Agriculture Reference
In-Depth Information
of contracts from Illinois and find that land with higher quality soil is less likely to be cash
rented. Both of these findings are consistent with predictions 4.2a and 4.2b.
Two control variables—ACRES and ROW*HAY—are included in all equations.
ROW*HAY was included because of the data overlap we mentioned earlier. In all cases
the coefficient for this variable was insignificantly different from zero. ACRES was in-
cluded to control for the possibility that the size of the farm influences contract choice. The
estimated coefficients for ACRES vary across the sample but show no statistically signifi-
cantly effect in the full sample. FARM INCOME is also included as a control in the smaller
samples.
29
The negative and statistically significant coefficients indicate that contracting
parties (farmer or landowner) with larger fractions of income from farming are less likely
to choose a cropshare contract.
Table 4.4 shows the estimated coefficients from a logit regression using the British
Columbia-Louisiana data. The table shows estimates for the full sample and for British
Columbia (155 contracts) and Louisiana (414 contracts) separately. The British Columbia-
Louisiana data sample sizes are sensitive to the inclusion of variables because those variables
relating to farm capital and wealth contain a number of missing observations. We estimated
the equations in table 4.4 without these variables and obtained results for the remaining
ones that were similar to those presented.
The estimates are similar to those found in table 4.3. For instance, as expected the
estimated coefficients for HAY are negative and (usually) statistically significant. The coef-
ficients for INSTITUTION are always negative and statistically significant. The estimated
coefficients for IRRIGATED, however, are less supportive than in the Nebraska-South
Dakota data. Only in the full and British Columbia samples are the estimates negative and
they are not statistically significant. For the Louisiana sample, the estimates are actually
positive. A plausible explanation is that for these data the IRRIGATED variable is dom-
inated by flood irrigated rice that does not capture the soil exploitation phenomenon the
same way it does for the Great Plains. The estimates for ROW CROP are, as expected, pos-
itive and statistically significant, except for the British Columbia sample.
30
Two additional
variables, RICE and TREES, are used to measure the potential for soil exploitation. Rice,
though not a row crop, is a crop for which soil degradation can be severe because of weed
and disease problems; hence, it is expected to be shared.
Fruit trees, however, provide a novel test of our hypothesis, because the source of
exploitation are the trees themselves, and farmers can extract more fruit in a given year at the
expense of crops two or three years down the road by improper pruning. The share contract
dampens this incentive and encourages proper tree maintenance.
31
Thus, we predict, and
find, that the coefficients are positive (and statistically significant) for RICE and TREES.
Orchards are found in British Columbia, but not in our Louisiana data. Hence the variable
TREES is left out of the Louisiana regression.
