Posted Nov 06, 2016 at 9:40 AM by Maurice Berger
Why are Nate Silver's numbers more bullish for Trump over at 538, in contrast to virtually all other polling analysis sites (including this one)? Here's an interesting take on why 538's election model differs from all
other polling analysis websites.
Ultimately, there is something strange about Nate Silver's methodology in this cycle, as he goes into
each poll and "adjusts" it's numbers, mostly in favor of Trump. Is he
hedging his bets: remember, he loudly pontificated that Trump could not get the GOP nomination. (We, at PollTrack.com,
began to predict a sharp trend towards Trump in August 2015.) Silver did so
not based on numbers but his personal opinion, as a pundit reading the
GOP electorate. (He also confidently claimed that the Cubs would not win
the World Series.)
So now he appears to be bending over backwards not
to underestimate Trump. (despite the fact that the general electorate is
FAR more diverse and complex than the Republican base). He's adjusting
the internals of each poll, thus possibly skewing the Electoral College
to show a closer race than other analysis sites (including PollTrack).
This doesn't mean that Silver is necessarily wrong, but these
adjustments are not explained. How does a poll from an organization with
an A+ accuracy rating (according to Silver himself) go from +1 Clinton
in Florida to +3 Trump after 538's adjustment of it. No explanation. And
frankly, illogical given the quality of the polling of the organization
PollTrack never goes inside polls to adjust their results. If we detect a clearly faulty methodology, we drop the survey from our polling
average. This is standard practice. To do otherwise is extremely
unorthodox. Maybe Silver is on to something, but
without explaining these "adjustments," his conclusions run counter to
virtually all of the other polling analysis websites
Posted Feb 26, 2013 at 8:36 AM by Maurice Berger
In an interview withStudent Life, Nate Silver suggests that he might end his election forecasting after the 2014 or 2016
elections should his projections "actually influence the elections'
outcome." Silver continues: "The polls can certainly affect elections
at times. I hope people don't take the forecasts too seriously. You'd
rather have an experiment where you record it off from the actual
voters, in a sense, but we'll see. If it gets really weird in 2014, in
2016, then maybe I'll stop doing it. I don't want to influence the
democratic process in a negative way."
Posted Nov 29, 2012 at 9:24 AM by Maurice Berger
Here is a fascinating analysis of how the Obama campaign gauged its relative strengths and weakness through internal polls. Mark Blumenthal focuses on the Obama campaign polling operation and notes they their view of the state of the race was local rather than national. Rather than taking nation-wide polls, the campaign
limited its surveys to 11 battleground states (Colorado, Florida,
Nevada, New Hampshire, North Carolina, Ohio, Pennsylvania, Virginia and
Wisconsin), conducting them at regular intervals throughout the campaign. Campaign manager Jim Messina says this gave him a deeper understanding of
"how we were doing, where we were doing it, where we were moving --
which is why I knew that most of the public polls you were seeing were
Posted Nov 06, 2012 at 3:05 PM by Maurice Berger
Last night, PollTrack posted its FINAL Today's Map and Tomorrow's Map. The final tall: 303-Obama/235-Romney. The last state called: Florida, which according to our poll averaging is very, very close.
Posted Nov 01, 2012 at 9:14 AM by Maurice Berger
PollTrack's aggregate of the national polls in the
race as of 10:00 AM EDT this morning shows a stable race. Once again, a word of caution:
several trackers have suspended operations due to Hurricane Sandy, thus
making it more difficult to access the current state of the race.
President Obama now
stands at 47.6% and Mitt Romney at 47.2%--for a lead for Obama of +0.4%.
