I had given up on pollsters. Hillary Clinton lost in 2016, after polls predicted she would win. At the time, both my husband and I had a sense that Trunp might win. In the 2020 election, pollsters predicted a stronger showing by Joe Biden than was actually the case. We all held our breath, and he did indeed win, but not by as much as predicted. With the 2024 election, the feeling of impending doom that Trump would win again returned, and it was borne out by the final results, in spite of some polls that suggested Harris would win.
I found Strength in Numbers after watching a discussion you had with Paul Krugman and decided to follow you on Substack when you launched Strength in Numbers. I quickly upgraded to paid because you explain your methods and how you arrive at results. I do not understand the statistical details as well as others (my degrees are in English literature), but I do respect that conclusions are data driven, which is important in every field.
I'm not linking to these posts to make any particular point about AI, but rather to illustrate Nate's commitment to intellectual honesty and epistemic humility. Or lack thereof.
As my Dad used to always tell me, don't let the bastards get you down. You do admirable work, I trust it, and that's why I'm a paying subscriber. Someone who is really trying to be constructive would pick up the phone and have a conversation.
Social media is many things, but gutless is certainly near the top of the list. Along with incredibly boring.
I also have a question for Morris. Polls seem to either poll likely voters or registered voters. Could you explain what the advantages versus disadvantage of each group is and whether it makes any significant difference as the elections get closer. In other words do polls using likely voters become more accurate the closer to an election we get or do both poll types simply become more accurate at about the same rate?
One of the comments complained about Morris having a political bias. I hope Morris will address that issue sometime in one of his future articles. I think everyone has some form of bias politically and pollsters should at least check their processes and tell us what they think are their bias and how they address these bias in their polling. I know my bias are very anti-Trump, anti-oligarch, anti-trickle down economics, and pro education and science to name a few.
The feeling is mutual! One of the reasons I chose to write this Substack instead of taking another mainstream job is to develop personal relationships with readers. It’s very rewarding.
Nate silver is attempting to create a brand. When one is engaged in branding one makes decisions deferring to power and wealth.
Elliot knows the Talking Heads on social media are threatened by him. Ignoring them and letting them make the noise is great and very perceptive. It is in line with his suggestion we turn off social media. Carry on Elliott.
I've never been a social media user, so I can't comment there, but I can tell you that I like your willingness to engage with your readers and respond to our interests.
Speaking of which, has anyone done any recent polling analysis on ICE lately? It feels like the Epstein files are critical, but they are also obscuring other critical things. How do you feel about it?
I followed Nate Silver from 2008. There were signs accruing, but I finally canceled my subscription to his Substack over his non-data driven take on immigration that he never bothered to revise. That's when I found yours.
In #1- I can’t take my eye off the Texas Republican giant red circle. ~63% is low for Trump approval rating amongst Republicans in TX! Lower than FL and CA Republicans!
Keep it up, GEM. Ignore the opinionologists spewing BS.
Also - I still despise everyone who voted for the orange demented idiotic thieving criminal turd, in spite of all evidencethat predicted what has actually happened since January. Your penance is to vote dem in the next 30 elections. Maybe we can undo some of the immense damage that he and his troop of cretins have inflicted on this country.
Your newsletter is one of the best things I've ever come across on Substack. Keep up the great work, and keep dispelling the simplistic narratives of the pundit/strategist/insider class.
During Trump's first term, I largely bought into the "move to the middle" mantra, especially in rural states, but the 2018 midterms proved to me how misguided that was. There were a handful of red state Democratic senators up for re-election (Claire McCaskill, Heidi Heitkamp, Joe Donnelly) who all seemed to spend more time running against their own party than against Trump, and they all lost handily. Not saying that progressives would have done any better in those states, but at least they would have provided a needed contrast between the two parties and had a better chance of motivating voters repulsed by Trump/MAGA.
Florida is an interesting example of where moderation just hasn't worked at all for Democrats. In 2018, they narrowly lost governor (with a very liberal candidate) and senate races. In the years since, they've tried to stem the red tide in the state by consistently nominating moderates, like former Republican Charlie Christ, for statewide office, and they've gotten blown out each time. Turnout on the left has plummeted. Not only are moderate Dems unable to peel away Trump/conservative voters for the simple fact they have a "D" next to their name, but they do nothing to fire up younger, more liberal voters. Not saying nominating AOC and Mamdani types would win them races in Florida right now, but could it be any worse than it has been in recent cycles? Has there ever been a state that has slid more rapidly from purple to red than Florida despite repeated efforts by Democrats to "move to the middle"?
