predicted signs and are statistically significant. Taken together with the AGE coefficients,
the evidence in support of prediction 3.1b is strong in our data.
We use ACRES and VALUE as variables that control for the value of the contracted
plot of land, and we use FAMILY and INFORMATION to control for information the
parties have about each other. We expect that as assets becomes more valuable, the costs
of damage, theft, and poor farming become higher and the contracts are less likely to be
oral. Thus we expect the estimated coefficients for ACRES and VALUE to be negative. In
both equations the estimated coefficient for ACRES is negative and statistically significant.
VALUE is only available for the Nebraska-South Dakota data and has a negative and
statistically significant coefficient as well. 21 We have no predictions for the coefficients
on FAMILY and INFORMATION because it is not clear whether the information measured
by these variables enhances or inhibits contracting. We find a positive coefficient for
INFORMATION indicating more information leads to oral contracts. The coefficients
for FAMILY, however, differ between the two data sets. For Nebraska-South Dakota the
coefficient is positive (more likely to be oral) but the coefficient is insignificantly different
from zero in the British Columbia-Louisiana data.
Annual or Multiyear?
The choice between annual and multiyear contracts can be used to test predictions 3.2a
and 3.2b and thus estimate the importance of specific assets and market enforcement in
determining contract complexity. This contract choice is more directly related to specific
assets than to reputation, while the opposite is true of the choice between oral and written
contracts. The primary benefit of multiyear farmland contracts is more likely to be the
reduction of contract renewal costs. The primary cost of a multiyear contract is that it
requires additional clauses so that it can be adjusted to changing conditions. These clauses
also require additional enforcement effort.
We test predictions 3.2a and 3.2b by using logit regression analysis to examine the
decision to use an annual or multiyear lease. Table 3.3 shows the coefficient estimates for
these equations. Again, because the data sets do not contain the same variables, we estimate
different specifications for the Nebraska-South Dakota and British Columbia-Louisiana
IRRIGATED and TREES are again used as variables to indicate the presence of specific
assets. An annual contract is less likely to mitigate this possibility; thus prediction 3.2a
implies that IRRIGATED and TREES (only available for the British Columbia-Louisiana
data) will have negative coefficients. Our findings are mixed for these predictions. The
estimated coefficients for IRRIGATED is negative for the Nebraska-South Dakota equation
but are positive for the British Columbia-Louisiana equation. One possible explanation for