Skip to content
 

2020: What happened?

Yair Ghitza and Jonathan Robinson write:

Based on our analysis to date, the following are our 10 big takeaways for election results for the top-of-the-ticket presidential race:

1. This was the most diverse electorate ever. The voting electorate continues to become more diverse, and 2020 was the most racially diverse electorate ever. This was due to big turnout increases in communities of color, particularly among Latino and Asian voters. The electorate was 72% white, compared to 74% white in 2016 and 77% white in 2008. This composition shift comes mostly from the decline of white voters without a college degree, who have dropped from 51% of the electorate in 2008 to 44% in 2020.

2. Biden and Harris won with a multiracial coalition. The Biden-Harris ticket benefitted from a diverse set of supporters: 39% of their coalition were voters of color, and the remaining 61% were split fairly evenly between white voters with and without a college degree. They also made significant gains among white voters compared to 2016, particularly among white college and white suburban voters, who have shown a solid and consistent backlash against Trump’s Republican party. The Trump-Pence ticket was more homogenous . . . only 15% voters of color.

3. Latino voters continued to favor Democrats, but Republicans made inroads with Latino voters, too. Along with massive increases in turnout, Latino vote share as a whole swung towards Trump by 8 points in two-way vote share compared to 2016, though Biden-Harris still enjoyed solid majority (61%) support among this group. . . .

4. Black voter turnout increased substantially, resulting in significant gains for Democrats, despite a modest overall drop in Democratic support levels. . . . This dynamic – many more voters turning out but at a slightly lower Democratic margin – resulted in more net Democratic votes from Black voters in 2020 than in 2016, particularly in several key battleground states. For both Black and Latino voters, we discuss how an expanding electorate might bring marginal voters into the electorate at slightly lower support levels.

5. Asian-American and Pacific Islander voters saw the largest relative increase in turnout, which benefited Democrats. Even in a high turnout year, AAPI voters had a remarkable jump in turnout, the biggest increase among all groups by race. The number of AAPI voters increased 39% from 2016, reaching 62% overall turnout for this group. AAPI voters remain strongly supportive of Democrats, delivering a 67% vote share to the Biden-Harris ticket, largely consistent with past elections.

6. The urban-rural voting divide continues to be important, with suburbs growing more Democratic and more racially diverse. The relationship between urbanity and voting is essentially as strong as ever, though it did not grow wider in 2020 than in recent years. . . . The Biden-Harris ticket maintained gains in the suburbs that began earlier in the Trump presidency. These gains are not all about white suburban voters, as is sometimes misunderstood. Suburbs are increasingly racially diverse, which accounts for part of the change in voting patterns.

7. Women remain critical to the Democratic coalition. Women comprise 54% of the electorate overall and an even larger majority of the electorate among Black (59%) and Latino (56%) voters. Overall, women voters of color supported the Democratic ticket at a rate of 79% while support among white women was 48%. We find a 10-point gender gap, with women supporting Democrats more than men fairly consistently across races. White college-educated women in particular have shifted against Trump, moving from 50% Democratic support in 2012 to 58% in 2020, a trend that began in 2016 and continued in 2018 and 2020.

8. Young voters drove record-breaking turnout. 2020’s historic voter turnout gains were primarily driven by young voters. 18-29 year olds grew from 15% (2016) to 16% (2020) of the voting electorate, but the generational changes have been even more dramatic. Millennials and Gen Z now account for 31% of voters, up from 23% in 2016 and 14% in 2008. . . .

9. New voters made a big difference, especially in Sunbelt swing states. . . . Nationally, 14% of voters were first-time voters, who we haven’t seen vote in a previous even-year general election. This understates the change from 2016, however, due to many first-time 2018 voters and other sources of year-to-year turnover. When we compare state-by-state electorates from 2016 to 2020, 29% of voters were new presidential voters in their state in 2020. Some of these voters registered and voted for the first time in 2018, others were brand-new in 2020 or moved from out of state. . . .

