Conservatives are way ahead of Labour in polls of the United Kingdom — which may have been behind Prime Minister Theresa May’s decision to declare a “snap” general election for June 8 in the hopes of increasing her party’s majority in Parliament.
But May’s move isn’t risk free. In fact, if the final polls come within a couple of percentage points of the actual vote on June 8, it will be something of an upset. Polls in the U.K. have a history of inaccurate performance, with average errors that more than double the ones in American presidential elections.
American journalists considered it a huge failure for the polls last year when Donald Trump won the White House. But polls in U.S. general elections have mostly been pretty good. Final polling averages in the U.S. have missed the presidential popular vote by only 2 percentage points, on average, since 1968. And contrary to the conventional wisdom, Trump’s performance wasn’t an exception. He beat his national polls by only 1 to 2 percentage points in losing the popular vote to Hillary Clinton. He beat his polls in the average swing state by just 2 to 3 points in winning the Electoral College. Far from having committed an extraordinary error, the polls had correctly pointed toward a competitive race in the Electoral College, although the pundits had mostly ignored this data.
If you don’t appreciate how good we’ve had it in the U.S., take a look across the Atlantic. In the table below, I’ve collected data on how accurate U.K. polls have been in general elections dating back to 1979, along with last year’s Brexit referendum and Scotland’s 2014 independence referendum (which only Scottish residents voted on). The chart shows the polling average at two intervals: first, 50 days out from the election1 — which is about how far we are from the snap election now — and second, in the final week before the election.2 Then, it compares the polling averages to the actual result.
| AVG. POLLS 50 DAYS OUT | AVG. POLLS WEEK BEFORE | ACTUAL | |||||||
|---|---|---|---|---|---|---|---|---|---|
| YEAR | VOTE | LEADER | MARGIN | ERROR | LEADER | MARGIN | ERROR | WINNER | MARGIN |
| 2016 | Brexit | Remain | +3.7 | 7.5 | Remain | +2.4 | 6.2 | Leave | +3.8 |
| 2015 | Parliament | Lab. | +0.1 | 6.7 | Con. | +1.0 | 5.6 | Con. | +6.6 |
| 2014 | Scotland | No | +9.0 | 1.6 | No | +3.3 | 7.3 | No | +10.6 |
| 2010 | Parliament | Con. | +6.3 | 0.8 | Con. | +7.7 | 0.6 | Con. | +7.1 |
| 2005 | Parliament | Lab. | +7.0 | 4.2 | Lab. | +8.8 | 6.0 | Lab. | +2.8 |
| 2001 | Parliament | Lab. | +20.6 | 11.6 | Lab. | +15.0 | 6.0 | Lab. | +9.0 |
| 1997 | Parliament | Lab. | +24.1 | 11.6 | Lab. | +18.1 | 5.6 | Lab. | +12.5 |
| 1992 | Parliament | Con. | +0.1 | 7.4 | Lab. | +2.0 | 9.5 | Con. | +7.5 |
| 1987 | Parliament | Con. | +11.4 | 0.0 | Con. | +9.3 | 2.1 | Con. | +11.4 |
| 1983 | Parliament | Con. | +7.3 | 7.5 | Con. | +18.7 | 3.9 | Con. | +14.8 |
| 1979 | Parliament | Con. | +13.8 | 6.8 | Con. | +5.2 | 1.8 | Con. | +7.0 |
| Avg. | 6.0 | 5.0 | |||||||
On average, U.K. polls this far out have missed the final margin by 6 percentage points. And they don’t get all that much more accurate as you go along — the final polling average has missed the result by 5 points. The experience in Brexit last year — when the polls missed the final margin by 4 points according to the Huffington Post polling average or 6 points according to the method I described above — wasn’t a big outlier by U.K. standards. The same goes for the previous U.K. general election in 2015, when they underestimated Conservatives by around 6 points. Polls in 2010 were quite good in diagnosing the Conservative-Labour margin, although they considerably overestimated Liberal Democrats’ performance.
May’s Conservatives do have a massive lead, with recent polls showing them 9 to 21 points ahead of Labour and their unpopular leader, Jeremy Corbyn. Also, while the polls in the U.K. haven’t been very accurate, they’ve tended to underestimate Conservatives rather than Labour in the past. (See also: the Shy Tory Factor.)
But if polls are missing election outcomes by 5 or 6 points on average, that means the margin of error (or 95 percent confidence interval) is very large indeed. Specifically, a 6-point average error in forecasting the final margin translates to a true margin of error of plus or minus 13 to 15 percentage points, depending on how you calculate it.
Exactly how sure Conservatives are of retaining a majority of seats in Parliament is a more complicated question than who will take the most votes. This stuff is tricky to model — as we’ve learned the hard way — because the results are greatly dependent on where the vote is distributed and how third parties perform. The Scottish National Party is likely to continue to control the vast majority of the 59 seats from Scotland, which gives Conservatives less margin for error in the event the election tightens. Bookmakers give 20-to-1 odds against a Labour majority, but only 5-to-1 odds against a hung Parliament — about the same odds that many of them offered on “Leave” winning the Brexit vote last year.3
So while May’s Tories are undoubtedly in a good position, this move isn’t totally without risk. Public opinion in the U.K. can change in a hurry, and pollsters have long had trouble nailing it down.