At a press conference in Washington on Tuesday, June 3, Frank Newport released the results of an extensive review of Gallup’s poor polling performance in 2012, when it consistently over-estimated Romney’s electoral support.
The extent to which Gallup’s results were an outlier during the campaign was illustrated by Mark Blumenthal with this graph on Huffington Post:
Among the many factors the review team investigated, four seemed to provide the explanation:
- a faulty likely voter model, which tries to predict which respondents will actually vote;
- mis-representation of voters by geographical region, which led to an under-representation of Democrats;
- over-representation of non-minority voters, caused largely by a faulty question designed to measure race; and
- a faulty sampling design, which used listed telephone numbers as a basis for drawing landline numbers, rather than a random-digit-dialing process that ensures the inclusion of both listed and non-listed numbers.
While Gallup appears to be transparent in sharing its results with the public, on Gallup’s website itself you may have difficulty finding any mention of the review. You need to find Newport’s blog, which three days after the press conference still doesn’t mention the report’s findings. A May 9 blog post promised to provide more information after the June press conference. The report itself is posted on the website, though there is no mention of it on the website that I can see. I found the report by using the search engine. If you don’t know it’s there, it doesn’t appear as though you would find it.