These are my quick thoughts. I’m sure I’ve missed a lot, so feel free to add your perspectives in comments.
1. The outcome
In 2016, Hillary Clinton narrowly won the popular vote and lost in the electoral college. In 2020 the vote shifted to the Democrats. In 2024 there was a swing back to Donald Trump, who this time may well have won the popular as well as the electoral vote. The voting patterns of the states didn’t move much relative to each other, and there was a comparable swing in congressional races, which makes sense given partisan polarization. As with Biden in 2020, Trump’s victory in 2024 was unambiguous but a narrow win compared to the twentieth-century norm. Recent elections have been fought on narrow battlegrounds, with incumbents getting thrown out each time.
What happened yesterday? The first answer from political science is the economy. More specifically, dissatisfaction with the economy. More generally, dissatisfaction manifesting itself in anti-incumbent feeling. Ever since Rosenstone’s 1983 book, we’ve thought that party, not candidate or campaign, is what matters most in presidential elections, and recent elections do not contradict that view. We’ll never know what would’ve happened had Biden stayed on as the Democratic nominee, but the ultimate election outcome might have been similar—with the difference that the loss could’ve been easily pinned on him, whereas now there is no such easy answer. In that way, this election is similar to 2008, in which the incumbent party took its best shot and lost. The Republican party could’ve reacted to 2008 by moving to the center, but they didn’t. It’s not clear what will happen with the Democratic party going forward. Democratic officeholders and voters seem to be pretty uniformly anti-Republican and anti-Trump, however the many endorsements of Harris from Republicans, most notably from former Trump appointees, would seem to create the opportunity for an anti-Trump coalition with moderate policy positions. I guess this could be the ideal outcome for conservative Republicans who’ve supported Trump, if they could move the opposition toward the center so as to mitigate policy swings following expected in-party losses in upcoming off-year elections.
All the above is about the national and state-level votes. There also have been much noted changes in the parties’ bases of demographic support, something that doesn’t typically come up in models of election outcomes. There seem to have been big changes in the breakdowns of vote by education and sex during the past few elections. It’s never completely clear to me how this would be expected to affect parties’ positioning—in electoral terms, their goal is to get more votes, not votes of any particular demographic groups—but it reflects changes in public opinion.
2. Polls and forecasts
There’s not much to say here. The polls had a level of nonsampling error that was comparable to that in the three previous elections—in 2012, Republicans did better in the polls than in the election; in 2016, 2020, and 2024, polling error was in the Democrats’ direction. The most natural explanation of the error is differential nonresponse—Democrats being more likely than Republicans to answer the polls in these recent elections, to an extent beyond any demographic adjustment. This did not happen in recent off-year elections, so maybe this can be attributed to Trump attracting occasional voters who would not otherwise turn out to vote.
Another factor is voter enthusiasm. The Harris campaign in 2024 was active, with many large rallies and a general sense of active participation. Enthusiasm is good for voter turnout but maybe even better for survey response: after all, something like 60% of people vote, and survey response rates are much much lower. This is related to differential nonresponse. It’s not clear how to best adjust for this in a survey, and it’s possible that likely-voter screening could be counterproductive, if enthusiastic voters are already being sampled.
The forecasts did what they were designed to do: they aggregated economic and polling information, did not overreact to individual polls, and allowed for a potentially large nonsampling errors, leading to wide uncertainty intervals. The only systematic way to do much better with this information would be to include a prior belief in the direction, not just the possible magnitude, of the polling errors. In retrospect there are reasons to have done so; on the other hand, pollsters were aware of the bias from before and were trying to adjust for it, so there was a concern about overcorrection. To put it another way: the polls were in aggregate off in 2024 as they were in 2020 and 2016 (although not in 2018 and 2022); the errors in 2024 had affected the forecasts less, partly because forecasters built the scale of possible polling errors into their models, and partly because the polls, along with the historically-based models, were pointing toward something close to an electoral college tie, so the forecasts happened to produce something close to even odds.
3. Looking forward
Politics isn’t just about elections. With the Republicans in something approaching complete control of the U.S. government (the executive branch, the Supreme Court, the Senate, and possibly the House of Representatives), much of national politics will depend on debates within the Republican party. What are their priorities? How will they resolve intra-party struggles. I assume that all factions of the party can agree on certain tax and spending cuts. Beyond that, I’m not sure.
The closest analogy in recent American politics is the Democratic sweep of Congress and the presidency in 2006-2008, where the limiting factor on policy initiatives were the moderates in the Senate. But back then the Democrats only controlled two of the three branches of government. Then as now, the majority of Supreme Court judges were Republicans, and that constrained the executive and legislative branches’ room to maneuver. The last time there was full single-party control of the federal government was in the 1930s-40s, when the limiting factor was the influence of conservative southern Democrats, and in the mid-1960s, when the Democratic party was so dominant that the United States was sometimes characterized as having a “one-and-half-party system,” similar to that of Japan and other countries that have democratic elections but with a single dominant party that sets the tone, and where much of the politics involves non-ideological struggles between factions of the major party.
The Democrats are not a minor party, though. They consistently win close to or more than half the national vote, they represent nearly half of Congress, and control many states. So their decisions matter too—partly from the now-looming threat of a Democratic resurgence in 2026 and 2028 and partly because, depending on what is on offer from the Republicans, they can attempt to unite in obstruction or attempt cross-party compromise.
I’m offering neither predictions or prescriptions for any factions of either party here. I’m just trying to emphasize that politics isn’t just about elections. The results of elections create the conditions for further politics, aspects of bargaining and decision making on which I have no particular expertise.
I think this election will be as significant as Biden’s poor debate performance: it’s an objective and clear demonstration that many Democrat’s beliefs about the nature of the country and the electorate are flat out wrong. And that Democrats need to shift course.
Information is the most informative when it is against your current prior about the world. And boy is this election a shock. Trump is the single most unpresidential candidate (by traditional metrics) to run in modern history and a majority of the voting population preferred him.
One constant thing is mockery of Allan Lichtman and his hilariously unscientific “key” system for predicting elections based on fundamentals. But I think it’s meaningful that the keys failed so badly here. This is a new era of American politics.
Ethan:
For reasons I’ve discussed before, I don’t think “the keys” ever made sense. But, yes, I agree that when the election is not extremely close and there’s no obvious explanation for the loss, that motivates reconsideration. It’s tricky because the reconsideration can go in different directions. The Republicans lost badly in 2008, but that was explainable as an inevitable consequence of the recession. Then they lost in 2012, which was more of a stunner, and they did recalibrate, but not by moving to the center. By analogy, one can imagine the Democrats shifting course by becoming a more centrist or conservative party (which, again, could in some sense validate the strategy of conservative Trump-haters such as Mitch McConnell to stay on the Trump train), or one could imagine them moving hard left in a mirror image of the Republican swing since 2016—although this could be more difficult for them, as they don’t have the court system on their side. Or they could just hold tight, work hard in state governments, and wait for a swing in public opinion in forthcoming elections. In the short term, though, I’d guess that most of the politics that matter will happen within the Republican party.
Is it a shock? Frankly I think polling, forecasts, and political scientists did pretty well this election cycle.
Somebody:
I don’t think the election outcome was a shock—it was well within everybody’s forecasts. And you can say that forecasts did well in the sense that they allowed for the possibility of systematic polling error.
Did the polls do “pretty well”? As I discussed in my article, Failure and Success in Political Polling and Election Forecasting, it kinda depends on how you look at it. On one hand, predicting public opinion within about 2 percentage points is pretty damn impressive and in most cases will give you a good sense of the outcome and the magnitude of victory. On the other hand, if the election is super-close—which it has been for most of the presidential elections since 2000, a 2-percentage-point systematic error is a problem. An unavoidable problem, but a problem nonetheless.
To put it another way, you could say the polls did pretty well—as long as they were coupled with an understanding of their potential nonsampling error, and indeed they were in this election cycle.
Guess Alan Lickman won’t be featured in NYT anymore.
