Working Paper: NBER ID: w31984
Authors: Honey Batra; Amanda Michaud; Simon Mongey
Abstract: We present six facts that characterize the little wage information contained in the universe of online job posts in the U.S. First, wage information is rare: only 14% of posts contain any wage information and the minority of these (6%) have a point wage. The majority (8%) feature a range of wages that are on average wide, spanning 28% of the midpoint (e.g. $21-28/hr or $32,000-$42,000/yr). Second, information varies systematically along the occupation-wage gradient. Third, posted wages are 40% higher than wages in BLS data in low-wage occupations and 20% lower than BLS data in high-wage occupations. Fourth, among the wages that are posted, high wage firms are more opaque, with more and wider ranges. Fifth, there is zero correlation between wage information and local labor market tightness. Sixth, of the top 20 posting private firms, none have any wage information in more than 2% of their posts. Our findings caution against treating wage data from job postings as a stand-in for administrative data. We provide an example of bias in econometric inference that worsens as wage information falls.
Keywords: No keywords provided
JEL Codes: E20; J30
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| lack of wage information (J31) | significant biases in econometric inference (C51) |
| observed wage data in job postings (J31) | not a reliable substitute for administrative wage data (J30) |
| systematic differences in wage information (J31) | misrepresent the actual wage landscape (J31) |
| lack of correlation between wage information and local labor market conditions (J31) | complicates interpretation of job posting data (J60) |