Biden Pollster Commits Massive Self-Own, Admits to Poll-Herding on Twitter
Photo courtesy of Getty
Earlier this week, you may have seen a Monmouth Poll that looked very different from other Democratic primary polls we’ve seen lately, in that Bernie Sanders and Elizabeth Warren led presumptive frontrunner Joe Biden:
NATIONAL POLL: Early #2020Dem preference:
20% @BernieSanders (? 6 pts from June)
20% @EWarren (? 5)
19% @JoeBiden (? 13)8% @KamalaHarris
4% @CoryBooker
4% @PeteButtigieg
3% @AndrewYang
2% @JulianCastro
2% @BetoORourke
2% @MarWilliamsonhttps://t.co/AQcJPs8d22— MonmouthPoll (@MonmouthPoll) August 26, 2019
That created a big splash, not least among Biden supporters and establishment Democrats who really, really didn’t want to see those results. The backlash was so loud that Monmouth was forced to release a statement essentially saying “these things happen.”
Statement from poll director @PollsterPatrick
regarding this week’s poll: [date corrected] pic.twitter.com/Xelpw1pt9j— MonmouthPoll (@MonmouthPoll) August 28, 2019
“In the end, we must put out the numbers we have,” Monmouth concluded, which is quite sensible. After all, you can’t skew the numbers, and you can’t simply throw out polls you don’t like…right?
Not so fast, says John Anzalone, a partner at ALG Polling, which has polled for Obama, Hillary Clinton, and—per his Twitter bio—Joe Biden. For Anzalone, the statement wasn’t good enough. This one’s screenshotted for the inevitable moment when he deletes it:
“When our firm comes out of the field and we believe we have an outlier we shit can the numbers and redo the poll at our expense. Period,” Anzalone said.
Ummm….
That’s called poll-herding! A number of respondents tried to point this out:
always a good idea to keep resampling until it says what you expect it to, that’s why the clients pay you the big bucks
— Matt Bruenig (@MattBruenig) August 28, 2019
man and people wonder why the polls were so off in 2016
it’s just confirmation bias, trying to find justification for a preexisting viewpoint
— Jesse, Your Squadmate (@loquaciusSA) August 28, 2019
One might interpret this as you saying that when polls don’t give you the desired result, you redo them until they do
— Eoin Higgins (@EoinHiggins_) August 28, 2019
Yiiiiiiiiiiiiiikes
— brendon bouzard (@brendonbouzard) August 28, 2019
“When the data doesn’t match my hypothesis, there is a problem with the data.”
— Derek (@ricketyoldshack) August 28, 2019
This is the most public admission of pollster herding I’ve ever seen…@gelliottmorris@NateSilver538@Nate_Cohn@PollsAndVotes@AAPOR
— Dan Rosenheck (@DanRosenheck) August 28, 2019
Now, Anzalone works for a company that polls for individual clients, so this might not exactly be “unethical,” since his results don’t go public. But it’s certainly stupid, and if I were Joe Biden or any of his other clients, I’d be distinctly worried about trusting any results he gave me. As several users pointed out, he polled for Hillary Clinton in 2016, and many of the polls associated with that race were off. Is Anzalone saying that it would have been okay to discard the polls he didn’t like so that he only presented Clinton with good news?
That aside, the idea that Monmouth should discard any polls that don’t conform to other polls is madness, and presents a slippery slope argument that would compel polling outlets to make ideological decisions, rather than preserving the integrity of their data. Polls are meant to reflect reality, not to influence it.