Skip to main content

The RAND presidential poll is an interesting one because it polls the same group of people week after week.  Each participant is polled once a week, and the total is averaged over the week.  There are two important limitations to this methodology that need to be borne in mind:

  1. The original sample has sampling error, like all polls, but that error is propagated throughout the poll, whereas if fresh samples are drawn, it will cancel out over time.  For this reason it is misleading to consider the absolute margin as a good indicator of the state of the race - the same non-representativeness of the original sample will be present in every total.  However, what this means is that its usefulness lies in the direction of the trends, which will tend to be far less noisy than when a new sample is drawn each day, whilst the opposite is true of conventional polls (sampling error cancels out over time, but makes the trends noisy).
  2. Polling people regularly will itself tend to focus their minds to the race - as a result, by now, this is probably a group of people that have shown an atypical degree of interest in the race. However, it occurs to me that in that sense it may mimic the opinions of those in swing states who have been much more heavily targetted with news, conventional polls, and ads.  So perhaps RAND may be regarded as a kind of proxy measure of attitudes in swing states.

So what I did was to plot the RAND margin, relative to its own baseline, i.e. with the margin artificially set to zero (in fact the Pollster model has Obama 1.4 points ahead on that date, which is what the RAND poll originally showed, so this is conservative), and simply tag a few key events.

And it seems to me (looking through my telescope from the UK) that a very clear pattern emerges (see below):

Romney got a small bounce from his VP announcement, and from the RNC convention.
He took a major hit from the release of the 47% video, from which he was recovering around the time of the first debate.  

Obama got a big bounce from the DNC convention, and caught the 47% video wave just as the DNC wave was subsiding.  After the first two presidential debates, Obama got a very temporary (immediate response of people polled the day after?) upswing, followed by a decline, although Biden's debate was followed by a steadier rise for Obama (possibly because the hit Obama took from the first debate was subsiding anyway).  Then, since the final debate, Obama's margin has risen to almost peak levels.

So, if RAND can be regarded as a measure of how people well-tuned to the race have reacted to key events, I'm really hopeful.  The message I'm taking home is that among attentive voters:

  • Obama's margin has a net rise since July
  • His gains have exceeded his setbacks
  • He has gained since the final debate
  • These trends may be a good proxy for voters in swing states

Onwards and upwards!

EMAIL TO A FRIEND X
Your Email has been sent.
You must add at least one tag to this diary before publishing it.

Add keywords that describe this diary. Separate multiple keywords with commas.
Tagging tips - Search For Tags - Browse For Tags

?

More Tagging tips:

A tag is a way to search for this diary. If someone is searching for "Barack Obama," is this a diary they'd be trying to find?

Use a person's full name, without any title. Senator Obama may become President Obama, and Michelle Obama might run for office.

If your diary covers an election or elected official, use election tags, which are generally the state abbreviation followed by the office. CA-01 is the first district House seat. CA-Sen covers both senate races. NY-GOV covers the New York governor's race.

Tags do not compound: that is, "education reform" is a completely different tag from "education". A tag like "reform" alone is probably not meaningful.

Consider if one or more of these tags fits your diary: Civil Rights, Community, Congress, Culture, Economy, Education, Elections, Energy, Environment, Health Care, International, Labor, Law, Media, Meta, National Security, Science, Transportation, or White House. If your diary is specific to a state, consider adding the state (California, Texas, etc). Keep in mind, though, that there are many wonderful and important diaries that don't fit in any of these tags. Don't worry if yours doesn't.

You can add a private note to this diary when hotlisting it:
Are you sure you want to remove this diary from your hotlist?
Are you sure you want to remove your recommendation? You can only recommend a diary once, so you will not be able to re-recommend it afterwards.
Rescue this diary, and add a note:
Are you sure you want to remove this diary from Rescue?
Choose where to republish this diary. The diary will be added to the queue for that group. Publish it from the queue to make it appear.

You must be a member of a group to use this feature.

Add a quick update to your diary without changing the diary itself:
Are you sure you want to remove this diary?
(The diary will be removed from the site and returned to your drafts for further editing.)
(The diary will be removed.)
Are you sure you want to save these changes to the published diary?

