Gerhard Tintner: Review of Allen and Bowley’s Family Expenditure (1936)

Gerhard Tintner’s review evaluates Family Expenditure: A Study of its Variation as an exemplary integration of economic theory, statistical technique, and detailed knowledge of evidence. His opening praise establishes the standard by which he judges the book:

This book is one of the few studies of economic subjects which can be called "econometric" in the true sense of the word.

The book draws on 21 collections of family-budget data from 13 European and American countries, alongside an original investigation conducted by William Beveridge. Tintner follows its progression from average expenditure, through individual variation around averages, to demand theory and a mathematical appendix. He particularly welcomes the Beveridge material on English professional and middle-class households and hopes for its fuller publication.

The review’s principal reservation concerns the simplifying assumption connecting theory to measurement. Allen and Bowley adopt the recent Hicks–Allen theory of value but use a linear preference scale, equivalent in utility terms to a quadratic utility function, throughout their investigation. Tintner acknowledges its unexpectedly broad success while identifying a missed opportunity to test its limits:

It would have been interesting if the authors had investigated them with the help of the known statistical tests for linearity which involve a comparison between the correlation coefficient and the correlation ratio.

For Tintner, departures from linearity are exceptional, but their statistical examination would strengthen the study’s theoretical foundation. His criticism therefore complements his praise of the book’s clear diagrams and its use of least squares, the χ² test, frequency distributions, and multiple correlation.

The substantive results range across income elasticity, the relative urgency of needs, relations between commodities, and differences in tastes across classes and nationalities. Expenditure variation also bears on the reliability of cost-of-living indices. Tintner finds the frequency distributions especially interesting: normal distributions occur more often than expected, while Gibrat’s distribution is useful where logarithms, rather than the observations themselves, are normally distributed. What matters is that these techniques remain answerable to economic questions:

They never calculate statistical parameters for their own sake but are well aware of the economic problems which their study tries to solve.

Tintner locates the book’s achievement less in surprising discoveries than in establishing familiar findings through sound economic reasoning and reliable statistical methods. Its quantitative estimates, despite random variation, improve substantially on earlier work. The review thus makes a methodological argument through its assessment of a particular study: successful mathematical economics requires methods suited to its subject, not borrowed scientific prestige.

It shows how it can be done successfully, not by aping the natural scientists but by the development and application of methods which are particularly fitted for the difficult task of economic and social problems.