Women’s employment, income and divorce in West Germany: a causal approach

Journal for Labour Market Research

Table 4 Sensitivity analysis for unobserved heterogeneity (based on the logit model)

	Marginal employment		Regular employment
	e^γ	Significance level	e^γ	Significance level
Main sample
At day t₀	1.70	0.048	2.00	0.046
At day t₃₆₅	1.65	0.045	1.75	0.048
At day t₇₃₀	1.52	0.045	1.75	0.045
At day t₁₀₉₅	1.38	0.041	1.80	0.048
Group A—women not regularly employed in t₋₂₁₈₉ to t₋₇₃₀
At day t₀	1.38	0.048	1.85	0.048
At day t₃₆₅	1.60	0.045	2.35	0.049
At day t₇₃₀	1.68	0.045	2.63	0.046
At day t₁₀₉₅	1.60	0.048	2.50	0.048

T₀ is the day the divorce file was opened. Cumulative days of regular employment for group A in t₋₂₁₈₉ to t₋₇₃₀ equals zero. E^ɣ and 1/e^ɣ provide sharp bounds for the selection into treatment. The hypothetical selection bias (due to unobserved or unmeasured confounders) within the bounds, however, does not drive the confidence intervals of the treatment effects from Table 2 (column 3 and 4) into zero (based on 5% significance level)