270 Communities and Social Capital
but unreported analysis and perhaps their findings still stand. If this is the case,
concern with such methodological issues has been so long-standing in the litera-
ture that the method for dealing with them should have been discussed.
The study also raises other questions about the analysis and the authors’ intent.
Barnes and Blevins (1992, p336) assert that they ‘wanted to analyze the potentially
different impact of farm structure on the well-being of counties with different
levels of farm dependency’. If this is the intent, then interaction terms (for farm
dependency by the farm structure independent variables) should have been inserted
directly into the regression model. A related point concerns their statement that
‘there is an important interaction effect between the farm and nonfarm variables’
and that this effect can be discerned by examining the amount of variance explained
(Barnes and Blevins, 1992, p345). However, interaction does not refer to the
increase in an R-squared coefficient that one gets by adding additional variables.
Interaction effects refer to the idea that the effect of independent variable A on the
dependent variable is conditioned by the level of independent variable B. Interac-
tion effects must be tested by adding an interaction term (A × B) and testing for its
significance. If the authors believe that there are important interactions in their
data, then these should have been tested for appropriately.
In addition to raising methodological questions, there is a more significant
problem. The Barnes and Blevins (1992) article does not advance the literature
conceptually but rather reflects the earlier, post-war period of research in which
analysts simply replicated the Goldschmidt hypothesis. There is little attempt to
explain from a theoretical standpoint why the effects of farm structure may vary in
different geographic contexts or to extend the topic through adding new literature
and insights. Arguments about the need to incorporate non-farm structure and to
control for farming dependency are repeated from prior work. Thus, no new
ground is covered and research slips back to an earlier era.
New Directions: Whither the Goldschmidt Hypothesis?
Goldschmidt’s (1978a) work has achieved a status seldom accorded other research.
The case study of Arvin and Dinuba has taken on mythic proportions and some
studies treat the relationships as if set in stone. Over a decade of research has
shown, however, that Goldschmidt’s hypothesis cannot be applied unconditionally
across different regional and historical settings. It is now also well-known that
researchers must acknowledge how the structure of agriculture and the factors that
shape well-being have changed over time. The literature in the Goldschmidt tradi-
tion appears to be exhausting itself, particularly if one looks at the proliferation of
studies and regional research projects in the late 1970s–early 1980s as compared
with the present. Rather than resurrecting old debates, researchers should focus on
the more significant questions raised by the Goldschmidt (1978a) study. How are
economy and society linked? What do changes in the economy mean for rural