Just added to my radio archive (click on date for link):
November 25, 2021 Alex Vitale, just out with an updated edition of The End of Policing, on what cops really do and how we can get rid of them • Barry Eichengreen, co-author of In Defense of Public Debt, on the very long history of public borrowing
There are certain things that people say that sound so true that others repeat them credulously without feeling the need to cite evidence. Two covid-era favorites: everybody’s working from home (WFH). And people have decamped en masse for the hinterlands, thanks to WFH. Neither is really true.
I wrote about the slim WFH numbers in September. In July, which was the most recent month available then, 13.2% of the employed were teleworking, the Bureau of Labor Statistics’ favored term. In October, that had fallen to 11.6% (graph below). Their ranks were still dominated by highly credentialed professional and managerial workers. The miserably paid couriers who brought (and still bring) them food and other essentials were most certainly not working from home, though they easily fall out of some people’s conception of “everybody.”
Even if “everybody” wasn’t teleworking, lots were. But there’s no support at all for the mass decampment story. Freshly released data from the Census Bureau shows just 8.4% of Americans moved between 2020 and 2021, the smallest share since the Bureau started counting in 1948. That was down almost a full percentage point from the previous pair of years, a large decline by recent historical standards. But as the graph below shows, the share has been declining steadily since 1985’s 20.2%, when high oil prices and deindustrialization drove movers south and west in search of jobs. (Unfortunately for the energy migrants, oil prices collapsed in 1986.)
Also not surviving a confrontation with Census data: the claim that, freed from the need to go to an office, covid refugees headed for the hills (and beaches) in large numbers. They didn’t, as the next graph shows.
Yes, people have been moving out of cities for decades, but the net out-migration of 0.5% of the national population was smaller than the two previous years, and over a quarter less than the 1986–1999 average. (Reminder: these numbers count only in- and out-migration, not changes in population levels, which are also affected by births and deaths.) As the trendline shows, urban out-migration has been in a slowing trend since this data series began. Suburbs gained migrants, but at a rate slightly lower than the previous eight years. Metro areas—the combination of suburban areas and the “principal” cities associated with them— have seen steady, if declining, in-migration for all of the last 35 years. And nonmetro areas, supposedly the recipients of all those terrified urban refugees, lost migrants between 2020 and 2021—more, in fact, than any year since 2014.* That trendline ambles very lazily downward, though it’s hard to distinguish from the axis, because it’s so close to 0.
A larger question here, pandemic aside, is what happened to American mobility? Pulling up stakes and moving a long distance in search of fresh opportunity used to be a foundational national myth. Like many others, it needs to be retired.
* The Census Bureau reminds us that nonmetropolian areas are not the same as “rural” areas, since there are many rural-feeling regions that are nonetheless technically part of metro areas. Census has a point, but on the other hand, there are social differences between thinly populated areas that are close to major conurbations and those that aren’t. For example, parts of Westchester County, just north of New York City, may look rural, but it’s amenity-plentiful rich people’s country, and a short train ride away from the big city.
As marvelous as it would be to see a revival of labor militancy, people got a little ahead of things calling last month “Striketober.” According to Bureau of Labor Statistics (BLS) stats, it was a blip by historical standards.
Here’s a graph of the number of workers involved in strikes or lockouts (the BLS counts them together) since 2000. There were 57 months with higher numbers of workers off the job. At the high point of this graph, May 2018, there were over fourteen times as many workers on strike as there were last month.
Here’s another measurement—what the BLS, in nice Victorian fashion, calls “days of idleness” as a percent of total days worked throughout the economy. It was 0.01% in October, a level that’s been matched in 39 other months since January 2000. And as the bottom graph shows, back in the old days when strikes were frequent, lost workdays were many times 0.01%. Before 1980, the low was 0.07%, set in 1957. From 1948 to 1979, it averaged 0.16%. In 1959, just two years after the pre-neoliberal era low, it was 0.43%, the series high.
People criticize the BLS data because it covers only large strikes, those involving over 1,000 workers or more. The government used to publish data on smaller strikes, but it looks to have disappeared from the Federal Mediation and Conciliation Service’s website after not having been updated for years (and it was always in a very user-unfriendly format). But when I looked at what data there was in 2018, it showed the same pattern of decline as the large strike data.
There’s certainly some promising labor agitation going on, like efforts to organize Amazon and Starbucks, and there’s plenty of atomized discontent floating around. But there’s no strike wave yet.
