Wednesday, May 30, 2018

Why would anyone bother talking about polls ever again after the Irish abortion vote?


Last Thursday the Wall Street Journal ran an article about the Irish abortion vote, saying that it was expected to be close and included these polling results in support of that statement.



Per the footnotes, the poll was fielded by the Irish Times and Ipsos MRBI, both reputable companies.  The sample size was 1,200, a fairly decent size that should have yielded a statistically significant result.  And the margin of error was 3%.  So at best, support for “Yes” should have been 47%. (Hannon & Rocca, 208)

And yet, when the votes were counted 66.4% voted “Yes.” (McDonald, Graham-Harrison & Baker, 2018)

So what went wrong here?

For starters, the poll was taken on May 16, a week and a half before the actual vote.  If all the undecideds voted “Yes,” that brings the total to 61% - 64% if you include the margin of error.  That’s closer.  But what are the odds that all the undecideds voted “Yes?”

More likely the problem was the same thing that went wrong with the polls for Brexit and the U.S. Presidential election.  Our polling techniques are broken. 
 


They rely on companies being able to get a reliable sample from calling people on landlines.  But these days when you rely on landlines, you get a disproportionate amount of older, poorer, less educated voters.  They are anything but a random sample.

Adding online surveys to the mix can help capture younger voters, but here too, the people who respond are not necessarily representative of the whole.  And, deciding how to merge the data from the different sources presents problems as well. (Knutson, 2016)

So, where does that leave us? 

When looking at research, pay attention to sample size, sponsoring companies, and dates.  But even if all those criteria look good, it’s probably better not to rely on poll results anymore.  They could be just more Fake News.



Hannon, P. & Rocca, F. (2018, May 23)  Catholic Church Takes a Background Role Ahead of Ireland’s Abortion Vote.  wsj.com.  Retrieved May 29, 2018, from

McDonald, H., Graham-Harrison, E., Baker, S. (2018, May 26)  Ireland votes by landslide to legalise abortion.  theguardian.com.  Retrieved May 29, 2018, from

Knutson, R. (2016, October 27)  Pollsters Scramble As Fewer People Take Their Phone Calls. wsj.com  Retrieved May 29, 2018, from https://www.wsj.com/articles/pollsters-scramble-as-fewer-people-take-their-phone-calls-1477577832
 


No comments: