Concentration and instability: Again

Concentration and instability: Again

Journal of Development Economics 33 (1990) CONCENTRATION 149-151. AND North-Holland INSTABILITY: AGAIN J. LOVE University c?fStrathclyde, G...

161KB Sizes 3 Downloads 74 Views

Journal

of Development

Economics

33 (1990)

CONCENTRATION

149-151.

AND

North-Holland

INSTABILITY:

AGAIN

J. LOVE University c?fStrathclyde,

Glasgow G4 OLN, UK

Final version received July 1988

In my 1986 paper in this Journal on the causal influence of commodity concentration on export instability I sought to shift discussion away from a cross-section framework and to devise a time-series approach. The results obtained from the time-series model provide support for a priori reasoning and contrast with the inconclusive results of earlier work. Masse11 (1990) is now concerned with the manner in which that paper proceeded: specifically, he is concerned that the shift from cross-section to time-series analysis involves the use of an invalid measure of instability. Following customary procedures for definition, I drew a distinction between actual and trend changes in earnings and instability was equated with deviations from trend. The time series of deviations from trend, both positive and negative, were then used as the dependent variable. Masse11 contends this is counter-intuitive on the grounds that ‘a large positive deviation from trend indicates high instability and a large negative deviation signifies low instability’. There should be no presumption, however, that a large negative deviation from trend signifies low instability. By definition, instability comprises both positive and negative deviations. In cross-section studies instability is described by a summary statistic, typically a sample estimate of the variance of deviations from trend where the process of squaring deviations from trend renders all deviations positive. The degree of instability is then identified by drawing comparisons across the summary statistics obtained for a sample of countries. ‘Low’ instability is judged against some benchmark such as the sample mean or inclusion towards the bottom end of the range of sample values. The shift to examining the circumstances of a particular country on a time-series basis obviously changes the reference point for relativities signified by the terms ‘high’ and ‘low’. Masse11 appears to predicate his contention of counterintuition on the view that negative deviations imply ‘lower’ instability than positive deviations. That is an incorrect interpretation of the manner in 03043878/90/$03.50

0

199GElsevier

Science Publishers

B.V. (North-Holland)

150

J. Love, Concentration

and instability

which I sought to represent instability. In my paper a large negative deviation is taken to represent an equal amount of instability as an equally large positive deviation. ‘Higher’ and ‘lower’ values are seen in terms of proximity to trend values, whether deviations are positive or negative. The sign on the deviation is not indicative of ‘high’ or ‘low’ instability but an indicator of whether in any given year the couhtry experiences an above- or below-trend change in earnings. With respect to the specific argument about the causal influence of commodity concentration it seems desirable to identify positive and negative deviations. If the conventional view that commodity concentration induces export instability is correct then, assuming we begin from a point where actual and trend earnings coincide and we have a given degree of commodity concentration subsequent above-trend earnings from a country’s major product(s) will result in higher measured commodity concentration and a positive deviation from trend of total earnings. Below-trend earnings from a country’s major export(s) will produce both lower commodity concentration and a negative deviation from trend in the earnings aggregate. It is difficult to see the value of Massell’s recommendation that absolute values of deviations from trend in total earnings should be employed. Below-trend earnings from a country’s major product(s) would be reflected in lower measured concentration and a positive deviation from trend in total earnings. Time-series testing then becomes meaningless. Masse11 does suggest, of course, that the time-series approach is invalid since the results simply reflect a statistical artifact. Paraphrasing my argument Masse11 points out, for example with respect to below-trend earnings, that ‘if the leading export(s) is (are) below trend, both the Gini coefficient and total export earnings will also tend to be below trend’. Statements such as this are, however, clearly conditional. The main feature of statistical artifacts is their definitional inevitability not their conditionality. In the section of my paper following that paraphrased by Masse11 conditions are set out under which other outcomes are possible. Where, for year t, U, and C2 represent the deviation from trend in total earnings and the Gini measure of concentration respectively, it is argued that: ‘If for example, earnings from the major export(s) exhibit relatively stable (along trend) changes, as suggested to explain the lack of significant results in cross-section studies, changes in U, and C2 will be determined by the behaviour of minor exports. Thus, windfalls (shortfalls) in earnings from minor exports will result in higher (lower) values for u, being associated with lower (higher) values for C”.’ [Love (1986, p. 243).] Perhaps what is at issue is not the definition step of formulating the nature of the underlying

of instability but the prior argument for the individual

J. Lace,

country. That, Rejoinder.

however,

Concrntration

is not

what

and instability

Masse11 sets out

151

to establish

in his

References Love, J., 1986, Commodity concentration and export earnings instability: A shift from cross-section to time-series analysis, Journal of Development Economics 24, 239-248. Massell, Benton F., 1990, Concentration and instability revisited, Journal of Development Economics, this issue.