Niche post: Doing a lit review, it's amazing how confused the literature seemed to be about identification in BLP (demand side) given how quickly it became the workhorse. As recently as this year (!) I've had smart people claim that some parameters of BLP aren't identified 1/

My impression is that this was maybe a concern 20 years ago, but it really shouldn't be at this point. We want to estimate demand functions wrt price. BLP-style demand is just one nice way to allow for nonlinearities in that function 2/

I'm contractually obligated to mention FRAC, which actually makes this crystal clear. If you approximate BLP this way, you see that the mysterious nonlinear parameters are just coefficients on a regressor that looks a lot like p^2 (plus some extra terms) 3/

So identification in the simplest BLP-style demand model is about as simple as a logit model w/ linear and quad terms in price. Nothing mysterious

I'm sure this post is too niche, but we need more IO method content on this website. Feel free to argue if I'm oversimplifying! 4/4