Posted Aug 20, 2009 at 1:37 AM by Maurice Berger
At a recent statistical convention, the polling director of SurveyUSA, Jay Leve--one of the most accurate pollsters in recent cycles--had some shocking news for his peers: traditional polling methodologies, as we know them, may soon be doomed. Specifically, he was referring to the standard methodology of reaching potential voters--through landlines (and more recently, cell phones). Concluding his presentation, reports the National Journal Online, Leve summed up the problem: All phone polling, he said, depends on a set of assumptions: "You're at home; you have a [home] phone; your phone
has a hard-coded area code and exchange which means I know where you
are; ... you're waiting for your phone to ring; when it rings you'll
answer it; it's OK for me to interrupt you; you're happy to talk to me;
whatever you're doing is less important than talking to me; and I won't
take no for an answer -- I'm going to keep calling back until you talk
Yet, as it now stands, the current reality for pollsters is often much different:
"In fact, you don't have a home phone; your number can
ring anywhere in the world; you're not waiting for your phone to ring;
nobody calls you on the phone anyway they text you or IM you; when your
phone rings you don't answer it -- your time is precious, you have
competing interests, you resent calls from strangers, you're on one or
more do-not-call lists, and 20 minutes [the length of many pollsters'
interviews] is an eternity." Leve then concluded: "If you
look at where we are here in 2009 [with phone polling]," he said, "it's
over... this is the end. Something else has got to come along."
Posted Jun 03, 2009 at 2:22 AM by Maurice Berger
A reader, Derek Fields, writes the following to PollTrack's political director:
I haven't seen the specific wording of the Gallup poll, but I wonder
whether they ask any questions that separate the issue of legal
protections for "united" gays versus the religious overtones of the
term "marriage" My understanding is that when a pollster asks a
question that addresses the civil protections without introducing the
term marriage, support for gay unions jumps substantially.
Given the strong support generational divide in the poll numbers, I
would speculate that the days when a majority opposes gay marriage in
this country are severely limited.
Given the descrepany in recent polling, Derek is undoutedly correct. The very wording of a question within a survey--especially a controversial one--can dramatically alter the overall result. As for the second point, fresh polling absolutely backs up Derek's assumption about future attitudes about gay marriage. The recent Gallup survey, for example, reports that a "majority of 18- to 29-year-olds think gay or lesbian couples
should be allowed to legally marry, while support reaches only as high
as 40% among the three older age groups." The overall numbers for support of gay marriage amomng younger voters hovers around the 60% mark--a clear harbinger of future trends in the United States.
Posted Nov 07, 2008 at 12:34 AM by Maurice Berger
How close did polling organizations come in predicting the outcome of Election 2008? A few were right on the money. A few, such as CBS News/NY Times, Zogby, and Newsweek--each showing Obama with a double digit lead--were way off.
Dr. Costas Panagopoulos, Ph.D, of Fordham University's Department of
Political Science, undertook a detailed analysis of the results of the
top polling organizations. His findings indicate that not all surveys
were good at analyzing and reporting the relative strength of the
candidates. Professor Panagopoulos writes: "I analyze poll estimates
from 23 polling organizations. Four of these polls appear to have
overestimated McCain support (indicated with a * below), while most
polls (17) overestimated Obama strength. Pre-election projections for
two organizations’ final polls—Rasmussen and Pew—were perfectly in
agreement with the actual election result (**)." (The list follows
below.) One problem with this analysis, though, is that it was completed before final results were in. According to the Associated Press, Obama's final number has inched upward slightly, to nearly 53%, bringing the final total to 52.7% to 46.0%. In this case two other pollsters actually tied for first place: Ipsos/McClathcy and CNN/Opinion Research (both tied at 6th place in the Panagopoulos analysis). The more important conclusion from the Fordham survey, however, is not its top numbers but the idea that most pollsters overestimated the Democrat's support. In the past five cycles, this skewing towards blue has tended to be the case. Why? Perhaps because voters who "refuse" to be interviewed fall into demographic categories that favor Republicans--older, male, white, and rural. In this cycle, one other factor may also be at play: Democrats were FAR more enthusiastic than Republicans and thus were less equivocal, tended to have decided early, and were more willing to share their views with pollsters. Given the tendency of the least accurate pollsters to overestimate Obama support, the so-called "Bradley Effect" may have been operative as well. For Dr. Panagopoulos's analysis click here.
