I tried the Cook Political Report’s Demographic Swingometer
Everyone has heard of the website 270 to win. If you haven’t, you can check it out HERE. It’s a quick and easy way to make electoral college maps.
The Cook Political Report have taken this to the next level and developed a free to use tool where you can change the voting behaviours — including party ID and turnout — of various demographic groups and see the impact this will have on certain states on the electoral map. You can access that HERE.
I spent some time sifting through what I would refer to as “possibilities, not probabilities”. Or, in other words, scenarios that I could realistically see happening on any given day, but not necessarily ones that I would be willing to bet my own money on.
Feel free to disagree with any of my assumptions, but also note that I’m absolutely not saying any of the below outcomes will definitely happen, just that they are broadly within the realms of possibility.
The starting point of this tool is, and I quote, “the results of the previous election, adjusted for demographic change since 2016”.
Below is a table of the tool’s fundamental assumptions. I think the first glaringly obvious thing is the vote share always equals a perfect 100% meaning there is no consideration for third party vote. I know the percentage of third party voting is projected to be way down this year, but as I will prove, even the smallest shifts in these numbers can have a material affect. The Cook Political Report notes “votes for independent and third-party candidates are counted as non-votes in turnout figures”.
The second observation is that there are only five demographic groups which is particularly troubling for the last group. For example, I don’t know who they classify as “other”. In an ideal world, it would have been really cool if they could add more groups, but I understand that adds to difficulty of putting this together, especially where data is sparse.
The above assumptions produce a map that gives Joe Biden 307 electoral college votes, and Donald Trump 231. On the face of it, that seems like a reasonable assumption, but it gets there in a more unconventional way. These numbers mean Biden wins Florida but loses Arizona. Maine’s second district and Nebraska’s second district goes Republican. North Carolina also stays red. This is not necessarily in line with the polling.
White, non-college educated
I remember reading a piece of research that concluded there might be nearly half a million non-college education white male voters in Wisconsin alone who didn’t vote in 2016. To contextualise just how large a number that is, Trump carried the state by a mere 22,000.
Of all the demographics the tool has data for, the white non-college educated group is the only one where Republicans command a healthy lead. It is predicted Trump carriers them by nearly 40%. Unsurprisingly considering he carried that group by just over 40% in 2016, securing 67% of their vote, according to the Pew Research Center.
Even with an increased share of that vote (up 2% points), Trump still loses the election, however. But dramatic changes occur quickly after.
If Trump gets 70% of this vote and nothing else changes, then he wins the election. Biden would flip Pennsylvania, but Trump would hold Wisconsin and Michigan, giving him 290 electoral college votes. If that goes up to 72%, Minnesota and Pennsylvania both go into Trump’s column as well. Interestingly, the next material change comes at 74% where Nevada flips, then Colorado at 78%. Both unlikely.
Fear not, the Biden campaign. If Trump’s share of the vote falls just one point, to 68%, then turnout amongst that group will have to increase a whopping 7% points for him to experience the aforementioned gains, something that I feel is possible but unlikely.
If Trump hit 69% vote share, turnout would only have to increase 5 % before Trump clears 270. If both things happen and Trump gets — say — 72% of their vote while achieving turnout of 60%, not only does he comfortably win, but he gets over 70 million votes, adding over 7 million from 2016.
A final observation of the white non-college educated vote is that, according to the Edison exit poll data of the 2016 election, this group only represented 33% of voters, whereas the Cook Political Report projects them to represent 43% this time. I understand, after Trump’s success, the desire to make sure this group is thoroughly accounted for, but that does seem a little high to me, especially because it is a fundamentally declining demographic.
White, college educated
The Cook Political Report starts from a very bold position here. According to the Pew Research Center, Trump actually won this group in 2016 by 4% points, beating Clinton 49/45. While the assumption that Trump loses 3% of his support is very reasonable, a Democrat gain of nearly 10 points is bolder. The tool also predicts this group will represent only 31% of the electorate, whereas the 2016 Edison exit poll had that figure at 37%. If non-college whites are a decreasing demographic group, then I find it hard to believe college education whites would also fall, especially by 6 points.
Nonetheless, there is evidence to suggest Biden is in a strong position here. Which, incidentally, would be fairly remarkable as not only did Trump win this group, but so did Romney (56%) and McCain (51%), both by majorities, as opposed to Trump’s simple plurality.