My own 2024 predictions were based on the idea that the 2020 and 2016 patterns of polling errors were a semi-stable Trump-specific phenomenon that would be repeated in 2024 since he was once again on the ticket. It’s early but it seems I was able to get the electoral college vote exactly right, and made generally smaller vote margin errors than Nate Silver, 538, or the Economist. I don’t think this approach will be replicable in future elections not involving Trump, and as you say the history of errors shows that they vary in direction, but my relative success may suggest that the polls continue to be informative, but more effort needs to go into modeling the direction of the error. Elections have been so close of late that who wins and by how much is almost entirely determined by this question.
In case anyone is interested, I posted my predictions here:
https://www.econjobrumors.com/topic/what-the-polls-really-say-about-november
These were generated at the beginning of October, based only on September polls. I did a refresh on election eve after adding October and November polls, but only shared this with friends via Instagram so no public link. The electoral vote prediction didn’t change but the vote margins got a bit Trumpier across the board, including the popular vote switching to a narrow Trump win.
Ram:
Yes, that makes sense. We considered something similar for the Economist model (see the appendix of this paper) but in the rush to put together the model in time, it was decided not to include it.
Boy was I wrong to doubt that Selzer was a regular Iowan Oracle… Oh wait, she ended up being off by a catastrophic 16 pts.
Sm:
See P.P.S. here. It was never about anyone being an oracle; it’s much more about any poll having both sampling and non-sampling error.
It was not a surprise that the economy was the #1 issue and deciding factor in the election result. The surprise (for me) was that it was only the perception of the economy and not the real economy. Notwithstanding the numerous discussions we have had about difficulties measuring prices and incomes, the disastrous economy that drove so many people’s votes simply isn’t real. Yes, eggs are expensive and mortgage rates are strangling new home buyers, but first time home buyers are far fewer than existing homeowners on low rate mortgages which add to the wealth in the current high interest rate environment. Still, eggs are expensive, but so are people to mow my lawn and plow my snow, and those are the people whose incomes are outpacing inflation. What continues to amaze me is that if you insist that the economy is terrible and say it enough times, then plenty of people believe it. Compare today’s economy with past periods of dissatisfaction with economic conditions: it’s hard to find the terrible economy today that so many people point to.
The one area where things are now worse than in the past is inequality. That I agree with – but it’s hard to see how either party is likely to do much to deal with that (and, in my view, even less likely from the Rs than the Ds).
I know my claims about prices and incomes are debatable, so if you think I am wrong about those then just say so – I won’t go down that rabbit hole, so just ignore my comment.
Dale:
I agree, and debate continues regarding the role of the real economy, perceptions of the economy, and reports of the economy. Data and theory both support the idea that, with increased polarization, the economy will be less predictive of electoral outcomes than in the past; also, though, it’s possible that different aspects of the economy will be important, and this can be connected to the declining role of the mass news media in communicating reports about the economy.
The “economy” means different things to different people, and it seems like you are focusing on more short term issues like current inflation, current mortgage rates, etc. (others might throw in the booming stock market). It seems plausible longer term structural issues play just as much, if not more, of a role in how people feel about the economy. For example, high costs for medical care, education, childcare, and more, on top of the inequality you mentioned. I don’t see either party offering any real solutions to these problems.
Dale, as inequality increases, the number of people suffering considerably goes up, potentially exponentially. At one point I fit a maximum entropy model for income with an average of 1 and given gini. If you look at the number of people in the range say [0,0.1] as a function of gini, it’s at least sort of exponential with gini. It’s essentially 0 with gini less than 0.25 and it’s quite high by the time you’re at say gini of 0.45 or so where US is. I don’t have the numbers to hand right now, but it wasn’t that hard to do the calculations. I just created a radial basis function for log(p(x)) on x in [0,10] and then maximized entropy subject to constraints of integrating to 1 on the interval 0 to 10, averaging 1, and the given gini. Using Optimization.jl in julia it took a few seconds to do for gini values between about 0.2 and 0.5
The bigger inequality gets the less any kind of single number summary of “the economy” means anything. It’s particularly true if we don’t follow individual people through time. Economists tend to look at things like quintiles of income without ever considering how through time they don’t represent the same people!
“Dale, as inequality increases, the number of people suffering considerably goes up”
Impossible that is always truth . I would say that every growth period is started by a significant increase in inequality.
If billionaire wins more 2 billion and worker wins more 20k both got better despite inequality increased assuming inflation did not cut meaningfully.
Yes, I should have said “as inequality increases for a fixed average income”. That is, comparing two places, one with an average income of say $60,000 and gini coefficient of say 30% and another with average income $60,000 and gini 40%, the one with gini 40% has more people suffering. This is always true because to make the gini larger you have to spread the incomes out more, but to keep average constant you have to spread them out in both directions around the average in a balanced way.
“that it was only the perception of the economy and not the real economy”
In 2020, most Americans were living paycheck to paycheck with virtually no emergency fund. They were experiencing already historically unaffordable housing, education, childcare and so on costs. Then in four years time they got a 50% increase in housing, food, transportation, and other costs while getting a 2% annual pay increase if they’re lucky (that’s more than I got).
Dismissing that as “perception of the economy” is a fundamental mistake. Probably the single biggest mistake of the election.
I’m pretty bearish on Bidenomics, but what you’re describing is literally impossible. If they were living paycheck to paycheck, then got a 50% increase in cost of living and a 2% annual pay increase, then they’re dead. What exactly do you think “paycheck to paycheck” means?
Not if they put it on credit cards and other loans:
https://fred.stlouisfed.org/series/RCCCBBALTOT
I agree that households are more leveraged and that’s not great, but real median net worth increased from 2019-2022.
https://www.federalreserve.gov/publications/files/scf23.pdf
Again, I’m on the side of “things are pretty tough” because I don’t think increases in cost of housing and what Amartya Sen considers “costs of basic social functioning” can be reasonably averaged out by more consumption goods and nonsense. But there’s simply no way to reconcile the idea that 50% of Americans were living paycheck to paycheck, then cost of living increased by 50% and pay increased by 2%. In addition to those ideas together being falsifiable, each of those numbers individually is obviously made up. I really think the truth is bad enough without having to make things up.
I’ve been baffled by much the same thing. I’m not a Bidenomics booster for much the same reason as the other replies; I don’t think that you can cancel out much more expensive housing with cheaper movies and clothes to get a net increase in welfare.
However, on some very basic, very simple datapoints, people’s perceptions are wildly out of touch. If I ask people how much a big mac has increased in price from 2019 to 2023, pretty much everyone thinks it’s at least a 50% increase and up to a 150% increase. The answer is like 20% depending on time and location. If I ask people how much increased in the 4 years before that, they generally think it was like 0-2% when the answer is like 12%. This is a very literal, concrete question about numbers people make personal contact with regularly, and the answers are very easy to verify. Yet, seemingly, people think that inflation materialized suddenly in 2020 and skyrocketed wildly out of control, even as they continue to purchase many big macs. The disturbing thing is that many of the people I’m talking about are VERY numerate. It’s hard for me to escape the fact that people generally don’t even really think about numbers at all and just kind of absorb the headline collective consciousness.
I think people who live paycheck to paycheck, and are right on the edge financially, are very strongly aware of the numbers coming in and going out. And that’s the majority of the country.
Two factors to consider.
One is the ideological signal in the surveys of attitudes about the economy – suggesting that people’s view a out the economybshoikd be expected to be tightly correlated to he actual economy..
The second is that apparently surveys show a generalizeable divergence between how people view their own economic situation compared to “the economy” – with views about “the economy” being more pessimistic, relatively.
Our those together and you’ve mind if got a mess. And I think that mess is reflected in this year’s election where, at least to some degree, the dissatisfaction about “the economy” reflects other, underlying causal mechanisms – for example, a manifestation of other feelings (antipathies) in surveys on “the economy,” or an inherent pattern of blaming the current administration for a combo of them having done stuff (that’s never good enough) and not having done stuff.
Maybe the supposed advantage of incumbency has morphed into something quite different – an inherent penalty for being the incumbent (as seen in many elections recently in other countries). Things change.
A new record for typos in that comment.
