Social Desirability Bias

It must be strong today, something like 5 points.

If so, you’ll be hearing calls for the end of the secret ballot.

Advertisements
This entry was posted in Uncategorized. Bookmark the permalink.

134 Responses to Social Desirability Bias

  1. Difference Maker says:

    TRUMP 2016

  2. magusjanus says:

    Greg, by “it must be”, do you mean:
    A- In order for Trump to win, it has to be that strong
    OR
    B- that in your assessment, it IS that strong today
    ?
    (or both I suppose, but trying to get at what you meant)

  3. Erik says:

    I seem to recall the 2008 election had social desirability bias coming and going repeatedly at various stages of the process in various strengths. Props to you for staking a claim; I’m just going to admit ignorance here.

    Also, I’m thinking that social desirability bias might get hidden by fraud. People say they’ll vote for Historic First XX President but don’t do it; but Historic First XX President gets same amount of votes as promises. How live a concern is such fraud, would you say?

  4. AppSocRes says:

    Unfortunately, what is socially desirable varies from person to person. The desiccated remnants of the 1960s-1970s feminist movement are going to be voting en bloc for Clinton because the first woman president is more important than survival of the Republic or even avoiding an increased likelihood of a nuclear war with Russia. That kind of non-reasoning will drag a lot of others in its wake. OTOH, the desire to avoid confrontation with the bien pensant and their brown shirt thugs on the streets has probably suppressed the poll estimates of Trump support by better than five percent. I think there’s a reasonable probability that we will see MSM reps openly in tears late tonight as the election results start trending.

    • The betting markets have Clinton a heavy favorite at 5 to 1. I have bet with confidence on the favorite in past elections right before the election but I am not touching this one. If you like Trump then you can open up an account at 5 dimes and make +490% on your money. We can turn on the boob tube at 9PM, that is when the west coast polls close, and the real reporting starts.

      I will be drinking if Hillary wins, I’ll be drinking if Trump wins, and I’ll be drinking to celebrate the end of this nonsense. My money stays in my back pocket this election, but I trust betting markets a hell of a lot more than I do the supposed experts OR myself.

    • The Z Blog says:

      Funny, but I was on the Obama train early too. I still think that was an easy call. Dems like young and hip and he was young and hip. It was their turn and Bush was a disaster so it was the perfect time for the young hip guy to be the nominee.

      This time? I picked Trump very early to be the GOP nominee, but I am much more pessimistic about the country than you. America looks like it is ready to quit. It is an exhausted idea.

  5. The Z Blog says:

    I have a strong desire to be pessimistic about this election, but it has been so weird that I’m firmly in the “I have no bleeping idea” category this morning. I have clients in government research that are voting Trump. These are Ph.D.s and female. Their reason? Republicans tend to spend more on their stuff. First female President? They don’t care.

    On the other hand, I have friends who were not all that excited for Obama, but they are over the moon for Clinton. They were also over the moon for the Fake Indian in Massachusetts. Their reasons? Men are bad for women and these are men saying this stuff.

    I went to vote on the way to the office. The lines were huge and looked like a job site. The working class, if nothing else, is motivated to vote. It reminded me of 2008 when I was the only honky in line to vote. I doubt it means anything as the people reluctantly voting Clinton will still vote, just later in the day.

    I don’t know. I do know I skipped 2012 after standing in-line for 30 minutes watching a bus full of little brown guys handed provisional ballots. I would not bother voting today if not for the chance to poke a finger in the eye of the man. So, I will vote today.

    • epoch2013 says:

      ” I would not bother voting today if not for the chance to poke a finger in the eye of the man.”

      I think compulsory vote was meant to prevent this from influencing votes too much. At least the majority knows it has a mandate of the people that way.

  6. b.s.a. says:

    I’d believe this if Trump supporters hadn’t been so open over the past year, much more open than Hillary supporters (as opposed to Hillary nose-holders). You can’t say Trump’s not dominating the election memorabilia market. I mean, would a shy Trump voter not see all these hats, shirts, and signs, and see that about 40% of the country is for sure behind his candidate, and stop feeling as shy? And then answer truthfully in the polls? This late in the game?

    I’ll defer to Greg’s opinion over mine here (and most everywhere else), but if anyone would like to set me straight…

  7. dearieme says:

    I don’t believe the polls. They haven’t polled the dead. It might all be decided by the Norwegian Blue vote.

