November 9 - 12:14 am PST
I've been live blogging this evening on Facebook to my friends. Here's my last entry:
Some final thoughts.
At this point it is almost certain our next president will be Donald Trump. Republicans will retain control of both the Senate and House. Donald Trump will appoint a new justice to the Supreme Court that Democrats will find difficult at best to block, and probably a replacement for Justice RBG when her seat opens up in a few years.
There are a few small bright spots in tonight's results. Democrats picked up seats in the House and Senate. Not many, but there were a few. And Sherriff Arpaio went down in Arizona, something that should have happened a decade ago.
Right now Democrats are in shock -- we really didn't see this coming. Some voices are already beginning on my Facebook feed and on tv to say this wouldn't have happened to Bernie. Would it? He might have won, or he might have lost even bigger. We'll never know.
What we do know is that barring a miracle which ain't forthcoming, Trump will be our next president. Our government will take a sharp tack to the right. It has happened before, most notably in 1980 and 2000. We survived those, we'll survive this.
There is a good chance Hillary will win the popular vote tonight. If so, it'll be the sixth time in seven elections this has happened, and the second time during that period Democrats lost in spite of it. You're liable to hear more calls for the abolition of the electoral college. There's no real chance of this happening. It would take an act of Congress, and right now the GOP isn't inclined to change anything that has worked for them.
Now let's go back 25-30 years. In the 80's it was the Reagan era. Partisanship was much lower, so Democrats were much more inclined to negotiate with him in Congress. AIDS was a horrific problem which was killing tens of thousands of people annually. And at the time you openly saw pundits and religious figures on television saying AIDS was a good thing because it was punishing "degenerates and drug users".
Because so many people were dying and there was so little hope, a community of direct action sprung up. You may have heard of one -- ACT UP, the AIDS Coalition To Unleash Power. It was extremely effective and achieved many of its goals. They chained themselves to the Golden Gate Bridge and blocked traffic in front of Wall Street. Always with a specific message they wanted to get out, and did wonders for raising awareness about issues no one wanted to care about.
ACT UP and the other direct action groups died out after the election of President Clinton. Activists looked at him as their savior and everyone thought the world would become a better place. It did... kinda, eventually, but it took years, far more than the Clinton's two terms in office. And many of their goals remain unfulfilled today.
During the Obama administration two groups sprung up. The first was the Occupy Wall Street movement, which didn't do much other than sit around in parks while smoking pot in drum circles. Frankly, they were a disgrace to activists everywhere. Sitting around doing nothing doesn't evoke change.
The second group was the Tea Party movement. They registered voters, lobbied their members of Congress, and challenged party members who didn't agree with them. They were surprisingly effective, and although they were eventually coopted by the Republican party and died out, their anger and legacy remains. That anger was much of what is ushering in Donald Trump tonight.
So at this point, three thoughts.
First, don't blame people who voted for Trump. They have many legitimate concerns -- for many reasons our government has failed to address them. This made them angry, and their anger was channeled into power.
Second, if all you do is sit around on Facebook bitching about the election, you're not doing much more to effect change than the members of Occupy Wall Street. If you believe change can be made through the system, do what the Tea Party initially did and register people to vote and become active politically. If you believe the system is inherently broken, then join the direct action groups that I predict will spring up under a Trump administration. Our country has a proud tradition of civil disobedience -- become a part of it.
Three and finally, there's a (false) story about the Pottery Barn motto -- "You break it, you buy it". Barring an astonishing upset tonight, the Republicans will now control the House, Senate, Presidency, and before long the Supreme Court. It means they now own the system. That means they can no longer sit in Congress being the party of obstruction. They want to repeal Obamacare? They now have that power. But they've never been able to decide upon what they want to replace it with. That will no longer be a choice. The new Congress is sworn in on January 3, 2017. Donald Trump will become president on January 20. It'll be up to them to decide what they're going to replace Obamacare with. And there are many members of their own caucus who don't know how to build anything -- they only know how to disrupt.
I predict you'll see an initial honeymoon for Republicans followed by some deep and bitter infighting within the party as they struggle over what they believe.
Either way, 2018 will come. It's a rough map for Democrats in the Senate -- they're defending the bulk of the seats as the Republicans were this time around. But if the GOP thinks they can just do nothing for two years and cruise to victory they have a rude awakening coming.
And there will be 2020. Democrats will start lining up next week to challenge Trump.
The world will go on. If you don't like what happened tonight, be a part of the change you seek. And that doesn't mean you can blame "Republican bigots" or "dumb Democrats who didn't support Bernie". That's what Occupy Wall Street would have done. Be better than them. The Tea Party eventually won Trump the presidency. What will you do with your anger?
November 8 - 2:40 pm
I can't find any more polls so this will stand as our final prediction -- Clinton: 323, Trump: 215. It's worth noting that 538 came up with the exact same forecast. If I read their article correctly, their number crunching suggests a 60% chance they'll have one or more states wrong. So we'll see how right we are this evening.
They are correct though, many of these dates are very close together -- Ohio, Florida, and North Carolina most notably. There is very little polling for the Maine 2nd CD, so that could easily be wrong as well.
The first precincts have closed in the presidential race. Under New Hampshire law, towns with less than 100 residents can open their polls at midnight and close them as soon as all votes are cast. Here are the results of several such towns.
Clinton led in the first town to complete it's voting: Dixville's Notch. There she received 4 votes to Trump's 2 and Johnson's 1. Mitt Romney also received one write-in vote. In Hart's Location, Clinton won the town with 17 votes to Trump's 14. Johnson received 3 votes. Bernie Sanders and John Kasich both received one write-in vote. These were not enough to overcome Trump's lead in Millsfield where he crushed Clinton 16-4.
