Variables that increase the likelihood of share tenancy (relative to fixed-rent) are highland ownership
(18%), proportion of female adults (201%), bullock ownership (10%) and the proportion of students
(45%).
The second stage (leasing) equations show that farming skills do not affect the extent of land
leased conditional on contract choice. The results also show that tenants who own more land are able to
lease-in more. This contradicts the land adjustment hypothesis within the group of tenants. In addition,
primarily agricultural households lease-in larger parcels while those with a more educated head of
households and those who engage in livestock or fishery activity lease-in smaller parcels. These effects
are consistent with opportunity cost arguments. Finally, tractor ownership has a positive effect on the
extent of leasing at least for share tenants. Table 5 re-estimates the multinomial logit model with two
alternative skill indices: 1) using the linear production function, and 2) using the frontier without asset
ownership variables. These estimates can be used as a simple test of whether the strong distributional
assumptions of the frontier method and potential correlation with risk aversion affect our conclusions. In
both these alternative models, the coefficient of the skill index is still strongly positive for fixed-rent
contracts and insignificant for share-rents. In fact, with a linear index, the t-statistics are even larger than
when the frontier-based index is used. Other coefficient estimates are also not qualitatively affected.
This simple test gives us confidence that our conclusion does not depend crucially on the structural
assumptions of the frontier model.
Bivariate Probit Models
Table 6 reports the results of the bivariate probit model with truncated sample. The probit for
leasing-in shows a large skill effect and strong evidence in support of land and input adjustment.32 The
32 If all leasing in is classified as share or fixed rent contracts , the first stage equation should have the same estimates
and standard errors regardless of whether the second stage is for share-rent or fixed-rent. However, our sample
38