I dunno, I don’t think I’ve purchased a big mac ever in the last 24 years. But I’ve got a family of 4 with two teenage boys, so I can tell you I used to take my family out to dinner for say thai or mexican food for around $55 in 2019 ish and now it’s around $90. They eat more but not as much more as you might think, 10 year olds who do soccer 3 times a week eat a lot too.
I’ve fortunately got a fixed mortgage, but I can tell you a house for a family of 4 in this area rents for about $5500/mo and in 2019 that would have been probably $3000 ish… If I look at the case-shiller index for housing in Los Angeles, https://fred.stlouisfed.org/series/LXXRSA house price index now is 441.6 while in late 2019 it was 288.8 or a factor of 1.53, combined with increased interest rates, undoubtedly monthly payments are up more than 1.5. So inflation of the core expenses of a 4 person family since 2019 has been a factor of around 1.6-1.7 ish.
Now, if I look at CPI all items, https://fred.stlouisfed.org/series/CPIAUCSL I see today it’s around 314.7 and in late 2019 it was 257 ish so it acknowledges about a factor of about 1.22
So basically, get out of here with that 1.22. Economists as a profession, and with respect to macroeconomic issues are a laughable bunch of fools and Democrats having adopted a neoliberal view of Economists as technocratic know it alls who should drive policy are getting bitten by the fact that eventually everyone can see the emperor has no clothes.
Daniel –
The increased prices you described may be good for explaining why middle- to upper-class whites would he angry about “the economy” – except apparently they aren’t particularly angry about it or at least didn’t express that through their vote, especially if they’re female and white with a 4-year degree.
So actually, the cost of Big Mac may be more instructive here.
I’ve seen reports that the wage/inflation ratio grew favorably particularly for lower wage workers.
This reminds me of the Brexit leave vote which as supposedly an expression of anger about economic resentment – except actually the”leave” vote didnt particularly correlate with things like decline in earnings or increases in unemployment..
Joshua, sure, unfortunately I don’t have personal anecdotes to bring to bear on the question of what say lower middle class hispanic families in Texas think or whatever. And while I’ve done a number of things to measure cost of living in the past using ACS data and the like, I don’t have those analyses for this time period.
Fortunately Blair Fix has said a lot of useful stuff on this topic and you can read about it online https://economicsfromthetopdown.com/2022/12/15/inflation-everywhere-and-always-differential/
Inflation just isn’t well described by a single number, nor is “real income” and the persistent tendency to quote single numbers instead of discussing distributions and the transport through time of individuals at different points in the distribution is one of the most laughable parts of macroecon. It’s like trying to discuss the gas giant Jupiter without discussing turbulence.
Blair also has a lot to say that I agree with about the situation specifically with housing prices and income inequality
https://economicsfromthetopdown.com/2024/10/23/the-american-housing-crisis-a-theft-not-a-shortage/
Basically that while average incomes are increasing, the number of people in grinding poverty is ALSO increasing. Both are possible thanks to the miracle of distributional thinking.
Daniel –
I’ll check out your links. Meanwhile, this chart speaks volumes. Pretty stunning, imo. Hard to imagine how the trends are consistent with any actual measures of economic health.
https://x.com/AlecStapp/status/1854243109418209739
The ideological signal in views about “the economy” is far from unique to this year’s election cycle.
https://x.com/AlecStapp/status/1854243109418209739
We are not going to resolve this debate. But a couple of observations that need to be dealt with: many of the outrageous cost of living situations that Daniel refers to come from California – a state where Harris prevailed. Rents in many of the swing state areas do not share that experience. Households living “paycheck to paycheck” deserve some more analysis. It is easy to end up in that situation if you spend beyond your means, which happens to be an American institution. Even the idea that you should always be better off than you were 4 years ago (despite COVID and other major events) presumes a lot. I don’t deny the fact that peoples’ expectations have not matched reality, but at least part of the problem is there expectations.
None of this is to deny that there is real suffering and a significant segment of the population that has suffered economically (although I still think most of them were suffering before the recent bout with inflation, which was actually short-lived by historical standards). As far as I can tell, those segments mostly voted for Harris. I still haven’t seen convincing evidence that the groups that went decisively for Trump were groups that were suffering economically.
We can get bogged down in the weeds and examine all sorts of data, but I think we will fill the blog with a bunch of distracting and inconclusive material. The responses I would be interested in are clear statements about what segments of the population have truly been suffering due to inflation and how that relates to their voting patterns in this election.
What I find most disturbing about this election (and it doesn’t depend on who won) is the prevailing belief that the role of government is to put money in people’s pockets. That was the appeal made by both parties. It seems that the idea that government exists to better our collective well-being, that there are public purposes that private interests cannot address, and that we should want leaders who exhibit personal traits of compassion, intelligence, belief that truth matters, and that bullying is inappropriate – is all but dead for around half of the population (probably a run on sentence). People are motivated by narrow self-interest, but I thought our big brains were supposed to offer some protection against the worst of those impacts. That seems foreign to so many people.
> So basically, get out of here with that 1.22. Economists as a profession, and with respect to macroeconomic issues are a laughable bunch of fools and Democrats having adopted a neoliberal view of Economists as technocratic know it alls who should drive policy are getting bitten by the fact that eventually everyone can see the emperor has no clothes.
Inflation in most developed countries has been something like 10% for two years, so a total of 20%. I read an interesting article a few weeks ago about how it has probably been much higher than officially stated because of quality being reduced or sizes going down and also being which is hard to measure but could push up that 20% total to something much higher. E.g. a chocolate bar might still cost £1 but go from 30% cocoa to 20%, have milk changed for palm oil, or have new packaging so it stays £1 but goes from 100g to 80g.
On the flip side, he also said that this means inflation in the 90s and early 2000s would have been much lower than stated because the price rises often came with quality increases.
David. as discussed in the Blair Fix blog link on inflation a few posts back, it just **barely** makes sense to discuss inflation as a single number. It’s a little like saying the velocity of a hurricane is 3 miles per hour because that’s how fast the eye is moving towards the mainland. The graphs Blair put on his post show that products have
Some products have gone down 40% others are up 120%, whether your net expenses are up 1.1x since 2019 or 1.8x since 2019 is all down to what kind of basket of goods your family needs and where you live and soforth. The standard deviation of price rise was about equal to the mean price rise over this period. That’s like saying your Weight is 150 lbs +- 150 lbs.
Fortunately he’s got all the graphs and references to all the sources and things in his post.
” mortgage rates are strangling new home buyers, but first time home buyers are far fewer than existing homeowners on low rate mortgages which add to the wealth in the current high interest rate environment”
This claim is worth looking at too. It seriously underestimates the pain the housing market is causing. Here is some specific issues:
-skyrocketing housing leads to skyrocketing rents.
-something like 80% of the population can not afford the average house. There’s no indication this is temporary. That has a strong effect on peoples future plans even if they aren’t looking to buy now.
-housing insurance and taxes have skyrocketed too. So even if you own your home outright, you can still have much higher housing costs. Especially those on fixed incomes.
-most people who already owned a home find they can’t move unless they want to trade in a 2% mortgage for a 7% one and much higher taxes.
Was this true in 2022? That year, the Democrats greatly over performed polls. They were expected to lose three Senate seats, for instance, in NV, PA, and GA. (Google 538’s forecast for the Senate that year.) They won all three. If these quantitative economic data were a primary concern to voters this year, then they should have been for the party in power in the Senate, in 22. This, obviously, wasn’t the case.
What effect this had on any election is a separate question. I was denying to claim that this is just perception.
To take one telling example, even if a retiree owns their home outright, the increases in property taxes + home insurance + maintenance costs can be enough to make their homes unaffordable. There’s clear evidence this is already happening in parts of the US.
To claim this is just false perceptions is ridiculous.
There is good reason to believe that statistical aggregates failed to reflect important changes in real incomes and living standards during the post-pandemic period; see: https://www.ineteconomics.org/perspectives/blog/inet-research-and-the-2024-election
Incidentally, I think Tom Ferguson has become one of the most insightful analysts of the US political economy writing in more or less real time.
