Table 4. Final random effects generalized least squares linear regression model of direct abortion reporting on within respondent difference in list experiment treatment and control list response latencies within interviewers1.
| β | 95% CI | |
|---|---|---|
| Direct abortion question response (reference no) | ||
| Yes | 5.11** | 0.21,10.00 |
| Age (reference 15–19) | ||
| 20–29 | -1.31 | -3.32,0.71 |
| 30–39 | -2.03 | -4.72,0.67 |
| 40–49 | -3.89** | -7.04,-0.75 |
| Marital status (reference currently married/cohabiting) | ||
| Divorced/widowed | 0.87 | -2.91,4.65 |
| Never married | -0.61 | -3.44,2.21 |
| Schooling (reference never attended) | ||
| Primary school | -0.93 | -2.26,1.33 |
| Secondary school | -0.54 | -2.96,1.89 |
| Higher education | 0.33 | -2.10,2.76 |
| Wealth quintile (reference poorest) | ||
| Middle poorest | 0.69 | -1.39,2.77 |
| Middle | 0.5 | -1.68,2.67 |
| Middle wealthiest | 1.8 | -0.70,4.30 |
| Wealthiest | 1.39 | -1.58,4.35 |
| Caste (reference scheduled caste) | ||
| Scheduled tribe | -2.46** | -4.70,-0.22 |
| Other backward caste | -2.03 | -4.55,0.49 |
| Generate caste | 0.00 | -2.90,2.91 |
| Religion (reference Hindu) | ||
| Muslim | -1.01 | -3.67,1.65 |
| Other religion | -1.53 | -6.19,3.14 |
| Parity (reference 0) | ||
| 1–2 | 0.22 | -1.99,2.43 |
| 3–4 | 1.11 | -1.47,3.70 |
| 5+ | 3.19* | -0.19,6.58 |
| Interviewer acquainted (reference no) | ||
| Yes | -0.25 | -2.00,1.51 |
| Interviewer age (reference 15–19) | ||
| 20–29 | -2.99** | -5.84,-0.15 |
| 30–39 | -1.75 | -5.48,1.98 |
| 40–49 | -1.94 | -9.91,6.04 |
| Interviewer education (reference secondary or technical) | ||
| University | -1.67 | -3.75,0.40 |
| Masters or doctoral | 1.23 | -0.96,3.42 |
| Interviewer ever married (reference no) | ||
| Yes | 0.86 | -1.87,3.60 |
| Interviewer parity (reference 0) | ||
| 1–2 | 0.71 | -1.61,3.02 |
| 3–4 | 3.06 | -1.08,7.20 |
| Interviewer thinks abortion illegal (reference no) | ||
| Yes | 0.11 | -1.64,1.86 |
| Constant | -1.84 | -4.59,0.92 |
| Rho | 0.03 | |
| AIC | 47582.22 | |
| N | 5,298 | |
1Model applies survey weights that account for the complex sampling design and non-response, representing the inverse probability of selection. Used Taylor linearization method to calculate robust standard errors that account for clustering among respondents in the same EA.
* denotes p<0.10,
** denotes p<0.05, and *** denotes p<0.01.