Agriculture Reference
In-Depth Information
comparisons are shown in table A.6. In many cases the means are nearly identical; in all cases they are within one
standard deviation of each other. We conclude that our samples are quite representative of agriculture in British
Columbia, Louisiana, Nebraska, and South Dakota.
A.3
Crop Yield Variability Data
Data on yield variability come from state and provincial agricultural statistical offices. We collected a times series
(1975-1991) of per-acre yields for each crop at the county and parish level for Louisiana, Nebraska, and South
Dakota. The precise number of years varied across crops because of data availability. The most common unit of
measure is bushels and tons, although these vary by crop and jurisdiction. For Nebraska and South Dakota, the
same data were collected for “regions,” which from five to ten counties and their compositions were drawn from
the respective state department of agriculture crop reporting systems. Regional data was unavailable for Louisiana
crops. For British Columbia, yield data are only available for each of the eight “Census Agricultural Regions”
most of which are larger and more heterogeneous than the American states, were used in the study. As a result,
these regions are of little use for the risk tests that we used for the U.S. data. Table A.7 shows the variable means
for the crop yield data, while table A.7 shows the distribution of yield variability (measured both by coefficient of
variation and by standard deviation) for some of the major widespread crops in the four jurisdictions.
OLS Estimation of Cropsharing at the State-Province Level
Table A.9 shows OLS estimation of three separate equations for each state or province. The dependent variables
in these equations are the fraction of cropshare contracts, the fraction of cropshare acres among leased acres, and
the fraction of cropshare acres among all acres farmed. These represent three different measures of what it might
mean to increase sharing when a crop becomes more uncertain. Overall, the regression estimates fail to reject the
null hypothesis that CV does not influence contract choice; in other words, we find no support for the hypothesis
that higher yield variability leads to more cropsharing. We also pooled the data and obtained the OLS estimates
found in table A.10, where absolute t-statistics are in parentheses, the adjusted
2
= 0.42, and the overall F-value
= 4.97. FRACTION CROPSHARE = the fraction of all land in cropshare contracts, while the other independent
variables are state-provincial dummies for four of the five samples.
R
Price-Yield Correlations
In chapter 6 we discussed the possibility that prices and yields might be negatively correlated. Table A.11 shows
the price-yield correlations for various crops and across various regions. In general, the table shows no statistically
significant correlation.
Additional Risk Regressions
In chapter 6 we estimated many regressions to test the relationship between crop yield variation and contract
choice. We also estimated several different specifications using this microlevel data, which are shown in table
A.12. The dependent variable in our estimated equation, CONTRACT, is dichotomous (1 if cropshare contract, 0
if cash rent), so we use logit regression. Data limitations prevent us from using CVs for each crop for state- and
province-level samples, so we instead use crop dummy variables to estimate the effects of exogenous variability
on contract choice.
In table A.12, equations (1) - (6) show that crop riskiness does not successfully explain the use of share contracts
among Nebraska, South Dakota, British Columbia, and Louisiana farmers. Of the twenty-eight estimated crop
dummy coefficients in equations (1) - (5), only four are consistent with risk sharing. In all equations the left-out
crop dummy is WHEAT. Each equation includes a dummy variable OTHER CROPS whose parameter estimates
are not reported because the crops included in this dummy vary across regions: In Nebraska it might be rye or
sunflowers, while in Louisiana it might be sweet potatoes. As such, comparisons of it with WHEAT have little
meaning. Testing the risk-sharing hypothesis requires ranking crops by CV for the appropriate state or province.
Because the crops and the CV rankings vary across these jurisdictions, it is not possible to visually compare the
