Knowledge that Transforms

of economic time series similar to the observed high frequency series. A transformation that converts these series to serially uncorrelated stationary time series would therefore introduce the same conversion to the residuals. Although the discussion in this paper has been limited to the problem of distribution, a similar treatment can be given to the problem of interpolation and extrapolation by related series. As has been shown in the paper of Chow-Lin, the three problems can be treated simultaneously by properly defining the transformation matrix B.

The essence of Mirus' (1980) argument is that the endogeneity of interest rates and the spot exchange rate undermine the official forward position as an identifying restriction on the absolute slopes of the arbitrage (a) and speculation (,B) schedules in the market for forward exchange. My explicit assumption (on page 137) was that the spot exchange rate is indeed endogenous, which was why instrumental variables estimation was necessary. However, unlike Mirus, I assumed that the interest rates were largely predetermined on the grounds that monetary flows are large relative to overseas financial flows, i.e., the capital account is the tail and the domestic financial market is the dog. Thus I allowed only for the endogeneity of S, and insofar as r and ru. should have been endogenized too, Mirus' criticism is correct in principle. I doubt, however, whether in the U.K. case this makes much difference since Hutton (1977) reports that recognizing the endogeneity of r makes little difference. Even if we were to accept Mirus' argument I am not sure that his equation (1) is correct for a number of reasons. First, he implies that X is an extra parameter when as we shall see it is in fact a function of the parameters of the model I described in my paper. Thus no new parameters have to be identified econometrically. Secondly, if the expected return on forward speculation rises, i.e., if Se rises relative to F, why does Mirus assume that it is both interest rates and exchange rates that adjust and not the exchange rate alone? Thirdly, it is difficult to interpret his equation (1) when X = 1 which might be taken as the case of perfect capital mobility. In a small open economy ru. may be taken as exogenous. However, if we endogenize r and S on a common basis, as Mirus in fact does, we may use my original model to derive the correct reduced form relationship between F, Se and G. Equation (2) expressed the market clearing forward rate as

faces, however elastic it may be. But it is necessary to look beyond the elasticity of the curve to determine whether or not this is the case. Or consider two demand estimation situations, identical in terms of the supply elasticities involved. Differences in the variability of supply could make for large differences in the quality and appropriateness of OLS estimates. Yet a researcher relying on elasticity would be totally unaware of the differences in OLS bias in the two situations. Finally, concern about elasticities turns attention from the central question in choosing an estimator in the face of possible OLS bias: the relative shifting in the equations involved. That this is an important factor is well known, having been pointed out by Working as long ago as 1927, and by many others since. However, this essential factor is often overlooked in empirical situations.

Rolf Mirus, Arbitrage, Speculation, and Official Forward Intervention: The Cases of Sterling and the Canadian Dollar: A Comment, The Review of Economics and Statistics, Vol. 63, No. 1 (Feb., 1981), p. 147

In this paper we have shown that output effects can have a significant impact on estimated import price elasticities for aggregate classifications. Our results indicate that output effects could reduce estimated import price elasticities for Canada by as much as 10% to 15% for some aggregate classifications. These r&sults tend to understate the importance of output effects for estimated elasticities to the extent that they do not include any output effects of changes in exports and they tend to overstate the importance of these effects to the extent that supply curves for domestic production are less than perfectly elastic. However, they provide order-of-magnitude estimates of the bias caused by output effects when import price elasticities estimated for aggregate classifications are applied to disaggregate classifications of imports. These results also indicate the degree to which a general equilibrium approach may understate the impact on trade of an exogenous change in relative prices by doublecounting within-class output effects. Similarly, they indicate the degree to which import price elasticities estimated for very disaggregate classifications may overstate the case against elasticity pessimism because these disaggregate estimates account for fewer output effects. REFERENCES

Even if You's method for accounting for capital employment could be successfully implemented, his claim that his results support Okun's original estimate of the link between unemployment and the GNP gap would be incorrect. His claim of support for a three to one link between the employment ratio and output gap results from a failure to distinguish the response of real GNP and of PBS output to a change in the employment ratio. When his estimates are corrected for this, the results are in line with the significantly smaller link which others have recently observed. An alternative method of including capital in the production function-used by Rasche and Tatom (1977), among others-indicates that the Federal Reserve Board index of capacity utilization is an adequate measure of the capital employment ratio. Tatom (1980) has shown that the capital-labor ratio is procyclical in the context of this approach-a claim made by You, but not supportable by his analysis.

part of the analysis undertaken in Smyth (1980). A comparison of the mean values of X for the countries with the positive /3s, those with /3s that are negative and significant at the 5% level, and all those with negative /3s fails to support Otani's hypothesis. The mean values of X for the countries with positive /3s are greater than for the corresponding means for the negative /8 countries in three of the four comparisons, and trivially less (0.28 compared to 0.29) for Xs when the significantly negative /8 countries are used in the comparison. Table 1 also reports the coefficients obtained when /8 and XB and /8 and XS are correlated. The Otani hypothesis predicts negative correlation. However, the product moment and rank correlation coefficients are all positive; none are significant at the 5% level. The evidence, then, does not support Otani's explanation of the inter-country differences in his estimates of the relationship between real wages and output in the manufacturing sectors of industrialized countries. REFERENCES