Today’s PPP poll is released with highly questionable data.

Today, PPP release a poll looking at the Florida governor’s race and found that in a hypothetical match up, Charlie Crist would defeat Rick Scott by a margin of 52% to 40%. While this numbers seem kosher on its face value, there are very serious questions regarding the results which might make this poll obsolete. Before we go into the large errors, let’s look at some other aspects first.

1. Poll conducted of “voters”, not “likely voters”  – In midterm elections, Democratic voter turn out, as well as liberal voter turnout, usually suffers. Therefore, those who are likely to vote in the 2014 election will more than likely be more conservative than the electorate in general, who will be more liberal.

2. Liberals over-sampled – In a typical mid-term election, only about 19% of the electorate identifies themselves as being liberal. This number is consistent in exit polls in the 2002, 2006 and 2010 elections. In this sample, 30% of those who responded identified themselves as either “very liberal” or “somewhat liberal”. While this might be a correct calculation of “voters” in Florida, it is, historically, 11% more than exit polls indicate, which could indicate the discrepancy between “voters” and “likely voters”

3. Older voters under-sampled  – In 2010, voters over the age of 65 were 35% of the total electorate. In 2006, those 60 and older were 34% of the electorate. In the PPP poll, only 26% of those sampled were over the age of 65. This 9% drop  takes away from Rick Scott. In a hypothetical match up between Scott and Crist among those 65 and older, Crist only wins by 2%, well within the +/-4.4 margin of error of the sample. Therefore, historically, the 65+ numbers should be much higher.

4. Democrats over-sampled – Historically, Republicans usually are tied or ahead of Democrats in midterms elections as far as the composition of voters that turn out. In a “Democratic wave year”, Republicans were 40% of the electorate while Democrats were 36%. Oddly, in a “Republican wave year”, both Democrats and Republicans were at 36% each of the total electorate. In this sample, Democrats have a 5% advantage over Republicans.

So, these are just some of the small “hiccups” that I noticed in the poll. While the poll itself could be an accurate reflection of the state overall, it does not seem to be an accurate reflection of “likely voters”. So, let’s move to the big questions:

1. Rich and Iorio hated by liberals  – Yes, doesn’t it seem odd? Here we have two Democrats, with Senator Nan Rich being one of the most liberal Democrats in the Florida Senate, and people who identify themselves as “very liberal” hate her? In the PPP poll, of those who identify themselves as “very liberal” only 6% approve of Nan Rich, while 33% disapprove of Senator Rich. For Pam Iorio, it isn’t much better. With the same group, 20% approved of her while 41% disapprove. Again, seems extremely incorrect.

2. Rich and Iorio hated by Democrats  – Much like the liberal flaw, Democrats seem to dislike these two as well, which, again, seems odd. Of Democrats, 10% approve of Nan Rich while 20% disapprove of her. As for Pam Iorio, 16% of Democrats approve of her while 27% disapprove. What is even more odd is that Republicans approve of Pam Iorio more than Democrats. In addition to that, Pam Iorio has a higher approval rating among Republicans (21%), than disapproval rating (20%), even though it is in the margin of error. And while the poll does identify that a large number of those surveyed were “not sure” about either candidate, both the ideology and party totals are very much out of line.

These two highly questionable result should have led the PPP to reexamine their survey and data, but they decided to run with it. So, where did the error happen? If we are to look at the question order as well as the question wording, the PPP made a possible error. Here is the question order of the first five questions:

Q1 Do you approve or disapprove of Governor Rick Scott’s job performance?

Approve …………….. .33%
Disapprove…………. .57%
Not sure …………….. .10%

Q2 Do you have a favorable or unfavorable opinion of Charlie Crist?

Favorable……………………………………………….. 46%
Unfavorable ……………………………………………. 43%
Not sure …………………………………………………. 11%

Q3 Do you have a favorable or unfavorable opinion of Pam Iorio?

Favorable……………………………………………….. 19%
Unfavorable ……………………………………………. 23%
Not sure …………………………………………………. 58%

Q4 Do you have a favorable or unfavorable opinion of Nan Rich?

Favorable……………………………………………….. 8%
Unfavorable ……………………………………………. 18%
Not sure …………………………………………………. 74%

Q5 Do you have a favorable or unfavorable opinion of Alex Sink?

Favorable……………………………………………….. 31%
Unfavorable ……………………………………………. 30%
Not sure …………………………………………………. 39%

In the case of Rick Scott, Alex Sink and Charlie Crist, most voters that were surveyed probably know their party identification. In the case of Pam Iorio and Nan Rich, this might not be the case. Because the Nan Rich and Pam Iorio questions were followed up by the current Republican governor and a former Republican governor, those being surveyed might have thought that both Iorio and Rich were Republicans as well. This would explain why the Democrat/Republican as well as the Liberal/Conservative numbers appear inaccurate. If the wording of the question included the party identification of all the candidates, then this poll might have come to conclusions that more accurately indicate the voters’ true feelings.

The next question that needs to be asked is if the ordering of the approval rating questions, with lack of party identification, led to a socially desirable answer on the follow match up polls against Rick Scott. After a liberal or Democratic voter said that they “disapproved” of Iorio and Rich, which at the same time they “approved” Crist, could this lead to more “not sure” responses in the match up polls? While only 8% are unsure in a Crist/Scott match up, 19% were unsure in an Iorio/Scott match up and 21% unsure in a Rich/Scott match up. Therefore, a respondent who might feel embarrassed that they didn’t know that Rich or Iorio were Democrats might say “not sure” in the match up poll as a socially desirable answer to their ignorance of party identification of these two candidates.