Just added to my radio archive (click on date for link):
November 4, 2021 Sheryll Cashin, author of White Space, Black Hood, on the origins, mechanisms, and effects of residential segregation, mostly by race but also by class • Peter Victor and Robert Pollin debate the virtues of “degrowth” in avoiding climate catastrophe
Just added to my radio archive (click on date for link):
October 28, 2021 Samuel Moyn, co-author of this article, on the reactionary history of the Supreme Court and how to democratize it • Deepak Bhargava, one of the editors of Immigration Matters, on immigration policy, historical, current, and future
Yesterday’s post about how people were finding it harder to pay the bills didn’t get into any demographic detail. Time to do just that.
According to the Census Bureau’s biweekly Household Pulse Survey, as of the two weeks ending October 11, 47.7% of adults were having no difficulties paying their bills, down just over 2 percentage points from May, which was the best period in the covid era. The numbers bounce around some from one survey to the next, no doubt an indeterminate mix of noise and trend, so it’s a good practice to average a couple to smooth things out.
First, bill-paying capacities by demographic group—and not just by the usual categories like race, income, sex, and education. In August, the Bureau started asking questions about gender and sexual orientation in the survey, a nice increase in the richness of the data.
Averaging the two surveys covering September 13 through October 11, you get a pattern like the one graphed below. Almost 48% of all adults reported no difficulties in paying their bills. Cisgender men come in a few points above that, and cis women a few below. Transgender and “none of these” come in well below. Among cis men, almost five times as many have no trouble as report finding “very difficult.” (To keep the number and density of the graphs down, I’m only showing the no trouble figures.) Among cis women, it’s three-and-a-half times. For trans and “none of these,” the numbers are almost equal. Being gay or lesbian puts you about 3 points below average, but being bi puts you 18 below.
Results by race, education, and income hold no surprises. Whites are 7 points above average, and Asians are 6; Hispanics or Latinos, 14 points below; blacks, over 17 points below. (The classification scheme and labels are all the government’s choice, not mine.) Not finishing high school puts you 23 points below; having a bachelor’s or more, 17 points above. Gaps correlate perfectly with income ranges—though it’s striking that over 40% of those with incomes between $75,000 and $99,999 are not comfortably paying their bills—and 13% of those over $200,000.
And now the changes since spring, graphed below. Most demographic groups have been finding it harder to pay bills than it was five or six months ago. From the averages of the May surveys to the September 13–October 11 average, the share of people reporting no difficulties paying their bills fell by 2.2 percentage points. Here the damage looks more widely distributed than you might expect. (Gender and sexual orientation can’t be compared because they weren’t part of the survey in May.) Hispanics or Latinos found it slightly easier, but whites had a bigger negative shift than blacks. Those who didn’t finish high school took a hit of over 5 points—but those with some college or more did a bit worse than high school grads. And changes by income class didn’t follow the usual stairstep pattern—the categories just above the median did worse than those below them (though of course their absolute levels are lower, as the graph above makes clear).
As I said in yesterday’s post, these aren’t enormous shifts, but they’re going in the wrong direction. Despite gains in employment—strong at first, now faltering some—more people look to be struggling.
This all makes you wonder about The Great Resignation—record-high quit rates and a stubborn refusal by many to take any crappy job on offer, which is getting uncommonly close coverage. Can it continue as prices rise and savings are drawn down? Or is something breaking in the governing order? It’s not every day you read a New York Times columnist—even one as good as Farhad Manjoo—interviewing anti-work scholar Kathi Weeks and writing words like these:
I’ve been reading /antiwork for months, and I’ve been surprised to find myself joining in the visceral thrill of seeing people wrest the reins of their lives from the soul-sucking, health-destroying maw of capitalism.
A problem is that capitalism usually makes it very hard to reclaim one’s life from it. Having to pay the bills is usually a work discipline without any appeal. Maybe something is happening though.
Since April 2020, the Census Bureau, in collaboration with several other official statistical agencies. has been conducting a biweekly survey of people’s material well-being called the Household Pulse Survey. It asks questions about employment, income, food availability, mortgage and rent status, health, and the ease of paying bills, among other things. There’s a lot in these surveys, but for now I want to take a look at just a couple things: how hard people are finding it to pay their bills and where the money is coming from.
There have been several “waves” of this survey, and response rates and questions have varied between them, complicating longer-term comparisons. In most of what follows I’m going to look mainly at responses since May 2021. I will say that in the months leading up to May, households found it progressively a little easier to get by than they did last summer. For example, in August and September 2020, about 44% of households said they were having no difficulties paying their bills. That rose to around 50% in May 2021. The share describing it as “very difficult” fell from about 14% to 10%. The share drawing on savings, running up the credit cards, or borrowing from friends and family fell.
