Of the 182 observations, 43 (23.6%) have machinery ownership and operating costs which are
less than the custom rates.
Results
All four models were estimated using the entropy procedure in-sample, and then were
tested with the jackknife estimation procedure to derive out-of-sample model statistics. Table 2
displays the estimated coefficients from the in-sample models. The KAS state average custom
rates, and other calculated priors were included for a reference of how the estimated coefficients
relate to the reported priors, and are reported in the KAS column.
From the in-sample model statistics (Table 2) the reciprocal model has the highest CC
(0.874) and the lowest RMSE (27,598). The out-of-sample model statistics (Table 3) show that
the reciprocal model dominates, with the highest CC (0.869) and the lowest RMSE (28,204). As
such, the reciprocal model will be used in the remaining discussion of this research. On average,
across operations, the field operation coefficients decreased by 1.4% from the priors (custom
rates) for the reciprocal model. Considering this, and the average size Kansas farm in this
research (1,188 harvested acres), 25.5% I 1.255=1.241+I 33.027-------1-------I-0.014
^ ^ harvested acres J
would need to be added to published custom rates to arrive at the true cost to own and operate
machinery.
The estimated coefficients reported in Table 2 are the custom rates estimated by these
models, and the respective scale adjustment factors to take into account farm size. Two ways to
estimate a farm’s per unit machinery costs are available. The first option would be to multiply
the estimated coefficient for the operation of interest times the scale factor adjustment for the
farm taking into account the number of harvested acres of the farm. This results in the expected
11