I have a comment or two on the article by David Leal et al., “The Latino Vote in 2004” (PS,January 2005, 41–49). First, the William C. Velazquez Institute exit poll was not a national exit poll. Their poll was done in only 14 states and in 56 precincts with high concentrations of Hispanics. The Bush vote is an artifact of their sample. A more comprehensive sample would give different results.

The WCVI presidential vote does not represent Hispanic voters, even in their 14 states, who do not live in high concentration Hispanic areas. The limited disclosure of the WCVI methodology said they only sampled high concentration Hispanic areas. One of the surprising things I learned years ago at the Census Bureau is that a majority of all African Americans lived in low density black areas. To put it another way, a majority of blacks lived in precincts that were less than 25% black. The same is true for Hispanics. Whatever the exact percentage, these Hispanics were omitted from the WCVI study.

Also omitted from the WCVI exit poll were votes cast absentee or by early voters. These voters were more for Bush than Kerry and were very different than those Hispanics voting at polling places on Election Day. Election Day Hispanic voters were almost 2-to-1 for Kerry over Bush. A number of the WCVI 14 states have very large absentee votes or votes cast before Election Day at polling places set up for the purpose.

WCVI claims they were in states with 90% of all Hispanics nationwide. This is a slight overstatement of the Hispanics voting in their 14 states. A clearer statement would have been that the WCVI exit poll only targeted 40% of all Hispanics nationwide. The others were excluded for the reasons cited above. This assumes they sampled precincts that were 20% or more Hispanic. If the minimum Hispanic concentration in a precinct was greater, then the targeted population decreases. In these precincts Bush received 31% of the Hispanic vote, which is about what WCVI's exit poll gave him.

The article by Leal on the Hispanic vote only relies in part on the WCVI exit poll, but their argument contradicts not two exit polls as they claim, but three. The NEP poll they cite was a national exit poll of 250 precincts. The Los Angeles Times national exit poll sampled 125 precincts. The third exit poll I referred to is a weighted sum across all 50 states and DC from the NEP state exit polls. The state and national exit polls do not represent the same voters. Summing across the 50 states and DC under-represents Hispanics slightly as it omits Hispanics in a few states with a very small Hispanic population. In those states NEP did not ask voters to identify themselves as Hispanic. Even so, the sum across the states produces a Hispanic Republican vote of 40%. However, that 40% vote for Bush is based on over 4,471 Hispanic interviews nationwide. The 44% Bush vote from the national exit poll is based on 250 precincts and 1,037 Hispanic respondents. Both figures include absentee voters. Any estimate of Hispanic vote from an exit poll will have a big variance due to clustering. The 40% figure, when it was announced by NBC at a conference following the election, was taken as a correction to the 44% figure. That is not the case. The 40% and 44% Bush vote figures are based on two different sets of exit polls conducted by Edison Media Research and Mitofsky International for NEP. We believe both to be reliable estimates of the Hispanic vote. They both show a significant increase over the Hispanic vote for Bush in 2000. The Los Angeles Times exit poll was consistent with these estimates. It had Bush receiving 45% of the Hispanic vote.

The comment in the Leal article about the under-representation of urban areas in the NEP exit polls is an artifact of how urban is defined. There is no standard definition of a suburb. In the NEP exit polls a suburb is the balance of an MSA outside the principle city(s). By any definition suburbs are still part of an urban area. Excluded from the suburbs in the NEP scheme are the central city and other smaller cities in the MSA that might have been considered suburban by many demographers. These smaller cities, when outside the MSA, are classified as rural by NEP. They are only classified as small cities when in an MSA.

I think the article by David Leal et al. is wrong in its conclusion. The Los Angeles Times and NEP exit polls were a more appropriate method for estimating the Hispanic vote than the preelection surveys the authors used as the basis of their arguments. Most of these surveys were taken months before the election. The methods for sampling Hispanic voters are not given. Shall we accept as an article of faith that all these surveys used probability sampling methods to reach Hispanic voters? The vote Leal et al. cite averages 60%-Kerry, 32%-Bush. What did the other 8% of Hispanic voters do? Also overlooked is that there was a campaign in progress. Campaigns sometimes change votes. The article ignores this.

If one wants access to the respondent data from the NEP exit polls they have been archived at ICPSR at the University of Michigan and at the Roper Center at the University of Connecticut. The methods statements are available at www.exit-poll.net along with the questionnaires, the completion rates, and an evaluation of the NEP exit polls.

Editor's note: David L. Leal, Matt A. Barreto, Jongho Lee, and Rodlofo de la Garza will respond to Mr. Mitofsky's comments in the July issue of PS.