MO-SEN 2018: The Megathread

[MarkD]

What do new voter registration number in Missouri indicate? (do they favor Dems. Repubs, about even , etc)?

In Missouri, we don't register to vote by party. Every registered voter is, officially, an independent.

[jamestroll]

I can not believe we will see clean missouri, 12 dollar minimum wage, and medical weed pass yet an incumbent Democratic Senator could well lose in a Democratic wave.

Don't forget right to work getting overwhelmingly rejected!

It's almost as if Republicans vote based on identity politics, white resentment, etc. rather than Randian economic philosophy, union busting, or prohibition.

Lmao haha yes.

I am not expecting this but it's possible that an unqualified candidate could defeat an excellent incumbent auditor as well here.

[DataGuy]

I was noticing today that 538 still has McCaskill as the 61% favorite here. I think she's probably in greater danger than that. She hasn't led a single poll since late last month, and I've heard that internal campaign polls are consistently showing a clear lead for Hawley.

My model's formula is showing this result right now:

Josh Hawley (R): 48.62% ✓
Claire McCaskill (D): 46.70%
Other candidates: 4.68%

This seat is looking like the most plausible Republican pickup besides North Dakota.

[mcmikk]

This race has been pretty much a dead heat for the past year. It will probably come down just to national environment, which is why Claire probably takes this one.

[I Can Now Die Happy]

I am really hoping it's at least Hawley+5. Claire put out that anti-male video a while back and she also was once of those eye-roll worthy Game of Thrones watchers who claimed they'd stop watching because of the Sansa-Ramsay incident.

[Bidenworth2020]

I am really hoping it's at least Hawley+5. Claire put out that anti-male video a while back and she also was once of those eye-roll worthy Game of Thrones watchers who claimed they'd stop watching because of the Sansa-Ramsay incident.

umm... did she actually say this? lol

[RI]

I am really hoping it's at least Hawley+5. Claire put out that anti-male video a while back and she also was once of those eye-roll worthy Game of Thrones watchers who claimed they'd stop watching because of the Sansa-Ramsay incident.

umm... did she actually say this? lol

https://twitter.com/clairecmc/status/600636817239605249

[RussFeingoldWasRobbed]

I can not believe we will see clean missouri, 12 dollar minimum wage, and medical weed pass yet an incumbent Democratic Senator could well lose in a Democratic wave.

Don't forget right to work getting overwhelmingly rejected!

It's almost as if Republicans vote based on identity politics, white resentment, etc. rather than Randian economic philosophy, union busting, or prohibition.

But muh Bernie won WWC voters in the dem primary so they are #populist (:

[Hindsight was 2020]

I am really hoping it's at least Hawley+5. Claire put out that anti-male video a while back and she also was once of those eye-roll worthy Game of Thrones watchers who claimed they'd stop watching because of the Sansa-Ramsay incident.

Good that plot line was disgusting. It happened to a different character in the books, it was solely done for shock value, and the plot to get her into the scenario made no sense

[Ebsy]

lol now random blue avatars are just posting what numbers their highly rigorous models show occurring. What a sad joke this forum has devolved into.

[Calthrina950]

lol now random blue avatars are just posting what numbers their highly rigorous models show occurring. What a sad joke this forum has devolved into.

DataGuy doesn't come across as a fire-breathing partisan to me. He had Sinema winning in the prediction he put up on the Arizona thread.

[Ebsy]

lol now random blue avatars are just posting what numbers their highly rigorous models show occurring. What a sad joke this forum has devolved into.

DataGuy doesn't come across as a fire-breathing partisan to me. He had Sinema winning in the prediction he put up on the Arizona thread.

Two words: who cares.

[I Can Now Die Happy]

I am really hoping it's at least Hawley+5. Claire put out that anti-male video a while back and she also was once of those eye-roll worthy Game of Thrones watchers who claimed they'd stop watching because of the Sansa-Ramsay incident.

Good that plot line was disgusting. It happened to a different character in the books, it was solely done for shock value, and the plot to get her into the scenario made no sense

I think Season 5 in general was trash. The whole Jeyne Poole thing in the books was superior, and so was everything else in AFFC and ADwD.

[I Can Now Die Happy]

I am really hoping it's at least Hawley+5. Claire put out that anti-male video a while back and she also was once of those eye-roll worthy Game of Thrones watchers who claimed they'd stop watching because of the Sansa-Ramsay incident.

