Election polling: why is it so difficult?

preview_player
Показать описание
Election polling is tricky. While pollsters' successes can fill headlines, if they get it wrong, their failures do too. Is there a better way to predict elections?

00:00 - What’s the best way to predict elections?
00:48 - How were elections predicted in the past?
02:36 - How do modern day polls work?
04:32 - Why polls miss the mark
07:14 - How does statistical modelling work?
08:30 - Our French election model

Рекомендации по теме
Комментарии
Автор

I am interested in how the polls can actually influence the outcome of an election. We all know that voters like to back horses that they believe can win. Donors prefer to place their money where it is likely to contribute to success.

Is there any research on this topic?

J-A-A-K
Автор

Beautifully produced and edited! Kudos to The Economist for giving us these informative, topical, and aesthetically pleasing videos.

antheacao
Автор

One thing that they did not discuss at 5:30 is that the trust in the government in those states was very low. With more people losing trust it would make it much harder to know who would go to vote and who would not go. A documentary by Michael Moore named "Fahrenheit 11/9" goas more in depth on that specific topic.

robbydomino
Автор

Thanks for another great video, Economist!

arcofspira
Автор

Wasn't expecting those tarot cards to flip over to show faces of presidents. Nice

fortune_roses
Автор

This was an amazing video. Really enjoyed the methodology walk through toward the end there and providing the source code. So cool

babaksanaee
Автор

Very well done video! Love the new style!

riley
Автор

I feel like the 2020 US Presidential Election would be a more interesting case study than 2016; in the former, unlike in 2016, the national polls were incredibly inaccurate

jacobwatts
Автор

US elections are so messed up. I believe most countries do elections more or less like in my country, Colombia. In the presidential election, anyone who was able to become a candidate can participate. Every citizen votes directly. If anyone gets 50% plus one of the votes that's it. If nobody could achieve that percentage then a second round with the two highest voted candidates takes place. Again, it's one vote per citizen; whomever gets highest number of votes wins. What in the US is called "popular vote", that is in fact the vote that should count.

IRosamelia
Автор

Fascinating, I didn't know that poll forecasting is very similar to weather forecasting!

antonyjohnson
Автор

Never knew where the term Gallop Polling came from!

toasted_.coconut
Автор

In portugal we have "the golden parish"
its a parish whose polls are consistently similar to the national survey and with a similar demographic composition to the national one. On the eve of the elections, the golden parishes are analyzed again and again.

Duck-wcde
Автор

Working out the future leader works in systems of single winner constituencies or presidential systems but in proportional systems the leader is determined by deals between parties and can't be predicted in the same way.

catmonarchist
Автор

presidential election in the philippines is coming up in may, we are very excited what will be the result. have a peaceful election there in france

josemarirobledo
Автор

Maron will win because all other parties and voters from the left to the center will band together to prevent a Le Pen win. It always happens in French politics.

gtfoutube
Автор

if there are two to choose from - pick that one, or that one

importantname
Автор

"Negative" voting voting Is devilishly hard to poll, because the effect is difficult to quantify....
A positive vote is a simple binomial, while the negative has a third option, whether to simply withhold completely.
In 2016, with *both* candidates 'negative' ratings so high, it's a wonder the final numbers came close to the polling at all.

floydblandston
Автор

Great explanation, but the point about needing to get a sample size that looks like the population needs to be clarified. In election polling, you need a sample that looks like the VOTING population, not the overall population. In the US this has been increasingly difficult over the past three election cycles. The problem is simple (you need to know who is going to turnout in order to make an accurate prediction based on your poll results), but the solution is notoriously difficult when turnout AND poll-response demographics vary wildly between elections and parties.

ichifish
Автор

Great use of Tarot, im sure that ms Abramovic would be proud. Your world in 2017 magazine cover did a great job at "predicting the future"

KnifedbyButter
Автор

So... What is the forecast for Colombia?

jonathanaguilar