The rapid growth of the Latino electorate and their presence in swing states like Colorado, Nevada, New Mexico and Florida ensure that news organizations will be fed a steady diet of Latino polling in the lead up to 2012. Many of these polls will have been conducted by researchers who attempt to maximize the efficiency of their calls by only dialing voters with Hispanic surnames. The assumption behind this methodology is either that a preponderance of Latino voters have Hispanic surnames or that Latino voters with Hispanic surnames do not vote differently than Latino voters who do not have Hispanic surnames. This methodology begs two questions: 1) how many Latino voters have non-Hispanic surnames?; and 2) do Latino voters with non-Hispanic surnames have demographic and political differences relative to Latino voters with Hispanic surnames?
The only way to answer those questions is by drawing a sample of every voter in a jurisdiction regardless of surname and asking whether the voter self-identifies as Hispanic. The four statewide polls of Latino voters I have conducted in that manner suggest that the assumptions on which Hispanic surname polling relies do not hold up. Significant numbers of Latino voters do not have Hispanic surnames. Furthermore, Latinos that do not have Hispanic surnames are more likely to be U.S.-born Latinos rather than immigrants, resulting in predictable demographic and political differences compared to voters with Hispanic surnames. As a result, any poll of Latino voters that relies only on voters with Hispanic surnames is introducing bias even before the first interview is conducted.
My Polls that Screened on Ethnicity
The states I polled without any reliance on Hispanic surname collectively contain more than a third of the nation’s Latino voters. In each one, I drew a sample based on every voter and asked each respondent whether he or she is Hispanic. While rigorous, this methodology is also inefficient and expensive. But it is the only way to ensure that the sample has a truly representative number of Hispanics that do not have Hispanic surnames. The California poll was self-funded; the New Mexico poll was commissioned by the 2008 Obama presidential campaign; and the Arizona and Nevada polls were commissioned by the Annenberg Public Policy Center. The sample for each poll came from a different list provider: Political Data Inc. (California), Catalist (New Mexico), and the respective Democratic state parties for Arizona and Nevada. The source of the poll is significant because every file vendor has a different way of trying to identify Hispanic voters. For example, different list companies will have more or less exhaustive dictionaries of Hispanic surnames. To allow for vendor differences, the phrasing I will use in addition to “having a Hispanic surname” is “tagged as Hispanic” to take into account the various methods companies like Catalist use to make their Hispanic lists more robust. The source with the highest match was Political Data Inc. – 62 percent of the Latino voters I interviewed in California were tagged as Hispanic.[i]
Voters with Hispanic Surnames Vote Differently
In each poll, Hispanic voters with Hispanic surnames vote differently than Hispanic voters that do not have Hispanic surnames. For example, in my 2006 post-election poll of California Latino voters, voters who were not tagged as Hispanic said they voted for Arnold Schwarzenegger (43 percent Schwarzenegger – 36 percent Angelides) while voters who were tagged as Hispanic said they voted for the Democratic nominee, Phil Angelides (48 percent Angelides – 29 percent Schwarzenegger). In all 4 states, tagged Hispanics are more Democratic and less Republican than non-tagged Hispanics.
The finding that voters with Hispanic surnames have different political preferences than voters without Hispanic surnames makes sense. My research suggests that all immigrant populations move to the mean over generations. U.S.-born Vietnamese are not as predominantly Republican as Vietnamese immigrant voters, U.S.-born Cubans are not as predominantly Republican as Cuban immigrant voters and U.S.-born Mexicans are not as predominantly Democratic as Mexican immigrant voters. In my surveys, only 17 to 18 percent of immigrant voters in all four states self-id’ed as Republican while U.S.-born Latino voters ranged between 24 and 31 percent. Knowing that Latino voters move to the mean over generations is the first step towards understanding why voters with Hispanic surnames have different political preferences.
The second step is knowing that U.S.-born Hispanic voters are less likely to have Hispanic surnames than are immigrant voters. This is what one would expect owing to intermarriage and Anglicization. In all four of my polls, immigrants are more likely to be tagged as Hispanic than are U.S.-born Latino voters. Consequently, polls that depend on voters with Hispanic surnames in states with largely Mexican-American Latino populations can be expected to oversample immigrant populations and, as a result, to overstate Democratic strength.
