Table S3.

Effect of yellow on the accident rate with drivers’ demographic characteristics as control variables using Aid=α0+βYellowid+α1Xid+εid

VariablesParameter estimatesRobust standard error
Yellow–0.000303***(0.000050)
Driving experience in 10 years–0.000226***(0.000046)
Age in 10 years0.000115***(0.000035)
Relief driver–0.001007***(0.000056)
Years of education0.000020(0.000013)
Driver’s residence fixed effectsYes
Observations2,862,669
R20.0002
  • The dependent variable Aid{0,1} is whether driver i had an accident on day d, and the independent variables include whether driver i’s taxi was yellow on day d, the driver’s driving experience and age, whether the driver was a relief driver on that day, and the driver’s residence fixed effect dummies (defined by the first two digits of the driver’s home zip code). The analysis is done at the daily level to account for drivers switching colors on any day of a month. The number of observations is based on the number of drivers with active contracts that day, among the 3,341 drivers. The per-day coefficients can be scaled by a factor of 365/12 to convert them to a per-month estimated level for easy comparison with the other tables. R2 is the coefficient of determination. ***P < 0.0001.