I took a look at one of the papers from the site you linked (though the 100+ pages deterred me from reading carefully). While it raises many interesting questions, I am quite unimpressed. First of all, the paper sets up a strawman and focuses on tearing that down: the idea that labor markets have undergone some major transformation. It distracts from the question of measurement of prices and incomes that are certainly worth some debate. I noticed one argument, that the increases in real wages at the bottom of the distribution don’t account for the “fact” that much of this is compensation for increased risk (e.g. jobs where people are exposed to more germs, such as hairdressers). However, one of their other findings is that real weekly incomes have fallen for many low income people even as their wages have increased – this is due to less hours worked. They then fail to mention that less hours worked means more leisure time and that many of the reduced hours of work are voluntary.
Rather than debate these points in detail – this is not the place for that – I only mean to point out that the research commits many of the problems we have discussed on this blog. Cherry picked arguments, measurement issues, hyped arguments, etc. I do not mean the research is wrong, only that it is far from clear whether low income workers have fared better or worse in the past 4 years than the 4 years before that. Even that comparison is a bit strained – why is the measure whether you’ve gained more in the last 4 years than the prior 4 years, rather than whether you have gained or not. Surely, I can remember some “good old days” when I did better than recently.
None of this explains to me why the groups that voted heavily for Trump do not appear to be the groups in the lowest decile of the distribution (if the exit polls are to be believed). It looks to me like the lowest decile was Harris territory. I find it hard to see that Trump’s victory was due to the votes of the worst off part of the American economy. Certainly the wealthy voted disproportionately for Harris, but I’d like to see how the Trump-heavy parts of the income distribution actually fared during the past 4 years. It is not obvious to me.
Regarding voting patterns, I was looking at this from the exit polls:
less than $30K $30-$50K $50-100K $100-$200K $200K+
Harris 50% 45% 46% 51% 51%
Trump 46% 53% 51% 47% 45%
Of course, these interact with education and race (among others), but my question concerns the disaggregated state of the economy in relation to the voting patterns. It is far from clear to me.
The above thread reinforces Dale’s point that we’re talking so much more about perceptions of the economy than about the economy itself, in a way that I think is different from 1980, 1992, 2008, and other elections where the out-party clearly gained from an economic crisis. Even beyond understandable partisan differences in the long-term economic trajectory, the short-term performance of the economy has become a political football, and this was perhaps particularly harmful to the Democrats in this year’s election, because people on both the left and right found flaws in the economy.
The remarkable thing to me was that the message from Trump was “your economic situation is bad, and I can fix it” and the message from Harris was “no it isn’t” rather than “I can fix it better.” Rather than being given policy choices, voter were given reality choices. And if you didn’t think your economic situation was all that bad, you might still prefer the candidate who runs on improving it.
In past elections, both sides acknowledged a similar state of the economy, and gave the voters policy choices.
Regardless of perceptions of the current economy, I think a key aspect that has not been mentionned yet is that democrats vastly underestimated the number of people who strongly reject the status quo (to avoid saying the system). This is something that can be observed in many countries.
One other thing to ponder is how/why the “economy” relates to the huge gender difference in the vote. Why do men find the economy so much worse than women? This appears to be particularly true for Hispanics and Whites. And in particular age ranges. Those patterns (dependent, of course, on whether the exit polling is accurate) are hard to explain in relation to the real economic conditions.
Dale:
Not just gender. Views on the economy also differ a lot by partisanship!
Methinks the definition of “the economy” has a bit of a Rorschach test in it.
This year, there’s been a correlation between a low approval rating for an incumbent party’s leader, the perception of the economy, and the likelihood of the incumbent party either losing seats or losing power. Consider, France, Austria, the UK, Portugal, the Netherlands, India, and South Africa almost all had unpopular party leaders with similarly low approval ratings and voters saying that the economy was poor (the exception was India’s Modi who had high approval ratings but low job growth). I don’t have historical data to generalize from this point and I also don’t know if this is mentioned in Rosenstone’s book.
Nonetheless, the reason I’m mentioning it is a way to quantify the perception of the economy. Low approval ratings that correlate with the perception of the economy can be used as a prior, say, with economic data to quantify how economic data may be interpreted. Specifically, the more that a party leader’s low approval rating correlates with a similar country’s party leader, then this increases the likelihood that economic data is going to be interpreted as worse than it quantitatively is.
Can you be more quantitative about whether these results are surprising in your models? E.g., from your election model and given the polling data you had, what was the probability that Trump would win by this at least this much? or win all swing states?
It’s an interesting question. I can’t see details about the prediction from The Economist but it seems that Trump winning all the competitive states (the likely 312/226 final outcome) was the most probable outcome for 538 and for Nate Silve (with probabilities in the 5-10% range, I think).
Yes, and it was the single most probable outcome (at a 6.3% chance) in Silver’s model by a pretty wide margin.
Chart here:
https://x.com/pescami/status/1854293570540241132
It’s fascinating to read the comments there, to see how different people process that information to assess whether Silver’s “tossup” forecast was a failure or not.
If you set aside partisan attacks, I think the biggest issue is that most people just don’t understand probability or statistical models. I don’t mean it to sound condescending: I don’t have a great grasp either of what these models are doing, but I know there’s much I don’t really understand about them so I try not to go on a near-jerk reaction.
I think Silver’s last post was saying something like, “Out of 80,000 simulations, Harris wins 40,012 times and loses 39,988 times”. Lakshya Jain tweeted the night before the election the following:
“
The pre-election data is somewhat *unclear*. The median outcome is a close election. But the *modal* outcome is not necessarily that.
There’s a ~50% chance that Harris or Trump just sweep all 7 swing states. Polling error is correlated.
”
This blog shows that working scientists have problems using statistics in their own area. How much hope is there that an average person who just pays attention every four years would be able to even understand what Silver and Jain are actually saying, let alone the success of a model against a specific outcome on election night? They’re going to scoff when they hear 6.3%
(I don’t know where Jain got his number about a sweep, by the way)
Yes, also on the Economist page, 312-226 is one of the most likely outcomes—basically the most probable, conditional on there being a polling error in that direction.
Also, someone asked me how it was that our forecast moved from 50% to 56% Harris win probability on the final day. I’m not sure–I just helped them set up the forecast model, and when we were doing that, we weren’t thinking so much about what would be happening just before the election. My guess is that this has to do with the different error terms in our model. We have an error term for day-to-day changes in public opinion and an error term for systematic polling error. On election day that first error term becomes zero, so fluctuations in late polls are particularly informative. Given what we think about possible last-minute opinion changes, this seems reasonable, although in this case of course it didn’t help.
Andrew:
This makes a lot of sense, except that you say it “seems reasonable”. I know next to nothing about American politics or political forecasting, but this last-minute-sensitivity seems like a bug to me, not a feature. You mention “last minute opinion changes” but even if that is a reasonable explanation, it seems to have swung the forecast in the wrong direction. Do you (or anyone else) mind expanding a bit on why you think it’s reasonable?
9p:
I’m not sure. To understand what’s going on here, I’d want to play around with the model a bit, change the inputs and see how sensitive the inferences are to those final polls. Just speaking generally, I think it makes sense to allow for a time series of opinion changes, which will imply that later polls will be more informative. But it’s hard for me to say much about the details of this case.
Also potentially relevant: 82 million votes were early votes. That’s somewhere around more than half the total votes cast for president. Final-week polls ought to be devalued because they reflect fewer of the still-potential voters.
John:
Good point. From a modeling standpoint, the right way to do this is not to literally underweight later polls but rather to report inference on an average of public opinion during the voting period, rather than just reporting inference for opinion on election day.
That said, none of this would’ve fixed the large and systematic nonsampling error, which our model allowed for probabilistically (which is why the range of possibilities included the actual election outcome) but not directionally (which is why the forecast was off by something like 2 percentage points of the national vote).
As the natural scientist in the room (except I think Daniel used to be one), what strikes me is that all of you are talking about politics as if we were dealing with a stationary system. We aren’t, because of it the links between political and natural systems. This election is going to put us a long way farther behind in efforts to deal with climate change, which is going to force mass migrations that are going promote xenophobia and so boost autocrats.
From this point of view, the election of 2000 looks a lot more important. Gore would have pushed us towards dealing with climate change when it would have been a lot easier to do.