That all began turning around in May. Graphed below are the changes in households’ experience paying their bills since then. In May (averaging that month’s surveys), 49.9% of households said they had no difficulty paying their bills. In early October, that fell to 47.7%, a decline of 2.2 points. The share reporting it “a little” difficult rose slightly, and those reporting it “somewhat” difficult rose a bit more—but those reporting it “very” difficult rose from 10.4% to 12.2%. Combine “somewhat” and “very” and it rises from 26.6% to 28.6%. These are not massive changes, but they’re not what you’d expect in a period when employment rose by 2.6 million jobs.
And, as the next graph shows, there are some distressing changes in where the money’s been coming from. There’s been a rise in those reporting “regular, like pre-pandemic,” which is what you’d expect in a period of rising employment. But there’s also been an increasing reliance on borrowing, liquidating assets, and government aid (other than unemployment insurance, which has been deeply cut, and stimulus payments, which are a fading memory). It’s good the government aid is there, but given the cheesiness of the US welfare state, there’s typically not much of it.
These are not enormous changes—a few percentage points (though every percentage point shift represents about 2.5 million adults). But they’re mostly in the wrong direction.
Between February and April 2020, the US economy lost over 22 million jobs, almost 15% of total employment. That was by far the largest job loss since the early years of the Great Depression. Between 1929 and 1932 or 1933 (depending on whose numbers you use, since there are no solid, official stats), 20–25% of jobs disappeared (again, depending on whose numbers you use). Since World War II, however, the worst contraction, the Great Recession of 2008–2009, killed just over 6% of all jobs—a big number, but well short of 15%.
Since April 2020, we’ve recovered just over three-quarters of the jobs lost early last year. That’s a healthy chunk, but it still leaves employment over 3% below February 2020’s level, a loss that that would qualify as a sharp recession in more normal times. The recovery remains highly incomplete.
I haven’t mentioned unemployment as a measure of labor market damage because it tells only a partial story. To be counted as unemployed, you have to be actively looking for work, and many people just aren’t looking, either out of discouragement or disgust with the jobs on offer. The labor force—the sum of the employed and the officially unemployed—is over 3 million, or 2%, smaller than it was before covid hit. While the unemployment rate of 4.8% in September isn’t terrible by historical standards, and is well below April 2020’s peak of 14.8%, the 3+ million who’ve withdrawn from the labor force wouldn’t be properly accounted for. In any case, we’re still a long way from February 2020’s 3.5%.
A better measure than unemployment under these circumstances is the employment/population ratio (EPOP), the share of the adult population employed for pay. It was 61.1% in February 2020, fell almost 10 points to 51.3% two months later, and has since recovered to 58.7% as of September. Again, it’s a substantial but still very incomplete recovery. That 58.7% neighborhood was around where the EPOP was in the depths of the Great Recession.
Of course, the trajectories of job loss and subsequent recovery vary by demographic. The graphs below, known as spiders in the trade, are not as familiar a form as line, bar, or pie graphs, but they’re a very useful way to depict the job loss and recovery of the last year and a half. Let’s start by indexing the level of the EPOP for each demographic group in February 2020 to 100. That number isn’t meaningful in itself—it’s just a way of comparing subsequent levels to that starting point. The outside ring, the light blue (turquoise? this site says it’s “downy”) represents that 100 for all demographic groups. The innermost ring, the violetish one (or “east side” according to that same site), represents the April 2020 lows. And the dark blue, or “wild blue yonder,” represents September 2021, the most recent month available.
It’s no surprise that job losses were very unequally distributed. The EPOP for everyone fell from 61.1% in February 2020 to 51.3% two months later, a decline of 16%. (That’s not graphed. Note that that 16% decline refers to the change from 61.1% to 51.3%, a decline of 16%, which is different from the 9.8 percentage point decline.) For men, it went from 66.8% to 57.2%, or a decline of over 14% (again, the percent change from 66.8 to 57.2, not the 9.6 point decline derived by subtracting 57.2 from 66.8). That loss is visible on the inner ring of the graph on the upper left, where the graph along the “all men” axis hits 86. For women, the decline was steeper—almost 18%, which is graphed as 82 on the next point clockwise on the innermost ring. Hardest hit in the first grouping were Latina women, down over 23%. It’s not shocking that white men were the least hard hit, down just over 13%, but that’s still quite a blow.
