Commentary:
Election Day Notes:
Note#4: The door closes on PA and WI, state where Romney had a chance to win, but clearly did not. The republicans spoke of two paths to victory, one with OH and one without. The one without was always a false road in our opinion, as it depended on either PA (legitimately in play based on recent polls but not on recent voting history) as a substitute for OH or Romney running almost a sweep of other swing states. We felt the other swings (VT, WI, NV, IA) might hop on board a Romney momentum train, but that would assume a victory in OH too. With NC, FL and VA too close too call, it makes a harder narrative to believe that that all the other swings will fall in line with Romney. Right now, Romney's path to victory is squeak out FL, VA, NC and OH, then grab CO. A loss in any of those and the election should fall to Obama.
Note #3: Most of what we are seeing come in we would categorize as "2008-light" type turnout. This means you should focus on the 2008, '40/08 average and the '08/'10 models for many of the states. With NC, VA and FL all tight at this point, that favors Obama.
Note #2: Early voting numbers were released for CO today. When we place them in our models and reweight for the way democrats, republicans and independents have polled recently, we see this data as projecting to about a 0.5% favorability for Romney. When ananlyzed in comparison to 2008 data, a R+2.3% model would be indicated. Please note, the number of assumption and other factors involved does not lend itself to a high degree of confience, but we analyze it based on the data at hand and present it to (1) see how it holds up as today's voting gets reported and (2) provide insights that can be compared to other information in other states to see if any trends are developing. From a modeling standpoint a R+2.3% electorate would be favorable to Romney.
Note#1: Update: We have not seen any indication that the earlier report was accurate. As we watch this state things seem more in line with 2008 light type turnout which would favor Obama. Prior: Earlier today some information was released in OH that seems to indicate a Romney lead of 92,000 votes in early voting. Based on a quick analysis of some data at hand, this would indicate either (1) democrats are not supporting Obama at the levels that republicans are supporting Romney (i.e. Romney is getting more lane changes in his column) or (2) that the rule of thumb that 80% of undecideds go for the challenger is holding true. If this data ends up being correct, Romney would be modeled to carry the state.
Special Commentary: Standing on the eve of the election with what may be our last polling updates of the season, we offer how we will be using the data contained on these pages to analyze the election as it unfolds. As stated previosuly, we look at these polls and look for the degree to which a model favors a candidate as opposed to the absolute percentages indicated by the reweighting. Take Florida as an example. Models indicate an Obama win with 2008 style turnout or D+3% and beyond. So we'll be paying attention to exit polling and watching the poll numbers come in. If FL is competitive for Obama, it may indicate turnout more along the lines of '08 or 04/08 or 08/10 levels. that would favor Obama. We'll probably have that more conclusively in VA and NC when their polls close and declarations of winners are made in the media. If Romney looks strong there, it would indicate that perhaps the 2008 style will not be achieved and a 2010 type electorate is emerging. Look to those three state, margin of victory should provide great insight early on as to what type of night we will see.
11/5 Evening update: FL Gravis looks good for Obama. OH Gravis looks better for Romney.
11/5 Good data from swings states brings several of them into clearer focus. See state tabs for more details. Afternoon update: More polls and states conitnue to fall into better focus.
MN update: We analyzed the PPP poll, and as we've been saying all election cycle, models only show MN going Romney in scenarios where he rides a wave and independents/undecideds go strong for him on election night. Polling data does not suggest a Romney win in the state.
11/4 Five new swing state polls added today. Models get moved. PPP has a crush on Romney it appears when reweighting their numbers. See state tabs for more detail. Lots of interesting tidbits in these polls. Evening update: added PA Morning Call poll. See PA for details. Late night update for IA, NH, OH and VA.
11/3 Evening update: Swing polls added in FLA, IA and NH. Models shift in 2 of 3.
11/3 OH poll confirms current models when reweighted. Obama continues to show strength in 2008 based models, Romney in 2004 and 2010 models, with averages being purple.
11/2 Evening update: 9 new polls see state tabs for more details. Most confirmational one state has changes.