10. There are still millions of non-voters who could cast ballots in future elections. Electorates are dynamic, and millions of voters drop out and join the electorate with each midterm and presidential election. 2020 saw millions of first-time voters, even though campaigns and civic engagement organizations were not able to run traditional registration, canvassing or get-out-the-vote operations due to the pandemic. At the same time, over 70 million eligible citizens did not cast a vote in 2020 and may cast ballots in the future, further changing the electoral landscape.

Lots more detail at the link.

I guess this recent success for the Democrats is one reason that so many Republican party leaders are going all-in on not counting all the votes. Scary times.

P.S. Yair adds this graph:

and writes:

If you look from 2008 to 2012, for example, you see almost no growth in # of votes among youngest generations (when voter turnout went down by 3 points overall, as percent of eligible population). The more general point is that if turnout goes up overall (as it did in 2020), older generations are already voting at higher rates and there’s a ceiling effect, and the larger gains are among younger generations.

That may be obvious to you, but we’ve gotten a ton of feedback that this was totally unclear to other people. As a comparison, here’s the % of electorate that are made up of young voters (18-29) in each of those elections, 2008 to 2020: 16%, 15%, 15%, 16%. The vast majority of people who study / think / write about elections think in those terms, and it looks like close to nothing is happening.

I agree that the growth in Gen Z looks really high because half of them were ineligible in 2016, and we say exactly that in the paper! I also agree that a lot of this is a natural progression as people age, but you’ll (eventually) see that part of this was unique to the increased turnout in 2020.

36 Comments

  1. D Kane says:

    > so many Republican party leaders are going all-in on not counting all the votes

    Is that a neutral/reasonable description of the Republican view on election policy? For example, Hawaii requires that mail-in ballots arrive by Election Day. Are the Democrats who run Hawaii in favor of “not counting all the votes?”

  2. jim says:

    “39% of their coalition were voters of color”

    What does “of color” mean? I’ve never seen a list of sub groups that are included in this group. Is a “Latino” person white or “of color”? Is a Mexican person “Latino” or “Hispanic”? These terms are used as though they are all mutually exclusive subgroups but I don’t think they are, so the comparisons could be misrepresented.

    “Biden and Harris won with a multiracial coalition…The Trump-Pence ticket was more homogenous . . . only 15% voters of color.”

    “White” people are all one race? The English and Italians and Russians? Are Lebanese and Palestinians “white”, or “of color”? What about Arabs?

    • somebody says:

      Seconded. I’m pretty sure this means anyone who can’t pass as a WASP, but I don’t know how informative that is about someone’s politics anymore. Shortly after the election the seeming disintegration of the existing racialized voting blocs attracted some attention, mostly around Cuban Latino voters but also around a broad realignment. Recent immigrants, especially from overseas countries, are pretty strongly bifurcating depending on their country of origin. Among Asians in particular, a strong international coalition against PRC imperialism has responded to Trump’s one big good thing and driven a formerly staunchly blue demographic quite a bit more red. I don’t know where or whether things are going to crystalize, but without such crystallization I think it’ll help to be really specific.

      • Andrew says:

        Somebody:

        1. No, “of color” does not mean “anyone who can’t pass as a WASP.” As I said in my comment to Jim, I think it refers to everyone who in their survey responses do not call themselves non-Hispanic white. This includes lots of people who would never pass as a WASP.

        2. You should read the report. They write a lot about subgroups of voters. In particular, they find that Asian Americans and Pacific Islanders gave Biden 67% support in 2020, pretty much identical to the 68% that Clinton received in 2016 and the 66% that Obama received in 2012. So I don’t see any support here for your claim that this “formerly staunchly blue demographic” became “quite a bit more red.”

        Regarding point #2: This sort of thing is why Yair did this report in the first place. You can learn a lot from the numbers! Sometimes what you’ve heard somewhere, or you think is true, isn’t. (Of course their report could have mistakes too. My point here is that we can often learn from data, beyond what we might think based on guessing and stories.)

        • somebody says:

          > 2. You should read the report. They write a lot about subgroups of voters. In particular, they find that Asian Americans and Pacific Islanders gave Biden 67% support in 2020, pretty much identical to the 68% that Clinton received in 2016 and the 66% that Obama received in 2012. So I don’t see any support here for your claim that this “formerly staunchly blue demographic” became “quite a bit more red.”

