The U.S. Doesn’t Have A Monopoly On Polling Problems

A report on British polling highlights the challenges election pollsters face worldwide.

Polling has a record of 1-1 in the 2016 presidential primary contests so far, with a miss in Iowa and success in New Hampshire.

Media and pollsters worldwide are watching the U.S. presidential polls. “What happened in the US ... has similarities with Kenya,” Magie Ireri wrote in All Africa last week, comparing the polling miss in Iowa to a miss in the polling prior to Kenya’s 2013 presidential election. “As we get closer to the 2017 elections … Kenyans need to be acquainted with some factors that could result in a significant difference between opinion polls and the actual election outcome.”

Some of the factors that Ireri lists are identical to those U.S. pollsters point to when electoral outcomes differ from what polls predicted -- turnout levels and voters changing their minds at the last minute were huge contributors to the miss in Iowa.

Polls are under close scrutiny in part because the past year has not been the best for the polling industry worldwide. 2015 was full of polling mishaps: In March, polls missed the outcome of the Israeli Knesset elections. In May, polls leading up to the British parliamentary elections indicated a close race, but the Conservatives won a clear majority. In July, polls grossly understated the degree to which the Greek people would reject austerity measures. Canadians surprised pollsters with a Liberal Party majority in October. In November, U.S. polling organizations botched the outcome of the Kentucky governor’s race.

What happened in the US ... has similarities with Kenya. Magie Ireri in All Africa

The British polling failure created the biggest splash of the 2015 blunders. Most of the other misses could be attributed to a dearth of high-quality polling data and complicated electoral situations. In the U.K., however, there were 92 polls conducted between the official start of the campaign on March 30 and the election on May 7. Nearly all of them showed the Conservative Party and the Labour Party neck and neck, and a few even had the Labour Party ahead.

The Conservatives’ 6.5 percentage point win was enough of a surprise to prompt the British Polling Council to launch an inquiry into what went wrong. The preliminary results of the report were released in late January, with the promise of a full report sometime in March.

The short answer to what went wrong is that the polls talked to too many of the same type of voter -- politically engaged people, and particularly young politically engaged people who supported the Labour Party. On Election Day, a variety of people turned out to vote -- engaged and not. Those not politically engaged who had often been missed by pollsters tended to favor the Conservative Party. Late opinion shifts also contributed to the problem, but the primary issue seems to be that the polls’ samples weren’t representative of who would actually vote, so they didn’t anticipate the Conservative win.

These initial findings from the U.K. inquiry are similar to what the American Association for Public Opinion Research found in 2008 when they investigated how all of the New Hampshire pre-primary polls that year missed a surge of support for Hillary Clinton over Barack Obama. That report concluded that late opinion shifts, incorrect estimates of who would vote and difficult-to-reach voters were to blame.

The initial release from the British polling inquiry says that the final report will make recommendations for improving polls, which the AAPOR report on New Hampshire did not do. Those recommendations could prove vital to the field beyond the U.K. -- especially since eight years after the New Hampshire 2008 debacle, not much has changed.

Polls still end before voting begins, turnout models are often flawed in their predictions of who will vote, and hard-to-reach voters are still hard to reach. Without changes, these patterns will continue throughout 2016 and into 2017 in the U.S., Kenya and beyond.