44
Table 7. Program impact on off-farm and farm jobs, treatment indicator interacted with
quartile dummies of asset holdings, 1999 and 2004.
|
Household |
Individual | |||
|
(1 ) off-farm |
(2) farm |
(3) off-farm |
(4) farm | |
|
poorest in asset value in 1999 |
0.515 |
-0.431 |
0.198 |
-0.164 |
|
(dummy) x treatment x year2004 |
(2.54)** |
(1.93)* |
(2.49)** |
(1.97)** |
|
second poorest in asset value in 1999 |
0.331 |
-0.341 |
0.197 |
-0.082 |
|
(dummy) x treatment x year2004 |
(1.64) |
(1.54) |
(2.49)** |
(1.03) |
|
second richest in asset value in 1999 |
0.197 |
-0.521 |
0.115 |
-0.115 |
|
(dummy) x treatment x year2004 |
(0.96) |
(2.32)** |
(1.50) |
(1.39) |
|
Richest in asset value in 1999 |
0.091 |
-0.399 |
0.105 |
-0.161 |
|
(dummy) x treatment x year2004 |
(0.45) |
(1.79)* |
(1.39) |
(1.93)* |
|
treatment |
-0.107 |
-0.003 |
-0.041 |
0.012 |
|
(0.82) |
(0.02) |
(0.78) |
(0.23) | |
|
year 2004 dummy |
0.163 |
0.492 |
0.062 |
0.106 |
|
(0.93) |
(2.55)** |
(1.04) |
(1.62) | |
|
year 2004 * Yangtze basin |
0.156 |
-0.205 |
0.052 |
-0.048 |
|
(1.55) |
(1.86)* |
(1.66)* |
(1.32) | |
|
household size |
0.109 |
0.097 |
0.011 |
-0.013 |
|
(5.34)*** |
(4.03)*** |
(1.77)* |
(2.03)** | |
|
total land holdings |
-0.000 |
-0.000 |
-0.000 |
-0.000 |
|
(0.02) |
(0.09) |
(0.39) |
(0.40) | |
|
household members with off-farm |
0.608 |
0.746 | ||
|
work in 1999 |
(14.36)*** |
(25.89)*** | ||
|
household members working on-farm |
0.687 |
0.730 | ||
|
in 1999 |
(20.84)*** |
(29.00)*** | ||
|
Constant |
-0.142 |
0.343 | ||
|
(0.93) |
(2.01)** | |||
|
Observations |
528 |
528 |
1,928 |
1,930 |
|
R-square |
0.41 |
0.56 | ||
Absolute value of t-statistics in parentheses in models (1) and (2); z-statistics in (3) and (4).
* significant at 10 percent level; ** at 5 percent level; *** at 1 percent level
Notes: In models (1) and (2), the dependent variables are the number of household members with (1)
off-farm work or (2) farm work. In models (3) and (4), the dependent variables are 1 = individual has (3)
off-farm work or (4) farm work and 0 = not. Columns (3) and (4) report the marginal effects of a probit
model and the standard errors are clustered at the household level.