Yeah, I bet it isn’t a simple linear model, at least. But I also wonder if a model that takes the effects you’ve identified into account could be trained. I guess we’d have to have some historical source of the true GDP numbers, though.