)

and other studies discussed below.

Numerous studies have since substantially refined and extended Moskowitz and

Grinblatt’s (1999) initial industry momentum findings. Swinkels (2002), using 40

Datastream industry indices over the 1973–2000 period, confirms that many skip-a-

month industry momentum strategies with 6- to 12-month formation and holding periods

are significantly profitable in the United States and even more so in Europe, but not in

Japan (one of the few countries where individual stock momentum also is not

significantly profitable). Giannikos and Ji (2007) examine industry momentum strategies

in 37 countries over a similar period; they exclusively use 6-month formation and

holding periods with a 1-month lag. Although they do not report significant profitability

for every country, they do find significant profits for every region when results are

aggregated across the Americas (excluding the United States), Europe, and Asia.

Moreover, Giannikos and Ji show that all of this profitability is essentially accounted for

320 The Journal of Financial Research