second probit equation, for contract choice conditional on leasing-in, reveals difference between the two
types of contracts similar to those found in the multinomial logit model. In addition, our hypothesis that
tenants with greater skill enter fixed-rent contracts is supported although with less than 95% confidence.
Among tenants, those who own relatively more paddy land obtain fixed-rent contracts, while those who
own a relatively large number of bullocks and have a male head of household get share contracts. The
land extent equations show that households that have more family labor and own more tractors, and those
that are primary agricultural lease-in more land as sharecroppers. On the contrary, there appears to be
very little adjustment of inputs among fixed-rent tenants. In fact, the fixed-rent tenants that own more
paddy land lease-in larger extents. However, all models of contract choice agree that higher skill levels
increase the probability of entering a fixed-rent contract relative to share-tenancy as well as self-
cultivation. Table 7 reports the bivariate probit variant with a censored sample.33 Here again, the strong
skill effect in the probability of obtaining a fixed-rent contract is observed.
A Comparison of Contract Types
The theoretical model argued that sharecropping is a second-best efficient solution to a double
incentive problem that arises when the tenants are relatively unskilled. In the empirical analysis so far, we
established that only the relatively skilled tenants obtain fixed rent contracts. In this section, we use the
stochastic frontier estimates [Table 2] to directly establish the efficiency effects of different contract
types.
included some observations where both share and fixed rent is 0 although hire is equal to 1. This causes the slight
difference in first stage estimates and standard errors when both equations are jointly estimated.
33 Only the contract choice equations are reported for the sake of brevity.
39