Ever since Inauguration Day, the public’s faith in the president has been eroding. According to the latest Gallup surveys, only about 38 percent of Americans approve of Trump’s performance, which is down roughly 5 points from February. Vox’s own polling, conducted in partnership with SurveyMonkey, shows a similar drop — from 46 percent at the outset of his term to 41 percent in July.

For the president’s critics, these slipping approval numbers seem like vindication. They show that Americans aren’t blind to the disorder in the White House — that at least some Trump supporters are second-guessing their president.

But what does this mean in practical terms? American politics, by design, has never perfectly followed public opinion. After all, Trump never had a majority of Americans rooting for him. He won the Electoral College but lost the popular vote. What’s important isn’t his popularity nationwide but his approval rate in key parts of the country.

Democrats, for one, hope that the growing disenchantment with the president will translate into congressional victories come 2018. But according to an analysis of more than 50,000 respondents of Vox/SurveyMonkey polling, the prospects of a midterm Democratic surge still look shaky.

Most of the decline in Trump approval, it turns out, occurred in congressional districts that are already solidly red or solidly blue. In the closest Republican districts, opinions of the president haven’t budged at all over the past six months. Overall, Americans may be growing weary of Trump’s chaotic administration — but it seems the Americans who will matter the most politically next year have yet to change their minds.

Close Republican districts are an anomaly

Over the past six months, Vox and SurveyMonkey have been polling Americans about their opinions on the president and the economy. By grouping people according to where they live, we can get a sense of Trump approval in different regions. The nation has 435 voting congressional districts, but the most important ones are those that occupy the middle of the political spectrum. These are the swing districts that determine the relative clout of the parties in the House.

In solid Democratic districts — places where the Democratic House candidate won overwhelmingly in 2016 — Trump approval slipped from an average of 32 percent in February through April to 29 percent in May through July. Likewise, in solid Republican districts, Trump approval fell from 56 percent to 53 percent. Though these are statistically significant declines, they don’t mean much politically, because these districts will practically never be in play.

In the 50 closest congressional districts on the Republican side, though, Trump’s approval rating has held steady at 47 percent. Meanwhile, among the 50 closest districts on the Democratic side, Trump approval seems to have declined from 42 percent to 39 percent (though it’s still too close to call).

This pattern becomes clearer when we focus on people who say they “strongly” approve of the president. In close Democratic districts, that percentage fell noticeably over the past six months, from 23 percent to 20 percent. But in close Republican districts, enthusiasm for the president hasn’t changed: 26 percent of people continue to strongly believe that Trump is doing a good job.

In other words, the polling suggests that Democrats might be tightening their grip on close districts they won in 2016, but they haven’t made any progress on the districts they hope to flip in 2018 — at least not if Trump’s approval is any indicator.

Most Republicans turning against Trump live in Democratic districts

Close congressional districts are close because they contain an even mix of people from both sides of the aisle. Opinions about the president are extremely polarized by political affiliation — among Republicans and right-leaning independents, the Trump approval rate remains well over 80 percent. Among Democrats and left-leaning independents, the approval rate is less than 10 percent.

Even Republicans are starting to change their minds, however. In February, 90 percent of right-leaning Americans approved of Trump. By July, that number had fallen to 86 percent.

But it seems that most of those Republicans turning against the president hail from Democratic congressional districts: In safe Democratic districts, the Republican approval rate has fallen from 86 percent to 82 percent. And in close Democratic districts, it’s fallen from 89 percent to 84 percent. These are statistically significant changes.

Here’s the bad news for Democrats: Republicans in close Republican districts remain upbeat about Trump. There, approval rates appear to have dipped from 91 to 89 percent, but it’s a difference that is too small to be statistically significant.

Again, the phenomenon is easier to see when we look at those whose are especially enthusiastic about the president. In close Democratic districts, the percentage of Republicans who say they “strongly” approve of the president fell dramatically, from 59 percent in the first three months of his term to 51 percent in the most recent three months.

