This misses an important point: Consumers had to spend money to see the movies they saw. They formed an opinion, whether the author considers it informed or not, on where to spend their money. So, if 1,000 paid to see movie A, and only 100 paid to see movie B, there is no reason to assume that if those who chose to spend their money on A were made to watch B as well, they would change their votes in large numbers ... regardless of the fun the author had with discrete choice models and data mining.