Our unit of randomization is the police station: treatment stations mandated to start WHDs will be compared to a control group with no WHDs. All treatment stations receive the first three components: creation of a desk, training, and community outreach. We have randomized the fourth component (allocation of additional female personnel) as a separate treatment arm. This will allow us to gain leverage on the question of whether or not the presence of additional female officers has an independent effect, above and beyond the effect of the WHD.
The Madhya Pradesh Police selected ten districts - with nearly 300 police stations - across the state to be representative of geography, demographics, and socio-economic conditions. After removing specialized police stations (headquarters, cybercells, etc.), we have 180 police stations in our sample. Within each district, we stratified police stations by geography (urban and rural) and by the first principal component of a vector of police station characteristics (number of assigned officers; number of registered cases; population served). We randomized our two treatment and control arms within these strata, with 61 police stations in the first treatment arm (WHDs with first three components), 59 in the second (WHDs with first three components + female police officers assigned), and 60 in the control arm.