The mailbag
Gerrymandering makes for interesting mail! Here are some excerpts from activists, a journalist, political scientist, and a few redistricters. ...
The week ahead for PEC
A preview of the week ahead. I plan to have:
(1) An expert briefing on wireless phone sampling, courtesy of Ed Freeland, director of the Princeton Survey Research Center.
(2) A Presidential election prediction challenge to other forecasters.
(3) House and Senate updates, of course. That’s where the action is.
Suggest more topics below. Have a good weekend!
Topics:
Sam,
Can’t wait to see what you’ve come up with. On another front, I’ve visited 538.com to see the latest from Nate Silver. To my dismay, his forecast model has dropped BO because of the latest USA manufacturing report. He’s really getting hit hard by posters who write that although the report is bad it trumps the most recent polls showing a widening for BO including CBS/NYT polls where he’s taking a slim lead over who could best handle the economy. I know you’ve said that he uses different variances in his models that are more noise than anything. (Not sure if that’s exactly what you’ve said but somewhat along those lines). At one point when the job numbers were somewhat good in July and the DOW skyrocketed he made adjustments to reflect that. The Market is still going up but no mention of it at all. Followed him the last election but I’m starting to have my doubts. Not sure if it’s because he’s joined NYT or what but something is not sitting well with me with him this time round. What’s your spin? Also, what’s your opinion on this by Dick Morris from RCP?
Thanks,
http://www.realclearpolitics.com/articles/2012/09/14/why_romney_will_win_115452.html
The nowcast is still at 8.2 for Romney. Why is this? Because the forecast uses simulation runs and umm….like 7 “economic indicators”.
The nowcast is polldata only, its a snapshot.
Dr. Wang says the economy is already baked into the polls.
I agree.
What Obama demos are going to switch to Romney if the economy tanks, and what Romney demos are going to switch to Obama if the economy booms?
Ans: indistinguishable from zero.
I think Nate’s economic indicators are just adding noise. In a way, Nate is counting the economy twice.
Hello Iseeurfuture,
thank you for link to the Dick Morris article. I just read it and Mr Morris, the way I understand has 2 points.
First point – in second paragraph he states:
“It appears that the bulk of the Obama post-convention bounce has been in blue states where his left-oriented convention stirred up the enthusiasm of an already committed group of voters.”
but on:
http://www.colleyrankings.com/election2012/
I see that
Flip: FL (29) from Toss-Up to Obama
Flip: VA (13) from Romney to Obama
Another point he is making is that people who say they’ll vote for Obama might not show up, but Obama is winning also in polls of likely voters.
have a nice day,
vladimir
xxxxxxxxxxxxxxxxxx
How about something on the “likely voter” models that everyone is breaking out? What are they? How do they differ between pollsters? What systematic uncertainty is associated with them?
Also, how does demographic weighting work when some subsets have close to zero entries? This could tie in with your cellphone sampling discussion.
A likely voter is someone who says they are going to vote when polled.
Most pollsters also ask for a level of intensity, like 1 to five or 1 to 7.
So they can run kinetic multidimensional scaling test perhaps.
lol.
But you see the problem with cell onlies and people who dont answer their landlines for polls (like you, Amitabh, I suspect). How to tell if someone is a likely voter if they are never asked if they are likely to vote?
A captive expert! Very exciting, Dr. Wang.
I also love this site, and especially the new comment threading. Where else can one read about ferromagnetics and cognitive neurosci on one blog?
wheeler cat: I believe that the selection of the ‘likely voter’ varies. Confessed likelihood is one method. I have heard of some inquiring at to voting in the last three presidential elections (assuming their old enough). Also, some take demographics in account. My personal experience was just after I had had a couple of weeks discussion on polling in one of my sociology classes. So, I paid attention to the questions asked and we discussed the questions in class. Otherwise, as Dr Wang’s book probably says, there is no way that I could remember the questions. They asked name, race, sex, and who I was voting for. Yet, when the poll was published, they had likely voters and all voters. Then, after the results, we discussed how they might have determined likely voters (the concensous, I don’t remember).
I just wanted to state for the record, in case the barbarian hordes arrive in the comments section here one day, what a true pleasure it has been to read the commentary on this site by Dr. Wang and all of you others. It does this dyed in the wool geek’s heart good to read such quality insights, day in and day out. I salute you all.
