The paper's arguments hinge on several methodological flaws, and it dismisses valid empirical alternatives. Plus: Gerrymandering, and the Republicans' lost lead on economics
“I have a half-baked thesis that the more uncertainty surrounds some empirical claim, the more heated the argument about that claim. It should be the opposite, I think. If the evidence is pretty mixed or uncertain, then that should (maybe?) short-circuit the argument. The right view is basically “it’s hard to know.” …
“So, do I think there are zero electoral benefits to moderation? Small benefits? Maybe even a counterintuitive penalty for moderation? I’m not sure! My point is that we’re having heated debates about this question that seem vastly out of proportion to our actual ability to answer the question. The correct strategy, for now, is to embrace the uncertainty.”
The NYT has always had a status quo lean (which favors - surprise - white men). Thanks for pointing out the ongoing holes in their political analysis which far too many people rely upon.
Am becoming more convinced the US has to split into 2 nations: for short-hand, put it split into red and blue states.
What % of the American people support / would favor and want to proceed to separate the two feuding camps into a pro-democracy nation and an anti-democracy nation?
We need a divorce.
And I don't see why we couldn't hold a National Referendum to decide the issue...maybe a 55% threshold to determine it one way or the other. (The average highest popular vote margin since 1964, with a 53% threshold, is 57%--but a president has received even as much as 53% only twice since 1988 (GHW Bush in '88, and B. Obama in '08); Nixon tallied 61% in '72, and Reagan 59% in '84).
Probably a bad idea--but, otherwise, I see a majority of the people having to live under autocratic rule.
Disheartening to see the NYT make basic data analysis errors on such an important question. The most important idea I emphasized when teaching undergrad students in my Psychology Research Methods course was that psychological phenomena are COMPLEX. Hard to explain with one variable. The key thing to do when trying to establish that one variable has a causal effect on human performance (eg voting) was to rule out the possibility that the effect was actually due to a "third" variable that was correlated with (confounded with) the var you are interested. My smart undergrads got this. The NYT does not. But wait I suspect the the NYT writers of poll/stats related article do understand this. So this leaves another explanation....
Excellent article and an excellent explanation of the average voter. Does anyone poll or analyze why so many voters seem to vote against their self-interest? Farmers being a group that seem to support Trump yet many of his actions are directly against farmers' self-interests. Tariffs that destroy foreign markets and buying foreign beef being two obvious actions by Trump that are adverse to many farmers.
Thanks for your work, and for providing the percentage (10%) of voters who make a choice depending on how 'moderate' or 'progressive' a candidate is. I didn't bother to read the NYT piece, considering its premise to be based on a fiction.
Hold up. The main thrust of your first point is that most of the advantage that the NYT sees falls away when you adjust for incumbency, since most incumbents are moderates. I think using this as the basis for arguing that moderation doesn't help is flawed at best.
Incumbents are, by definition, the candidate who won the last election. I think the fact that most incumbents are moderates (again, the basis of your argument) is already a damning point against your argument. Restating it, it means that people who have won the previous election (between 2-6 years prior) have more moderate views than those who did not win the previous election, and are better at winning the next election. By trying to control for incumbency when incumbency is so highly correlated with electoral victory, you're throwing the baby out with the bathwater.
Secondly, I think most people, including the NYT use Harris's win percentage because Harris is a known constant in every district and state. Your comparison is to a "replacement Congressional Democrat," who is apparently already a lot more moderate than the national face of the party. ("Democratic voters in suburban Michigan or southern Louisiana would never elect a candidate as progressive as Harris... for a House district; they would nominate a Democrat closer to the median Democratic congressperson.") That is to say, people as progressive as Kamala Harris - who, I will note, a lot of commentators already lambaste as too moderate - wouldn't even get through a primary in a lot of these districts. I think you've already conceded the larger point, even while trying to argue the opposite.
And besides, the median Democratic congressperson is also relatively moderate incumbent (see my previous point) - in effect, your argument boils down to, moderates wouldn't enjoy as much of an advantage in electoral victory if you compared against an imaginary candidate who was more moderate than Harris.
The other thing is a much deeper flaw that I think is in both yours and NYT's argument: the districts/states that matter in terms of holding the House and Senate and winning the Electoral College are not the average of all the states. They are very distinct districts with ideological slants that are decidedly to the right of the Democratic party. Candidates in these states often run against the Democratic party because the national party is so out of step with the voters they need to win. They are being dragged down because the national image of the party is either not moderate or, at least, not welcoming enough to moderates. And that is the real problem.
"...empirical studies show that upwards of just 10% of voters have a coherent mental model of where the parties stand on most issues."
If only there were a Nobel Prize for political analysis, the committee could just follow the lead of the Economics prize in 2017 - "Thaler’s work shows that assuming human beings are predictably irrational is the most rational approach to studying their behavior." wrote Derek Thompson in the Atlantic.
While Elliott waits for his prize, I'll also note the common fallacy of assuming that everyone is a data nerd like us. Breaking news - they are not. SAD!
