When the U.S. Census Bureau released its 2016 voting and election survey data last week, the numbers painted a grim picture of Hispanic voting gains in the 2016 presidential election: Just 48,000 more Hispanic Texans went to the polls compared with 2012, a dismal 2.5 percent increase, according to the federal count.
The census data resulted in a flurry of headlines indicating that Latino voters in Texas failed to boost their participation in the 2016 presidential election, despite the polarizing presence of candidate Donald Trump.
But the Census Bureau numbers tell only one part of the story, and they do a poor job of actually counting voters, experts say.
Data gleaned from Texas secretary of state voter rolls, which reflect actual votes cast by Texans, show that voters with Spanish surnames increased by nearly 400,000 from 2012 to 2016. That represents a rise of 29 percent, outpacing the non-Spanish surname turnout increase of 9 percent. It’s also more than 10 times higher than the jump reported by the federal government.
The state and federal data also tell two different stories when it comes to voter registration: According to the census, the number of registered Hispanic voters grew by less than 1 percent in 2016, from 2,652,000 to 2,654,000.
The state data show an increase of registered voters with Spanish surnames of 20 percent, from 2.9 million to 3.5 million.
The two data sets come from very different places.
State officials use actual voter rolls to count Hispanic voters based on Spanish surnames. The methodology isn’t foolproof. For example, Hispanic women who marry men and take their non-Hispanic last names would not be counted (and neither would their children). That potentially could lead to an undercount of the Hispanic voting population.
U.S. Census Bureau data, on the other hand, is based on surveys taken after the election. The bureau conducts 60,000 interviews, both in person and by phone, across the country to arrive at its estimates.
The hard numbers determined by the Census Bureau are problematic. For example, the bureau estimates that 9.6 million Texans voted in 2016. But the secretary of state’s numbers show that just over 9 million people actually went to the polls.
Matt Barreto, co-founder of the polling and research firm Latino Decisions, said the census samples were likely “too small and not geographically balanced.”
“I would always trust the actual official voter file and precinct data on this,” he said.
Rice University political scientist Mark P. Jones, one of the state’s leading experts on voter demographics, said that while the state numbers are reliable in measuring the same thing over time, the census estimate comes from sometimes undependable interviews.
“People tend to overreport, rather than underreport, having voted in past elections when answering surveys,” Jones said. “Some people mistakenly remember having voted when they did not, while others accurately recall they did not vote but do not want to admit it and therefore say they voted when they did not.”
His theory on the Census Bureau’s estimate of anemic Hispanic voter growth is that a disproportionately large share of Texas Hispanics overreported in 2012, thus camouflaging gains in 2016.
That might explain the sizable 2012 discrepancy in which the Census Bureau counted 1.9 million Hispanic voters in Texas compared with just 1.4 million Spanish-surname voters counted by state officials.
While the state data paint a less dismal picture of Hispanic voting patterns, they show that Hispanic turnout continues to be problematic. The share of the electorate with a Spanish surname increased from 17.2 percent in 2012 to 19.4 percent in 2016, but Hispanics make up 38 percent of the Texas population.
Turnout of Hispanic registered voters increased from 47 percent to 50 percent, though it still trailed non-Hispanic voters’ turnout, which was 63 percent, down from 65.4 percent in 2012.
“The Texas electorate becomes more Latino and less Anglo with every passing electoral cycle,” Jones said earlier this year. “But the increase is fueled primarily by natural demographic trends rather than by a dramatic spike in participation rates among Latinos.”