When you look at state-level election results, fundraising, and the national environment, surprising state polls actually make a lot of sense. Your weekly political data roundup for October 4, 2026.
As always, love your deep explanations and broad field of vision. At minimum, having some sens that I might have justified reason for optimism is helping me sleep at night. Plenty of time to fret about next steps in November.
I understand that historically polling errors don't show much of a statistical pattern. But since 2016, especially in the Midwest, we do see a pattern. Furthermore, on several dimensions we know that the Trump era has been distinctive (usually not in a good way). Appealing to the historical absence of a pattern does very little to assuage my concern that the polling might be favoring Dems in the Midwest yet again. Our understanding of the Trump political coalition explains why that might be the case — low-trust, conspiracy-minded voters, etc.
An excellent post. This Substack is my go to source for information on this election. I have an annual subscription and urge others to do the same.
Now for a technical question. Errors are normally assumed to be uncorrelleated between samples. This is rarely the case with time series data. For this reason, the normal parametric statistical techniques are invalid for use with this data. In other words, they are not Best Unbiased LinearEstimators (BLUE). Autoregressive Integrated Moving Average (ARIMA) techniques are one way to remove this problem.
Few understand this issue and it is often a source of misleading predictions. Some of the effects noted in this great post are probably the result of this effect. Does anyone have any estimate of how correlated the errors are in polling data? Has anyone applied ARIMA models to this data?
You often comment that Democrats have a lot of good (call that running better than average in a general election) candidates. But it is also true that Republicans have a lot of bad (call that running worse than average in a general election) candidates. You often comment on particularly bad Republican candidates like Ken Paxton (in military terms KP are appropriate initials); but the current crop of Republican candidates are bad generally.
I think that there is a reason this has happened and it does not bode well for the Republican future. The old time "solid" Republicans (New England Republicans or Midwest Republicans) have no place in the current Republican Party. They have moved or been pushed to the sidelines or become Independents or even Democrats. Those remaining tend toward the extreme. They run in Republican primaries in which those Republican voters that tend toward the extreme have disproportionate voting power. And this tendency toward the extreme is multiplied by Donald Trump's endorsements of the most MAGA candidates (instead of candidates who will run well in a general election). The result is a crop of candidates who are polling (and will probably run) worse than average. This has produced races in which a candidate who runs better than average in a general election is opposing a candidate who runs worse than average in a general election.
Hasn't it been established that many state polls, even by
poliical parties or small pollster firms aren't as accurate as more broad polls on issues by large media organizations. If so, then most state based polling averages are mostly a shot in the dark. Given that, isn't just averaging a group of state polls on a state race, merely guessing at a possible reality that may not be close to reality when the election reality is posted?
I think we will test different specs of fundamentals for 2028, that allows the “personal vote” to activate for non-incumbents with a track record, and also formally incorporates other predictors of the national environment beyond the generic ballot
How can we take this polling data on US Senate and House races and apply it to the state House and Senate races in, for example, my beloved North Carolina?
Elliott has done this by using Census data to determine the characteristics of each District and using polling data matched to those characteristics from a multitude of StrengthInNumbers/Verasight polls. I'm sure that one of the members who is wonkier than I can provide a reference as they so frequently do when Elliott and David are on live.
There are plenty of things MattY could rightfully be called, but "economics blogger" is not one of them. That's a bit of an affront to actual economists and economics bloggers. Noah Smith -now that's an "economics blogger"
An excellent article, empirically based rather than opinion based on 'priors'. Anything can happen in November, as repeated so often - and the post election analysis can consider actual polling bias, should it occur. The real danger is people with agendas using historically out of context polling misses to seek to invalidate results which don't accord with their 'adjusted' predictions.
Really glad for this column. I had been thinking that since 2016 it often seems Republicans have a way of changing the stakes in the last month or even days leading up to the election. It’s hard to trust polls but I feel a little better after this.
It seems like there is more movement in presidential approval and generic ballot this year than in previous midterms, and if that continues through election day, polls from four weeks out are not likely to be that accurate, right?
I also would like to have a sense of how the Democratic House candidates are doing in Indiana. I know it is an uphill climb, but some candidates are working very hard.
Indiana has a very important and competitive race for Secretary of State--the person who oversees elections. The Republican, Max Engling, came out with a poll showing him in the lead. The Democrat, Beau Bayh, countered with a poll that showed them tied. There is also a Libertarian candidate (they usually take about 5%) and a new Lincoln Party candidate who is a former Republican and former mayor of Indianapolis.
As always, love your deep explanations and broad field of vision. At minimum, having some sens that I might have justified reason for optimism is helping me sleep at night. Plenty of time to fret about next steps in November.