If Biden slipped to 50%, then Trump wins with over 300 electoral college votes, include 4 from New Hampshire. However, if Biden gets up to 57% he begins to get into landslide territory, securing 350 electoral college votes, including carrying Georgia’s 16, and beating Trump by nearly 10 million votes.
Turnout amongst this group is projected to be high compared to others. Which, according to the tool, as the turnout amongst that group increases, it gives Biden more votes, but has no material impact on the electoral college, shy of giving Biden the projected win anyway.
A final note on turnout, however. Because, while it is comparatively high, it would still be lower than 2016 (79.%) and 2012 (77.6%) according to the Center for American Progress. Turnout amongst that group was significantly higher in WI, PA, and MI in 2016 too: 86.1%, 80.6%, and 83.8% respectively.
I think this could be the most important group. A U.K news channel, Channel 4, recently presented their findings into the Trump campaign’s efforts to suppress the Black/African American vote in 2016. Similarly, Trump is courting Black/African American voters in an attempt to chip away at the margin of which Democrats win this group. I have written before about the how important shortening those losing margins will be.
It was one of the most significant under-performances within Hillary Clinton’s electorate. The Washington Post claims that 11% of Black/African American Obama 2012 voters stayed at home in 2016. The extent to which Trump’s performance amongst this group is a trend will be an interesting question. While the GOP performed poorly amongst them in 2008, there was a steady uptake in 2012 and 2016, adding about 2% points each year.
Where you sit on that question will have big implications for how you use the Cook Political Report’s tool. The tool predicts Trump to win 8% of votes from this group — a result I suspect his campaign would be disappointed with. But, because it does not account for third party voting, it therefore gives Biden a huge 92% of their vote, as opposed to Hillary’s 88%.
Once again, however, these numbers are a relatively new development. According to the Pew Research Center, in 2004, while Kerry matched Hilary’s 88%, Bush secured 11% of the group’s vote, reducing the margin to 77 points. While that was good enough for Bush to secure re-election, the population growth amongst the demographic means that should the 2004 numbers be mirrored exactly, Biden still wins 307 electoral college votes, partly through being able to hold Florida.
If we assume third party voting will be lower, then the magic number will is somewhere around 84/85%, which puts Michigan and Florida into Trump’s column and gets him over the line. To drop so significantly from Clinton’s number is, I suspect, unlikely. A more likely lever that could be pulled would be around turnout, not least because, as we have seen, there are concerted efforts to suppress this group’s vote from the GOP.
The tool estimates turnout amongst this group to be 57%. As the Center for American Progress highlights, turnout fell “more than 4 points from 62.1 percent to 57.7 percent” from the 2012 to 2016 elections. The assumption that turnout remains at this level seems slightly conservative to me. The tool predicts that if this group’s turnout climbs to 62% then Biden would win Georgia, anything less than 62% though, and no state swings either way.
Alarm bells went off in September when a poll in Miami Danes county revealed Biden was underperforming there. In 2016, Clinton beat Trump there by 30 points, whereas the polls appeared to show the Democrat lead reduced to just 17. Usual caveats apply: the poll was only of 500 people and had a 4.4% MOE, but it spoke to the idea Trump thinks it can do better amongst this group this year. The New York Times goes as far to say Hispanic/Latino men are “part of Trump’s base”.
How much of this was due to factors not relating to Trump though is interesting. The Republican share of the vote remained virtually identical to 2012, and significantly lower than 2008, when Arizona Senator, John McCain was on the ticket.
Expectedly, the two states that are impacted the most by this group’s voting behaviour are Arizona and Florida. The Cook Political Report’s tool projects Trump to, yet again, win 28% of their vote, with Biden winning the remaining 72%. Again, this number is high — higher than Obama’s performance in 2012 — but I suspect that is due to the lack of third party voting integration into the data.
That is almost immaterial though. The vote share would have to get to an unprecedented 57/43 split before Trump wins an addition state other than Arizona and Florida. Similarly, if everything else stays the same, the Trump would have to slip 5 points, to just 23% of their vote for Biden to extend his lead by winning Arizona.
Turnout and the proportion of the electorate this group represents is, as always, important to explore. And it’s where I think some of the fundamentals are inexplicably conservative. For example, the Cook Political Report estimates the group will make up 9.9% of the entire electorate, whereas they represented 11% in 2016. I cannot really understand the drop, not least because US Hispanic/Latino population grew by nearly one million in 2019 alone.
