Agriculture Reference
In-Depth Information
Table 5.8
Logit regression coefficient estimates on pooled data: Nebraska and South Dakota (1986)
(dependent variable=1iffarmer pays all input costs; = 0 if farmer pays input costs proportional to his output
share)
Independent variables
Full sample
Farmer sample
Predicted sign
CONSTANT
2.962
2.205
(30.07)
∗
(16.69)
∗
MARKET
−
1.648
−
1.241
−
(
−
35.43)
∗
(
−
20.46)
∗
HIGH VALUE
−
1.661
−
1.149
−
(
−
17.49)
∗
(
−
9.62)
∗
LEASED ACRES
−
0.415
−
(
−
4.12)
∗
Observations
12,485
3,940
2
Model
χ
1,693.41
562.21
Degrees of freedom
(2)
(3)
Note:
t-statistics in parentheses.
∗
significant at the 5 percent level (one-tailed test for coefficients with predicted signs).
less expensive to measure and to divide, so the estimated coefficient for MARKET is
expected to have a negative coefficient.
15
The estimated coefficients, in both specifications,
are negative and statistically significant, supporting prediction 5.5. We also estimated these
equations by modifying the MARKET dummy to include irrigation energy as a market
input. This specification slightly decreased the size of the MARKET coefficients, but it did
not appreciably alter the other estimates or their standard errors.
Prediction 5.5 can also be confronted using the variable LEASED ACRES, which mea-
sures the fraction of farm acres that the farmer leases in the observed contract. If the farmer
leases all of his acres from one landowner, then his ability to shirk on shared input use
is severely curtailed. However, if a farmer has other farm acres, either owned outright or
leased from another landowner, it is possible for him to divert shared inputs from one plot
of land to another and bill the landowner for the extra costs. LEASED ACRES is used only
in a sample in which farmers responded to the survey. As prediction 5.5 implies, we found
the estimated coefficient for LEASED ACRES to be negative and statistically significant.
This finding indicates that farmers who lease from different landowners or also own some
land are less likely to use a contract in which input costs are shared.
Predictions 5.5 and 5.6 are tested again in table 5.9 which shows separate logit estimation
for each of the eight crop inputs. As in the pooled data set, we estimated two equations for
each input. Again, the estimated coefficients for HIGH VALUE and LEASED ACRES are
all predicted to be negative. Since the dummy variable MARKET cannot be defined in the
