Election 2021 in Context: Looking at Electorates
The electoral calendar affects who the electorate is.
Election 2021 Context, Part 1: New Jersey
Election 2021 Context, Part 2: Virginia
So, how could it be that a state that voted one way for president could turn around and vote another way for state offices? It must be because of messaging, or candidate quality, right?
While I would not entirely discount such issues, the reality is that the main explanation is less who was running and what they were saying and more who was voting.
Let’s look at Virgina:
| Virgina | Popular Vote for President | Popular Vote for Governor | Difference |
| 2020/2021* | 4,460,524 | 3,254,649 | 1,205,875 |
| 2016/2017 | 3,982,752 | 2,608,608 | 1,374,144 |
| 2012/2013 | 3,854,489 | 2,240,314 | 1,614,175 |
| 2008/2009 | 3,723,260 | 1,985,103 | 1,738,157 |
| 2004/2005 | 3,198,367 | 1,983,778 | 1,214,589 |
| 2000/2001 | 2,739,447 | 1,886,721 | 852,726 |
Now, it is certainly true that over time populations grow and one has to be cautious about making comparisons of absolute totals election to election because the pool of voters grows over time (especially over a four-year period). But we are speaking here of only a one-year difference. The pattern is pretty clear: the electorate that shows up to vote for governor is substantially different from the electorate that shows ups to elect the president. The average difference is ~1.3 million fewer voters in the gubernatorial elections than the presidential ones.
Look, any two electorates (even in a two-round election in the same cycle) are not identical, but when you start taking over a million voters out of a given pool, you are very much dealing with different electorates, so trying to judge Election Day 2020 and Election Day 2021 as if all that has changed is the date plus some pet issue is a significant error (although one made in the American media all the time). Or, to put it another way: trying to make sweeping, grandiose claims about national politics in these circumstances is not a great idea.
Here’s New Jersey for more context (but only the last three cycles):
| New Jersey | Popular Vote for President | Popular Vote for Governor | Difference |
| 2020/2021* | 4,549,353 | 2,505,080 | 2,044,273 |
| 2016/2017 | 3,874,046 | 2,147,415 | 1,726,631 |
| 2012/2013 | 3,640,292 | 2,120,866 | 1,519,426 |
(The data for governor’s races was not as accessible for NJ as for VA, so I got lazy as I think that the point is pretty well made without further data).
I would note that NJ is more reliably Democratic, at higher levels than Virginia, and hence the off-year election does not have as strong a counter-WH outcome (that is, the skew in NJ is more pro-D).
Regardless, all of this shows pretty clearly that the electoral calendar affects who the electorate is. This is a fundamental issue going on here, and it makes it very difficult to make definitive statements about how messaging, or even candidate quality, affects outcomes when comparing two fairly close elections even in the same state.
It is entirely possible (indeed, quite likely), for example, that ceteris paribus the electorate that turned out in November of 2020 would have elected McAuliffe, and that the electorate that turned out in November of 2021 would have voted for Trump.
As such, when elections for specific offices are held matters. A major reason (if not the main reason) that Virginia has the pattern that I detailed in Part 2 of this series is when it elects its governor and other state offices. The same is true for patterns we see in mid-terms for Congress (especially the House, since Senate patterns are also influenced by the fact only 1/3rd of the chamber is up in a given two-year period).
And, sure, things happen in a year that affect and influence how voters vote (as well as who shows up), but the reality is that presidential elections are the ones that draw out the most voters (and even then, comparatively, the US has low voters turnout) and therefore those are the most representative elections that we have.
One of the ironies of American democracy, that I have noted before, is that we have too many elections. The practical upshot of too many elections (for, quite frankly, too many offices) is that citizens are not adequately engaged and therefore do not participate. I mean, it seems more democratic to have more elections, rather than less, but if that means fewer voters, it is really more democratic? And while it sounds more democratic to elect, say, the county coroner, is it really democratically desirable or necessary?
One can argue, as I suspect someone reading this has already thought, that citizens ought to be more active and ought to pay more attention. All well and good, but you have to design institutions for the human race you have, not the one you might like to have.
