So what's going on here? Are the polls simply wrong? Is there some margin of error something or other?
The short answer is that we don't know. Over at the New York Times's Upshot blog, Nate Cohn has a nice exegesis of the poll results that determines that Clinton probably won more of the Hispanic vote based on what voting patterns looked like in heavily Hispanic areas and based on some of the glitchiness of the entrance-poll system. Cohn knows what he's talking about, and his points are fair.
But we can add some more context which, unfortunately, probably throws more shadow over things than anything.
It's important to note, first of all, that polling shortly before the caucuses began showed a surprisingly close race among nonwhite voters. In her write-up of CNN-ORC's poll from last week, CNN polling director Jennifer Agiesta wrote that, "although the pool of potential caucusgoers in Nevada is more racially diverse than those who participated in Iowa or New Hampshire, the racial divide among likely caucusgoers isn't nearly as stark as among voters in South Carolina, with both white and non-white voters about evenly divided between the two candidates."
CNN's numbers didn't feature large enough samples to separate blacks and Hispanics individually. But according to the entrance polls, about 13 percent of voters were black and 19 percent were Hispanic. If blacks heavily favored Clinton and Hispanics were a bigger part of the nonwhite population, that would suggest that they slightly favored Sanders.
If you match precinct results (as shared by the Nevada Democratic Party) with the Hispanic population in the populous area around Las Vegas, you can see that there isn't a 1 to 1 correlation between strong support for Clinton (dark green being her winning 100 percent of delegates) and heavy Hispanic populations (dark orange). It's more mixed than that; some areas have strong support for Sanders (dark pink being his winning all delegates).
The caucus process, winnowing hundreds of votes down to a handful of delegates, introduces a level of uncertainty. But there doesn't seem to be a link between a heavy Hispanic population and heavy Clinton support.
Entrance polls are indeed imprecise tools. The company that conducts them, Edison Media Research, identifies precincts to question voters about their choices, and it weights the results of those interviews against the final results. Cohn notes that they're imprecise — but so does Joe Lenski, executive vice president of Edison Research, as he did when we reached him by email over the weekend.
First of all, he noted that the margin of error was plus-or-minus seven percentage points, which is about the margin by which Sanders was preferred by that group in its final weighting.
Lenski also noted that the same dramatic split in age that we've seen in Democratic voting to date also played out in the Hispanic vote. Younger Hispanic voters were much more likely to back Sanders than older voters — a split we've seen among white voters and, as the Wall Street Journal reports, among black voters, too. Age is a much better predictor of how someone will vote than race or ethnicity. Among the younger Hispanic voters, Lenski said, Sanders won about 4 out of 5 votes. Among older Hispanic voters, Clinton won two-thirds — with hefty margins of error on both.
As a general rule, younger voters turn out less heavily than older ones do — and younger Hispanics are among those the least likely to vote in the country, according to data from the Census Bureau.
On the other hand, people 35 and younger are a larger percentage of the Hispanic vote than that age group is of any other racial or ethnic group, as data from Pew Research showed in January.
To cut to the chase, then: Entrance polls are often imprecise, and in this case there was a large margin of error. But while younger Hispanics tend not to vote often, more Hispanics (at least nationally) are younger — and that group heavily favored Bernie Sanders. Older Hispanics, meanwhile, preferred Hillary Clinton -- though apparently not as broadly as did black voters.
Future polls will indicate how much the overall Hispanic vote prefers Clinton or Sanders. The lesson from Saturday, if nothing else, appears to be that it's the black vote that will act as Clinton's reliable firewall, if nothing else. And in Nevada, that was enough.
Scott Clement contributed to this report.