A word on Hawaii:* Shock Poll: Romney closes in Hawaii/Gains +8% from last poll ** OK, so obviously we are joking with the headline. Partly. We analyzed the HI poll partly because we think numbers are fun to crunch, partly because it was something new to do after weeks of analyzing the same few states intensely and partly out of curiousity. For the curious, Romney does show a lead in the models through D+2%. Unfortunately for Romney the state was D+16%, D+25% and D+20% in '04, '08 and '10. So I wouldn't go expecting a late inning campaign stop by Obama in his home state (birthers: read Hawaii not Kenya here). On the other hand, it wouldn't shock us if team Obama did send Joe Biden to Hawaii for the next several days. Or Kenya. For longer. But on a serious note, the take away from Hawaii may just be that the national polling favoring Romney is correct, and that Romney has made republican gains in non-swing states. To the extent that Hawaii may be repeating itself in other non-swing states, this poll would lend anecdotal evidence there and further the possibility of a split between the popular and electoral votes.
11/2 Our apologies for the site crashing. This close to the election, to borrow a phrase, it was not optimal. Including yesterday's polls we could not upload, there are 11 new polls on the site today and plenty of movement. The down time did give us a chance to revisit some national polling. At this point national polling matters only in that if the press coverage influences those who want to vote for a winner or discourages those who would otherwise vote from going to the polls it can be impactful. But as Al Gore will tell you (and perhaps Romney this year), the electoral vote is all that matters. The national polls are the candy coated treat, the state polls have more nutritional value as far as the models are concerned.
11/2 NV polling favors Romney strongly today showing leads into D+4% models now. This state will turn on the get out the vote effort by each side. Those with a view similar to or approaching 2008 have hope on the democrat side. Those with an '04 or '10 viewpoint of the electorate can see a Romney victory in the state.
A word on using the models: There are many methodological factors across polls that influence their final projections. In that sense, comparing polls from different polling companies is like apples and oranges on some levels. A good example being the classification of what constitutes a likely voter or who is included as part of a party versus labeled independent. Our reweighted models take projections to a tenth of a percent and can imply an exactness that perhaps is unwarranted. There exists a variety of assumptions in the underlying polls and assumptions we make based on the depth of internals available. As such, while we present numbers such as 2.3%, we look at the resulting scale and trend of the results in our modeling when determining if a model should fall in either candidate's column. So a 2.3% and a 2.5%should be considered equivalent and show a weak lead for the candidate. A 6.3% value would imply greater strength for the candidate and more confidence in the lead for the candidate as modeled in the reweighting. A +0.1% and a -0.1% would be considered as even or too close to call. We use the reweighted results to look for relative strength, consistency across polling and trends in movement of the numbers.
11/1 New polls in swing states added today. See state tabs for details on the movement in the models.
10/31 Evening update: Movement in the electoral vote for the D+1%, D+2%, D+3% and 2008 models including movement in Ohio and Wisconsin. See state tabs for more detail. See Michigan tab for a special note there as well.
Another word on Minnesota: The KSTP/SurveyUSA poll was released and reviewed today. While the data is incomplete, there are some things we can start to infer with regards to the state. Right now, we see the state breaking down along turnout lines. Meaning that a party's advantage in turnout should be matched in the polls. Historically, MN was D+3% in 2004 and D+4% in 2008. We see the republicans as needing to get into R+% territory to have a shot at the state, something that would buck the history of the last two presidential cycles.
10/31 Lots of polls added today for the swing states. No movement in any of the models to report, but they do contain some interesting data. Of note, the independents. In the seven polls updated today, Romney leads Obama in every one of them with independents. The polls favor Romney: (R/O) 45/40 and 49/44 in FL, 49/43 and 48/37 in OH, 50/38 in NC, and 59/33 and 57/36 in VA.
10/30 Florida updated for the SurveyUSA model. Polling is consistent with what we have beeen seeing in the state for some time.
