Donald Trump is a big underdog to win re-election
This article answers a simple question: what is Donald J. Trump’s chance of winning the 2020 Presidential election? I first assess Trump’s chances of getting reelected with his current approval rating (spoiler: not good) and then explore the chances his popularity improves enough to win another term.
Trump isn’t currently popular enough to win re-election, he needs to win people over.
Table 1. lists each President’s Gallup approval rating at time of re-election (the average of their rating in the last poll before and first poll after the election) and the percent of the combined Democrat-Republican vote they won.
Doing a simple one variable regression yields R² of .83 — i.e across the sample a President’s approval explained 83% of the variation in his re-election vote percentage.
We can use this to estimate a President’s chance of winning the popular vote as a function of his approval rating.
According to the model Trump’s current 40% approval, taken forward to 04 November 2020, wouldn’t cut the mustard. But could it? Why might the model be wrong?
- Your calculation looks at the popular vote but Presidents are decided by the Electoral College. Republicans have an advantage in the Electoral College as demonstrated by 2000 & 2016.
Democrats held the Electoral College edge as recently as 2012, 2008 and 2004. Electoral College edges are fleeting almost by nature —there’s no gerrymandering and strength in small states is of limited advantage: California still provides more than 17 times the electoral votes of Wyoming, after all*.
This means parties can, unlike in the House and Senate, easily plug Electoral College leaks from one cycle to the next. Democrats would fix their Electoral College simply by… nominating someone more popular in the Rust Belt!
*(Remove two electoral votes from each state, making state’s Electoral College contribution proportional & Trump wins 2016’s electors 246–190. His edge had little to do with small state strength.)
2. We’re in a new paradigm — everyone is so unpopular 42% approval will do much better than before.
Every political cycle is filled with 100 examples of ‘new paradigms’, 99 of which don’t hold up. If 42% if the new 52% then wasn’t Barack Obama’s 52% approval rating in 2012 the new 62%? And why did it only yield only 52% of the two party vote (~what the model predicted!)?
3. A third party candidate will achieve great results (maybe after Democrats nominate a far left candidate), this will help Trump win with a 40% approval rating.
What’s your evidence Democrats will tilt far left? What’s your evidence third parties will perform well (and why didn’t they in 2016?)? What’s your evidence third party candidates would help rather than hinder Trump?
4. The polls underestimate Trump’s popularity.
2016 polls predicted Hillary Clinton would best Donald Trump by 3.2% — not far from her actual margin of 2.1%. 2018 polls did even better. They were excellent in special elections. If you think Gallup has it wrong and prefer FiveThirtyEight’s polling average it gets Trump to 42% — still short of what he’s likely to need.
5. As we get closer to the election people will re-interpret the question as whether they intend to vote for him rather than whether they approve of him, this will put him over the finish line.
Maybe but again, where’s your proof? Trump’s approval rating has a lower number of ‘Not sure’s’ than previous Presidents (see below) and his “strong disapproval” numbers are stratospheric. Voters are less likely to tell pollsters they plan to vote for Trump in 2020 than that they approve of him.
These “shy Trump voters” must be very very shy indeed :/
6. Your model sucks.
Maybe. Should I have use net approval? Was it right to include the average of Harry S. Truman’s June 1948 & January 1949 approvals? Should I have top coded Presidential Approval at 60%?
Perhaps I made the wrong call on some of these. If I did it affects the model’s predicted chance of Trump getting re-elected … if he has a 47% approval. At 40% he’s in big trouble no matter your model, (I think he’d have more chance than the model suggests at 40% for reasons I explain later. But he’d be a big underdog and the most likely outcome would be Trump losing in a landslide.)
7. He’ll cheat.
This is a legitimate concern. Voter suppression, gerrymandering, blocking judicial appointments and lame duck power grabs all suggest the GOP’s respect for democracy is waning. Headed by a authoritarian-fetishist who could conceivably need to win another term to stay out of jail, it’s unlikely Trump & the GOP will meekly march t’ward electoral drubbing.
American democracy under Trump has held up better I expected. But it also held for over a decade in Turkey under then Prime Minister Recep Erdogan before he and his family were caught embezzling millions of dollars. Faced with the possibility of losing an election along with his and his family’s ill gotten gains and freedom (this may sound familiar…), Erdogan turned Turkey into a dictatorship.
