IN SEARCH OF AN EXPLANATION OF COMMERCIAL BANK SHORT‐RUN PORTFOLIO SELECTION*
| DOI | http://doi.org/10.1111/j.1468-0084.1980.mp42004002.x |
| Author | ANTHONY S. COURAKIS |
| Date | 01 November 1980 |
| Published date | 01 November 1980 |
IN SEARCH OF AN EXPLANATION OF COMMERCIAL BANK
SHORT-RUN PORTFOLIO SELECTION*
By ANTHONY S. COURAKIS
INTRODUCTION
The purpose of this study is to explore the implications of certain issues
disregarded in applications of portfolio theory and to examine the usefulness of a
number of mean-variance models of portfolio selection in explaining the asset
choices of London Clearing Banks during the period January 1959 to July 1967.
As with previous studies in this tradition,' the specific aim is to explain behaviour
vis-à-vis those assets which are amenable to short-term management. Unlike
earlier studies however the analysis presented here:
recognizes the importance of anticipated changes in the remaining components
of the intermediary's portfolio,
proceeds to a systematic examination of the issue and implications of proxying
expected returns from actual rates,
relates empirically to the behaviour of individual institutions and should,
accordingly, be thought to provide a more accurate evaluation of a theory
based on individual bank maximizing decisions.
I. ASSUMPTIONS, LIMITATIONS AND PRESUMPTIONS
In the traditional analysis of choice under 'uncertainty' it is assumed that,
subject to its balance sheet constraint, the decision-taking unit aims to maximize
the expected utility of wealth at some terminal date. Underlying the approach is
the idea that the determinants affecting choice of alternative portfolios can be
reduced to expected return and risk, where the former is the mean of the probability
distribution of return and the latter is usually, though sometimes half heartedly,2
approximated by the variance of that distribution. On the additional assumption
that the decision taking unit is a price (yield) taker in every market, decisions
regarding alternative portfolios are quantity-setting decisions and stochastic
variability in end-period wealth is thus entirely due to prices (yields) being
stochastic.
* An earlier version of this paper was presented to the 1975 Money Study Group Conference in
honour of Professor G. L. S. Shackle, under the title 'Portfolio Choice in the Small, the Large and the
Many'. I am grateful to the Director of the Institute for providing me with an office and research
facilities without which the demanding task of estimation of the results here presented would not have
been feasible, to the Houblon Norman Foundation for financial support that enabled the speedier
execution of the final stage of estimation, and to Nick Weinreb for computing assistance at this final
stage. My thanks are also due to Bob Bacon,John Flemming,Jim Mirrlces, Steve Nickell, Adrian Smith
and the Editors for their comments.
1 See Parkjn el al. (1970b) and also Courakis (1974).
2See Hicks (1934-5) and (1962).
305
306 BULLETIN
Insofar as such a model is to serve as the basis of analysis of actual behaviour,
its single period nature is obviously restrictive, since it implies that the decision
taking units choices are independent of all eventualities other than those pertaining to
the probability distribution of returns on the various assets at the specific moment in time
to which end of period wealth relates.3 Yet, as we have been reminded by Merton, in
the absence of transactions costs and indivisibilities in exchange investors would
prefer to revise their assets at any time, whether they actually do so or not. And
though in reality such costs and indivisibilities do exist and provide one reason
given for the finite trading interval and single period models, this method of
avoiding the problem of transactions costs is not satisfactory since a 'proper
solution' to it would almost certainly show the trading intervals to be stochastic
and of non-constant length.4
Unfortunately in the present state of our knowledge one cannot even pretend
that a 'proper solution' to discontinuities exists. We are thus left with the choice of
either assuming no adjustment cost and opting for continuous trading or reverting
to the discrete time model. But in practice this choice is more apparent than real
since the literature on continuous time and multiperiod models of portfolio selection
(however valuable in answering the question of under what assumptions would it
be valid to pursue a single period analysis when the true behaviour is characterized
by multi-period horizons) has relied on assumptions that yield demand relationships
empirically identical to those resulting from single period models.5 At the same
time it can be argued that given experience regarding 'normal' variation in relative
yields and adjustment costs, a definite decision interval and institutional procedures
reflecting this interval are far from uncommon. Furthermore one may define
sequential structures of decisions where allocations to groups of assets reflect
discrete-time choices of different frequency from those pertaining to intra-group
allocations. For a financial intermediary such a process does not seem
unreasonable ;6 but even then certain features of the paradigm remain obnoxious.
In general this restrictibn is interpreted to imply that the investor acts 'as if' intra-period
transactions are disregarded, in the sense that such transactions are barred; but it is sufficient to assume
that beginning period decisions, including intra-period trades, are binding throughout the period, or that
intra-period trades are independent of the value of the portfolio at the point in time to which such
adjustments relate.
quoted from Merton (1973), p. 869.
5i for example the continuous time model presented in Flemming (1974). The demand
functions for the various assets there derived are of the form
(V/Wo)=Gy+B
where V is a vector of quantities of the various assets, y are the expected yields on these assets, Wo is
the decision taking unit's initial wealth, i.e. the sum of the various assets, a is the risk aversion parameter,
and G and B are functions of the inverse of the variance covariance matrix of returnsand hence
identical, to the demand functions derived from a single period model in which the investor is assumed
to maximize the expected utility of return of his portfolio in terms of a negative exponential utility
function, i.e. max E[U(R)] = E[U( W/ Wo)], where W is end-of-period wealth, subject to tV= Wo where i
is a sumation vector, and granted preferences defined by
U(Ñ)=bc e
and normally distributed stochastic returns on V.
61 am grateful to David Lomax, Economic Adviser of National Westminster, and John Luck,
Treasurer of Williams & Glyn, for useful discussions on this issue.
To continue reading
Request your trial