The overall verdict? Trump has 32 votes in New Hampshire so far, Clinton has 25.
FYI, here's a list of the poll closing times for election day.
Survey Monkey updated their polls with one more day of data into their rolling poll, so I updated their most recent polls to reflect the new dates / numbers. There were no significant changes, just the occasional up or down by 1%, none of which had any impact on our projection.
November 7 - 8:37 pm
I am pleased to announce my "secret" project: the Election Map Tool. Yeah, the name sucks, but it's a great toy to play with on election night. It analyzes all of the possible maps leading toward the four likely endings of the election: Trump wins, Clinton wins, Trump and Clinton tie 269-269, and neither candidate reaches 270 votes because Evan McMullin wins in Utah.
The tool allows you to "call" the states as the evening goes on, either using your own imagination or when the tv networks do, at which time you'll see a decreasing number of electoral maps available for each scenario. Once all maps have been eliminated for rival scenarios the election is over, even if the networks haven't called it yet.
So have fun playing with the tool. Share it with your friends, and know there are no guarantees that it won't break down if the server crashes from overuse during the evening.
The Google polls are typed in. No unexpected flips. Google is insisting to the bitter end that Hillary will win Kansas. I'll doubt we'll ever get a public explanation of what went wrong with their prediction, but if they ever want to have any political business in Kansas they'll certainly have to fix the issue.
Either way, I've typed in so many polls today that our predictions are pretty well set in stone now. Only the Maine-2 and Nebraska-2 races are ones I have a high level of uncertainty about. There hasn't much polling about the former, and virtually none about the latter. In particular the Maine-2 prediction might be wrong.
538's podcast said they expect the presidential race to be called tomorrow night around 11:00-11:15. Basically, long enough for the 11pm poll closing states to be closed, checked for any aberrant results, and announced to the public. I, however, think we'll be able to call the race even sooner. See my next announcement (above) as to why...
The final Google poll is out. Sure enough, it has one more day of data than everyone else's. This means the three-day average will roll a day off from each. I'll type them in now -- let's see if their odd skews causes any unexpected flips.
I've updated the most recent Ipsos tracking polls of New Mexico, Maine, Hawaii, Delaware, Idaho, Montana, South Dakota, Arizona, and New Hampshire so that they would be included in today's average. Because they're updated and not newly added, they don't appear in the "new polls" listing on the right side of the screen. They're in our database though and you can check them out by looking at any state. Most notably, the Arizona poll shifted slightly more Trump leaning.
Update: Just did the same for their polls of Maryland, South Carolina, Washington, Oregon, Tennessee, Minnesota, Kentucky, Missouri, Colorado, Connecticut, and Wisconsin.
... and sure enough, the aberration in Arizona and Nevada worked itself out.
Along with a mess of Ipsos polls that just came in, I updated the information on a couple of old ones, most notably North Carolina. These are rolling polls that change a percent of their sample every day, so if I'm not careful I'll duplicate the data under consideration. To avoid this I updated the previously entered survey for a couple of states. Most aren't competitive, but the change in North Carolina gave Trump a slight boost in the state.
Polls are coming in fast and furious right now. At present our model finds Clinton to have a strong electoral win ahead of her, but there is one odd abberation: Nevada is going Republican while the more red Arizona is going Democratic. This defies most logic about the state, but perhaps it reflects the difference in their Hispanic populations? Or the model could be wrong, of course. More polls will arrive as the day goes on (and probably into tomorrow) so there's a good chance this won't be our final forecast.
Evan McMullin has faded a bit in recent polls since he peaked at 31% on October 19. In the last five polls of Utah he has trailed Trump by 12%, 5%, 16%, 15%, and 9%. Three-way races are notoriously difficult to poll, but the evidence suggests that he'll be duking it out with Hillary for second place.
For the second time this election season Trump is winning in our forecast. Pennsylvania flipped to Trump just now thanks to a Trafalgar Group poll with a large sample size. More data will be entered today though, and as I've said the forecast is highly volatile at the moment. It'll be interesting to see how it settles by tomorrow.
Update (just a few minutes later): That didn't last long. A 50-state poll from Survey Monkey has sent Pennsylvania back into the hands (or clutches, depending on your opinion of her) of Clinton, along with Florida, Arizona, and North Carolina. Expect some of these to tilt back when the final Google polls are added later or tomorrow.
November 6 - 9:47 pm
Could this make a difference? An Democratic member of the elector college from Washington says he won't vote for Clinton even if she wins the popular vote.
I'm 99.8% done with my new project. It's in beta testing now and I'm fixing a couple of final issues. I'll have it ready to present tomorrow.
In the meantime, I cleaned up the map display at the top a bit. The gray box is still extended to the right, but I've moved the widgets back to where they can be clearly seen. This'll be my patch to see the site through election day.
Otherwise, lots of polls being added are labeled as "final", meaning no more from that pollster about that state are forthcoming. Expect a mess of polls on Monday and a smattering on Tuesday, then all that remains is the counting of the ballots.
November 5 - 11:58 pm
Doing more work on the site tonight. It will go up and down as I work on it. Such is the downside of not having a development server, but I really want to get this project done by election night.
November 4 - 7:00 pm
Work on the new section of the site continues. It's required I widen the display a bit. If I had more time I'd pretty it all up, but for now the site may look a little wonky. The new feature will be well worth the inconvenience though. Watch for details in a day or two.
