Farm Family Response to a Negative Income Tax: Discussion

Abstract

The income maintenance experiments have been unique in terms of public policy development. Prior to the development of new legislation on income-conditioned transfers, Congress was willing to fund an experimental version of such a program, including the analysis of behavioral changes induced by the experiment. The primary reason is the divergence of opinion about the impact on hours of work and earned income of household members covered by such a plan. The inclusion of a rural experiment in the group of income maintenance experiments has meant that poor rural people have not been forgotten in the planning of new welfare legislation. Also, it has provided an opportunity to develop administrative procedures for handling problems that are somewhat unique to rural areas-a large proportion of selfemployed persons and rather thin asset markets. These problems seem manageable. Given that the primary objective of the experiment was to determine the effect on hours of work, I am puzzled about the small amount of resources that were allocated to obtaining data on the hours of farm work by farm family members. Weekly hours of farm work over a year by farm operators, by their wives, and by other farm household members have a relatively large variance compared with weekly hours by wage workers. Part of this variation is seasonal, but much of it is intraseasonal. Drawing upon experimental design theory, one quickly recognizes that the estimate of mean weekly hours for a quarter based upon a sample size of one has a large sampling error. Thus, one need not fall back on defects of the respondents, as Primus does in suggesting a Hawthorne-type effect, to be skeptical of levels and differences in reported hours of farm work by farm household members.1 Th se measurement-error problems with reported hours have undermined much of the empirical analysis that was oriented toward the primary objective of the experiment, including the research reported by Primus and Kerachsky.2 The research methodology employed in the three empirical papers troubles me as an economist and an empirical researcher. As scholars, we must carefully blend economic theory and econometric techniques in our data analysis in order to advance knowledge and understanding of systematic social behavior, but the path to this end is not always well marked. I have three major concerns. First, some empirical relationships reported have weak foundations in economic theory. It may be an interesting question as to how experimental and control farm households differ in their crop and livestock inventory holding (sales) patterns, debt-to-asset ratios, sources of credit, etc., but I see no convincing reasoning behind the so-called control variables included in