Good that plot line was disgusting. It happened to a different character in the books, it was solely done for shock value, and the plot to get her into the scenario made no sense

You're completely right.

I think Season 5 in general was trash. The whole Jeyne Poole thing in the books was superior, and so was everything else in AFFC and ADwD.

[Calthrina950]

lol now random blue avatars are just posting what numbers their highly rigorous models show occurring. What a sad joke this forum has devolved into.

DataGuy doesn't come across as a fire-breathing partisan to me. He had Sinema winning in the prediction he put up on the Arizona thread.

Two words: who cares.

Are you saying that users can't post predictions on here?

[Bidenworth2020]

dataguy clearly isn't making biased predictions, I would be curious to see his model though

[DataGuy]

Of course I have my personal opinions on political issues, but when it comes to numbers I am absolutely willing to predict either side winning. If you ask me, Kevin Cramer and Josh Hawley have the advantage now. So do Kyrsten Sinema and Joe Donnelly.

I'm still in the final stages of making my model, but I will certainly give complete details on my methodology when I post my complete Senate predictions around Election Day.

It's really not some top-secret magical hocus-pocus. The model takes polls, averages them, adjusts them, and includes some other factors like fundamentals and aggregated third-party ratings. Its basic ingredients are like those of FiveThirtyEight's "Deluxe" model, but it applies the data somewhat differently.

First of all, I use mean-reverted bias instead of house effects to adjust polls. I do not believe that house effect is the best way to adjust polls, because it assumes by default that the polling average is right and makes all polls conform to it. Mean-reverted bias judges polls based on their historical track records relative to actual election results.

Second, I consider the trend lines in the polling averages. Basically, I give candidates a "bonus" if they clearly have the momentum. To my knowledge, other models do not do this. In applying this factor retroactively to past elections, I found that doing so would have helped foresee several upsets. A candidate who is clearly losing steam tends to finish lower than their final polling average, because the trend continues into Election Day.

I also have a different way of calculating "fundamentals." I restrict my fundamentals to true numerical fundamentals, namely the changes in partisan advantage both nationally and locally. If a certain state has moved sharply toward one party since the last election, that counteracts or augments the national movement since then.

I do not include fundraising as a fundamental, since it is becoming less and less predictive of the final results. Nor do I attempt to judge nebulous factors such as "candidate quality" or "scandals." What is a high-quality candidate or a scandal to one person might not be to another. I also consider "incumbency" as baked in when considering the partisan changes from a previous election, so I do not include it as a stand-alone factor. This is perhaps the biggest difference between my model and 538's. I think 538 attempts to quantify factors that are really subjective.

Finally, I include a broader range of "expert" Senate ratings as compared to FiveThirtyEight. They use only about three, while I use seven or eight.

I then give all these factors due weight, with polls getting more weight as more data is available. However, I have an absolute upper limit on the weight that polls can receive. The other two factors are always guaranteed an appreciable share of the weight.

As for minor aesthetic components, my model lists percentages to the hundredths place instead of the tenths place. It will also attempt to forecast specific vote tallies and turnout totals. This part is pretty much experimental.

I know I basically just wrote a full-length article, but I do think it's important people know how I'm doing my calculations and how my model is different from the other ones. Whenever someone asks me about my model from now until Election Day, I will refer them to this post. If there's demand for it, I will get into the specific mathematical details once I have posted all of my final predictions.

I'm aware that some people don't like the election data wonks, but I believe in data and I'm proud of it. It is not always 100% right, nor do I expect it to be. Nevertheless, it's a safer bet than subjective and biased gut feelings. Even in 2016, despite all of the talk of huge polling errors, Trump was within the margin of error in several key states that he ended up winning. The problem is typically not the data, but rather how commentators present the data in an overly confident and insufficiently detailed way.

[Xeuma]

I would love a full report on the mathematical workings.

[Antonio the Sixth]

Of course I have my personal opinions on political issues, but when it comes to numbers I am absolutely willing to predict either side winning. If you ask me, Kevin Cramer and Josh Hawley have the advantage now. So do Kyrsten Sinema and Joe Donnelly.

I'm still in the final stages of making my model, but I will certainly give complete details on my methodology when I post my complete Senate predictions around Election Day.

It's really not some top-secret magical hocus-pocus. The model takes polls, averages them, adjusts them, and includes some other factors like fundamentals and aggregated third-party ratings. Its basic ingredients are like those of FiveThirtyEight's "Deluxe" model, but it applies the data somewhat differently.

