Identifying Rates of Emigration in the United States Using Administrative Earnings Records

Table 4

Logit-hazard model regressions investigating the determinants of annual emigration, odds ratios.

Panel A. Emigration from the USA
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Sex (dummy, 1 = men)	1.040	1.092	1.036	1.086	1.084	1.070	1.065	1.065
Age	1.022	1.022	0.977	0.986	0.984	1.044	—	—
Age squared	—	—	1.001	1.000	1.000	1.000	—	1.001
Average earnings^a	0.983	0.982	0.984	0.982	0.983	0.987	0.987	0.987
Difference in growth rates
of GDP per capita^b	—	—	—	—	1.003	1.004	1.004	1.004
Constant	0.010	0.007	0.023	0.013	0.013	0.006	0.003	0.002
Country dummy variables	No	Yes	No	Yes	Yes	Yes	Yes	Yes
Year dummy variables^c	No	No	No	No	No	Yes	Yes	Yes
Age dummy variables	No	No	No	No	No	No	Yes	Yes
Number of observations	3,672,140	3,672,140	3,672,140	3,672,140	3,474,397	3,375,883	3,327,699	3,327,699
Panel B. Emigration from the US Social Security System
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Sex (dummy, 1 = men)	1.084	1.136	1.084	1.138	1.142	1.118	1.111	1.111
Age	0.999	0.999	1.006	1.014	1.013	1.094	—	—
Age squared	—	—	1.000	1.000	1.000	0.999	—	1.001
Average earnings^a	0.982	0.981	0.982	0.981	0.982	0.987	0.987	0.987
Difference in growth rates
of GDP per capita^b	—	—	—	—	1.002	1.002	1.002	1.002
Constant	0.018	0.013	0.016	0.010	0.011	0.003	0.003	0.003
Country dummy variables	No	Yes	No	Yes	Yes	Yes	Yes	Yes
Year dummy variables^c	No	No	No	No	No	Yes	Yes	Yes
Age dummy variables	No	No	No	No	No	No	Yes	Yes
Number of observations	3,672,140	3,671,838	3,672,140	3,671,838	3,474,397	3,375,883	3,327,699	3,327,699

Source. author's calculations.
Note. the dependent variable is the odds of the immigrant emigrating from the United States in any year.
^aDefined as average earnings over the past three years, in 2007 CPI-U-RS-adjusted dollars.
^bAnnual percentage change in US GDP per capita minus annual percentage change in GDP per capita in the country or region of origin.
^cDuration-interval-specific dummy variables for each observation.
All estimates are statistically significant (with heteroskedastically-consistent standard errors) at the 1-percent level, except for the coefficient on age in columns (1) and (2) of Panel B and the GDP variable in Panel B, all of which are statistically significant at the 10 percent level.