Can Industry‐level Trade Linkage Predict Stock Returns?

• Date: 01 April 2020
• DOI: http://doi.org/10.1111/ajfs.12292
• Author: Tae‐Hoon Lim
• Published date: 01 April 2020

Can Industry-level Trade Linkage Predict

Stock Returns?*

Tae-Hoon Lim**

Division of Language and Trade, Hankuk University of Foreign Studies, Seoul, Republic of Korea

Received 31 July 2019; Received in current form (1

st

revision) 15 January 2020; Accepted 18 January 2020

Abstract

In this paper, I test whether cross-predictability exists among trade-linked industries across

international borders and explore possible explanations for this. I find strong evidence of

cross-border stock return predictability among trade-linked industries. A trading strategy of

buying industry portfolios for which trade-linked industry had high returns, and shorting

industry portfolios for which trade-linked industry had low returns, yields an annualized

return of 12%. Such returns cannot be explained by known risk factors and are different from

industry momentum. I find some evidence that counters the information segmentation expla-

nation for cross-predictability and find support for illiquidity as a new channel of explanation.

Keywords Return predictability; Customer–supplier relations; Trade linkage; International

equity; Institutional ownership

JEL Classification: G15, G17, F37

1. Introduction

International trade volume and its shares of the GDP of many countries have been

growing steadily over the past few decades

1

. More importantly, this increase in

international trade activity has served to strengthen economic linkages between

*This paper is based on my PhD dissertation. I am grateful to my advisor David T. Ng and

my committee members Edith Liu and Sanjeev Bhojraj for their encouragement and guid-

ance. I am also grateful for comments from Andrew Karolyi, George Gao, Kee-Hong Bae,

Louis Gagnon, Selim Topaloglu, an anonymous referee, and seminar participants at Cornell

University, Konkuk University, Korea Institute of Finance, Korea Institute for International

Economic Policy, Pacific Investment Management Company, Queen’s University, and CAFM

2014. All remaining errors are mine. This work was supported by the Hankuk University of

Foreign Studies Research Fund.

**Corresponding author: Division of Language and Trade, HUFS, 107 Imun-ro, Dongdae-

mun-gu, Seoul 02450, Republic of Korea. Tel: +82-2-2173-8807, email: thlim@hufs.ac.kr.

1

The World Bank (http://data.worldbank.org/): Export volume index, Exports of goods and

services (% of GDP).

Asia-Pacific Journal of Financial Studies (2020) 49, 234–271 doi:10.1111/ajfs.12292

234 ©2020 Korean Securities Association

industries in multiple countries. In such an environment, it is possible that infor-

mation originating from international trade partners of an industry can predict

future returns for that industry. In this paper, I test whether future retur ns of

industry portfolios can be predicted using past information from trading partners.

Moreover, I c utilize trading partner relationships between industries to characterize

and explore possible explanations for cross-predictability.

Previous research on return predictability among economically linked industries

has mostly been focused within a single country. Menzly and Ozbas (2010) show

that stock returns of economically related industries can cross-predict each other’s

returns in United States stock markets. Similarly, Cohen and Frazzini (2008) inves-

tigate whether firm level public information on customer and supplier relationships

can be used to obtain abnormal returns. Shahrur et al. (2010) extend the scope of

studies of predictability to an international setting to examine equities in 22 devel-

oped markets but only utilize within country economic linkages, assuming that

these countries have very similar within country economic linkages as those in the

United States.

However, as inter-industry relationships extend beyond national borders, inter-

national interdependence of industries warrants further investigation of this issue in

an international setting. Moreover, the international setting of this paper provides

the additional benefit of providing a more suitable testing ground for existing theo-

ries of cross-predictability, since the conditions upon which these theories rely, such

as informational segmentation, illiquidity, and other market friction, are more natu-

ral in international settings.

Recent empirical literature finds some evidence of return predictability among

internationally linked stocks, but this literature relies on country-level trade data to

identify international economic linkages. Rizova (2013) examines the interdepen-

dence of country-level trade relationships and country-level equity market perfor-

mance. Albuquerque et al. (2015) find return predictability of firms with high trade

credit based on returns of customer countries identified by country-level trade data.

Since the international economic linkages considered in the previous literature are

aggregated at the country level, it is difficult to fully investigate the role of interna-

tional economic linkages between industries. In this paper, I bring in a new data

source, the GTAP (Global Trade Analysis Project), which can describe international

economic linkages with much more detail than previously considered.

The GTAP provides data on spending on imported and exported goods for each

industry and country pair around the world. These data, widely used in interna-

tional trade literature but never in the finance literature, enable us to look not only

at the industry-level breakdown of exported and imported goods and services, but

also specific industries’ dependences on particular imported goods. For example,

the data describe quantities of iron and steel products imported from Japan to

Korea. Moreover, it reports amounts of this iron and steel consumed by related

Korean industries. The rich structure of the data help us understand relationships

Trade Linkage and Return Predictability

©2020 Korean Securities Association 235

among industries across countries. More importantly, such a broad cross section of

economically linked industries enables us to relate sources of cross-predictability to

their customers and suppliers.

To quantify degrees of international linkages between industries, I consider

international trade flows and imported goods usage by industries. Based on such

linkages, I can quantify how an industry in one country is related to other indus-

tries around the world. I construct related industry portfolios and examine whether

industry portfolio returns can be predicted by past returns of internationally related

industries. Also, bilateral relationships between related industries allow for new pos-

sibilities for testing existing theories. In particular, we have access to a cross section

of related industries with varying levels of international trade relationships, which

can be used along with other relationships between two industries, such as institu-

tional co-ownership and analyst co-coverage. This data structure enables us to break

down the predictor variable into several pieces and analyze whether there is a vary-

ing level of predictability along certain criteria, such as co-ownership or co-cover-

age, in addition to the international trade link.

Overall, I find strong evidence for cross-border stock return predictability

among trade-linked industries. A trading strategy of buying industry portfolios for

trade-linked industries that had high returns, and shorting industry portfolios for

trade-linked industries with low returns, yields annualized returns of 12%. Such

returns cannot be explained by known risk factors, and are different from industry

momentum. I find some evidence against the leading explanation that posits infor-

mation segmentation as the only reason for this cross-predictability, and find sup-

port for illiquidity as a new channel of explanation.

My paper makes the following three contributions to the literature. First, I

uncover effects on returns of industry-level trade linkages around the world. Sec-

ond, I test whether relative information is efficiently priced across countries and

industries. Third, by selecting an international setting in which there is natural

information segmentation across countries and great variability of liquidity, I obtain

a better testing ground for these theories.

I find the following four empirical results. First, as noted above, I find that self-

financing trading strategies based on past information from economically related

industries yield significant premiums.

Second, I analyze the characteristics of past returns that are most powerful in

predicting an industry’s return. If information segmentation is the main explanation

for cross-predictability, then returns from obscure or ignored stocks are more likely

to carry more weight in predicting related industry returns in foreign countries. In

contrast, I find that the strongest predictive power comes from past returns of eco-

nomically linked industries that share greater degrees of (1) institutional ownership

and (2) analyst coverage. Such industries are likely better -known and familiar to

investors. This suggests that information segmentation does not fully explain cro ss-

predictability.

T.-H. Lim

236 ©2020 Korean Securities Association