But among Republicans living in close Republican districts, the rate of enthusiastic Trump supporters has held at around 56 percent. That contrasts with what’s happening in solid Republican districts, where the rate of Republicans who strongly approve of the president has fallen from 61 percent to 56 percent.

It appears that in congressional districts held by Republicans, there remains a core group of right-leaning voters who don’t seem concerned about the messy dramas emanating from the White House.

The pattern is the same when we narrow the focus to the closest 25 congressional districts on either side. (The pattern also persists when we look at close counties, which rules out gerrymandering as an explanation, since county lines are not regularly redrawn for electoral reasons.)

So what’s happening here?

It’s important to point out that the data set we’re using is massive, but not without its caveats. Our 50,000 respondents come from online surveys, and while this a nationally representative sample, it might not precisely capture the opinions of those who shun the web. Furthermore, our results end in the first week of July (our August survey is currently in the field), so they don’t reflect the latest shifts in public opinion. A lot happened in July — the Don Trump Jr. emails, the failed Senate health care bill, the brief, wondrous White House career of Anthony Scaramucci — and other polling suggests that these controversies have caused Trump approval to slip further in recent weeks.

What’s safe to say is that Trump approval has remained surprisingly resilient in the places that matter most to Democrats right now: the districts they hope to win in 2018.

Our polling data also casts a different light on the special House elections of the past few months, where Democrats have made startling gains. In the high-profile race for Georgia’s Sixth, for instance, Democratic challenger Jon Ossoff nearly defeated Republican Karen Handel, even though the district went Republican by a 62-38 margin in 2016. (For reference, GA-6 doesn’t even count as one of the 50 closest Republican districts.)

Many pundits took the close Ossoff-Handel race as a sign that backlash against Trump would trigger a Democratic surge in 2018. Overall, though, our surveys suggest that Democrats will have a tough time in vulnerable Republican districts, where opinions about Trump have been slow to change — particularly among the Republicans and swing voters that Democrats will need to rely on to win seats.

All of this seems to be another sign that Americans are divided, not only by ideology but by the information they consume and the company they keep. Trump approval might be declining across the nation, but not so much in the Republican enclaves that Democrats desperately hope to flip.

Sean Spicer, the embattled and increasingly invisible White House press secretary, resigned on Friday morning shortly after the president offered Wall Street financier Anthony Scaramucci the job of White House communications director. The New York Times, which first reported Spicer’s resignation, writes that Spicer “vehemently disagreed” with Scaramucci’s appointment as his new boss, and that he quit in protest.

These developments appear to be the first steps in a long-promised communications shakeup at the White House, which has struggled to stay focused amid the unfolding Russia collusion story.

Scaramucci is a polarizing figure among the warring factions in Trump’s administration. According to Axios, White House Chief of Staff Reince Priebus and chief strategist Steve Bannon opposed the appointment, while Ivanka Trump and Jared Kushner supported it. As communications director, Scaramucci would be stepping into a position that has been vacant for several weeks since the resignation of Mike Dubke, a Spicer ally, in May.

One of Scaramucci’s first tasks, it seems, will be to search for a new White House press secretary. In his six months on the job, Sean Spicer developed a reputation as a fierce Trump defender, willing to sacrifice his own credibility in service of the president’s narratives. Spicer’s rocky relationship with the press became a hindrance, but what really seemed to diminish his stature at the White House were his series of gaffes that turned him into a talk-show punchline and internet meme.

Spicer’s departure had been gossiped about for months, with rumors that the president had considered promoting him to a position out of the heat of the cameras. In the meantime, Spicer has spent less and less time in the public eye. His deputy Sarah Huckabee Sanders led most of the daily briefings with the White House press corps in recent weeks.

A few years ago, Seattle lawmakers embarked on a bold experiment in public policy. Between 2014 and 2017, the city hiked its minimum wage from $9.47 to $15 an hour — a plan so aggressive that economists weren’t sure what would happen. Could employers afford the spike in labor costs? Or would this drive low-wage jobs out of Seattle?

