Suppose you make a profit of P0 in a market with no medallion system, a profit of P1 in a market with a medallion system when you have no medallion, and a profit of P2 in a market with a medallion system when you have a medallion. Then that sale suggests that P2-P1 is at least $600k, but that doesn't tell us anything about P2-P0.

Probably P1 is approximately 0, at least for someone who's already got a substantial legal taxi business, since it's illegal to operate a taxi with no medallion. So we know that P2 is at least $600k. But we know nothing at all about P0; it might be just the same, or much less, or much more.

And, actually, we don't even really know that P2-P1 is at least $600k. The person who paid that much might be an idiot (unlikely) or a speculator who thinks others will foolishly be prepared to pay even more (not so unlikely); or he may think that he'll keep the medallion for a few years and then sell it for approximately the same amount. In that case, our lower bound on P2-P1 is something like 5% of $600k per year (taking 5% as a ballpark interest rate).

So: sorry, but I'm totally unconvinced. Am I missing something?