The purposes of this paper are twofold. First, we examine, using a unique dataset collected in the Tunisian village of El Oulja, what might be termed "all or nothing" separability (Benjamin, 1992) by testing, while controling for plot characteristics (Udry, 1996), the proposition that farm input use is independent of household characteristics. Second, we test for separability in the context of a model of class structure based on the seminal work of Eswaran and Kotwal (1986). In order to do so, we construct a model of class structure which offers an appealing representation of the typology of household types in the village when they are classified in terms of (i) their hiring in of wage labor, and (ii) their hiring out of family labor. We use a two-stage estimation technique in which we first estimate the probability of class membership as a function of household characteristics using discrete choice methods. In the second stage we estimate labor intensity per hectare using a Lee-Heckman procedure in which the inverse-Mill ratio from the first stage is included as an additional explanatory variable. As with the case of the test for "all or nothing" separability, the test for selective separability involves exclusion restrictions on household characteristics, although these are now conditional on class membership. Our empirical results strongly support the selective separability hypothesis as well as our theoretical model: most of the "action" in terms of non-separability stems, as one would expect from the model, from the class of "self-cultivators" who neither hire in wage labor nor hire out family labor. Our paper extends the work of DeJanvry, Sadoulet and Benjamin (1996) on Mexican ejidatarios to plot- as opposed to household-level estimation. Moreover, our paper provides an elaboration on and an empirical bridge to theoretical models, such as Roemer (1982), Eswaran and Kotwal (1986), and Carter and Zimmerman (1995) which examine how credit constraints and imperfections on factor markets shape the class structure of the agrarian economy. The implications and scope for government intervention in the context of market imperfections are also examined.
