What an Approval Rating Actually Measures

A presidential approval rating sounds like a verdict. It is treated as one in the press and discussed as one in editorial pages. The number goes up, the president is doing well; the number goes down, the president is in trouble. That reading is intuitive, and it is mostly wrong about what the number actually is.

Approval ratings measure a specific thing, in a specific way, with a specific set of caveats. Once you understand what the question actually asks and what shapes the answer, the readings become more useful and less dramatic. A lot of what looks like a presidency rising or falling is just an instrument doing what instruments do.

The question almost never changes

Most major approval polls use a version of the same wording: "Do you approve or disapprove of the way [the president] is handling his job as president?" Pollsters use this near-identical question precisely so the answer can be compared across time, across presidents, and across pollsters. The stability of the wording is the whole point.

It also means the answer is constrained in a particular way. The respondent is being asked about overall job performance, not about a specific decision, not about whether they like the president personally, not about whether they will vote for him. Some respondents conflate those things and answer for one when asked the other. Pollsters know this and treat the noise as part of the measurement.

When you see approval rating coverage that treats a one-week change as meaningful, what is usually happening is that the noise floor is being reported as a signal. Single-week swings of one to two points in either direction are within sampling error and within the day-to-day variation of who answers a phone call.

What most of the number actually is

On any given day, a sitting president’s approval rating is mostly two things added together: a partisan baseline, and a sentiment overlay.

The partisan baseline is the share of the country that will approve of any president from their party and disapprove of any president from the other party, almost regardless of what happens. For modern presidents that floor and ceiling are both relatively close. A Democratic president can usually count on something like eighty to ninety percent approval from self-identified Democrats and around five to fifteen percent approval from self-identified Republicans. The reverse holds for Republican presidents. That partisan structure now accounts for the great majority of the rating, before anything specific to the president has been considered.

The sentiment overlay is everything else — the small share of voters who genuinely shift their answer based on conditions. They tend to be the same kinds of voters who are described as independents or persuadable, and they respond mostly to two things: the economy as they personally experience it, and major news events that override their normal political filter. Wars, recessions, scandals, and crises move the overlay. Most other news barely touches it.

Once you accept that breakdown, modern approval ratings start to make sense. A president pinned in the low forties is not usually being judged on specific decisions; he is mostly hitting his floor, with little upward pull from the overlay. A president temporarily in the high fifties has caught a sentiment wave that is almost always temporary.

Why the band has narrowed

Approval ratings used to swing further than they do now. Eisenhower hit the seventies. Both Bushes briefly cleared eighty after major events. Carter and Truman dropped into the twenties. Those swings happened because the partisan baseline was lower — fewer voters answered automatically along party lines, and more voters were genuinely open to changing their evaluation.

That has changed. Partisan identity now structures opinion on more questions than it used to, including questions like "is the economy good," which were once asked answered partly on the basis of personal experience and are now answered largely on the basis of party. The result is that approval ratings have a much shorter elastic range. Modern presidents almost never break sixty for any sustained period, and almost never drop below thirty-five even at their worst.

You can see this just by overlaying the ratings curves of recent presidents. The lines look different in their middle stretches, but they nearly converge at their floors and ceilings, because those floors and ceilings are now being set by partisan baseline rather than performance.

Approval is not predictive of reelection — most of the time

There is a common assumption that approval ratings predict elections, and a softer version of that assumption is roughly correct: a sitting president with sustained approval above fifty is almost always reelected, and one with sustained approval below forty is almost always not. Between those bounds, approval is a weak predictor.

The reason is that elections are choices between two candidates, and approval is a unary measurement of one. A president with forty-four percent approval can still win reelection comfortably if the alternative is unpopular for other reasons. A president with forty-eight percent approval can still lose if the challenger’s coalition turns out at higher rates. Approval is one input among several, and in close elections it usually is not the deciding one.

There is also a timing issue. The approval rating that correlates best with reelection is the one taken in late summer or early fall of the election year, after both parties have nominated their candidates and after the economy has produced most of the data the electorate will see. Approval ratings a year or two before an election are interesting context but not much more.

How event response actually shows up

When something big happens, approval usually moves in a recognizable shape. There is an initial swing of a few points in one direction — up for unifying events, down for scandals — followed by a slower return toward the previous baseline over several weeks. The size of the initial swing is mostly proportional to how surprising the event was, not how important. Long-running scandals barely move the number because the partisan filter has already absorbed them. Sudden events move it more because the filter has not had time to engage.

The return to baseline is the part that surprises people. Even severe drops tend to claw back most of the way over a month or two, as the news cycle moves on and partisan respondents reabsorb the news into their pre-existing frame. The exception is when an event genuinely shifts a respondent’s underlying assessment of the president — that is rare, and you can usually only see it in retrospect.

Reporters covering an event-driven move in approval often treat the initial swing as the new normal. It almost never is.

A useful way to read the number

Two habits make approval coverage more informative. First, look at the rolling average over four to six weeks rather than any single poll. The average smooths sampling variation and lets the slope of the line emerge. The slope tells you something the spot value does not.

Second, watch the gap between the president’s approval and the share of respondents identifying with his party. That gap is the closest available proxy for the sentiment overlay — the part that is actually moving. A president whose approval is running five points below his party’s share has some persuadable voters disengaging. A president whose approval is running three points above is gaining sentiment beyond his floor. The gap usually tells you more than the number itself.

Done with those two habits, an approval rating becomes what it is supposed to be: a useful, stable, fairly modest instrument for tracking the state of public opinion about a specific person doing a specific job. Not a verdict, not a forecast, not the headline it gets played as. Just the measurement, doing its job.