First of all, I use mean-reverted bias instead of house effects to adjust polls. I do not believe that house effect is the best way to adjust polls, because it assumes by default that the polling average is right and makes all polls conform to it. Mean-reverted bias judges polls based on their historical track records relative to actual election results.

Second, I consider the trend lines in the polling averages. Basically, I give candidates a "bonus" if they clearly have the momentum. To my knowledge, other models do not do this. In applying this factor retroactively to past elections, I found that doing so would have helped foresee several upsets. A candidate who is clearly losing steam tends to finish lower than their final polling average, because the trend continues into Election Day.

I also have a different way of calculating "fundamentals." I restrict my fundamentals to true numerical fundamentals, namely the changes in partisan advantage both nationally and locally. If a certain state has moved sharply toward one party since the last election, that counteracts or augments the national movement since then.

I do not include fundraising as a fundamental, since it is becoming less and less predictive of the final results. Nor do I attempt to judge nebulous factors such as "candidate quality" or "scandals." What is a high-quality candidate or a scandal to one person might not be to another. I also consider "incumbency" as baked in when considering the partisan changes from a previous election, so I do not include it as a stand-alone factor. This is perhaps the biggest difference between my model and 538's. I think 538 attempts to quantify factors that are really subjective.

Finally, I include a broader range of "expert" Senate ratings as compared to FiveThirtyEight. They use only about three, while I use seven or eight.

I then give all these factors due weight, with polls getting more weight as more data is available. However, I have an absolute upper limit on the weight that polls can receive. The other two factors are always guaranteed an appreciable share of the weight.

As for minor aesthetic components, my model lists percentages to the hundredths place instead of the tenths place. It will also attempt to forecast specific vote tallies and turnout totals. This part is pretty much experimental.

I know I basically just wrote a full-length article, but I do think it's important people know how I'm doing my calculations and how my model is different from the other ones. Whenever someone asks me about my model from now until Election Day, I will refer them to this post. If there's demand for it, I will get into the specific mathematical details once I have posted all of my final predictions.

I'm aware that some people don't like the election data wonks, but I believe in data and I'm proud of it. It is not always 100% right, nor do I expect it to be. Nevertheless, it's a safer bet than subjective and biased gut feelings. Even in 2016, despite all of the talk of huge polling errors, Trump was within the margin of error in several key states that he ended up winning. The problem is typically not the data, but rather how commentators present the data in an overly confident and insufficiently detailed way.

Credit where it's due, this seems like a serious endeavor. I'd be curious to see how your model would backtest in past elections (although backtesting is not always a great predictor of future accuracy).

Also, do you have estimates of the uncertainty too (ie, not just the point estimates of your results but also the confidence intervals that go with it)? If so, what probability distribution are you using to model it? I'd be curious to see your probability numbers to compare them to 538's.

[SCNCmod]

What do those who live in Missouri feel about the state of the race.  Is everything still up in the air... is one side pulling a little ahead, or do you have a feeling in general regarding who will pull out the win?

[MarkD]

What do those who live in Missouri feel about the state of the race.  Is everything still up in the air... is one side pulling a little ahead, or do you have a feeling in general regarding who will pull out the win?

I have the race as Lean Hawley. That's the way the polling indicates it is playing out. I have a hard time seeing McCaskill being able to win any of the rural counties in the state, nor St. Charles or Greene Counties. She won't be able to overcome the strong R trend that was displayed in 2014 and 2016, and I don't see a Democratic "wave" coming in this year.

[Holy Unifying Centrist]

What do those who live in Missouri feel about the state of the race.  Is everything still up in the air... is one side pulling a little ahead, or do you have a feeling in general regarding who will pull out the win?

I have the race as Lean Hawley. That's the way the polling indicates it is playing out. I have a hard time seeing McCaskill being able to win any of the rural counties in the state, nor St. Charles or Greene Counties. She won't be able to overcome the strong R trend that was displayed in 2014 and 2016, and I don't see a Democratic "wave" coming in this year.

The limited recent polling we have clearly indicates a Tossup

[Ebsy]

Very clearly the race is close and has been all the way through. Hawley has run an extremely lackluster campaign but outside groups have spent a ton of money attacking McCaskill. Should come down to how each performs in the St. Louis and KC suburbs.

[Wisconsin+17]

Poll out, Hawley up by seven now.

[IceSpear]

Poll Internal out, Hawley up by seven now.

Fixed.