A word on Oregon: With the recent release of several polls for the state we dediced to take a quick look-see at the state. We reviewed three polls covering the period from 10/16 to 10/25 with varying degrees of internals disclosed. Some things to note from the polls, Obama appears to be leading among independents who comprise approximately 25% of the vote. Neither party has of yet nailed down it's base with both republicans and democrats indicating voting support for their candidate in the 80%'s (so there is some wiggle room left for Romney). This state went R+2% in 2004, D+9% in 2008 and D+10% in 2010. The two polls with internals were D+8% and D+6%. The Hoffman Research Group poll is more favorable to Romney showing positively for him through the D+2% model, but real Romney strength is not achieved until the R+2% or R+3% models similar to 2004. If the republicans can get to R+2%/2004 levels, they have a shot at the state. However, the polling data and recent history tends to not support that viewpoint. At the moment, we do see Romney closing the 16% gap Obama had in 2008 over McCain, but don't see this state as knocking on the swing state door just yet.
10/29 Two FL polls and a CO added mid-day. Obama moves the D+5% into his column in FL and no changes result in CO.
A word on Minnesota: There has been some squawking in the news regarding Minnesota recently with regard to improved polling for Romney and Romney campaign ad spends in the state. We have been tracking MN informally for some time. MN is a difficult state to model effectively for several reasons. First, we have no exit polling data for 2010 in our models, rendering the 2010 and '08/'10 average models ineffective in modeling outcomes for the state. Secondly, we have three polls to assist with our analysis, with one having good internals but being dated, another lacking good internals and the last being the most timely but having internals that conflict with disclosed results in our models. In other words we don't trust the polling data we are using. What is clear from the polls is that MN is much closer than it was four years ago, but we see nothing in our analysis that predicts a Romney victory in the state currently. Is he close enough to make a run at the state? Possibly, we just don't have enough good data to model that at the moment. Until we see some solid polling data to the contrary, we don't think ancedotal evidence is strong enough to realistically see MN as a red state at this time. We have regarded MN as a state that could go Romney if there is a tidal wave of swing state Romney support that breaks in the final days his way, but probably not a state that lone wolf supports Romney.
10/28 The Cimci Enquirer poll in OH confirms current polling models. There are no updates to the models. Evening update sees two polls apiece in NH and OH added. In OH D+3% moves Romney and in NH Romney picks up a couple of other models.
10/27 Romney sees strength in national polls when reweighted. Obama sees strength in PA battleground state level polling. National polls at this point are more a gauge of overall mood more than anything else. In the details of the state by state remodling is where the electoral vote will be won or lost. VA update sees a strong poll for Romney in VA with the models all indicating to Romney for this one poll. Overall, Romney reclaims the D+5% model yielded yesterday in VA. Obama holds the 2008 model but no others in the state.
10/26 Evening update: Plenty of polls to update this evening. CO moves Romney direction in the 2008 and D/R+0% models. OH models break 2:1 in favor of Romney with the CNN tipping in Obama's direction. No movement in the models for this state though. NC sees the 2008 model move Romney as these two polls reweight off the charts for Romney. VA is favorable to Obama with D+5% model moving his way.
Special commentary on momentum: Because we have received numerous responses regarding our use of momentum it is worth highlighting how we apply the term. To us, momentum means that polls are trending favorably to a candidate (be that nationally, regionally or at a state level). While a candidate may get favorable headlines in the paper or pick up endorsements, fill stadiums or not, or get positive treatment on television, as far as our model is concerned we do not considered that as momentum. We consider momentum similar to potential and kinetic energy in our models. The events described above would be like "potential momentum" but only when they show up as positive measurable data in polling samples indicating favorable polling for a candidate do we consider them "kinetic momentum" and thus label a candidate as having momentum. Secondly, momentum indicates only improved polling for a candidate in modeling, not a lead. If a candidate was losing in the polls 99% to 1% and over the course of days polling changed to 98%/2%, then 97%/3%, then 96%/4% we would say the candidate has momentum, but would not be implying leadership in the polls.
10/26 FL Sunshine poll is more favorable to Obama than the prior Gravis poll. But still in line with the Romney ahead narrative. NO changes in the state for our models based on this poll. IA Gravis was a D+6% poll which reweighted favorably to Romney flipping the 2004, R+1% and R+2% modesl his direction. From a modeling standpoint this tightens the state but Obama remians the favorite overall here.
