St. Pete Polls…are they any good?

Over the last few weeks, political observers in Florida have been fascinated by a new organization calling themselves St. Pete Poll. In addition to the presidential and senate polls we see nearly on a daily basis, St. Pete Polls have gone further, giving Floridians an insight into congressional, state house and state senate races. What makes this interesting is that there is now somewhat of a snapshot of races that would otherwise be ignored.

Even with this being said, are these St. Pete Polls any good?

Over the last month, St. Pete Polls have conducted two polls on congressional and state-level races. One was conducted on the first weekend of September and the other conducted in the first week of October. What is interesting is that some polls show a vast difference between previous polls, yet nothing earth-shattering has happened. Why is this the case?

The first set of polls that I want to look at is the State Senate District 8 polls. In September, Republican Dorothy Hukill was beating Democrat Frank Bruno by .9%, well within the poll’s margin or error. This makes sense, since most people expect this to be a battleground race. But when the same poll was conducted this month, Hukill had a 20% lead. What happened to Frank Bruno? Did he die? Did he get caught up in a scandal that only people in the 8th District know about? What happened? In my 20+ years of being involved in politics, I have never seen a poll, even at the most local level, shift so drastically over the course of one month without some highly-charged independent variable being thrown in. There is no reason that this poll should have shifted so drastically, and no explanation from St. Pete Polls to the reason for the drastic shift.

The next poll that raises some eyebrows will be that of Congressional District 26, where ethically challenged Congressman David Rivera faces off against Democrat Joe Garcia. On the surface, the results look fine, which is 43.4% for Rivera and 40.5% for Garcia (though almost all other polls in this race have favored Garcia regularly). These are the weighed results. The non-weighed results give it a dead 42.1%-42.1% tie. On the outside, this looks like it is going to be a very close race, which is what most people expect.

When looking at the Rivera vs. Garcia race, the questions aren’t about the the face value of the poll, but the details within the poll. Of course, this poll is weighed to make up for any lack of under or over-sampling of a certain demographic. But in this case, the number is staggering. Of the total respondents to the St. Pete Poll, only 23.2% were Hispanic. Yet, the district’s Hispanic voting age population is 69%. Even if that number were to be weighed to reflect the electorate in District 26, it is still way too small of a sample to get an accurate assessment of the voting intentions of the district.

Of course, this leads to whites (who St. Pete Polls describe specifically as “White, Not Hispanic”) who only make up 20% of the voting age population of the district, are over-sampled, comprising of 65% of respondents. With whites being over-sampled, the likelihood of their political leaning in this district is probably highly accurate. Unfortunately, this means that the Hispanic respondent results could be highly inaccurate. True, it is said that Hispanics in Miami-Dade County are hard to poll, but this is still a large discrepancy even for majority Cuban district.

Even though some polls do show some pretty big discrepancies, others seem to be more accurate. For example, in State Senate District 14 race between Republican William McBride and Democrat Darren Soto, whites were sampled more than Hispanics, but not drastically like in the Rivera-Garcia race. Therefore, when the poll is weighed, there is more of a likelihood of the poll being accurate.

Where the most reliable polls seem to be coming from is, not surprisingly, Pinellas County. These polls haven’t moved as drastically as the Hukill-Bruno poll. As far as the sampling, because much of Pinellas is white, the weighing of the polls by race is much easier. There have been very few issues with the Pinellas polls regarding the data compiled.

All other polls around the state fall in a category between the Bruno-Hukill/Garcia-Rivera races to the Pinellas County races. Therefore, each poll needs to be examined on an individual basis to determine if it is an accurate depiction of the electorate.

When I attended the University of Utah, I was taught by one of the most highly respected pollsters in the nation, Dan Jones. While he does most of his business in Utah nowadays, he used to poll nationwide. One day we were talking about polling in class, primarily about good and bad polling. Dr. Jones said that polls conducted by the media are usually very cheap polls. The only reason the media does them is so they can keep their readership interested. Therefore, they aren’t nearly as accurate as what the campaigns are seeing.

In that same conversation, we discussed bad polls being released. He truly thought that if a poll is a bad poll that it should not be released. To say that Dr. Jones has never released a bad poll would be inaccurate, he definitely has. This is where I think St. Pete Polls should truly do some soul searching. They should ask themselves “if this poll shows such a discrepancy compared to our previous poll, or if our sample is way off, should we release it?” I truly feel that St. Pete Polls could have kept some credibility by just keeping some of these polls in a folder on their desk instead of releasing them.

So, back to our question…are the St. Pete Polls any good? While they might not be there now, hopefully they will be. The concept and the idea of polling these smaller races is fascinating. If St. Pete Polls can work a little harder on putting out quality polls that show the true reflection of the electorate, then they can become a vital part of Florida politics. If they don’t work hard, and continue to release polls that can easily be picked apart as inaccurate, then there is a likelihood of them being the Florida version of Zogby International.


2 thoughts on “St. Pete Polls…are they any good?

  1. Thank you very much for your analysis of our polling, we are always evolving how we conduct our polls and how we present the results. Our goal is to be as open as possible and conduct polls on more localized races within Florida.

    We have received a lot of feedback on the last round of polls, some of it positive, some of it negative, and both of those from both sides of the political spectrum.

    The last round of polls was conducted using only active past voters from the last four years. If a person registered in the last 3 months, there is no chance they would be called. Also, using this restriction on the sample group skews both the sample and the overall demographics toward the Republican side on almost all races because voter turnout among Republicans is almost always higher than voter turnout among Democrats and No Party voters.

    On the SD8 race, we did give an explanation of how the vote changed in that race, it was given in the notes of the results report:

    “The results for the SD8 race between Hukill and Bruno have changed significantly since we polled this race last month. The reason for the change is the shift in party crossover support. In our September poll, both candidates saw their other-party support at about the same levels. In this new poll, Hukill’s support from Democrats is now triple the level of Republican support that Bruno has. During this poll, we also asked the respondents if they consider themselves to be a Republican, Democrat or Independent, and the self-identified party affiliation support in this race was at the same rate as the state supplied registered party support was.”

    In addition to that explanation, the October poll uses active voter demographics, which have Democrats at 39% and Republicans at 40% while the registered voter demographics that were used in the September poll show the breakdown as Democrats 38%, Republicans 36%. Hukill has an advantage in this more recent poll because the members of her party vote at a higher rate than the opposing party, and those active voter demographics are what we used in this October poll.

    As for CD26, the voting age population may be 69% Hispanic, but the active voter population self-identifying as Hispanic is only 55%. This effect is similar in other majority-minority districts as well, and it does have a significant effect on poll results when active voter demographics are used instead of total resident population demographics. I acknowledge that we do need to do a better job of polling Hispanics, which can be very challenging without knowing what language to first present to them.

    We have some very interesting plans for a new poll we are going to conduct next week, we are going to try something very different. I look forward to reading your analysis of it.

Leave a Reply

Please log in using one of these methods to post your comment: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s