Agriculture Reference
In-Depth Information
The counties and parishes in our data set closely approximate the conditions described by
the model in equations (6.6)-(6.8). First, each county or parish has several hundred or more
farmers using nearly identical technology. The mean number of farms per county or parish
is 456 for Louisiana, 650 for Nebraska, and 551 for South Dakota. Second, these crops are
sold in competitive world markets, so individual farmer output does not have price effects
and farmers face the same price variation. Third, the regions for which data are available are
reasonably homogeneous areas where idiosyncratic risks are not important.
22
As a result,
for the crops and regions we examine, variability in average regional yields (measured as
the standard deviation or coefficient of variation) is a strong, though not perfect, measure
of variability in the random input.
To implement this approach, we collected ten to fifteen years of time series data on the
variability in crop yields for thirteen different crops for the region of each observed contract.
Crop yield is calculated as total crop output in a region divided by total acres of the crop in the
region. By doing this we measure exogenous variability for each contract choice observation
in our data set. We calculate yield variability, which approximates exogenous variability
(
2
), in two ways: (1) STD (
Y
t
), the standard deviation of yield over time; and (2) CV(
Y
t
),
the coefficient of variation of yield over time. For the Nebraska-South Dakota contract
data, we calculate these measures at the county level and for larger, relatively homogeneous
regions that include several counties. For Louisiana, we use only parish (county) yield data,
and for British Columbia we use only yield data from rather large and less homogeneous
geographic regions. The definitions for these data are found in appendix A, along with their
means (and standard errors) for each crop.
23
The calculations and sources for these data are
explained in appendix A.
σ
6.3
Empirical Analysis: Risk Sharing and Contract Choice
To test these predictions we must couple data on exogenous variability with microlevel data
on individual farm contracts. We begin this section by describing the data and examining
some preliminary findings. We then examine the extent to which risk parameters (derived
previously) explain contact features by estimating their effects on contract choice. We also
examine related risk predictions based on wealth, institutions, and the extent of futures
markets. All variables used in this chapter are defined in table 6.3.
Farm-Level Contract Data
In addition to the data used in earlier chapters, some of the information here comes from an
additional survey, the
1979 British Columbia Farming Lease Survey
. As before, appendix
