Agriculture Reference
In-Depth Information
shares equaling the cropshare is 65 (25 out of 389), for 60-40 it is 6 (37 out of 580), and for
67-33 (13 out of 530) it is 2. This trend is robust for market and nonmarket inputs and for
all cropshare rules.
Estimation of Contract Choice
To estimate the choice between pure cropshare and input-output cropshare contracts, the
contract data are reorganized so that input shares are dichotomous; that is, input costs
are either fully borne by the farmer or shared proportionally. Thus, for any contract
i
the
complete model is
q
i
=
Xβ
+
i
(5.9)
1,
q
i
=
if
1
q
i
=
(5.10)
q
i
=
s
∗
,
0,
if
where
X
is a row vector of explanatory variables including the constant,
β
is a column
q
i
vector of unknown coefficients, and
denotes the true
underlying input share that is observed as either shared or not shared. The equality
i
is an error term. In equation 5.9,
0
means that the farmer's share of input costs equals his output share; the similarly observable
equality
q
i
=
1 means that the farmer pays all the input costs. Tables 5.8 and 5.9 show the
results of logit estimation of the model given by equations 5.9 and 5.10. In all estimates
the focus is on testing predictions 5.5 and 5.6. In the pooled sample used in table 5.8, the
dependent variable (QSHARE) equals one if the contract is a pure cropshare arrangement
and zero if the contract shares input costs and output in the same proportion. In table 5.9 the
dependent variable is also dichotomous but is specific to each of the eight inputs, so that the
variable equals one if the specific input (for example, fertilizer) is paid for by the farmer. It
equals zero if the costs are shared in the same proportion as output.
For table 5.8, a new pooled data set was created in which each input for each contract
became an observation. This resulted in 12,485 observations on input-sharing rules. Two
logit equations estimate the probability that the cost of any input will be fully borne by
the farmer rather than shared with the landowner. Prediction 5.6 is tested using the variable
HIGH VALUE, which equals 1 if the crop is either irrigated or a row crop, and is intended to
measure the value of the land attributes. The estimated coefficient is predicted to be negative.
The problem of land exploitation is more serious for highly valued land than for poor land.
By sharing input costs, the farmer's share of the output can be reduced (
q
i
=
l
∗
),
thus reducing the farmer's incentive to exploit the soil. In both specifications in table 5.8,
the estimated coefficient is negative and statistically significant, supporting prediction 5.6.
Prediction 5.5 is confronted using the variable MARKET, which is a dummy indicating
whether or not an input is purchased in the market. Inputs purchased in the market are
l
s
gets closer to
