Table 1

Robustness of Basic Sorting Patterns

				Sample				Firm Size
	Baseline (1)	ME Corrected (2)	Uniformly Distributed Skills (3)	Men 30–45 (4)	+ Women 30–35 (5)	+ Public-Sector Firms (6)	- Low-Skilled Men^{^a} (7)	≥50 Employees (8)	≥100 Employees (9)	Plants (10)
Panel A: Cognitive Skill
Between-firm variance
1986	0.173	0.197	0.173		0.141	0.166	0.143	0.151	0.139	0.209
1997	0.225	0.258	0.226	0.198	0.173	0.230	0.203	0.200	0.186	0.265
2008	0.229	0.264	0.232	0.228	0.178	0.233	0.206	0.191	0.172	0.264
Within-firm variance
1986	0.840	0.817	0.836		0.863	0.848	0.747	0.875	0.890	0.811
1997	0.747	0.714	0.758	0.781	0.799	0.751	0.697	0.787	0.803	0.722
2008	0.721	0.686	0.734	0.734	0.773	0.722	0.690	0.755	0.768	0.693
Panel B: Noncognitive Skill
Between-firm variance
1986	0.081	0.111	0.079		0.072	0.076	0.070	0.061	0.055	0.099
1997	0.109	0.156	0.114	0.097	0.094	0.111	0.096	0.088	0.080	0.135
2008	0.110	0.156	0.113	0.108	0.092	0.112	0.091	0.088	0.081	0.130
Within-firm variance
1986	0.872	0.841	0.887		0.894	0.879	0.829	0.889	0.895	0.849
1997	0.786	0.739	0.822	0.806	0.828	0.790	0.744	0.803	0.805	0.763
2008	0.801	0.755	0.825	0.794	0.842	0.801	0.762	0.814	0.816	0.782

Notes: The table shows the variance decompositions outlined in Section III of the paper. See Online Appendix Table C1 for the results from simulated variance components under random allocation of workers to firms (for the baseline case).
↵a. The share of the sample with low education falls dramatically over time. The cutoff, ten years or fewer of education, implies that 35 percent are dropped in 1986, compared to 14 percent in 1997 and 8 percent in 2008. This is due to the rapid expansion of upper-secondary education.