I guess you could reconstruct my assumption(s) in terms of DGP, but I think it’s more clear to do the following reconstruction, with two assumptions:

- Movie ratings express movie quality.
- Most people think/experience most movies as being of an average quality.

As it is with all assumptions, you don’t have to justify them. That’s why they are called “assumptions” in the first place.

If you could justify them, you wouldn’t have to **assume** them anymore.

Now, building upon these two assumptions, I have argued that movie ratings should have a normal distribution.

So I have justified why movie ratings should have a normal distribution only in the context of assuming (1) and (2) as true.

This is one way to do it, as you indirectly remarked. If you want to reach a conclusion where the distribution it’s left skewed, you’ll have to assume that:

(i) Movie ratings express movie quality.

(ii) Most people think/experience most movies as being of a bad quality.

I did considered different assumptions on which I could build my whole reasoning. And the one that seemed the safest was the one I chose to work with.

Now let’s say you build your reasoning departing from (i) and (ii), and reach a different conclusion than mine (say you come to recommend the tomatometer).

What we have now, it’s basically two rival theoretical models. We should try to determine which is better. One way to do it is to find a way to derive predictions from them, and see which one works better.

Another way is to try to build different argumentations and see if we reach the same conclusion, like I did with taking the Fandango variable as a negative reference.

P.S. I explained why (2) cannot be proven true or false in this response and this response.

And thanks for taking the time to let me know what you think!