MODELS OF PRIMARY PRICE INDICES
| Published date | 01 August 1987 |
| Date | 01 August 1987 |
| Author | L. Alan Winters |
| DOI | http://doi.org/10.1111/j.1468-0084.1987.mp49003004.x |
OXFORD BULLETIN OF ECONOMICS AND STATISTICS, 49, 3(1987)
0305-9049 $3.00
MODELS OF PRIMARY PRICE INDICES
L. Alan Winters*
In recent years much attention has been focused on developing countries'
exports of manufactures, in terms of both their market penetration of
OECD markets and their growing share of developing countries' total
exports. This has not been misplaced, but it has sometimes obscured
the equally important role that exports of primary commodities play.
Around one-third of developing countries' export revenues, and prob-
ably an even higher proportion of their GDP, derive from non-fuel
primary commodities. The behaviour of primary commodity markets is
also of interest to industrial countries: first, as Winters and Yu (1985)
show, developing countries' export receipts quickly get translated into
demand for industrial countries' goods; second, industrial countries
have significant primary exports themselves; third, commodity prices
are an important determinant of industrial countries' inflation; and
fourth, fluctuations in primary prices induce significant variability in
industrial countries' terms of trade and real income. Thus primary
commodity markets play an important role in both the world economy
and in models and projections of it.
This paper is addressed to the need of both the World Bank and the
OECD Secretariat to explain and forecast aggregate commodity prices
in the course of their macro-economic projection exercises. In the
World Bank, detailed price forecasts for 34 individual commodities,
accounting for about 80 per cent of developing countries' non-fuel
primary exports, are made roughly twice a year. These contain large
amounts of commodity-specific information and expert judgement, and
they are used to define a 'base-case' projection for aggregate commodity
price indices. Clearly, however, such detailed projections cannot be re-
calibrated for every alternative world macro-economic scenario con-
sidered in the course of a projection exercise, and equally clearly, it is
not satisfactory to assume that commodity prices are independent of
other macroeconomic developments. We need, therefore, a set of func-
tions which relate aggregate primary prices to various macroeconomic
phenomena, and which may be used simply and cheaply to adjust the
base-case projections for changes in the macroeconomic environment.
* This paper is based on work undertaken in the Global Analysis and Projections Division
of the World Bank. I am grateful to Peter Mjovic and his staff for their help and comments.
I am also grateful to the editors for their extensive comments. Naturally the views expressed
must not be attributed to the World Bank and I remain solely responsible for all the paper's
remaining shortcomings.
307
308 BULLETIN
This paper reports such a set of functions for three broad groups of
primary commodities - food, non-food agricultural products and
metals and minerals. Starting from a simple theoretical specification, it
estimates general equations for the three commodity price indices and
then simplifies them to derive, as far as possible, robust and parsi-
monious empirical models. The latter are subjected to a wide range of
tests of specification and structural stability and are compared with
three other models taken from the literature. These tests suggest that,
with a few minor reservations, the models derived here do offer an
acceptable representation of the economic process determining primary
prices.
I. THE MODEL
The model adopted here is economically very simple. For a typical
primary commodity we express demand as a function of economic
activity, y, and relative prices, p. Ignoring, for now, any dynamics, we
can write:
qD qD(yp) (1)
Supply is taken to be a simple function of relative prices and, for agri-
cultural goods, major climatic or biological shocks (d)
qSrqS(p,d) (2)
Stockholding, h, is expressed as a function of the overall size of the
market (qS or qD) real interest rates, r, and possibly expected prices, pe
h=h(qD,r,pe). (3)
The role of expected prices, pe, is uncertain, for a correctly measured
real interest rate should capture all the speculative components of the
stock decision.
Assuming that expected prices may be either ignored or substituted
out by appealing to rational or autoregressive expectations, and using
market clearing, (1) to (3) imply a price equation:
p = p[L(p), L(y), L(r), d] (4)
where L (x) denotes general lag functions of x.
Equation (4) is specified very loosely, embodying none of the a priori
constraints that might appear to be desirable from (1) to (3). There are
four reasons for this. First, for perfectly measured individual commodi-
ties, (1) to (3) would indeed embody some a priori constraints that
could be useful econometrically, but we are dealing with imprecise data
aggregated over both commodities and time. Second, the relationship of
expectations to observable data may be very complex. Third, the price
To continue reading
Request your trial