Wednesday, October 19, 2011

Problems with Azimuth Polling Group

Originally published in Red Racing Horses.


Did you see the poll from Azimuth Polling Group claiming a tossup between Solicitor General Ted Cruz and Lt. Gov. David Dewhurst in the Republican senatorial primary in Texas? Let's take a look.

Azimuth Polling Group. 10/12 - 10/17/2011. MoE +/- 3%. 844 Republican voters in Texas. 
Ted Cruz- 32%
David Dewhurst- 31%
Tom Leppert- 8%
Lela Pittenger- 5%
Elizabeth Ames Jones- 4%
Undecided- 14%
(Also polled were Glenn Addison [3%], Andre Castanuela [1%], and Curtis Cleaver [0%])
  • Lela Pittenger, one of the unknowns, has 5% to Railroad Commissioner Elizabeth Ames Jones' 4%.
  • The methodology includes both telephone and email interviews. Email is notoriously dificult to use effectively in polling.
  • Other polls have shown Cruz, Lepeprt, and Jones to be significantly lagging behind Dewhurst in name recognition; Cruz here has to be suddenly well-known to poll so strongly (for comparison, 71% of voters had no opinion of Cruz in a September PPP poll).
  • Most importantly, Cruz is polling outrageously higher than usual here- compare to the same September PPP poll, where Cruz had 12% to Dewhurst's 41%).
Azimuth is new, but it has a few public polls under its belt. Check out this poll of New Hampshire from September, where Paul trails Romney 26/23 and Johnson boasts 8%. Or another poll from the end of May where Paul leads Perry in Texas.

The pollster claims their polls are different; they do not only use voter rolls but take the pulse of activists and party operatives more than anything else. From the pollster:
 ...our sample focused on highly involved Republican voters with clusters in the most politically active Republican areas of the state and using lists taken not only from voter rolls but also from  other sources likely to identify voters whose awareness of candidates and issues is substantially higher.  Basically, they  [PPP] polled voters and we polled more of the grassroots party activists who will influence those voters.
With our more informed and involved sample it is almost certain that pure name recognition played much less of a role in our poll.  This means that our results may be more indicative of future trends when voters have had time to become more informed about the candidates and issues, while the PPP poll may be a better snapshot of a broader segment of the voter base at this moment.
Fine, except that libertarians are doing excessively well. The polls would be useful if they accurately gauged support within the movers and shakers of the party (although I am deeply skeptical a poll could do that); instead, it overstates a certain faction's support.

Furthermore, we do not have a good methodology provided of how they select which activists to question. What qualifies one to be included in the pool of potential respondents? How are those contacts gathered [what lists in particular]? Is the population large enough to draw an adequate (ie representative) sample? All of this is disturbingly lacking.

This article explains the situation after the Texas presidential primary poll cited earlier.
The domain is registered to Dave Nalle of Austin, TX. But who is Dave Nalle? Dave Nalle is a Ron Paul supporter and  is Chairman of the Republican Liberty Caucus, a group that promotes Libertarianism within the Republican Party. In fact, if you go there, you’ll see an article written by Dave Nalle entitled “Don’t Believe the Hype.  Meet the Real Rick Perry.” The RLC also endorses just one current elected official that is running for president. Any guess as to who that is? Yup. Ron Paul. You can also find a great many posts by Nalle which stakes out a lot of which you can hear from Rep. Paul.
The article is wrong insofar that Nalle is a Johnson supporter (the RLC has informally backed Johnson this cycle, at least at the leadership level), but either way he works in the libertarian faction of the party. He also happens to endorse Ted Cruz' campaign, as I noted in August. Pittenger, who performed excessively well in this poll, is a very libertarian candidate. Cruz has also been working to round up libertarian support in the state, particularly with his endorsement from Senator Rand Paul. This leaves one with two potential conclusions:

1) Nalle essentially ran an internal poll for Cruz intended to get a strong result. A common trend in internal polling is surprising results where dark horses barely beat the incumbent (like this one from BJ Lawson in NC-04 in 2010).

2) Azimuth is consistently loading the deck with libertarians. The political circle the group deals with (particularly when gauging 'activist' support) is giving a false representation of who influences politics in the state. 

This does not even get into the fact they use email to poll voters. Given the flaws above, I am skeptical the email interviews were conducted in a sufficiently rigorous manner (a counter-example is YouGov, an outfit that has used email effectively).

In sum, Azimuth Polling Group has significant difficulties with its polling. Its results should be treated with skepticism.

No comments:

Post a Comment