THE DEMAND FOR MONEY IN THE UNITED KINGDOM, 1875–1913
| Date | 01 August 1987 |
| DOI | http://doi.org/10.1111/j.1468-0084.1987.mp49003001.x |
| Author | Jan Tore Klovland |
| Published date | 01 August 1987 |
OXFORD BULLETiN
of
ECONOMICS and STATISTICS
OXFORD BULLETIN OF ECONOMICS AND STATISTICS, 49,3(1987)
0305-9049 S3.00
THE DEMAND FOR MONEY IN THE UNITED
KINGDOM, 1875-1913
Jan Tore Klovland*
I. INTRODUCTION
The interest in the compilation of historical data on the money stock
and the estimation of long-run demand-for-money functions for the
United Kingdom seems to be stronger than ever. In particular, the
recent publication of the long-awaited study of monetary relationships
over the past century in the United States and the United Kingdom by
Friedman and Schwartz (1982) has given further impetus to the study
of this subject.1 Hendry and Ericsson (1983) have undertaken a pro-
found re-examination of the econometric properties of their demand-
for-money function for the United Kingdom. The startling conclusion
reached by Hendry and Ericsson is that the ship so confidently embarked
by Friedman and Schwartz in order to cross the Atlantic is found not
to be seaworthy at all, due to a number of inherent instabilities.2
There are several reasons for undertaking a separate analysis of the
demand for money in the United Kingdom in the period 1875 to 1913.
The pre-1914 gold standard era seems to be a period of relative smooth
waters, in the sense of being free from disruptions like major wars or
abrupt changes in the financial system. Furthermore, Mills and Wood
(1978) found that income caused money (in the Granger sense) for this
* I would like to thank Forrest Capie, Neil Ericsson, David Hendry, Terence Mills and a
referee for helpful comments.
Recent contributions include Artis and Lewis (1984), Bordo (1981), Capie and Rodrik-
Bali (1985), Capie and Webber (1983-85), Huffman and Lothian (1980), and Mills and Wood
(1982). Earlier research on this subject was pioneered by Higonnet (1957), Kavanagh and
Walters (1966), Laidler (1971) and Sheppard (1971).
2Longbottom and Holly (1985) have subsequently presented some further empirical evi-
dence on these issues, which seems to go some way in establishing a demand-for-money function
with more satisfactory properties, but the equations still exhibit instability over the most recent
years.
251
Volume 49 August 1987 No. 3
252 BULLETIN
period, which makes it natural to approach the modelling of the relation-
ship between money, income and interest rates in terms of a demand-
for-money equation. The properties of the estimated relationship from
this period may thus serve as a benchmark for assessing results from
subsequent periods. Finally, it will be argued in some detail below that
the annual data series covering this period which were used by Friedman
and Schwartz and other researchers are not the best available ones in
several respects. This proposition does not apply solely to the stock of
money itself, to which Goodhart (1982) has already drawn attention,
but can be extended to estimates of the Consol yield, the short-term
interest rate as well as the own yield on money. A second major purpose
of this paper is thus to examine to what extent the use of alternative
data series affects the main properties of the empirical demand-for-
money function.
H. THE BASIC MODEL
The general theoretical framework of the demand for money in Friedman
and Schwartz (1982) (henceforth referred to as FS) is adopted here.
Their empirical model relates the per capita money stock (M) to the
price level (P), real per capital income (Y), a short-term interest rate
(RS) and a long-term interest rate (RL). In some specifications the
differential yield between short-term substitute assets and money (RN)
is used instead of RS, and in some cases RL is omitted from the equa-
tion. In the following, annual, rather than phase average, data will be
used, the merits of which are evident from the analysis in Hendry and
Ericsson (1983). The estimation period is 1875 to 1913.
The data used in this section are those given in FS, table 4.9, except
that their net national product is replaced by the more widely used
gross domestic product at constant factor cost.3 A similar substitution
is made for the price deflator. Neither of these changes affected the
results in any significant way; for a further comparison of the alternative
income measures, see Section VI below. Initially, both logarithmic and
untransformed values of the interest rate variables were used. The
logarithmic version was chosen, since residual variance was slightly
lower in most cases, but parameter estimates were little sensitive to the
choice of functional form.
Starting from an unrestricted autoregressive distributed lag specifica-
tion of the real per capita demand for money, the following equation
was obtained (henceforth logarithmic values of the variables are denoted
by lower case letters, White's (1980) heteroskedastic-consistent esti-
mates of standard errors in parentheses):4
The main data series are tabulated and further defined in the data appendix.
All equations were estimated using the GIVE programme (Oxford version 1.2); see Hendry
and Srba (1980) for a general description.
To continue reading
Request your trial