On Monday, researchers at the University of Washington announced that Seattle’s efforts may indeed have backfired. According to their calculations, when the minimum wage reached $13 an hour in 2016, Seattle employers substantially cut back on low-paying jobs. Even though average pay went up for people at the bottom, there was less work, so they became worse off overall. Between job losses and reductions in hours, Seattle new minimum wage policy may have cost low-wage workers about $125 per month each, the researchers say.

Though the paper has not yet been peer-reviewed, it is already the focus of a fierce debate because it seems to contradict more than two decades of economics research showing that the benefits of small to moderate increases in the minimum wage largely outweigh the costs. Just last week, a competing study from economists at the University of California Berkeley found that the restaurant industry in Seattle — a big source of low-wage jobs — did not shrink its labor force despite raising wages to comply with the new law.

But the University of Washington study differs from previous studies in a significant way: It takes advantage of a much richer source of data. The researchers had access to state records for individual employees, which included how much time they worked and how much they were paid. Because this kind of information has been hard to come by in the past, much of what we know about the minimum wage has relied on research about specific low-wage sectors (like food service) or specific groups of lowly paid people (like teens).

In contrast, the University of Washington researchers were able to track low-wage workers across all parts of the economy, which may explain why they found different results.

“There aren’t a lot of good data resources that will tell you who’s working what wage,” said Jacob Vigdor, a professor of public policy at the University of Washington and a co-author on the paper. “Previous studies have tried to work around this limitation by focusing on the restaurant industry, focusing on teenage employment, with the hypothesis that most teenagers are in low-wage jobs, that most restaurant workers are in low-wage jobs.”

“Here in Washington state, we have data that tell us how everybody’s pay breaks down,” he continued. “The data that we’re using is pretty rare.”

The University of Washington study has significant limitations of its own. The detailed data set it uses has two major blind spots. First, it excludes independent contractors — like Uber and Lyft drivers, who started working in Seattle around 2014. Even though independent contractors aren’t covered by the city’s minimum wage law, that kind of gig economy work may have served as hidden safety net for low-wage workers squeezed out of Seattle’s regular economy.

Second, the study excludes many people employed by larger businesses with multiple locations in Washington state, because researchers could not figure out which of those employees were based in Seattle. For this reason, about 38 percent of Washington state workers had to be left out of the study, possibly skewing the calculations.

The study could revolutionize minimum wage research

It is of course dangerous to overgeneralize from a single paper about a single city, particularly a paper that has not passed the full gauntlet of peer scrutiny. But the University of Washington study is an unusually meticulous look at an issue that has come to the fore in recent years with the Fight for $15 and other campaigns to establish a living wage. Seattle is one of many jurisdictions, including New York, Chicago, and California, that have led the nation in raising their minimum wage to significant new heights.

Some have interpreted the University of Washington paper as evidence that these efforts have gone too far. Though there are sound theoretical reasons why a modest hike in the minimum wage might not depress employment, most economists agree that if you set the minimum wage high enough, at some point you will start destroying jobs. It may be that Seattle has finally reached that breaking point.

But then again, prevailing wages in Seattle were high to begin. A $13 minimum wage may seem exorbitant compared with the federal minimum wage of $7.50, but most workers in Seattle were already earning more than that. Arindrajit Dube, an economist at the University of Massachusetts Amherst who studies the minimum wage, has pointed out that Seattle’s new minimum wage is not all that lavish when you take into account the city’s high overall wages. In fact, in relative terms, it seems to be roughly as generous as the federal minimum wage in 1968, back before the value of the federal minimum wage was eroded by inflation.

The far more controversial interpretation of the University of Washington study is that it might overturn — or at least throw into doubt — much of the past two decades of research on the minimum wage. The study seems to highlight a limitation with minimum wage studies that focus on industries dominated by low-wage workers — which is a popular methodology.

When the University of Washington researchers used their data to look at only the restaurant industry, they arrived at results that lined up with previous work: Minimum wage hikes did not noticeably affect overall employment in food services in Seattle. This also agrees with the finding from the Berkeley study, which used methods similar to past research on the minimum wage.

