Sergio Bendixen disagrees with our poll analysis
11/19/2012
Prominent pollster Sergio Bendixen of Bendixen & Amandi International disagrees with Crowley Political Report's view of recent polling of Cuban-American voters. That story, which appeared here and in Columbia Journalism Review noted considerably confusion in the reporting of polls and questioned the disparity among pollsters. I think Mr. Bendixen may have missed the point.
Since Mr. Bendixen spent time writing a lengthy rebuttal (which he added as a comment to the original story) - it seems only fair to post his reply here:
From Sergio Bendixen
When I first came to the United States from Peru in 1961, Ricky Ricardo was the only television personality I could identify with. So any time his image is used to challenge anything Cuban, I must come to his defense.
The recent Crowley Political Report is missing some important information which has obviously clouded the judgment of its author when it comes to his conclusions.
1.) Most - if not all - researchers and pollsters agree that if you want to accurately measure the voting behavior of a demographic group, exit polling is the best way to do it, as long as it includes interviews with absentee and other early voters. Pre-election polls cannot predict accurately which voters will show up at the polls and which voters will not. Precinct analysis of voting results cannot reveal the voting behavior of Cuban voters because there are no "pure Cuban" precincts in Miami-Dade County or any other county in the United States. The top 50 "Cuban precincts" in Miami-Dade County (according to the U. S. Census) have a substantial percentage (20% to 50%) of African American, White Anglo and non-Cuban Hispanic registered voters within them.
2.) The only organizations that conducted an exit poll in Florida among Cuban voters were Edison Research (for the major television networks and AP), the Pew Hispanic Center and Bendixen & Amandi International (B&A: my firm). All of the other polls and studies that Mr. Crowley mentions in his report were either pre-election polls or post-election precinct analysis of Miami-Dade "Cuban precincts." Those include the FIU/Miami Herald/El Nuevo Herald poll of likely Florida Hispanic voters conducted by Professor Eduardo Gamarra in October, the FIU post-election precinct analysis utilizing "ecological regression" methodology conducted by Professor Dario Moreno (incorrectly identified in the Crowley report as an exit poll), and the Latino Decisions "eve of the election" Florida Hispanic poll. The report also cites an electoral analysis by Professor Ben Bishin of the University of California at Riverside.
3.) Mr. Crowley misses an obvious pattern in the data. The three statewide exit polls show extremely similar results. All of them show Obama and Romney splitting the statewide Cuban vote (Edison and Pew had Obama at 49% and Bendixen & Amandi had Obama at 48%). In other words, all three exit polls had the same finding - Cuban voters supported a Democratic presidential candidate at a historic level in 2012.
4.) The FIU post-election precinct analysis is consistent with the Bendixen & Amandi exit poll results. Professor Moreno estimated that the vote for Obama in "Cuban precincts" in Miami-Dade County was between 41 percent and 45 percent. The B&A exit poll interviewed more than 2,000 Cuban voters statewide. It indicated that Obama received 44 percent of the Cuban vote in Miami-Dade County and 59 percent in other Florida counties. Cuban voters in Hillsborough (Tampa), Orange (Orlando), Osceola (Kissimmee) and Broward (Ft. Lauderdale) Counties represent 25% of the statewide Cuban electorate and B&A interviewed more than 500 of them.
The Edison and Pew exit polls did not report regional numbers because their sample of Cuban voters was much smaller than ours.
5.) The results of the FIU/Miami Herald/El Nuevo Herald October poll of Florida Hispanic voters can be ignored not only because Professor Gamarra utilizes discredited computerized phone interviews for his polls but also because it obviously could not have measured changes in the presidential choice of Cuban American voters over the last weeks of the intense 2012 campaign. For example, a couple of controversial Spanish-language television ads during the last week of the campaign could have easily had a major impact on Cuban voting patterns.
6.) The Latino Decisions "election eve" poll is more credible. Nevertheless, the B&A exit poll shows a substantial difference between absentee and early Cuban voters (Obama at 39%) and Cuban Election Day voters (Obama at 52%). It is probable that this "election eve" poll was not able to measure the powerful impact of the Obama GOTV operation or the negative impact on elderly Cuban Romney supporters of the 3 to 4 hour lines at the polls on November 6th.
7.) The analysis by Professor Ben Bishin seems to be based on the FIU precinct analysis. After reading his blog on "The Monkey Cage," it is clear that it is also based on incorrect information. For example, he does not seem to know (or he conveniently ignores) the fact that the figures published by the three exit poll organizations are based on statewide interviews while the FIU precinct analysis is limited to Miami-Dade County.
8.) It is true that B&A works mostly for Democratic candidates. You should also know that B&A conducted all of the polling research for the Obama Hispanic campaign and produced many of its Spanish-language television ads. Nevertheless, it is important to make clear that the exit poll of Florida Hispanic voters was financed entirely and independently by the firm
So Mr. Crowley, I think you got some "splainin’ to do."
Comments