by Paul A. Samuelson
AMONG the most striking uniformities yet uncovered in economic data are the relationships between various categories of expenditure and family income. Their regularity is substantiated by studies which go back as far as the nineteenth-century investigations of Le Play and Engel.1 In fact, so strong are these income effects that it is very difficult to find empirically the influence of price, the variable customarily related to demand by the economic theorist.
In recent years business cycle theorists have tended more and more to be of the opinion that investment is the strategic moving factor underlying fluctuations and determining the level of the system. This view implies as a corollary that consumption expenditure should be related passively to income. This is a fundamental assumption not only of the Keynesian system (e. g., the doctrine of the multiplier), but of most other schools as well.
Recent statistical material provides the opportunity to test this relationship, and numerous attempts have been made. Three general methods have been employed:2 (a) the analysis of budgetary data, representing a cross section of the different income levels at the same time;3 (b) the use of time series of national income, consumption, capital formation, etc.;4 (c) more or less plausible____________________