  8. Frank says:

    I think they should do away with the secret ballot. What’s the point? People should be held accountable for the people they didn’t know at all that they voted for.

    Maybe people would actually think about it that way.

    • lemmy caution says:

      the secret ballot is to prevent your boss from firing you if he knew how you voted. (which is totally legal)

      • albatross says:

        More broadly, there is all kinds of social pressure that could be brought to bear on people with minority opinions, from firing to beating up to police hassles to social ostracism. It’s hard to see how that improves the quality of your vote.

  9. DataExplorer says:

    According to Nate Silver’s poll summary, a fair few of the polls that have Clinton leading are anonymous internet polls. I can understand why some Trump supporters would be unwilling to admit supporting him on the telephone with their wife within earshot, but I don’t get the social desirability bias coming into play on an online poll.
    http://fivethirtyeight.com/features/election-update-clinton-gains-and-the-polls-magically-converge/

  10. TWS says:

    Fraud will decide this election. I’ve seen democrat fraud first hand when I was a kid and it is not like they have suddenly discovered moral rectitude and civic mindedness.

  11. DK says:

    That’s pretty bold prediction, Greg! I’d predict only about 2% in the same direction as you.

  12. Douglas Knight says:

    Is that your prediction, that the polls are wrong by 5 points? Do you also predict that the exit polls will be wrong by 5 points?

    And what is your distribution? Is 5 points your over-under?

    (Apologies if this is a double-post.)

  13. MawBTS says:

    Every undercard says they overperform their poll numbers. I can’t see any reason to think this is the true in Trump’s case.

    Polling data is bad on blacks and minorities, but they’re Clinton supporters.

    • deuce says:

      Aren’t those voting blocs also computer illiterate just like Trump supporters? Digital savvy made the difference in this election. The Alt-Media, especially on youtube and twitter, CRUSHED the corrupt hag you were evidently were rooting for. BTW, I made $200 betting on DJT.

      All that said, please enlighten us further, Rainbow Nostradamus.

  14. I work among psychologists, social workers, psychiatrists, occupational therapists – I think social desirability bias would be fascinating to measure here with secret ballot. No one would dare be openly for Trump. One nurse was appalled at a Trump sticker in the parking lot and waited to see who got into the car. My experience – witnessing both the mentally ill, those who care for them, and the folks who make the news – is that conservative violence is essentially defensive “Come and take it!” Cold dead fingers, etc. Liberal violence, whether sneaky or intimidating, tends to be more aggressive. That has been stark this time around.

  15. DB says:

    Holy f***ing shit.

    • Unladen Swallow says:

      I second that motion. I figured it was certainly bigger than the pollsters were saying, but I didn’t think the bias would amount to 5 percent of the total electorate, I think Silver had Trump’s chances of winning at 28.6 percent less than a day ago, you knocked it out of the park, Greg.

      • ursiform says:

        28.6% wasn’t a trivial chance for Trump. And Silver defended that high a percentage because of the possibility of there being correlated errors in the polls. The people giving Trump a ~1% chance have a lot more to be embarrassed by than Silver does.

        • Unladen Swallow says:

          Oh, I’m aware, and it had been steadily climbing, 10 days out from the election it had been 18 percent, but still being the most visible of liberal quantitative social scientists studying elections, he has egg on his face, particularly since he considerably underestimated Trump throughout the campaign.

          I think back in November or December 2015 he gave him an only 1 percent chance of winning the nomination. He also botched the 2015 UK elections, arguing for a virtual dead heat between Conservative and Labour, when the Conservatives actually gained enough seats to dump their coalition partner the Liberal Democrats.

          • ursiform says:

            A 28.6% chance of winning doesn’t mean a loss. It means that if you ran 1000 elections at those odds a candidate with a 28.6% chance of winning would win about 286 times, and lose about 714 times. You can’t tell from one election whether the projection was wrong or not. Nor can you tell how well someone makes forecasts by cherry-picking cases where they were wrong. You have to look at the statistics over many cases and see if their predictions were consistent with the odds they quoted.

            Your argument is like saying the weatherman is wrong if he projects a 30% chance of rain, and it rains. And, by the way, you remember three other days when the lesser probability happened, too, so clearly he’s a charlatan.

  16. Misdreavus says:

    Gregory M. Cochran, PhD — prophet of God.

