Post by Moses on Jan 1, 2005 9:13:59 GMT -5
www.wcvi.org/latino_voter_research/polls/national/2004/flores.html#_ftnref1
Dr. Henry Flores, PhD
St. Maryís University
All data presented is election day minus absentee ballots
Introduction
There has been disagreement as to the levels of national Latino support enjoyed by both President Bush and Senator Kerry in the 2004 Presidential elections. The National Election Poll (NEP) and Los Angles Times exit polls indicated that 44% and 45% of the Latino electorate respectively voted for President Bush on election-day. However, sample design flaws have called those findings into question.
The William C. Vel·squez Institute (WCVI), the non-profit, non-partisan sister organization of the Southwest Voter Registration and Educational Project (SVREP), founded by Willie Velasquez in 1985 conducted the first national exit poll of Latinos on November 2, 2004. Never before in the history of presidential exit polls had an organization conducted a poll exclusively of Latino voters. The findings of this historical poll contradicted those of the NEP and Los Angles Times. The WCVI poll found that only 33.0% of Latinos queried on election-day voted for President Bush while 65.4% supported Senator Kerry. Another 1.4% of Latinos polled indicated that they had cast their votes for Ralph Nader and his vice-presidential running mate.
What was most important about the WCVI findings was not that they differed greatly from those of the NEP and LA Times polls, but that they only varied slightly from five other pre-election polls. Additionally, these findings indicate that Latinos had changed their voting patterns little since the 2000 general election. In 2000 both the WCVI and an independent research source found that President Bush had received only 35% of the Latino national vote. When factoring in the margin of error for both polls the 2004 findings of 33.0% support for President Bush indicate that Latinos had continued to vote in the same manner as they had historically.
The most important question underlying the disagreement between the findings of the NEP, Los Angeles Times and WCVI polls is why the findings of Latino support for the presidential candidates differed so markedly? The twelve to thirteen point variance goes well beyond any possible margin of error. The answer may be found in the methodologies utilized in designing the sampling frame. In other words, if one compares the samples of all three polls one will find why a difference exists between the pollsí respective outcomes.
Methodological Considerations
The Los Angeles Times Sample
The Los Angeles Times poll sample was constructed out of two samples, a nationwide and a California state sample. Both were drawn separately and then combined to increase the number of Latino respondents. The total sample was comprised of 5,154 interviewees questioned at 136 polling places across the nation. Of the total sample, 3,357 or 65.1% were California voters who were interviewed at 50 polling places.(1) The Times indicated that the ì[sample] precincts were chosen based on the pattern of turnout in past primary elections.î The survey was self-administered, was confidential and was conducted in either English or Spanish. It was also noted that the margin of error was a plus or minus three percent for all voters but that for some subgroups the margin was higher.(2) This was as much detail available on how the sampling frame was constructed. The gathered demographics indicated that 79% of the sample was comprised of voters who identified themselves as White, another 10% were Black, 5% of the sample identified themselves as Latino, and 3% considered themselves Asian.
Several fundamental flaws are initially clear concerning the construction of the Los Angeles Times poll. First, it is inappropriate, methodologically, to draw two independent samples and then combine them into one larger one without considering controlling for the margin of errors of both samples. This requires conducting a statistical test insuring the complete independence of both samples. It is not clear whether this step was taken including the exact test utilized. Secondly, the manner in which the samples were drawn and then combined resulted in California voters being heavily over sampled, including the Latino voter sub sample. Finally, if one performs the simple arithmetic function of determining the size of the Latino sub sample, one discovers that the 5% Latino sub sample represents a total of 258 cases. If one also considers that Latinos comprise 7.3% of all registered voters nationally, then the Los Angles Times poll under sampled Latino voters by approximately 32%. So the Latino sub sample was both under sampled and biased toward California voters. This last contention is only a suspicion because of the lack of information concerning the sampleís construction available from the Los Angles Times indicating what percentage of the Latino sub sample was drawn from the California sample and what percentage was drawn from the remainder of the United States. An additional item of important information that is lacking is how many total number of Latino interviews beyond California were conducted, in what states they were drawn from, and whether these interviews were drawn from suburban, rural or central city areas.
