Thursday, September 22, 2016

UH releases poll with 10-point Clinton lead in Harris County

If you recall, I made fun of the lady at the beauty shop over a week ago for advancing this exact rumor.  So she gets to feed me some crow.

Poll results released today by the University of Houston Hobby School of Public Affairs show that Hillary Clinton has a 10-point lead over Donald Trump among registered voters in Harris County, the largest county in Texas and third largest in the nation. Clinton leads Trump 42 percent to 32 percent, with nine percent supporting Libertarian Party candidate Gary Johnson, two percent backing the Green Party’s Jill Stein, and 15 percent undecided.

If the numbers hold, it would represent the widest margin of victory for a Democratic presidential nominee in Harris County since 1964, the year Lyndon B. Johnson was elected president. He received 59.5 percent of the Harris County vote.

Clinton’s lead narrows to only four points, 43 percent to 39 percent, among voters who say they are extremely likely to cast a ballot this November.

The "extremely likely" screen, with the four-point lead, is probably the closest to accurate, FWIW.  Even if ten is the margin, I simply don't think it can hold all the way to Election Day, but it's more than enough to give blue partisan hopes a big boost.

The rest of the poll's results, however, won't (bold is mine).

The UH Hobby School poll finds no evidence of national Republican concern that Trump’s unpopularity within the party will negatively effect down ballot races. In the race for Harris County District Attorney, among the voters extremely likely to cast a ballot this fall, incumbent Republican Devon Anderson narrowly bested Democrat Kim Ogg, 30 percent to 29 percent, while incumbent Republican Ron Hickman led Democrat Ed Gonzalez 36 percent to 30 percent in the contest for sheriff. A plurality of voters is unsure about their preference for district attorney (47 percent) and sheriff (36 percent).

A year ago during the mayor's race, UH returned a similarly strange polling result:  Sylvester Turner with a ten-point lead, Bill King and Adrian Garcia tied for second, Chris Bell a close fourth, and a massive quantity of undecideds.  The November general election results were wildly different, with Turner 32%, King 25, Garcia 17, Ben Hall 9, and Bell 7.

I pooh-poohed their poll then, and I'm pooh-pooh-ing these two county results now.

Update II (9/23): Charles says I'm confusing my Cougar polls here.  Fair enough; I sit corrected.  I still pooh on both.

For one thing, that's an enormous number of GOP split tickets.  Even with the high unsure/undecideds, something appears to be very off if Hickman is leading Gonzalez by six while Clinton leads Trump by ten.  I would have expected the opposite in the sheriff's race to be true, in fact, irrespective of the presidential.

This is a data point, but an awfully strange one.

Update (9/23): The mighty Kuffner has weighed in, and disapproves of the poll to some greater degree than I do.   No issues with his take.  But I thought about these numbers awhile overnight, and did some back-of-the-envelope math: if 100% of the Harris County electorate represents 1,000,000 voters (Charles will probably have a guess at turnout later and it will be higher than this; my number is extremely conservative) then by virtue of UH's poll, Clinton is ahead of Trump by 100,000 votes, and Ron Hickman leads Ed Gonzalez in the race for sheriff by 60K.

This doesn't seem plausible in a sampling that is 50% D, 45% R, and 5% I (page 5).  It would suggest there is a very large number of Democrats in Harris County -- joined by Republicans, for that matter --  who would be splitting their tickets Clinton and Hickman.  (Significantly less so Clinton and Anderson in the DA's race; everybody understands that one will be very close all the way to Election Night).

Could there really be tens of thousands of blue and red partisans in the county voting Hillary for president and Ron Hickman for sheriff?  I'm just not buying that.

1 comment:

Gadfly said...

Maybe this is a spin on the old Bradley effect. A lot of "good" Republicans, even to anonymous pollsters, are ashamed to admit voting for him?