I agree with you on "moderation." Why would Republicans vote for a moderate Democrat when the Republican already ticks their boxes?
That said, I think that some of the terminology is loosely used by corporate media and political pundits. Both Spanberger and Sherrill have been called "moderates," but neither is a Donnelly. (I live in Indiana, so I know.)
Moderation is the latest false path for Dems, it may be worse than neo-liberalism ala Clinton or the no drama Obama years. National health insurance. Liveable national min wage. Carpe Diem! If Sweden can do it, so can we.
How much of the change in Trump approval among independents is driven by composition effects? Presumably the population of folks who identify as independent in 2025 is somewhat different (smaller? Even knowing the size diff would be good perspective) compared to the folks who called themselves independent in 2017.
Good Q and one I answered on email a few times also, so you’re not the only one clocking this. I believe a good amount of the change in Trump’s approval among independents is due to changes in party affiliation (Rs who don’t like Trump in 2017-2025 changing their ID to independent), but since I don’t have panel data to look at the actual switchers, cannot come up with a precise estimate.
I think it right and proper to share methodological questions or disagreements and to debate approaches. It is unacceptable to use ad hominem attacks. I will continue to subscribe to SIN over the coming years because I appreciate Elliott's rational, empirically based analysis. He won't always be right, but that's not the point. Anyway, my question to all (as a UK citizen) is - are the the poor approval ratings for Trump different this time around? He's consistently survived things that politicians and Royals this side of the pond would/do not. The new 'investigation' into the Epstein situation will prevent the release of the files for some time, whether this reduces the political potency of the issue???????
Would love to see Elliott's take on this question, but his approval ratings seem to be largely in line with where they were in his first term. But Trump 2.0 seems to be a far different (and worse) figure than Trump 1.0, when he at least at times showed the pretense (thanks to sensible people around him) to abide by democratic norms and constitutional guardrails. That's all been thrown out the window since time around, which may exacerbate the intensity of disapproval (the first time around, I think a lot of voters gave a pass to his erratic behavior because the country didn't seem much different from the Obama years). Also, the issue of inflation/prices was not nearly as salient in his first term, thanks to decades of low inflation, as it is now. Seems to be that has a greater chance of impacting his approval/popularity long term than was the case in his first term.
My take here is that political scientists for decades have argued party sorting and affective polarization would enable parties to pursue more ideologically extreme agendas and maybe even erode democratic legitimacy. We are seeing validation of those theories
What is the remedy for affective polarization, or is there one in this climate? The Ezra Kleins of the world seem to think that if liberals just meet rural/conservative voters "where they are," they can make inroads, but that doesn't seem likely to me. At some point, you have to have the guts to stand for what you believe in, and call out anti-democratic, fascist behavior, and let the chips fall where they may.
that claimed to use the same methods as they were. That was until I discovered the graph produced by Fiftyplusone. Though the two sources make reference to the same polls and have a similar appearance, Fiftyplusone consistently shows an approval Gap if it is two to three points greater than the one shown by the Times. For example, the NYT currently shows Trump's net approval at -14.0 as opposed to -16.1 at Fiftyplusone. What is Fiftyplusone doing that is different from what the New York Times is doing and why do you believe that the the Fiftyplusone methodology is more accurate?
The simple explanation is that the 50+1 methodology adjusts polls back to the population of all adults, since that’s the reference population in the historical data from firms like Gallup and others. When there’s a big difference in attitudes between registered voters, likely voters, and all adults, this manifests itself as a difference across the aggregators.
FWIW, this is how my average worked for Biden too, and all had him less popular than other averages. These days the all adult population tends to disapprove, and say “don’t know,” and higher rates than RVs and LVs.
Note that for election polls, the average is adjusted to be among likely voters, not all adults or RVs. This is one thing I think led to our more accurate averages in VA and NJ.
I had given up on pollsters. Hillary Clinton lost in 2016, after polls predicted she would win. At the time, both my husband and I had a sense that Trunp might win. In the 2020 election, pollsters predicted a stronger showing by Joe Biden than was actually the case. We all held our breath, and he did indeed win, but not by as much as predicted. With the 2024 election, the feeling of impending doom that Trump would win again returned, and it was borne out by the final results, in spite of some polls that suggested Harris would win.