  17. Space Ghost says:

    You win again, Dr. Cochran

  18. Difference Maker says:

    Golden age cometh

    I knew this day was likely
    Almost wish I wasn’t a US citizen so that I could have made a true killing in the markets

  19. ziel says:

    So Greg was right again – about the vote. More ominously, I think he’ll be right about his 2nd prediction as wel – we should all start monitoring the op-ed pages for the inevitable “time to re-evaluate how we vote” and “transparency should extend to how we vote as well” ‘think’ pieces.

    • pavetack says:

      So far, it’s been “Let’s get rid of the Electoral College”, not the secret ballot. Perhaps they’ve realized they can import Democratic voters, but they can’t distribute them equally.

      • ziel says:

        but they’ll soon catch on that getting rid of the electoral college is impossible – the constitution pretty much guarantees that any such effort is self-defeating. But there’s no constitutional guarantee to a secret ballot – that’s controlled by state laws, so any given state can eliminate it by simple legislation.

        • Ilya says:

          Actually, it’s quite possible it will go that way. A good justification that they could use is convenience and security, achieved via enabling people to vote electronically, via smartphones. The technology is there already.

          At first, it could be pushed as optional. Once established, it will pave a way for it to become mandatory.

          • albatross says:

            Mail in ballots already are widespread, and they make coercion/social influence a lot easier. We can all fill out our mail-in ballots together, with the boss standing right there to make sure everyone does their civic duty.

    • gcochran9 says:

      Well, it may take a while for them to notice what happened.

  20. JimLahey says:

    The breathless commentators on ABC discussed the exact phrase “Social Desirability Bias” in their coverage last night, maybe around 11:30pm EST.

  21. Yudi says:

    It looks like Trump lost the popular vote, so an attempt to get rid of the electoral college is more likely than weakening the secret ballot.

    • Toddy Cat says:

      With a Republican House, Senate, and Supreme Court, good luck with that…

    • ziel says:

      See my comment above – not possible to get rid of the electoral college. would take an amendment, passed by 2/3 of the very states that benefit from it.

      • Jim says:

        The Supreme Court could rule that the Fourteenth Amendment overrides the provisions in the Constitution regarding the Electoral College and renders the Electoral College unconstitutional. This argument doesn’t make any sense but when has that stopped the Supreme Court?

    • dearieme says:

      Nobody knows who won the popular vote. Too much cheating.

      • dearieme says:

        Though of course it’s only a guess that the cheating might outweigh the reported difference in votes. Is there any decent way of estimating the number of votes cast by cheating?

      • gcochran9 says:

        I don’t see much sign of it.

        • ziel says:

          I always read that there’s no evidence of significant election fraud, but I never see it explained how they know that. I assume there are some statistical techniques to find likely incidences of fraud?

          • lemmy caution says:

            You register to vote; then, when you vote, they write your name down. It is pretty easy to catch people.

            • Ziel says:

              But there are lots of non voters out there and people who can scan the rolls and ID those who aren’t around anymore – particularly in ‘certain neighborhoods’ – and then send people around to vote in their place. One person could easily vote a dozen times this way. This doesn’t seem too hard to pull off. But perhaps you can’t run up enough votes this way to make it worthwhile

              • lemmy caution says:

                You would have to vote in the place of people who have not registered in some other location. How would you know?

        • Josh says:

          I wonder, if trump had precommited to not challenging the results, would the outcome have been different?

        • Anonymous says:

          She only beat him by 200,000 votes. Can you imagine, among every major city of every swing state, that over 200,000 votes were falsified in total?

          • gcochran9 says:

            I think not. Getting things basically wrong, for example thinking that there is lots of voter fraud today when there’s not, keeps you from understanding the situation. Varying amounts of fraud could explain any pattern.

            Same with the people that think all the polls are deliberately skewed: I’ve never seen any evidence for that, not with who you gonna vote for questions. On policy questions, sure, by the way the question is phrased.

            If you understand that fraud is generally insignificant, and that the polls are not deliberately biased, where then do the occasional big systematic errors come from? From people lying to the pollsters and/or the exit pollsters. Because it seems that many or all of the powers that be, the Establishment, disapprove of their choice. It can be be good old-fashioned fear.

