Employment trajectories in Germany: do firm characteristics and regional disparities matter?

Journal for Labour Market Research

Table 3 Cross-level effects on the highest education level and the local accumulation of human capital (three-level logistic random intercept models)

Independent variables	1999					2002
	Exit from job	Interfirm career path			Unemployment	Exit from job	Interfirm career path			Unemployment
	Exit from job	Upward mobility	Lateral mobility	Downward mobility	Unemployment	Exit from job	Upward mobility	Lateral mobility	Downward mobility	Unemployment
Cross-level-Effects: Highest education level × Local accumulation of human capital
No vocational training (1 = yes) × Share of students	0.006^∗∗∗	0.013^∗∗∗	0.009^∗∗	0.006	0.001	0.005^∗∗∗	0.007	0.013^∗∗∗	0.014^∗∗∗	−0.000
Secondary school and vocational training (1 = yes) × Share of students	−0.004^∗∗	−0.003	0.005	0.001	−0.002	0.001	0.010^∗∗∗	0.005	0.002	−0.002
Advanced secondary school and vocational training (1 = yes) × Share of students	0.007^∗∗∗	0.024^∗∗∗	0.013^∗∗	0.004	−0.007	−0.001	0.008	−0.003	0.009	−0.004
University degree (1 = yes) × Share of students	−0.004^∗∗	0.008^∗∗	0.009^∗∗	−0.015^∗∗∗	−0.011^∗∗	−0.006^∗∗∗	−0.006	−0.013^∗∗∗	−0.006	−0.000
Episodes (persons)	370779	370,779	370,779	370,779	370,779	363,339	363,339	363,339	363,339	363,339
Episodes (firms)	1836	1836	1836	1836	1836	2140	2140	2140	2140	2140
Episodes (regions)	97	97	97	97	97	97	97	97	97	97

Notes: The dependent variables are coded as dummy variables. In the columns 1 and 6 the value 1 represents a job exit; the value 0 denotes that the employee remains in the firm. In the columns 2–5 and 7–10 the value 1 indicates one of the destination states, respectively, whereas the value 0 subsumes all of the other employment states. In each regression the whole sample is used. Regression coefficients are reported after three-level logistic random intercept estimations. Models additionally contain all explanatory variables which are reported in Table 2.
^∗ p<0.05; ^∗∗ p<0.01; ^∗∗∗ p<0.001
Source: Linked Employer–Employee Data (LIAB); own calculations