Washington Post Poll vs Catania Internal Poll

post reply Forum Index » Political Discussion

1 2 ► Go

Go to page...

09/25/2014 10:56

Subject: Washington Post Poll vs Catania Internal Poll

quote

jsteele
Site Admin
Joined: 11/12/2007 23:38
Messages: 50050
Offline

Is it possible that the NBC4/Washington Post/Marist poll of the mayoral election showing Muriel Bowser leading David Catania by 17 points is wrong? Most of us would probably consider that to be unlikely, but the Catania campaign is making a strong argument to the contrary. The Catania argument -- which has some support -- is that the NBC4/Washington Post/Marist poll artificially inflates Bowser's support as a result of a wrongly-modeled poll sample.

The Marist poll results followed a Catania internal poll showing the two candidates separated by just 3 points and within the margin of error. While many -- including me -- are inclined to discount internal polling results, I noticed that the Catania campaign was still standing by its polling numbers. Moreover, in a Washington City Paper article, Will Sommer linked to this article:

http://www.slate.com/articles/news_and_politics/victory_lab/2011/12/_likely_voters_lie_why_private_campaign_polls_get_such_different_results_from_public_media_polls_.html

which gives credence to the Catania campaign's position.

The Slate article was written by Sasha Issenberg who has gained quite a reputation for understanding how campaigns use data. He is the author of "The Victory Lab: The Secret Science of Winning Campaigns". So, his thoughts on this issue are credible. As Issenberg explains it, "a large methodological gap has now opened between the surveys candidates and their strategists see and the ones you do." The "methodological gap" involves the "likely-voter screen" used by pollsters to select likely voters for poll samples. Research has shown that poll responders are not particularly accurate in reporting their voter-registration status or likelihood to vote. While discrepancies in these two matters might not be all that important to media outlets, they are crucial to campaign strategists. Hence, campaign pollsters no longer rely on self-reporting. As Issenberg says, "campaign pollsters usually begin with a list of registered voters available from state election authorities." Hence, no need to rely on respondents to accurately answer whether they are registered. Similarly, voting histories are used to determine which respondents are likely voters rather than relying on self-reporting.

I talked to Catania Campaign Manager Ben Young about the campaign's internal poll in light of the Issenberg article. He stressed the registered and likely voter methodological differences between the polls, but explained several other differences as well. All polls are based on a sample of voters. How that sample is modeled to reflect who will actually show up at the polls is crucial to the validity of a poll's findings. The NBC4/Washington Post/Marist poll simply models its sample based on the current DC population. In fact, in his article, Issenberg quotes Washington Post polling manager Peyton Craighill as saying, "The main thing is we want to know what everybody thinks, not just voters." As a result, according to Young, the NBC4/Washington Post/Marist poll sample does not accurately reflect the demographic makeup of those who historically vote in DC local elections.

According to Young the population of those who historically vote in DC local elections skews older and female. I don't know why females are more likely to vote than males and I didn't delve in to it, but apparently that's what the data shows. The bias toward older voters is easier to understand. Young voters don't tend to follow local politics closely, are therefore not as well informed about local politics, and turn out to vote in lower numbers than older voters. As a result, Young believes the NBC4/Washington Post/Marist poll over-sampled young and male voters.

It is the over-sampling of young voters that Young emphasized as being problematic to the NBC4/Washington Post/Marist poll's accuracy. Not only are such poll respondents -- based on historic patterns -- less likely to actually vote, but they are also less likely than older voters to be informed about the campaign. As such, they may or may not recognize the names of the candidates when asked by a pollster, but they would certainly recognize party affiliations. In Democratic Party-dominated DC, an uninformed voter is most likely to select the Democrat. This results in artificial inflation of the Democrat's numbers. Young says this is supported by the fact that the NBC4/Washington Post/Marist poll and the Catania internal poll found similar numbers for Catania and Carol Schwartz. The NBC4/Washington Post/Marist poll had Catania at 26 and Schwartz at 16. The Catania poll had Catania at 25 and Schwartz at 14. In contrast, the NBC4/Washington Post/Marist poll showed Bowser at 43 compared to 28 in the Catania poll. Young contends that difference in favor of Bowser is artificial. Moreover, Young says that he has heard that the Bowser campaign's internal polls also show a very close race. He suggests that question could be settled if the Bowser campaign would release those results.

We may never know which poll was more accurate. The findings may be superseded by events and polls can have a self-fulfilling effect. But, there seems to be enough substance to the Catania campaign's argument that we may want to take the NBC4/Washington Post/Marist poll with a bit more salt than we might ordinarily do.

Disclaimer: I am a Catania supporter and I interviewed Catania Campaign Manager Ben Young for this article. However, I chose to write this article on my own initiative and have not cleared the article with the campaign in advance. This article reflects my own thinking based on the sources I credited. All errors are my own (even spelling and grammar mistakes which I can't even justifiably blame on autocorrect).

Edit: I correct one typo. Probably a few more remain.

profile p.m.