If your data is getting better, then you make better predictions.

If your model is a good one you, able to detect more subtle and sophisticated variation in the data, then you may be able to make better predictions.

If your model more accurately forecasts past demand, from a better set of data, or better model formulation, then it will likely predict future demand better, that is how they are tested.

This is modeling on steroids, so yes some caution is warranted. But since this is what I was trained to do, and did in my second career, I find it very interesting. It is an interesting new approach, using a new tool which may or may not improve future predictions, by being able to incorporate more sophisticated and subtle changes in the variables affecting future demand.

I've used various multi-variable time series models for rail passenger demand forecasting, price elasticity, demographic and population changes, auto ownership, travel time, all go into sophisticated regression models, and try to make realistic forecasts of future demand. Regional travel demand models use similar methods. Good data is one key to often bad forecasting since it takes time and money to collect it. Then there are some atrocious models like many of the current regional transportation models which we know are bad, and why (poor data collection, poor model validation, and poor calibration) and yet are required to use, by law.

Or, you can have a system planned just like the CHSRA, with almost no use of any such means or methods, and no real analysis of the markets to be served, at what price points, for what demographic, serving what trip types..... And, we see just how well that is working for us, right?

If you are making decisions to spend $550 million of stockholders money on a new RR yard these are nice things to know....if you can.

To all of those reading this, not yet at an age where you are well beyond much hope of redemption, and are still in school, remember to take your calculus series, it will triple your expected lifetime earnings.