Hi all,

I want to develop my model to assess the relationship between traffic accident and repair of traffic light.

In each regions (there are 100 regions), some of the traffic lights are broken. The number of lights and and broken lights varies across the region.

Thus, government have been repairing the lights based on the occurrence of traffic accidents and the their regular examination.

It means that if the accidents occurred, then the repair is likely to occur (but not 100%) and if they find the light is broken, then they repair it later soon.

In this kind of situation, I assume that accidents affect repair action so that I expect to see the adverse causal relationship between them.

But, "repair" also affects future accidents probability (not occur) in the region. Then, "repair" does not hold as a treatment.

In this case, what kind of model I can use to capture the true relationship between accidents and repair?

"repair" is continuous variable taking the number of repaired lights in region r in year t
"accident" is continuous variable taking the number of accidents in region r in year t
"examine" is continuous variable taking the the number examination in region r in year t

reg accident repair examine

seems too simple to me

Can you advise me on this?

Thanks in advance!