My Myth of the Rational Voter argues that elections are surveys.  The essence of a survey is that you state an opinion, secure in the knowledge that your stated opinion is non-binding.  While there remains an off-chance your vote changes political outcomes, there’s also a off-chance your survey response changes political outcomes!  Indeed, a response to a nationally representative survey is probably more likely to sway policy, because a survey respondent is one voice out of thousands instead of one voice out of tens of millions. 

For a comic illustration of this insight, see this fascinating story about registration for the American Independent Party (AIP) in California.  From the LA Times:

With nearly half a million registered members, the American
Independent Party is bigger than all of California’s other minor parties
combined. The ultraconservative party’s platform opposes abortion
rights and same sex marriage, and calls for building a fence along the
entire United States border…

But a Times investigation has found that a majority of its
members have registered with the party in error. Nearly three in four
people did not realize they had joined the party…

What went wrong?  Voters treated their party registration forms about as seriously as any other survey:

Voters from all walks of life were confused by the use of the word
“independent” in the party’s name, according to The Times analysis.

Residents of rural and urban communities, students and
business owners and top Hollywood celebrities with known Democratic
leanings — including Sugar Ray Leonard, Demi Moore and Emma Stone — were
among those who believed they were declaring that they preferred no
party affiliation when they checked the box for the American Independent
Party.

[…]

Of the 500 AIP voters surveyed by a bipartisan team of pollsters, fewer
than 4% could correctly identify their own registration as a member of
the American Independent Party.

How is this different from product confusion on, say, Amazon?  Two big ways.  First, customers have an incentive to check their work, because ordering the wrong product is selfishly costly.  Second, customers can easily check their work, because they directly experience their purchase once it arrives in the mail.  When voters face the choice to go AIP, in contrast, they have no selfish incentive to review their order.  And since each voter is just a grain of sand on the beach of politics, they can overlook their error indefinitely – or at least until the LA Times comes calling.

P.S. The AIP registration expose also neatly illustrates the principle that rare survey answers tend to be even rarer than they look.  If a survey found 2% of Americans favored abolishing the minimum wage, for example, we should suspect that many of those who said “abolish” misunderstood the question or misspoke.