HN Explorer

Ask HN: Political Polling Methodology?

4 points by forkandwait ↗ HN
Is there anyone here who can describe in some mathematical detail how an election forecast is calculated?

Obviously, I wonder why the latest presidential race forecasts were so consistently wrong, but I am curious in general too.

I imagine for every response ("yes, I will vote for Trump") there are weights for how representative is that response, how likely, and what is the associated turnout. But I am sure it is more complex.

7 comments

[ 3.3 ms ] story [ 29.6 ms ] thread
pcunite
>> why the race forecasts were so wrong

Because America is not just New York, Silicon Valley, and Hollywood. The media, pollers, and forecasters shared a narrative that they wanted to be true, not one that was actually grounded in investigated reality. We should all take from this experience to seek for truth, the way it is, not the way we wished it could be. Save that for opinion polling.

Anecdotal:

The media was telling me no one in their right mind could side with Trump. Okay. I was watching both candidates online and there would be like 70,000 watching Trump and maybe 3,000 watching Hillary. Personalities like Lady Gaga don't equate to 60,000 people.

Trump had more twitter followers, more FaceBook likes, and more people at his rallies. All the data was there. It is just that the media did not want to report that. And finding out the reasoning behind that ... well, that would be an interesting conversation.

blairbeckwith
I think that – sadly – a lot of people assumed that the majority of people watching the rallies and following on social media were there for the entertainment value. I think some were, but it seems like most were there sincerely.
dumbmatter
http://fivethirtyeight.com/features/a-users-guide-to-fivethi... is a decent overview of the most popular forecast. There's plenty of other relevant content on that site, too.
matt_s
That group was featured on ABC news coverage last night in the US. They predicted Clinton would win 70% to 20% early in the evening.

I didn't read what they said about their process but they were completely off too.

dragonwriter
> They predicted Clinton would win 70% to 20% early in the evening.

No, they predicted about a 70%:30% chance of Clinton vs. Trump win, not a 70% to 20% win by Clinton.

> I didn't read what they said about their process but they were completely off too.

It wasn't so much that they were completely off as that some people mistake a claim of probability > 50% with a claim of certainty.

Sure, the most-likely winner map was wrong in a number of states, but the reason 538 had a 30% chance of Trump winning rather than much lower chances given by other models was that 538 takes into account the fact that deviations in results from polling in different states tend to be correlated rather than independent.

ryanmccullagh
You'll most likely find that methods are proprietary. Some companies are in this space. For example, a company I was just rejected from claimed they have a monopoly on polling. A quick search could reveal who this company is. Hint: Eric Schmidt is on the board.
zzzcpan
> Obviously, I wonder why the latest presidential race forecasts were so consistently wrong

I think it's rather obvious, some people didn't want to participate in the polls, likely out of shame for their candidate, fear of being shamed for their choice in front of their peers, etc.