Sunday, July 17, 2016

Clinton vs. Trump:
The pre-convention landscape

In previous years, I made the assumption that people aren't paying much attention until after the conventions and are not fully committed until after Labor Day. But looking at my Twitter and Facebook feeds, this assumption doesn't hold up. It is a tired cliche to call the current presidential election "the most important of our lifetime", but we really do have a stark contrast between the candidates, and the inclusion of Donald Trump makes this easily the weirdest election I have seen personally, which takes us back to 1968.

My personal feelings: I am definitely anti-Trump and not keenly pro-Hillary. I know more than a few people on the left that take every negative view of Clinton as gospel, and that makes me wonder about her chances, but my training is as a mathematician. I would rather speak with data available to back up my positions rather than anecdote or even a well thought out think piece. I certainly read as much as I can get my hands on about the election, but I'm putting the numbers forward in these posts.

How good are the numbers? They are getting better. I now have polls from 26 states in May, June and July, the months when the presumptive nominees have been clear. Of course, I'd like to have some polling in all 51 contests being held in the Electoral College, but there are some where prior elections and demographics are good enough to be confident of the outcome. Donald Trump is not going to win the District of Columbia, and Hillary Clinton should not rely on electoral votes from North or South Dakota, just to give a few examples.

Any surprises so far? Yes. If all polls are weighted equally, Oregon and Utah could be battlegrounds. Utah as a battleground would be horrendous news for the Republicans and Battleground Oregon would be bad news for the Democrats.

Do you give all polls equal weight? If they are current and they ask respondents if they are registered to vote, yes I do. I've done poll aggregation in four presidential elections now, and if I see a brand name I don't know, I'm wary, but I still count the numbers. If the unknown pollster identifies as Republican, I'm more disgusted than wary, but I still count them. In 2008 and 2012, the Republican pollsters were generally so far off from the truth, their clients should have had a case to sue for fraud. Oregon is a battleground due to only having two polls and one of them is from an off-brand Republican outfit.

What do you count as "battleground" states? That will change from post to post. When a state has a Confidence of Victory number for the leader between 50% to 60%, that is a clear battleground. If the number is over 60% and under 90%, I consider that to be "leaning" and over 90% is solid.

Where does that put us right now? Ah, good now we get to the numbers.

Total electoral votes in Solidly Trump: 158

Leaning Trump: GA AZ
Total electoral votes in Leaning Trump: 27 (grand total:185)

Battleground favoring Trump: IA UT
Total electoral votes in Battleground Trump: 12 (grand total:197)

Toss-up state: OH
Electoral in Toss-up state: 18

Total electoral votes in Solidly Clinton: 257

Leaning Clinton: NH OR PA
Total electoral votes in Leaning Clinton: 31 (grand total: 288)

Battleground favoring Clinton: NV FL
Total electoral votes in Battleground Clinton: 35 (grand total: 323)

Current totals: Clinton 323, Trump 197, Toss-up 18
Totals from last post: Clinton 336, Trump 187, Toss-up 15
Comparison to 2012: Obama 332, Romney 206

Okay, back to regular size font not in bold. Trump has had a positive couple weeks, but he still hasn't caught up with Romney's numbers from 2008, which means he is still losing and losing by a lot.

Things can change. If they couldn't change, I wouldn't be writing this blog. Trump is currently telegraphing a strategy of focusing on New York and California, which are currently extreme long shots and are demographically stacked against him. The "easiest" path for Trump is to win all his Solid and Leaning States, hold on to Iowa and Utah, take Ohio from the toss-up and score what are now upsets in Nevada, Florida and Pennsylvania to get to 270 exactly. Oregon could help by replacing another state with 6 or 7 electoral votes.

Currently, Nate Silver is giving Trump a 35% chance of winning the election. Looking at most likely paths, I would put that number at less than 5% currently, but I don't put much stock in such numbers so far out from Election Day. 

Mr. Silver is better known than I am, but he's not better at this. I got more stuff right than he did in both 2008 and 2012. The best known poll aggregator I trust is Sam Wang at Princeton. Here is a link to his website.

Back at the end of July when both conventions will be history.



  1. "Of course, I'd like to have some polling in all 51 contests being held in the Electoral College..."

    Technically aren't there 56 contests being held, since Maine and Nebraska aren't winner-take-all?

    Do you have a summary of your methodology and how it differs from Sam Wang's? How do you calculate "confidence of victory" from the polling data?

    1. I do split up Maine and Nebraska if the polling data shows up.

      My method is very much like Wang's. It surprises me my Probability of Victory is so much higher than his right now. We may get more in line when the number of polls increases over the next few months.

      I'll publish a FAQ this week. Thanks for stopping by.

    2. Is your probability of victory really that much higher? He doesn't seem to publish an exact statistic for probability of victory if the election were held today, but his histogram shows what looks like around a 99% probability of Clinton winning:

    3. The numbers he has at the top of the blog are
      Random Drift 65%
      Bayesian 80%

      You have to click on a link to see his prediction of 99%, which he then discounts.

    4. Right, but those are predictions for November, so they're not really comparable to your snapshot of the polls today: they're answering different questions.

    5. Or maybe I'm misunderstanding what you mean by "a snapshot of what polls say now"; your "confidence of victory" estimate is based on what the election result would be if it were held today, right?

  2. This comment has been removed by a blog administrator.