November 3 - 7:44 pm
I corrected a data entry error in South Carolina from a few days ago. I reversed Hillary and Trump's poll numbers. The state is now back to leans Trump and not toss-up.
The flip to Trump lasted about eight hours. New Hampshire (and the election) is now flipped back to Clinton.
There are lots of election polls out there -- over 2000 of them. To make sense of this data there are "poll aggregators" like this website. There are numerous others, 538 being perhaps the most famous of them.
270towin is a bit different. It's a poll aggregator aggregator. It looks at multiple poll aggregators to predict the outcome of the presidential election. Check it out -- interesting site.
For the first time we are projecting this election to be tied. If correct, the race now goes to the US House of Representatives where every state delegation casts one vote. It's hard to imagine in this situation the House would not elect Donald Trump.
November 1 - 4:56 pm
I'm doing a bit of maintenance on the polls.html page, so it may be screwy for the next hour or so.
Edit: The maintenance is complete. The error preventing the polls.html page from displaying all polls by date has been fixed.
The October 4-14 YouGov poll of California showed some support for Glorida Estela La Riva, the nominee from the "Peace and Freedom" party. Only one person (Hispanic male, age 30-44, liberal, high school education) in the survey indicated support for her. Unfortunately this still rounds to 0% in the poll, so I haven't added her to our listings. I'll be sure to keep my eyes open in case she has a late surge that threatens Clinton's supremacy in California.
I'm beginning to think that if Trump has an unusually good night next Tuesday my model will be looking pretty good. Google releases their weekly polls on Tuesday, meaning they'll release their final poll next Tuesday (Monday if they decide to cutoff the data period by a day so they can release it before the election). If they release their poll on Tuesday, it'll bump up the 3-day average by a day in every date, meaning any poll that finishes its data gathering period on Friday, November 4 will be eliminated from the the average. I have my suspicions this will be a fair number of polls, so I may be getting a pro-Google skew in my results. And considering those results are often screwy (Google still insists Hillary has a strong lead in Kansas), it's likely to throw off my results.
Either way, the good news for Trump in today's polls is that he's closer to winning the election than he's ever been before on this site. Winning any more states worth 5+ electoral votes (4+ if we assume he'd win in the House) wins him the presidency. The good news for Clinton is that Trump has no good prospects for winning that state. Michigan may be his best bet, and he has been campaigning there. So we may be in for an interesting time on Tuesday.
All this being said, we are the most pro-Trump of any polling analysis site I know. Without the Google surveys it might be another story. (I may run an analysis without Google later tonight just to see what happens.)
October 29 - 7:56 pm
An interesting flip in Alaska this evening. Now that we're down to a three-day polling average you'll see anomalies like this crop up, although they should settle themselves out by election day. Like in Alaska, where the new poll showing Clinton leading by 4 points is the only poll in the three-day average. The less populated states are particularly vulnerable to odd flips like this, as the polls are so much less frequent and the 50-state polls going on sometimes fail to achieve our minimum sample size of 100 to be listed. Needless to say, our model has a funky result at present, but it'll resolve itself soon. Just watch.
October 26 - 10:10 pm
Just an FYI that starting October 29 a poll will be considered obsolete once three days have expired. With the large number of polls being released for swing states, this guarantees we'll be using only current data as we go into election day. The change will take effect for each state as its first poll is added with an end date of at least Oct 29.
October 21 - 2:49 pm
Lucid put out a correction on several its October 10 polls of Pennsylvania, Iowa, Ohio, and Indiana. I corrected the numbers. I don't believe anything flipped because of this, but I wanted to mention there were some moderate changes. Details of the changes are available on their site.
October 20 - 12:00 pm
After pondering yesterday's issue about McMullin in Utah, I've decided the simplest way to address the issue is to no longer consider any poll of Utah that does not include McMullin. Such polls would be prone to a massive amount of error anyway (ex: the recent CVoter poll of Utah with Trump leading Clinton 65 to 30), so they're of only marginal use at best.
The last poll of Utah without McMullin is dated Oct 16. Meaning that once we reach October 23 (Sunday), the next poll of Utah to be entered will cause that poll to be scrolled out of consideration. At that point we'll have an accurate (or at least as accurate as these things ever get) idea of what's going on in the state.
October 19 - 1:35 pm
For the first time ever on this site an independent candidate is leading in a poll. Today's poll of Utah has McMullin leading with 31%. I'd have to significantly redo the design of the front page of the site to create room for a third column featuring McMullin's poll numbers. I'm not willing to do it yet, but if he continues to lead in polls then I will. So for now I added code to remove the coloration from the Trump (the 2nd place candidate) in the right column so it doesn't look like he leads.
Now it's time for a talk about methodology. Herein I have a problem:
My site basically converts a candidate's support in a poll into raw votes. For example, if Trump has 40% support in poll with 600 respondents, then he receives 240 votes. The "current" polls for a state are all computed this way. The candidate with the largest number of current votes is then considered the winner in that state. Depending on the degree to which they're winning, the state can also range anywhere from strong Republican to strong Democrat.
Now here's the kicker: this assumes that all polls ask about all candidates who might be leading in a poll. Until now that's always made sense. It's not much of a presidential poll if it doesn't include the Republican on the list of candidates. The rise of a third leading candidate in Utah now breaks that assumption. Not all polls of Utah include McMullin on the ballot. And when he's not listed, he's considered to have received zero votes. Meaning that the polls of Utah that don't list him as a possible choice (Google and Cvoter which poll all 50 states at once) will prevent him from ever being considered as leading in Utah.