Overall, the EPOP has recovered to just over 96% of its February 2020 level, though Latina women, at almost 93%, are the least recovered, and black women, at just over 93%, are the second-least. Latino men, though hit hard by the downdraft, have recovered pretty strongly, to just over 96% of the pre-pandemic high. The weaker showing for black and Latina women reflects in part their heavy representation in government employment, where the recovery has lagged the private sector’s. Overall, however, despite some talk about gender disparities, there’s little difference in the recovery rates overall between men and women.
By education, the initial damage correlates precisely with attainment. Those who didn’t complete high school saw their EPOPs fall by 24%. For those who graduated from high school, the decline was 20%; for those with some college, 15%; and for those with a bachelor’s degree or higher, the decline was just under 9%. (There’s just no basis for the notion, circulated in authoritative-seeming memes, that higher education is of no benefit in the job market.) Recoveries also correlate with the level of educational attainment, though those without high school diplomas (a small group, just 6% of employment) have done slightly better than those with. College grads have regained almost 98% of their losses. Peter Thiel aside, credentials do pay.
By age, teens got hit hardest of all, losing almost a third of their EPOP, but they’ve since recovered all their losses. So-called “prime-age” workers, those aged 25–54, got hit least hard. The oldest cohort, 65 and over—another small group, just 7% of employment—were also hit pretty hard, and are the least recovered of the four groups (no doubt many of them have just retired rather than fight their way back into employment).
Loss and recovery rates don’t say anything about levels, and EPOPs very widely by demographic as the next graph shows. The vertical dotted line marks the overall employment rate of 58.7% in September. Men are above it and women are below. The most employed demographic in the race/gender/ethnicity grouping is Latino men, at 75.6%. White men are a distant second, at 67.2%. At 54.0%, white women are the least employed group. By education, the stairstep correlation applies again, with the EPOP rising with each level of attainment.
Why does the recovery remain so incomplete? Many employers, especially in restaurants and retail, went out of business, and workers have no place to return to as the crisis eases. People are still reluctant to go to restaurants and hotels. Hotels were at 61.6% of capacity in September, down from 67.2% in September 2019. Restaurants were operating about 9% below 2019 capacity in September. Retail foot traffic hasn’t fully recovered either. Some employers are still uncertain about the future and unwilling to commit themselves to hiring permanent staff.
And workers remain notoriously reluctant to take any crappy job that comes along. A survey by the Indeed Hiring Lab found just a third of the unemployed actively searching for work, little changed from June. You can’t blame pandemic-expanded unemployment insurance benefits for that reluctance anymore either—they’re now gone everywhere, not just in Republican-run states.
You have to wonder how people are paying the bills. Some aren’t, for sure. But the latest edition of the Federal Reserve’s distributional financial accounts, which look at the financial status of households at various levels of the wealth distribution, found that at the end of the second quarter, they had almost $3.5 trillion in checking accounts and cash, $2.4 trillion more in than they averaged in 2019. The top 1% unsurprisingly accounts for a lot of the increase, but other segments of the distribution are flush as well. Even the bottom 50% was packing some reserves, with an average of $3,744, or 78%, more than their 2019 average. The next 40% has $11,405 more, two-and-a-half times as much. Much of that is no doubt the lingering result of stimmy checks and other covid supports. That’s certainly not enough to fund a life of leisure, but it can sustain a few months’ pickiness—especially if you have an employed spouse or partner (a factor cited by a third of respondents in the Indeed survey).
It looks like the economy is experiencing something like long covid too.
Millions of unfilled job openings, workers quitting en masse, soaring wages (at least in some sectors)—wild time in the job market. Here are some graphs to make the point.
The Bureau of Labor Statistics does a monthly Job Openings and Labor Turnover Survey (JOLTS) that queries employers on unfilled openings, hires, firings, and quits. August’s data was released this morning, October 12.
Here’s a graph of openings, expressed as a percent of current employment, in the private sector as a whole and in accommodations and food services (hotels, casinos, restaurants, and bars, hereafter AFS), the sector that’s grabbed a lot of attention as a site of worker discontent.
Although the openings rate came down in August from July’s all-time high, we’ve never seen anything like it since the survey began in December 2000. (The graph starts in January 2001 for appearance’s sake.) Overall private-sector openings were 6.6% of employment—and at 10.2%, the AFS sector was even more extraordinary than the overall rate.
And a lot of workers are quitting. No downturn in August here: the two graphed quit rates—3.3% overall and 6.8% in AFS—set records by a comfortable margin.
