BOOK REVIEWS
J
221
for prediction purposes without regard to the average age of the average mouth that is to eat all these foodstuffs, but the basic reason, I believe, lies in the methodology adopted in this study. Economists' faith in sophisticated forecasting methods such as those employed here which seek out underlying relationships with the guidance of economic theory and the use of econometric techniques - rests on the ability to determine what these relationships are on the basis of past data and on the belief that the explanatory variables (income, population and price in this study) can be forecast with greater accuracy than the variable one is primarily concerned with, in this case the demand for commodities. If either or both of these twin pillars are suspect it may well be that a naive method, such as extrapolation of past trends, is both cheaper in terms of research resources and more accurate. In the present case there would seem to be strong reasons for suspecting the basis of the study. Regarding the econometrics one can fault the specification of demand relationships which yield improbable price elasticities already discussed; the absence of any tests of and methods of dealing with the serial correlation likely to occur in time series analysis; the absence of coefficients of determination which would indicate the explanatory power of the demand equations; and the use of an explanatory variable which is itself a speculative 'guesstimate' of per capita income and certain to give rise to biased estimates of per capita income coefficients of the errors-in-the-variables type.
As for the assumption that the explanatory variables per capita income and population — can be predicted with greater accuracy than the demand for commodities, one must admire the authors' optimism. Hong Kong's demographers greatly overestimated the 1966 population through a failure to take into account age-specific birth rates. Even granted the improved professionalism in this area over the last five years it remains true that changes in age of marriage and recourse to birth control can throw out demographic projections. Per capita income projections are far shakier since we do not even know what the trend to date has been with much certainty. Faced with these data problems the authors might well have opted for a naive forecast on the basis that things will stay as they are, only becoming more so! Future requirements of the various commodities could have been predicted on the basis of past values alone, using some function which is found to fit past data best. Such a method would not tell us anything about the underlying forces working
BOOK REVIEWS
J
221
for prediction purposes without regard to the average age of the average mouth that is to eat all these foodstuffs, but the basic reason, I believe, lies in the methodology adopted in this study. Economists' faith in sophisticated forecasting methods methods such as those employed here which seek out underlying relationships with the guidance of economic theory and the use of econometric techniques - rests on the ability to determine what these relationships are on the basis of past data and on the belief that the explanatory variables (income, population and price in this study) can be forecast with greater accuracy than the variable one is primarily concerned with, in this case the demand for commodities. If either or both of these twin pillars are suspect it may well be that a naive method, such as extrapolation of past trends, is both cheaper in terms of research resources and more accurate. In the present case there would seem to be strong reasons for suspecting the basis of the study. Regard- ing the econometrics one can fault the specification of demand rela- tionships which yield improbable the price elasticities already dis- cussed; the absence of any tests of and methods of dealing with the serial correlation likely to occur in time series analysis; the absence of coefficients of determination which would indicate the explana- tory power of the demand equations; and the use of an explanatory variable which is itself a speculative 'guesstimate' of per capita income and certain to give rise to biased estimates of per capita income coefficients of the errors-in-the-variables type.
As for the assumption that the explanatory variables
per capita income and population — can be predicted with greater accuracy than the demand for commodities, one must admire the authors' optimism. Hong Kong's demographers greatly overestimated the 1966 population through a failure to take into account age-specific birth rates. Even granted the improved professionalism in this area over the last five years it remains true that changes in age of marri- age and recourse to birth control can throw out demographic pro- jections. Per capita income projections are far shakier since we do not even know what the trend to date has been with much certainty. Faced with these data problems the authors might well have opted for a naive forecast on the basis that things will stay as they are, only becoming more so! Future requirements of the various commodities could have been predicted on the basis of past values alone, using some function which is found to fit past data best. Such a method would not tell us anything about the underlying forces working
No comments yet.
Private notes are available after approval.