I understand that historically polling errors don't show much of a statistical pattern. But since 2016, especially in the Midwest, we do see a pattern. Furthermore, on several dimensions we know that the Trump era has been distinctive (usually not in a good way). Appealing to the historical absence of a pattern does very little to assuage my concern that the polling might be favoring Dems in the Midwest yet again. Our understanding of the Trump political coalition explains why that might be the case — low-trust, conspiracy-minded voters, etc.
An excellent post. This Substack is my go to source for information on this election. I have an annual subscription and urge others to do the same.
Now for a technical question. Errors are normally assumed to be uncorrelleated between samples. This is rarely the case with time series data. For this reason, the normal parametric statistical techniques are invalid for use with this data. In other words, they are not Best Unbiased LinearEstimators (BLUE). Autoregressive Integrated Moving Average (ARIMA) techniques are one way to remove this problem.
Few understand this issue and it is often a source of misleading predictions. Some of the effects noted in this great post are probably the result of this effect. Does anyone have any estimate of how correlated the errors are in polling data? Has anyone applied ARIMA models to this data?
Hi Russ, You're right. I've just renewed my annual subscription too - SIN and 50+1 are excellent sites.
You often comment that Democrats have a lot of good (call that running better than average in a general election) candidates. But it is also true that Republicans have a lot of bad (call that running worse than average in a general election) candidates. You often comment on particularly bad Republican candidates like Ken Paxton (in military terms KP are appropriate initials); but the current crop of Republican candidates are bad generally.
I think that there is a reason this has happened and it does not bode well for the Republican future. The old time "solid" Republicans (New England Republicans or Midwest Republicans) have no place in the current Republican Party. They have moved or been pushed to the sidelines or become Independents or even Democrats. Those remaining tend toward the extreme. They run in Republican primaries in which those Republican voters that tend toward the extreme have disproportionate voting power. And this tendency toward the extreme is multiplied by Donald Trump's endorsements of the most MAGA candidates (instead of candidates who will run well in a general election). The result is a crop of candidates who are polling (and will probably run) worse than average. This has produced races in which a candidate who runs better than average in a general election is opposing a candidate who runs worse than average in a general election.
Hasn't it been established that many state polls, even by
poliical parties or small pollster firms aren't as accurate as more broad polls on issues by large media organizations. If so, then most state based polling averages are mostly a shot in the dark. Given that, isn't just averaging a group of state polls on a state race, merely guessing at a possible reality that may not be close to reality when the election reality is posted?
As an incumbent House member for nearly 6 years, I'd argue that Ashley Hinson should be weighted somewhat as an incumbent.
I think we will test different specs of fundamentals for 2028, that allows the “personal vote” to activate for non-incumbents with a track record, and also formally incorporates other predictors of the national environment beyond the generic ballot
How can we take this polling data on US Senate and House races and apply it to the state House and Senate races in, for example, my beloved North Carolina?
Elliott has done this by using Census data to determine the characteristics of each District and using polling data matched to those characteristics from a multitude of StrengthInNumbers/Verasight polls. I'm sure that one of the members who is wonkier than I can provide a reference as they so frequently do when Elliott and David are on live.
There are plenty of things MattY could rightfully be called, but "economics blogger" is not one of them. That's a bit of an affront to actual economists and economics bloggers. Noah Smith -now that's an "economics blogger"
Well to be fair, that is how he is identified via Wikipedia, and I believe how he referred to himself for some time https://en.wikipedia.org/wiki/Matthew_Yglesias
hah. fair enough. 'is not' maybe should have been 'should not be'
Matt Yglesias, a wojak incarnate animated by the potion from Fantasia
An excellent article, empirically based rather than opinion based on 'priors'. Anything can happen in November, as repeated so often - and the post election analysis can consider actual polling bias, should it occur. The real danger is people with agendas using historically out of context polling misses to seek to invalidate results which don't accord with their 'adjusted' predictions.
Really glad for this column. I had been thinking that since 2016 it often seems Republicans have a way of changing the stakes in the last month or even days leading up to the election. It’s hard to trust polls but I feel a little better after this.
It seems like there is more movement in presidential approval and generic ballot this year than in previous midterms, and if that continues through election day, polls from four weeks out are not likely to be that accurate, right?
Typically polls move toward the out-party in the closing stretch for midterms. No reason to expect otherwise this year IMO
Is Indiana so far red that potentials are not even considered?
As far as I’m aware, there is no U.S. Senate election in Indiana this year.
What about congressional candidates
I also would like to have a sense of how the Democratic House candidates are doing in Indiana. I know it is an uphill climb, but some candidates are working very hard.
Indiana has a very important and competitive race for Secretary of State--the person who oversees elections. The Republican, Max Engling, came out with a poll showing him in the lead. The Democrat, Beau Bayh, countered with a poll that showed them tied. There is also a Libertarian candidate (they usually take about 5%) and a new Lincoln Party candidate who is a former Republican and former mayor of Indianapolis.