          This kind of thing is actually exactly what I mean. Perhaps my phrasing wasn’t precise enough, but there’s definitely a rightward (or Trumpward) shift among Asian immigrants.

          This report focuses a lot on Latino neighborhoods, but Asian neighborhoods are definitely in there
          https://www.nytimes.com/interactive/2020/12/20/us/politics/election-hispanics-asians-voting.html

          the Asian demographic LOOKS static because it averages out with higher political mobilization with the younger, native born population. Look at the difference between the native and foreign-born numbers here
          http://aapidata.com/blog/biden-trump-gap-very-high-among-younger-and-native-born-asian-americans/

          So if you’re looking to see the future of the Asian American vote, it’s up to whether the population will be primarily new immigrants or children of Asian Americans.

          I know that any categorization can be picked apart if you look at it closely enough, and these classifications are just a helpful dimensionality reduction, but I really do think things are changing here in a way that makes this particular set of categories less helpful than they used to be.

        • Clyde Schechter says:

          “I think it refers to everyone who in their survey responses do not call themselves non-Hispanic white. “

          I’m not familiar with this particular survey, but certainly in health care, self-identification is the standard way of ascertaining race and ethnicity.

          Of course, it has its problems. It does not provide consistent answers to questions like those raised by jim and somebody. And in longitudinal data, a substantial fraction of respondents will respond differently over time. In addition, the standard racial groupings offered as response options are losing their relevance over time as the number of people who are multiracial increases.

          It’s a serious and important problem in my view. We cannot develop fair policies without good data. But it is difficult to do that, and increasingly so over time.

          • jim says:

            “And in longitudinal data, a substantial fraction of respondents will respond differently over time.”

            So people “of color” become…er…less colorful as they grow older, become more secure and financially established? :) I believe this is true. When people become successful their peer group changes and gradually they come to identify with that peer group. It’s not so much that they recognize a distinct difference in the “color” of their peer group. It’s more that the labels for the groups, however expressed, identify to some degree by culture rather than by race. When you have a management job and a decent place in the suburbs you’ve worked hard to get, you might not feel like identifying with people campaigning for higher min wage and workers rights, but you have a limited number of labels open to you.

    • Andrew says:

      Jim:

      I think this is based on survey responses. “White” is a characterization that people can give to themselves (as with survey responses) or can be given from others (as in police reports). “White,” “Black,” etc., are conventionally referred to as races, but “race” does not have any precise definition. I sometimes define race as divisions of ethnicity that happen to be salient in current society. So, in the old days, English and Irish or Czech and Slovak used to be considered different races, but now they’re considered to be ethnic groups, not races. Similarly, Americans typically think of Africans (or perhaps sub-Saharan Africans) as a single category, but there are lots of differences between people from different parts of sub-Saharan Africa.

      So, the answer to your question, “White people are all one race?”, is that it depends on the definition. According to the U.S. census, yes. When Yair or I write about how white people vote, we’re typically talking about non-Hispanic whites, as defined by people’s survey responses. Nothing deeper than that. I think that people from the Middle East are typically considered white in these definitions, even though Lebanon, Palestine, Iran, etc., are geographically in Asia.

      Finally, I’m not 100% sure how Yair is using the phrase “of color,” but I’m guessing he’s referring to everyone who’s not non-Hispanic white. Again, nothing magic about these definitions, and, indeed, if you look at the report you’ll see lots of discussion about how these are not monolithic blocs.

      • jim says:

        Great thanks Andrew.

        As I was writing that comment it occurred to me that there’s probably a standard among demographers / poly sci people (which is I guess what you described) for how these terms are used, in which case I would just be out of the loop.

      • Yair says:

        Yep this is correct, it is based on self-reported data, and “voters of color” include everyone except white voters. I agree all of these groups shouldn’t be treated as a monolith, and in the report we separate things out into particular racial groups in many places! But sometimes it makes sense to group things together (with further details after that), so that’s the terminology we used. FWIW in the past I’ve written reports that label that group as “non-white,” but have gotten feedback that is even worse.