As a result the contention that the Los Angles Times made that based upon their exit poll 45% of Latino voters nationally supported President Bush must be considered carefully because of the problematic nature of the Latino sub sample.
The National Election Poll (NEP)
The National Election Poll was commissioned by a consortium of media corporations that included ABC, Associated Press, CBS, CNN, Fox, and NBC to provide election day exit polling for use as news outlets affiliated with the consortium reported the election returns throughout the evening of Election Day. The data were collected by Edison Media Research/Mitofsky International who interviewed 13,110 individuals who exited 250 polling places around the nation.(3) The Latino sub sample was comprised of 1,031 individuals representing 7.8% of the entire sample. This was reflective of the proportion of Latinos who are registered voters nationally. The Latino sample, however, was not accurate in several regards. For instance, the sample was drawn principally from geographical regions having populations of less than 500,000 with only 25% or 258 coming from areas having populations of over 500,000. Additionally, 33% of the entire Latino sample was drawn from the Southern Region of the United States. Essentially, the NEP Latino sample did not reflect the normal residential patterns of Latinos nationally in that the Brookings Center on Urban & Metropolitan Policy and the Pew Hispanic Center estimated during the 2000 census that fully 52% of Latinos resided in ìEstablished Latino Metros.î(4)
Other demographic characteristics of the NEP Latino sample bring into question the integrity of the NEP exit pollís findings. For instance, 46% of those Latinos interviewed indicated that they had completed at least a Bachelorís degree. This percentage is even higher than the overall national average and highly erroneous. Additionally, only 39.3% of the Latino interviewees identified themselves as being of Mexican origin or ancestry. Data provided in the Pew Hispanic Centerís ì2004 National Survey of Latinos: Politics and Civic Participationî which was conducted in July 2004 found that only 16% of Latino registered voters and 10% of all Latinos had achieved a Bachelorís degree or higher. The same survey found that 60% of registered Latinos identified themselves as having Mexican ancestry.(5)
(continued)
Dr. Henry Flores, PhD
St. Maryís University
All data presented is election day minus absentee ballots
Introduction
There has been disagreement as to the levels of national Latino support enjoyed by both President Bush and Senator Kerry in the 2004 Presidential elections. The National Election Poll (NEP) and Los Angles Times exit polls indicated that 44% and 45% of the Latino electorate respectively voted for President Bush on election-day. However, sample design flaws have called those findings into question.
The William C. Vel·squez Institute (WCVI), the non-profit, non-partisan sister organization of the Southwest Voter Registration and Educational Project (SVREP), founded by Willie Velasquez in 1985 conducted the first national exit poll of Latinos on November 2, 2004. Never before in the history of presidential exit polls had an organization conducted a poll exclusively of Latino voters. The findings of this historical poll contradicted those of the NEP and Los Angles Times. The WCVI poll found that only 33.0% of Latinos queried on election-day voted for President Bush while 65.4% supported Senator Kerry. Another 1.4% of Latinos polled indicated that they had cast their votes for Ralph Nader and his vice-presidential running mate.
What was most important about the WCVI findings was not that they differed greatly from those of the NEP and LA Times polls, but that they only varied slightly from five other pre-election polls. Additionally, these findings indicate that Latinos had changed their voting patterns little since the 2000 general election. In 2000 both the WCVI and an independent research source found that President Bush had received only 35% of the Latino national vote. When factoring in the margin of error for both polls the 2004 findings of 33.0% support for President Bush indicate that Latinos had continued to vote in the same manner as they had historically.
The most important question underlying the disagreement between the findings of the NEP, Los Angeles Times and WCVI polls is why the findings of Latino support for the presidential candidates differed so markedly? The twelve to thirteen point variance goes well beyond any possible margin of error. The answer may be found in the methodologies utilized in designing the sampling frame. In other words, if one compares the samples of all three polls one will find why a difference exists between the pollsí respective outcomes.