10/25 CO, NV and VA polls added. CO polls split and leave the models unchanged. NV are D+6% andD+7% but do reweight favorably for Obama moving the D+3%, D+2% and 08/10 average model back to Obama's column. VA poll is favorable to Romney and shifts D+4% and D+5% models to romney leaving only the 2008 model favorable to Obama in the state. Bottom line is Romney locking down the south which we had modeled for him for some time. Obama is gaining ground in what are still swing states. So polling momentum remains on Obama's side.
FL Gravis poll moves the D+5% model to Romney. Romney now leads in all FL models other than 2008. NC PPP poll flips the 2008 model to him. This is the only model now predicting Obama victory in the state.
OH Time poll added. Complex poll that seems to indicate when analyzed and reweighted a plus to Obama. The recent trend has been towards Obama both nationally and in a number of swing states. This poll continues the trend. We move the 08/10 average and D+3% models back to Obama in the state. See more detail in the OH state tab.
10/24 SurveyUSA OH poll is a D+7% weighted poll which would match 2008 turnout. When reweighted this poll is favorable to Romney in most models and confirms current modeling. See OH tab for more detail.
10/23 Two national polls. ABC is lightly favorable for Romney. CBS News goes more strongly to Obama. At this point, we are still including the national polls, but frankly the swing state electoral projections are more meaningful indicators of paths to victory. From an analysis standpoint, we'd take one solid poll from OH over 10 new national polls. NH and NV polls released and are also favorable to Obama. NV reweighted moves the D+4% and D+5% back to Obama. NH moves R+3% model back to Obama. As of now, we see NH shading blue and NV shading red.
10/22 update: Update: OH, PA and National polls added. IBD/TIPP national models would have this showing Obama for 2008, 04/08 and 08/10 averages. Suffolk OH moves D+4% model back to Obama in the state. PA polls do not change models. See PA for more insight into today's polls.
10/22 National, FL and OH updates today. No changes to the models, but polls do confirm recent trends. Monmouth/SurveyUSA poll reweights favorably for Romney with only the 2008 model favoring Obama. OH poll is much better for Obama, not enough to flip models just yet. But is worth noting and keeping an eye on trends in the state. Once again, debate tonight looms and will be a reset to polls and a test of current campaign momentum.
Special commentary: Sometimes people ask how can two polls on the same day give such different results. the 10/20 FL polls are a good example of this. You have three polls which have the following regarding party ID: D/R% even, D+1% and D+9%. That's right, one polling firm weights dems and repubs even in turnout and another sees dems +9% in turnout. This is why a RCP average ends up just being a bad average. Re-weighting the polls as we do here puts them all on common ground for scenario analysis. The SurveyUSA poll ends up being a very favorable poll for Romney once the bias is removed. You can see how it moves the poll back in line with the Fox News poll when re-weighted. So two polls with very different headlines of who is winning actually confirm the same thing once re-weighted.
10/20 Lots of updates today. National polling models out slightly favorable to Romney. In FL the three polls when viewed together tend to confirm the story as is. The 2008 model does flip back to Obama in the state. In OH the Fox News poll is a D+8% poll but when re-weighted favors Romney. Inside polling numbers show independents breaking 52/28 Romney here. Something to keep an eye on in future polling. The Gravis poll is even at face, but off the charts Romney when reweighted. Even 2008 modeling is Romney +2% when reweighted. Given these two polls, the D+2% and D+3% models now moves to Romney's column and moves the electoral vote to 272 and winning status for Romney in those two models. the OH train keeps chugging. This time with PPP added and D+4% falls to Romney. D+5% could reasonably be argued here as well, but we won't call it at the moment.
10/19 PPP comes in with two polls today that are favorable on face to Romney but when re-weighted tell a different tale. These two polls are good for Obama when re-weighted with the R+1% and R+2% switching to Obama in the models for both states. In IA the 2004 model and in NH the 2010 models are captured by Obama here as well. See the state pages for more detail on why these two polls favor Obama.
