By Gary Segura and Gabriel Sanchez
Whit Ayers, of Ayres/McHenry, responded to the criticisms raised of his new poll of Latinos by making three points.
1) That language was not the fifth question but that they used call-backs.
2) That my comparison of nativity figures to the Latino National Survey is comparing registered voters to all respondents, which includes non-citizens, and is hence an inaccurate comparison.
3) That national demography is not informative for evaluating a statewide survey.
With respect to the first point, call-backs produce low Spanish language response rates, a point we have made on a number of occasions. Good survey practice is to complete the interview when you make a successful contact. Calling respondents back dramatically lowers incidence and completion rates. That only 2% of respondents completed the survey in Spanish in a state where a supermajority of Latino citizens are foreign born and speak Spanish as a first language suggests serious mis-representation.
With respect to the comparison to the Latino National Survey, Mr. Ayres is misstating the original post. In the original post, Gary Segura said “the Latino National Survey (the largest academic survey of the Latino population of the US ever conducted) estimated that foreign-born first generation respondents were 73.5% of Florida’s citizen registered population, 21.2% of Colorado’s registered citizens and 16.3% of New Mexico’s.” (emphasis added). The numbers Gary drew from the LNS are just among citizens registered to vote (that is, the same as Mr. Ayres), not among all Latino adults, non-citizens included.
Third, Mr. Ayres notes that national data may not reflect state-level differences. To address this point we analyzed state-level data from the current population survey. In Florida, among Mr Ayres’ sample, 56% earned incomes about $50k. In the Current Population Survey November 2010 supplement (the government’s official estimate of registration and voting), looking at just Latino registered voters from Florida, the actual number is under 43%. Likewise, in New Mexico, Ayers has 42% of his respondents over $50k in income while the CPS has 36.6%, and in Colorado, Ayres reports 54% of his respondents have over $50k in income while the official estimate from the Current Population Survey is 38.4%. Again, these comparisons are apples-to-apples, CPS data are restricted to Latino citizen registered voters over 18 years of age, identical to his sampling frame.
Similarly, the education numbers state-by-state are a problem in the Ayres data. Ayres/McHenry’s data estimate college completion at 40% each for Colorado and New Mexico and 61% in Florida. The Current Population Survey estimates, state-by-state, are 24% in Florida, 16.3% in New Mexico, and 14.7% in Colorado. By any estimation, and directly comparing the state level results from the Current Population Survey for registered Latino citizens to the Ayers/McHenry data, we find that their data over-estimate college completion by 24 to 37 percentage points.
Gary Segura is Political Director, Latino Decisions, and Professor of Political Science at Stanford University. Gabriel Sanchez is an Associate Professor of Political Science at the University of New Mexico and Research Director for Latino Decisions.