CYCLICAL PRICE BEHAVIOUR AND CONCENTRATION: A TIME SERIES ANALYSIS*
| Author | Howard N. Ross,Joshua Krausz |
| Published date | 01 August 1985 |
| DOI | http://doi.org/10.1111/j.1468-0084.1985.mp47003004.x |
| Date | 01 August 1985 |
OXFORD BULLETIN OF ECONOMICS AND STATISTICS, 47, 3 (1985)
0305-9049 $3.00
CYCLICAL PRICE BEHAVIOUR AND
CONCENTRATION: A TIME SERIES ANALYSIS*
Howard N. Ross and Joshua Krausz
I. INTRODUCTION
If relative prices are sticky when real aggregate demand fluctuates.
markets will not clear, and consequently, the cyclical variation in real
output and employment will increase. This familiar proposition, linking
the microeconomics of relative prices to the macroeconomics of aggre-
gate supply, explicitly rejects the view of markets characterized by
instantaneous clearing in a regime of flexible prices. Price rigidity in
commodity markets has been rationalized in a variety of optimizing
models as a reaction to uncertainty and incomplete information. A
more traditional justification, by no means independent of the others,
lies in the exercise of monopoly and oligopoly power. Controversial as
that explanation is, it is accessible to quantitative verification, and has
provoked a large body of research on the 'administered price hypothesis'.
A contemporary version of the thesis states that after adjusting for
differences in industry specific parameters such as the variable costs of
production and demand, prices in concentrated makets do not fall as
much as in unconcentrated markets during short lived economic con trac-
tions, and rise more gently in the early phases of recoveries. A recurrent
interpretation of this behaviour is that oligopolists choose to ignore or
partially ignore temporary demand shocks in their pricing decisions
while more competitive sellers have no such option.
The wealth of statistical studies on the subject notwithstanding, we
find there is no systematic examination of the effects of market struc-
ture on price movements in the expansions and contractions of the
postwar business cycles in the United States. The studies that have been
conducted are mainly cross-section experiments on annual data in
arbitrary time periods that bear little relationship to the actual turning
points of the cycle. Their results are inconsistent if only because the
regression coefficient of concentration is found to be either significant
or insignificant, or has an unexpected sign, depending on the sample
and the time period.'
* The authors are indebted to David F. Hendry for many helpful suggestions.
'H. J. DePodwin and R. T. Selden (1963), L. W. Weiss (1966), J. Dalton (1973) and S.
Lustgarten (1975); for Canadian data, W. Sellekaerts and R. Lesage (1973), and for the EEC
countries, L. Phlips (1969). The studies for the United Sates were conducted primarily in
231
232 BULLETIN
We propose to augment these findings by using monthly data to
examine price movements in the year following the peak and trough of
each of the six reference cycles between 1948-80. The focus on the
short run coincides with the average duration of postwar recessions, and
isolates more clearly the effects of the turning points on the longer
lived recoveries. For this purpose, we use a non-parametric time series
test, the H-statistic, developed by Holbrook Working and his associates
at the Food Research Institute (Brinegar, 1970). J. J. McRae and F.
Tapon (1979) imaginatively adapted the statistic to detect the presence
of administered prices in Canadian data.2 With this technique, it is pos-
sible to describe the time path of prices, and to distinguish between
them according to how market prices and administered prices are
expected to behave within a business cycle. It can demonstrate, for
example, if an increase in price between two points in time occurs
monotonically over the interval or in a series of relatively smooth
reversals - a sequence of price decreases and increases - or in an
unsmooth pattern of discrete jumps. The time path of price adjustment
is a neglected aspect of cyclical price research since investigators have
been interested primarily in cost adjusted prices, and cost information
is unavailable except on an annual basis. If the time path of prices can
be expressed as a valid measure of price adjustment, it can help to
resolve some of the conflicting findings on prices, and a step will have
been taken beyond the convention of perceiving relative prices only
through the amplitude of price change.
II. THE METHODOLOGY AND THE DATA
Price movements are measured by the H-statistic as ranges over speci-
fied intervals and subintervals of time, and are compared to expected
values drawn from an a priori distribution which is a random walk
process. Working called this distribution a 'random difference series' to
indicate that although the elements of the distribution are not them-
selves random, their first differences are. A major advantage of this test
response to Gardiner Means' testimony before the Subcommittee on Antitrust and Monopoly
of the US Senate Judiciary Committee (1957, 1959) that an 'administered price inflation'
contributed to increases in the price level between 1953 and 1959. Means' claims and the
counterclaims of his critics were essentially directed to the long run rather than to the cyclical
adjustment of prices for which the administered price hypothesis is relevant. A common defi-
ciency in these cross-section studies is the omission of the recessions, 1953-54 and 1957-58,
because of data inconsistencies with later periods due to census redefinition of industries in
1958. Yet, these two recessions ought to provide singular opportunities for a demonstration
of the hypothesis since they were sharp, brief contractions in a non-inflationary context during
which the price level failed to decline.
2study differs significantly from McRae and Tapon's in at least one important respect.
They assume the business cycle is a Monte Carlo process, and test for administered prices over
the cycle and over a long period. We make no such assumption, and test within the cycle during
recessions and recoveries. Our short run approach yields more elaborate hypotheses for each
phase of the cycle, and we must account for a larger variety of paths of adjustment.
To continue reading
Request your trial