Excellent Nate Silver article on polling vs. state fundamentals

[Awsi Dooger]

Article focuses on the 2018 Tennessee and Mississippi senate races, but really the key points are instructive on state races everywhere, and not merely in 2018. This type of balancing act is in play all over the country this year, with state tendencies counteracting with polling and also candidate to candidate variables. Basically, we are asking questions like...if a Democrat leads polling by 5 points in a state historically with a 15 point red slant, how much weight do we give to each variable?:

https://fivethirtyeight.com/features/can-a-democrat-really-win-a-senate-race-in-tennessee-or-mississippi/

Some highlights:

* "What happened in these races? Polls came closer to the final margin about two-thirds of the time (in 18 of 28 cases). So if you had to choose between polls and fundamentals, you’d pick polls. However, the race moved in the direction of the fundamentals three-quarters of the time (in 21 of 28 cases). That is, if the Republican was doing better according to the fundamentals analysis than according to the polls, the Republican tended to gain ground 75 percent of the time, and likewise for the Democrat."

* "So the best forecast comes from taking a blend of (mostly) polls and (some) fundamentals. Exactly how much our model weights each component depends on the amount of polling and the amount of time left until Election Day. Essentially, the fundamentals calculation is treated as the equivalent of 1 or 2 recent, high-quality polls. So if there are 10 or 15 recent polls of a state, the fundamentals calculation has little influence. In states such as North Dakota where polling is fairly sparse, they can have more sway."

* "What that means for Tennessee is that any poll showing Bredesen tied or ahead — and perhaps even behind by 1-2 percentage points — is good news for Bredesen, because the model expects the race to revert toward Blackburn based on the fundamentals. With every new poll, it weights the fundamentals less and less."

***

I concede, this is my type of analysis. I started following closely/wagering on politics in 1992 and immediately realized I had to understand how demographics and voting trends varied in each state, so I wasn't guessing. Many hours in back rooms of a small library, both in Las Vegas and Miami.

My secret weapon was fundamentals. Others I was wagering against, either man-to-man or in a 16-man betting pool, relied almost exclusively on polling. I'm glad this Nate Silver article wasn't available at the time, otherwise my competitors would have understood what I was using. Granted, my stuff in the '90s wasn't nearly as sophisticated as Silver now, but not bad for hand research and a dab of Excel spreadsheets on my little Mac Quadra 610

[uponit7771]

1. Come on Nate, those "state tendencies counteracting with polling" happens mostly in Red States that

... have enacted Voter Suppression Measures (VSM) that are aimed at keeping economic and racial minorities from voting.

Telling the truth isn't being partisan

[Awsi Dooger]

2. Thankfully Nate deals with known quantities

Happy Adjusters make the world whatever they want it to be. Obviously this isn't their type of article

[uponit7771]

3. If that has been dealt with then that's rarely a reported factor, I rarely see them adjust for VSM

... red states.

Especially in the last election where each of the states HRC lost that were tight were all red VSM states.

That should've been front page headlines for months imho

[Scurrilous]

4. Thanks!

K&R

[DemocratSinceBirth]

5. The further out from an election rely on the fundamentals.

The closer to an election rely on the polls. That's why I remain bullish on Bill Nelson.

[Awsi Dooger]

10. Also, in lightly polled states rely on fundamentals

I was glad Nate made that point but I'm glad others were not aware of it over the years.

I have won several wagers in Alaska politics solely due to that aspect. The state is seldom polled, and when it is polled the margins invariably make no sense in comparison to the fundamentals. The Democrat is always overstated. I think they have a lousy model with too much faith in Anchorage.

Tony Knowles in 2004 is somewhat similar to Bredesen this year. Knowles was a very popular former governor and then sought the senate seat. Lisa Murkowski had been appointed senator by her father in 2002 when he won the governorship and gave up his senate seat. It screamed of nepotism so Lisa Murkowski was immediately unpopular. The legislature changed the law so it couldn't happen again.

All of those variables contributed to make Knowles the presumed winner. He led literally every poll for an entire year and the average was 4 points. But I didn't trust the polling. I had already seen too many examples of Alaska polling overstating the Democrat. I wagered on the fundamentals and it was an easy win for Murkowski...never in doubt. That was on the night Kerry lost so it was a devastating night but one of my theories held up. That's the way I looked at it.

Tennessee is somewhat more heavily polled but that state also has a major tendency in recent cycles to badly overstate the Democrat. That's why I'm not fully confident the Bredesen race is accurately portrayed by the polling.

Check out RealClearPolitics from the 2012 and 2016 Tennessee presidential races. Only one poll accurately portrayed the actual massive margins between Romney/Obama or Trump/Clinton. Everything else had the margin less than half of how it played out. Consequently I have one Las Vegas buddy who has been telling me for a year that Tennessee is the new Alaska, that they stopped polling the state once it turned more red and now that polling has resumed the models are simply screwed up:

https://www.realclearpolitics.com/epolls/2016/president/tn/tennessee_trump_vs_clinton-5911.html

https://www.realclearpolitics.com/epolls/2012/president/tn/tennessee_romney_vs_obama-2047.html

The 2014 polling also understated Lamar Alexander's edge in the senate race:

https://www.realclearpolitics.com/epolls/2014/senate/tn/tennessee_senate_alexander_vs_ball-5032.html

[DeminPennswoods]

6. "Fundamentals" are just a way for pollsters to hedge their bets

It's their "fudge factor" if you will. Statistics are based on mathematical truth.

[DemocratSinceBirth]

7. Nate is a forecaster, not a pollster.

[DeminPennswoods]

9. I was also

not politics, but also forecasting under uncertainty.

[RhodeIslandOne]

8. I feel like in Tennessee it comes down to....

....an utterly awful candidate in Blackburn versus a very good one who has won statewide election.

Most people will not vote for a media whore congressperson who goes on every national program disgracing the state outside their own district.

[Awsi Dooger]

11. I hope so but that state scares me

As detailed in post #10

I didn't see an exit poll in Tennessee in 2016 but the 2012 presidential exit poll had 73% from Tennessee labeling themselves either Very Conservative or Somewhat Conservative, and only 27% combined labeling themselves Moderate to Liberal.

Those are scary fundamental numbers for a senate race, to put it mildly

https://www.nytimes.com/elections/2012/primaries/states/tennessee/exit-polls.html

[Awsi Dooger]

12. Good contrast of fundamentals

As DemocratSinceBirth pointed out, Bill Nelson is in good shape considering fundamentals.

Now compare that race to the Beto race. In recent days/weeks the polling can pretend those two situations are similar. Every poll has Nelson slightly ahead or behind. Other than one weird Ipsos poll with a severely inflated Cruz favorable rating and therefore 9 point margin, the Texas polling recently also has Beto tied, slightly behind and one poll with a 2 point lead.

But are those two races really similar in terms of likely outcome? Fundamentals would scream no...not even close

* Beto is a challenger in a solid red state with no recent history of preferring any Democrats on the federal level, and 44% self-identified conservatives to 20% liberals. A Republican currently leads the governorship polling by 20 points and there is straight ticket voting

* Nelson is a long time incumbent in a swing state that is only slightly red and Obama carried twice. The state has 36% conservatives and 25% liberals. A Democrat leads the governorship polling and is now a 60/40 betting favorite.

***

Now, perhaps Beto wins and Nelson loses. Or some combination of Beto faring equal to or better than Nelson. But fundamentals hint strongly the other way, that polling is probably overstating Beto in terms of likelihood, and polling is probably understating Nelson's likelihood.

That's the way I have used this stuff for a quarter century. Doesn't always work, but it is the gambler's delight...more often than not.