All of this speaks to questions of when we should schedule elections and for which offices we should be voting. Citizens are more prone to show up to elect the national government, not local government. We can talk all day about the importance of local government, but the reality is that people pay more attention to national leadership.
(My ital.)
Wisdom.
I’ve long believed that a big part of the success of our Constitution was that it was written by men who had a pretty fair (jaundiced) view of their fellow humans. It was nice blend of idealism (the People!) and cynicism (power corrupts, trust no one.)
One system was predicated on the notion that you couldn’t really trust anyone because: human nature. Another system was predicated on an evolved human who would naturally act in ideologically appropriate ways.
Thank you. I had just read Sleeping Dogs comment in today’s Forum. Sleeping linked to an NYT column arguing it wasn’t turnout. As I had argued yesterday that VA really was all about turnout, this surprised me. So I started reading the NYT piece and realized he was comparing governor race turnout not to last years prez election, but to the previous governor race. I was starting to look for VA turnout numbers when I saw your post. I will happily cease looking for turnout history.
I will repeat my fearless prediction from yesterday, that GOPs are about to discover they’ve always known real election security requires holding state and local elections in odd numbered years.
@Michael Reynolds: I see conservatives fairly regularly comment that liberals believe man is perfectible. This surprises me, I have no idea where it comes from. I don’t recall any “liberal” say anything like that. I do recall seeing Hillary Clinton explicitly say, and I wish I’d saved it, that we can’t make people better, but that we can sometimes change circumstances and incentives.
@gVOR08: Well, though. On the topic of racial justice, we sort of do think that humans are perfectible, or maybe at least improvable.
There’s a policy portion of racial justice, but there’s also a personal behavior component. And the latter gets a lot more day-t0-day attention. You know, of the form, “he said what?!!”
In fact, many white people seem to think that we can eliminate racism. I don’t share that belief (in spite of being a white person), even though I think we can do a lot better than we are doing.
@gVOR08:
It’d be Marxists who seem to believe we can have new humans, and that is a definite strain of leftist thinking. Regular old liberals don’t so much believe man is perfectible, rather that man is educable. The first idea leads to starvation, secret police and concentration camps. The second idea leads more often to eye-rolling.
I wanted to chime in with Michael and say, yeah, this is so right, and wise. We need to believe in democracy AND make things easy for people to act in a way that reinforces democratic values, not hard.
And the hard numbers answer my question from a few days ago. There doesn’t need to be a single person switching their votes to accomplish this situation. It’s much more a matter of who stays home and who shows up.
And so I’m curious. How much were covid type measures still in place. Early voting? Mail-in voting?
I agree on the whole, Mr. Taylor, but those VA numbers also show the lowest difference in twenty years. Turnout was good. I do wish the country had a two day national holiday and coordinated elections more. Shoot, kids had Diwali off yesterday, but when it comes to getting time off to participate in your most important civic duty?? Coordinate that with your boss….
@gVOR08:
I think it’s because we want to stick them in re-education camps, but they want to stick us in death camps. The re-education camps are far more optimistic about human nature, because it assumes people can be re-educated.
Ok, seriously though, we believe that to some extent we can teach people to not be racist, homophobic, and all of that. Care for your neighbors, at least a little bit. Maybe we don’t think we can perfect people, but we can nudge them towards less worse through a combination of education and appeals to their better nature.
Conservatives… kind of don’t. They believe in a natural order where some people will get screwed over, and if they are the ones getting screwed over it’s because someone has messed with that natural order. Probably with CRT.
——
The liberal belief that you can make people better, mixed with our beliefs about what better is, likely makes us insufferable. Rural America is emptying out because of long term economic trends, destroying people’s home towns, and we would like those people left to be a little less racist, and maybe teach them a trade that will allow them to leave too.
We need to make Democrats who represent largely rural areas as much of the party leadership as the Chuck Schumer’s and Nancy Pelosis, to make sure rural voices get heard and have a seat at the table. Folks like Jon Tester, for instance.
I think it is unlikely the 2021 VA electorate would have voted for Trump. Exit polling showed Trump with a low approval number and Youngkin received many votes from Dems and Reps precisely because he wasn’t Trump as a way of making a point.