Sam, just wanted to say thank you for highlighting the importance of state races. I am passing these links on, encouraging donors to invest locally in the NC General Assembly. Unfortunately, recent experience in our state has illustrated how much damage can be done in only 2 years once the other party gets control. Assault on women’s rights, the environment, and the anti-marriage-equality-amendment all came down the pike since 2010. And that happened even with the courageous vetoes of Democratic Governor Bev Perdue. We’re working hard to undo that damage but it would have been much better if Dems could have kept control in the first place. I hope other states can learn from our experience!
As a life-long Democrat and resident of NC since 1992, I will say that my perception of soon to be former Democratic political control is a bit more dichotomous. I agree on the unfortunate turn with social issues; however, fiscal management has been notoriously awful especially during the Hunt administrations.
Having one party in power for the most part since Reconstruction has led to episodes of ineptitude and corruption.
Actually the Republications in the statehouse have worked closely with the Democrats this past year to faciliatate a pro-job and pro-growth agenda. Too bad we can’t clone cooperation on a national level.
Fortunately, the State is on track to be more progressive as our demographics trend blue.
Suggestion 1:
If the information is available, I would like to see a more detailed analysis of the shifting resource allocations of PACs and SuperPACS, and what these tell us about the race.
You’ve touched on this, but perhaps there is more to be said?
Suggestion 2:
Previously, you’ve linked to articles that in frightening detail describe how the GOP poured massive resources into winning state legislatures and governorships in 2010. Those investments paid off hugely! And as others have pointed out, the Republican Party has used those enlarged state-level platforms to try to dictate and drive the national debate –– with impressive success.
What is the situation today? And what are the overall prognoses? Are we likely to see a repeat of 2010 and an expansion of Republican state-level control, or are those efforts being countered in a meaningful way?
Suggestion 3:
Most of all, I would like to see an analysis of the robustness and integrity of the election process itself. I would be happy to detail what I mean by that. (Perhaps this latter topic might be even more appropriate a few weeks down the road?)
PS. Dr Wang, do you have graph plotting changes in the Meta-Margin over time? Perhaps labelled with key events?
I apolgize for being off-topic but I’m mesmerized by PEC and read it more than every day. I’m glad you’re nt a catblogger. You really made me care about the state races, and yours is nearly the only political opinion I care about. Do you think that there’s any chance that the President if reelected would nominate Caroline Kennedy to the Supreme Court, and if he did, would she be disqualified by the fact she is actually qualified [having written several books on Constitutional law and rights especially privacy] and therefore too much is known about her? I ask because if there’s any use to I might start a petition or hopefully do something else youll explain is more useful. But maybe I’m just wrong-headed, like when I think the candidates should be asked about killing the TSA. Elizabeth Farnsworth and her husband told me on Monday that these Occupy-like topics are campaign poison.] I’d really like to know what you think, as you know a lot about polling, and have a sense of public pinion which I do not. Answering my question would also give you a chance to tell everyone you wont answer more political question XD Thank you sir. My husband and I, former followers of 528, listen only to you.
I read that likely voter screens differ wildly between pollsters. One could just ask “do you intend to vote in Nov 2012?” but apparently that is considered far too naive.
Screening questions can be “Did you vote last time?” or “Can you tell me how to get to the polling place from your house?” Or some linear combination of such questions and demographic information.
My point is that likely voter screens are very model dependent. They start with an idea as to what sorts of registered voter is likely to vote. I cannot imagine how these screens can be entirely bias-free.
Ideally, the raw data would be available and you could apply your own screen (and even tighten or loosen the screen to see if biases were creeping in).
Sure, pollsters can look at the covariance of age, sex, race, polling history or whatever.
But this is where the problem starts.
Historically, cell-onlies and smartphone users are younger and not “likely voters”.
But cell-onlies are nearly a third of the electorate now and smartphone users are 47% for the first time.
This is an election of firsts.
Just a couple of very nuts-and-bolts suggestions.
1) Please resume the election-day prediction on the right hand side of the EV plot, or as its own EV plot. I’d like to see the variance shrink now that we’re approaching 40 days out.
2) If you’re going to add in a variance minimization technique, then please do it early, and let your readers know the details as far in advance as possible. Also, an explanation of why it’s needed near election day, but not far from election day would be helpful.
Thanks,
Marc
I was polled by the American Future Fund this week (on my cell). They asked me how likely I am to vote on a scale of 1 to 10. They also asked whether I voted last time and who for. That was the extent of the likely voter-ness of the poll. Saturday I had been polled by an unnamed source, who I assume was an Obama affiliated group, since I was contacted the next day to ask for volunteers. They had even less of a likely voter screen, only asking whether I would definitely vote, probably vote etc.
