By now, most denizens of this site realize that most polls, both national and state, do not accurately include Latinos in their models. For one things, Latinos heavily rely upon cell phones, and those polls which use automatic dialers cannot call cell phones (which may mean they under-sample young people as well). While some automated pollsters attempt to adjust their models to account for this, not all do.
Further, there is a significant percentage of pollsters who do not offer a bilingual option for those Spanish speaker who are not comfortable being interviewed in English. We have known for some time this result in (1) under-reporting Latinos as a share of the voting electorate, and (2) possibly tilting the distribution of those Latinos participating as inclined to vote for Republicans.
Latino Decisions has been running a poll for 11 weeks with bi-lingual interviewers, and since it does not auto-dial not needing the adjustment for cell phones, which are included. You can see the cumulative results here
A couple of things to note:
1. Rate of Hispanic participation they expect a total participation rate of 77% or more - in this blog post they note that 16% of Latinos have already voted and another 73% expect to vote. Latino Decisions expects at least 73% of registered Hispanic voters to participate, which would exceed the previous 72% record in the '96 election.
2. In the poll itself, immigration reform and the Dream Act almost equal the Economy as a key , trailing only by 47-43.
3. Among those likely to vote or who have already voted, the poll shows a split in favor Obama by 73-22 with 3% undecided.
The polls I have examined that give demographic cross-tabs tend to somewhat underestimate Latinos as a percentage of the vote, usually by 1-2% if Latino Decisions is to be believed. They also are more generous towards Romney than Latino Decisions projects - some showing Romney getting 1/3 of the Latino vote.
The effects are more noticeable in states with substantial Latino populations. Consider - PPP showed 7 point margin for Romney in AZ, with 16% of model being Hispanic. Latino Decisions predicts 20% will be Latino, with Romney getting only 16% of Latinos. That is a 6+-point difference in the top line add 3+ to Obama and take it from Romney.
“With 11 weeks of tracking, we are headed towards a record level of Latino votes for a Democratic presidential candidate,” said Matt Barreto, principal investigator for Latino Decisions. “If Latinos turnout at the high rates we are expecting, they could deliver Nevada, Colorado, Florida and Virginia to Obama.”
Assuming Obama wins OH, winning any two of those four states would effectively guarantee the election for Obama. And if Obama is winning all four - and as I noted above, possibly AZ as well - expect to see a significant down-ballot impact as well, given that
68% had already voted or were certain they would vote for Democrats, and another 4% saying they lean towards Democrats in the Congressional vote.
In my own predictions, I take seriously the data and judgment of Latino Decisions, which is why on Friday I predicted that Obama would carry all four of NV, CO, FL and VA. The only one of those in question in my mind is CO, given the uncertainty of how many left libertarians Gary Johnson might draw from Obama given the initiative to legalize marijuana. As of now I have not changed my mind.
What do you think?