14. As we noted in the introduction to this chapter, we assume the inputs are independent for several reasons.
An additional reason is that this assumption yields testable predictions supported by the available data (Friedman
15. Because θ ∼ ( 0, σ
) and because risk neutral parties maximized expected profits, the error term, θ , vanishes
from all first-order conditions.
16. Occasionally cash rent is paid in two or three installments during the growing season. This payment is a sunk
cost and thus does not influence the farmer's choice of e or l .
17. Although we use a single period model to address soil exploitation, recent studies have used intertemporal and
dynamic models (Dubois 2002; Ray 1999). Consider a simple intertemporal framework in which L is labor effort
is the current output with
is the forgone future output with
is the cost
L ∗ ) must satisfy
F + V = C , however, a short term cash rent
of effort with
0. The first best effort level (
L s >L ∗ . This framework incorporates our land exploitation incentive but
does not account for effort moral hazard or output monitoring.
18. This assumption and its implications are inconsistent with our data set. Dollars are rarely shared, and inputs are
not always shared in the same proportion. Eswaran and Kotwal (1985) assume that net revenue is shared, which
implies that input costs are shared in the same proportion as the output and that the costs of output division are
zero. See chapter 5 for a complete discussion on variations in output shares and the relationship between input and
19. Underreporting the quality and quantity of crops and livestock in share contracts is an ancient problem. In
the Old Testament, Genesis 30:37-43 records how Jacob became “exceedingly prosperous” at the expense of his
partner and father-in-law, Laban, by manipulating the breeding stock in such a way that he always received a
higher-quality share of the animals.
20. As noted in chapter 1, we assume that competition among farmers and landowners is strong enough that only
the most valuable contract is chosen. Contracts that failed to maximize joint wealth would simply not survive.
21. AGE, ACRES OWNED, and INSTITUTION are only available from surveys of farmers; ABSENT is only
available from surveys of landowners.
22. DENSITY is only used in the U.S. samples because the regions in British Columbia are too large to be
meaningful measures of urbanization.
23. We do this because the Nebraska-South Dakota data leave open the possibility for the ROW CROP and HAY
variables to have a slight overlap. In the British Columbia-Louisiana data, this design problem does not exist and
so the variable is excluded from estimates in table 4.4.
24. Many farmers now own semi-tractor trailers that allow them to haul grain hundreds of miles to large central
locations. This change will likely increase measurement costs of output and reduce the attraction of cropshare
25. Hay is often used as a livestock feed by the leasing farmer, making third party measurement even more difficult.
26. Irrigation for rice in Louisiana, however, does not fit this description, since it is for pest control rather than to
supplement soil moisture.
27. Indeed, for hay crops, the farmer's ability to deplete the soil by excessive tilling is minimal, because the soil
is seldom manipulated; the crop is simply harvested periodically.
28. The irrigation variable could also be explained by risk sharing. Irrigated crops are less variable and hence are
more likely to be cash rented under the assumption of risk aversion. In chapter 6 our data refute this interpretation.
29. The finding in the farmer sample might be interpreted as a refutation of risk sharing because it indicates that
farmers who are more specialized in agriculture choose contracts (cash rent) that put more residual ownership on
their actions (Allen and Lueck 1992b).
30. Few row crops are grown in British Columbia, so this finding is not surprising.
31. As we note later in our discussion of European agriculture, cropshare contracts have a centuries-old tradition
in vines and trees.
32. As noted previously, we devote chapter 6 to exhaustively testing risk-sharing predictions.
F = C and effort