The purpose of this paper is to propose alternative specifications which can be used for predicting imports by commodities for a large number of countries. Both "variety" and "scale" hypothesis have been made use of in suggesting simple cross-section relationships which can be easily used for predicting imports by commodities for any country of the world. Indeed, since detailed country characteristics are not taken into consideration in per capita income and population, it is likely that consistent under or overpredictions may result for a country when the estimated relationship is used. The extent of under or overprediction is reduced by using the variable elasticity formulation of the import demand function. Our estimates of cross-section relationships show remarkable stability in structural coefficients for the period 1967-1973 on the two-digit S.I.T.C. data. A set of eight different relationships is estimated and using 1970 cross-section relationships forecasts for 1973 and 1980 are given. Forecasts of imports of 45 commodities for 37 countries were derived and compared with the corresponding actual 1973 figures. The root mean square error predictive criterion suggests that per capita income and population in a constant elasticity form performs reasonably well.