Can someone please explain why this is an unscientific poll?
The fair attendees seem like a random distribution of Iowa voters.
It's not scientific as people self-select into casting a kernel. Fairgoers vs non-fairgoers is not random, and the fairgoers who cast a kernel compared to fairgoers who do not cast a kernel is not random.
Now, you could make the same argument about traditional polling; the difference is that pollsters try to weigh their sample to account for disparities between their sample and the population. They do this with varying degrees of success.