John:
Agreed, and pandemics are another such example of natural systems. Also these issues of nonstationarity and irreversibility arise in social processes too, even if we were in a world with no clear physical constraints. For example, think of the chains of consequences arising from decisions leading to the two Iraq wars, the World Trade Center attacks, the Ukraine war, etc.
>> This election is going to put us a long way farther behind in efforts to deal with climate change,
I’m actually not as convinced as most that this is true! There will certainly be less federal government effort to address climate change. But that is not necessarily the same thing if one considers the possibility that those efforts in the last few years have been ineffective*, or even counterproductive (by driving people to take the issue less seriously).
I do not think climate change can be meaningfully helped by top down policy in the US – there is too much of a time lag between emissions and visible climate effects. Within a 4 year term, voters will see the costs but not yet the benefits.
(I do actually think climate change is a real and serious problem! But I think the only possible solution is technological & manufacturing & infrastrucure advance to make low/zero carbon power – especially solar + battery storage – win *in the market*. If imposed, voters will rebel and nothing will be accomplished. Already, many rural Americans in areas depending on cattle ranching and oil/gas see climate policies as an existential threat not just to their economy but to their way of life (the cow methane issue is really, really politically poisonous in middle America).
*EG all the hydrogen stuff in the Inflation Reduction Act. Hydrogen is not a power source — there are no hydrogen mines — and as a power transport mechanism it makes no sense in a world with widespread lithium-ion batteries (and sodium batteries coming along). The “hydrogen economy” made sense in the 90s, but it does not make sense in the 2020s.
There have been substantial shifts to the right across many countries in Europe recently and Latin America, correct? Most notably France this summer? If the US elections are put in that light, then the result is very expected based on the trend around the world. While ‘right’ and ‘left’ in Europe and ‘right’ and ‘left’ in the US are likely very different, perhaps the ‘left’ is out of touch with a large proportion of the voting populace across the globe? I wonder why they don’t move more center.
The point about what left and right means is a good one to raise – Harris spent the critical days of her campaign appealing to the hypothetical (and in my opinion mythical) moderate Republican and explicitly eschewing or rejecting progressive policy positions (e.g., tough on border, tough on crime, allowing no criticism of Israel). In return, she made basically no inroads among Republican voters and convincingly lost the election.
In what ways could she have moved more “center”? Why would someone who prefers Republican policy positions and cultural signifiers vote for an imitation version of them instead of the real thing? In my opinion the American version of left/right is hopelessly confused (or purposefully obfuscated to exclude discussing any structural economic components that historically define left/right parties).
Anon:
Research has found a small electoral benefit to moderation. It made political sense for Harris, as someone with one of the furthest-left voting records in Congress, to establish her credentials as a moderate, and Republican support is one way to get there. That said, it also seems that candidate effects are pretty minor. It could well be that had Haris emphasized more left-wing policy positions, she would’ve done slightly worse.
Jd:
It’s complicated. One issue is that centrist positions are not always popular on their own terms. For example, one way for the Democrats to move toward the center on policy would be to abandon their defense of their health-care plan. But it’s not clear that voters would like that. I think that one reason the Democrats moved to the right on immigration specifically is that this is an issue where their previous position was to the left of the median voter.
Seriously, do you think that the official stance on immigration is at all relevant? Trump = immigration is a huge problem and requires drastic action. Once that belief is entrenched, I don’t think anybody cares or believes what Harris says about immigration. I would say that her moving her position had not effect at all on the outcome. The only thing that matters about immigration is how strongly voters believed the initial claims – claims that were largely unsubstantiated and mostly racist (there is probably a better term to use, but it doesn’t occur to me right now).
I find it hard to buy that the positions have much impact at all. It is all about beliefs and facts (mostly not in evidence). In fact, I think that shifting positions is taken by many people to indicate that the candidate cannot be trusted. One thing Trump does very well is to never ever show weakness or indecision (much like today’s post about researchers never finding that mistakes might change their results). You can even change your position, but it must be done with certainty and no excuses.
I seem to think that in my youth (which I can no longer remember accurately), that the party platforms actually meant something. I don’t see that any more (if it ever was the case).
Dale:
Yes, I think that candidates’ positions on issues, as viewed by the voters, do make a small electoral difference. Trump, like just about every other candidate, has changed his positions on many issues, sometimes sends contradictory messages to different audiences, and sometimes tries to keep his positions blurry so as not to alienate voters. Harris has done this too. It’s what politicians do.
Dale –
I seem to think that in my youth (which I can no longer remember accurately), that the party platforms actually meant something. I don’t see that any more (if it ever was the case).
It’s intersting to consider that recollection along with considering that there used to be more diversity within parties (or at least their representatives, if not necessarily their constituencies).
I don’t think anybody cares or believes what Harris says about immigration.
A critical aspect of that, imo, is that you just can’t assume any more that people have remotely congruent sets of information to assess issues or platforms..
The “lab leak” would be a good example. Or COVID vaccines. For many people, it’s an article of faith that a lab leak was a certainty, and/or that the vaxes killed millions – because they see no informational sources that say otherwise.
It’s not merely that people aren’t always really voting based on issues or the candidates’ actual platforms per se (and are more voting as an expression of identity and antipathy towards the “other”), but also to the extent that they are voting on issues they have almost entirely different views on what comprises the issues.
Anon – I wouldn’t suspect that there is anything that Harris could possibly do to magically transform her position from one of the most left leaning to center. Nobody would buy that. I would suspect her position was fixed beforehand in the eyes of most voters no matter what she said. That wasn’t what I was asking. I was asking why the left would run such a candidate in the first place. In the light of recent trends across the world, why not run people on the right side of the left party, rather than the left end of the spectrum? In any case, it seems that the Democratic party in the US has positioned itself very left, in a right leaning world. Not being an expert on political science myself, that seems rather absurd (at least if the assumption of a party is that its main goal is to win elections, which I think appears to be the case).
Jd:
1. I disagree that Harris’s position “was fixed beforehand in the eyes of most voters no matter what she said.” To get the endorsement of prominent Republicans is a costly signal of centrism. Most voters are going to vote for who they’ll vote for based on party alone, so candidate positioning won’t have much effect, but I don’t think it makes sense to think of the candidates on either party as having completely fixed ideological positioning.
2. In 2020, the Democratic party chose Biden as the nominee, and Biden is not on the left end of the spectrum. In 2024, Biden won in the primary elections, but then after he chose to step down, the Democrats chose to nominate the sitting vice-president.
3. One goal of a party is to win elections; another is to advance policy goals. The Republicans won in 2024 but have lost in several other recent elections. I expect that had they pursued more centrist policies, the Republicans would’ve done better at the polls, but that got in the way of their policy goals. The same thing goes for the Democrats. Indeed, to the extent that the swing from 2020 to 2024 was part of a general anti-incumbent movement, one can make the reasonable argument that the parties’ ideological positioning was not going to effect the outcome anyway, in which case there is a logic to parties taking more extreme positions. Many people have said this: it’s a connection between political polarization and the logic of party positioning.
Andrew – I wasn’t under the impression that those Republican endorsements represented a center shift as much as they did a condemnation of Trump. So I don’t agree that it was any signal to centrism whatsoever. It more corresponded to Harris’ campaign point that Trump was “unhinged”.
A better dividing line is pro-patriot act vs anti-patriot act.
This was activated after 9/11 and has been going since and is too top-secret for members of congress to look into:
https://en.wikipedia.org/wiki/United_States_federal_government_continuity_of_operations
Essentially it legalized the same slush fund games as used during Iran-Contra.
Just to clarify: this isn’t just the first time since WW2 that all incumbent parties in developed countries lost vote share.
It’s the first time since this data was first recorded in 1905. Essentially the first time in the history of democracy (universal suffrage began in 1894).
https://x.com/jburnmurdoch/status/1854559598574784631
Because the cost of shedding ideological committments is higher than the cost of losing elections.
Parties, which are institutions designed to have winning elections as their main criterion, have been severely weakened around the world. (There are a lot of reasons for this; see, e.g.. Martin Gurri.) Losing parties normally add to their coalitions, diluting ideological conformity to increase their chance of winning. But with weak party structures and gatekeeping, the indeological cost of that dilution is too high.