Anecdotal observation, from watching elections in Indiana the last eight years, is that Democratic candidates who tried for the "moderate" lane by saying that they would not vote for Pelosi as speaker, or that "Washington is broken," so we must "reach across the aisle" in Congress, do not fare well with some Democrats (who don't turn out to vote), or most Republicans (who do turn out to vote). The worst example was when Senator Joe Donnelly ran an ad for re-election to the Senate in 2022 in which he touted his ability to work with both sides, ending with "let's help President Trump build that wall." His opponent, Mike Braun, said, "That's when I knew I was going to win." In other words, why elect a Democrat if the Republican already fulfills the criteria?
At the same time, it is possible to put forth non-moderate views (I hesitate to call this one progressive) that help sink a campaign, such as the Democratic candidate for my House district in 2022, who when asked a question on U.S. defense policy in a debate, blurted out that what we really need is a Department of Peace. (Seriously, did no one prep this guy?) That was in a debate where the Republican candidate did not even show up, after entering the race late after the incumbent died in an accident, yet he won by a hefty margin, even with a Libertarian siphoning off some of the vote.
I was heartened that Democrats focusing more on specific issues in the 2024 governor's race and Attorney General's race pulled in 40% of the vote. Building on that might bring change to Indiana.
Good work! I do have a question on the baseline change though. It's very possible I am missing something here but it seems like changing out the baseline replacement candidate for a more moderate candidate than Harris would imply some truth to the notion of policy moderation effects. Is the issue just the size of the effect and the exact meaning of "moderation"? Is it possible that the effects are not linearly distributed (eg there are diminishing returns to moderation as you move away from the left most positions of the party)?
I agree with most of Elliott’s analysis here, but he makes a mistake by putting too much weight on the 95% confidence interval in judging the impact of various moderation measures. He argues that a moderation measure that fails the 95%-confidence standard simply has no impact at all. The 95% confidence interval may make sense in a context like drug testing where there may be important unknown adverse effects of concluding that a variable (e.g., use of a drug) is efficacious (think Thalidomide), but it is not likely to be relevant in a political context. Think of it this way: if you had a political strategy that had an 80% chance of increasing your probability of winning, would you adopt it? The reliance on the 95% confidence interval says “No”, you should only adopt the strategy if it has a 95% probability of increasing your chance to win. Many economists (such as Deirdre McCloskey) have criticized the slavish and unthinking adherence to the 95% confidence interval; Elliott should join their ranks.
Thanks for making your syntax available. Your use of a generalized linear model makes a lot of sense to me.
I think this is very wise from Prof. John Sides:
“I have a half-baked thesis that the more uncertainty surrounds some empirical claim, the more heated the argument about that claim. It should be the opposite, I think. If the evidence is pretty mixed or uncertain, then that should (maybe?) short-circuit the argument. The right view is basically “it’s hard to know.” …
“So, do I think there are zero electoral benefits to moderation? Small benefits? Maybe even a counterintuitive penalty for moderation? I’m not sure! My point is that we’re having heated debates about this question that seem vastly out of proportion to our actual ability to answer the question. The correct strategy, for now, is to embrace the uncertainty.”
Whole short piece here:
https://mailchi.mp/goodauthority.org/the-unmaking-of-a-viral-graph-15080486?e=aeca82172d
Which is not to say Elliott isn’t right to fight back!
The NYT has always had a status quo lean (which favors - surprise - white men). Thanks for pointing out the ongoing holes in their political analysis which far too many people rely upon.
Am becoming more convinced the US has to split into 2 nations: for short-hand, put it split into red and blue states.
What % of the American people support / would favor and want to proceed to separate the two feuding camps into a pro-democracy nation and an anti-democracy nation?
We need a divorce.
And I don't see why we couldn't hold a National Referendum to decide the issue...maybe a 55% threshold to determine it one way or the other. (The average highest popular vote margin since 1964, with a 53% threshold, is 57%--but a president has received even as much as 53% only twice since 1988 (GHW Bush in '88, and B. Obama in '08); Nixon tallied 61% in '72, and Reagan 59% in '84).
Probably a bad idea--but, otherwise, I see a majority of the people having to live under autocratic rule.
This post is why I cancelled my Times monthly subscription and used that money to subscribe to Strength in Numbers and Paul Krugman’s Substacks.
Disheartening to see the NYT make basic data analysis errors on such an important question. The most important idea I emphasized when teaching undergrad students in my Psychology Research Methods course was that psychological phenomena are COMPLEX. Hard to explain with one variable. The key thing to do when trying to establish that one variable has a causal effect on human performance (eg voting) was to rule out the possibility that the effect was actually due to a "third" variable that was correlated with (confounded with) the var you are interested. My smart undergrads got this. The NYT does not. But wait I suspect the the NYT writers of poll/stats related article do understand this. So this leaves another explanation....
Excellent article and an excellent explanation of the average voter. Does anyone poll or analyze why so many voters seem to vote against their self-interest? Farmers being a group that seem to support Trump yet many of his actions are directly against farmers' self-interests. Tariffs that destroy foreign markets and buying foreign beef being two obvious actions by Trump that are adverse to many farmers.
