OLS and IV regressions of log labor force participation rate and hourly wages of 60- to 69-year-olds on log relative cohort sizes for 16- to 24-year-olds and 25- to 49-year-olds, 1984–2016
| . | LFP . | Hourly Wages . | ||
|---|---|---|---|---|
| OLS . | IV . | OLS . | IV . | |
| Both Sexes | ||||
| ln(cohort size aged 60–69 / cohort size aged 16–24) | –0.120* | –0.459* | 0.024 | –0.239† |
| [0.047] | [0.188] | [0.039] | [0.135] | |
| ln(cohort size aged 60–69 / cohort size aged 25–49) | 0.077 | 0.597** | –0.010 | –0.074 |
| [0.055] | [0.192] | [0.058] | [0.174] | |
| First-stage F statistic | ||||
| Dependent variable = ln(cohort size aged 60–69 / cohort size aged 16–24) | 10.51 | 10.51 | ||
| Dependent variable = ln(cohort size aged 60–69 / cohort size aged 25–49) | 9.72 | 9.72 | ||
| Durbin-Wu-Hausman test for endogeneity (p value) | .006 | .083 | ||
| Men | ||||
| ln(cohort size aged 60–69 / cohort size aged 16–24) | –0.073 | –0.480** | 0.034 | –0.299 |
| [0.047] | [0.163] | [0.043] | [0.183] | |
| ln(cohort size aged 60–69 / cohort size aged 25–49) | 0.031 | 0.464* | 0.007 | –0.066 |
| [0.054] | [0.187] | [0.061] | [0.195] | |
| First-stage F statistic | ||||
| Dependent variable = ln(cohort size aged 60–69 / cohort size aged 16–24) | 9.20 | 9.20 | ||
| Dependent variable = ln(cohort size aged 60–69 / cohort size aged 25–49) | 14.20 | 14.20 | ||
| Durbin-Wu-Hausman test for endogeneity (p value) | .013 | .101 | ||
| Women | ||||
| ln(cohort size aged 60–69 / cohort size aged 16–24) | –0.104* | –0.444 | –0.045 | –0.242 |
| [0.049] | [0.271] | [0.041] | [0.149] | |
| ln(cohort size aged 60–69 / cohort size aged 25–49) | 0.124* | 0.770** | 0.052 | –0.056 |
| [0.058] | [0.252] | [0.057] | [0.180] | |
| First-stage F statistic | ||||
| Dependent variable = ln(cohort size aged 60–69 / cohort size aged 16–24) | 8.43 | 8.43 | ||
| Dependent variable = ln(cohort size aged 60–69 / cohort size aged 25–49) | 4.85 | 4.85 | ||
| Durbin-Wu-Hausman test for endogeneity (p value) | .002 | .129 | ||
| . | LFP . | Hourly Wages . | ||
|---|---|---|---|---|
| OLS . | IV . | OLS . | IV . | |
| Both Sexes | ||||
| ln(cohort size aged 60–69 / cohort size aged 16–24) | –0.120* | –0.459* | 0.024 | –0.239† |
| [0.047] | [0.188] | [0.039] | [0.135] | |
| ln(cohort size aged 60–69 / cohort size aged 25–49) | 0.077 | 0.597** | –0.010 | –0.074 |
| [0.055] | [0.192] | [0.058] | [0.174] | |
| First-stage F statistic | ||||
| Dependent variable = ln(cohort size aged 60–69 / cohort size aged 16–24) | 10.51 | 10.51 | ||
| Dependent variable = ln(cohort size aged 60–69 / cohort size aged 25–49) | 9.72 | 9.72 | ||
| Durbin-Wu-Hausman test for endogeneity (p value) | .006 | .083 | ||
| Men | ||||
| ln(cohort size aged 60–69 / cohort size aged 16–24) | –0.073 | –0.480** | 0.034 | –0.299 |
| [0.047] | [0.163] | [0.043] | [0.183] | |
| ln(cohort size aged 60–69 / cohort size aged 25–49) | 0.031 | 0.464* | 0.007 | –0.066 |
| [0.054] | [0.187] | [0.061] | [0.195] | |
| First-stage F statistic | ||||
| Dependent variable = ln(cohort size aged 60–69 / cohort size aged 16–24) | 9.20 | 9.20 | ||
| Dependent variable = ln(cohort size aged 60–69 / cohort size aged 25–49) | 14.20 | 14.20 | ||
| Durbin-Wu-Hausman test for endogeneity (p value) | .013 | .101 | ||
| Women | ||||
| ln(cohort size aged 60–69 / cohort size aged 16–24) | –0.104* | –0.444 | –0.045 | –0.242 |
| [0.049] | [0.271] | [0.041] | [0.149] | |
| ln(cohort size aged 60–69 / cohort size aged 25–49) | 0.124* | 0.770** | 0.052 | –0.056 |
| [0.058] | [0.252] | [0.057] | [0.180] | |
| First-stage F statistic | ||||
| Dependent variable = ln(cohort size aged 60–69 / cohort size aged 16–24) | 8.43 | 8.43 | ||
| Dependent variable = ln(cohort size aged 60–69 / cohort size aged 25–49) | 4.85 | 4.85 | ||
| Durbin-Wu-Hausman test for endogeneity (p value) | .002 | .129 | ||
Notes: Data source is described in notes to the figures, and specification details are described in notes to Table 1. The table reports estimates of Eq. (2). Regression is weighted by average state population through the sample period. Standard errors are clustered by state. The two IVs used are the total number of births for 60- to 69-year-olds divided by the total number of births for 16- to 24-year-olds by state and year and the total number of births for 60- to 69-year-olds divided by the total number of births for 25- to 49-year-olds by state and year. We exclude certain years and states with missing birth data for the cohorts because only a handful of states started reporting births in 1915. N = 1,326.
†p < .10; *p < .05; **p < .01