        Probably worth adding some specifics on a few points. One complication is that some self-reported data (surveys, registration forms) uses two questions to identify race — one question about race (White, Black, etc), and a second question asking whether the respondent is “of Hispanic origin,” with everybody saying yes to the latter coded as Hispanic. Alternatively, a single race question is sometimes used with “Hispanic” as one of the choices. The “single question” format typically results in a smaller estimate for Hispanic, and a higher one for White, because some people of Hispanic origin self-identify as White given this forced choice.

        Multi-racial is also complicated, because some surveys code multi-racial respondents as “Other race” and some choose one, for instance saying anybody who is both black and white are labeled as black.

        For us, we use self-reported data from voter registration forms where available (21% of registered voters), and then a model to approximate otherwise. The model is trained on a mix of these types of questions (we may change that in the future and just choose one coding scheme to be clear on these rules). That may be part of the reason we are a point or two higher on % white than the Current Population Survey (another popular data source for understanding composition of the electorate).

        It’s a tough question to figure out what is the “right” number because of those complications — see here for an academic study arguing that the CPS has some problems on this front: https://www.bernardfraga.com/s/CPS_AFS_2021.pdf

    • ie Rabinovitz says:

      An old acquaintance of mine named “Cohen” had grandparents named “Quinn”.
      At the time I was a youngster and thought it remarkable; and moreover the “Quinns” didn’t make the slightest effort to pass as “WASPS” either.
      I asked Mr. Quinn the provenance of the surname.
      It used to be “Kvinn”, he explained, winking as he served up this likely-story.

  3. Phil says:

    I either don’t understand the stuff about young voters, or I understand it and think the point is ridiculous. I fear it’s the latter, but I’m not sure.

    “When we look at composition of the electorate by age, we see a slight increase in the percent of the electorate under the age of 30, going from 15% in 2016 to 16% in 2020. While this increase directionally lines up with what we would expect, the change seems modest given the circumstances.” I’ll say.

    But then they go on to say that there were actually huge changes by “generation”. Hey, just look at the plots, the “Greatest Generation” (born 1928 and earlier) dropped by half! Gen Z (born after 2000) leapt upwards!

    WTF, how could that _not_ have happened? The “Greatest Generation” is dying off (quickly), and Gen Z includes four more years of eligible voters than it did four years ago!

    If you think the boundary years between “generations” are truly meaningful discontinuities in some way, something like this can make a difference, but to me it seems kinda silly.

    • somebody says:

      The full text of their analysis actually makes this even more confusing


      It turns out that looking at composition by age substantially understates the magnitude of the change in this election. Age data can be confusing because substantial portions of the electorate “age out” from one group into another. For instance, 6% of voters were between the ages of 30 and 33 in 2020, meaning they moved from the 18-29 group, a common cutoff for discussing “young” voters, to the 30-44 group.

      We can see what is happening more clearly by looking at birth year or generation. In short, 2020 accelerated a massive change in the composition of the electorate, with Millennials and Gen Z taking an increasingly prominent role in the future of American elections – a demographic change that is functionally permanent. Figure 16 shows the data year-by-year. Turnout increases were clearly largest among the youngest generation. In terms of raw number of votes, Gen Z increased their total from 2016 by nearly 300%, since many were under 18 and therefore ineligible in 2016. Millennials were the second largest increase at 27%. The turnout increases by generation continue to decline from there. The “break even” point is around age 70, with people born in the 1940s or earlier declining in terms of raw number of votes.

      Their point is rescued however by a continuous-time visualization and their last couple of sentences. Turnout increased for everybody except the oldest generation which is dying off, but turnout increased by much more for younger people including those who were eligible to vote in 2016. This pretty strongly implies that given an eligible voter in both 2016 and 2020, the difference in propensity to vote was negatively associated with age.

    • Andrew says:

      Phil:

      The missing piece here is that voting patterns vary by cohort rather by age; see here. So a changing of the cohorts makes a difference.

      • Phil says:

        Sure, but that seems irrelevant to my point/complaint. Or…well, maybe I just don’t like the way they frame the age-related issues.