I found Strength in Numbers after watching a discussion you had with Paul Krugman and decided to follow you on Substack when you launched Strength in Numbers. I quickly upgraded to paid because you explain your methods and how you arrive at results. I do not understand the statistical details as well as others (my degrees are in English literature), but I do respect that conclusions are data driven, which is important in every field.
Keep up the great work!
Here is Nate Silver responding to a post from prominent AI skeptic Ed Zitron in April: https://x.com/NateSilver538/status/1917250626901217542
Here is Nate Silver posting about AI last month: https://x.com/NateSilver538/status/1977058560401195435
I'm not linking to these posts to make any particular point about AI, but rather to illustrate Nate's commitment to intellectual honesty and epistemic humility. Or lack thereof.
As my Dad used to always tell me, don't let the bastards get you down. You do admirable work, I trust it, and that's why I'm a paying subscriber. Someone who is really trying to be constructive would pick up the phone and have a conversation.
Social media is many things, but gutless is certainly near the top of the list. Along with incredibly boring.
I also have a question for Morris. Polls seem to either poll likely voters or registered voters. Could you explain what the advantages versus disadvantage of each group is and whether it makes any significant difference as the elections get closer. In other words do polls using likely voters become more accurate the closer to an election we get or do both poll types simply become more accurate at about the same rate?
Great Q&A topic! Will address it at the next one.
One of the comments complained about Morris having a political bias. I hope Morris will address that issue sometime in one of his future articles. I think everyone has some form of bias politically and pollsters should at least check their processes and tell us what they think are their bias and how they address these bias in their polling. I know my bias are very anti-Trump, anti-oligarch, anti-trickle down economics, and pro education and science to name a few.
It’s a fair question and deserves a fair response!
Thanks. I really appreciate your responses. Its one of the things I like about your site.
The feeling is mutual! One of the reasons I chose to write this Substack instead of taking another mainstream job is to develop personal relationships with readers. It’s very rewarding.
Nate silver is attempting to create a brand. When one is engaged in branding one makes decisions deferring to power and wealth.
Elliot knows the Talking Heads on social media are threatened by him. Ignoring them and letting them make the noise is great and very perceptive. It is in line with his suggestion we turn off social media. Carry on Elliott.
I've never been a social media user, so I can't comment there, but I can tell you that I like your willingness to engage with your readers and respond to our interests.
Speaking of which, has anyone done any recent polling analysis on ICE lately? It feels like the Epstein files are critical, but they are also obscuring other critical things. How do you feel about it?
Not about ICE directly, but do have a question about preference for immigration/asylum policy in the next poll.
F Nate Silver then
I followed Nate Silver from 2008. There were signs accruing, but I finally canceled my subscription to his Substack over his non-data driven take on immigration that he never bothered to revise. That's when I found yours.
In #1- I can’t take my eye off the Texas Republican giant red circle. ~63% is low for Trump approval rating amongst Republicans in TX! Lower than FL and CA Republicans!
Keep it up, GEM. Ignore the opinionologists spewing BS.
Also - I still despise everyone who voted for the orange demented idiotic thieving criminal turd, in spite of all evidencethat predicted what has actually happened since January. Your penance is to vote dem in the next 30 elections. Maybe we can undo some of the immense damage that he and his troop of cretins have inflicted on this country.
Your newsletter is one of the best things I've ever come across on Substack. Keep up the great work, and keep dispelling the simplistic narratives of the pundit/strategist/insider class.
During Trump's first term, I largely bought into the "move to the middle" mantra, especially in rural states, but the 2018 midterms proved to me how misguided that was. There were a handful of red state Democratic senators up for re-election (Claire McCaskill, Heidi Heitkamp, Joe Donnelly) who all seemed to spend more time running against their own party than against Trump, and they all lost handily. Not saying that progressives would have done any better in those states, but at least they would have provided a needed contrast between the two parties and had a better chance of motivating voters repulsed by Trump/MAGA.
Florida is an interesting example of where moderation just hasn't worked at all for Democrats. In 2018, they narrowly lost governor (with a very liberal candidate) and senate races. In the years since, they've tried to stem the red tide in the state by consistently nominating moderates, like former Republican Charlie Christ, for statewide office, and they've gotten blown out each time. Turnout on the left has plummeted. Not only are moderate Dems unable to peel away Trump/conservative voters for the simple fact they have a "D" next to their name, but they do nothing to fire up younger, more liberal voters. Not saying nominating AOC and Mamdani types would win them races in Florida right now, but could it be any worse than it has been in recent cycles? Has there ever been a state that has slid more rapidly from purple to red than Florida despite repeated efforts by Democrats to "move to the middle"?
