RTA Ridership Model Update

Last month I built a statistical ridership model for Cleveland's RTA system. One major concession was that I excluded a service-level independent variable because I lacked sufficient data to include it in the model. Unsure how to resolve this issue, I recently received this email from someone at the Maryland Transit Administration:
The service level data not in your model is actually a huge factor driving ridership-if the system isn’t seen as robust due to meager service offerings, fare prices and population don’t matter. I think if you add service levels to your model, you’ll see that it adds a good chunk of explanatory power, and that it may even be collinear with population. As population departs, tax base declines, service has to be cut, and so on. For the share of transit system users that are transit-dependent, poor service means they will find other ways-ride sharing, moving, etc.-to get to where they need to go. Transit use is always optional, and the elasticity is different for different rider segments, but less service will always reduce ridership.
This explanation is actually more intuitive than one that high fares primarily drive ridership. And here is the worst news: next April, RTA will cut service more significantly than in recent memory.

So continues the death spiral of transit service in Cleveland...