            How strong is that fear, how large is the ‘social desirability bias’? Well, in the gay marriage referenda, those against got five to seven points higher at the ballot box than in polls. Then you have to think about how strong the social pressure against Trump was (stronger than the gay marriage thing, probably) and how it varies from place to place.

            So the real answer is useful and also depressing. The pollsters weren’t deliberately lying, which they might stop doing someday, say because of repentance or legal penalties. They just don’t understand people well enough to do their fucking job.

            • Anonymous says:

              I’m not assuming that there’s a lot of voter fraud. 200,000 is all it takes and that is definitely within the realm of possibility.

              http://www.sciencedirect.com/science/article/pii/S0261379414000973

              As an aside, the assertion that it’s purely a matter of dishonesty also discounts the possibility that pollsters accidentally undersampled a given demographic because that demographic had higher turnout than expected.

              • Josh says:

                What I can’t understand is why there isn’t more voter fraud. From what I understand, with many of the current voting machines, feasibility is not the problem, I can’t imagine scruples being the problem, and motive certainly isn’t the problem.

              • gcochran9 says:

                In places with two parties, the other party has a strong motive to catch you. Also, polls are imperfect, but they would detect any massive amount of cheating.

              • josh says:

                But polls wouldn’t detect a targeted small amount of cheating and I’m not very knowledgeable on the vote counting process, but would the Rs be able to detect if somebody at one of the big voting machine companies had uploaded viruses onto some of the voting machines in swing states? Its my understanding that there is no paper record against which they can check the electronic results and its not like they are going to look at the code. I’m not sure who besides the internet crazies would have the appropriate incentive to risk being called a kook and making a lot of of enemies to investigate such a fraud.

                Not saying this happens, but I’m still not sure why not.

            • AppSocRes says:

              Non-response/refusal rates have risen steadily over the years and the non-response/refusal rates are now so high that they are going to bias estimates and make statistically correct computations of sampling error almost impossible. Polling orgs are part of the establishment and I suspect that more c onservative Americans are more likely to show their disgust with the establishment by refusing to participate in the establishment’s games. This could easily bias a poll estimate of the proportion voting for a particular candidate by five percentage points. And because we are no longer dealing with a probability sample we cannot compute standard errors correctly and therefore we have no idea what is a statistical significant difference between proportion voting one way and proprtion voting the othe.

          • Difference Maker says:

            Think on how excited the Dems were to have the historic first time Hispanic voters. As we know they were perhaps not quite as monolithic or civic minded as might have been hoped for. But literally flooding the country with new voters to tip elections

      • saintonge235 says:

                At the moment, 11/14/2016, nobody knows who won the popular vote because the final figures haven’t been published.  They were still counting at least on Saturday.

  22. Yudi says:

    To those saying it’s impossible to eliminate the electoral college: this is an attempt to get rid of it, in effect: https://en.m.wikipedia.org/wiki/National_Popular_Vote_Interstate_Compact

    • RCB says:

      No one asked, but I’d prefer the popular vote to the current system.

      • Gringo says:

        The electoral college is preferable. Consider an election where the popular vote is close- like this election, or others like 2000 or 1960. If the popular vote is the way to decide the election, ALL the 50 states have to be recounted. That will get VERY messy. For example, even though Candidate A won Illinois by 500,000 votes, it would have to be recounted in the event of a close national popular vote. Think of all the dead voters that could be unearthed in a Cook County recount. Ignoring the possibility of additional fraud, it is very messy to do a recount, and much more messy for all 50 states.

        With the electoral college, the recount is more confined.

    • RCB says:

      (Meant that to be a general comment, not a reply to you.)

  23. Polls had Hillary Clinton winning this this by 3 to 5 points and what happens? Damn near a dead tie in the popular vote. So what happened? People lied when they were asked who they were going to vote for.

    Social desirability status.

    People were given too much shit when they said they were going to vote for Trump so they lied about it. Hats off to Greg Cochran for being just about the only person to predict this correctly. Damn near everybody else, myself included, thought Clinton was going to win this.
    Nobody gave Trump much of a chance to win the states of Wisconsin and Michigan, not even the Trump crowd, but he did. Polls had Clinton winning those states easily but they were way off.

    The Trump people will say after the fact that they knew it all along, but they didn’t.

    • Difference Maker says:

      I made money, so yes I did. I’m not one of the depressives at iSteve. I didn’t know about Wisconsin or Minnesota though, but that is from laziness, emotion and arrogance.

