Skip to main content
Menu
How We’re Tracking Joe Biden’s Approval Rating

If it’s possible to sum up a presidency in a single number, that number would be the president’s approval rating — or the share of Americans who approve of the job he’s doing. Arguably, that simple percentage can determine the fate of an entire presidency.

For instance, a high approval rating can marshal support for a president’s agenda and minimize his party’s losses in the midterm elections — not to mention help the president himself win reelection. But a low approval rating can be electoral poison and imply that a president has lost the mandate to govern entirely.


Partisans don’t just disagree, they hate one another | FiveThirtyEight

This, in a nutshell, is why we at FiveThirtyEight track the president’s approval rating (and its glass-half-empty cousin, disapproval rating) in real time — first for former President Donald Trump, and now for President Biden. According to our average of all the Biden-approval polls we have so far, Biden starts his administration with a 53.9 percent approval rating and a 35.1 percent disapproval rating.1 (If you look at only polls of likely or registered voters — which you can do using the dropdown menu in the top right of the interactive — the numbers are similar: 54.4 percent approval, 36.0 percent disapproval. Same with polls specifically of all adults: 53.2 percent approval, 31.6 percent disapproval.)

These are strong approval numbers compared with what we’ve grown accustomed to seeing over the past four years. But they also probably won’t stay that high, since presidents typically experience a “honeymoon” period of inflated popularity during their first few months in office. As my colleague Geoffrey Skelley wrote on Tuesday, some experts believe that political polarization has made presidential honeymoons a thing of the past, but, for now at least, Biden appears to be enjoying one: He has a +18.9-point net approval rating (approval rating minus disapproval rating) after winning the election by only 4.5 points.

[Related: An Updating Calculation Of President Joe Biden’s Approval Rating]

That said, that’s a less impressive honeymoon than most past presidents have enjoyed, suggesting that partisanship is taking its toll. For reference, at this point in their presidencies, Bill Clinton had a +36.3-point net approval rating, George W. Bush had a +32.0-point net approval rating and Barack Obama had a +39.3-point net approval rating. Trump is the only president to have started his administration with a lower net approval rating than Biden: +2.0 points on Jan. 27, 2017.2

Although Biden’s approval rating is somewhat on par with Clinton’s, Bush’s and Obama’s, his disapproval rating is much, much higher, reflecting the built-in animosity that many Americans already have for him. (You can compare Biden’s approval rating, disapproval rating and net approval rating to past presidents all the way back to Harry Truman by scrolling down to the bottom of our interactive.)

How do we boil down dozens of approval-rating polls into a single number? It’s not a simple average! We use an empirically tested, weighted average that accounts for poll quality and uncertainty. It’s the same methodology we used to calculate Trump’s approval-rating average and similar to the approach we take in our election forecasts and other polling averages; here’s a more detailed explanation.


Is Biden’s vision of unity possible? | FiveThirtyEight Politics Podcast

First, our indefatigable team of poll researchers collects every national poll of the president’s approval rating; we don’t disregard any legitimately conducted3 scientific polls (this is because we don’t want to be in the position of making subjective judgments about how “good” a poll needs to be in order to be worth including), although we do assign them different weights (more on that in a moment). You can see these individual polls listed just below the main graph on the approval-rating page and download them via the link at the very bottom of the page.

(One quick housekeeping note here: Sometimes, pollsters release numbers for the president’s approval rating among different populations — for example, all adults vs. registered voters vs. likely voters. In this case, we default to the result that represents the broadest swath of people — so adults over registered voters, registered voters over likely voters. However, as previously mentioned, we also have versions of the average that compute the president’s approval rating among only polls of adults or only polls of voters.)

Next, we determine how much weight to give each poll in our average. First, polls conducted by pollsters with higher FiveThirtyEight pollster ratings — a letter grade measuring how accurate and methodologically sound pollsters are4 — are given more weight. Second, polls with bigger sample sizes also count for more. Finally, we downweight polls by pollsters that survey approval rating very frequently (i.e., more than once per 20 days) so that no one pollster is exerting too much influence on the average.5 Each poll’s pollster rating, sample size and ultimate weight in our average are displayed next to it in our list.

[Confidence Interval: QAnon Is Not Going Anywhere]

From these weighted averages, we then calculate a trend line of the president’s approval and disapproval ratings over time using local polynomial regression — basically, drawing a smooth curve over the individual data points. (But not too smooth — you don’t want the average to be unresponsive to movement in the polls. We choose our smoothness settings6 based on what has historically best predicted a president’s future approval and disapproval ratings in polls since 1945, which visually turns out to not look very smooth at all.)

But wait! That first trend line we calculate isn’t the one you see on the page. Instead, we use the initial trend line to see if a given pollster’s polls are consistently better or worse for Biden than the weighted average — in other words, if the pollster has a “house effect.” Polls from pollsters that systematically over- or underestimate Biden are then adjusted in order to remove this house effect.7 For example, Republican-aligned pollster Rasmussen Reports has an anti-Biden (and had a pro-Trump) house effect that must be taken into account when judging its polls. Accordingly, our model adjusted its recent poll that gave Biden a 48 percent approval rating and 48 percent disapproval rating to 50 percent approval and 43 percent disapproval.8 Each poll’s adjusted approval and disapproval ratings are displayed in the rightmost column of the approval tracker’s list of polls, just after the poll’s raw, unadjusted numbers.

