Evaluating the polls

This year we got a set of polls for various Democratic primary races from the UH Hobby Center. Because of the nature of some of the races and the fact that we don’t usually get this kind of polling data, these numbers got tossed around a lot in the early voting period. I thought this was as good a time to go back and compare the poll numbers to the actual election results, so we can see how well they did. If we’re going to get this kind of polling in the future, it would be nice to know what kind of confidence level to have in them.

The first batch included the US Senate race (for Harris County only) and the four countywide offices. Here’s the breakdown:

US Senate poll numbers – Colin Allred 66%, Roland Gutierrez 7%
US Senate actual numbers – Colin Allred 64%, Roland Gutierrez 10% (Harris County only)

District Attorney poll numbers – Sean Teare 59%, Kim Ogg 21%
District Attorney actual numbers – Sean Teare 75%, Kim Ogg 25%

County Attorney poll numbers – Christian Menefee 41%, UA Lewis 7%
County Attorney actual numbers – Christian Menefee 70%, UA Lewis 30%

Sheriff poll numbers – Ed Gonzalez 63%, next best candidate 5%
Sheriff actual numbers – Ed Gonzalez 67%, next best candidate 16%

Tax Assessor poll numbers – Annette Ramirez 12%, next three candidates between 5 and 7%
Tax Assessor actual numbers – Annette Ramirez 41%, next three candidates between 15 and 18%

Gotta say, overall this is a strong performance. They underestimated Sean Teare, but had him well over 50% and with a dominant margin. Christian Menefee probably had less name recognition than some other candidates otherwise he might have scored higher, but again they got the basic fact of a large lead correct. I’ll give them a pass on the Tax Assessor race as name recognition there was low, but again they identified the leader and found a tight bunch after that. Really, this is close to a grade A performance.

The second batch had Congressional and legislative races, and it was more chaotic.

CD07 poll numbers – Rep. Lizzie Fletcher 78%, Pervez Agwan 11%
CD07 actual numbers – Rep. Lizzie Fletcher 82%, Pervez Agwan 18% (Harris County only)

CD18 poll numbers – Rep. Sheila Jackson Lee 43%, Amanda Edwards 38%, Robert Slater 3%
CD18 actual numbers – Rep. Sheila Jackson Lee 60%, Amanda Edwards 37%, Robert Slater 3%

SD15 poll numbers – Rep. Jarvis Johnson 18%, Molly Cook 18%, Todd Litton 14%, others less than 10%
SD15 actual numbers – Rep. Jarvis Johnson 36%, Molly Cook 21%, Todd Litton 16%, others less than 11%

HD139 poll numbers – Rosalind Caesar 12%, Angie Thibodeaux 10%, Charlene Ward Johnson 8%, Mo Jenkins 4%, Jerry Ford 4%
HD139 actual numbers – Angie Thibodeaux 33%, Charlene Ward Johnson 24%, Mo Jenkins 18%, Rosalind Caesar 17%, Jerry Ford 8%

HD142 poll numbers – Rep. Harold Dutton 38%, Danny Norris 7%, Joyce Marie Chatman 6%, Clint Dan Horn 6%
HD142 actual numbers – Rep. Harold Dutton 61%, Danny Norris 19%, Joyce Marie Chatman 15%, Clint Dan Horn 5%

HD146 poll numbers – Rep. Shawn Thierry 40%, Lauren Ashley Simmons 16%, Ashton Woods 4%
HD146 actual numbers – Lauren Ashley Simmons 49.5%, Rep. Shawn Thierry 45.4%, Ashton Woods 6%

Not quite as impressive. The CD07 poll is a little confusing because that district includes a piece of Fort Bend County as well, but the poll was only of Harris County voters. Harris is most of CD07 but Fort Bend was much friendlier to Agwan, who took 46% of the vote there. That made the actual final result 74-26 for Rep. Fletcher, but the Harris portion was accurate enough. That was the best result, with HD142 in second place – they got the big lead for Rep. Dutton, just not with a big enough number for him. But the basic idea was right, so close enough.

The rest? Meh. Their result in CD18 promoted a narrative that CD18 was close, maybe even too close to call, when it was not. My guess is they didn’t have a good model of voter ages – it was an old electorate, and the older voters strongly favored Rep. Jackson Lee.

SD15 did pick the top two contestants, but halved Rep. Johnson’s support, which in this context seems like a pretty big miss. The poll would make you think it was neck and neck between Johnson and Molly Cook with Todd Litton right behind, but Johnson had a substantial lead over Cook. Not sure if they didn’t have a good model of Black voters or if they just got a sufficiently significant no response rate to skew the final result.

I did say when I posted about these polls that the State House races were likely too unreliable due to small samples to really mean anything. The results speak for themselves, as they completely whiffed on the order (and thus the runoff participants) in HD139, and had Rep. Thierry up by a lot in HD146 when in fact she barely missed losing outright. It was close between her and Lauren Ashley Simmons, but they had the leader wrong. Again, I’m not sure what caused that, but those races are the toughest to call. I’m glad they did these, but this is the reminder to take them with the largest grains of salt.

So I’d give the first batch an A minus, and the second batch a C, maybe C minus. Kudos for the undertaking, I hope we see more of it, just remember to use individual poll results with great care and more than a little doubt.

Related Posts:

This entry was posted in Election 2024 and tagged , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , . Bookmark the permalink.