      We could suspect Michigan already from what we know of Bernie’s performance there

      Back to Wisconsin and Minnesota, with more energy, thought and sobriety, knowing about Trump’s visit to Minnesota would be a big clue. Secondly, investigating the actual demographics and situation on the ground rather than writing them all off as comfortable cucks and cheese heads prima facie would be valuable. You would research your investments, wouldn’t you

    • Difference Maker says:

      I’d have bet 100k last year if I could have

      • Difference Maker says:

        This is not to say that it was a sure thing. As we know it was certainly a close run and dramatic race, especially with the vote totals in state after state

        The primaries were easier, for a variety of reasons

    • saintonge235 says:

              The fact that Trump visited those states and campaigned showed he thought he had a chance.  Hell, he visited here in Minnesota!  And there were times it looked like he might carry the state!!! (I thought I’d wandered into the Twilight Zone.)

              Meanwhile, Clinton never visited Wisconsin, and didn’t do all that much campaigning in the Rust Belt.  She thought she had them in the bag.  WRONG.

  24. Observer says:

    So USC Dornsife/LA Times & IBD Polls got it right. Also, Professor Helmuth Norpoth with his ‘primary model’ and an AI prediction system.

    Have to say, as someone in a very liberal city, and a very liberal workplace, it’s an odd feeling being the one person secretly pleased that Trump got over the line 🙂

    http://primarymodel.com/

    http://www.cnbc.com/2016/10/28/donald-trump-will-win-the-election-and-is-more-popular-than-obama-in-2008-ai-system-finds.html

    • dearieme says:

      “being the one person “: howdja know?

      • Observer says:

        haha true, I suppose some people might have been faking their dismay. My observations over a period of time suggest that at my work, at least, most of them simply accept whatever John Oliver or the Guardian tells them.

    • gcochran9 says:
        <a href="https://www.flickr.com/photos/93779577@N00/5548068663">like this</a>
      
    • Sandgroper says:

      That’s clearly wrong. If you look at the popular vote numbers, Trump is nowhere near as popular as Obama in 2008. In fact, he’s less popular than any of the last three Republican candidates, win or lose, including McCain running with Palin. Trump did not win the election, Clinton lost it. If the Democrats had fielded a credible candidate who could get out the voters in the big swing states, they would have beaten Trump by a mile. Their insistence on fielding Clinton is what lost it for them, almost by default. People preferred not to vote than to vote for her. Not as bad as Kerry in 2004, but close.

      I stole these numbers from someone else, but they basically tell the story.
      2016 Clinton 59,582,654 Trump 59,343,508
      2012 Obama 65,915,795 Romney 60,933,504
      2008 Obama 69,498,516 McCain 59,948,323
      2004 Kerry 59,028,444 Bush 62,040,610

      • gcochran9 says:

        The acid test of popularity is whether you win.

        Every candidate has choices. It’s perfectly possible that the winning strategy for candidate A against candidate B results in high turnout for both candidates, while the winning strategy against candidate C depresses turnout for both candidates – but depresses it more for C than for A. Nor is there a unique measure of electability: it’s perfectly possible for the electorate to prefer A to B, B to C, and C to A.

        Nor is the electorate the same in different years, demographically or psychologically – you can’t step into the same river twice, yada yada.

        That said, sure, Clinton was a lousy candidate.

        • Tolmides says:

          By the time all the votes are counted, it’s likely that Clinton will be somewhere near Obama’s 2012 numbers, and that Trump will have beaten Bush’s 2004 numbers. Clinton was no Obama, but she wasn’t Kerry either.

      • ursiform says:

        The 2016 votes aren’t all counted yet, so your argument is premature.

  25. JerryC says:

    The other difficulty in polling, besides people lying to you about their intentions, is figuring out who is going to show up at the polls. I mean, everyone knows blacks will vote D at a ~90% rate, but how many will actually bother voting? That’s more difficult to get right.

  26. Cpluskx says:

    Next four years will be terrible for US. Republicans have everything (Presidency, House, Senate) they will pass their insane laws and since democrats have nothing they will be on streets Gezi Park style. Also say goodbye to limiting climate change to 2 degrees. Singularity in mid century or it’s over.

