Hillary Clinton won Nevada’s caucus last night by about 5.5 percentage points. We don’t have official voting results by race, but polls taken before voters caucused (“entrance polls”) suggested the following demographic breakdown:
The Clinton campaign and several journalists have raised questions about whether Sanders actually won the Latino vote. They argue that Clinton “won the parts of Nevada that are most heavily Latino” and that there are also reasons to doubt the accuracy of previous polling of Latino voters in Nevada.
It’s certainly possible that the entrance polls are unreliable. However, if that’s true, the Clinton campaign and these journalists should not make that claim selectively. The polls actually matched the final results in Nevada incredibly well, as shown in the graph below.
So if Clinton did better with Latino voters than the entrance polls suggest, she did worse with other demographic groups. Suppose, for example, that Clinton spokesperson Brian Fallon’s speculation that she actually won something more like 61 percent of the Latino vote is correct (that result would be way outside the poll’s margin of error, but let’s not worry about that at the moment). If we assume that the entrance polls got the Black vote and Other Non-White vote totals correct, that would mean that Clinton actually lost the White vote by 9 percentage points, not 2. If we assume that the entrance polls got the White vote correct and the Black vote wrong, then Sanders lost the Black vote by 15 percentage points, not 54. And if we assume that the entrance polls got both the Black and White votes correct, then Clinton, instead of winning the Other Non-White vote by 2 percentage points, lost it by 55. It’s probably most likely, if the entrance polls are indeed wrong, that none of the above numbers can be trusted.
The Clinton campaign would probably prefer to believe that she won the Latino vote while losing the White vote by more than the polls indicate. Sanders supporters (myself included) would probably prefer to believe the entrance polls or, at the very least, to believe that he still did very well among Latino voters while doing better among Black voters than the entrance polls indicate. The truth is that we cannot know for sure.
What we do know for sure, however, is that the entrance polls cannot be selectively wrong. The Clinton campaign and journalists reporting on the results would do well to remember that.
Update (2/23/16): The William C. Velasquez Institute, a nonprofit that has been “conduct[ing] research aimed at improving the level of political and economic participation in Latino and other underrepresented communities” since 1985, issued a strong statement today on this issue. An excerpt:
Simply put there is no relevant statistical inconsistency between Edison’s Entry Poll results for Latinos, Whites, and Blacks and the overall election results. Based on this fact WCVI concludes that there is no statistical basis to question the Latino vote breakdown between Secretary Clinton and Senator Sanders.
We note that some analysts have said that Secretary Clinton’s victories in heavily Latino precincts proved that she won the Latino vote. However[,] the methodology of using heavily Latino or “barrio” precincts to represent Latino voting behavior has been considered ineffective and discarded for more than 30 years due to non-barrio residential patterns…common among Latino voters since the 1980’s.
The reason Sanders probably won the Latino vote despite losing heavily Latino precincts? He did well among young Latino voters who may not live in majority-Latino areas (thanks to Jonathan Cohn for the heads up on that).
7 responses to “Let’s Not Selectively Doubt Nevada’s Entrance Polls”
Or, more likely, there was some error in both white and African-American samples. If 61% of Latinos voted for Clinton, and half of the difference is allocated to white caucus goers and half to African-Americans, then Clinton won the African-American vote by about 2 to 1, and lost the white vote by 8 percent instead of 2. That sounds very plausible to me, especially given that the Latini sample size is only 195 voters. The MOE for that sub sample must be pretty large.
Or it’s possible that the entrance polls were off slightly and that Sanders still won the Hispanic vote. If you consider that polling shows Sanders getting brutally crushed with regards to African American voters across all age groups, even millenials, by wide double-digit margins, then a possible explanation is that the sample for African Americans was very accurate while the numbers for Caucasians and Hispanics were a bit off. If you consider that the Clinton campaign has been in Nevada since April, then maybe her courting of casino workers helped her to close the gap with white voters, keeping the loss in the mid-single digits. In that scenario, Sanders wins the Hispanic vote by a small margin, and Clinton winning more of the Hispanic-heavy counties can be attributed to his support among Hispanics being less concentrated in specific areas than Clinton’s.
This makes a hell of a lot more sense than him suddenly doing better among African Americans for no reason whatsoever.
Thank you. Seriously.
Reblogged this on David R. Taylor-Thoughts on Education.
The main problem with the Clinton campaign’s explanation is that their argument relies on the ecological fallacy. Just because heavily Latino areas went for Clinton doesn’t mean that the majority of Latinos went for Clinton. As the pollster pointed out, Sanders cleaned up with younger Latinos, who apparently tend to live in whiter/more diverse areas. Without knowing where all of the other Latinos in NV live and how they voted, we can’t make any inferences about overall Latino support solely based on the returns from any particular precincts.
Great points! Someone else pointed the age breakdown out, too, which led to my update :).