Table 2.

Average treatment effect on the treated (ATT) of sobriety checkpoints on rates of traffic fatalities in Mexican municipalities

Model	ATT (95% CI)
Main model (n = 106)	− 12.3% (− 17.8;− 6.5)
Sensitivity 1: excluding Mexico City (n = 92)	− 8.5% (− 13.3;− 3.4)
Sensitivity 2: excluding municipalities where the program was discontinued (n = 89)	− 11.9% (− 18.1;− 5.2)
Sensitivity 3.1: using metropolitan areas and first municipality to adopt (n = 72)	− 22.9% (− 35.1;− 8.6)
Sensitivity 3.2: using metropolitan areas and core municipality (n = 72)	− 18.4% (− 27.2;− 8.8)
Sensitivity 4: negative control (cancer deaths as outcome) (n = 106)	1.2% (− 2.9;5.6)

Global effect estimates were exponentiated and calculated as a percent change, calculated as (exp(B)-1)*100. Models were adjusted for pre-treatment covariates combined with inverse probability weighting (IPW). In models using municipality-level data (i.e., main model and sensitivity models 1, 2, and 4), municipalities were clustered within metropolitan areas using clustered robust standard errors