Comment Preferences

  •  Thanks for the chart (3+ / 0-)
    Recommended by:
    Febble, TomP, George3

    I agree that the RAND methodology makes it sort of a swing state on steroids. It will be interesting to see how RAND matches up against the final national numbers. Probably leans O, but Romney certainly has solid representation in it.

  •  yeah, you're sick of polling (10+ / 0-)

    sure you are ;-P

    Good read of RAND.  I post it every day because no one else does. I think the poll is very informative and sets the boundaries (somewhere between RAND and Gallup) and I love it for trends (it mirrors the betting markets w/o the manipulation market).

    For both of us, the most interesting graph is 'who switched'.

    "Politics is the art of looking for trouble, finding it everywhere, diagnosing it incorrectly and applying the wrong remedies." - Groucho Marx

    by Greg Dworkin on Sat Oct 27, 2012 at 05:05:23 AM PDT

    •  The breakouts by voter characteristics (4+ / 0-)

      are nice too.

    •  I also find it useful (4+ / 0-)
      Recommended by:
      FredFred, ItsSimpleSimon, TomP, George3

      even with the limitations that may accrue b/c of its methodology.

      Where I might disagree with you just a bit is that given some of the known limitations of other polls, particularly in their likely voter model, my gut at this point is that Rand will wind up being far closer to reality than Gallup.   That is, the final result MAY be in between, but were I to take the difference at any moment, I think I would make reality at between 2/3 and 3/4 of the difference between the two, and not at the middle.  

      I have no scientific or mathematical basis for doing this, especially given the lack of a track record for this poll.  Consider it a working rule of thumb.

      And FWIW, if asked right now, this minute, how I think the final results will be, I am increasingly confident that the popular vote margin will be >2%, and lean towards around 4%.  While there is no direct correlation with results in individual states, for similar reasons I am also increasingly confident that Obama will have >303 electoral votes -  that is, for me the only remaining questions are FL, NC, 2nd CD in NE and maybe AZ.  

      Still want to see what happens for another 5 days or so.  That includes what if any impact the storm has, including how Obama and administration are perceived as responding.  Wondering about possible impact Sandy will have on campaigning -  Romney has already cancelled his Sunday Virginia Beach rally, for example, and I expect Obama will follow suit and cancel his event with Clinton for Prince William County.

      I like the additional insight the Rand poll gives, and believe it provides another dimension that helps understand impact of events like debates in greater depth than do snap polls.

      "We didn't set out to save the world; we set out to wonder how other people are doing and to reflect on how our actions affect other people's hearts." - Pema Chodron

      by teacherken on Sat Oct 27, 2012 at 05:14:52 AM PDT

      [ Parent ]

  •  heh (9+ / 0-)
    Please take RAND with the other 7 trackers (average/aggregate), use it for trends. The truth lies between RAND and Gallup.
    @DemFromCT via TweetDeck

    "Politics is the art of looking for trouble, finding it everywhere, diagnosing it incorrectly and applying the wrong remedies." - Groucho Marx

    by Greg Dworkin on Sat Oct 27, 2012 at 05:18:15 AM PDT

  •  RE "momentum" - while I'm on my soapbox... (8+ / 0-)

    I do hate that metaphor.  It has nothing to do with anything.  Most time-series data have a negative autocorrelations - the higher a data point is, the more likely it is that the next one will be lower.

    Basically, "regression to the mean".  However, sometimes you get positive feedback loops, where the fact of being high increase the probability that the next time you'll be higher still.  But that's not "momentum" - that's acceleration!

    It seems to me that what the RAND shows mostly is a negative autocorrelation - what goes up tends to come down, eventually.  However, it may also reflect a bit of positive feedback in the form of enthusiasm - people like to back a winner.

    But I suspect that in such a close race, that's not happening much, and the overall Obama rise simply reflects an underlying reality that he's the better candidate.

  •  The RAND poll's track record for being highly (0+ / 0-)

    responsive to political events in the race is undeniably very interesting.  On the face of it, this is jump the President has received in it really does seem to indicate good things should be coming.  But if a more significant bump for Obama doesn't materialize, at the very least, in RV  numbers, it's probably not worth putting too much more stock in this poll.

  •  How attentive voters.... (5+ / 0-)

    ...could fluctuate between Obama and Romney is beyond me. How anyone could have any working brain matter and  find something acceptable about the Romney candidacy is also beyond me. Of course I don't understand what makes people a-holes either.