@gVOR08:
It is the case that a basic tenet of classical liberalism (and modernity in general) is that humans can improve their conditions via applied reason. Conservatives of the classic variety are more skeptical of such things.
But perfectability? Not so much.
@Jc:
Sure, but we are talking over a million fewer voters. These are not the same electorates. And it stands to reason that the party out of power is more upset and motivated a year after the presidential election. The pattern in VA for decades is pretty stark in that regard.
@Raoul: If the pool of voters in 2021 was more GOP-leaning, then the likely outcome would have been a close Trump win. But, of course, I am playing unprovable counterfactuals here.
Had Trump been on the ballot, or if a true trumpkin had been running, turnout might have been different for that matter.
To be clear: I am not arguing that this is all we need to know about the election this week, but rather to note that it explains a lot more than most of the hot takes that are out there at the moment.
@Steven L. Taylor: It’s unmentioned in the OP but a fairly conspicuous factoid is that extra 500K votes in the VA presidential election in ’20 seems to match a roughly 500K gain in votes one year later in the governors race.
Has the population swollen? Is voting that much easier? Is some demographic remarkably excited?
Or is it the same sort of statistical anomaly as skirt lengths and the Dow Average?
Sure, but that is true for every election. 17 million more people voted in 2020 compared to 2016. Additionally, in the space of those four years, many voters died, others became eligible to vote. Some voted for the first time. Some opted not to vote. Some switched sides. In short, the electorate in 2020 was substantially different than 2016, yet that doesn’t stop us from comparing them.
Every electorate is different and the reality is that we don’t have good data to measure the differences from one electorate to another except in very broad and vague strokes. Instead, we compare aggregates and assume consistency.
Secondly, everyone knows that turnout for elections in non-presidential years is going to be less. Political actors and candidates should be taking that into account when they craft their campaign strategy, because that strategy has an effect on what the actual electorate will be. And if a candidate knows going in that the cyclical winds are against them, then they ought adjust accordingly.
In short, candidate selection and campaign strategy, and messaging are still two of the most critical factors for winning elections.
@Andy: It would seem that you are either downplaying, or ignoring, the dynamic that different types and timing has on turnout.
By going, for example, to 4-year intervals (and of the same type of election) you are ignoring most of what is in the OP, yes?
@Steven L. Taylor:
I did specifically say ” everyone knows that turnout for elections in non-presidential years is going to be less” so the idea that I’m ignoring or downplaying election timing on turnout isn’t true and it wasn’t my argument.
My broader point is this: if you want to maximize the representational significance of elections, you should elect everything on the same day.
To move state elections a year off from the presidential election is to create a new electorate as part of institutional choice, not because X amount of time has allowed for population changes.
@Andy: It comes across as either dismissive of the point or perhaps not seeing the significance because all of this is just what has always been done.
Another broad point is: we shouldn’t pretend (as the press and common citizens do all the time) that the most important shift from 2020 to 2021 was messaging when the most important shift from 2020 to 2021 was how the electoral calendar shapes who votes.
This is also true about any mid-term and why we see the patterns we see. I am pointing out that we shouldn’t pretend that the main change is message when it isn’t actually the main change.
Again, I am not saying that message and candidate quality don’t matter, but I am saying that if you schedule the game when Tech can’t or won’t send as many fans, the makeup of the crowd will favor State and what happens on the field is unlikely to change anyone’s rooting interests.
@Steven L. Taylor: I think you and Andy are talking past each other. You’re interested in the shift between 2020 and 2021, and he is interested in the shift between 2017 and 2021.
There might be an interesting post comparing the shift between 2016-2017 and the shift between 2020-2021.
(Also, here, your charts might be easier to understand with a percentage drop-off rather than a vote difference in the third column — it would make it much easier to see a pattern)
@Steven: Virginia’s population has indeed grown (by ~500,000) since 2010 but not enough to account for the differences shown here.
@Jc: Election Day is technically a state holiday in Virginia now, as of last year. Plus, the same Democratic governor and legislature that passed that change also added in no-excuse absentee voting for 45 days. So, turnout should indeed have been pretty high.