    • Kai says:

      As a European I usually find Democrats more reasonable than Republicans, at least the bible thumping, gun crazy variety…but lately the political correctness, blank statism and general hostility to heterosexual males of democrats has become unbearable, and spread to the whole western world, Europe being as much (more for some aspects) contaminated as the US. It I’d so prevalent and Uber the top that it reached a point where it has large problematic consequences, instead of being just a posturing among academics…This had to stop, and Trump looks more likely to stop this brand of insanity than Clinton.

    • Toddy Cat says:

      With all due respect, I have not heard Trump propose anything insane. Enforcing immigration laws, allowing people to defend themselves, and not shutting down the global economy over a few degrees temperature rise seem pretty sane to me. Of course, you can disagree with this, and many do, but there’s nothing “insane” about it. Calling anyone you disagree with “insane” is part of what got Trump elected in the first place…

    • Erik says:

      What events of this magnitude have you correctly predicted before?

    • I intensely dislike The Donald, I think he was a horrible choice for president. But I wish intelligent people would get a whole lot more detached from the whole stupid mess that is politics and look at the big picture. Let me quote Cochran “modern life is a set of shared illusions.” No where else is it more obviously a set of shared illusions than politics. It is a goddamn popularity contest where really stupid people are allowed to pick which awful salesmen is going to attempt to solve complex problems they don’t have a clue how to solve. Even if they did have a clue they are at a wheel of a ship that doesn’t turn the rudder. They think it does but it does almost nothing.

      Now we can go on to the particularly awful decisions Trump is threatening to do, I said he was an awful choice for president, but maybe the Orange man is a smarter salesmen than he appears, we dunno. He did do one hell of a good job of tapping into the pent up anger in this country, we have to give him that.

      This ramble does have a conclusion, let me get to it. Let’s get this blog back to the wonderful subjects that embrace science and history, and leave the political ideologists to babble at each other on the rest of the internet.

      • Luke says:

        You have that luxury I’m sure. Anyone who’s white, heterosexual and christian doesn’t have to fear for being who they are.

        Trump is a blow to science, knowledge and everything I value.

        Any action on climate change by the US is now dead in the water. What was already a difficult task, now becomes impossible under our new orange overlord.

        • another fred says:

          I can’t comment about your values. I did not vote for Trump. If you believe that Hillary or the Dems were on the side of science and/or knowledge you have a lot to learn.

          Or you could just go through clinging to your delusions.

  27. Greying Wanderer says:

    The polls were rigged for most of the campaign (e.g. Clinton ahead by 12 points) until the last week when they had to unrig them to maintain credibility going forward.

    That ABC Clinton+12 poll in particular was an obvious psyop as they tied it to a narrative of Clinton “coasting” which covered up for why she wasn’t actively campaigning – because she was sick.

    Apart from that yes, people conform to media peer pressure and media pressure was 90% anti-Trump, some conform completely and vote accordingly and some just lie to pollsters – and women more than men.

    You saw the same thing in Brexit after the MP’s murder – a dip in Leave voters but it turned out only half that dip was real and half was people keeping their head down to pollsters.

  28. nankoweap says:

    A woman I met last week used to own a polling company. Her theory: while “likely voters” is usually more useful than “registered voters” and is the gold standard as election day nears, this year is the exception. Disaffected white people who were non-voters in 2008 and 2012 would vote for Trump this year (she said) and they by definition were categorized as not being likely voters – and so were not being counted by the pollsters. So she was confident of a Trump win.

    • gcochran9 says:

      Fron what I hear, the exit polls were predicted a Clinton win. That largely eliminates the problem of predicting who’s gonna vote.

      • nankoweap says:

        The two concepts are not mutually exclusive. It could well be that many “unlikely voters” also turned out to be shy Trump voters who didn’t particularly want to be subjected to exit polls.

      • JerryC says:

        I’m no pollster, but I think that exit polls still have to be modelled on expected turnout, because they can’t interview everyone. If your expected turnout model is wrong, your exit poll will be wrong, too.

        • gcochran9 says:

          If you interview from precincts with various compositions, you’ll pick up the turnout changes in different groups. You detect the turnout changes and take them into account.

          Luntz, looking at the exit polls, thought that Hillary was the next President. He was wrong, because the exit polls were systematically wrong – but wrong in an approximately predictable way.

  29. brendan says:

    If prediction markets ever take off so that there were thousands of different social and tech questions you could bet on – nice and liquid so big bets were possible – Greg should quit this science stuff and start running money. Make a few billion. Then things get interesting.

