[KGZotU]

Right, there are several people arguing here against the entire project of naive modeling. The author's response is that you can criticise, but only in the form of providing a competing naive model.

From the post:

> If you disagree with these numbers or this distribution, there is no need for you to stop reading. Just download the code and change the numbers. Your change in these assumptions might completely invalidate my conclusions, or they might not matter at all. The best way to find out is to actually run the model.

Varying parameters and producing different predictions would not invalidate the model. What validates or invalidates a model is whether or not it actually predicts the behavior you're trying to model. Without a validity test, modeling is pure sophistry.

So instead of suggesting another naive model, let me suggest a way to augment the whole project: construct a model that successfully predicts actual, ongoing, real world behavior. Test your model against actual, real world policy changes. If it has verified predictive power, then you have some base to claim that it can predict novel policy changes.

Any given person's opinion may condense decades of personal experience. I disagree with the author's claim that (even naive) mathematical modeling is better than personal opinion. Given any rigorous definition of better, that is frankly an empirical claim that the author has left unsupported.