out the reduction of explanatory power of each variable, in order to detect the
emasculation of factors. Therefore, this regression should not be called a ‘model’.
DSA I = G (COSTi, QUALITYi, FTFi)
COSTi: The average cost per bag (60kg) of rice in i prefecture12
QUALITYi: The share of Jishu-ryutsu-mai (high quality rice) in total amount of
distributed rice in i prefecture
FTFi: The share of households with full-time professional farmers in all households
conducting agricultural activities in i prefecture
Expected signs are as follows:
∂DSA i / ∂COSTi > 0
∂DSA i / ∂QUALITYi < 0
∂DSA I / ∂FTFi < 0
This conceptual model and regression need to be converted to empirical ones. To
this end, functional form was considered. Theoretically, the effects of two variables in
the model should be multiplied by the degree of political pressure. In this sense,
double-log form seemed to be appropriate. With regard to the regression related to
bureaucrats’ seemingly discretionary allocation formula, there was no theoretical ground
to adopt such a double-log form. However, the same functional form was adopted for
the latter regression, considering that the consistency of functional forms enables us to
compare appropriately.
12 In Table 1, there is merely the name of factor, “Productivity” mentioned. No concrete
data can be specified from this abstract notion. But, in this paper, we interpret this
notion as a meaning of cost and the data related to cost was adopted as one of the
selected explanatory variables.
14