I live in Colorado too and I have never-ever been polled on my iPhone.
A week or two ago I was puzzled by a seeming discrepancy between the effect of likely-voter screens on national polls (where they seemed to make the result skew very much more Republican) and state polls (where nothing of the sort happened).
I’m now convinced it was probably just that the national polls were older, using pre-convention data. They seem to have mostly caught up.
I recently looked at a polling website that predicted democrats would lose control of the senate and remain in the minority in the house. This saddened me and prompted the following question that may be the basis for future discussions:
1. What is the historical precedent for presidential coat tails swaying house and senate races?
2. Are polls for senate and house races as accurate in predicting final outcomes? I would guess that most people are less well informed of these candidates positions, so the polls in these races may be more volatile.
Not gonna lie: the electoral challenge gets my blood going. Esquire recently asked in a poll who people thought would win in a Fist Fight–Obama or Romney. Most meaningless question ever–but I think if PEC is right (87+ percent for Obama) the race could use more excitement!
Looking forward to it–may InTrade take odds on the electoral challenge!
(My response to the Esquire poll http://politicalomnivore.blogspot.com/2012/09/and-in-this-corner.html)
I would much rather have a real choice between some reasonable, thoughtful positions. Imagine if Obama were running against Eisenhower or Rockerfeller, and they laid out their visions and ideas in full and trusted in citizens.
As it is, I have not had to think who I am voting for in such a long time, I wonder if I am still capable.
I have more choice in consumer goods than I have in political parties.
I was recently on shift at the Large Hadron Collider with a colleague from an eastern European country. He was old enough to have seen Soviet forces crush demonstrations when he was a child. Now he and his friends could sit and discuss policy proposals from several political parties (each with some sensible points and some pandering, to be sure).
One would have thought the US would go that way. A losing political party would go towards the center to regain support. And that should work, but for the primary system, which exerts considerable force away from this equilibrium.
PS: I guess because I didn’t go to a Division 1 school I didn’t develop “the sports teams from my educational institution are superior to the sports teams from your educational institution” mindset.
PPS: However, we do have the Beaver Cup, where we beat Caltech to a bloody pulp more often than not (sorry Sam).
“A losing political party would go towards the center to regain support.”
Amitabh, that is actually what is happening now. The demographic timer is going to force the GOP to begin to accommodate minorities as the vote share of white conservative christians declines.
It just happens very slowly, by design if you will.
The founders and framers built damping into the system…they most feared demagoguery and extreme oscillations in the system. For example, George Washington once said the Senate is the saucer that cools the “hot tea” of the House.
Boy howdy was he ever right.
And you cant have a ‘real choice’ between reasoned positions and have a “freed” market in information at the same time, because of asymmetrical ideology and the bell curve of IQ. Consumers of information will only pay for information that agrees with their worldview.
And its certainly possible that the Grand Experiment will be a fail, despite or perhaps because of how carefully the founders built the system.
But consider– citizenship is the only requirement to vote. You don’t even have to be able to read or write.
I think that is still huge.
Wheeler’s Cat, that is a high-minded statement. However, note that being able to read or write is also the threshold for offering comments on YouTube videos. And just look at those.
Honey Boo Boo for Prezident!!1!
Iseeurfuture, you should not worry about the transient ups and downs of the NYT model, for which there is no market pressure to make less noisy – they want eyeballs. There is nothing new to be said about it.
Take your stress and get on the treadmill for while. Or better yet, go find a close House race to offer your volunteer services.
Regarding LV and cell phones: tomorrow Ed Freeland will join us. Have at him!
Olav, PAC/SuperPAC spending is a good suggestion. OpenSecrets.org is the closest thing I have found to a one-stop compilation. Unfortunately, the NYT seems focused on Presidential race funding, which to me is no longer interesting.
In regard to election integrity, not to be Pollyanna-ish, but the US process has much more good than bad. A possible exception is the new wave of voter-ID laws. Some of these will probably be reviewed by the Supreme Court next year, which has shown an eagerness to re-interpret the Voting Rights Act of 1964. See scotusblog.com and Rick Hasen for more. Those laws are probably good for an approximately 1% benefit for Republicans where they are fully implemented. The Republican strategy is one of winning near-tied situations, and this is an example.
Plotting the Meta-Margin: click on the EV graph at right for a pleasant surprise.