Thanks for your work, and for providing the percentage (10%) of voters who make a choice depending on how 'moderate' or 'progressive' a candidate is. I didn't bother to read the NYT piece, considering its premise to be based on a fiction.
Wear their criticism as a badge of honor! So happy I canceled my subscription some time ago.
You rock.
Hold up. The main thrust of your first point is that most of the advantage that the NYT sees falls away when you adjust for incumbency, since most incumbents are moderates. I think using this as the basis for arguing that moderation doesn't help is flawed at best.
Incumbents are, by definition, the candidate who won the last election. I think the fact that most incumbents are moderates (again, the basis of your argument) is already a damning point against your argument. Restating it, it means that people who have won the previous election (between 2-6 years prior) have more moderate views than those who did not win the previous election, and are better at winning the next election. By trying to control for incumbency when incumbency is so highly correlated with electoral victory, you're throwing the baby out with the bathwater.
Secondly, I think most people, including the NYT use Harris's win percentage because Harris is a known constant in every district and state. Your comparison is to a "replacement Congressional Democrat," who is apparently already a lot more moderate than the national face of the party. ("Democratic voters in suburban Michigan or southern Louisiana would never elect a candidate as progressive as Harris... for a House district; they would nominate a Democrat closer to the median Democratic congressperson.") That is to say, people as progressive as Kamala Harris - who, I will note, a lot of commentators already lambaste as too moderate - wouldn't even get through a primary in a lot of these districts. I think you've already conceded the larger point, even while trying to argue the opposite.
And besides, the median Democratic congressperson is also relatively moderate incumbent (see my previous point) - in effect, your argument boils down to, moderates wouldn't enjoy as much of an advantage in electoral victory if you compared against an imaginary candidate who was more moderate than Harris.
The other thing is a much deeper flaw that I think is in both yours and NYT's argument: the districts/states that matter in terms of holding the House and Senate and winning the Electoral College are not the average of all the states. They are very distinct districts with ideological slants that are decidedly to the right of the Democratic party. Candidates in these states often run against the Democratic party because the national party is so out of step with the voters they need to win. They are being dragged down because the national image of the party is either not moderate or, at least, not welcoming enough to moderates. And that is the real problem.
"...empirical studies show that upwards of just 10% of voters have a coherent mental model of where the parties stand on most issues."
If only there were a Nobel Prize for political analysis, the committee could just follow the lead of the Economics prize in 2017 - "Thaler’s work shows that assuming human beings are predictably irrational is the most rational approach to studying their behavior." wrote Derek Thompson in the Atlantic.
While Elliott waits for his prize, I'll also note the common fallacy of assuming that everyone is a data nerd like us. Breaking news - they are not. SAD!
Anecdotal observation, from watching elections in Indiana the last eight years, is that Democratic candidates who tried for the "moderate" lane by saying that they would not vote for Pelosi as speaker, or that "Washington is broken," so we must "reach across the aisle" in Congress, do not fare well with some Democrats (who don't turn out to vote), or most Republicans (who do turn out to vote). The worst example was when Senator Joe Donnelly ran an ad for re-election to the Senate in 2022 in which he touted his ability to work with both sides, ending with "let's help President Trump build that wall." His opponent, Mike Braun, said, "That's when I knew I was going to win." In other words, why elect a Democrat if the Republican already fulfills the criteria?
At the same time, it is possible to put forth non-moderate views (I hesitate to call this one progressive) that help sink a campaign, such as the Democratic candidate for my House district in 2022, who when asked a question on U.S. defense policy in a debate, blurted out that what we really need is a Department of Peace. (Seriously, did no one prep this guy?) That was in a debate where the Republican candidate did not even show up, after entering the race late after the incumbent died in an accident, yet he won by a hefty margin, even with a Libertarian siphoning off some of the vote.
I was heartened that Democrats focusing more on specific issues in the 2024 governor's race and Attorney General's race pulled in 40% of the vote. Building on that might bring change to Indiana.
Good work! I do have a question on the baseline change though. It's very possible I am missing something here but it seems like changing out the baseline replacement candidate for a more moderate candidate than Harris would imply some truth to the notion of policy moderation effects. Is the issue just the size of the effect and the exact meaning of "moderation"? Is it possible that the effects are not linearly distributed (eg there are diminishing returns to moderation as you move away from the left most positions of the party)?
Bravo.
I agree with most of Elliott’s analysis here, but he makes a mistake by putting too much weight on the 95% confidence interval in judging the impact of various moderation measures. He argues that a moderation measure that fails the 95%-confidence standard simply has no impact at all. The 95% confidence interval may make sense in a context like drug testing where there may be important unknown adverse effects of concluding that a variable (e.g., use of a drug) is efficacious (think Thalidomide), but it is not likely to be relevant in a political context. Think of it this way: if you had a political strategy that had an 80% chance of increasing your probability of winning, would you adopt it? The reliance on the 95% confidence interval says “No”, you should only adopt the strategy if it has a 95% probability of increasing your chance to win. Many economists (such as Deirdre McCloskey) have criticized the slavish and unthinking adherence to the 95% confidence interval; Elliott should join their ranks.