As the weeks between now and Election Day slip between our fingers, Hillary Clinton’s nearly evaporated lead from mid-August is giving many Democrats and progressives unassuagable anxiety over what will happen on Nov. 8. But the dramatic shift in the polls — from Clinton being ahead of rival Donald Trump by roughly eight points in the weeks following the Democratic National Convention, according to the Real Clear Politics average, to the current one showing the two candidates within a point (or less) of each other — don’t explain the underlying phenomenon of what has changed. Understanding that, and how it interacts with the Bradley Effect, may offer some relief to Clinton supporters (and/or Trump opponents).
The Bradley Effect is generally applied when race is a factor in an election. As Ballotpedia explains, the Bradley Effect is when voters "tell pollsters that they are undecided or likely to vote for a minority candidate but will vote against the minority candidate on Election Day." Namely, white voters will tell pollsters they're voting for or are undecided about a minority candidate out of concern about seeming racist for voting against them. That racial concern isn't at play this presidential election, but the idea that voters may lie to pollsters about their support for Trump is.
Earlier in the election, pundits posited that such a twist on the Bradley Effect might help Trump, since voters were less likely to openly disclose their support of the reality TV star to another person (i.e. a pollster). That theory was echoed by Morning Consult, a polling firm which compared Trump responses in online polls to those in automated telephone polls and live phone interview polls. Shockingly (not), "Republicans are more likely to say they want Donald Trump in the White House if they are taking a poll online versus in a live telephone interview."
Trump campaign manager Kellyanne Conway revived this idea in late August, saying that there were a number of “undercover Trump voter[s],” who were less likely to support Trump in a poll, but would nevertheless vote for him. This “reverse Bradley Effect” may have enjoyed some popularity, but others argue there’s little evidence of this happening.
That doesn’t mean there isn’t some other manifestation of a Bradley Effect in this election. The current narrowing of the race has less to do with Trump winning over Clinton supporters and more to do with Clinton losing the support of young voters, many of whom are supporting third-party candidates Gary Johnson and Jill Stein instead. A lot of this is driven by the unpopularity of Clinton among millennials, who is second in unpopularity only to Trump himself.
But these one-time Clinton supporters have grown up in the era of incessant political polling, and I’d argue that they are more aware of the importance (or lack of importance) of polls than many pundits would give them credit for.
Consider a hypothetical Millennial voter who supports Libertarian candidate Gary Johnson, but also opposes Trump. If a pollster asked who they’d support, at this point in the race, that voter has every incentive to say Johnson, especially when third-party candidate participation in the presidential debates hinges on polls.
If the polls continue to remain tight (especially in certain swing states), I’d put good money on these voters making a tactical decision to vote for Clinton on Election Day. Granted, this is based more on my theorizing than hard evidence. However, my gut (and anecdotal experience) suggests that when push comes to shove, even many die-hard third-party supporters will ultimately cast their ballots for Clinton to block Trump — even if they're not telling that to pollsters.