A note on Rassmussen and Gallup. While polling data is generally slow so far today from other sources, these two organizations are out with many new polls. We do not include them in our analysis. This is not related to the quality of their polls, but because we don't have sufficient internals to make them useful as part of the data set. We do watch them though. Rassmussen has Romney up +5% in FL and +3% in VA with national polling even while Gallup see the national voting at Romney +6%. These polls are consistent with our models on the state level. We don't see other national polls re-weighted reflecting a 6 point Romney lead. The Hartford poll from today re-weighted shows Romney +2% for the 2004 and 2010 models. We would think a more realistic number if you wear your republican hat. The 2008 model favors Obama by 3.6% if that is your turnout viewpoint of the world.
10/18 OH polling continues the Romney momentum. Important to note that this poll was conducted prior to the second presidential debate. So while it is interesting and confirmational of recent trends, it remains to be seen if it will have staying power over the next few days. WI and IA models shift to Obama with CO R+1% model moving Romney. The WI and IA polls give Obama something to hang his hat on today.
10/17 The IBD/TIPP poll continues the Romney surge narrative. But the real story these days is in the swing state polling. We will see in the coming days what they indicate. NV polling is bad news for Obama when re-weighted. NV polling by SurveyUSA is interesting in that it allows direct comparison with their prior poll. At the top line SurveyUSA reports Obama's lead widening. But when you dig into the internals, this model is actually more pro-Romney than the prior one. The prior poll was a D+7% and this one was D+9% in the sample. While NV was D+8% in 2008, it was R+4% in 2004 and D+2% in 2010. The D+4% and D+5% flip in NV to Romney on this poll. The WI poll out of Marquette furthers the Romney narrative and flips the 2004, R/D+0%, and R+1% through R+5% his way.
10/16 National polling from Daily Kos has Romney +4%. This poll is friendly to Romney and has him favored in all models other than the 2008 version. New pols from PA, FL and NH. not a lot of updates to the models here. D+0% and D+1% in PA shift course to Romney, but largely PA remains a blue state, though less blue than a month ago, but blue still none-the-less. See the PA tab for a look at how the vote is shoring up along party lines and why current internals point to a tough road for Romney in PA. not impossible, but tough. IA and CO chime in today as well. No major shifts to report in either state.
10/15 Not much new on the national front. Re-weighted polling continues Romney momentum narrative with the exception of the 2008 model. In NC and FL new PPP polling does not change things for now. The PPP OH poll does move toward Romney and flips D+2% to D+5% and 08/10 average back to camp Obama. The IA and VA polls break Romney while the CO poll breaks Obama.
10/13 A word on the Behavioral Research Center AZ poll that is out today. We haven't tracked AZ as we've been modeling it red. But this poll has Obama up 2% at 44/42 so it is worthmentioning as it puts AZ into play. First, we never, love a pol where the candidates are both in the low 40's. Too much swing left to make any analysis deeply meaningful.
Second, reviewing the internals it appears, though not confirmed, that this underweights republicans versus historical exit polls. The last three elections cycles were 2004 Republican+14%, 2008 R+7% and 2010 R+8%. We don't model past R/D+5%, but it looks like plugging this into our models that anything beyond R+6% would mean Romney owns the state. It would not surprise me if this poll was less than R+6% for Obama to be winning here. There is a line in the internals that says "the indication this year is that Republicans, older voters , conservatives, Caucasians and liberals are the least likely to forgo voting. At the same time, minorities, including Latinos, younger voters and Independents are among the most likely to predict that they may not vote." You give me that, with the historical exit polls, in a traditional red state, with the recent national trends towards Romney and my estimatation that this is a republican underweighted poll, and I think it stretches credibility to think this is a blue state based on this poll.
10/12 Polls in FL shift the D+4%, D+3% and 2008 models into Romney's column. The NH poll is particularly interesting. It moves the 2010, and R+1% through R+5% to Romney. See the NH tabs for more detail as to why this poll is so interesting. On the national front the IBD/TIPP poll continues the trend for Romney. When modeled it shows Romney ahead in all models except 2008. This is particularly strong as even D+5% modeling does not get Obama over the top with this poll.
