We used these data to simulate election/polling outcomes at the population level. For typical (choose-better) choices, we show that these outcomes will be least representative of the population when the majority has a clear preference (A>B) but likes both less than the minority.