It's a breakdown of my methodology, and I have no immediate solution. I'll ponder this in the next day or two and, if I can come up with a good solution, will add it to the site.
October 18 - 6:41 am
The new FDU poll in New Jersey conducted an interesting experiment. Half the respondents were asked about the usual 4-way race. The other half were asked about Clinton and Trump along with a pair of less well known candidates, James Hedges from the Prohibition Party and Monica Moorehead from the Worker's World Party. Sure enough, support for third party candidates depends on who they are. Johnson and Stein totaled 10% support, while Hedges and Moorehead received only 2%. Moral of the story: it matters who the third party candidates are. When voters have not heard of the candidates, they're unlikely to offer any support.
I also wonder about a second possible moral: James Hedges received 2% while Monica Moorehead received 0%. I doubt 2% of New Jersey respondents have heard of him -- I'm in New Jersey and certainly haven't. Could some voters have potentially confused his name with Gary Johnson? Or do male unknown third-party candidate names receive more support than female names? Food for thought.
October 13 - 6:58 am
I'm doing a bit of programming on the site this morning. I do live testing of new code, so pardon if things get a bit wonky for the next hour or so. During this time the site may periodically have minor problems. Also, I'm playing with the Utah listings, so the recent Y2 poll may show erroneous results during this time.
Update: The upgrade is complete. The state listings can now display the proper color when a third party candidate leads in a poll. Some very(!) rare complex poll results may not display correct -- I'll fix it if and when they happen. Also, the large map at the top of the main page cannot yet handle a third party win of a state yet. If McMullin starts to win polls in Utah I'll get to work on it.
October 12 - 9:55 pm
The New York Times has a great article on problems polls experience when weighting voters.
Today's Y2 Analytics poll of Utah has Trump and Clinton tied at 26% and minor candidate Evan McMullin at a whopping 22%. Garry Johnson is in fourth place at 14%. I've never had to deal with a poll in which a third party candidate leads a state. If I see much more evidence this may actually happen I'll get to work preparing the site in case he actually leads in the state.
October 10 - 12:02 pm
It's worth a mention as to why I haven't banned Google from our listings. It's been suggested to me, especially in light of their wildly counter-intuitive results (Trump up significantly in Florida, North Carolina, and New Jersey, while Clinton is abnormally strong in Kansas and Missouri).
I don't remove Google because it's an incredibly bad habit to get into. Google's polls are almost certainly screwy, yes. But if I remove them from our listing what's to stop me next time I have a problem with a company? And the time after that?
In the end I risk reducing my analysis to one where I cherry pick which polls I like and which I don't. That's the mistake Romney supporters made in the last election when they started unskewing polls they didn't like. The end result was Karl Rove's election night meltdown on FOX, something I don't want to replicate in front of my family.
Not to mention I made a methodology and, for better or worse, it's something I'm stuck with. To change the rules mid-game is to go back on my own judgment in an attempt to influence the final projection of this website. That's not something I want to do. If I don't like the way the analysis works, then I can correct it for 2020. There might be a better option how it can be addressed in the methodology.
Either way, I have hopes it'll work out in the end. We'll know more on election day when our final projection is available and the final results pour in.
Just typed in a slew of polls resulting in a number of flips.
1. Arizona and Missouri flipped from Republican to Democrat.
2. North Carolina flipped from Republican to Democrat and back again.
3. Florida flipped from Democrat to Republican and back again.
As I said in a previous update, we're down to a 7-day average. It's going to cause these flips to occur more often. This is especially true because of the odd results in the Google polls. Take Missouri, for example. Every poll since Hillary's of the summer has Trump leading... except for the Google polls which are sure Hillary is 4-9 points up. These polls have a large margin of error which is throwing off my results in the states.
I'm grudgingly accepting that my forecast may be wrong on election day thanks to Google. But after the election I plan on running a special analysis to see how my projection would have been were it not for the Google polls. I'll be curious to see how accurate that forecast is.
Arizona appears to be a different matter. While Google does have Clinton up by 6 (an unlikely reality), Emerson college has her up by 2 and OH Predictive Insights found the state tied. So there may be something going on there related to Trump's comments about the former Miss Universe.
As for Florida and North Carolina, both are swing states that are relatively close. Google has found unexpected pro-trump leanings in both states explaining why the states are flipping back and forth rather than just static for Clinton.
I haven't given up hope though. Polls become more frequent as election day approaches, and with the wild swing in the polls Trump may be about to experience thanks to Friday's audio release about Trump, the final count in the electoral polls remains almost impossible to forecast at this point.
Larry Sabato: First of all, Ohio moves all the way from Leans Republican to Leans Democratic in the Electoral College... If Clinton wins Ohio, Trump has no path to victory. No Republican has ever won the White House without it, and, particularly in this election, Ohio's demographics should make it relatively low-hanging fruit for the Republicans. So we now have Trump as the underdog in Florida, North Carolina, and Ohio, three electoral vote-rich states that he absolutely needs to have any chance to win.
Larry Sabato also moved Nebraska's 2nd Congressional District from Toss-up to Leans Democratic. He's likely basing it this change on a lot more than just polling, although he may also have access to internal polls that I don't. I'll be curious to see if this changes on our analysis at some point as well.
October 8 - 11:58 pm
A few notes on a variety of things in no particular order.