These highly questionable results are better news for Iorio, Rich and, possibly, Sink. As of right now, 90% of those polled had an opinion of Charlie Crist. Therefore, these numbers might represent the highest that Crist will go in an election. Polls coming out in the next few months should indicate what kind of movement there might be once the spotlight is taken off of Crist.

On the flip side, 58% of voters don’t know who Pam Iorio is and 74% don’t know who Nan Rich is. And, if we include the possible error of question order and party identification when looking at approval ratings which is mentioned above, the amount of voters that do not know these candidates could be even higher. This means that in both Iorio and Rich could see higher numbers as time passes if they are able to get their names out there. While more people know who Alex Sink is, 39% of the people basically had no opinion of her. Therefore, her numbers can go up as well.

What does this poll mean?  Of course, we can’t take anything away from early polls. Even six days before the 2010 Republican primary for governor, Rick Scott only had a 30% approval rating while losing to Bill McCollum in the same Quinnipiac poll 44% to 35%. And, as we know, Rick Scott won by 3%. So, in a nutshell, match up polls as this point are irrelevant.

But if there were an overall takeaway message from this poll, it is that both Pam Iorio and Nan Rich need to work on name identification. While a large majority of people still not knowing who they are, as well as possibly mixing them up with Republicans, name ID should be their main goal over the next few months.

As for the Scott vs. Crist match up, right now Crist is in the lead, but that can change. With both candidates having well under a 50% approval rating in the polls, the match up between the two men seems to be more of a battle of “the lesser of two evils” than people voting in support of either candidate. When this is the case, usually Democratic voter turnout suffers, such as the case of Dukakis vs. Bush in 1988 or Kerry vs. Bush in 2004. Again, these are just early numbers. As soon as Crist announces that he is running for governor and the primary campaign starts to go in full swing, then we might see some more solid numbers.

Basically, even with the obvious errors by PPP, the election in Florida is still very fluid. Even with this being said, this poll isn’t representative of what the likely voter turnout will be in the 2014 election. Exit poll numbers from previous elections show that liberals and Democrats are over-sampled in this poll while those who are over the age of 65 are under-sampled. So, while this poll might be somewhat accurate of the feeling of the state in general, it will more than likely be quite different in showing the opinions of those who will actually show up to the polls.

For exit poll numbers referred to in this article, click here for 2006 and click here for 2010.

Advertisements

6 thoughts on “Today’s PPP poll is released with highly questionable data.

  1. 1.) This early in an election season they don’t tend to ask “likely voters” because there’s not enough data on who these likely voters are and it’s too far out for people to be able to predict with any accuracy whether or not they’ll vote. However, you’re spot on that that benefits Democrats

    2 and 4.) This was a big issue among conservatives in the last campaign, that liberals and democrats were over sampled. But poling companies don’t control for ideology of affiliation because that’s a moving variable that can change depending on the mood of the state. If you control for that you get that “unskewedpolls.com” baloney.

    PPP has always been very good and very accurate, I would be hesitant to leap to the conclusion that they put out a flawed study. No doubt question order matters, but my guess is they randomize their question delivery so that it won’t be an issue. I’m not sure though.

    Good article, maybe you should keep writing about Florida Politics 😉

  2. Crist’s negatives are sure to rise during the campaign as issues of corruption and his willingness to sacrifice principle for political expediency are reintroduced to the voters by the GOP.

    Scott losess to Iorio or Sink but beats Crist or Rich.

  3. In larger races, some do poll likely voters early, especially in high profile races. Some of the smaller races like Congress don’t. I would consider this a high profile race, and asking someone if they are a likely voter is a back-end question anyway, so it shouldn’t matter.

    As far as comparing me to the “unskewed poll guy”, I am far from it. If that was the case, I would have predicted Mitt Romney to win as well. But I did predict the presidential race 100% correct, even before your beloved Nate Silver 🙂 And, unlike Nate Silver, I gave you the SPECIFIC reason why Obama would win Florida, which proved to be the exact case (Nov 4th quote: “Oh, and a 4th, I don’t see Romney winning Hillsborough, especially with good early voting numbers for the Democrats. Hillsborough is Romney’s Ohio in Florida. If he doesn’t win Hillsborough, he doesn’t win the state.”) So there 😛

    As I said in this poll, these numbers might be accurate at telling us the view of the electorate in general (except for those “very liberal” and “liberal” numbers), and I really don’t doubt that. That unskewed polls guy just lowered the liberal bias on his polls because he just wanted liberal to be less than what he thought and wanted the numbers to say what he wanted them to say because he didn’t like the results.

    On the other hand, I am taking historical data and applying it to the numbers (even using Cronbach’s Alpha can show its validity of historical turnout numbers as consistent). The poll could be accurate regarding voter preference of all voters, but Florida does have the worst mid-term turnout of any state. So, applying historically accurate and consistent numbers to these isn’t skewing, but taking the current data and applying it to a hypothetical November 2014 election.

    As for PPP, I think they are the best, which is why I am surprised that they didn’t catch the word ordering problem. That is, as I see it, the only survey problem that exists. That is why I really didn’t call the first four concerns “errors”. I mostly put them there so that people don’t jump to the conclusion that these are Election Day results. It is PPP’s job to interpret the electorate in regards to their universe. It isn’t their job to apply those numbers to historical turnout data, and I wouldn’t ask them to do that. But Florida polls have historically been inaccurate in mid-term election (and by a large extent in most part).

  4. Dave, I thought that you are “… no longer going to be involved with writing opinion pieces about Florida or American politics.” I’m confused

  5. No, I am just doing analytical information, not opinion pieces. This is just an analytical look at the polls and, I guess, my “opinion” of that poll. I’m out of the “Charlie Crist sucks because…” business.

Leave a Reply

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

WordPress.com Logo

You are commenting using your WordPress.com 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