We didn't really 'add' 1.8 million jobs in July.
Last week, the Bureau of Labor Statistics (BLS) released their July jobs report. According to many business journalists, this was the third month of surprisingly good news in a row: The U.S. added 1.8 million jobs, based on a survey of around a third of all employers, and the unemployment rate dropped to 10.2%, based on a survey of ~40,000 households asking about the week of July 12.
Except...two days before that, the ADP Jobs Report, which surveys 400,000 businesses, counted a July increase of just 167,000 jobs. It’s not unusual for ADP to adjust their numbers up after initial reporting, but theirs wasn’t the only number disagreeing with the headline figures from BLS.
Half of all respondents to the Census Bureau’s new quick-and-dirty Household Pulse Survey reported a loss of employment income the week of July 12 -- about 126.6 million people. A third expected to lose more in the future. Even within the jobs report, the numbers didn’t align; the household survey estimated only 1.4 million more people were employed (and the unemployment rate for Black people remained a devastating 14.6%).
The majority of service and manufacturing businesses reported a reduction of employment in July, the fifth month of contraction in a row, according to the Institute for Supply Management’s July Report on Business
As for unemployment claims, about 30.8 million Americans collected unemployment across all programs during the week of July 11. The next week, it was 31.3. That works out to an “insured unemployment” of 11.6%, a metric the BLS specifically doesn’t use because it’s almost always lower than the real unemployment rate.
So what the fuck is happening?
Stop using the phrase “job creation” unless you mean it.
It’s incredibly hard to define ‘jobs’ and ‘employment’ in our union-free, gig-economy hellscape. And all our estimates are extra-squirrelly thanks to [the pandemic has fucked all reporting of everything] and [most models rely on comparisons to historical data, which are pretty useless right now], both of which I’m going to cover in more depth in upcoming newsletter issues.
But we can at the very least call bullshit when we see it. And pretending unemployed people have millions of jobs to go back to is bullshit.
Between July 2018 and 2019, payroll numbers in professional and business services (a catch-all category, mostly office jobs) rose by 369,000. Those jobs were ~created.~ Between February and July this year, the same sector was down 1.6 million jobs. Almost every one of the 170k jobs ‘added’ between June and July were ‘temporary help services,’ aka temps.
All told, less than a third of the 16.7 million full-time jobs lost in March and April were back by June, according to the household survey -- and almost none were added in July. (You can mess with the household and payroll survey data yourself.)
To deal with uncertainty, look at trends, not single numbers.
In a recent working paper, data scientists Yair Ghitza and Mark Steitz tried to model coronavirus job losses by neighborhood, using the microdata underlying BLS unemployment stats (the Upshot did a great piece on their work). They stressed that they have no background in the field, but are hoping to start a conversation about new ways to model unemployment data.
Their introduction explained better than I can why we should be careful right now:
“First, our estimates are modeled inferences, not reported data,” they wrote. “This caveat may be obvious for experienced modelers, but the allure of ‘big data’ and sophisticated methods often confuse people into assigning too much specificity to modeled inferences.”
“Second, the impact of COVID-19 on the labor force is an incredibly fast-moving situation, accompanied by many types of data and reporting problems...As such, we interpret our estimates, as well as data from other sources, as having a larger amount of uncertainty around them than usual. We believe analysts and policy-makers should not exclusively rely on any single piece of evidence, and would be better served by comparing different sources.”
Whatever comes next is someone’s fault.
Over the next few issues of this newsletter -- and with the help of actual experts! -- I’m going to dive into where estimates and models are failing us, and how we can do better. But in the meantime, I have to say something about how we’re talking about unemployment benefits.
Lots of self-described conservatives keep complaining that household incomes have increased under enhanced unemployment benefits, which both expanded access to the social safety net and added $600 a week to the take-home. They argue having a living wage will prevent people from going back to jobs that don’t pay a living wage.
I understand it hurts to imagine poor people eating and paying rent without working three jobs and unnecessarily exposing themselves to coronavirus. But America hasn’t been a manufacturing economy in a long, long time -- consumer spending makes up more than 2/3rds of the U.S. GDP (Bureau of Economic Analysis). And if the $600/week enhanced unemployment benefits are completely dropped, total household income in the U.S. will sink by $75 billion a month, according to Bill Adams, a senior economist at PNC Bank. We cannot produce our way out of this hole.
(Please don’t get me started on Trump’s ridiculous plan to sidestep Congress and divert disaster relief funds to pay people off for a month; the actual mechanics of it are a mess, and I’m not actually sure we’ll still be a country by the time anyone figures out how to do it.)
But Christ, I hope Congress gets its shit together. Even without an ongoing pandemic, we’ll still need massive federal help to survive this. Unemployment as a result of the 2008 crash didn’t peak until May 2010 (Brookings, fig. 3), a year after the recession officially ended. It’s hard to imagine a situation where things look better next month than they do this month (especially with schools reopening under circumstances that have left me, and every sane health expert, in literal open-mouthed horror).
People in power like to talk about the “the economy” and “payroll” and “fiscal responsibility” because it gives them distance from the human consequences of their actions. Now’s the time to stop debating economics like a natural science, and turn their attention to history. Nothing good comes from a ruling class letting their peasants starve.