        People born in the past thirty years have different political views than those born 50 years earlier. Sure. If you define a ‘cohort’ in terms of, say, who was president when they reached voting age, then voting by cohort is going to change as older voters die and younger voters enter the voting rolls. That’s entirely unsurprising, indeed it could hardly be otherwise. Is it not the case that in every election the most recent “generation” to reach voting age has a large increase in the number of them who vote, and the “generations” that contain people over 75 years has a large decline?

        To me the big story here is the one they instantly dismiss: the share of younger voters did not increase (defining ‘younger’ very generously to include everyone under 30). Every election we hear that the Democrats are energizing younger people, who are fed up with the lack of opportunities and the high student debt loads yada yada, and every election the dreams of a massive youth turnout fail to be realized.

        • Andrew says:

          Phil:

          Sure, but the difference here, compared to 20 or more years ago, is that there’s a huge gradient by age in how people vote. So replacing oldsters by youngsters will in itself have a large effect.

          • Yair says:

            Hi Phil — point taken that this may have been written in a confusing way. What we were trying to say was that the turnout changes by birthyear/generation are stark, and that looking at age data doesn’t capture it. If you look at other years, it’s certainly true that the oldest voters drop off disproportionately, but it’s not necessarily true that you see such stark diffs between generations.

            I … don’t think I can post an image here? But I’ll send one to Andy that maybe he can post, showing the same continuous chart (Figure 16) adding 2008 and 2012. You can see there that if you look from 2008 to 2012, for example, you see almost no growth in # of votes among youngest generations (when voter turnout went down by 3 points overall, as percent of eligible population). The more general point is that if turnout goes up overall (as it did in 2020), older generations are already voting at higher rates and there’s a ceiling effect, and the larger gains are among younger generations.

            Maybe that’s obvious to you? But FWIW we’ve gotten a ton of feedback that this was totally unclear to other people. As a comparison, here’s the % of electorate that are made up of young voters (18-29) in each of those elections, 2008 to 2020: 16%, 15%, 15%, 16%. The vast majority of people who study / think / write about elections think in those terms, and it looks like close to nothing is happening.

            BTW I agree that the growth in Gen Z looks really high because half of them were ineligible in 2016, and we say exactly that in the paper! I also agree that a lot of this is a natural progression as people age, but you’ll (eventually) see that part of this was unique to the increased turnout in 2020.

    • David Chorlian says:

      The graph by Yair may be intended to make this issue more clear. I think the graph would be more clear if
      1) The x-axis was reversed.
      2) The y-axis was the turnout of the voters born at an age as a percentage of the eligible voters of that age. This would eliminate the actual numbers, but the number of eligible voters could be plotted with the same x-axis but a different y-axis on the same graph as a background. As it is the graph conflates demographic changes with political engagement changes.
      3) To examine age effects directly, as opposed to the cohort effects, the same information could be plotted by age at voting rather than age at birth.

  4. Dzhaughn says:

    “I guess this recent success for the Democrats is one reason that so many Republican party leaders are going all-in on not counting all the votes. Scary times.”

    My counter-guess would be: leadership of both parties are posturing on elections to build favorable myths among their bases. Your sentence serves as a modest example of the mythical nature: non-falsifiability (“one reason” “so many”), application of a non-informative cliche (“not counting all the votes”), and the invocation of fear.

    Of course, it is easy to find more extreme examples, on both sides.

    • Andrew says:

      Dzhaughn:

      It’s not mythical at all. I just figured the readers of this blog are well enough informed not to need the details. Just look at the statements regarding the election coming from Donald Trump and Ted Cruz, for example, and look at what happened with Liz Cheney. “Going all-in on not counting all the votes” seems like an accurate description. This is not a cliche; unfortunately it’s reality. As to “fear,” yeah damn right I’m scared to have leaders of one of the country’s two major political parties taking the side of people who were trying to nullify the election. Fear is how we react to scary things.

    • Joshua says:

      Dzhaughn –

      Courts have found that some Republicans were intentionally suppressing votes (along racial lines), and some Republicams are on tape voicing that objective. Prolly some Republicans are actually triggered about widespread fraud – although evidence of such is lacking – and no doubt Democrats have their own forms of political expediency.

      But it seems blatantly obvious that the quote from Andrew that you excerpted is accurate and given so, I don’t understand why you wouldn’t find it concerning or (it seems) think that Democratic malfeasance somehow diminishes it’s significance.