It may still be too early to say for certain, but it seems that Washington State could be the only state where Kamala Harris is receiving a higher margin of support than Joe Biden did in 2020 (currently, the margin stands at 1%). I suspect several factors are contributing to this, which likely explain the shift in voting patterns. I’m hoping to explore these reasons further, and I’ve compiled an initial list of possible factors:
Ideological Migration – It’s possible that a higher proportion of Democrats are moving to Washington compared to independents or Republicans.
Higher Percentage of College Graduates – Washington has a relatively high percentage of college graduates, which could influence voting behavior.
Statistical Noise – Some of the margin could simply be due to randomness or minor fluctuations in the voting population.
Another theory, though harder to quantify, is that recent economic policies passed by Democrats—such as the Working Families Tax Credit, childcare subsidies, and long-term care programs—might have boosted their standing with voters. Some of these policies were directly tied to funding measures on the ballot, and I wonder if that affected public perception of Republicans in the state.
I haven’t fully worked out how to analyze the impact of these policies, but I’m open to any suggestions on how to approach this.
Let’s not forget that in 2020 Trump was the incumbent and in 2024 Harris/Biden is the incumbent. One possibility is that voters want to out whoever is in – all their gripes are channeled into the “new” candidate to replace the incumbent. In that sense, we would expect Harris to underperform relative to Biden. Of course there are myriad other differences, but one possibility is simply that it is difficult to retain the presidency now (Obama did it, but that was ages ago). It is hard to reconcile this with the incredible power of incumbency in Congressional races, so I’m not sure I believe this – but it is possible that there has not really been a “shift in preferences.”
Dale:
Regarding your last sentence: I haven’t looked that the data recently, but it’s my impression that the incumbency advantage in congressional races is much less than it used to be. It’s still positive—maybe 2% on average?—but not the 10% that it was in the 1980s.
I think this is likely. People are generally pessimistic about the future, though for very different reasons left vs right.
As for Congressional incumbency – isn’t there polling showing people generally approve of their own Rep but disapprove of Congress in general? I think this might be much like how many anti-government Americans are pro-police: the more “government” is identified with my community/state, the more it is liked; the more it is identified with an amorphous Washington DC power center seen as corrupt, the more it’s hated.
The country may reject the incumbent again and elect a Democrat in 2028 – though that may well depend on what the Trump Administration actually does. (There remains a possibility IMO that they do little that matters, and as post COVID inflation fades in the rearview mirror the country’s mood improves – especially if the Ukraine and/or Gaza conflicts are resolved by 2028.)
“Washington State could be the only state where Kamala Harris is receiving a higher margin of support than Joe Biden did in 2020”
Washington State has been trending Democratic for a couple decades. It seems weird now, but the redistricting after the 2010 census was coordinated by Democrats with a goal of creating some purple districts from what had been blue districts. Can you imagine? It didn’t end up working out, the two districts that were supposed to be purple quickly trended towards Democrats and the Republicans made no progress on the west side of the state.
The statewide “default” vote in Washington is 57-43 blue when neither candidate is particularly well known. Dave Reichert was the strongest Republican candidate for governor in decades but Bob Ferguson is our beloved former state AG. It is currently 56-44 for Ferguson, Reichert should be happy with his extra 1%.
Washington State is quite similar to a semi-socialist European country, and working folks benefit the same way those Europeans do, less inequality, longer and more healthy lives, strong social safety net, etc.
Great. But what we want from you right now is an expert consideration of the statistical anomalies behind the House results vs presidential vote in the swing states.
I will go even further. I don’t think most people understand the difference between ‘inflation’ and ‘price level’. Also monetary policy of raising interest rates may slow the economy to reduce overall inflation but for lower-income people with high flexible debt, it is inflation.
Not just inflation, it’s a mechanical government-policy-created transfer from poor people to rich people. It’s really no different than taxing the poor and giving payouts to the rich in effect.
Daniel –
Although it’s on Twitter – I thought you might find this worth watching.
https://x.com/smalls2672/status/1855327797041869212
That’s one of the saner takes I’ve seen.
This is 100% my take, and here is an economist who finally gets it! Hallelujah. Notice how she talks about the real problem being 50 ish years old, and that the issues have nothing to do with prices of computers and TVs and music and movies and clothing and such and everything to do with things like Houses, Childcare, Utilities, Transportation, Healthcare, College Tuition and soforth.
Somewhere there are posts on this blog where I explicitly graphed an “inflation index” based on a mixture of these components and showed that wages declined continuously for at least 20-30 years on that basis.
Then once an additional necessity became a massive target of inflation (food) it really triggered a lot of anxiety. It’s one thing to not be able to afford a house, a thing you normally plan for over decades, but not being able to afford to feed your children, a necessity that can kill you in a matter of weeks to months, that ramps up the economic stress to high levels. One thing economists typically don’t get is the importance of timescales. They explicitly teach everything as an equilibrium, which means time is point-like. But in reality Timescales for not eating to kill you is about 2 months. Timescales for living on the street unaided to kill you in Minneapolis is about 6 months(late spring to late fall, freezing to death), in southern California it’s probably 40 years.
So, it’s good to see an economist who gets it. Thanks for that link.
Relevant to this issue is another article by Blair Fix on the meaninglessness of “Real GDP” https://economicsfromthetopdown.com/2019/04/29/real-gdp-the-flawed-metric-at-the-heart-of-macroeconomics/#fix_2019
He makes his argument fairly well. This is why I’ve argued that an important denominator metric of “Real Income” should be one that primarily weights consumption that is physically required (typically by thermodynamic restrictions)… housing, food+water, transportation and communications, healthcare, education would be a good start. There is a quantity of these goods which is necessary to be compatible with stable life (ie. people not dying or inevitably declining in productive capacity). Income as a fraction of the price of this basic quantity of necessary maintenance goods is the appropriate measure of economic health. By those criteria, income has declined over the last 50 years. This is exactly what the video you linked on X says.
I spent about 6 months trying to formulate such an index in about 2017. The task exceeded what I could reasonably do on a volunteer basis, after months of work I shelved the project. But the process taught me a LOT about the real economy. I actually was able to fit these models pretty reasonably **some of the time**, but Stan would get caught out in left field on some chains. These days I’d probably do things differently having learned more about how to handle model issues. Importantly my model also was geographically dependent, that is, it varied continuously with say latitude and longitude within the continental US. It is TRULY ridiculuous to claim that economists “Real GDP” is a single number which aggregates the experiences of *all* of the US. The proper framework is that for every location in the US there is a *distribution* of a dimensionless ratio (nominal income) / (cost of thermodynamically required goods), and *at best* you can aggregate this distribution across space to produce a distribution of “real income”.
While there is much I agree with in the X post, I am not in favor of claims made about “most Americans.” That isn’t much better than the single number for Real GDP. Any single number is unable to capture the variation in experiences that people face. Daniel’s repeated attempt to derive more meaningful measures that the standard inflation and income measures are certainly admirable – but fraught with issues. Yes, housing is a necessity, but how much housing? Every time I am in the market to shop for houses, I am struck by the lack of reasonable size houses available. Americans (most?) purchase more housing than is necessary, I’d say. We all need food, but what kind of food and how much? We need utilities, but does that include 5G phones with unlimited data plans, CATV with ESPN sports channels, etc.? Defining a meaningful measure for what we truly need and then finding that data seems like an impossible task.
I will mention the Consumer Expenditure Survey data that BLS collects. I looked at it this morning – I don’t have the time to work with it, but there is much granular data that would be useful to analyze. The public use files contain individual household data with highly disaggregated expenditure patterns. And the data is available over time, so a number of these important questions can be addressed. If there is a complaint I have, it is that the analyses I would most like to see don’t seem to have been done by BLS, and the data is complex enough to make it a major undertaking. It is possible to compare spending patterns across the income distribution and over time: I have found little in the way of this in their research studies. Perhaps there is a political dimension to why not.
So, I’m reluctant to subscribe to the idea that most Americans have experienced a reduction in their standard of living over several decades. I think that is not likely to be true, although there are certainly some groups for which it holds. I do believe that we have all experienced meaningful decreases in our standards of living regarding our quality of life – the things like quality of leisure time, environmental quality, lack of stress, etc. Those things are not measured at all.
