How sensitive are matching estimates of active labor market policy effects to typically unobserved confounders?

Journal for Labour Market Research

Table 4 Heterogeneity by type of ALMP

	Regular employment			Real monthly earnings
	After 36 months			After 36 months
	Standard specification	Extended specification	Difference in estimates	Standard specification	Extended specification	Difference in estimates
Short-term training	0.030	0.025	0.004	75\(^{*}\)	70	5
Short-term training	(0.024)	(0.025)	(0.008)	(42)	(43)	(15)
In-firm training	0.116\(^{***}\)	0.105\(^{***}\)	0.011	253\(^{***}\)	250\(^{***}\)	3
In-firm training	(0.047)	(0.049)	(0.027)	(86)	(89)	(49)
Long-term training	0.141\(^{***}\)	0.110\(^{***}\)	0.031	285\(^{***}\)	217\(^{***}\)	68
Long-term training	(0.052)	(0.052)	(0.030)	(105)	(108)	59.106
One-Euro-Jobs	− 0.041	− 0.034	− 0.007	− 34	− 19	− 15
One-Euro-Jobs	(0.025)	(0.024)	(0.012)	(43)	(43)	(20)
Other ALMPs	0.103\(^{***}\)	0.100\(^{***}\)	0.003	153\(^{***}\)	133\(^{**}\)	20
Other ALMPs	(0.034)	(0.034)	(0.014)	(58)	(56)	(25)
Control variables
Typically observed covariates	\(\checkmark\)	\(\checkmark\)		\(\checkmark\)	\(\checkmark\)
Typically unobserved covariates		\(\checkmark\)			\(\checkmark\)

This table shows estimated ATTs. Standard errors are obtained via bootstrapping with 999 replications. Statistical significance based on the normal approximation on the 10/5/1% level is indicated by \(^{*}/^{**}/^{***}\). The kernel matching with additional regression adjustment employed uses a standard bandwidth of 0.06 (Leuven and Sianesi 2003)