Washington – In a new Arizona Republic piece, “Why exit polls are wrong about Latino voters in Arizona,” Stephen A. Nuño and Bryan Wilcox-Archuleta refute exit poll claims that 31% of Arizona Latinos cast votes in favor of Donald Trump. Relying on analysis of precincts – instead of counties – Nuño and Wilcox-Archuleta find that over 80% of Latinos in Arizona voted for Clinton, a number much closer to the Latino Decisions estimate of 84% to 12% in favor of Clinton.
To read the entire analysis by Nuño and Wilcox-Archuleta, click here, or see below:
“In 2012 Democrats lost the presidential election in Arizona by 9 points, and in 2016 they lost by just four.
As data from across the state shows, the Grand Canyon State may be headed for the prized battleground status in 2020 thanks to an overwhelming Democratic vote by Latinos. However, there has been some debate over the true Latino presidential vote, both nationally and in Arizona.
National exit polls claimed that 29 percent of Latinos voted for Trump. But the polling firm Latino Decisions conducted an election-eve poll that found Trump only garnered 18 percent of the Latino vote. In Arizona, exit polls reported 61 percent of Latinos voted for Clinton and 31 percent for Trump, while Latino Decisions reported a much larger Democratic advantage of 84 to 12 percent in favor of Clinton in Arizona.
Which number is correct?  
A few other analysts have recently looked at county voting patterns, but with 2.6 million voters and just 15 counties in Arizona, county-level data is inadequate and results in what sociologist William Robinson called the “ecological fallacy.”
Smaller units such as precincts are greatly preferable for accurately inferring voting patterns. Thus, we gathered data for 1,288 of 1,469 precincts across the state and found it’s statistically unlikely that 61 percent of Latinos voted for Clinton. Instead, we found strong evidence that over 80 percent of Latinos in Arizona voted for Clinton, a number much closer to the Latino Decisions estimate.
Part of the reason exit poll findings differ from Latino Decisions’ findings is because of the different methods used by the pollsters. While exit polls are useful to assess views of the whole electorate, the survey’s methods are not designed to accurately assess Latino voters in a state.
By their own admission, the pollsters behind exit polls admit that their approach is “not designed to yield very reliable estimates of the characteristics of small, geographically clustered demographic groups,” such as Latinos. This is largely because exit poll survey results come from a sample of a handful of precincts across the state that are not specifically selected to accurately capture Latinos.
As a result, past exit polls have over-represented the views of Latino voters with higher income and education levels than most of the Latino voting community and have failed to include enough Spanish-speaking Latino voters.
By contrast, the Latino Decisions’ election-eve poll relied on a random statewide sample of all Latino voters across each state with English and Spanish speakers available from the initial point of contact. The data was then weighted to match the U.S. Census Bureau’s data to ensure the correct geographic dispersion, age, education and gender of Latino voters in the state.
Anti-immigrant rhetoric did mobilize Latinos
Mitt Romney’s 9-point victory over President Obama in 2012 was cut to a 4-point victory by President-elect Donald Trump over Hillary Clinton. That’s a 5-point gain by Democrats and an increase of 140,000 votes over Obama, in part because of the growth of the Latino electorate in Arizona.
Voting rates in Arizona compiled by the US Elections Project show that the overall turnout rate among the voter eligible population in Arizona grew from 53 percent in 2012 to 56 percent in 2016.
Our precinct-by-precinct analysis suggests turnout was especially high in majority-Latino areas. Looking at the official election data for Arizona, our research concludes that a Latino vote of 31 percent for Trump, as the exit polls would have you believe, cannot possibly be correct.
Arizona could not have voted 5 points more Democratic with Latinos voting less Democratic. Instead, we see evidence of a growing Latino electorate in Arizona that has been mobilized by the anti-immigrant rhetoric and policies of Republican politicians.
More than 500,000 Latinos voted in 2016
In 2008, the U.S. Census reported that 291,000 Latinos voted in Arizona, growing to 400,000 in 2012 and an estimated 550,000 in 2016. That’s a growth rate of 89 percent in just eight years.
While the growth in the Latino voting community is in part a story of demographic growth, it also reflects the hard work of civic groups such as Promise Arizona, Mi Familia Vota and One Arizona, all of whom have been registering and mobilizing Latino voters. These voter engagement efforts have ramped up in part because of the backlash to legislation such as the anti-immigrant Senate Bill 1070 and the anti-Mexican studies program legislation, House Bill 2281.
To assess the validity of the exit poll claim that Trump won 31 percent of the Latino vote in Arizona, we turned to official precinct data from across the state.
We downloaded individual precinct results for every precinct across Arizona that was reported as of Nov. 22; 1,288 precincts had vote totals posted out of 1,469. In addition to the Clinton-Trump vote, we included the percent of all registrants who are Latino in each precinct from a combination of redistricting data and updated Spanish-surname voter file records by precinct.  We have made this data publicly available for download on our website for anyone to review and replicate.
Less than 15 percent of Latinos voted Trump
Relying on a social-science analysis called ecological inference developed by Harvard political scientist Gary King, our precinct analysis shows Latinos in Arizona voted at rates greater than 80 percent for Hillary Clinton in 2016. An estimated 550,000 Latino voters providing record support for the Democratic candidate is likely why Arizona was a narrow 4-point margin in 2016.
This trend is easily observable in this graphic, which plots all 1,288 precincts across Arizona on a basic X-Y scatterplot.  As the percentage of Latino voters in a given precinct increases, the vote for Hillary Clinton increases linearly.
The Latino vote for Hillary Clinton in Arizona increases linearly
Layered on top of the scatterplot is a weighted regression line that predicts greater than 80 percent Latino vote for Clinton, and less than 15 percent support for Trump. So how is it possible that the exit poll estimated 31 percent Latino vote for Trump when the precinct data suggests otherwise? We suggest it’s not possible.
As the Latino electorate continues to grow, Arizona could be one of the most competitive battleground states in 2020. And polling that accurately captures the growing and pivotal Latino electorate will be essential to good analysis in Arizona and beyond.”