Sabtu, 14 Desember 2019

What did the UK elections teach us about 2020? Trust the polls - CNN

Maybe you think Republican President Donald Trump has reason to smile after his friend, Conservative Prime Minister Boris Johnson, won a big mandate.
Maybe you think the crushing defeat of Jeremy Corbyn's Labour Party shows the peril of the Democrats potentially nominating Bernie Sanders.
Those theories may prove to be true, but I think the clearest lesson is staring us right in the face: The polls are still pretty good as we head into the 2020 presidential election in the US.
Take a look at the average of polls for the four parties that have earned at least 10 seats each in the House of Commons (the UK Parliament's lower House). The average of the final UK polls had the Conservatives winning 43% of the vote, Labour 33%, the Liberal Democrats 12% and the Scottish Nationals 4%.
The actual result was Conservatives taking 43.6%, Labour 32.2%, the Liberal Democrats 11.6% and the Scottish Nationals 3.9%. In other words, each of these parties got within 1 point of its final polled vote share.
This remarkably accurate result was better than we'd expect based on history. The final 2019 polling average missed the margin between Conservative and Labour by about 1.9 points. Since the 1945 election (i.e. the prior 20 UK general elections), the average final poll had missed by 3.9 points.
Indeed, despite a lot of cries that the polls are broken, the UK elections taking place during the Trump administration show that isn't true. Beyond this year, the difference between the Conservatives and Labour margin in the final 2017 polling average and election result was 4 points. In other words, it's right in line with what we'd expect, given the historical polling accuracy rates.
The US's own polls have likewise been fairly accurate during the Trump era. The average House, Senate and governor's polls were about a point more accurate in 2018 than they had been in similar elections over the prior 20 years. The same was generally true for House special election polling in the 2017-2018 cycle and the three governor elections of 2019 (Kentucky, Louisiana and Mississippi).
Another key point is that just because one side outperformed in the polls in the last election doesn't mean the same party will outperform in the next one. I know some people were expecting (and a lot of Labourites were hoping) that because the polls underestimated Labour in 2017 they would do the same in 2019. It didn't happen. The Conservatives were actually slightly underestimated.
Again, we saw this same lesson play out in the US over the past few years. After the polling underestimated the Republicans almost across the board in 2016, there was less of a systematic error in 2018. The polls slightly underestimated the Democrats on average. Now, the polls weren't perfect in 2018 in the US, but they were better than average and correctly projected a strong Democratic year. Similarly, the polls, if anything, underestimated the Democrats in the gubernatorial elections of 2019 and special elections over the course of 2017 and 2018.
The direction of the polling errors is most often random. If something is methodologically amiss in surveys, good pollsters tend to figure out what's wrong before the next election.
None of this guarantees that the final polls will correctly gauge who is going to win or lose in 2020. There are still margins of error, so someone slightly ahead in the polls may end up losing. Likewise, someone slightly behind may end up winning.
But in an era with a lot of disinformation out there, the polls continue to do a very good job of separating the signal from the noise.

Let's block ads! (Why?)


https://news.google.com/__i/rss/rd/articles/CBMiXWh0dHBzOi8vd3d3LmNubi5jb20vMjAxOS8xMi8xNC9wb2xpdGljcy91ay1lbGVjdGlvbi1wb2xsaW5nLTIwMjAtcHJlc2lkZW50aWFsLXJhY2UvaW5kZXguaHRtbNIBYWh0dHBzOi8vYW1wLmNubi5jb20vY25uLzIwMTkvMTIvMTQvcG9saXRpY3MvdWstZWxlY3Rpb24tcG9sbGluZy0yMDIwLXByZXNpZGVudGlhbC1yYWNlL2luZGV4Lmh0bWw?oc=5

2019-12-14 13:32:00Z
CAIiEOYp3jl4QIEIvYROdY_uFMgqGQgEKhAIACoHCAowocv1CjCSptoCMPrTpgU

Tidak ada komentar:

Posting Komentar