1. Trump's campaign began imploding mid-afternoon on October 7 when audio was released of him describing how he sexually harasses women. For the next few days polls that arrive will all have been taken completely or partially before that date, so we won't know for a while what effects will be. Furthermore, there's a debate later tonight (Sunday, Oct 9) that will have yet more effects. So I suspect few polls are going on this weekend and will be put into the field on Monday.
2. As of today (Oct 9) the website will reduce the time a poll is "current" to just 7 days. The date of a poll is considered to be the end date that the data was collected. Expect to see flips happen more often because of this. On October 29 the average time will drop to 3 days. All polls are judged by the last date their data was collected.
3. It will be interesting seeing what effect the Google polls have on our election. As I've been writing, Google is posting some unexpected results compared to the herd of other pollsters. This can be a good thing, as sometimes one pollster has their finger on something right the others are missing. On the flipside, their polls may be wrong resulting in significant error in the average. Ideally lots of polls occur so error is watered down by good data, but some states don't get a lot of polls.
So we may be seeing weird flips in places like Kansas where Google has insisted (in spite of all recent evidence to the contrary) that Hillary has a strong lead in the state. It'll also effect swing states like Iowa, which today flipped to Clinton because of the poll. Or in Florida where the state has remained in Trump's favor solely because of the Google polls which have an abnormally large sample size.
Switching to the 7-day cycle removes 50% of the weekly Google polls, but also halve the amount of time that polls can water down the results. So expect lots of interesting actions to occur because of them.
4. Trump's implosion may have effects in unexpected places. Utah, for example, never really liked Trump. It's entirely possible a third party candidate will lead in the state, in which case I'll need to do some fast coding to make this site work. Other states may become competitive. Alaska, which has relatively few polls, has a current poll saying that Trump only leads by three points. The implosion may well put the state into play or result in Hillary winning. Expect to see changes in other states as well.
National polls always come out before state polls do, and robo-polls (ie; automated phone polls) come back faster than live operator polls which are more accurate. So be patient as results come in. It takes a while for state polls to adapt to a changing situation, but they always do.
5. There are less than 30 days left until election day.
October 5 - 12:57 pm
Arizona just flipped from Republican to Democrat? These Google Surveys are going to be the death of me.
Update: Only lasted a day before it flipped back, although it was interesting that the new poll added was 2 points in Clinton's favor.
October 4 - 5:42 am
Hillary's debate bounce is starting to have an impact. Nevada has flipped from red to blue. I think Florida is not far off -- it would have already except for the odd Google poll results in the state.
October 2 - 8:10 pm
At this point we still don't know what effect last week's debate is having. There weren't a lot of polls this last week, and some were taken before the debate took place. Still others were taken over several days including both before and after the debate. National polls coming out are suggesting that Hillary is getting anywhere from a 2-3 point bounce, but we don't have evidence among state polls to back that up yet.
September 29 - 4:03 pm
Just a reminder that the website is currently doing a 14-day average of polls based on the date the last poll was added. Starting October 8 this will drop to 7 days.
September 28 - 10:21 pm
And we have our first poll of the three Nebraska Congressional districts. NE-2 remains only nine points apart so it's unlikely Clinton will be able to pick off a vote here. It's just one poll so one never knows, but initial results do not look promising.
September 26 - 9:12 pm
There we have it, folks. I just finished the last of the polling released before this evening's debate. A few states switched out of tossup status, most notably New Jersey and Kansas after the recent Google poll. No states flipped between Republican and Democrat on the election map. This election season hasn't seen a lot of those so far -- just 27 in all. I'll make a separate post about them sometime.
So with the campaign now entering the debate phase, we've been settled into a 272-266 electoral vote win for Clinton for a while. This could be slightly off though. There have been no polls of the Nebraska 2nd Congressional district, a potential electoral vote pickup for Clinton. I'm sure someone will put a poll in the field at some point (I'm looking at you, Omaha World-Herald), but until that point we'll have to wonder if there's a vote waiting to be had here. Probably not -- the Nebraska GOP revamped the district to make it more heavily Republican after Obama won its vote in 2008, but you never know.
A few polls may trickle in over the next few days, but it'll likely be late this week or early next week before we know what effect the debates have on the race.
Most forecasts have Clinton winning the election with anywhere from a 51% to 70% chance of winning. We're in agreement with this, as our model finds Clinton winning by 272-266. Not all models are in agreement though. The Washington Post explains one such model predicting a Trump win.
September 23 - 2:55 am
The New York Times has a fascinating article about real-time election day projections for 2016. The media has an unofficial rule against announcing these during the day, in part because their results are not always accurate, and in part because they don't want to be accused of skewing how voting progresses during the day. Rather, the data is the province of the campaigns so they can better utilize resources during the day. (Example: Is turnout better than expected in district A? Redirect your "don't forget to vote!" callers to district B.)
This year will be different. Political scientists, statisticians, and media companies will be providing updated information in real time from dawn to dusk for the presidential race and key Senate races. You likely won't find coverage of this on most of networks, but Slate has said they'll cover them. It should be interesting seeing what we learn.
Harry Enten from 538 presents the top-5 "must win" swing states for each candidate. In order of importance, they are...
Trump: Iowa, Ohio, Florida, North Carolina, and Nevada
Clinton: Michigan, Wisconsin, Minnesota, Virginia, Pennsylvania
Picking off a "must-win" state from the opponent's list while holding your own virtually guarantees victory.
Today's Google poll of Arkansas is why I dislike these national polls with small sample sizes in each state; they give erratic results. Today's pull has Hillary up a point in Arkansas with a 7.5% margin of error. No one, not even the Clinton campaign, thinks they are actually winning there or even have a shot at winning the state. Yet Google is sure of what it's finding as the last poll had the state tied. So either Google has tapped into the groundswell of support for their former First Lady, or something is messed with their questions or sampling. Or they're just unlucky with their findings in that state.