    "Good to be here, good to be anywhere." --Keith Richards

    by bradreiman on Sat Oct 27, 2012 at 05:28:53 AM PDT

    •  Well that's where the breakouts can be (0+ / 0-)

      interesting.  The other day there was one by income, and what struck me was how little preferences had changed amongst voters earning <$50,000 since the conventions.  All the volatility was amongst people earning <%50,000.  Which isn't that surprising, I don't think - those are the people for whom it really matters.

      And Romney's promising to create more jobs.

      •  was that first < supposed to be a >? (0+ / 0-)

        I didn't see the chart.

        I'm not sure that I buy your argument here, but then I'm not even positive what empirical result is under discussion. :)

        Election protection: there's an app for that! -- and a toll-free hotline: 866-OUR-VOTE
        Better Know Your Voting System with the Verifier!

        by HudsonValleyMark on Sat Oct 27, 2012 at 06:38:51 AM PDT

        [ Parent ]

        •  Yes it was :) (0+ / 0-)

          I've always been useless with left and right!

          I should have taken a screenshot of the breakout.

          •  OK, just checking! (0+ / 0-)

            I suppose there are at least two ways of looking at this. One is that -- almost the exact opposite of Romney's gibe about the 47% -- the people who are economically vulnerable are most receptive to his promises about jobs, and their vote preferences vary depending on how clearly those promises come through, relative to other campaign messages.

            The other is that low-income folks are (on average!) less politically aware -- in part because they aren't at all convinced that political outcomes really matter to them -- and more likely to be swayed by atmospheric perceptions of which candidate seems stronger at the moment.

            It's possible to combine those lines of argument -- probably crucial, really. Voters do care about jobs, and voters do care about strength. But to get things right, we would probably need to know a lot more than we do. So it's hard to get past a battle of broad brush strokes. :)

            Election protection: there's an app for that! -- and a toll-free hotline: 866-OUR-VOTE
            Better Know Your Voting System with the Verifier!

            by HudsonValleyMark on Sat Oct 27, 2012 at 07:28:52 AM PDT

            [ Parent ]

            •  I saw a video on TV this morning (1+ / 0-)
              Recommended by:
              HudsonValleyMark

              Of a man in a Romney T-shirt, at a Romney rally, in tears of desperation over unemployment.  And Romney was there saying we need change, I will cut taxes by 20% across the board, create 12 million jobs, and get the country working again".  

              I'd vote for that, if I believed it.

              •  yup (0+ / 0-)

                Although I agree with many of the criticisms of Obama's performance in the first debate, I think your point here suggests the limits. IMHO, no matter how tough Obama was, no matter how much he made eye contact, whatever, it would have been hard to "rebut" the thing where Romney looks into the camera and says, "Trust me, I can fix this for you."

                Election protection: there's an app for that! -- and a toll-free hotline: 866-OUR-VOTE
                Better Know Your Voting System with the Verifier!

                by HudsonValleyMark on Sat Oct 27, 2012 at 07:55:16 AM PDT

                [ Parent ]

                •  The weird thing to me (0+ / 0-)

                  is that essentially, that's all he's saying.  No other argument, just "I'm a businessman, I know how to balance budgets and create jobs".

                  •  smart, isn't it? (1+ / 0-)
                    Recommended by:
                    Febble

                    It's not as if he could expect to win on the merits. I mean, he needs to have a bit more in his repertoire than "trust me, I'm rich," but this ain't no Oxford Union, as you know. And most of the traditional news organizations are feeling a bit desperate as it is, and in no mood to risk alienating a large fraction of their audience by saying that one major-party candidate is actually a reptile. (There are other plausible attributions of motive, which coexist with fear, but I think fear probably suffices.)

                    Election protection: there's an app for that! -- and a toll-free hotline: 866-OUR-VOTE
                    Better Know Your Voting System with the Verifier!

                    by HudsonValleyMark on Sat Oct 27, 2012 at 09:04:13 AM PDT

                    [ Parent ]

    •  Polls should have an a-hole demographic (1+ / 0-)
      Recommended by:
      Febble

      I think we pretty well know what the results of that would be.

