220
BOOK REVIEWS
regressions and in four cases the coefficient is of the wrong sign. Relative price is discarded from the model when projections are made so that the entire weight of the study is thrown onto the income variable. Clearly this variable cannot stand the strain, as will be seen below. It is hard to accept that price does not matter to Hong Kong's cost-conscious consumers, to judge from recurrent public outcries over trivial price increases. Moreover, the study itself gives indirect support to the idea that price can explain variations in consumption. In the case of some commodities such as vegetables, the income coefficient is strongly positive. Bearing in mind the importance of vegetables in Hong Kong budgets, we would expect the income effects of price changes at least to be fairly significant. Would it not have been worth experimenting with alternative specifications of the demand equations which included price itself, rather than relative price, particularly since the latter showed insufficient variation to be picked up in the regressions? This method would have produced multicollinearity as both incomes and prices have strong trends but this could have been dealt with by substituting the independent cross-sectional estimates of income elasticities derived from the 1963/4 Household Expenditure Survey. The failure to make use of these estimates other than for checking purposes is in any case a fault of the study.
Because of the failure of prices as an explanatory variable, the study depended entirely on the estimated response of the various commodities to income changes, combined of course with projections of incomes over fifteen years. How successful have these predictions been? I have taken advantage of the lapse of time between the date of the latest data used (1965) and this review to check the accuracy of the 1970 projections for the three main commodities, rice, vegetables and cotton. Realised imports in 1970 differed from the forecasted figure by 10% in raw cotton (realised imports: 160,000 tons -- predicted imports: 178,000 tons), 33% in rice (realised imports: 327,000 tons predicted imports: 485,000 tons) and 53% in vegetables (realised imports: HK$197 million -- predicted imports: HK$129 million). Errors of prediction are proportional to the distance from the base data so that if errors are of this order after only five years they will in some cases be truly wild after fifteen.
Why have the authors been unsuccessful in performing an admittedly difficult job? One could cavil at the use of per capita income