Dale –
Interesting comment. Thanks
Certainly there are other measures like food insecurity which might help fill out the picture a bit. Or living below the poverty line. They’re not impervious to your critique but maybe they can help to account for how “standards” of living change over time in ways that makes an absolute measures difficult to get your arms around. Often I read (particularly from conservatives, although not as much lately unsurprisingly) that measuring poverty over time or across nationalities is misleading because standards change; poverty today vis a via living standards is different than it was in the past, or than it is in other, developing nations. That feels judgemental to say, given I don’t face poverty, but I do think it has to be accounted for in some fashion if we’re looking at trends in standards of living over time.
Especially when we being things like social security or access to healthcare and education into the mix. Maybe those haven’t changed much in the last 50 years – but then again 50 years ago may have been a kind of high water mark that wasn’t sustainable. At some point, is it unrealistic to expect a high water mark to continue as opposed to seeing a reversion to standards more consistent with the bulk of human history.
Then again, there is absolutely no real that I can see that we couldn’t sustain, or actually raise that high water mark. It’s not some inherent limitation in the system. It would merely take a relatively small degree of income redistribution to maintain that high water mark. (Where’s chipmunk? I need to look out for incoming objects after making that last statement.)
Dale, Joshua,
Dale is right that defining what constitutes the “necessary” amount of various goods is complicated. And he’s also very very right that it’s frustrating that BLS hasn’t really done much to tackle the question with the data they have, and that it’s a bit of a hurculean task (I know, I actually tackled it for about 6 months on a totally volunteer basis, I had to pull back because it dominated all my time and my family was suffering for it).
One nice thing about a Bayesian approach is we can easily include the uncertainty in the measure itself. So for example we can say that people need somewhere between 1000 and 3000 calories a day depending in age, size, and activity level, we can do a regression on age and sex and limit the uncertainty probably down to around 500 calories a day.
The USDA has various “food plans” which when I contacted them they refused to release any additional data on (like computer readable data tables) so everything is in PDFs on their website in a fairly useless form. But with some funding one could get undergrad research RAs or NGO employees to transfer that data to databases. The consumer expenditure surveys and the price data on individual items that goes into the CPI are available… with lots of holes in the data and weirdness but we could have analysts work on that problem.
Mechanical Engineers have data on “heating degree days” and “cooling degree days” that tell you something about how much load you could expect on an HVAC system to keep a residence at a safe temperature.
Yes, lots of houses are bigger than needed, but with information about the regional distribution of housing sizes we could do regressions that could partition a particular family in the ACS into a portion of rent that goes to the minimum square footage needed for a chosen fixed low standard of living, let’s say something like 250sqft per person, which is tighter than most families live in today, and discretionary consumption of additional space.
We can make an argument that with existing remote work technologies, we need about say 5Mbps per working adult upload and 10Mbps per working adult download as an internet connection in order to enable a family to be eligible for remote work opportunities. Doing a calculation you can see that you need about 3Mbps minimum in order to guarantee low latency on any one stream (this is a calculation associated with the MTU of 1500 bytes per packet and latency requirements for interactive audio)
Water per person is relatively easy to calculate a minimum for, you want say 2 gallons of drinking water, 1 shower, 4 toilet flushes, and a couple pots of spaghetti at a minimum per person. Add that up it’s probably 10 gallons a day ish.
Transportation… you need to get food to your house, you need to get to work and home, you need to get kids to school. It’s not unreasonable to estimate this minimum.
On each of these “minimal budgets” we can impose some bayesian quantification of uncertainty, like maybe the real quantity of water is lognormally distributed and we could be off by factors of 2 or 4. Maybe the transportation quantity is more well bounded, and a fairly tight Gamma distribution would be more appropriate… whatever.
The point is this… there’s a clear enough methodology we could be pursuing. It’s not like i’m the most brilliant person who ever lived or something, other people could have thought up “hey what are the kind of minimal standards of living we should consider as poverty level”.
Up until the mid 2010’s our definition of poverty in the country was something a woman from the USDA put on the back of a napkin in the 1960’s like, 3x the cost of the USDA basic food plan if I remember correctly. Not bad for a 1960’s back of the napkin… fucking horrific for the BLS a multi-billion-dollar-a-year government agency who has had 60 years since then and exponential growth of computing power so that my cell phone is faster than a 1980’s cray supercomputer.
The BLS + Census did some kind of “Supplemental poverty” measures, and I assume they work something like the above. They consistently show California as the highest percentage of people living in poverty. Stick that in your pipe and smoke it. It’s not because we don’t have a strong economy in terms of “GDP growth”, it’s because we have a highly differentially growing economy and inflation of basic cost of living goods like housing and food and childcare and such.
It appears that there will be around 5 million fewer votes in the 2024 Presidential election than there were in 2020 (I’ve included 3rd party candidates and the less than complete count at this time – primarily from California). Nobody seems to have mentioned this yet. Everything was pointing towards record turnout, but it doesn’t seem to have happened. Any ideas why? (aside from claiming that the millions of illegal votes were not cast this time around)
Dale:
I dunno, maybe in 2020 there was a pandemic and people had nothing else to do with their time, so they voted?
Today, the AP reports this about turnout: https://apnews.com/article/election-2024-voter-turnout-republicans-trump-harris-7ef18c115c8e1e76210820e0146bc3a5. The headlline bothers me – given all the hype and anticipation, I expected turnout to be higher than in 2020, not 6 million votes lower (despite it still being high by historical standards). Also, the group they provide showing state level comparisons I find incomprehensible. The coloring is supposed to show two comparisons: one with 2020 and the other with 2016, but each state only has 1 color. I’ve had my coffee today, so I can’t blame that if I am missing something obvious about the graph. But I looked at it for a couple of minutes and don’t understand it.
OK, a fresh look at the graph and I see the user error. You have to click on the 2020 or 2016 comparison button to toggle between them. My bad. Still, there could have been better instructions I think.
A contribution to the discussion of real vs perceived economic conditions: https://jzmazlish.substack.com/p/yes-inflation-made-the-median-voter?r=naug&triedRedirect=true (found this via Marginal Revolutions). It is perhaps the most illuminating analysis I’ve seen – and I might even change some of my thoughts. But one thing stood out to me. The column states, near the beginning,
“1. In no part of the income distribution did wages grow faster while Biden was President than they did 2012-2020.
a. This is true in the raw data, and even more stark after compositional adjustment.
b. In particular, the change in median incomes was well below its 2012-20 run-rate.”
Right from the start this muddies the water in my mind. The primary question is what happened to real income, did it go up or down. But this quote compares the change in real income with the prior 8 years – a different question. I just found that off-putting from the start. But the column does conduct a number of interesting a relevant comparisons. As I said, I might rethink some of my positions. I just get skeptical when analyses start with such indirect measures of a more direct question. Point 1 can be paraphrased as “what have you done for me lately?” Perhaps that is the bottom line for many voters.
Dale:
Yeah, using 2012 as a baseline doesn’t seem right, for two reasons:
1. If you’re comparing performance in Biden’s term vs. Trump’s first term, it would seem more natural to compare the 2020-2024 period to 2016-2020.
2. In 2012 the country was emerging from a recession, so, yeah, growth will look good compared to that baseline.
It’s possible that the author of that post picked the 2012 baseline as a way to find an effect in the desired direction, so as to come up with an explanation of why people view the economy negatively. This explanation might make sense—maybe people really to compare current trends to trends following an economic recovery—but it’s not obvious to me that this should be the case. In repeatedly comparing to the 2012-2020 period, the author is just baking this choice into the analysis.
Also, from a political standpoint there seems something iffy with talking about “that median growth in incomes was much weaker than it was during the prior two Presidencies is a major part of voter’s discontent,” without noting that “the prior two Presidencies” included Obama’s second term.
Dale –
But, the change in median wages is not what matters; it is the median change in wages that does.
Except what happened, supposedly, is that lower wage workers moved right while higher wage workers moved left. So I don’t think median “change is wages” is the key focus if the point is to understand voter behavior. If the point is merely look at the economics independently, well then fine.