On a more granular level, the Cook Political Report tool says the only way Democrats can flip Texas with an emphasis on the Hispanic/Latino vote is to have their turnout completely sky-rocket from a projected 45% to 74%, or to have Democrats secure 85% of their vote nationwide. Yet, they still only classify Texas as ‘Lean R’. Part of this, I think, is the tool underplaying the volume of Hispanic/Latino voters in the state, mirroring the nationwide fundamentals. Pew Research Center, for example, estimates the group represents 30% of the state’s electorate, whereas Cook Political Report only has this number at 22%. However, according to an AP VoteCast survey, conducted by NORC at the University of Chicago for the Associated Press and Fox News after the 2018 Texas Senate race, that also concluded 22% of the electorate was Hispanic/Latino. I would be surprised if this doesn’t doesn’t increase, at least to the mid-20s, if not 30%.
A phenomenon everyone has to be cautious of is the difference between pre-election and exit poll data, or in other words, “silent Trump” voters. As the Pew Research Center states “most pre-election polls found lower Latino support for Trump than the exit polls did”. This is not a Hispanic/Latino exclusive phenomenon, but that said, the difficulties polling this group are well reported.
In seeking to draw anything from the data, this is the group I found the most challenging. Firstly, the ambiguity of “other” is inherently problematic as one does not know who we are even talking about. Moreover, when talking about the ‘Asian vote’, that is also deeply ambiguous. An Asian American voter survey released in September 2020 splits the Asian American voters into seven sub-groups. In order of their support for Joe Biden, they are: Asian Indian (65%) Japanese (61%), Korean (57%), Chinese (56%), ASIAN AM (54%), Filipino (52%), Vietnamese (36%).
As a broad coalition, they are the fastest growing racial or ethnic group in the US electorate, representing an estimated 4.7% of the electorate this year. The Cook Political Report gives this group 5.6% of the electorate, so I have to assume that about ‘other’ makes up about 1%.
From the data available, which is limited, there is a confusing picture. The Asian American voter survey concludes that, while Democrats have leads amongst six of the seven sub-groups, their support is decreasing within all of these groups when compared to 2016. The same polling company found that in 2016, Democrats actually won Vietnamese Americans by 15 points; they are now losing by the same amount. A 30 point swing.
One of the easy things for both parties to do would to simply ignore this group. They do, after all, only represent less than 5% of the electorate. But, a huge word of caution is needed. There is now ample evidence that Black/African American voters, especially young males, do not have the historic ties to the Democrat party that the previous generations did. There is a feeling amongst them that they have been ignored by the party. Read HERE, HERE, and HERE. If we fast forward a few decades and the Asian American feel the same, it would be a horrible indictment for US party politics.
With conclusions difficult to draw from the available data, the presence of Asian American, Kamala Harris on the ticket, and the historically high turnout amongst the demographic in the 2018 midterms, could translate to a meaningful impact. I suspect that as the influence of this group grows, more detailed demographic breakdowns will come.
A final word of warning. All these changes that I have made have been done under the assumption everything else remains the same. An unlikely situation, I accept. It is, for example, unlikely to assume white non-college voters will turnout in higher numbers without Donald Trump also securing a higher proportion of them, thus exaggerating the impact. Similarly, it seems unlikely African American turnout would increase while Latino turnout would decrease.
An example of where independent movements create a distorted picture is clear amongst white, non-college educated. As I mention, the tool expects Trump to win 69% of this group, if it increases just 2 points, Trump secures a comfortable re-election win. But if, at the same time, if Black/African American turnout also goes up 2 points then Biden still wins regardless. The tool allows you to explore all of these interlinked possibilities.
Another specific example is in Arizona. For example, the Cook Political Report themselves already has Arizona as a ‘Lean D’ state, yet, when looking through the lens of Hispanics/Latinos alone, Trump would have to drop his share of the vote quite significantly in order for this to happen. If the right movements occur from other demographics, then the level of support necessary from Hispanics/Latinos in Arizona remains quite conservative for Biden to win the state.
The conclusion is a fairly boring one: that while Biden is in a very strong position, just a few small percentage shifts amongst key demographic groups could radically change the electorate in key states, enough to secure Donald Trump re-election.