23
area to total land holdings, (2) the number of years in the program, and (3) an interaction term between
the ratio of retired to total land area and the duration of the program.
The results in most specifications of the model reveal that greater intensity of program
participation increases off-farm labor participation (Table 3, columns 2-6). Specifically, a larger retired
land area and a higher ratio of retired land to total holdings lead to an increase in off-farm labor
participation. The results imply that a household composed of five adults that retires an additional 10 mu
of its cultivated land to the Grain for Green program will increase off-farm work by 0.5 persons (0.01 ×
10 × 5) (column 2). Likewise, when a household of five adults allocates 40 percent more of its cultivated
area to the program, the household will increase off-farm work by 0.5 persons (0.284 × 0.4 × 5—column
3). In the sample, the average number of adults per household is four. Duration in the program, by itself,
is not associated with greater off-farm labor participation (column 4), but when the program area and
duration are jointly considered, the longer a household has been in the program, the greater its increases
in off-farm labor participation (columns 5-6).
At the individual level, only the ratio of program area to total land holdings is associated with a
greater propensity to work off-farm (Table 5, columns 2-6). These results suggest that a household (that
the individual belongs to) retires all of its cultivated land will increase the likelihood of an adult member
working off-farm by nearly 10 percent (column 3).
Interestingly, we find that program intensity matters for changes in off-farm labor participation
but not for changes in work on the farm regardless of whether the data is at the household or individual
level (Tables 4 and 6, columns 2-6). This result may be driven by the binary nature of the measure of