ECONOMIC FORCES IN THE LONDON STOCK MARKET
| Author | Michael Beenstock,Kam‐Fai Chan |
| Published date | 01 February 1988 |
| Date | 01 February 1988 |
| DOI | http://doi.org/10.1111/j.1468-0084.1988.mp50001002.x |
OXFORD BULLETIN OF ECONOMICS AND STATISTICS, 50,1(1988)
0305-9049 S3.00
ECONOMIC FORCES IN THE LONDON STOCK
MARKET
Michael Beenstock and Kam-Fai Chan
I. INTRODUCTION
In a previous paper, Beenstock and Chan (1986), we sought to test the Arbi-
trage Pricing Theory (APT) in the context of the market for UK securities.
Using monthly observations on rates of returns for 220 different securities
over the period 1961-81, we found that there was considerable empirical
support for the APT. Our empirical methodology was based on Roll and
Ross (1980) in that we used factor analysis to estimate the sensitivity of
individual securities to different systematic risk factors. These factor analytic
techniques and our data suggested that approximately 20 risk factors are
present in the market for UK securities.
In contrast, Roll and Ross (1980) concluded that in the market for US
securities, only four factors were present. Dhrymes (1984) has observed that
the number of factors appears to depend on the number of securities and that
if the sample size is increased, the estimated number of factors tends to rise.
Beenstock and Chan found similar properties in their UK data. The apparent
arbitrariness of the number of factors threatens to undermine the testability
of APT.
The basic problem lies with the methodology of factor analysis in which
the factors cannot be interpreted. Unless it is known what the factors repre-
sent, it is impossible to distinguish systematic and idiosyncratic risk factors. In
short, although the APT has been exclusively 'tested' using factor analysis, it
hardly provides a satisfactory basis for empirical enquiry.
In this paper we therefore propose an alternative methodology for testing
APT that does not rely on factor analysis and which we illustrate with UK
data. Instead we try to identify variables which economic theory suggests will
influence security returns. Some of these variables are company specific while
others affect all companies but to different degrees. The former are idiosyn-
cratic, while the latter are systematic. Forecast errors of the systematic
variables will generate systematic risk. The test of APT then consists of see-
ing whether expected returns depend linearly on the sensitivity of returns to
innovations in the systematic variables. Because we deal throughout with
economic variables, questions of factor interpretation do not arise. Of course,
one can contest the appropriate class of systematic variables, but at least this
puts the debate onto a more transparent footing.
27
To continue reading
Request your trial