America’s strong institutions, Democratic House control and the Trump’s incompetence should salve worries but we shouldn’t rule out malfeasance pushing Trump over the edge in a tight election.
7. He’ll pivot.
Really? Are we talking about the same guy? Trump says racist and sexist stuff because … he holds racist and sexist views, and the impulse control of a rabid puppy. It’s not complicated! Is he going to stop watching ten hours of cable news every day? Display even a basic understanding of a single policy issue? Speak in coherent sentences? Come on.
8. Who cares? Isn’t there like zero correlation between a President’s approval rating during his first midterms and his approval rating during re-election?
Well yes, and no.
Prima facie yes. There’s minimal obvious link between a President’s midterm and re-election approvals.
The four Presidents with the lowest midterm approval rating won re-election (three in landslides) and the three President’s with the highest midterm approval ratings batted .33. Great news for Trump, right? Well, no…
Trump’s approval rating is the 1% of “new paradigms” that are actually new paradigms. We’ve never seen stability like it. His Gallup weekly average approval has never surpassed 45% — the sort of minimum level he needs for a good shot at re-election.
Below is each President’s highest approval and highest disapproval in their first three months of office, via Gallup’s polls/weekly averages.
(Note Barack Obama started very popular, repudiating arguments that Trump’s unpopularity is attributable to increased partisanship.)
People eventually ended up disliking Jimmy Carter & George Bush Sr.’s enough to boot them out, but it wasn’t intrinsic. There was nothing of Carter of Bush’s essence to prevent them winning re-election with 400+ electoral votes, they lost because the economy was in a recession.
In his first three months Gerald Ford took over a scandal ridden administration, issued an unpopular pardon for Nixon, got whacked in the midterms, in the middle of a deep recession and still maxed out at 33% disapproval! Trump started at 47% disapproval — his best ever score! We’ve never seen anything like this. A huge chunk of American viscerally loathes Trump and this is at least as big a story as the Midwestern diner patrons who love Trump™ American broadsheets are oh-so fond or profiling. Trump is really unpopular!
Let’s return to midterm and re-election approval ratings. What if you add in the Consumer Confidence Index? Still no pattern?
You’ll note a few things.
i) The five Presidents who saw CCI improve between midterms and the general won re-election.
ii) The two Presidents who saw CCI decline between midterms and the general lost re-election.
iii) The three highest CCIs coincided with landslide re-elections. The slightly above average CCI saw a close re-election. Two of the below average CCIs saw the President defeated in a landslide. i.e it’s the economy, stupid.
(Obama’s performance in 2012 is probably more impressive than commonly credited for)
iv) Changes in CCI and approval rating moved in the same direction for six of the seven Presidents. The exception is George W. Bush whose post 9/11 popularity faded despite improving consumer sentiment.
We can conclude an improving economy would boost Trump’s approval and hence his chance at re-election. How realistic is this? Well you might have noticed something else
iv) Consumer confidence is really fucking high right now.
9. The economy is still going to be good in 2020 right? Two years isn’t a long time.
It isn’t?
Let’s graph CCI for every election from 1968 on against CCI two years later. So 1968, 1970 … 2016 (x axis) against 1970, 1972 … 2018 (y axis).
No discernible pattern and note the two election years (1968 & 2000) with Consumer Confidence close to 2018’s were followed two years later by below average Consumer Confidence.
10. Please regress Consumer Confidence now on Consumer Confidence x months ago, adjust for seasonal affects and plot the Adjusted R²s.
Great idea!
In English — Consumer Confidence now tells about half the story of Consumer Confidence next year — economic forecasting twelve months out is tough but as a rule the economy doesn’t change dramatically over a year (see graph below).
Two years out R² drops to ~.12 —the state of today’s economy doesn’t have much predictive power about the economy in November 2020.
For an idea of what R² of .12 means, it’s roughly how much predictive power knowing an NFL team’s results (and only their results) this year have on forecasting their results next year.
The table below list how the regular season wins of each NFL team coming off a 13–3 season from 2000 on.
