Agriculture Reference
In-Depth Information
Table 6.2
Summary of incentives for risk sharing
Effort moral hazard
Land moral hazard
Risk sharing
Contract costs
Cropshare contracts
Yes
No
Yes
High
Cash rent contracts
No
No
No
Low
Implementing the Predictions
In the standard principal-agent model, the farmer's utility depends on the level and variance
of income derived from the land contract. The farmer's income, in turn, depends on the
price of the crop (
)
will influence the choice of contract, yet the above predictions ignore income variability
(because we normalized the output price to unity) and focus exclusively on output (crop
yield) variability. In practice, we are able to ignore price and income variability in the tests
we perform because of the way we develop our contract data sample.
16
In our tests of predictions 6.3 and 6.5, we use contract data in which all farmers grow
the same crop. Moreover, all of these crops are sold in world markets in which individual
farmers are price takers. This means that there is no variance in price across farmers and
that income variance is equivalent to yield variance. Generally if
P
) times the crop quantity (
Y
). In principle, the variance in income (
PY
P
and
Y
are independent
2
2
2
2
2
2
P
2
is the variance and
then
is the mean. This
implies that data on the mean and variance of output prices would be needed to test the
standard risk-sharing model. When price is constant,
Var(PY)
=
σ
P
σ
Y
+
σ
P
µ
Y
+
σ
Y
µ
, where
σ
µ
2
2
Y
, and price data
are still required to measure the variance in income. However, if price has no variability
across the observations because of price-taking markets, then two possibilities emerge.
First, if price is literally constant then
Var(PY)
=
P
σ
2
Y
is a scale parameter and
incorporating price data in a regression would simply rescale the estimated coefficients on
yield variability. Second, if price is random, but the same for all observations, then the
regression constant incorporates the price data.
Even when the test conditions are so controlled that price variability can be ignored, other
issues for the appropriate measure of yield variability remain. Testing predictions 6.3 and
6.5 requires data on the variance in the random input,
Var(PY)
=
kσ
, where
k
2
. We measure this exogenous vari-
ability by using data on crop yields.
17
Still, successfully conducting these tests has potential
problems because production (crop yield) variability has two sources: 1) exogenous vari-
ability that cannot be influenced by the contracting parties (variability of
σ
2
), and (2)
endogenous variability that is influenced by the actions of any contracting party (variability
of
θ
or
σ
e
). The impediment to performing tests of risk sharing lie in the difficulty of finding a