Methodological Considerations
The Los Angeles Times Sample
The Los Angeles Times poll sample was constructed out of two samples, a nationwide and a California state sample. Both were drawn separately and then combined to increase the number of Latino respondents. The total sample was comprised of 5,154 interviewees questioned at 136 polling places across the nation. Of the total sample, 3,357 or 65.1% were California voters who were interviewed at 50 polling places.(1) The Times indicated that the ì[sample] precincts were chosen based on the pattern of turnout in past primary elections.î The survey was self-administered, was confidential and was conducted in either English or Spanish. It was also noted that the margin of error was a plus or minus three percent for all voters but that for some subgroups the margin was higher.(2) This was as much detail available on how the sampling frame was constructed. The gathered demographics indicated that 79% of the sample was comprised of voters who identified themselves as White, another 10% were Black, 5% of the sample identified themselves as Latino, and 3% considered themselves Asian.
Several fundamental flaws are initially clear concerning the construction of the Los Angeles Times poll. First, it is inappropriate, methodologically, to draw two independent samples and then combine them into one larger one without considering controlling for the margin of errors of both samples. This requires conducting a statistical test insuring the complete independence of both samples. It is not clear whether this step was taken including the exact test utilized. Secondly, the manner in which the samples were drawn and then combined resulted in California voters being heavily over sampled, including the Latino voter sub sample. Finally, if one performs the simple arithmetic function of determining the size of the Latino sub sample, one discovers that the 5% Latino sub sample represents a total of 258 cases. If one also considers that Latinos comprise 7.3% of all registered voters nationally, then the Los Angles Times poll under sampled Latino voters by approximately 32%. So the Latino sub sample was both under sampled and biased toward California voters. This last contention is only a suspicion because of the lack of information concerning the sampleís construction available from the Los Angles Times indicating what percentage of the Latino sub sample was drawn from the California sample and what percentage was drawn from the remainder of the United States. An additional item of important information that is lacking is how many total number of Latino interviews beyond California were conducted, in what states they were drawn from, and whether these interviews were drawn from suburban, rural or central city areas.
As a result the contention that the Los Angles Times made that based upon their exit poll 45% of Latino voters nationally supported President Bush must be considered carefully because of the problematic nature of the Latino sub sample.
The National Election Poll (NEP)
The National Election Poll was commissioned by a consortium of media corporations that included ABC, Associated Press, CBS, CNN, Fox, and NBC to provide election day exit polling for use as news outlets affiliated with the consortium reported the election returns throughout the evening of Election Day. The data were collected by Edison Media Research/Mitofsky International who interviewed 13,110 individuals who exited 250 polling places around the nation.(3) The Latino sub sample was comprised of 1,031 individuals representing 7.8% of the entire sample. This was reflective of the proportion of Latinos who are registered voters nationally. The Latino sample, however, was not accurate in several regards. For instance, the sample was drawn principally from geographical regions having populations of less than 500,000 with only 25% or 258 coming from areas having populations of over 500,000. Additionally, 33% of the entire Latino sample was drawn from the Southern Region of the United States. Essentially, the NEP Latino sample did not reflect the normal residential patterns of Latinos nationally in that the Brookings Center on Urban & Metropolitan Policy and the Pew Hispanic Center estimated during the 2000 census that fully 52% of Latinos resided in ìEstablished Latino Metros.î(4)
Other demographic characteristics of the NEP Latino sample bring into question the integrity of the NEP exit pollís findings. For instance, 46% of those Latinos interviewed indicated that they had completed at least a Bachelorís degree. This percentage is even higher than the overall national average and highly erroneous. Additionally, only 39.3% of the Latino interviewees identified themselves as being of Mexican origin or ancestry. Data provided in the Pew Hispanic Centerís ì2004 National Survey of Latinos: Politics and Civic Participationî which was conducted in July 2004 found that only 16% of Latino registered voters and 10% of all Latinos had achieved a Bachelorís degree or higher. The same survey found that 60% of registered Latinos identified themselves as having Mexican ancestry.(5)
(continued)