@Gustopher:
They could well be.
The pattern that I am trying to show is the drop-off from pres to gov elections in raw votes each cycle. The point, that perhaps is not as clear as I think it is, is that by shifting the gov election one year after the pres election you guarantee a significantly different electorate.
@Steven L. Taylor:
turnout in off cycle elections is not the main change either. The messaging in VA did play just as significant a role as an off cycle election. The effect of instant media and it’s role on messaging outside of actual campaign messaging is quite large nowadays. When looking at what’s different this time around as we always do, not a ton of core things have changed so we fall back on the core things without thinking about how much the playing field has changed.
@Jc:
This is possible, but my question to you is how do you know this to be the case?
It always feels like there is some monocausal reason why an election turned out as it did due to whatever the dominant media narrative is. And, in close elections, there are likely any number of factors to explain wins and loses.
All I am asking for people to recognize is that the sequencing of the election, and the clear pattern is has wrought for decades, is pretty important.
And as much as I would like it to be the case that elections were won or lost solely on who had the best message and the best candidate, this just isn’t the case. (And that is a general observation and not just about this race).
@Gustopher:
Not exactly. My point is that every election has a different electorate and I just used 2016 vs 2021 as an example.
But I see I didn’t explain myself fully enough to be clear, so let me be more detailed:
For example this opening:
My point is that “who is voting” is different in every election and who shows up to vote is influenced by a variety of factors. That is true whether elections are spaced weeks or years apart. So logically, I see no reason why one can’t also say that outcomes in Presidential election years is also “less who was running and what they were saying and more who was voting.”
Related to that is this point I made:
So, what is the difference between 2020 and 2021 in Virginia besides the numerical difference of 1.2 million fewer voters? How, exactly, does the mere fact of fewer voters prove that issues and candidates are of minor importance?
The problem is that we don’t know have the data to compare the 2020 electorate and 2021 electorate with any kind of precision. Is the lower turnout dispositive? Maybe, but I think to make that case you’ve got to show the work.
So, I’m not denying that off-year elections with their decreased turnout don’t change election dynamics – quite the opposite! But unless you are able to compare the two electorates, then you can’t really predict the effects, much less conclude how relevant candidate selection, campaigns, and issues were comparatively.
@Steven L. Taylor:
Yes, you would definitely get a different electorate. But as I noted, every four years you get a substantially different electorate too. What is the practical difference in terms of the relative importance of candidates and issues?
Finally, I think I’ve said here many times over the years that I’m not a fan of weird election dates. My preference would be for two-year cycles aligning with Congressional races. But I don’t think that’s going to make a huge difference. It’s not like voters will suddenly care more or will be more informed about state and local elections simply because it happens during a Presidential election year.
@Steven L. Taylor: I am good with ” This is possible” 🙂
But yes, I don’t know without exit polling 3 million plus Virginians. It is impossible to know. But there is that vibe in VA, you see the people, the reactions, the social networks…the climate was different for sure. And yes, I said vibe lol. How is that for empirical evidence!
@James Joyner:
What? Are you trying to imply that vote-by-mail doesn’t automatically throw all elections to the Ds?
@James Joyner:
“Technically” seems to be doing some heavy lifting here. How many businesses are required to be closed so that their employees can vote?
@Andy: I wouldn’t presume to speak for Dr. T, but allow me to elaborate a little.
First, nobody’s discussing 2021 v 2017 except some fool at NYT as noted above@gVOR08:. The whole discussion is how did VA go from Biden up by 10% to Youngkin up 1.4% in one year. (All numbers IIRC.)
A big chunk of the answer is simply that 4.5 million Virginians voted in the 2020 presidential election, only 3.3 million in 2021. It’s a commonplace that D turnout drops off more than R in off-year elections. There are lots of reasons, but the fall off is disproportionally minorities, the poor, and young people. Who skew D. The off-year electorate is significantly more R than the prez year electorate. I haven’t seen any numbers, but it probably cost McAuliffe at least half of Biden’s margin before anything else happened.
Channeling some Garett Jones today. Very nice series of posts by the way. This one had me thinking about Albert Hirschman, which inspired these somewhat jumbled thoughts.
