The title of the NPR article would be more accurate as "Cities See a Reduction in Ambulance Usage After Uber is Introduced".
The paper looks at 700 different cities, each with a different Uber market entry time over a two year period. The researchers also control for some seasonality effects.
Perhaps more covariates could have been considered, but this seems like a pretty decent design to measure a before/after causal effect.
Oh great. They let people do it in the tiny portion of the market that is normal desktop apps. They still keep a tight leash on the much larger mobile app market.
