11
recall data for 1999. If the results from the analysis using the subsample of households are consistent the
results from the analysis using the full sample, it would suggest that that recall bias is limited.
Off-farm labor allocation
By 2004, a large share of participating household members has reallocated their time to
off-farm work (Figure 3). In the 2005 survey, enumerators asked each respondent what the participating
household did with the time that was freed up after implementation of the program. According to
tabulations of the data, the largest share of respondents replied that they had reallocated the time of
household members to off-farm work (32 percent). The second most frequent response was that
households had allocated more labor to their remaining cultivated land (29 percent). In addition,
respondents stated that they had invested this freed labor time in leisure time (or time spent at home—11
percent) and (in conjunction with the in-kind grain compensation) to increase the scope of their livestock
enterprises (9 percent).
Descriptive statistics from the household data showed that off-farm labor allocation was
increasing for both participating and nonparticipating households (Figure 4, Panel A). From 1999
through 2004, individuals with off-farm jobs increased 13 percent for participating households and 8
percent for nonparticipating households. Because off farm employment is changing for both types of
households, it is clear that in order to evaluate convincingly the impact of the program on off-farm labor,
we need to control for the time effect and thus cannot simply compare postprogram levels of off-farm
work between the two groups. Among the individuals that had off-farm employment in 2002, we find
that 42 percent had jobs that were not local (implying that they were part of the migrant labor force and