Table 5.
Selected estimation results for substance use outcomesa
| Estimation method | Substance use problem severity at 12 months follow-up |
Abstinence at 12 months follow-up |
||
|---|---|---|---|---|
| OLS | Two-stage least-squares | Probit | Bivariate probit | |
| Treatment initiation | −0.435 | −2.399 | 0.486** | 0.987 |
| (0.247) | (1.688) | (0.155) | (0.788) | |
| [0.189] | [0.363] | |||
| Site 1b | −0.245 (0.304) | −0.237 (0.341) | −0.252 (0.253) | −0.212 (0.257) |
| Site 3b | −0.180 (0.264) | −0.007 (0.330) | −0.006 (0.181) | −0.004 (0.214) |
| Site 4b | −0.160 (0.270) | 0.080 (0.366) | −0.074 (0.180) | −0.070 (0.212) |
| Percentage with less than high school degree | −0.275 (1.077) | 0.340 (1.161) | 0.189 (0.866) | −0.059 (0.939) |
| Percentage unemployed | 4.137 (4.462) | 2.589 (4.368) | −1.279 (2.341) | −1.162 (2.505) |
| Median household income (1999) | −0.042 (0.053) | −0.041 (0.066) | 0.029 (0.047) | 0.024 (0.047) |
| Female | −0.137 (0.216) | −0.294 (0.231) | 0.190 (0.146) | 0.224 (0.149) |
| Age | −0.072 (0.075) | −0.070 (0.090) | −0.129* (0.054) | −0.147* (0.060) |
| African Americanc | −0.268 (0.221) | −0.001 (0.352) | 0.176 (0.216) | 0.159 (0.248) |
| Hispanicc | 0.094 (0.260) | 0.066 (0.298) | −0.019 (0.179) | −0.038 (0.195) |
| Other ethnicityc | 0.112 (0.304) | −0.131 (0.343) | 0.090 (0.196) | 0.191 (0.222) |
| Lived in residential program | −0.221 (0.403) | −0.374 (0.391) | −0.032 (0.282) | 0.031 (0.282) |
| Substance use problem severity at baseline | 0.027 (0.034) | 0.047 (0.037) | −0.035 (0.021) | −0.037 (0.022) |
| Parent graduated college | 0.035 (0.260) | −0.015 (0.285) | 0.205 (0.164) | 0.177 (0.160) |
| Familial alcohol/drug problems | 0.084 (0.186) | 0.153 (0.218) | 0.028 (0.138) | −0.008 (0.140) |
| Referral from legal system | 0.127 (0.222) | 0.206 (0.254) | −0.278 (0.153) | −0.298 (0.160) |
| Missing geocoded data | 1.077 (0.737) | 0.903 (0.956) | −0.318 (0.361) | −0.187 (0.398) |
| Missing parent’s education | 0.192 (0.416) | −0.799 (0.654) | – | – |
| Test of exogeneity of treatment initiation (p value)d | (0.230) | (0.549) | ||
| N | 383 | 363 | 384 | 364 |
*
Statistically significant at p<0.05
**
Statistically significant at p<0.01
a
Coefficients, robust standard errors (in parentheses), and marginal effects (in brackets) are reported. Instrumental variables include distance from residence to treatment facility and parental assessment of difficulty traveling to treatment (see Table 3).
b
Comparison is Site 2 (the largest site).
c
Comparison is whites.
d
Wald test employed to test H0: ρ=0 for bivariate probit models. p value of C statistic is reported for two-stage least-squares models. The null for this test is that treatment initiation is uncorrelated with the residuals in the outcome equation.