In the 1940s, the sociologist Paul Lazarsfeld and his colleagues explored how the media affects political views by comparing people's opinions (as measured by surveys) to the news and advertisements they were exposed to. The investigators expected to find evidence that media messages had immediate, powerful, and intuitive effects on people's political views. Instead, they found that opinions were largely stable and invariant to media messages. You could face a barrage of the Madison Avenue pitches proclaiming the virtues of either President Franklin Roosevelt or his Republican challenger, but if six months in advance you were inclined to vote for one of those men, in November that was who you'd probably vote for. Very few people changed their preferences over the course of the campaign.

The same finding held throughout the broadcast era: There was very little relationship between people's intended choices and the messaging they encountered. Whatever change did occur usually took the form of people aligning their candidate preferences with their underlying party affiliation.
...

Perhaps my favorite example of this occurred when Herman Cain ran for the Republican presidential nomination. His successful tenure as CEO of Godfather's Pizza was a primary talking point on the campaign trail. Coincidentally, YouGov's BrandIndex already happened to be surveying the public about the restaurant's brand favorability. When Cain's campaign began, Republicans and Democrats viewed the pizza chain similarly, but as the country learned about Cain, opinions of Godfather's polarized: Democrats began to view the chain more negatively while Republicans did the opposite. By the height of Cain's popularity in late 2011, Republicans and Democrats differed by 25 points (on a scale ranging from -100 to 100) in their view of the company. It was the same pizza, but suddenly people's political loyalties played a major role in determining how much they liked it.