10/11 update - Another national poll and a FL poll to add. The national is more of the same with Romney holding to a lead by 1%. In Florida, the poll is a +4% demcrat model that has Romney up big with many strong internals. This shifts the 04/08 and 08/10 averages along with D+2% models into Romney's column.
10/11 The narrative of the pollsters seems to be Obama has reached the critical 50+% mark in polling. While there areplenty of new polls today, the internals are lacking. As a result, not much weight is applied as it relates to changes in the model. The following changes were made: CO 2008 to Romney and PA 2004 to Obama and NV 08/10 and D+3% go Romney. The internals not used in modeling for PA were very friendly to Obama. See the PA tab for more details.
10/10 SurveyUSA morning poll from Ohio is a +6% democrat model showing Obama up 1%. Re-weighted, this pretty much confirms the Romney lead here. No change to any of the modeled predictions, but does inch Ohio further into Romney territory. Nevada also chimes in with a poll that when re-weighted slightly favors Romney allowing us to flip the D+2%, D+1% and D/R+0% to Romney.
10/9 Updated with new afternoon polls. In NC there is confirmation that Romney is in control. In the polls favor Romney in the 2004 model along with D+0% and all R+% models, but only 2004 model is flipped to Romney pending further polling. Ohio CNN poll pulls the tide back towards Obama. We are leaving the calls in place from earlier 10/9commentary, but could easily flip the prediction models to Obama with new polling showing even the most modest of Obama favorability. In Colorado, polling shows trend to Romney, but not enough to flip current models giving the state to Obama. As with Ohio, Colorado could flip to Romney very easily, but we'd need to see a poll point that way to tip the scales before shifting the prediction models for electoral votes.
10/9 ARG has just released a new poll for OH and this one is a shocker. In a poll where the internals favor democrats by +9%, yes not a typo, Romney is polling at +1%. In analyzing the polls, you generally watch them do a slow wave in a new direction. And that is what they have done the last few days. But this poll is a giant cresting wave. We'll see if it lasts, but it sure adds validity to the other polling data that has moved towards Romney. Romney now is in control of OH for all models other than the 2008.
10/8 National polling is interesting regarding the Pew poll. While it is encouraging for Romney to have a poll showing him with the lead, the models show the poll as largely status quo for Obama. The poll is +2.8% republican, so it is largely inconsequential when adjusted for the historical turnout models. Having said that, the internals regarding indpendents and women largely are a boost to Romney. We will see if these hold and their influence on polls going forward. Would I rather be Romney than Obama in the Pew poll? Yes. Does this poll inflict critical damage to an Obama re-election campaign? Probably not either.
If the Pew is not as influential as others may believe, the Michigan and PA polls do show real movement in the models toward Romney. R+1%, R+2% and 2004 model now back to Romney in PA. D+1%, D+0% amd R+1% to R+5% flip to Roney in MI.
The VA polling is also more of the same in that State with only the D+1% model flipping back to the Obama column.
10/6 Strong polling for Romney in the Gravis poll for Colorado. Romeny flips the 04/08, and R+1%-R+5% back to his column. R+3%, R+4% and R+5% models, while representing the republican extreme, do show Romeny over the 300 electoral vote fence for time this election cycle.
10/5 Romney up in three swing states: 47%/46% in Ohio, 48%/45% in VA and 49%/46% in FL. This is significant given it is post debate polling. However, the President received good jobs data today. So it remains to be seen if this is a shift or a bounce for Romney.
10/4 The data continues to demonstrate the importance of the get out the vote effort this year. Models indicate a Romney path to victory that is more narrow when considered against the President's possibilities.
In general Obama maintains more flexibility to come off of 2008 numbers than Romney has to come off 2004 or 2010 numbers.
Modeling shows some battleground states coming into consistent focus. It remains to be seen what the effect of the first presidential debate will be on the subsequent polling. Check back for frequent updates.
Inside Polling
Re-weighted Polling Electoral Count