And Kansas? Hillary up by 16 points? I sincerely doubt this. Yet all six Google polls have found Hillary leading in the state. True, Kansas has an above-average Mormon population which isn't sold on Trump, but I hardly think today's lead of 16 points is remotely accurate... up from 8 points in their last poll.
Either way, I account for some of these erratic polls by not including any poll with at least a sample size of 100. After all, I don't want to call a state where few polls visit (say, Hawaii) for the wrong candidate just because a screwy Google poll with 50 voters and a 25 point margin of error comes out at the last minute.
If I spend 200+ hours doing the data entry on this website only to discover my results on election day are screwy because of Google's polls I am going to be one pissed off hobbyist.
September 21 - 10:02 pm
[Humor] Wow, the pollsters aren't kidding around any more.
September 20 - 11:15 pm
With the flip of Nevada from Democrat to Republican, Trump rises to 266 electoral votes on our election map, the highest we have projected him since the election season began. Clinton still wins at 272 votes, but if the election gets any closer the projection may shift in Trump's favor for the first time.
The New York Times has an article about how the exact same data provides different results when assembled by different pollsters.
September 19 - 5:33 am
Small tidbit. An fellow named Richard Duncan showed up on today's Suffolk University poll of Ohio at 1% support. He's an obscure independent candidate from Ohio who has a small but dedicated following in one of the counties there.
September 14 - 7:24 am
Gary Johnson's campaign said yesterday that it will appear on all 50 states plus the District of Columbia this fall. The Wall Street Journal is reporting this is the first time in two decades this has happened, but I've read from multiple other sources that he was on the ballot in all 50 states last time. I'm not sure which is correct.
September 13 - 3:52 pm
The Tulchin/Moore polls of California I just entered are not duplicate polls. They did a 2400+ sample that was split in half with each half receiving different questions. Half got a 2-way matchup, half got a 4.
September 12 - 7:33 pm
The Washington Post says data scientists, journamlists, and Silicon Valley entrepreneurs are teaming up to provide live election day projections in a manner not traditionally seen in American media. This will start in the early morning when polls open and go all day until the polls close and the results are tabulated.
September 11 - 12:19 am
Our last area to be polled has finally arrived -- Washington, DC. The weekly Google polls were all under our minimum sample size of 100, so I averaged four weekly polls into one larger one (with appropriate weighting for each sample size). End result: Hillary is beating Trump there about 2:1. Nothing surprising, but it's nice to finally have a poll for DC.
September 10 - 12:52 am
We've been incorporating a lot of 50-state polls lately into our listings and will be adding more as time goes on. 538 has a good article on the value of these polls and their reliability in the grand scheme of things.
These are national polls that survey, say, 10,000 people across the country. They then use the data to also provide state-by-state results as well. Some of the less populated states have small (
September 9 - 3:52 pm
This sort of thing drives me batty. Morning Consult released a national poll with more than 18,000 registered voters surveyed. As part of their effort, they provided data for each state along with the margin of error... but no sample sizes. So I've got a whole slew of data I'd love to enter into my database, but without a critical piece of data necessary to do so.
Either way, some interesting findings in their research such as an exact 40.3%-40.3% tie in Iowa. Give it a looksie if you're interested.
Today marks the last day of our 21-day polling averages. Starting tomorrow, whenever a poll is added to a state the average will only consider the most recent 14 days worth of polls.
September 6 - 1:38 pm
With today's Survey Monkey polls, we now have a presidential poll in every state. Some specific areas are exceptions. For example, there is not yet a poll of the District of Columbia. Also, I don't have a breakdown in the individual Congressional districts of Nebraska (and the single poll for Maine's is outdated). Nebraska and Maine allocate electoral votes in part to the winner of each Congressional district, so it's important to see who leads in each place.
September 3 - 3:07 pm
We had an odd flip in North Carolina today. A new poll with Hillary leading by 5 points was added resulting in the state flipping to Republican. It's because an earlier poll of North Carolina scrolled off our 21-day average, and just under 21 days ago there was an outlier poll with Trump leading by 12 points. That poll will scroll off starting with the next poll added after Monday, so unless it's an unusually strong Trump poll for Trummp the state will flip Democratic again.
It's also worth noting that as election day grows closer, the length of time we'll consider in poll averages will drop. Starting September 8 the model will only consider the most recent two weeks of polls starting whenever the next poll for each state is added.
Overall though, we're seeing a slight swing across the map toward Trump. Clinton is still handily winning the election, but he's a bit more competitive overall. This mirrors some changes we're seeing on the national level as well as the lower "Trump-implosion" factor in the news lately.
August 31 - 12:18 am
Survey Monkey: Should pollsters poll support for Gary Johnson and Jill Stein? Pro: They're on the ballot in most states. Con: Third party candidates (especially during the summer) have wildly inflated levels of support in polls compared to how they do in November.
Also of note, Stein voters overwhelmingly support Clinton as their second choice (60% to 17% with 24% not sure). Johnson voters only narrowly support Trump as second pick (39% to 36% and 25%).
PBS Newshour: Hillary Clinton has more three times the number of campaign offices in key states than Donald Trump.
Stuart Rothenberg: An electoral blowout is not likely, but a sizable Clinton victory is.
August 29 - 5:09 pm
The editor of Talking Points Memo has issues with today's Emerson polls.
