Talk Elections

I've created a spreadsheet (http://docs.google.com/spreadsheets/d/14iZUZzIlFwMMuHeCly7MXnPOPN7OQRxRfZKF3qErC7w) that takes the partisanship of states and districts and the current generic ballot margin to predict the results of elections. It uses a logistic function to turn the predicted margin of victory into the probability of victory.

()

To find the overall probability of winning the presidential election or control of a legislative chamber, I took the probability of the tipping point state or district (US House: NE-2, US Senate: NC, US Pres.: WI, 2019 FL House: Not up). This isn't a perfect measure as it would be completely off in extreme circumstances, but I think it should suffice given the complexity of determining the prediction otherwise.

If you have any US elections you'd like me to add, leave them below. Feedback and questions are appreciated.

Polls of the generic ballot are courtesy of FiveThirtyEight (https://projects.fivethirtyeight.com/polls/generic-ballot/).
Partisanship data are courtesy of Daily (http://docs.google.com/spreadsheets/d/17yr9mcAtuUdNjI9NEPYKxXsEldzzQ2ZaDwEAbnPRyS4) Kos (http://dailykos.com/stories/2017/2/6/1629653/-Daily-Kos-Elections-2016-presidential-results-for-congressional-and-legislative-districts), Florida Division of Elections (http://dos.myflorida.com/elections/data-statistics/elections-data/precinct-level-election-results), and Federal Election Commission (http://transition.fec.gov/pubrec/fe2016/2016presgeresults.pdf).

I think your ratings are messed up. In the 2020 presidential tab, TX, OH, GA, NC, and AZ are all rated safe D.

Thanks for pointing that out. I simply plugged the wrong inputs into those formulas. It didn't affect the actual probabilities listed, just the ratings.

There we go. Very interesting project, I'll be interested to see how it performs.

I've updated the predictions to use FiveThirtyEight's partisan lean metric (with updated data from 2018). The partisan lean was formerly the difference between how a state or district voted and how the nation voted in the 2016 presidential, now it's the difference between how a state voted in 2016, 2012, and in state legislative races (I used the lower chamber of the legislature if it was applicable; 2015-2018). I've not run the numbers for congressional districts or state legislative districts so those values still use the 2016 formula.

Predictions Spreadsheet:
https://docs.google.com/spreadsheets/d/14iZUZzIlFwMMuHeCly7MXnPOPN7OQRxRfZKF3qErC7w/edit
Partisanship Spreadsheet:
https://docs.google.com/spreadsheets/d/1WfjI5caNbi3RbsNUtb8dT7ylzh6yJrlT9DJoVVSsNp0/edit

Let me know if you see any mistakes.

LA-GOV 2019 isn't Safe R.

KY, LA and MS aren't safe R

But the rest are okay

They're the way they are because gubernatorial elections don't tend to correlate as well with partisanship as federal or state legislative elections do. But you're right. I'd take the gubernatorial races with a grain of salt.

What's happened to the file?