I agree with you on "moderation." Why would Republicans vote for a moderate Democrat when the Republican already ticks their boxes?
That said, I think that some of the terminology is loosely used by corporate media and political pundits. Both Spanberger and Sherrill have been called "moderates," but neither is a Donnelly. (I live in Indiana, so I know.)
Moderation is the latest false path for Dems, it may be worse than neo-liberalism ala Clinton or the no drama Obama years. National health insurance. Liveable national min wage. Carpe Diem! If Sweden can do it, so can we.
How much of the change in Trump approval among independents is driven by composition effects? Presumably the population of folks who identify as independent in 2025 is somewhat different (smaller? Even knowing the size diff would be good perspective) compared to the folks who called themselves independent in 2017.
Good Q and one I answered on email a few times also, so you’re not the only one clocking this. I believe a good amount of the change in Trump’s approval among independents is due to changes in party affiliation (Rs who don’t like Trump in 2017-2025 changing their ID to independent), but since I don’t have panel data to look at the actual switchers, cannot come up with a precise estimate.
I think it right and proper to share methodological questions or disagreements and to debate approaches. It is unacceptable to use ad hominem attacks. I will continue to subscribe to SIN over the coming years because I appreciate Elliott's rational, empirically based analysis. He won't always be right, but that's not the point. Anyway, my question to all (as a UK citizen) is - are the the poor approval ratings for Trump different this time around? He's consistently survived things that politicians and Royals this side of the pond would/do not. The new 'investigation' into the Epstein situation will prevent the release of the files for some time, whether this reduces the political potency of the issue???????
Would love to see Elliott's take on this question, but his approval ratings seem to be largely in line with where they were in his first term. But Trump 2.0 seems to be a far different (and worse) figure than Trump 1.0, when he at least at times showed the pretense (thanks to sensible people around him) to abide by democratic norms and constitutional guardrails. That's all been thrown out the window since time around, which may exacerbate the intensity of disapproval (the first time around, I think a lot of voters gave a pass to his erratic behavior because the country didn't seem much different from the Obama years). Also, the issue of inflation/prices was not nearly as salient in his first term, thanks to decades of low inflation, as it is now. Seems to be that has a greater chance of impacting his approval/popularity long term than was the case in his first term.
My take here is that political scientists for decades have argued party sorting and affective polarization would enable parties to pursue more ideologically extreme agendas and maybe even erode democratic legitimacy. We are seeing validation of those theories
What is the remedy for affective polarization, or is there one in this climate? The Ezra Kleins of the world seem to think that if liberals just meet rural/conservative voters "where they are," they can make inroads, but that doesn't seem likely to me. At some point, you have to have the guts to stand for what you believe in, and call out anti-democratic, fascist behavior, and let the chips fall where they may.
Thank you for doing this work. I read your column regularly and it improves my understanding of the political landscape.
After FiveThirtyEight went off the air I started following a presidential approval graph produced by the New York Times,
https://www.nytimes.com/interactive/polls/donald-trump-approval-rating-polls.html?smid=nytcore-android-share
that claimed to use the same methods as they were. That was until I discovered the graph produced by Fiftyplusone. Though the two sources make reference to the same polls and have a similar appearance, Fiftyplusone consistently shows an approval Gap if it is two to three points greater than the one shown by the Times. For example, the NYT currently shows Trump's net approval at -14.0 as opposed to -16.1 at Fiftyplusone. What is Fiftyplusone doing that is different from what the New York Times is doing and why do you believe that the the Fiftyplusone methodology is more accurate?
Please respond to this comment, Elliott.
Hi there! Good question and the methodology is explained here:
https://fiftyplusone.news/methodology
The simple explanation is that the 50+1 methodology adjusts polls back to the population of all adults, since that’s the reference population in the historical data from firms like Gallup and others. When there’s a big difference in attitudes between registered voters, likely voters, and all adults, this manifests itself as a difference across the aggregators.
FWIW, this is how my average worked for Biden too, and all had him less popular than other averages. These days the all adult population tends to disapprove, and say “don’t know,” and higher rates than RVs and LVs.
Note that for election polls, the average is adjusted to be among likely voters, not all adults or RVs. This is one thing I think led to our more accurate averages in VA and NJ.
Thanks. That clarifies things for me