August 24 - 9:59 pm
A huge data drop of polls this evening. Ipsos/Reuters is running the largest ongoing series of presidential polls ever with more than 10,000 voters surveyed each week. Some of the data posted here was learned from FiveThirtyEight's list of polls, as Reuters didn't make the sample sizes public. I presume 538 has enough clout they were able to ask for the information.
With this data dump our list of states for which we have no polls has grown smaller; we have no polls for Hawaii, North Dakota, Rhode Island, South Dakota, Wyoming, and the District of Columbia.
August 23 - 11:44 pm
The latest poll of South Carolina from The Feldman Group was distributed by the South Carolina Democratic Party to the Rachel Maddow show. She reported the poll was tied 39-39 in a 4-way matchup, but never reported Johnson or Stein's support or the margin of error. Studying a screenshot she showed of the poll I was able to determine it was 16% undecided. So while I don't have Stein or Johnson listed for that poll, know that they are the numbers from the 4-way matchup and not the 2-way.
Update: Huffington Pollster is reporting their support is 5% an 1%.
The New York Times: The electoral battlegrounds of Arizona and Ohio are to begin voting on Oct. 12, nearly four weeks before Election Day. And North Carolina and Florida will be underway before Halloween.
The New York Times: Landslides do not really happen in presidential elections anymore.
Evan McMullin finally appears on our list of candidates. A few days ago he scored 0% in a poll so there was no need to add him, but today he's at 9% in Utah. He's a former CIA operations officer and chief policy director for the House Republican Conference. He's also a Mormon which plays well in Utah. Although he'll appear on the ballot in multiple states, at this point he's essentially a protest vote focusing on a one state strategy.
Small disclosure: this site's software isn't equipped to handle what happens if a third party candidate is winning a state. So if he shows any chance of winning a state I'm going to have a lot of work to do if this site is going to handle it correctly. So if I'm crossing my fingers that he never threatens to take the lead in Utah, it's not me wishing ill of his campaign -- I'm just trying to save myself 20-30 hours of programming time.
August 22 - 1:36 am
I added a mess of new polls this evening, mostly polls from the past few month. They're mostly "obsolete" polls as they're prior to our current 21 day cutoff for our current forecast, but a pair of polls in Kentucky did cause a bit of a shift... from strong Republican to leans Republican and back again.
Special thanks to fivethirtyeight for maintaining an extensive public database of polls. I normally have a wide variety of sources where I learn about polls, but a couple of times each election season I like to poke around at the other analysts to learn what I'm missing. To my surprise it was quite a bit, albeit quite a few of these were internal campaign polls from partisan pollsters that received relatively little media attention outside of their individual states.
With these new listings we now have polls in 41 states. We have no polls for Alaska, Hawaii, Maryland, Montana, Nebraska, North Dakota, Rhode Island, South Dakota, Wyoming, and the District of Columbia.
Update (hours later): And we now have a poll for Alaska. Unsurprisingly, Trump is winning.
I just found a Missouri poll from Survey USA I erroneously recorded under Kansas. I deleted the Kansas entry and re-added the poll to our listings -- Kansas is unchanged but Missouri flipped from tossup to leans Republican. I also deleted a Wisconsin poll from May as it turned out to be a poll of a single Congressional district and not the entire state. This does not change our current projection of the state.
To the four or five of you who actually read this site, I offer my apologies for these errors.
August 18 - 8:02 pm
The Suffolk University poll of Nevada adds two new candidates to statewide polling for the first time. The first is Darrell Castle from the Independent American Party. The second is Rocky De La Fuente, an unaffiliated candidate. Both scored 1% in this poll.
August 17 - 5:32 am
Reports are coming out of Indiana that an internal poll from the John Gregg (D) gubernatorial campaign has Trump and Clinton tied at 44%. Unfortunately I have no details beyond this so I'm not able to add it to our listings, but this is bad if not unexpected news for Trump. Indiana has yet to be publicly polled in the presidential race.
The first poll to include Evan McMullin is now posted -- PPP's poll of Texas. Unfortunately he scored less than half a percent support so PPP rounded him to zero. As such, I haven't added McMullin to our listings. McMullin is the conservative alternative for Donald Trump being pushed by some on the right. Basically, a conservative protest vote against Trump.
August 15 - 2:33 pm
August 10 marks the 90-day point until the election. As such, In a Democracy is now tightening up what they consider to be a "current" poll. Previously, any poll within the last 30 days was considered when averaging out polls to determine which candidate is ahead in a state. Now that we're within 90 days, that time limit has tightened to 21 days. As the election draws closer that time period will become less and less as polls are released with increasing frequency. The cutoff days and times are:
Prior to August 9 -- 30 days
August 10 to September 8 -- 21 days
September 9 to October 8 -- 14 days
October 9 to October 28 -- 7 days
October 29 and after -- 3 days
August 11 - 11:16 pm
Talking Points Memo has interesting pair of graphs comparing the 2012 and 2016 races. Spoiler: Trump is doing worse than Romney was doing.
Politico explains that Trump is having a second polling problem this year. Polls using a likely voter model are swinging against him more than those with a registered voter model. This is reverse of conventional wisdom in which Republican candidates do better with likely over registered voters.
A year ago analysts would have said that it the House is highly unlikely to be in play. Larry Sabato still agrees that the House is likely out of reach, but says the Democrats are making progress toward taking it with a current projection of a 10-15 seat gain. By way of reference, Democrats need 30 seats to take the House. The Wall Street Journal reports that Democrats are actively working toward increasing the number of competitive seats.
