Hillary Clinton won the Nevada Democratic caucuses on Saturday. She did so thanks largely to her strength in Clark County, the home of Las Vegas and the most heavily Latino part of the state.
That's important to note, because entrance polls showed Bernie Sanders winning among Latino voters — by a shockingly wide margin.
An early wave of entrance polling showed Sanders beating Clinton by 11 points with Latinos. A later wave showed him winning by 8.
What really happened? We might not ever know for sure. But if you look at the possible scenarios — and at the entrance and exit polls' record with Latinos — the most plausible conclusion is that the entrance polls didn't correctly predict Nevada's Latino vote.
Clinton won the most heavily Latino parts of the state
Official voting results don't break down votes by race, so we can't match up the real outcome in Nevada up against the entrance polls. (They're entrance polls because in caucus states, voters are polled on their way into the caucus, rather than on their way out of the voting booth.)
But what we do know from the official voting results — broken down by caucus site and by region — is that that Clinton won the parts of Nevada that are most heavily Latino.
The most heavily Latino county in the state — Clark County — was Clinton's stronghold. With two-thirds of its precincts reporting, Clinton had a 10-point margin over Sanders — much wider than either candidate's margin of victory elsewhere in the state.
The New York Times analysis of caucus results, with 74 percent of precincts reporting, showed Clinton winning "More Hispanic" precincts in the state (if narrowly), while Sanders (even more narrowly) was winning "Less Hispanic" precincts.
And Clinton swept the at-large "casino caucuses" in Las Vegas — attended by the heavily Latino workforce on the strip.
The only explanation for the entrance polls would be that Clinton consistently won the parts of Nevada where the most Latinos happen to be — by overwhelmingly winning the non-Latino vote there, while Sanders won the Latino vote.
That is extremely unlikely. It is more likely that Hillary Clinton won the most Latino parts of Nevada because Hillary Clinton won Nevada's Latinos.
Entrance and exit polls have a bad track record in predicting Latino voter behavior
Here's the thing you need to understand about entrance and exit polls: They are designed to help media outlets figure out, as quickly as possible, who is likely to win the state. They are not designed to help them figure out who won specific constituencies.
So when entrance pollsters picked caucus sites to put surveyors at, says political science professor David Damore of the University of Nevada Las Vegas, they picked sites that would be representative of the state — not necessarily sites that would be representative of the Latino vote. The Latinos who attended those sites might have been representative of Latinos in the rest of the state, or they might not have been.
This is far from the first time that entrance or exit polls have gotten the Latino vote wrong. In 2012, the exit polls in the general election — which is theoretically easier to poll than a caucus — overstated the percentage of Latinos who voted for Mitt Romney.
And in 2010, in Nevada's tightly contested Senate race — where Latinos were widely credited with saving Harry Reid's reelection against Republican challenger Sharron Angle — the official exit poll said that 30 percent of Latinos had voted for Angle. That would have meant that substantially more Latinos had voted for Angle, who ran her campaign on fear of unauthorized immigrants, than had voted for the pro–immigration reform John McCain in 2008.
An exit poll exclusively of Latino voters conducted by the polling firm Latino Decisions, meanwhile, found that Reid had won an overwhelming 90 percent of the Latino vote — and Angle had won only 8 percent.
Unfortunately, there was no Latino-specific entrance poll conducted in the 2016 caucuses. But if there had been, it's reasonable to think that it would have showed Hillary Clinton winning the Latino vote.
This isn't an indictment of the entrance polls. Again, they're not designed to get the Latino vote right. But it should be a caution for anyone who's tempted to use the entrance polls to claim that Sanders got a moral victory in Nevada by winning 54 percent of Latino voters. That might have happened, but it probably did not.