Multinomial logit model of retirement path choice
| Base Category = Full Retirement | Model 1 | Model 2 | ||
|---|---|---|---|---|
| Partial Retirement | Unretirement | Partial Retirement | Unretirement | |
| Demographics & health (preretirement) | ||||
| Retirement age – 62 | 0.024 (0.031) | −0.068 (0.034) | −0.028 (0.025) | −0.127 (0.028) |
| I(retirement age ≥ 62) | −0.341 (0.201) | 0.153 (0.228) | −0.488 (0.194) | 0.087 (0.210) |
| Retirement age — 62 × I(retirement age ≥ 62) | −0.102 (0.100) | 0.020 (0.109) | 0.035 (0.100) | 0.073 (0.113) |
| Male | 0.414 (0.147) | 0.411 (0.168) | 0.468 (0.132) | 0.398 (0.151) |
| Black | −0.364 (0.216) | 0.462 (0.206) | 0.020 (0.176) | 0.284 (0.189) |
| Flispanic | −0.221 (0.303) | 0.080 (0.339) | 0.050 (0.228) | 0.080 (0.252) |
| Other | −0.485 (0.534) | −0.360 (0.500) | −0.230 (0.484) | −0.096 (0.417) |
| Married | 0.175 (0.179) | −0.097 (0.199) | 0.179 (0.154) | −0.108 (0.171) |
| I(education ≤ 12 years) | −0.257 (0.153) | 0.091 (0.175) | −0.116 (0.136) | 0.043 (0.155) |
| Fair or poor health (self-reported) | −0.352 (0.207) | −0.894 (0.241) | −0.494 (0.180) | −0.854 (0.203) |
| Number of health conditions | −0.099 (0.065) | −0.120 (0.077) | −0.077 (0.058) | −0.131 (0.064) |
| Retirement resources (preretirement) | ||||
| Log income | 0.081 (0.054) | 0.076 (0.059) | 0.099 (0.058) | 0.129 (0.063) |
| ASINH net worth | −0.059 (0.056) | −0.022 (0.061) | −0.038 (0.046) | −0.031 (0.049) |
| Self-employed | 0.544 (0.234) | 0.261 (0.279) | 0.604 (0.186) | 0.267 (0.230) |
| Employer pension | −0.611 (0.187) | −0.135 (0.199) | −0.450 (0.151) | −0.301 (0.165) |
| Employer offers retiree health insurance | −0.214 (0.199) | −0.120 (0.233) | −0.198 (0.176) | −0.323 (0.193) |
| Occupation (preretirement) | ||||
| Managerial/professional specialty | 0.630 (0.246) | 0.600 (0.278) | 0.709 (0.207) | 0.349 (0.232) |
| Sales/admin support | 0.556 (0.248) | 0.474 (0.277) | 0.574 (0.207) | 0.395 (0.227) |
| Services | 0.433 (0.274) | 0.454 (0.305) | 0.505 (0.221) | 0.054 (0.252) |
| Precision production/craft/repair | 0.560 (0.263) | 0.600 (0.287) | 0.405 (0.220) | 0.207 (0.240) |
| Retirement planning (preretirement) | ||||
| Short planning horizon | 0.032 (0.170) | 0.168 (0.175) | −0.016 (0.144) | 0.064 (0.157) |
| Plans to keep working in retirement | 0.875 (0.161) | 0.993 (0.188) | 0.894 (0.143) | 0.820 (0.161) |
| Worried about not having enough income | 0.332 (0.164) | 0.156 (0.204) | 0.277 (0.151) | 0.340 (0.179) |
| Worried about not being productive | 0.169 (0.169) | 0.351 (0.206) | −0.002 (0.150) | 0.514 (0.173) |
| Changes in resources & perceptions (postretirement) | ||||
| Net worth (nonhousing) drops by 25% or more | −0.144 (0.156) | 0.069 (0.168) | 0.024 (0.135) | 0.165 (0.143) |
| Stock value drops by 25% or more | 0.035 (0.177) | −0.001 (0.202) | 0.009 (0.162) | −0.008 (0.181) |
| Became more worried about income | 0.661 (0.298) | 0.108 (0.345) | 0.134 (0.232) | 0.162 (0.256) |
| Became less worried about income | −0.748 (0.238) | −0.118 (0.240) | −0.556 (0.204) | −0.300 (0.220) |
| Became more worried about not being productive | 0.075 (0.269) | 0.519 (0.259) | 0.242 (0.198) | 0.398 (0.211) |
| Became less worried about not being productive | −0.449 (0.271) | −0.223 (0.285) | −0.184 (0.234) | −0.261 (0.250) |
| Fiealth shock | −0.527 (0.161) | −0.778 (0.191) | −0.639 (0.138) | −0.698 (0.160) |
| OOP medical expenses jump by 25% or more | 0.051 (0.127) | −0.065 (0.143) | 0.056 (0.112) | −0.076 (0.126) |
| Lost health insurance | 0.043 (0.196) | 0.174 (0.207) | −0.150 (0.170) | 0.273 (0.171) |
| Pseudo R-Squared | 0.128 | 0.118 | ||
| Number of observations | 1,896 | 2,309 | ||
Notes: Sample is all individuals observed at least four years after first retirement. Model 1 uses hours/self-report definition of retirement and Model 2 uses hours-only definition. Multinomial logit coefficients are reported. Standard errors in parentheses and clustered at household level. Each model also includes an intercept, retirement calendar year dummies, and six dummy variables for missing values on variables in first model, and nine such dummies in second model (most not statistically significant). The omitted occupational category is Operators and Laborers. I(.) denotes the indicator function. ASINH is the inverse hyperbolic sine function. In neither model does a generalized Hausman test reject IIA.