Table 2 Regressions of income by worker race before and after the rating-scale change

	Model 1 (before rating-scale change)				Model 2 (after rating-scale change, all workers)				Model 3 (workers who joined after the rating-scale change)
	b (s.e.)	t	P	95% CI	b (s.e.)	T	P	95% CI	b (s.e.)	t	P	95% CI
Non-white	−14.777 (1.502)	−9.84	<0.001	−17.72, −11.83	−13.373 (3.090)	−4.33	<0.001	−19.43, −7.32	11.434 (8.392)	1.36	0.173	−5.02, 27.89
Constant	158.867 (1.038)	153.06	<0.001	156.83, 160.90	172.838 (2.019)	85.62	<0.001	168.88, 176.79	172.734 (4.276)	40.39	<0.001	164.35, 181.12
Location FE	Y				Y				Y
Service category FE	Y				Y				Y
r²	0.040				0.031				0.044
n	68,684				18,404				3,489

The unit of analysis is a completed job for which an income rate was applied. Income is the normalized income a worker received from a completed job. Non-white takes a value of 1 if the worker was perceived to be a racial minority and 0 if the worker was perceived to be white. Model 1 focuses on jobs completed during the five-star scale; model 2 focuses on jobs completed after the rating-scale change to thumbs up/down; and model 3 focuses on jobs completed by workers who joined after the rating-scale change to thumbs up/down. Estimates are from ordinary least-squares regressions and robust standard errors are clustered at the customer level. All tests are two-sided.

Quick links

Search