Implicit in a lot of these discussions is the assumption that more voters is better. You gesture at this above, but the assumption goes unexamined. Indeed, the only thing that is really considered is whether we would be better off voting for more/fewer offices. And the primary point you are making is about the timing of such elections. Your points on these items are well-taken.
Left unaddressed, though, is whether we would be better off with more/fewer voters. Or perhaps more narrowly, whether “voice” is the only option that could lead to an optimal outcome (however one defines it in a democracy). That is, “voice” is assumed to be the only (ethical, respectable, viable) option here.
“Exit” isn’t considered at all. It is only obliquely referenced when discussing who is voting. To be sure, there’s plenty of discussion of why people don’t vote (eg, lazy, uninterested, disenfranchised, etc.). But this is different than “exit” per se. This seems to me an important unaddressed option that we ought consider.
@Michael Cain:
Hence the technically! I don’t think any state in the union requires any business to close for any holiday. All a state holiday does is close state entities–including schools.
@gVOR08:
Vote-by-mail tends to redound to the benefit of Ds for a variety of reasons. But Rs can still win, obviously.
@Mimai:
I’m not sure I follow. Are there arguments that we’d be better off if only the well-informed showed up to vote? Sure. But voting is relatively difficult, especially for those who earn an hourly wage, lack flexible hours, lack transportation, etc. So making people vote repeatedly in order to be heard will impact some more than others.
Regardless, the point of the post (and the series of posts) is simply to explain why the outcome of an election in November 2021 would be different from one in November 2020 even independent of the candidates/campaign/issues that always seem to be at the center of the analysis.
@Mimai: I think that a fundamental goal of representative democracy is to represent the various interests of the society and that one of the ways, especially in our system, to increase representativeness is to have as good of a sample of the populace as is possible via the ballot box.
I think that issue stands apart from an analysis of why people might not vote (although I would note that a lot of people don’t vote because they feel as if they aren’t heard, i.e., not represented).
I will return, however, to my core point. If the system is designed to elect the one set of offices on Day1 and another set of offices on Day2, it is, by definition, creating a circumstance wherein Day1 with have Electorate1 and Day2 with have Electortate2.
From a democratic theory POV this can be problematic if is it known that D1 will always have more voters than D2. And, further, if the distance between D1 and D2 is a mere year, one might reasonably ask why D1 and D2 aren’t merged to take advantage of the turnout in D1. (It also would save the state a chunk of change).
The point is not simply that more voters are better, but that the choices made about when to assemble the voters has to effect of changing who is making representational choices.
This all made the more problematic when people pretend like E1 and E2 should be judged as identical when trying to understand how each group responded to messaging and candidates.
There is also a broader perversion here of national politics in general wherein if there aren’t massive policy shifts/successes in less than a year into a presidency that it is deemed a failure and we start upsetting the partisan applecart. It puts us in a perpetual state of election, reaction, election, reaction and it is yet another way in which governing is made difficult and biases the system to dysfunction.
@James Joyner:
Your inability to follow is entirely my fault – a combination of jumbled thoughts and imprecise writing. I wasn’t speaking to the issue of well-informed vs. less-informed voters. And I fully agree with your point about barriers to voting.
(side point, this brings to mind the distinction between fuel and friction, and how different people seem to privilege one or the other when trying to “get out the vote,” without thinking deeply about when and for whom fuel vs. friction ought be emphasized)
Rather, I was trying to point out that people may “vote with their feet” rather than with their selection at the ballot box, such as it is. And rather than summarily assume these folks are apathetic (need more fuel!) or face too many barriers (reduce friction!), we ought also consider that non-voting can be a vote in and of itself. It’s a vote of “exit” for a particular election. Or all of them.
I don’t think this consideration is counter to the points that you are emphasizing or to Steven’s. Rather, I think it’s another factor to consider when trying to make sense of specific election results, longitudinal patterns, and voting cycles.
@Steven L. Taylor:
Agree with your summary and the reminder of your main point. I did not misunderstand it. Rather, I wanted to spotlight something that I think lies below the surface of many of these discussions.