1. Rasmussen (11/1-3)**
1. Pew (10/29-11/1)**
2. YouGov/Polimetrix (10/18-11/1)
3. Harris Interactive (10/20-27)
4. GWU (Lake/Tarrance) (11/2-3)*
5. Diageo/Hotline (10/31-11/2)*
5. ARG (10/25-27)*
6. CNN (10/30-11/1)
6. Ipsos/McClatchy (10/30-11/1)
7. DailyKos.com (D)/Research 2000 (11/1-3)
8. AP/Yahoo/KN (10/17-27)
9. Democracy Corps (D) (10/30-11/2)
10. FOX (11/1-2)
11. Economist/YouGov (10/25-27)
12. IBD/TIPP (11/1-3)
13. NBC/WSJ (11/1-2)
14. ABC/Post (10/30-11/2)
15. Marist College (11/3)
16. CBS (10/31-11/2)
17. Gallup (10/31-11/2)
18. Reuters/ C-SPAN/ Zogby (10/31-11/3)
19. CBS/Times (10/25-29)
20. Newsweek (10/22-23)
Posted Nov 03, 2008 at 3:22 AM by Maurice Berger
Throughout the day, PollTrack will be providing analysis about the three hidden variables that could effect both turn out in and outcome of tomorrow's election:  Party weighting in polls.  The youth vote.  The so-called "Bradley Effect." Factor #1: One Thing to keep in mind about the today's final numbers--especially is VERY close races--most public opinion surveys in this cycle have tended to weight the party affiliation of likely voters in a way that skews to the Democrats by an historical degree. NBC/WSJ this morning gives the Democrats a +10% advantage in its national numbers this morning. Such figures suggests an historical realignment of the electorate that is virtually unprecedented over the past fifty years. If the Republican turnout should be greater than these polls suggest--and as a few surveys believe--the race could actually draw closer, especially in states that are already very close at this point, including Florida, North Carolina, Virginia, and Ohio. Will Obama benefit from a record turnout of Democratic voters? If so, he could win an electoral landslide, if not, things could get a bit closer.
Posted Oct 22, 2008 at 6:16 AM by Maurice Berger
Today's PollTrack average of the daily tracking polls shows a slight uptick for Obama: 49.1% to 43.6%, +5.5%. What is interesting about this averaging is that the distance between McCain and Obama narrows even more when the Zogby survey results--which have been extremely erratic over the past few weeks--are dropped: 48.5 % to 44%, giving Obama an aggregate lead of 4.5%. Of the six polls in our sample, five show the race stable (Rasmussen) or tightening slightly (Gallup, Hotline/FD, IBD/TIPP) and one rates it a virtual tie (GWU/Battleground), with Obama up by +2%, 49% to 47%. Combined with the high number of still persuadable voters, the race remains somewhat competitive, with the decided edge going to Obama. The periodic surveys are equally inconclusive, with Obama registering as little as a 1% lead (Associated Press/GfK) to as much as a +10% lead, NBC News/Wall Street Journal. The probable reason for this variation: the race remains close among voters who are certain of their choice. The more polls include "leaners," the greater benefit to Obama, who now leads with persuadable voters (who say they may still change their minds).
Posted Oct 17, 2008 at 2:13 AM by Maurice Berger
One way pollsters process raw data from samples is to filter it through party weighting models. In other words, a model that organizes voters by party and then weights the sample to reflect the percentage of likely voters from each party (as well as independents). With Democrat enthusiasm up this year, most pollsters give the party a considerable edge. In Rasmussen's weighting, for example Democrats outnumber Republicans 39.3% to 33.0%. For Zogby, it's closer: 38% to 36%. This weighting, in part (but several others factors are also at play), is responsible for the large swing in national numbers, from an Obama lead of +2% to +14%. PollTrack wonders: as national results are clearly narrowing, is Republican interest in the election gaining on Democratic? While Obama's campaign has fired up certain demographic groups --African-Americans, single women, young voters, for example--what of the traditional Republican constituencies: the over 65 set, Evangelicals, conservative Christians, and older white men? The latter groups tend to have exceptionally high turn out, literally making the difference for George W. Bush in 2004. There is anecdotal evidence that Evangelical voters, for a range of reasons--from anxiety about Obama to excitement about Sarah Palin--are growing increasingly enthusiastic about the Republican ticket. While some periodic polls (like CBS News/New York Times and Pew) show a very large lead for the Democrat, are these surveys underestimating the potential turn out of groups that--as a rule--vote in consistently and often extraordinarily numbers? Pollster John Zogby notes: "What troubles me is when I see some of my colleagues have 27% of the
respondents that are Republicans. That's just not America, period. [Party
affiliation fluctuates over time] it doesn't change "day-to-day, and it
never fluctuates by eight points in a short time period." Will the 2008 election break the mold--resulting in an unprecedented jump in Democratic turn out--or will Republican and conservative voters also show up in significant numbers, thus drawing the race much closer (especially in battleground states that already tilt Republican, such as MO, NC, FL, NV, CO, and OH)?