Ms. Jay Sheckley, thank you for writing. I am not a Supreme Court watcher; for that see SCOTUSblog.com. They are amazing. I will get out on a limb here and say that Caroline Kennedy’s odds for joining SCOTUS are not much better than mine.
Marc: thanks, taking it under advisement.
Mark in Missouri, In Presidential years, the party holding the White House is somewhat advantaged. For a recent article on the subject see this work by Bafumi, Erikson, and Wlezien.
Generally, Election-Eve Senate and House polls are accurate. See my retrospectives of 2008 and 2010 in the left sidebar. Somewhat worse are midterm House preference polls, which can be flaky, probably because it is so hard to estimate likely voters when there is hardly any news coverage and attention is low.
Amitabh, you have my every sympathy. Paul Starr was saying to me that he thought the Cold War kept the US internally focused on a shared factual narrative, and now it’s gone. I am not sure if that is the cause of our deteriorating national conversation, but the idea has a certain appeal. Before I spent a year in the Senate, my nostalgic view of past times was that I was Madly for Adlai. Afterwards I was a Johnson man.
In regard to why parties don’t move to the center, Amitabh and Wheeler’s Cat, I recommend Off Center by Hacker and Pierson. It’s mentioned in this rather good recent NYRB article on Our Weird National Politics.
MAT, Amy, and everyone: thank you all. It is a good discussion here, isn’t it. Remember this moment.
Truly I do not understand why this site isn’t swarmed with low information conservative trolls like Blumenthal and Silver.
I think they fear you, Dr. Wang.
As do we all.
😉
This may be outside the scope of this blog, but I would like to see in depth attention to geopolitics and demographics. To date, nobody has tackled that issue well (538 has only just scratched the surface), yet it’s one of the most useful aspects of political information there is
“… Suggest more topics below… ”
Hello,
well, I was waiting for somebody more knowledgeable and also better in English to suggest, … well, my suggestion is a list of mistakes (and/or contentious mistakes? 🙂 journalist, pollsters and aggregators do. I learned of many by reading this site, but would appreciate list of the fallacies … .
thank you for consideration,
vladimir
Dr Wang… could you please opine on the U of I model and methodology. I am confident that it is not correct, just not sure the exact reason why.
http://electionanalytics.cs.illinois.edu/election12/index.html
Additionally, I just wanted to say thank you for providing a great website loaded with useful information that is trustworthy.
Looking forward to the forecaster challenge… 332-206 barring any craziness from now until November. Romney may get slight bump if Ryan performs well in the VP debate; however, at the end of the day Obama squeaks out Florida, Virginia, and Wisconsin narrowly losing North Carolina. Nevada, New Hampshire, Michigan, Colorado, Ohio, and Pennsylvania firmly go blue even with suppression.
Thank you. Your assessment seems right.
At first glance they don’t seem all that bad. It’s a snapshot they give, not a prediction. It is not obvious what their poll selection rule is, or for conversion to a win probability. Their Bayesian component is unnecessary but undamaging.
I would like to talk about black swan events, if we could.
I am thinking more and more that the Klein anti-islam film is a black swan that is going to affect the campaign. I do not think this is a black swan that will help Romney– IPOF it is hurting him.
For one thing, I think Romney is going to lose another news cycle this week, which he can ill afford to do.
6 americans have died so far. That tends to focus the electorate’s attention.
“Heavily armed Taliban fighters on Friday stormed a strongly fortified air base in Helmand province where Britain’s Prince Harry is deployed, killing two US Marines in an assault the militia said was to avenge the anti-Islam video.
Six US fighter jets, costing tens of millions of dollars, and three refuelling stations were destroyed, clocking up unprecedented material losses for the Western military in Afghanistan where they have been fighting for 10 years against insurgents.
A total of 17 people have died in violence linked to the film so far, including four Americans killed in Benghazi, 11 protesters who died as police battled to defend US missions from mobs in Egypt, Lebanon, Sudan, Tunisia and Yemen, and the two US soldiers in Afghanistan.”
I think the forecaster challenge should include both an accuracy component and a (absence of) volatility component. Maybe you take the mean prediction of each forcaster over the last 6 months, and also calculate the standard deviation of each of them over that period.
People like those guys in Colorado just for data search
Oops.
People like those guys in Colorado who just predicted a particular number would get a great score for the low volatility but we have to wait and see if they were accurate.
Another forecaster might give numbers that swings up and down but averages the rigjt number.