  5. Cliff says:

    “Going all-in on not counting all the votes”

    yes, details are important.

    ‘Not counting invalid votes’ should also be a top concern in any legitimate election — and that was the prime Republican (and others) 2020 concern. Winners of that election found it of little concern, unsurprisingly.

    Nobody here can honestly state that there was NO fraud in the 2020 Election across the the vast U.S.
    Nor can they honestly state how much or little fraud might have occurred.
    There is no reliable mechanism in large U.S. elections to detect or measure fraud in the casting/counting/reporting of votes.
    That is a huge problem for American democracy, but partisan, power politics always rules the day (and year and century).

    • Andrew says:

      Cliff:

      We’re not talking about 1 or 2 votes; we’re talking about attempts to throw out thousands or millions of votes. The claims underlying these attempts have been baseless; for details, see here.

      • Roger says:

        That article does not say anything about throwing out millions of votes. It argues that fraud was not proved. There are Republicans who strive for a more fair and transparent election, even if fraud was not proved.

        It is mainly the Democrats, not the Republicans, who are seeking radical changes to election law.

        • Andrew says:

          Roger:

          Ted Cruz, Donald Trump, etc., were supporting efforts to throw out thousands or millions of votes. The article I linked to explains that there is no evidence in favor of those claims. To say that fraud was “not proved” is to way understate the case here. “Not proved” could imply there is strong but not 100% convincing evidence, for example saying that it was “not proved” that O. J. Simpson killed two people. Regarding the claims of massive fraud, it would be more accurate to say there is “no evidence” or “no real evidence,” in the same way that there is no evidence that child trafficking ever took place at Comet Ping Pong.

          • Zhou Fang says:

            Furthermore it’s blatantly obvious that Republican attempts to find “fraud” is restricted solely to alleged fraud that could increase Republican political power, frequently at the (supposedly coincidental) cost of preventing much larger numbers of entirely valid voters from being able to vote. There are no Republicans striving for a more “fair” election, the entire process, even if there’s fraud, is designed to make things more unfair.

            • Zhou Fang says:

              Like, from a scientific perspective, a good faith attempt to seek out fraud would focus firstly on sites and races showing discrepancies from pre-election polling, which is predominantly downballot races where the republicans did much better than expected, or say, Florida. Even if you adopt the strategy of checking for fraud where it might make a difference, you would strive to spend equivalent effort on checking for fraud in marginal locations where the GOP won. The methodology being applied by the GOP and its operatives is entirely invalid and would not pass any standards for academic inquiry.

              • ie Rabinovitz says:

                I’m sure you know well that standards for academic inquiry and standards for revolutionary activity are different genera.

        • ie Rabinovitz says:

          It has not been proved against all plausible objection that martians do not celebrate martian christmas by eating oysters with green-marmalade and fried martian potatoes in lairs beneath the surface of the martian moon phobos. Those who dismiss this possibility out-of-hand have some explaining to do for themselves. The gall!

    • ie Rabinovitz says:

      Excuse me sir, but the pedant that I am cannot resist “helping out” — though I am sure it will be of no help at all.

      “Nobody here can honestly state that there was NO fraud in the 2020 Election across the the vast U.S.”

      There cannot have been *no* fraud, therefore it is likely that there was fraud of a monumental scale?

      “Nor can they honestly state how much or little fraud might have occurred.”

      Were the many courts which heard the claims of fraud so stupid that they failed to see this alleged iceberg of fraud just beneath the surface? Or were they part of the same conspiracy which was alleged to have perpetrated the fraud?

      “There is no reliable mechanism in large U.S. elections to detect or measure fraud in the casting/counting/reporting of votes.”

      Were the recounts and audits which were called into being by these claims of fraud uniformly faulty or fraudulent themselves?

    • wow says:

      > There is no reliable mechanism in large U.S. elections to detect or measure fraud in the casting/counting/reporting of votes.

      You probably support the Arizona audit where they’re searching for bamboo ballots lol

Leave a Reply

Where can you find the best CBD products? CBD gummies made with vegan ingredients and CBD oils that are lab tested and 100% organic? Click here.