Supposedly, lower wage workers saw the most income growth relative to inflation. Yet they also were, supposedly, were the voters who were most angry about inflation.
Ezra Klein has the theory that’s because voters blame the government for inflation and believe themselves to be most responsible for increases.
Regardless, I don’t think it makes sense to evaluate perceptions of economic performance merely on economic performance. I know that sounds condescending, but the information environment has to be considered, imo.
Instead of arguing over what exactly to compare the median income to, try this:
“The poverty rate ticked up, as did the jobless rate. The number of Americans spending more than 30 percent of their income on rent climbed. The delinquency rate on credit cards surged, as did the share of families struggling to afford enough nutritious food, as did the rate of homelessness.”
“Interest rates were a problem too. The mortgage rate more than doubled during the Biden-Harris years, making credit-card balances, car payments, and homes unaffordable….”
Links to sources for each of these claims are in this article:
https://www.theatlantic.com/ideas/archive/2024/11/biden-harris-economy-election-loss/680592/
It very much is objective economic distress.
I only checked the first stated “fact” that “The poverty rate ticked up, as did the jobless rate.” According to Census data, the poverty rate from 2016 through 2023 was: 12.7%, 12.3%, 11.8%, 10.5%, 11.1%, 11.5%, 11.6%, 11.5%. The seasonally adjusted unemployment rate was 4.8% in January 2016, 14.8% in April 2020, and 4.1% in October 2024. Of course, all of these measures have numerous alternatives, but I don’t buy your statement of “objective economic distress.”
Dale, I didn’t check your sources. but here’s the supplemental poverty measure for 2023, I believe this is the main source we should use, as the supplemental measure is vastly more sophisticated than the earlier “official” measure:
https://www2.census.gov/library/publications/2024/demo/p60-283.pdf
Page 8, see how childhood poverty SURGED since 2021. I about doubled for black and hispanic families, from around 10% to around 20%, it about tripled for asians, from 5% to 14%
On page 9 supplemental poverty measure goes from about 8% to about 13% overall.
When you double or triple the number of children in poverty you’re bound to see a major backlash, it affects both the families themselves, as well as their friends and relatives perception of the world. “My sister’s kids go hungry every night” is not “yeah the economy is doing great”. Ten or twenty percent of families in poverty means everyone knows someone whose kids are suffering and a lot more of them than in 2021 when Biden took office.
Daniel
As I said, there are many measures that could be chosen – I am not saying I favor any one of them – I was objecting to the characterization as being “objective.” Still, the most interesting thing to me is that the groups you cite are not the groups that voted for Trump, as far as I can tell. Yes, there has been real suffering over the past 4 years, but I still think that the groups that suffered the most did not vote for Trump.
but I still think that the groups that suffered the most did not vote for Trump.
I must admit I got sloppy above as my favorite is when people say “working class voters surged to Trump.”
Some 90% of black voters voted for Trump, and a large percentage of them are working class (certainly they are disproportionately represented among “the working class.”) Or consider what it’s like for single-parent (woman-headed) households, who are probably disproportionately black and who presumably are under high levels of economic stress.
That certainly illustrates why there’s a lot of guesswork in going from “objective measures” through “perceptions of the economy” to explaining voting behavior.
Methinks there’s lots of room for confirmation bias.
Sorry – voted for Harris, obviously.
Dale, you’re absolutely right about the disconnect in voting. I have much stronger opinions about the measurement of the economy than I do about the motivations of trump voters.
If we start measuring the economy better though, we could have better policies that lead to better outcomes for most people. Until that happens we will continue to have measurements mainly designed to see how well the top 1000 billionaires are doing, or how some nebulous median is doing that doesn’t correspond to a person but rather a particular non-static slice of people.
Also: Delinquency rates on credit cards: 2.66% (1Q 2020), 1.54% (3Q 2021), 3.25% (2Q 2024). Delinquency rates on single family mortgages: 2.35% (1Q 2020), 1.73% (2Q 2024). Many more such measures and it is easy to cherry pick particular times and rates to show quite different patterns. Far from what I’d call “objective economic distress.” I’m not denying that their was distress nor am I stating that things got better or worse – all I am pointing out is that the data is far from as clear as you claim it to be.
Dale, your first example here proves my point (covid stimulus decreased credit card debt initially and then the stimulus went away). Maybe I should say “objective economic distress for poor people”. But all the sources linked in the article I linked above are FRED data, they’re not random numbers but full graphs. They pretty much all show an uptick in distress after the stimulus/pandemic programs disappeared (and yes, use supplemental and not regular poverty).
As Daniel repeatedly says, increasing inequality means topline GDP/unemployment numbers are less and less useful. Have you heard the phrase “K-shaped recovery?” It means the economy became great for rich people and dropped off a cliff for poor people. This is, I suspect, the explanation for the mortgage delinquency rate being lower: homeowners are richer and can pay off the increased mortgage rates easier.
As for the black Democrat vote share that someone mentioned upthread, it’s been slowly decreasing since at least 2012. We’ll have to wait till more granular precinct-level data is available, since exit polls aren’t reliable for subpopulations, but I would be surprised if that decrease has stopped. Similar trends hold for other minority groups.
anon
2/3 of Americans own homes, so let’s not equate home ownership with only the wealthy. I’m not going to debate whether the poor have suffered – and I certainly believe inequality is a major problem. But the upticks you refer to were upticks after all that Covid relief ended. Of course, many people lost out when those payments went away. But some of the graphs show steady declines prior to that uptick. So, the picture depends on where you choose as time points to compare (as well as who you try to compare and what you try to compare). Also, the decreasing support for Democrats in certain groups is a complex matter – there is a distinct gender difference there, and some of the hardest economic conditions have faced groups that voted for Harris heavily (at least I think that is the case). I have no doubt that some groups have not fared well in the past 4 years and have lost ground relative to where they were in the prior administration. But which groups and how many is an open question, in my mind.
And as pointed out indirectly at one point in the article Dale linked, if you define the low wage earners as something like the 20%tile of wages earned… Then you fire everyone who makes less than a million dollars a year, then the 20%tile of wages will go WAY WAY UP.. Yay! economic progress!
Except for 99.9% of the population they now have zero income.
The point being, unless you track the changes that are experienced individually, you’re often just missing the point. This is the point of “change in median wages” vs “median change in wages” in the article.
Good examples though of the kinds of economic distress people are experiencing. I continue to be completely baffled that Economists can be so stupid. They seem to have settled on “these three statistics matter” and turned their brains off in the 1970’s or so.
There was a group of polls that seemed to push Harris pretty hard and then tack back to be more reasonable closer to the election.
https://www.realclearpolling.com/latest-polls/national-presidential-election (look at the time series of Morning Consult, national polls from 01 OCT to just before the election):
+5, +5, +4, +4, +3, +2.
If that is a real trend, then there should be some analysis of what was going on. Why did the electorate move to Trump versus Harris as the election got closer? But there’s zero discussion of this trend. And if anything the news cycle and media coverage (e.g. PR thing) were more negative to Trump late. But the MC poll time series went the opposite direction.
I think (Bayesian guess) that the issue was not really the electorate moving away from Harris as the vote neared. But that her previous polls were a mirage. Lot of little levers that MC could use to pull the polls back to look more reasonable as the election neared (sample weighting, etc.) And my Bayesian guess is their method (to include sampling, weighting etc.) changed as the election neared. Didn’t want to look like that IA pollster after all! But still put out something that made East Coast media consumers (NYT readers slant left) a little more happy during most of the media cycle time.
I’d love to see the internals from Harris campaign. There were indications that they were not that confident. Also, we haven’t heard anything from them about the whole electorate moving away from her as the race closed.
This tactic of making the last poll point more reasonable is sort of last minute herding. And is very useful with the vast amount of people like you who only look at the closing poll. Not the time series.
Anyhow…sure I’m speculating. But if we live in the world of multiple hypotheses, I can think of two.
1. The electorate really moved right as the election closed. Or…
2. MC was fiddling with the levers over the time of the race and juicing Harris before, but tacking to reasonableness at the end.