Agriculture Reference
In-Depth Information
of variation has no units and many of the crops we examine are measured in different units.
For example, hay is measured in tons while wheat is measured in bushels. Second, using
the coefficient of variation controls for differences in means even for those crops that are
measured in the same units. This is important when comparing dryland and irrigated crops,
where irrigation always increases average yields.
Using this measure, one finds that a greater CV indicates a more variable crop and thus
is predicted to be a crop that is more often governed by share contracts rather than cash
leases. For each state or province the crop CVs are listed in ascending order, from top to
bottom. Table 6.4, however, shows there is no clear relationship between the use of share
contracts and crops with inherently high CVs. In particular, table 6.4 shows the state-wide
and province-wide CVs and the prevalence of share contracts as fraction of all contracts, as
a fraction of leased acreage, and as a fraction of all farmland. Simple inspection of the data
shows no obvious positive correlation: Moving down each column of crops for a given region
shows no monotonic increase in the fraction of cropshare contracts. In appendix A (tables
A.9 and A.10) we present OLS estimates of the effect of CV on the extent of cropsharing
and confirm the intuition gained from simply visually inspecting table 6.4. Although these
numbers are crude and preliminary, they do not support the risk-sharing thesis.
24
Consider, for example, the case of sugarcane in Louisiana. Sugarcane has one of the
least variable crop yields of any crop in our data set (CV = 0.099), yet sugarcane land is
overwhelmingly cropshared (78% of all leases and 81% of all leased acres). By the risk-
sharing thesis, sugarcane is expected to be a crop that should be cash rented relatively more
often than other crops. Throughout the rest of this chapter we conduct more sophisticated
and exhaustive tests of the risk-sharing model; however, the results never differ from the
simple observations of table 6.4. There is no systematic evidence that contract choice
depends on risk sharing.
The Effect of Risk on Contract Choice
Although the previous analysis does not support the risk-sharing predictions, these infer-
ences are limited because of the level of data aggregation. In particular, because these states
and provinces are large geographic areas, heterogeneity could bias the estimates using crop
dummies. A more precise test is to examine the extent of share contracts for a single crop
across regions where natural parameters such as weather and pests directly influence the
yield variability of the crop. By selecting a sample of land contracts for which crops are the
same, we can examine how natural variability affects contract choice. Where the data allow
it, we separate irrigated plots from dryland plots even for the same crops because irrigation
uses different technology and reduces the yield variance. Natural variability for a homo-
geneous locale is measured using both CV and STD for two sizes of geographic regions.
