In 2012, a weekly tracking poll found that just 55% of Latino voters were certain to vote for President Obama with another 9% leaning toward him, but not certain. All told, it was 64% for Obama with 12% undecided. By Election Day, the same poll found Obama with 73% of the Latino vote, a gain of 9 points. Today, a similar pre-election tracking poll finds Biden running at about the same rate as Obama in 2012, at 65% support.
Many of the polls cited in stories about Biden’s lack of Latino support are notorious for missing the mark on basic polling methodology with respect to Hispanics, often basing their results on small samples that will garner a margin of error of plus or minus 10 points, which is never reliable.
A foundational practice of accurate Latino polling is offering interviews in Spanish. This is not a regular practice of many mainstream pollsters. Reputable pollsters who know how to accurately poll Latinos consistently find 40% or more of Hispanic respondents in Florida and in Texas opt to take surveys in Spanish. If pollsters do not offer Spanish-speaking interviews, their surveys are suspect.