The estimation method adopted on two regressions is principally OLS. But as
mentioned before, in some cases where hetero-skedasticity was observed, White hetero-
skedasticity consistent covariance was employed. Estimation tests were conducted by a
cross-prefecture multiple regression over twenty-two years. Data were collected from
official statistics and government-published data on the rice set-aside program
(“Suiden-riyou Saihen-taisaku Jisseki-chousa Kekkahyou” [Surveys of the situation of
the implementation of the rice set-aside program] (1980-2001)).
With regard to explanatory variables, PF and FTF, the data from statistics of the
Agricultural Census (1975-2000) were employed. The data related to EI and COST
were calculated from statistics of the “Surveys of Rice Production Cost” (1975-2001).
The data of QUALITY were obtained from the “Beika ni Kansuru Shiryo” [documents
relating to rice price] (1979-2001) released from the Food Agency. As for PF, extremely
small farms, which were operated for the purpose of hobby farming, were excluded. As
for FTF, full-time but elderly farmers, i.e. over sixty, were excluded. As for COST, since
the data in years when severe cold weather damaged the yield of rice were abnormal,
the data in such years were replaced by those in normal years.
In determining a dependent variable, there were several data constraints. The rice
set-aside program started in 1971, however, the data in the first decade are not available.
Therefore the observation period is from 1980 to 2001. The degree of set-aside
allocation was represented by the figures of set-aside acreage divided by total acreage of
paddy fields. This figure was extremely high in Tokyo, Osaka and Kanagawa
prefectures because land conversion to non-farm use from farm use has rapidly
progressed due to urbanization in these areas. Therefore, these prefectures were
excluded from the sample.
Results
16