Posted Oct 16, 2008 at 1:02 AM by Maurice Berger
With one poll showing a slight Obama advantage in the state--and others indicating a McCain lead--West Virginia is turning out to be difficult to read. PollTrack continues to rate the state "Leaning Republican" on Today's Map. The Public Policy Polling organization thinks it has the answer: the difficultly of securing accurate and representative polling samples in the state. PPP writes: "If someone can get me a random sample of people who voted in the 2004
general election, 2006 general election, or 2008 primary in West
Virginia then we will poll it. Concern about being able to get a sample
of sufficient quality there is what makes us, and I'm guessing other
companies that do registration based sampling, hesitant to poll there.
That's not a problem with most other states."
Posted Oct 15, 2008 at 2:08 AM by Maurice Berger
Another reason for the broad variations in polling results (beyond differences in polling models and methodology) is the unusually dramatic, indeed traumatic, news cycles of late. The economic meltdown has injected a big dose of uncertainty and fear into the emotional lives of voters. And nervous voters tend to make impulsive or tentative political decisions. As John McCain's pollster Bill McInturff observes: ""The
financial tsunami has produced one of the most difficult and volatile times to
conduct polling in modern times. During these uniquely volatile
last few weeks, I have seen as much day-to-day movement as I have witnessed in
my 20 plus year career as a pollster." The erratic polling engendered by the "financial tsunami" may persist as long as the economic crisis in first and foremost on the minds of voters.
Posted Oct 15, 2008 at 1:21 AM by Maurice Berger
It looks like polling organizations are having difficulty determining likely voters this cycle. With so many newly registered voters--as well as a significant increase in younger voters during the primary season--some pollsters worry about using classic models and questions for determining a respondent's likelihood of voting. Will younger voters, for example, show up on Election Day, or--as in virtually every presidential cycle in recent years--will they stay home? Will African-American voter participation increase or stay the same? (Georgia election officials report an enormous black turn-out in early voting; Ohio reports the opposite: a relatively modest number of African American voters at this point.) Will newly registered voters show up? The problem is so daunting, that the Gallup organization is releasing three tallies in its daily tracking poll:  Registered: all registered voters,  Traditional Likely: likely voters determined by the "traditional" Gallup methodology, "which takes into account the intention to
vote in the current election as well as [respondents'] self-reported voting history," and  Expanded Likely: only voters who "self-profess likelihood to vote in 2008, [without factoring in] whether respondents have voted in past elections." Given the extraordinary spread in recent surveys--from Obama +14% (CBS News/New York Times) to Obama +2% (IBD/TIPP)--variations in models used to determine likely voters and voter enthusiasm may, in part, be to blame.
Posted Oct 07, 2008 at 5:13 AM by Maurice Berger
With today's Gallup and Rasmussen daily tracking polls showing Obama up +9% and +8 respectively and Hotline/FD and CBS News reporting a Democratic advantage of only +2% or +3%, it's clear that national polling is contradictory. The discrepancy might be due to party affiliation weighting or variations in likely voter models. It could be that some polls are picking up a trend others are missing. Whatever the reason, PollTrack will be watching these numbers very carefully.
Posted Sep 25, 2008 at 1:21 AM by Maurice Berger
Do pollsters under-represent younger voters by excluding from their samples voters who use cell phones exclusively? Pew Research seems to think so:
After including cell phone-only households in three recent polls, the organization notes "a virtually identical pattern is seen across all three surveys:
In each case, including cell phone interviews resulted in slightly more
support for Obama and slightly less for McCain, a consistent difference
of two-to-three points in the margin." PollTrack wonders: why then did many public opinion surveys during the Democratic primary season routinely OVER estimate Obama's actual support? And to what extent are pollsters' attempts to weight their samples to correct this deficit solving or adding to the problem? Another question: how do we evaluate Pew's reported discrepancy if the election is not until November and there are no hard results against which to gauge their polling estimates?