From there, we just rinse and repeat: The adjusted polling numbers are used to calculate a new trend line, which is used to calculate new adjusted polling numbers, which is used to calculate another new trend line — and so on. The final result, once the cycle is complete, is the main approval-and-disapproval-rating graph you see in our interactive. And you can use that graph to check not only Biden’s average approval and disapproval ratings as they stand today, but also what they were at the end of any day during his administration.9 (Note that those daily ratings are based on polls released by that date, not necessarily polls conducted by that date; we don’t go back and recalculate the average for past days once more data becomes available.)

You can also download Biden’s average approval ratings for every day of his term, as well as every poll that goes into the calculation, by clicking the appropriate links at the very bottom of the page. This document also includes upper- and lower-bound estimates for Biden’s approval and disapproval ratings,10 which are depicted in the interactive by the shaded green and orange areas around the main trend lines. This represents the fact that there is uncertainty in our approval-rating average: Both polls and our average have a margin of error.

[Biden’s Initial Batch Of Executive Actions Is Popular]

We calculate this uncertainty by measuring how well our approval-rating estimates for past presidents (all the way back to Truman) have predicted future polls of their approval rating. Things that make the confidence intervals wider (i.e., things that make us less certain) include a dearth of polling, a high level of disagreement in the polls we do have, and a lot of volatility in a president’s approval rating over the long term. Things that make the confidence intervals tighter include lots of polls, highly consistent polls and a very stable long-term average.

We set the width of our confidence intervals such that 90 percent of future polls should fall within that range. And we even offer a tentative forecast for which direction the averages will trend; toggle the switch that says “Today” to “4 years” at the lower right of the graph in order to see it.

Because approval ratings have historically tended to revert to the mean, and also to deteriorate slightly over the course of a president’s term, we expect Biden’s approval rating to decline and his disapproval rating to rise (as represented by the dotted lines on the graph). But as you can see, the 90-percent confidence interval for both approval and disapproval gets much wider the further you go into the future, meaning a wide range of outcomes are possible for Biden’s long-term popularity. Even in this age of intense polarization, circumstances and actions can still affect the president’s approval rating, so Biden’s political future is at least partly in his own hands.

[In 2020, Some Places Snapped Back To The Left, Others Shifted Further To The Right]

So that sounds like a pretty good reason, if we do say so ourselves, to bookmark our Biden approval tracker and check back on it often. And if you have any questions about our methodology, comments about the interactive, or missing polls we need to add, don’t hesitate to drop us a line at polls@fivethirtyeight.com.


‘Biden could have more of a wind at his back than you might think’: Silver


Confidence Interval: QAnon is not going anywhere | FiveThirtyEight

Footnotes

  1. All numbers in this article are as of 5:45 p.m. Eastern on Jan. 27.

  2. Many presidents did not have enough polls early in their administrations to calculate an approval-rating average for them on day eight, but a direct comparison will be possible for those presidents eventually.

  3. However, as always, we do ignore polls we know or suspect are fabricated from whole cloth.

  4. Our pollster ratings are based on how accurate their polls within the final 21 days of an election have been in elections since 1998, and also whether the pollster’s methodology meets the FiveThirtyEight “gold standard.” For a nitty-gritty rundown of how we calculate pollster ratings, see this article. In addition, note that our current pollster ratings were last updated in May 2020; we plan to update them with the polls and results of the 2020 general election later this year.

  5. This includes daily or weekly tracking polls, which we give special treatment to ensure that our average does not double-count individual interviews.

  6. For those interested in the gory details: The model uses three second-degree (quadratic) smoothers, with bandwidths of 10, 20 and 30 days, and averages these three estimates together. The bandwidth reflects the number of days used to calculate the polynomial: For instance, a 20-day bandwith applied on Feb. 27 would mean that polls from Feb. 7 to Feb. 27 are used in the calculation. The model also requires that a minimum of five polls be used in the calculation, so it will lengthen the time intervals under consideration in the event of sparse data.

  7. The key words here are “consistently” and “systematically.” We won’t adjust a poll too much if its pollster has only surveyed Biden’s approval rating once and the result was an outlier; that could be due to sampling error or some other benign explanation. But if a pollster keeps over- or undershooting the average in poll after poll, the house-effects adjustment will be bigger.

  8. House effects are calculated separately for approval ratings and disapproval ratings, and therefore they are adjusted separately as well.

  9. Except the first three, when not enough polls had been released yet.

  10. “Approve_hi,” “approve_lo,” “disapprove_hi” and “disapprove_lo” in the CSV.

Nathaniel Rakich is a senior elections analyst at FiveThirtyEight.

Comments