Early poll in the Houston Mayor’s Race warrants a second look

Last week I received an analysis prepared by the Carreno Group of a survey of Houston voters done by the National Hispanic Professional Organization (NHPO), which describes itself as a “networking, non-profit, membership-based organization.” The headline on the summary was “Parker Has Commanding Lead in the 2009 Mayoral Election.” Since this was the first survey I have seen not associated with a particular mayoral campaign, it is a good idea to briefly discuss polling in the upcoming city elections, and apply some of the general principles to this particular voter survey.

Polls have been a mainstay of election campaigns for more than a half century. Candidates and political parties use their own surveys as essential planning tools, news media commission independent polls to keep track of the “horse race” and explore issues that matter to voters, and private groups like NHPO sometimes sponsor election surveys for various reasons. In high profile contests like the 2007-2008 presidential race, 15 or more survey outfits were releasing their findings every week or two. With so much data available, a good strategy for the consumer of polls in such cases is simply to average the results to get an idea of how the race is moving along. If there is a problem with a specific survey or polling operation, averaging largely addresses that possibility. Unfortunately, in cases like the Houston mayoral election, we have relatively few polls publicly available, so consumers need to look carefully at each data set.

What should people be looking for in such cases? First, who did the poll? Do they have a track record, and is that record good, fair, or poor? Second, are the technical aspects of the survey fully described? When was the survey done, using what method of contacting voters, and how many surveys were completed with what error margin at a given confidence level. Is the questionnaire available for review so that one can see exactly how questions were worded and the order in which they were asked? And, are we given the basic demographic breakdowns of persons interviewed? How many whites, blacks, and Latinos are in the sample? What was the partisan composition of respondents? Their education level, and so on? If, this information is available, it gives one more confidence in the data presented, although all consumers of singular polls should keep in mind that even the best are just snapshots of fast changing landscapes.

Turning back to the NHPO survey, we note it measures up in some areas. We are given the dates when the data was collected (April 14 – May 1), via the Internet, with ZIP code addresses used to delete non-City of Houston respondents. The summary of responses does not have the exact questions used, but we are given some basic demographic information about the 736 individuals included in the mayor poll results, which showed City Controller Annise Parker with 48.8%, followed by former Houston City Attorney Gene Locke at 29.6%, City Council Member Peter Brown at 13.9%, and retired military officer Roy Morales at 7.7%. After looking over the summary, there are clearly major problems with this particular survey aside from the fact that NHPO has no track record in this area and Internet-based surveys raise significant issues in diverse communities where many voters cannot be contacted using this methodology. Let me zero in on two very evident problems with the poll.

First, the demographics in the NHPO sample bear little resemblance to the actual electorate in Houston mayoral races. In competitive mayor elections like 2009, recent history suggests the voters will be about 48% Anglo or non-Hispanic whites, 32% black, 15% Latino, and five percent Asian-Other. The NHPO demographic breakdown of survey takers is OK on Anglos (43.1%) and Asian/Others (5%), but wildly off the mark on Hispanics (41.4%) and blacks (just 10.5%). Since we have a credible black candidate, Gene Locke, in the race, this under-representation of African Americans seriously understates his likely support level.

Second, the data summary we are given does not add up. We are told that Anglo voters are breaking 44.6% for Parker, followed by Roy Morales at 28.4%, with Locke and Brown third and fourth. We are also informed, as noted, that Anglos constituted 43.1% of the total sample. Assuming these data are correct, we should be able to calculate for a candidate like Roy Morales what percentage of the total vote for mayor is made up of Anglos supporting him by multiplying the Anglo vote share (43.1%) by his share percentage of same (28.4%). When we do that, we get 12.2% as the result. But, when we look at the total mayor vote breakdown, Morales gets just 7.7% when his Anglo support alone is well above this percentage. This does not make any sense, and when a poll’s internal numbers do not add up, readers are well advised to look elsewhere for readings on how the mayor’s race is actually shaping up.

By Richard Murray

Comments

[Greg Wythe]

The poll was technically adminstered by SurveyMonkey.com. See here for the overall results:

http://www.surveymonkey.com/sr.aspx?sm=o145JZ5OdtNN3f6nZkLw72U8diuq1CLMtFB82Y_2fio84_3d

The means of distribution for the link were, basically: whoever the group emailed a link to, whoever the candidates emailed links to, and whoever got it from various email lists that it was distributed through.

This doesn't even compare to other polls conducted online in terms of reliability.

I'm informed that the ethnic crosstabs were calculated according to the entire universe of respondants, not just the Houston ones. That would definitely explain why the numbers don't add up. But it raises a bigger question of why anyone would want to release that information with their name attached to it.

[out of town]

That survey monkey poll was what this was? I was sent that poll link numerous times by the Parker campaign. I don't live in Houston, but I made up a Houston zip code to use. I mostly took the survey to see what it was about. I made up answers to nearly every question including who I would support. I treated the poll the way it deserved to be treated - as something that was silly.

I never imagined the Parker campaign would tout this as a real poll, since it was clearly very, very unscientific.

[SAM]

I also received this survey via Annise Parker emails... I can recall two different emails I received from her with this survey, encouraging that “I vote for Annise!”

[Craig Adams]

Isn’t it ironic that the resident of NHPO, Ben Mendez, is a Parker supporter? ...there’s another flaw with this “poll”. Sad that Annise attached her name to yet another flawed survey!

[Bob]

Hey Sam and Out of Town--

So, could you tell us whether you're with Gene, Peter, or Roy? I'd bet you're with Peter "Pedro" Brown or Gene given how poor their showings were. Produce the emails you claim or quit lying. Of course, Roy's team was caught trying to game the thing, too. It's an online poll! Everyone knows it's not like a real poll!

And no candidate has pushed this thing as legitimate. Well, except Gene pushing this blog post, of course...That's really too bad, though. Murry already said Gene needs 95%+ of the black vote to be viable, and this one only says he has 68% and that's worse than Turner did. Oops. Heckuva job, Team Locke. Course, as the song says, "I've come to expect that from you."

And, Roy...well, bless his heart...

And last, but not least, Greg. Does your boss and Bill White's Senate campaign know you're inserting yourself with unfounded claims ("I've been informed") into the mayor's race? Be Careful; only trouble lies down that road.

[AR]

Good job clarifying the poll Dr. Murray! Maybe some people should stick to what they know and leave the analysis of political polls to the pros. LOL!!!