    •  different voters are polled on different days (0+ / 0-)

      So it's not the same voters changing their minds, it's a different subgroup of the larger whole group.

  •  Great job on this post and your chart, Febble! (2+ / 0-)
    Recommended by:
    Febble, TX Freethinker

    I too have been watching Rand with interest this election cycle and I'm anxious to see how all this shakes out post election.  

    As we express our gratitude, we must never forget that the highest appreciation is not to utter words, but to live by them. John F. Kennedy

    by JaxDem on Sat Oct 27, 2012 at 05:41:06 AM PDT

  •  Rand is one of the best ways to poll (3+ / 0-)

    As you point out correctly.

    Complaints against it

    1. It's been untested: actually it's been tested in a peer review article

    2. It's new. So what ras and Gallup have been around a while and they both suck. Watch ras switch 3 days before election so their end results are close to the actual election.

    3. They attach prob. Weights to likely voters. But this is a much better system than ruling out likely voters on ore existing different conditions depending on pollsters whims

    4. They use the same sample. As long as the sample is representative there's nothing wrong with this. If the sample changes you don't know if the results are bc of trend changes or sample changes

    "Four seconds is the longest wait " -Sleater-Kinney

    by delphil on Sat Oct 27, 2012 at 06:00:56 AM PDT

    •  one thing about 4. (3+ / 0-)
      They use the same sample. As long as the sample is representative there's nothing wrong with this.
      What is arguably "wrong" is that people who are asked every week how likely they are to vote, and how likely they are to vote for each candidate, aren't likely to be quite as inattentive as the typical inattentive voter. So the measure begins to influence what is being measured.

      I don't think that is a fatal objection, by any means. I'm glad to have both approaches in use.

      Election protection: there's an app for that! -- and a toll-free hotline: 866-OUR-VOTE
      Better Know Your Voting System with the Verifier!

      by HudsonValleyMark on Sat Oct 27, 2012 at 06:36:19 AM PDT

      [ Parent ]

    •  Also no poll is likely to be representative (1+ / 0-)

      But this one propagates the initial error, whereas with new samples they (should) cancel out.

      That's why I subtracted the initial margin from all the margins in the plot, so that the absolute values are meaningless - my bars just show changes from the start of the poll.

      •  You have this completely backward (0+ / 0-)

        Errors don't "cancel" out in cross section polls. In theory sample error should be independent across polls. In practice poll construction and administration introduces systematic biases that do not cancel out - hence all those discussions of cell phone vs landline and non response bias etc.

        Until we get actual election results no one really knows whether their random sample is actually representative of the electorate, so each pollster makes a judgement about what combination of methods will yield a "random" sample. That accounts for most of the systematic bias in the traditional polls' RV results. Then they make a judgement about voter participation which is tantamount to guessing who will win and impose that through a likely voter screen that has no methodological basis at all.

        Rand weights their 3500 person sample by demographics and 2008 voting behavior, so there is no chance that they sampled too many men or women or Latinos or whites or democrats or republicans. That makes the sample here vastly more reliable than even the average of traditional polls which don't have past voting behavior as a measure on which to weight and are relying on question random sampling methods to ensure the sample is representative.

        All the margin of error in Rand comes from the danger that the individuals in the sample that represented unique demographic and partisan subpopulations are not in fact representative. In 2008 nearly 1% of voters were blacks who supported McCain. So the reported Rand results will be weighted so that 1% of the panel are black McCain voters. That may only be a couple individuals who might be all voting Romney, or all Obama or 67-33 Romney( 3 split2-1) . Unfortunately changes in the reported preferences or intention to vote of these under sampled populations or dropping in an out of the seven day sample will dramatically change the election forecast. The composition effects from high weight individuals dropping out and reentering are evident in the one day spikes that happen so frequently.

        Bottom line is that the trend and the initial ( or any other daily) election forecast are equally subject to errors in the sample particularly of potentially under sampled populations. The greatest strength of the poll is that it is calibrated to past voting behavior, which eliminates the biggest systematic biases in traditional polls.

        •  Cleary non-sampling error doesn't (0+ / 0-)

          cancel out - sorry I should have made that clearer.  My point was simply that with cross-sectional polls, the sampling error should cancel out over time, whereas in this poll it won't.  

