On the measurement of tasks: does expert data get it right?

Journal for Labour Market Research

Table 6 Task measures as wage predictors: survey vs expert data

Dependent variable: log hourly real wage	(1)	(2)	(3)	(4)	(5)	(6)
Abstract (Occ.)	0.94***			0.60***
Abstract (Occ.)	(0.06)			(0.06)
Routine (Occ.)	0.42***			0.24***
Routine (Occ.)	(0.06)			(0.07)
Abstract (Ind.)		0.53***		0.36***	0.37***	0.38***
Abstract (Ind.)		(0.03)		(0.04)	(0.03)	(0.03)
Routine (Ind.)		0.30***		0.19***	0.20***	0.20***
Routine (Ind.)		(0.03)		(0.04)	(0.03)	(0.04)
Abstract (Exp., Occ.)			0.46***		0.32***
Abstract (Exp., Occ.)			(0.03)		(0.03)
Routine (Exp., Occ.)			0.21***		0.14***
Routine (Exp., Occ.)			(0.03)		(0.03)
Survey tasks (Occupational)	\(\checkmark\)			\(\checkmark\)
Survey tasks (Individual)		\(\checkmark\)		\(\checkmark\)	\(\checkmark\)	\(\checkmark\)
Expert tasks (Occupational)			\(\checkmark\)		\(\checkmark\)
Occupation Dummies						\(\checkmark\)
F (Task Measures, Occ.)	150.79		192.44	52.42	82.90
F-pval (Task Measures, Occ.)	(0.00)		(0.00)	(0.00)	(0.00)
F (Task Measures, Ind.)		145.90		57.48	64.90	66.32
F-pval (Task Measures, Ind.)		(0.00)		(0.00)	(0.00)	(0.00)
Adj. \(R^{2}\)	0.19	0.19	0.19	0.20	0.20	0.22
AIC	49008.01	48969.22	48992.12	48790.82	48730.88	48219.27
BIC	49501.93	49463.13	49494.27	49301.21	49241.26	49832.73
Observations	27777	27777	27777	27777	27777	27777

Robust standard errors in parentheses
* \(p<0.10\), ** \(p<0.05\), *** \(p<0.01\)
This output is based on (10). The first two rows display coefficients based on occupational averages derived from individual responses in the employment surveys (“Occ.”). Point estimates corresponding to those individual responses are displayed in the third and fourth row (“Ind.”). Lastly, the last two rows display coefficients based on occupational averages derived from the Expert database (“Exp.”). All specifications include controls for gender, age, age squared, a dummy for living in an urban area, education dummies, occupational tenure, firm tenure, squared tenure for each dimension of experience, and a categorical variable reflecting firm size. Each regression is weighted by sample weights. The omitted task category is “Manual”, task returns are thus relative to performing manual tasks. I use BIBB/BAuA data that has been collected in 2011–12 and 2017–18, and data from BERUFENET, covering the years 2011–13. For the BIBB/BAuA data, see Hall et al. (2020b) and (Hall et al. (2020a), respectively. For the BERUFENET data, see Dengler et al. (2014)