Can we trust the polls?

[Vern]

Since 2016, when all the polling data were off. I have lost trust in polls. I was wondering what you guys think. Can we really trust the polls? I want to, but in the back of my mind I just can't. Was it just a freak event. Or is polling getting harder to do.

[136or142]

Since 2016, when all the polling data were off. I have lost trust in polls. I was wondering what you guys think. Can we really trust the polls? I want to, but in the back of my mind I just can't. Was it just a freak event. Or is polling getting harder to do.

Despite the headlines, polling in most countries and in even in most elections in the United States seems to be getting better.

In the case of Trump and 2016, I think the main problem was that the one time Trump seemed to actually tell the truth - in this case that he would get millions of people out to the polls who had rarely if ever voted before - nobody, at least no pollster, actually believed that would happen.

So, I think the problem with the polls, and this was mainly a problem at the state level, was that the polls' 'likely voter' screens were way off.

[GM Team Member and Deputy PPT WB]

always remember: polls have such a thing as error. Never only take polls into consideration.

[136or142]

always remember: polls have such a thing as error. Never only take polls into consideration.

The pollsters, the good ones anyway, are constantly looking to see where they were off, if they were off. I think this is the main reason the polls for the last year and a half or so have been largely accurate.

[ON Progressive]

Even the state polls in 2016 weren’t overly off, outside a couple of Midwestern states. All these averages are from RCP:
PA - Polling average Clinton +1.9, Trump +0.7 actual result (error of 2.6%)
MI - Polling average Clinton +3.6, Trump +0.3 actual result (error of 3.9%)
FL - Polling average Trump +0.2, Trump +1.2 actual result (error of 1%)
IA - Polling average Trump +3, Trump +9.5 actual result (error of 6.5%)
NH - Polling average Clinton +0.6, Clinton +0.3 actual result (error of 0.3%)
NC - Polling average Trump +1, Trump +3.7 actual result (error of 2.7%)
ME - Polling average Clinton +4.5, Clinton +2.9 actual result (error of 1.6%)
AZ - Polling average Trump +4, Trump +3.5 actual result (error of 0.5%)
CO - Polling average Clinton +2.9, Clinton +4.9 actual result (2% pro-D error)
OH - Polling average Trump +2.2, Trump +8.1 actual result (error of 5.9%)
WI - Polling average Clinton +6.5, Trump +0.7 actual result (error of 7.2%)
NV - Polling average Trump +0.8, Clinton +2.4 actual result (3.2% pro-D error)
VA - Polling average Clinton +5, Clinton +5.4 actual result (0.4% pro-D error)
GA - Polling average Trump +4.8, Trump +5.1 actual result (error of 0.3%)

Those combined errors sum up to a 38.1% error in 14 states, an average error of 2.7%. Take out errors that favoured Dems, and you get an average error of 2.95%. That is a very normal error. The only places that the polls were egriously off were IA/OH/WI, and the polls had Trump winning two of those anyway.

[Xing]

I mean, basically you can either trust the polls, trust your gut feelings, or factor in both. Many people whose gut told them Trump would win are probably largely ignoring polls completely this time around, but I don't think trusting one's gut always works out.

[Virginiá]

Even the state polls in 2016 weren’t overly off, outside a couple of Midwestern states. All these averages are from RCP:

[...]

The "polls were wrong" myth is easily the most annoying one about 2016 (to me anyhow). I think the reason it sticks around so persistently despite not really holding that much water is because there was this attitude that Trump wasn't going to win, and so when he did, certain people kind of internalized any possible errors they heard mentioned much more than they should have. The national polls in the end were just fine (do people forget polls have margins of error?) and like you said, not all states were that bad. It's just a handful that seem to drive this narrative.

In the end, I always like to remind myself that it's really irrelevant what the average joe thinks about 2016's polling. If the lesson they want to take away is that because they heard somewhere that 2016's polling was off big time, that anything is possible and that 2018 = #RedWave and 2020 = Trumpslide, then by all means, go ahead. It doesn't make it so, and that kind of complacency on the right very much welcomed by a leftie like me.

[Amenhotep Bakari-Sellers]

It's still early and polls don't mean much. Polls have showed rebound by the GOP, but still trail in states they should be leading in like FL. Democratic takeover is still likely in Nov.

[○∙◄☻¥tπ[╪AV┼cVê└]

Polling has gotten harder, and one of the best pollsters, California's Field poll, shut down.

[136or142]

Even the state polls in 2016 weren’t overly off, outside a couple of Midwestern states. All these averages are from RCP:

[...]

The "polls were wrong" myth is easily the most annoying one about 2016 (to me anyhow). I think the reason it sticks around so persistently despite not really holding that much water is because there was this attitude that Trump wasn't going to win, and so when he did, certain people kind of internalized any possible errors they heard mentioned much more than they should have. The national polls in the end were just fine (do people forget polls have margins of error?) and like you said, not all states were that bad. It's just a handful that seem to drive this narrative.

In the end, I always like to remind myself that it's really irrelevant what the average joe thinks about 2016's polling. If the lesson they want to take away is that because they heard somewhere that 2016's polling was off big time, that anything is possible and that 2018 = #RedWave and 2020 = Trumpslide, then by all means, go ahead. It doesn't make it so, and that kind of complacency on the right very much welcomed by a leftie like me.

I think that's actually much more the case of Brexit than the 2016 U.S Presidential election.  If the one rogue poll that showed a large lead for the 'stay' side in Brexit is removed, the polls actually showed a dead heat which fits with the final result of around 52-48%

That the 'political and economic elites' in the U.K took these tied polls as a guaranteed 'stay' really just showed how out of touch they were, and said nothing about the polling, which was largely accurate.  I personally blame this on the British first past the post system which works (most of the time) as an attempt to suppress votes for extremists (by making extremists vote for the more mainstream parties if they want to have any real say) but, does so at a cost of not providing anything near to a reflection of actual popular sentiment, if not in the party vote share, certainly in the seat share.

The problem here, is that from the state polls presented, first, if there is a 'meta survey' of all the polls, the margin of error goes down, so Michigan should be included in being 'way off' second, all of the major errors were in the same direction (a bias in favor of Hillary Clinton.)  So, while the overall polling wasn't too bad, it did lead to an incorrect election result prediction.

[Antonio the Sixth]

Actually, there's no sign that polls are getting less accurate on average.

The truth is, polls have always been and continue to be a very rough indicator. So taking them with a grain of salt is a good idea.

[Progressive Pessimist]

Yes but they matter much more the closer they are to election day. Technically the polls we have this far out are almost instantly irrelevant. They just exist to give us an idea of how the race is going but they are not a predictive guarantee.

[RussFeingoldWasRobbed]

As long as we take margin of error into account, I think so.

[Amenhotep Bakari-Sellers]

There's definitely conflicting polls out there, showing Dem incumbent senators in trouble, as well as GOP incumbents, now Cruz is only ahead by five. I say the are polls that are contradictory.

[Holy Unifying Centrist]

There's definitely conflicting polls out there, showing Dem incumbent senators in trouble, as well as GOP incumbents, now Cruz is only ahead by five. I say the are polls that are contradictory.

The average is Cruz + 8. Always look at the average, not the individual poll. You'd think someone who has been on here for 10+ years would understand this.

I don't expect that race to become more competitive until Beto actually starts running ads... although the most likely result is Cruz + 6.

[Amenhotep Bakari-Sellers]

I was looking at his vulnerablility, it's definitely a race now, as before, it wasn't worth money spent