Table A1.
Additive Model Estimates for Direct Effects of Grandparents’ Education on Grandchildren’s Education based on Mixed-Effects Models with Random Intercepts (Regular) and Marginal Structural Models (MSM)
| Full Sample |
African Americans |
Whites |
||||
|---|---|---|---|---|---|---|
| Regular | MSM | Regular | MSM | Regular | MSM | |
| Grandparents, generation 1 | ||||||
| Years of schooling | 0.015 (0.013) |
0.056*** (0.012) |
0.020 (0.018) |
0.039* (0.018) |
0.008 (0.019) |
0.055*** (0.017) |
| Family structure (ref: two-parent) |
||||||
| One-parent, divorced | −0.210** (0.081) |
−0.337*** (0.090) |
−0.007 (0.109) |
−0.090 (0.123) |
−0.408** (0.118) |
−0.574*** (0.124) |
| One-parent, unmarried | 0.239† (0.140) |
0.077 (0.154) |
0.291† (0.162) |
0.186 (0.177) |
0.268 (0.276) |
0.022 (0.287) |
| Disability | −0.223** (0.075) |
- | −0.067 (0.099) |
- | −0.347*** (0.107) |
- |
| Occupational status (socioeconomic index) | 0.003 (0.002) |
- | 0.001 (0.005) |
- | 0.002 (0.003) |
- |
| Own home | 0.067 (0.084) |
- | 0.106 (0.106) |
- | 0.039 (0.133) |
- |
| Average family income during G2’s childhood | 4.57*10−6*** (1.27*10−6) |
- | 3.45*10−6 (2.98*10−6) |
- | 3.66*10−6** (1.40*10−6) |
- |
| Parents, generation 2 | ||||||
| Years of schooling | 0.256*** (0.017) |
0.304*** (0.018) |
0.198*** (0.025) |
0.239*** (0.027) |
0.296*** (0.024) |
0.361*** (0.024) |
| Family structure (ref: two-parent) |
||||||
| One-parent, divorced | −0.477*** (0.068) |
−0.501*** (0.074) |
−0.517*** (0.114) |
−0.657*** (0.111) |
−0.334*** (0.087) |
−0.344*** (0.100) |
| One-parent, unmarried | −0.498*** (0.079) |
−0.547*** (0.087) |
−0.450*** (0.103) |
−0.614*** (0.105) |
−0.500*** (0.155) |
−0.591*** (0.165) |
| Disability | −0.171** (0.057) |
- | −0.107 (0.083) |
- | −0.188* (0.080) |
- |
| Occupational status (socioeconomic index) | 0.009*** (0.002) |
- | 0.004 (0.004) |
- | 0.010*** (0.003) |
- |
| Own home | 0.288*** (0.079) |
- | 0.197 (0.100) |
- | 0.282* (0.143) |
- |
| Average family income during G2’s childhood | −3.46*10−7† (1.94*10−7) |
- | 7.05*10−6** (2.11*10−6) |
- | −4.06*10−7* (1.96*10−7) |
- |
| Grandchildren, generation 3 | ||||||
| African Americans | 0.019 (0.086) |
−0.222** (0.084) |
- | - | - | - |
| Female | 0.670*** (0.051) |
0.675*** (0.053) |
0.871*** (0.075) |
0.876*** (0.078) |
0.497*** (0.069) |
0.507*** (0.073) |
| Age group (ref: 25–34) | ||||||
| 35–44 | −0.088* (0.059) |
−0.089 (0.067) |
−0.045 (0.087) |
−0.056 (0.096) |
−0.115 (0.080) |
−0.117 (0.093) |
| 45–54 | 0.002 (0.111) |
0.018 (0.119) |
0.028 (0.143) |
0.004 (0.149) |
−0.051 (0.179) |
0.035* (0.186) |
| 55–65 | 0.442* (0.177) |
0.531** (0.191) |
0.459* (0.222) |
0.497* (0.238) |
0.387 (0.298) |
0.496* (0.248) |
| Religion (ref: Catholic) | ||||||
| Jewish | 0.255 (0.276) |
0.620* (0.291) |
−0.100 (0.848) |
0.136 (0.453) |
0.259 (0.295) |
0.545† (0.311) |
| Protestant | 0.010 (0.095) |
−0.068 (0.100) |
0.251 (0.199) |
0.144 (0.216) |
−0.030 (0.110) |
−0.110 (0.111) |
| Others | −0.116 (0.154) |
−0.228 (0.172) |
0.335 (0.288) |
0.217 (0.311) |
−0.190 (0.183) |
−0.306 (0.199) |
| Region (ref: Northeast) | ||||||
| North central | 0.124 (0.113) |
0.117 (0.125) |
−0.110 (0.221) |
−0.159 (0.254) |
0.228† (0.133) |
0.219 (0.136) |
| South | −0.086 (0.108) |
−0.104 (0.119) |
0.018 (0.200) |
−0.066 (0.225) |
−0.176 (0.131) |
−0.171 (0.140) |
| West | −0.135 (0.126) |
−0.140 (0.135) |
−0.112 (0.260) |
−0.219 (0.270) |
−0.151 (0.144) |
−0.124 (0.151) |
| Intercept | 8.726*** (0.274) |
8.645*** (0.300) |
8.845*** (0.455) |
9.118*** (0.531) |
8.519*** (0.372) |
8.005*** (0.363) |
| Number of family lineages | 1,485 | 1,485 | 586 | 586 | 899 | 899 |
| Number of observations | 5,357 | 5,357 | 2,525 | 2,525 | 2,832 | 2,832 |
Data sources: Multigenerational linked data from Panel Study of Income Dynamics, 1968–2013.
Notes: Figures in parentheses are robust standard errors from 20 imputed samples.
p < .1,
p < .05,
p < .01,
p < .001 (two-tailed tests).
The final marginal structural models do not include the covariates, including grandparents’ and parents’ disability status, occupational status, family income, and homeownership, because they are used to construct the inverse probability weights and do not bias our estimates of exposure variables, namely grandparents’ education, after the weighting.