          Rand weights their 3500 person sample by demographics and 2008 voting behavior, so there is no chance that they sampled too many men or women or Latinos or whites or democrats or republicans. That makes the sample here vastly more reliable than even the average of traditional polls which don't have past voting behavior as a measure on which to weight and are relying on question random sampling methods to ensure the sample is representative.
          Well, possibly.  Possibly not. "2008 voting behaviour" is itself a biased measure (unless they actually have contemporaneous records) - in that there is good evidence that people tend to remember what comports with their current position.  And all polling samples are demographically weighted.

          But I agree that their sample is probably a better-than average one (albeit affected by the very fact of repeated polls), just as I think that I have been conservative to Obama by baselining the margin at zero.  I guess my point is that even with heavy provisos and conservative assumptions, the news is good :)

          •  They do have contemporaneous info (1+ / 0-)
            Recommended by:
            Febble

            For a big chunk of the panel. Plus they asked again in spring of this year and could compare responses to check for any systematic bias. There is a small group that wouldn't respond ( which seems odd cause they are agreeing to disclose their voting intent) and of course new and non voters that are again weighted to match the population share.

            In various papers they argue quite convincingly that the largest source of polling errors is sampling error on partisan ID. Obviously you can't weight for current ID cause that is so strongly correlated with the variable of interest, but weighting by historical voting behavior improves both precision and accuracy enormously.

  •  Obama intention to vote all time high (1+ / 0-)
    Recommended by:
    Febble

    They track the participant's intention to vote, as today it is at 84.5%, which is an alltime high for this sample.

    Of course these folks are more engaged than most, but that is still a great sign.

  •  Meanwhile Raz out with a whopper this AM (0+ / 0-)

    R 50  O 46

  •  ras pretty much sticking with Romney (1+ / 0-)

    Getting late in the game for ras to fall in line. either Romney is going to win the popular vote or Ras is giving up pretending to be independent.
    I guess late next week Obama can start making a major comeback and  Ras will have it tied election day so that way he can't lose.

  •  I'm hoping it's right . . . but (1+ / 0-)
    Recommended by:
    Febble

    one of the things in this election in particular is that there seems to be a significant correlation between how informed voters are and how they're voting, with the high information voters skewing Democratic.  

    And in this environment, when RAND or anyone else does a strict experiment, as opposed to your typical opinion poll quasi-experimental design, with the same 3,500 people over several months, it's almost impossible to stop them altering their behavior as a result of the experiment itself.

    What I'm saying is I think there's a real possibility that RAND subjects are paying more attention to the election in general as a result of their participation, and, over time, it's gradually skewing them in Obama's direction.

  •  Nice derivative plot - The RAND method (1+ / 0-)
    Recommended by:
    Febble

    allows you to look at the attitudinal shifts in a fixed population. The one question I'd have about their approach is whether it will, in the end, provide a better measure of the electorate's mind as a whole than traditional polling.

    One of the failings of traditional polling is that they have low response rates. RAND's panel has a much higher response rate - the panel members have already allowed that they are willing to answer the same questions, week in and week out.

    Does that make RAND panel members unrepresentative of the population in full? On willingness to respond, yes. On opinion presented in responses, unknown.

    On other factors appropriate to generating a representative sample RAND has surely done their homework. Their methodological paper lays out their goals for sample characteristics.

    RAND overcomes one key implied difference - that internet access is required to participate - so some demographics would be harder to include, "off-gridders" of one type or another. RAND accomplishes this by first using traditional dialing to solicit participation - those agreeing to take part who do not have internet access are afforded the same through RAND. Nice bonus really.

    In a nutshell, RAND finds ways to make responding easy - an active rather than passive approach.

    More details on RAND's life panel history are found here.

    The response rates are, as alluded earlier, stellar- and attrition is relatively low.

    Last, on demographics, RAND is very transparent on how they collect demographic information. Weighting methods are clearly described.

    Which leads me to the last very big difference between RAND ALP and traditional polling. They are not trying to assess or weight for party affiliation - at all - as far as I can determine. This makes them very different from virtually every other pollster I can think of.

    Will it make them more successful? Does it introduce a source of inaccuracy - or eliminate one that a pollster weighting for party-ID carries?

    All in all a very interesting method.

Subscribe or Donate to support Daily Kos.

Click here for the mobile view of the site