  30. dux.ie says:

    Recalculated from the interim raw data from wikipedia. Relative to the
    2012 results, Clinton seemed only to have small relative gains in percentages in
    6 states all of which might be immaterial as in 2 of the 6 Dem were already
    the majority while the other 4 they were not enough to become the majority.
    And for the rest she appeared to be coasting on past Dem achievements.

    DemPct16-DemPct12|DemPct16|DemPct12|RepPct16|State
    3.11|27.81|24.7|46.8|Utah
    1.92|43.32|41.4|52.43|Texas
    1.26|61.46|60.2|33.25|California
    0.72|45.32|44.6|49.64|Arizona
    0.24|45.74|45.5|51.2|Georgia
    0.11|60.81|60.7|33.52|Massachusetts

    -0.36|90.54|90.9|4.02|DC
    -0.63|55.57|56.2|38.17|Washington
    -1.71|49.49|51.2|44.68|Virginia
    -1.87|36.13|38.0|57.22|Kansas
    -2.16|38.44|40.6|58.09|Louisiana
    -2.19|55.41|57.6|39.41|Illinois
    -2.21|47.79|50.0|49.06|Florida
    -2.27|46.13|48.4|49.9|NorthCarolina
    -2.5|59.5|62.0|34.79|Maryland
    -3.08|37.72|40.8|52.89|Alaska
    -3.25|33.65|36.9|60.59|Arkansas
    -3.39|40.71|44.1|54.9|SouthCarolina
    -3.53|54.77|58.3|42.03|NewJersey
    -3.85|34.55|38.4|62.89|Alabama
    -4.04|33.96|38.0|60.33|Nebraska
    -4.06|39.74|43.8|58.32|Mississippi
    -4.2|34.9|39.1|61.06|Tennessee
    -4.24|53.86|58.1|41.73|Connecticut
    -4.27|28.93|33.2|65.33|Oklahoma
    -4.31|49.89|54.2|39.49|Oregon
    -4.38|47.62|52.0|47.25|NewHampshire
    -4.5|47.6|52.1|48.84|Pennsylvania
    -4.51|47.89|52.4|45.53|Nevada
    -4.59|46.91|51.5|44.8|Colorado
    -4.74|48.26|53.0|40.05|NewMexico
    -5.11|32.69|37.8|62.54|Kentucky
    -5.14|27.46|32.6|59.32|Idaho
    -5.25|53.35|58.6|41.92|Delaware
    -5.48|61.12|66.6|32.64|Vermont
    -5.51|57.89|63.4|36.84|NewYork
    -5.73|35.97|41.7|56.52|Montana
    -5.96|46.94|52.9|47.87|Wisconsin
    -6.03|37.87|43.9|57.12|Indiana
    -6.17|21.63|27.8|67.4|Wyoming
    -6.29|46.41|52.7|44.95|Minnesota
    -6.56|37.84|44.4|56.88|Missouri
    -6.87|47.33|54.2|47.6|Michigan
    -7.19|43.51|50.7|52.06|Ohio
    -8.16|31.74|39.9|61.53|SouthDakota
    -8.46|47.84|56.3|45.16|Maine
    -8.87|53.83|62.7|39.46|RhodeIsland
    -9.03|26.47|35.5|68.65|WestVirginia
    -9.52|60.98|70.5|29.44|Hawaii
    -10.29|41.71|52.0|51.21|Iowa
    -11.46|27.24|38.7|62.95|NorthDakota

  31. Wieland says:

    Another insight. Write another book please. Could be about anything.

  32. Cato says:

    Our dean gave us a long talk about polling the day after the election (he is a marketing research guy). He told us that the best prediction was made by a pollster who asked two questions: 1) who will you vote for? 2) who will your neighbors vote for? From these two questions the pollster was able to neutralize the “shy Trump” effect, and make an accurate prediction. This was the kind of meeting where I had to squirm and clench my lips and restrain myself from mentioning the prediction made here. I’m very impressed at the accuracy of this prediction–it’s what I thought three months ago, but then I did the usual inside-the-system thing and allowed myself to re-embrace Babylon.

    • gcochran9 says:

      I think that was the Trafalgar group.

      My favorite comment was Nate Silver suggesting that there might well be people reluctant to expose their preferences – Hillary supporters. It’s possible that being a gay Jewish baseball nut and poker enthusiast does not, in itself, attune you to the zeitgeist of Middle America.

