Leave the polling and analysis to professionals - UPDATED

It came to my attention that a blogger has taken issue with my post about a recent poll conducted by Survey USA and commissioned by CBS 4 TV in Miami that showed McCain leading by wide margins among Hispanics in Florida. Here's an excerpt:

Mr. Gomez seemed to be ecstatic with the results of the Survey USA poll (because they "blew away" the myths of the evil mainstream media), but he didn't look at the sample demographics. The poll showed that Hispanics in Florida picked Sen. McCain over Obama at 55 - 36% (with 9% undecided). But, the Survey USA poll vastly under-represents Hispanics in Florida.

According to US Census data and estimates, Hispanics in Florida currently make up about 20-22% of the population. The Survey USA poll has Hispanics at 13%. This is closer to estimates from 1990(!). Also, Survey USA over-represents the 65 and older Florida population. They have it at 27%, when actual estimates are around 18%.

But, these relevant facts don't bother Mr. Gomez. He's busy celebrating how the mainstream media got "blown away."

Let's start with the first point which that this not-so anonymous blogger makes. He takes issue with the demographic make-up of the respondents in the poll, saying that they don't conform to the real population of Floridians. Right off the bat he's heading down an incorrect path.

The purpose of a political poll such as this one is not to gauge the opinions of the total population but to gauge the opinions of likely voters. To be a likely voter you must be eligible to vote in the first place, and then you need to be registered to vote. There's other criteria too and every pollster has a different formula for calculating likelihood of voting but you have to be registered at a bare minimum

As we all know voter eligibility and registration is not uniform across all demographic segments. For one thing many Hispanics in Florida are immigrants. Immigrants who have not gained citizenship are not eligible to vote. Therefore we know that the pool of potential Hispanic voters is smaller than the Hispanic population.

The next issue is of course registration. There are many people who are eligible to vote but not registered. These are people that for whatever reason are less engaged in politics. Guess who happens to be less engaged in politics? Young people. Guess who are really engaged in politics? Old people. It's not surprising that a poll of likely voters contains what appears on the surface to be a disproportionate number of older respondents.

But here's the thing, even if demographic weighting of the overall poll were out of whack as this Mensa Club member suggests, that does not affect the results WITHIN EACH component demographic segment. And even this idiot acknowledges that "There's no doubt that McCain is leading in Florida." He quotes another poll to say that gives credence to the Survey USA poll. But it's the overall poll numbers that would be affected by an error in weighting the proportion of Hispanics (that he suspects) not the results WITHIN the Hispanic sample themselves.

Let me ask a rhetorical question: Who is more likely to know how to weight a political poll, a company whose job it is to conduct such polls and has been doing it for eighteen years with clients that include media companies from around the country or some commie back bench blogger with an ax to grind?

UPDATE: I should have thought of looking earlier but the Florida Division of Elections posts voter registration statistics by ethnicity on its web site. It turns out that Florida has 10,618,519 registered voters and 1,242,509 of them are Hispanic. That's 12%. The Survey USA poll in question had a sample that was 13% Hispanic. The commie back bench blogger felt it should be 20%-22%. So there's the answer to my question.

Posted by Henry Louis Gomez at September 26, 2008 02:00 PM