But the UW researchers had the advantage of being able to delve deeper. Instead of looking at the entire food services industry, they could focus specifically on lower-wage restaurant workers. Among these employees, the minimum wage indeed led to lost jobs or cutbacks in hours, the researchers found. They suggest that past research may have missed this effect because it could not distinguish between higher-paid and lower-paid restaurant workers.

This finding will continue to be debated in the coming months, but for now, despite its flaws, it is still one of the most persuasive pieces of evidence yet for the view that minimum wage policies hurt workers at the bottom of the economy. This, of course, was long the prevailing view among economists until empirical research in the 1990s showed that minimum wage hikes didn’t seem to have much of a measurable effect on employment — or even if some jobs disappeared, workers still benefited overall.

David Card, an economics professor at Berkeley, authored one of those papers, a famous study with Princeton economist Alan Krueger showing that New Jersey’s fast-food industry did not shed jobs when the state increased its minimum wage in 1992. Card and Krueger’s paper has long been cited as the seminal piece of evidence in the recent consensus that the minimum wage might be good for workers.

Now, with more detailed data, the picture is getting more complicated.

Card, who was traveling, said in an email that he had not yet read the paper, but cautioned overinterpreting this single study. “My guess is it will take some time to get this straightened out by professionals and that in the end the media will interpret it as an ‘unresolved dispute,’” he said.

Nothing’s official yet, but it seems the president will finally sideline White House press secretary Sean Spicer, the stormy spokesperson who often endured ridicule for blurring the lines between truth, fiction, and wishful thinking.

In typical fashion, the news so far is based on leaks. The White House has not made an announcement, but Politico and Bloomberg each reported yesterday that Trump is searching for candidates to replace Spicer at the podium, according to two unnamed administration sources. Spicer himself is expected to stay on, perhaps in a more senior — but less visible — position.

For weeks now, the president has reportedly expressed disappointment in his communications team, particularly for how it handled his firing of FBI Director James Comey, which resulted in more backlash than was expected. Communications director Mike Dubke stepped down in May, and many rumored that Spicer would be next, as soon as the White House found a replacement.

In the past Trump has praised Spicer’s combative daily press briefings, which have become a spectacle of daytime television. According to the Washington Post, Trump remarked in March, “I’m not firing Sean Spicer. That guy gets great ratings. Everyone tunes in.”

But in recent weeks, particularly as the Russia scandal heated up, Spicer has played a diminished role in public. The press briefings have been shorter, and have more frequently taken place off-camera. Yesterday, outlets were barred from even broadcasting the audio.

These tactics signal that the White House is anxious to take attention away from the briefings, which have turned into a ritual of resistance and stonewalling. Reporters ask tough questions, and Spicer deflects, or tells an outright lie. These daily encounters are no longer informative, and as theater, are no longer flattering for the administration.

As I wrote a few weeks ago: 

In a different era, the White House press briefings were low-key and rarely televised live. They were a chance for the press secretary to deliver updates, clear up day-to-day misconceptions, and to take the temperature of the nation’s top reporters. Under the Trump administration, the briefings have become belligerent, blustery, and increasingly divorced from reality. 

Spicer has resisted answering even simple questions. Asked repeatedly, on different days, whether the president believes in climate change, Spicer responded with the same excuse: “I have not had the opportunity to specifically talk to the president about that.”

Trump himself has questioned the value of the press briefings. “As a very active President with lots of things happening, it is not possible for my surrogates to stand at podium with perfect accuracy!” he said on Twitter in May. “Maybe the best thing to do would be to cancel all future ‘press briefings’ and hand out written responses for the sake of accuracy???”

Meanwhile, reporters are getting their real facts from others inside the White House. This is perhaps the leakiest administration ever, with newspapers regularly publishing scoops involving dozens of anonymous White House officials. Trump has complained about the press using unnamed sources, but the problem is internal. That administration officials are so eager to leak to the press speaks to an unprecedented level of mismanagement and mistrust in the White House.

Case in point: As of Tuesday morning, there is still no official word about Spicer’s status, or the candidates lined up to replace him as press secretary. But the gossip is already out.