Discussion: What We Do and Do Not Know about Food Habits of American Families

Abstract

further stratification is feasible. For example, use age as well as size, employment status of members including the housewife, and confine the analysis to a single city to eliminate possible price effects. Such things would tend to increase the cost of the data. question therefore arises as to whether refinement of estimate is worth the additional cost. It may be that the crude estimates currently available as to income effects are adequate for most purposes. There seems no doubt that the increase in pounds of food and nutrients with income is very low, that some of the increase in food expenditures occurs because of rise in quality and that the income elasticity of the bundle of things now arbitrarily defined as food is appreciably less than 1.0. With data available the estimate probably can be improved somewhat. To do so careful attention needs to be given to the characteristics of what is being measured, to the heart of the scientific method, and to the tools of estimation, i.e. statistical methods. issue of how great accuracy is needed depends on why the knowledge is needed. Certainly a refined estimate of the income effect may not improve a forecast unless it is matched by a refined estimate of cross elasticity of demand for various foods with insights on changes other than prices such as those wrought by the new technology and by consumer education.30 Those striving to improve the estimate of income effects on food consumption can take some comfort from the difficulties encountered by others striving to estimate its effect on total consumption expenditures, and from the observation by Lynd that The stubborn unavoidable fact that confronts the social scientist at every point is the presence, in every institutional trait that he seeks to analyze, of a subtly graded, unevenly distributed and continually changing array of behavior.31