An Empirical Analysis of the Relationship between Inflation and Economic Growth in 12 Countries, 1950-1969. A Comment

Abstract

This comment shows that some of the results in Paldam's article An Empirical Analysis of the Relationship between Inflation and Economic Growth in 12 Countries, 1950-69, The Swedish Journal of Economics, Vol. 75, 1973, No. 4, pp. 420-27, [2], are unreliable due to the applied methods of analysis and to the kinds of errors in variables which are likely to be present in his data. The conclusions drawn by Paldam are summarized in five points (Paldam [2], pp. 425-26): (1) In the long run there is no (or at most a weak) relation between price increases and real growth as regards GNP (as well as private consumption). (2) In the very short run-i.e. within the same year-the relation is negative; weakly negative as regards GNP (and clearly negative as regards private consumption). (3) After particularly large (small) real growth there will normally follow particularly large (small) price increases during the next few years. As regards GNP the effect is greatest after one year (and as regards private consumption not until two years afterwards). (4) After a particularly large (small) price increase there will normally follow particularly small (large) real growth during the next few years. (5) More inertia is evident in the price increases than in real growth. The study is based on five data series, collected for each country. On the basis of these series, fifteen series of growth rates have been formed. In order to simplify the discussion we will consider the following case. Let Yt be GNP in current prices and Qt be GNP in fixed prices, while Pt is the GNP deflator. Obviously Qt YtIPt. The study by Paldam [2] can then be presented in a simplified version as a study of correlations between Pt-, and q t-T2 t = 1, 2, 3 ..., T; m, -r2 =0, 1, 2, where Pt is the inflation rate and qt the rate of growth in real GNP. It is a well-known result that errors in variables imply biases (see Johnston [1], Ch. 9), but it seems that in this case the biases will become particularly