Table 1 Regressions of highest rating by worker race and rating-scale change

From: Scale dichotomization reduces customer racial discrimination and income inequality

 

Model 1

Model 2

Model 3

 

b (s.e.)

t

P

95% CI

b (s.e.)

T

P

95% CI

b (s.e.)

t

P

95% CI

Non-white

−0.035 (0.003)

−10.18

<0.001

−0.04, −0.03

−0.018 (0.003)

−5.08

<0.001

−0.02, −0.01

−0.018 (0.004)

−4.02

<0.001

−0.03, −0.01

Rating-scale change

0.090 (0.003)

31.05

<0.001

0.08, 0.10

0.081 (0.003)

27.72

<0.001

0.08, 0.09

0.054 (0.005)

11.26

<0.001

0.04, 0.06

Non-white × rating-scale change

0.030 (0.005)

6.08

<0.001

0.02, 0.04

0.032 (0.005)

6.32

<0.001

0.02, 0.04

0.039 (0.008)

5.18

<0.001

0.02, 0.05

Experienced worker

    

−0.001 (0.003)

−0.41

0.685

−0.01, 0.004

−0.001 (0.004)

−0.17

0.863

−0.01, 0.01

Highly rated worker

    

0.055 (0.003)

18.32

<0.001

0.05, 0.06

0.054 (0.004)

13.78

<0.001

0.05, 0.06

Experienced customer

    

−0.023 (0.003)

−7.49

<0.001

−0.03, −0.02

    

Constant

0.869 (0.002)

417.16

<0.001

0.86, 0.87

0.839 (0.003)

243.35

<0.001

0.83, 0.85

0.828 (0.004)

219.39

<0.001

0.82, 0.84

Location FE

N

Y

N

Service category FE

N

Y

Y

Customer FE

N

N

Y

r2

0.017

0.034

0.371

n

69,971

69,971

51,387

  1. The unit of analysis is a submitted rating. Highest rating takes the value of 1 if the customer submitted a five-star or a thumbs-up rating, and 0 otherwise. Non-white takes the value of 1 if the worker was perceived to be a racial minority and 0 if the worker was perceived to be white. Rating-scale change takes the value of 1 for the period when customers rated workers using the thumbs up/down scale and 0 for the period when customers rated workers using the five-star scale. Estimates are from ordinary least-squares regressions (linear probability models) and robust standard errors are clustered at the customer level. All tests are two-sided.
Back to article page