Saturday, August 13, 2016

Trump v. Clinton, State and National polls

On a message board I frequent a poster asked this question, “Trump is averaging 42% in national polls and polls that have him at 32 or 33 in Colorado and Virginia seem right to everyone?”

Trump isn’t at 42% in national polls. He’s at 37.6% in 4-way polls. If you look only at 2 way polls, Trump is at 41%, however. And he is at 33% in one Virginia poll and 32% in a Colorado poll by the same pollster.

My preference is for 4-way polls, since there’ll be at least 3 names on the ballot in every state and 4 in many of them. So why include voters for Trump who won’t vote for him? All 4 candidates aren’t always included in state polls, but there are also a higher percentage of undecideds in those polls.

I decided to see if the state and national polls were compatible. I found they were, but first my methodology.

I took the number of votes each state had in the 2012 Presidential election. Of course the states won’t have that exact number but it’ll approximate proportionality. Next I took the percentages Clinton and Trump got in the RCP averages for state polls. In some cases several polls were used. In others only one was used. Doing this can lead to some unreliable numbers, as the single poll might not be accurate and might be a month or more old. I allocated votes based on these percentages and put any undecided/third party votes in the third party column. This may inflate the third party totals but as long as I wasn’t assigning the votes to Clinton or Trump it wouldn’t effect their margin. (Note: 3rd party/undecided actually has the most votes in Utah and Vermont)

If there were no polls I used percentages calculated by fivethirtyeight. I don’t know their methodology but using their projection beats just guessing.

The states where I used polls are highlighted in yellow on the linked spreadsheet. The states where I used fivethirtyeight’s projection are in blue.

Totaling up all these votes got me a Clinton win of 43.4%-38.1%. That’s roughly a point off the current national polling of Clinton 44.0%-37.6%. That’s really close when you consider the state and national polls were taken at different times by different pollsters. A few new state polls could increase Clinton’s margin to the national average.

One reason Trump is doing better in the state polls is that he’s doing well in California and New York, two states in the top four in votes. Romney lost California by 23 and New York by 28. Trump is losing those two by 16 and 13. Two recent New York polls had Trump that close. That results in a net gain of 2.1 million votes by Trump. Clinton has a net gain of 4 million in the remaining states. Fivethirtyeight predicts a 21.5 Clinton win in New York and a 22.8 point win in California. If I input those margins I come up with a Clinton 6.3% win, almost identical to her 6.4% lead in the national polls.

Clinton’s national lead of 5.2% in the state polls and 6.4% in the national polls is 1.4%-2.6% larger than Barack Obama’s 3.8% win in 2012. If the states moved uniformly Mrs. Clinton would win Colorado by roughly 7 points and Virginia by 6. Yet she’s ahead by 11 and 10. This could be just a lack of the same pollster polling both at the same time or it could be because Donald Trump is going to genuinely change the map. He isn’t your typical Republican and it makes sense that his support isn’t going to uniformly move from Romney’s. Trump may lose New York by a much smaller margin than Romney while losing elsewhere by larger margins.

The state polls don’t show a dramatic electoral difference. Clinton picks up Georgia and North Carolina while losing Iowa. Iowa and Georgia are so close in the polls that they could easily be won by the same party that won them in 2012.

No comments:

Post a Comment