This is sort of represents our best guess as to consumer confidence in 2020: high likely to mean-revert at least a little; more likely to above average than below (but both v. possible); a wide range of outcomes — difficult to predict.
Trump’s smarter advisors note the danger of a recession in 2020 — and such a scenario would indeed get very ugly for the 45th President and his party. But they should also worry about a “9–7” or “10–6” economy, these still push Trump, who needs to win over voters, in the wrong direction.
11. Democrats will overstep their ‘Presidential harassment’ which will earn Trump sympathy popularity a la Bill Clinton.
Really? First people are more willing to forgive if you start off popular (Clinton) than not (Trump)! (Or “When you’re a star they let you do it.”). Most importantly: Kenneth Starr’s team went on actual witch hunts and after four years found … the President lied about an affair.
Democrats have known for months that Trump … lied about an affair with the added spice of directing his lawyer to commit a felony but chose not to run on impeachments in the midterms and are talking it down now.
Seriously, look at this list of incompetence, cruelty, corruption and criminality, add these, then note both lists came before the New York Times exposed Trump’s “self-made man” story as a fabrication masking decades of criminality.
The Donald’s administration could henceforth run with probity that makes the Virgin Mary look like Emperor Nero and it would still face a steady downpour of dropping shoes from crimes and misdeeds already committed, right through November 2020. Thinking is no more comparable to the Clinton administration is than The Deer Hunter is to Bambi.
12. You’ve convinced me. The data is clear — Trump has 0% chance of re-election.
Any conclusion from a dataset of only eleven Presidents will be, at best, al dente. We don’t know enough to say Trump can’t win re-election. There are many ways Trump’s approval could rise — I don’t think any of them are likely but we can’t rule them out.
Trump will always be a dick but he could ‘pivot’ to being less of a dick (e.g handing over control of his Twitter account a la the closing fortnight of 2016’s campaign). On policy, a well executed centrist ‘pivot’ would boost Trump’s approval if it turns out his unpopularity is due to ideology more than personality.
A surge in Jihadist attacks on European cities like that of 2016 might see Trump’s xenophobia play better in 2020 than 2018, especially if one of the attacks hit the U.S. Lefter field, perhaps Trump’s chooses a new Vice Presidential candidate and boosts his campaign contra political scientists’ conventional wisdom about the importance of V.P picks (btw Trump is in a big enough hole that ‘pot stirring’ moves such a new V.P are probably good. (Whether Trump should piss off Mike Pence when it’s widely assumed Presidents can’t self-pardon is another matter)).
Using a proxy such as unemployment, which mean-reverts slower than Consumer Confidence would paint a rosier picture of the economy in November 2020.
Even with a 40% approval rating, Trump could win re-election if Democrats nominate a similarly unpopular candidate. The obvious analogue is 2016 but we also see echos in 1980: two weeks before the election President Carter, despite approval in the mid 30s, held a narrow polling lead over challenger Ronald Reagan.
The Gipper’s better-than-expected performance in the campaign’s only debate assured voters enough for Reagan to win a landslide. But mid-October’s polling is enough to convince me Trump win oftener than the model’s suggested 1% with a 40% approval (he’d definitely still be a substantial underdog).
Trump’s path to re-election runs through a vertiginous act of funambulism, in stormy conditions. He needs the economy to hold up, Mueller to turn up empty, examination of his taxes to blank (or not get off the ground), no other investigation to hit a home run, to competently handle major geopolitical crises (or face none) and then hope the Democrats don’t nominate a strong candidate (I don’t he’ll ever be liked enough to beat someone as popular as Obama — Democrats nominating a mediocre/bad candidate is almost a necessary condition).
I suspect media reluctance to point out Trump’s long odds comes from getting their fingers burnt in 2016. But Trump was a massive underdog. He almost certainly would have lost, heavily, facing any plausible Democratic nominee other than Hillary Clinton. Likewise had the FBI leaked what they knew about Trump’s Russia connections before the election. Or if FBI’s director hadn’t written that letter. He hit the jackpot, and it’s easy to think he’ll do so again. But as (good) gamblers know, luck has no memory. It’s unlikely he pulls off the same trick twice.