Did 90 Percent of Latinos Vote for Reid in Nevada?
Post-election controversies about the accuracy of the Hispanic results in exit polls are a common occurrence and 2010 was no exception. The discrepancy that got the most coverage was Senator Reid’s Latino support in the CNN and the Latino Decisions exit polls. As reported in Newsweek, “According to election-eve polling and analysis by Latino Decisions, a surveying firm, Hispanics chose Reid over Angle 90 percent to 8 percent—an astounding margin. CNN’s exit polls showed a significantly smaller spread, with Reid winning 68 percent to Angle’s 30 percent.”[ii]The fact that Latino Decisions appears to call only voters with Hispanic surnames[iii]may explain some of the difference.
Nevada is a state whose Latino voters mostly have Mexico as a country of origin.[iv]If the sample of Latino Decisions’ exit poll was drawn only from Latinos with Hispanic last names, then they oversampled Latino immigrants.[v]Latino immigrants in Nevada tend to be stronger Democrats than are later generation, U.S.-born Latinos. Because later generation Latinos are more likely to have non-Hispanic surnames and Latino Decisions did not interview anybody with non-Hispanic surnames, they left out the Latinos that are most likely to self-identify as Republicans and therefore to have voted for Sharron Angle. As a result, their exit poll is likely to have overstated Harry Reid’s performance with Latino voters.
News organizations would be wise to consume Hispanic-surnamed polls with a grain of salt. Not only does the methodology already skew today, but it will skew even more with each passing year as the proportion of U.S.-born to immigrant voters rises and more Latino voters lose their Hispanic surname. Pollsters who only interview Latino voters with Hispanic surnames have something to contribute in terms of an overall look at Latino voters, but their methodology compromises the accuracy of their results.
[i] The challenge in identifying Hispanic voters by surname can be seen in the last names of the Latino voters in my New Mexico poll that were not tagged as Hispanic by Catalist (the appendix to this memo). No matter how good the surname dictionary, voters with last names like Blair and Sizemore are not likely to be tagged as Hispanic.
[ii] Arian Campos-Flores, “Did Hispanics Save Harry Reid?”, Newsweek, November 3, 2010.
[iii] Latino Decisions, “Methodology,” Latino Decisions, http://latinodecisions.wordpress.com/methodology/ (accessed February 20, 2011).
[iv] Sixty-seven percent in my poll.
[v] It would interesting to know but difficult to assess the quality of Hispanic surname dictionary used by Latino Decisions. As seen in the Table on page 2, the file given to me by the Nevada Democratic Party in 2008 had only a 27 percent match. In other words, doing a poll using a sample drawn from voters tagged as Hispanic on that file would mean that 73 percent of Latino voters would have had no chance of being called because of their surname. Even a sample drawn from the file with the best match (California’s) would result in a poll where 38 percent of Latino voters have no chance of being called. Anyway you slice it, Latino voters with last names like Richardson are very unlikely to be tagged as Hispanic. It is not a Hispanic surname. Latino Decisions argues on their website that “95% of all Latinos in the U.S. have a Spanish-surname.” Voters, though, are a small and distinctive subset of Latinos in the U.S. Among other things, voters are more likely to have been born in the U.S., better educated, more likely to speak English and, as my polls demonstrate, less likely to have a Hispanic surname.
On Tuesday, Sept. 27, 2011, Andre Pineda passed away. His passing came as shock to everyone who came in contact with him. I am comforted with the thought that he is busy in heaven polling the angels.
Andre Pineda is CEO and Founder of Pineda Consulting; a public opinion and strategic communications firm based in Passadena, CA with over 20 years of experience at local, statewide, federal and international level. His clients have included the Obama Presidential Campaign, the Democratic National Committee, the National Education Association and companies like Visa and Symantec.
Pineda Consulting 891 Adelaide Drive Pasadena CA 91104 626 4827929 www.PinedaConsulting.com