The Relative Predictability of Stock Markets in the Americas

• Published date: 01 April 2016
• DOI: http://doi.org/10.1002/ijfe.1536
• Date: 01 April 2016

THE RELATIVE PREDICTABILITY OF STOCK MARKETS IN THE

AMERICAS

GRAHAM SMITH*

,†

and ANETA DYAKOVA

Department of Economics, SOAS, University of London, Thornhaugh Street, London WC1H 0XG, UK

ABSTRACT

The degree of return predictability is measured for seven Latin American stock markets and those in Canada and the United

States using three finite-sample variance ratio tests. Daily data for the period beginning in February 1994 and ending in

December 2011 are used in a fixed-length rolling window to capture short-lived predictability, track changes in predictability

through time and rank markets by relative predictability. Overall, the degree of return predictability varies widely. The most

predictable are those located in Chile and Peru; the least predictable are in Argentina and Brazil. Predictability has decreased

for all of those stock markets examined, except those located in Ecuador and the United States. Predictability largely coincides

with times of crisis. Copyright © 2015 John Wiley & Sons, Ltd.

Received 8 November 2012; Revised 27 August 2015; Accepted 22 September2015

JEL CODE: C12; G1; G14

KEY WORDS: American stock markets; relative efficiency; return predictability; variance ratio test

1. INTRODUCTION

Stock return predictability has received considerable attention because of its relationship with the weak-form

market Efficient Markets Hypothesis (EMH), in which prices fully reflect all of the available information in histor-

ical prices. Most of the previous empirical works testing the EMH for stock markets in the Americas have carried

out tests for a particular market over a single period. This conventional, absolute efficiency approach leads to the

inference that a stock market either is or is not weak form efficient. This absolute efficiency concept is limited.

Campbell et al. (1997, p 24) have noted that relative efficiency—the efficiency of one market measured against

another—is both more relevant and potentially much more useful than the traditional all-or-nothing view. In this

paper, tests are implemented with a fixed-length rolling window. By repeatedly testing the EMH, the rolling

window detects changes in predictability and is used to measure relative predictability. Within this framework,

relative predictability is quantified by comparing the relative frequencies of rejecting the martingale hypothesis.

In weak-form efficient markets, price changes are random and, equivalently, successive one-period returns are

independent. One characterization of this is the random walk hypothesis, which requires independently and iden-

tically distributed (IID) returns. However, because IID returns lack widespread empirical support in stock markets,

the less restrictive martingale hypothesis is more appropriate and is used here. This hypothesis shares the charac-

teristic of independent returns but permits dependence in higher moments, including conditionally heteroscedastic

returns, and so is a more general version of the random walk hypothesis. The concept of relative predictability can

be used with individual stocks, groups of stocks or market indices so it is useful for both local investors and

international investment funds. The purpose of this paper is to test the martingale hypothesis using stock price in-

dices for seven Latin American equity markets located in Argentina, Brazil, Chile, Ecuador, Mexico, Peru and

Venezuela, together with those in Canada and the United States, and rank them according to their relative predictability.

*Correspondence to: Graham Smith, Department of Economics, SOAS, University of London, Thornhaugh Street, London, WC1H 0XG, UK.

†

E-mail: graham.smith@soas.ac.uk

Copyright © 2015 John Wiley & Sons, Ltd.

INTERNATIONAL JOURNAL OF FINANCE & ECONOMICS

Int. J. Fin. Econ. 21: 131–142 (2016)

Published online 3 November 2015 in Wiley Online Library

(wileyonlinelibrary.com). DOI: 10.1002/ijfe.1536