Table 7 Mediation effect regression and Bootstrap test results.

From: Will policy target groups choose to withdraw from rural housing land? A survey in Chengdu, China

Mediator variable

Variables

Regression equation

Bootstrap test (95% confidence interval)

Proportion of total effect that is mediated

S → T(a) (First stage model)

S → M(b) (Second stage model)

S → M → T(a’) (Third stage model)

S → M → T(c) (Third stage model)

Expected value

Farmland dependence

0.528** (0.215)

0.367** (0.187)

0.396* (0.214)

0.776*** (0.116)

(0.0014, 0.0309)

0.1800

Urban housing

−0.085* (0.046)

−0.132*** (0.043)

−0.085** (0.043)

0.776*** (0.116)

(−0.0077, −0.0010)

0.2030

Control variables

Yes

Yes

Yes

Yes

  

Observations

1515

1515

1515

1515

  
  1. Robust standard errors are in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1. a is the total effect of S on T, b is the effect of S on M, a’ is the direct effect of S on T after controlling for M, and c is the effect of M on T after controlling for S. S represents farmland dependence or urban housing, M is the mediator variable (expected value), and T represents villagers’ withdrawal choices. The first, second, and third stage models are Eq. (5), Eq. (6), and Eq. (7) in the model specification, respectively.
Back to article page