Your parenthetical (bolded by me) is related to my point, but not quite it (note, I don’t expect you to drop your point and PAY ATTENTION TO ME AND WHAT I WANT TO DISCUSS…I’m merely typing into the wind).
Citizens feeling that they are not well represented by their, er, representatives is indeed an important issue. Likewise, many citizens do not feel they are well represented by the structure of governance itself.
You’ve written a ton about this. And I’ve learned a lot from your writings. I was using your recent posts to elevate the exit vs. voice perspective, which I don’t see, er, represented (explicitly) in these discussions.
Finally, you write:
Perversion indeed! But that’s where the entertainment lies. And we must be entertained.
@gVOR08:
I think that general trend is probably true to some extent. But the question is how big of an effect did it have? Until I see some data I am not willing do downplay the effects of issues and candidate selection as minor factors, especially considering that issues and candidates are major factors in determining turnout.
And the other point is that the composition of the electorate is always different. Unless one can show some data, stating that the composition of the electorate was decisive in one election but wasn’t in another election seems like wishcasting. The example I gave was 2016 vs 2020 where the electorate increased by 17%, or about two-and-a-half times the popular vote margin between Biden and Trump. How big of a factor was that compared to candidates and issues? I don’t know and I don’t think we have the data to answer that question. Similarly, we don’t have the data to answer the question of how big an impact off-year turnout had in the VA race. At least that I know of.
@Andy:
But there’s tons of data that off-year elections have lower turnout than Presidential-year elections and that the deficit almost always rebounds to the benefit of the GOP.
@Mimai: FWIW, I would agree that the question of why American voters don’t turn out in higher numbers is an important one.
@Andy:
Well, I think that the data are quite strong that the main driver of turnout is which offices are being contested and that the make-up of those electorates matters.
While I cannot prove it, but I can logically infer that McAuliffe would have won the governorship even with his education gaffe if the election had been run in 2020 instead of 2021. The partisan environment was simply different.
And I am not arguing that issues and candidates are “minor factors” but I am arguing two things:
1. The choice of when elections are held matters and affects outcomes because it affects turnout. I have provided data for VA in the post. But there really is not disputing the empirical reality of the basic point.
2. This means that whatever narrative one wishes to apply to a given electoral outcome is probably not as deeply true as one thinks it is. More importantly, the narrative that the media and politicians glom onto are almost always overly simplistic.
Back to #1 for a second: the multi-decade pattern in VA suggests that any R versus any D with any issue set I had a really good chance of producing the same basic outcome. I fully realize past performance is not indicative of future outcomes and all of that, but given the pattern of a drop in the number of voters and the fact that the out-party tends to be more motivated to vote a year out has meant in the past that candidate quality and issues were not as important as when the elections were scheduled.
I suspect if I did the research, however, we would find some version of hot takes about what was wrong with the loser and how issues X, Y, or Z are now the pivotal ones for the mid-terms.
I will say this, although I know a lot of people don’t like/believe/accept it: a cardboard cutout of an R versus a cardboard cutout of a D gives you are least a baseline electoral prediction in a given district or state with a major factor determining the winner being who shows up to vote.
I exaggerate, but not by much.
@James Joyner:
Off-year elections always benefit the GoP? Or did you mean to say the party that is out of power?
@Steven L. Taylor:
I don’t disagree with that.
What I’m skeptical about are the follow-on analysis and conclusions regarding when the make-up of the electorate is decisive or the most important factor and when it isn’t. How much of the ~15 point swing can be attributed to which factors? Some of it is undoubtedly due to the pattern of off-year elections benefitting the opposition party. But I think claiming that all or most of it is due to that factor needs some better evidence.
Sure, but everything was different. A campaign of running against Trump would be much more successful when Trump was on the ticket. The salience of issues were different. The mood of the country was different. Covid and economic conditions were different. It’s not merely that the electorate was larger.
Sure, but cardboard cutouts don’t actually run for office. The experience with Trump highlights the weakness of comparing two generic candidates and then making conclusions.
Let’s try something else:
Let’s suppose that McAullife had run a better campaign and won. In which case what would your analysis be? Would it still be focused on the importance of electorate composition as the most dispositive factor?