      I say that thinking that Silver was genuinely trying to get the right answer. He understands the virtue of averaging polls. He just doesn’t understand anything about people, or this country.

      But then maybe he’s been closer to the Zeitgeist other years.

      • A Erickson Cornish says:

        All true, though Silver and most polling aggregates had HRC by 3.5, and it looks likely to finish HRC by ~1. Thus the final result appears to be about equidistant between the polls and Greg’s estimate of Trump by 1.5 (I’m inferring that estimate from the original post). He did get the direction of the error right, of course, and likely the source as well, though it isn’t immediately apparent (to me at least) why this systemic error would be strongest in the predominantly Rust Belt states where it was. After all, weren’t those of us in urban centers on the coasts subject to much stronger social pressure than, say, voters in those rural counties that swung and won him WI, PA, etc.?

        • gcochran9 says:

          The size of the systemic error ought to be proportional to the size of the Trump vote ( all else equal) – which was smaller in places like California. As for the error in the state polls, I didn’t see many polls in California, thought (correctly) not to be in play.

          More research is needed !

          • Douglas Knight says:

            Andrew Gelman has a scatterplot of poll error against forecast. There is a strong positive correlation, but the intercept is not zero — the effect is not proportional to the Trump vote. For example, states close to 50% had a 2 point error, while states close to 60% had a 4 point error. Why do you predict that? It’s the opposite of what social desirability bias predicts, isn’t it? Alternately, I could imagine a social desirability bias to vote the same way as last time, suggesting that we look at swing against error. I haven’t seen such a graph, but I don’t think it would have a strong correlation.

          • ursiform says:

            There was a lot of polling in California, including by well established organizations with good track records. Much more than in most states. But the presidential poll results didn’t receive much attention because, as you mention, the result wasn’t in doubt.

  33. Luke says:

    Greg.
    human beings in general are more receptive to news they want to hear and information that confirm their biases.

    However, just from looking at the quantity of fake news on the alt-right, are conservatives more receptive to fake news more so than people on the left? I know that by sheer number, they simply can’t all be stupid. But so much of the fake news goes against things/facts that you can readily find from reliable sources.

    Crime data can be found on police websites, fbi databases.

    Do people on the right know that they read fake news, but simply ignores logic and accept misinformation because it comforts them?

    They seem to not be able to discern opinions from fact. What every college student knows in research, that primary sources are more reliable than secondary, that expert opinion more likely grounded in facts than amateur.

    It reminds me of 1984, or the Matrix, yet this is the reality of the US.

    • Dahlia says:

      From what I observed, the fake news was passed around by the “low info” Trump voters, whereas Hillary voters, low info and “smart”, were taken in by the obviously erroneous (e.g., Science says 99% chance of Hillary win).
      Apples to apples and all that, I’m not aware of smart Trump supporters being taken in by either type of mistake.
      Helmut Norpath, no idea of his politics, was extremely humble in post-election interviews despite getting it right. Most smart people weren’t like the “smart” people pre-election, crowing smugly all over the place.

    • Michael says:

      So much of so-called main stream news is also fake, so it’s hard to sort out which groups believe more fake news. For example, the “hands up don’t shoot” story was entirely fabricated, but is still being sold as a real story. Every liberal believes it. Lots of people died because of it. I don’t know of any fake alt-right news story that caused a death.

  34. Gabriel M says:

    It’s amazing to me how many Dems read this blog. HINT: you can’t be a Christian and read blogs about how the Trinity is a bunch of crap and makes no sense, and you can’t be a liberal and believe in HBD.

    • Ursiform says:

      At least not if you get to define the terms.

    • Jim says:

      Your comparison of liberalism with Christianity is apt. Both are religions or ideologies. HBD at least as discussed by most people commenting on this blog is an attempt to understand reality. Reality cares nothing about our moral preferences.

      Ideologues, whether right or left, seem unable to grasp that reality may not always cooperate with one’s wishes.

      Actually I think one can be at least a certain kind of liberal and still recognize HBD realities. I think I recall Jayman describing himself as a liberal. But let him speak for himself.

      • Gabriel M says:

        If 99% of your co-coreligionists would consider you evil for expressing your opinions to them, it is a good sign that you are in an untenable position.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s