characteristics of the farmer, the land, the climate and, in the case of share-tenancy, the landlord.22 For
example, landlords may compensate for the low skills of the less able tenants by providing some or all of
the skilled inputs in a sharecropping arrangement. Therefore, two farms can have the same technical
efficiency even though the tenants’ skill levels are quite different. An accurate measure of farming skills
must isolate the farmer’s contribution to technical efficiency. The one-sided error term, u, can be
expressed as follows:
U = - log TE1 = a 1 Fi + ( ads + afdf ) Li + a 3 H1 + ei [48]
where F is a vector of farmer characteristics, L is a vector of landlord’s characteristics and H is a vector
of geo-climatic variables that include soil quality, gradients, irrigation, rainfall, humidity and temperature. ds
and df are dummy variables for share and fixed-rent tenancy and ei is a random i.i.d. error term which is
uncorrelated with the included variables.
Farmer skills can be defined as the farmer’s contribution to technical efficiency, or a1Fi in
equation [48]. In order to fully capture the many dimensions of skill, the vector Fi includes not only
variables such as education and age, but other farmer characteristics that are likely to be correlated with
unobserved aspects of skill such as motivation and ambition. Ownership of non-farm assets and types of
non-farm activity are included as potential indicators of such attributes. Since the weights of each factor
are econometrically derived, the best approach is to include as large a set of farmer characteristics as
possible in the efficiency equation. A consistent estimate of skill can be obtained only if the parameter a1
is consistently estimated. Unfortunately, since our data set does not match tenants with landlords,
landlord’s characteristics cannot be included in equation [48]. Because tenants and landlords are matched
in a systematic non-random manner, the estimate of a1 will thus be biased. We avoid this problem by
estimating the frontier only for the self-cultivators to obtain the coefficient estimates and then using these
22 Some studies have interpreted the technical efficiency term as a measure of skills. Kirkley et al. [1998], for example,
argue that differences in technical efficiency account for differences in skills between fishing boat captains who use
27