The quit rate got very low during and after the Great Recession, roughly 2008–2010; when the world is falling apart, workers are generally reluctant to leave their jobs. The quit rate is usually read as a sign of worker confidence. Bosses and financiers don’t like it when workers get too cocky—it makes them harder to push around—so central bankers often push the economy into recession in order to readjust their attitudes. The quit rate’s behavior lately, though, seems less about an economic boom making workers cocky and more about them losing patience with shitty jobs (especially in AFS) where supervisors often scream, pay stinks, covid is rampant, and customers threaten to punch you out for asking for proof of vaccination.
Here’s a longer-term perspective on the quit rate. Pre-2001 data are estimated from the share of job leavers among the unemployed (a figure reported every month with the employment release). As the graph shows, it looks like the August record would hold even if you take it back 54 years.
Employers look to be boosting pay in an effort to lure workers back on the job. Here’s a graph of the annual change in average hourly earnings (AHE) for all private sector workers and AFS. There are a lot of gyrations at the right end of the graph that need explanation. When the pandemic first hit, the AFS sector largely shut down, so the wage data covered a fraction of the sector’s usual employees. That strangeness is visible in the dip of the light blue (turquoise?) line into negative territory. More broadly, low wage workers were far more likely to lose their jobs than higher-wage ones. That made for a spike in AHE largely driven by this compositional effect (as economists say). As they returned, wages appeared to dip, but it was just the flip side of the compositional effect.
But recent behavior looks not to be driven by compositional effects, but by the elusiveness of hirable workers. For the year ending in August, hourly earnings in AFS were up over 15%—like nothing we’ve seen before. Wages for the broader workforce are up just 5.5%—high by the standards of the last 30 years, but nothing like they are in AFS.
Sadly, though, that 5.5% increase is equal to the current rate of inflation, meaning that real wages are essentially unchanged. But hotel and restaurant workers look to be booking gains well ahead of inflation, at least for now.
Employers, right-wing politicians, and the pundits who speak for them have been claiming that the expansion of unemployment insurance (UI) benefits to counter covid woes was hurting job growth. By making it possible to refuse crappy jobs, or maybe even not to work at all, that sort of public sector generosity was making the working class too picky. They’re a lazy lot, you know, and need a good kick in the ass to get them to perform their class duty of laboring for the boss.
Problem is there’s just no evidence that expanded UI benefits hurt job growth. They ended nationwide on September 6, and job growth last month was the weakest of the year—just 194,000, almost half a million below the average of the previous six months.
Comparing job growth in states that cut benefits before the federal deadline with those that didn’t underscores the point. State employment data isn’t released until two weeks after the national data, so we won’t have September data until October 22. But the data through August is revealing.
Graphed below are rates of job growth in August and for the previous three months for cutters (states who cut expanded benefits before the federal termination) vs. non-cutters (those that didn’t). Cutters are divided into those who cut early, before June 26, and those that cut later. The roster is at the bottom of this page.
As the graph shows, non-cutters saw faster job growth than cutters for both August and the three months preceding. In August, national employment grew 0.20%. Among cutters, it grew 0.12%; among non-cutters, 0.25% (These are based on totals of each of the categories, no simple averages of state growth rates.) Within the cutters, the early birds actually saw a small loss of jobs, 0.04%. The April–July figures tell a similar story, with cutters (0.45%) lagging non-cutters (0.61%), and the early cutters (0.29%) doing worse than those later to the cruel game (0.49%).
A look at the roster of cutters below shows that many of the states were covid hotspots over the summer. And there’s a big disparity in vaccination rates: cutters have an average vaccination rate of 50% (early: 48%, late: 51%) compared with 60% among non-cutters. Maybe covid is more of a problem than generous income supports.
Also, there’s no good reason for the feds to have ended expanded benefits. They’ve helped millions cope with a still-broken job market. But it was an additional stroke of cruelty to cut them before the Washington did. I was going to call it pointless cruelty, but cruelty was the point.
Early cutters Alabama, Alaska, Idaho, Indiana, Iowa, Mississippi, Missouri, Nebraska, New Hampshire, North Dakota, West Virginia, Wyoming.
Late cutters Arizona, Arkansas, Florida, Georgia, Louisiana, Maryland, Montana, Ohio, Oklahoma, South Carolina, South Dakota, Tennessee, Texas, Utah.
Just added to my radio archive (click on date for link):
October 7, 2021 Nancy MacLean, author of this paper, on how Milton Friedman’s war on public education fit nicely with Southern massive resistance to desegregation • Klaus Jacob, a geophysicist, on how we can live with rising seas and heavier rains