The "Unskewed Poll" nuts are back. As you may recall, in 2012 it was a theory among many Republicans that the polls were biased against Mitt Romney. Someone braniac had the "brilliant" idea of reweighting professional polls toward something much in favor of Mitt Romney. This idea gained traction on FOX News and the conservative media echo chamber with the end result that on election night many Republicans were surprised (read: utterly shocked) when Romney lost. Even Karl Rove fell victim of this when he underwent his epic meltdown on FOX.
A new bunch of conservatives had decided they don't like the current trends in the polls toward Hillary so they've reweighted the polls again. You can find the results on Long Room. I wish very much I could watch the election results come in with the folks from that site while we point at our respective models on how they're doing. As Harry Enten from 538 explained, professional pollsters put their reputation and careers on the line when they release poll results. To suggest they're actively conspiring against Trump is nuts.
Washington Post: The Clinton campaign is expanding into Georgia and Arizona.
August 9 - 5:57 pm
Stuart Rothenberg: In May, I wrote that Clinton had a "decisive" advantage, an assessment I reiterated in July. But now that the conventions have passed and the race stabilized, Clinton's advantage has gone from decisive to overwhelming.
Donald Trump: Why should I change when I'm winning?
August 8 - 2:34 am
Talking Points Memo: Hillary's campaign is focusing on the traditional battleground states and is not actively campaigning for achievable-but-difficult states like Georgia or Arizona.
August 5 - 11:58 am
Talking Points Memo has updated their electoral scoreboard. Clinton: 305, Trump: 179, Toss-up 54.
Politico: The 2016 battleground map may soon have one less state on it... with just over three months until Election Day, Clinton and her allies are showing outward signs of self-assurance here by all but cutting Colorado out of their recent ad buys.
August 2 - 8:16 pm
I'm not the only one with an electoral scorecard. Matter of fact, I'm probably the littlest frog in a very big pond running one. Talking Points Memo just introduced theirs. Hillary leads Trump 271 to 168 with 99 electoral votes up for grab. As a reminder, only 270 votes are necessary to be elected president.
The New York Times: Even as Mr. Trump has ticked up in national polls in recent weeks, senior Republicans say his path to the 270 Electoral College votes needed for election has remained narrow - and may have grown even more precarious. It now looks exceedingly difficult for him to assemble even the barest Electoral College majority without beating Hillary Clinton in a trifecta of the biggest swing states: Florida, Ohio and Pennsylvania.
Trump adviser Paul Manafort claims that both Connecticut and Oregon are within research for Trump. My thought is that this is a really lame fake-out attempt to convince the Hillary campaign to waste dollars defending those two states, and I don't think they or any independent analysts remotely believe those states are in danger of supporting Trump.
July 28 - 2:47 pm
Larry Sabato: ... we're sticking with our current Electoral College ratings, which show Clinton with 347 electoral votes safe, likely, and leaning to her, with 191 safe/likely/leaning Trump. We agree that if the election were held today, it would almost certainly be closer than that - and that Trump could very well win. But the election isn't being held today - it's actually still about 100 days away.
July 19 - 8:30 pm
The New York Times has an incredibly cool interactive tool for tracking the 1024 possible pathways to victory for both Hillary and Trump through the ten (at present) swing states. Scroll down the page to find it. If the swing states should change by election night I very much hope they update the tool. Twould make great fun on election night itself.
July 18 - 5:46 pm
Kansas has flipped back to normal today. I'm pretty sure the poll that threw it off a month ago was either the result of random chance or polling error, but either way it's fixed now.
July 14 - 2:35 pm
The Gravis Marketing poll of Michigan has an odd finding to it. The three-option presidential question has results about what you'd expect: 48% for Clinton, 41% for Trump, and 11% for "other". But when question is re-asked also providing Gary Johnson and Jill Stein as questions, the numbers change in a way you wouldn't expect: 37% for Clinton, 34% for Trump, 2% for Johnson, 1% for Stein, and a whopping 26% for other.
I have no explanation for this. When "undecided" is not presented as an option, it's common for a respondent to select other. There is often a small protest vote to be found in this category -- die-hard Sanders supporters or moderate Republicans who refuse to vote for their party's nominated candidate.
But why would "other" increase from 11% to 26% when the most common third-party candidates are listed? This makes no sense to me. The Gravis polls of Iowa and Colorado taken at the same time returned no such finding. One might attribute it to just a random event, but the margin of error for the Michigan poll is 2.4% so a 15% increase is highly statistically significant.
Is it improbable noise or is there something odd going on in Michigan? Or could the poll have a methodological problem that isn't readily apparent? I have no idea and am wishing I knew for sure because it's such an odd finding to me...
July 13 - 6:05 pm
Nate Silver discusses when to freak out about new polls.
July 1 - 11:01 pm
The Atlantic: Can Hillary Clinton Turn Red States Blue?
Popular website 538 posted their electoral model. Spoiler: Hillary is winning.
Washington Post: The GOP's massive electoral map problem isn't getting any better.
June 29 - 9:19 pm
Charlie Cook: Trump needs a trick of fates to win.
Lots of bad news for Donald Trump lately. Here's more: Democrats are taking a good look at their (small) chances of taking the House and thinking it may be doable after all.
June 21 - 7:38 pm
Washington Post: "Donald Trump has boldly declared that he will change the nation's electoral map. But as the general election begins, he faces a steep, almost impossible climb to the 270 electoral votes needed to win the White House... Even the best-case scenario for Trump looks very uphill."
The latest polls
There are no new polls to list.
View the archive of polls.