Agriculture Reference
In-Depth Information
Table 6.7 reports two sets of estimates for both British Columbia and Louisiana. These
equations use the same sample as did the equations reported in tables 6.5 and 6.6, simply
adding the additional wealth variables. The control variables include the crop dummies
and the variables that measure various characteristics of farmers, land, and landowners
(see appendix A). The upper panel shows logit parameter estimates of contract choice,
and the lower panel shows tobit parameter estimates of the farmer's contracted share of
output under cropsharing. For both logit and tobit estimation we use either (1) the aggregate
level of wealth (WEALTH), or (2) the component parts (BUILDINGS, EQUIPMENT,
and LAND) of total wealth. Under DARA the coefficient signs for WEALTH and other
variables measuring wealth should be negative for the logit estimates, but positive for the
tobit estimates.
Overall, as table 6.7 shows, the estimates give only limited support to the predictions
based on DARA. The logit estimates of the WEALTH coefficient are insignificantly different
from zero for both British Columbia and Louisiana, failing to support DARA. When the
wealth measure is broken into its component parts, the estimates become less consistent.
For British Columbia none of the estimates are significantly different from zero, although
the signs are not the same. For Louisiana all coefficient estimates are significant, but the
results are mixed with negative effects for BUILDINGS and LAND and a positive effect
for EQUIPMENT. Many of these estimates are consistent with effects of capital constraints
discussed in chapter 4, where we found evidence that as wealth increased (making capital
constraints less binding) contracts were less likely to be share arrangements.
The estimates from the tobit share equations in the lower panel of table 6.7 give similarly
mixed results. Like the logit estimates, the tobit estimates of the WEALTH coefficient are not
significantly different from zero for both British Columbia and Louisiana, failing to support
DARA. When WEALTH is broken into its three components, the results remain unfavorable
to DARA. Of the six estimated coefficients, only one is positive and statistically significant,
while four are insignificantly different from zero. To summarize, we find no consistent
evidence supporting the effect of wealth on contract choice under a model of risk sharing.
Further Tests: Futures Markets, Large Landowners, and Off-Farm Income
The risk-sharing predictions examined thus far have all been explicitly derived from the
principal-agent model presented at the beginning of this chapter. There are, however,
additional predictions consistent with this model that can be tested with our data, using
the variables FUTURES MARKET, INSTITUTION, and FARM INCOME. The tests that
we perform using these variables are scattered across many of the estimated equations in
this chapter and are sometimes, but not always, presented in the tables connected with the
previous tests. In this section we discuss this final set of risk-sharing tests.
