TABLE 7.
Effect of covariates on health care utilization and moral hazard of having a voluntary deductible: parametric polynomial specification
| Variable | (1) | (2) | (3) | (4) | (5) | (6) | (7) | (8) |
|---|---|---|---|---|---|---|---|---|
| Specialist visits | GP visits | Mental health care visits | Days in the hospital | |||||
| β 0 | (β 1–β 0) | β 0 | (β 1–β 0) | β 0 | (β 1–β 0) | β 0 | (β 1–β 0) | |
| Risk aversion | −0.009 (0.023) | −0.028 (0.144) | −0.012 (0.017) | 0.039 (0.083) | 0.090 (0.057) | −0.221 (0.218) | 0.099 * (0.059) | −0.519 (0.444) |
| Male | 0.003 (0.053) | −0.383 (0.349) | −0.159 *** (0.049) | 0.155 (0.251) | 0.147 (0.132) | −0.973 * (0.549) | 0.218 * (0.129) | −1.404 (0.870) |
| Age | −0.006 (0.011) | −0.018 (0.086) | 0.000 (0.008) | −0.008 (0.044) | 0.002 (0.020) | 0.144 (0.115) | −0.030 (0.020) | 0.239 (0.167) |
| Age2 | 0.000 (0.000) | 0.000 (0.001) | 0.000 (0.000) | 0.000 (0.000) | −0.000 (0.000) | −0.001 (0.001) | 0.000 ** (0.000) | −0.003 (0.002) |
| Employed | −0.037 (0.098) | 0.033 (0.578) | −0.002 (0.069) | −0.271 (0.404) | 0.273 (0.189) | −1.657 * (0.856) | −0.528 ** (0.227) | 1.103 (1.928) |
| Educ.mid | −0.002 (0.091) | −0.839 (0.637) | −0.009 (0.049) | −0.289 (0.353) | −0.345 ** (0.153) | 1.506 ** (0.737) | −0.082 (0.136) | −0.542 (1.089) |
| Educ.high | −0.020 (0.107) | −0.244 (0.721) | 0.054 (0.086) | −0.532 (0.445) | −0.043 (0.179) | 1.021 (0.850) | −0.481 *** (0.173) | 1.454 (1.173) |
| Married | 0.037 (0.080) | −0.356 (0.498) | −0.085 (0.056) | 0.373 (0.302) | −0.264 (0.199) | 1.603 * (0.862) | −0.234 (0.170) | 0.701 (1.068) |
| Good health | −0.188 ** (0.079) | 0.330 (0.436) | −0.174 *** (0.065) | 0.267 (0.273) | −0.727 *** (0.209) | 2.707 *** (0.864) | −0.115 (0.171) | 0.279 (0.986) |
| MHI5 | −0.009 *** (0.002) | 0.030 * (0.016) | −0.005 *** (0.001) | −0.014 * (0.008) | −0.017 *** (0.004) | 0.046 ** (0.020) | −0.002 (0.005) | −0.003 (0.031) |
| Chronic cond. | 0.292 *** (0.081) | −0.633 (0.642) | 0.240 *** (0.071) | −0.581 (0.401) | −0.152 (0.174) | 0.493 (0.861) | 0.424 *** (0.165) | −1.028 (1.322) |
| Smokes | −0.149 * (0.080) | 1.258 ** (0.498) | −0.092 (0.062) | 0.328 (0.320) | −0.006 (0.196) | 0.511 (0.736) | 0.176 (0.110) | −1.121 (0.793) |
| Log income | −0.108 (0.069) | 0.280 (0.409) | −0.108 ** (0.049) | 0.215 (0.243) | −0.084 (0.201) | −0.267 (0.821) | 0.200 (0.188) | −0.970 (1.323) |
| p1 | −3.270 (5.505) | −3.672 (3.302) | −4.966 (8.993) | 15.872 (16.275) | ||||
| p2 | −1.659 (14.487) | 8.507 (10.019) | 1.911 (11.305) | −32.470 (37.379) | ||||
| p3 | −4.607 (18.431) | −8.703 (12.295) | −2.450 (10.606) | 22.289 (49.657) | ||||
| Joint test p1 p2 p3, p value | 0.4001 | 0.7363 | 0.8568 | 0.4712 | ||||
| Joint test (β 1− β 0 ), p value | 0.1116 | 0.1975 | 0.0000 | 0.6611 | ||||
| E(Y1‐Y0)@X | −6.649 (8.063) | −3.343 (4.855) | −1.445 (3.409) | 0.075 (21.823) | ||||
| Constant | 2.532 *** (0.714) | 2.353 *** (0.473) | 3.340 ** (1.703) | −1.136 (2.040) | ||||
| Observations | 5,712 | 5,712 | 9,783 | 9,783 | 856 | 856 | 1,506 | 1,506 |
Note: The generalized Roy model is fit in two stages: a first‐stage probit regression to obtain the propensity score p, followed by a linear regression of log‐health care utilization measure on a constant, covariates, and their interactions with p, and the polynomial terms of p (p1, p2, and p3). Bootstrapped standard errors clustered at the household level are presented in parentheses. The specifications include wave dummies.
***
p < 0.01.
**
p < 0.05.
*
p < 0.1.