July saw the recovery stall out in many ways. The unemployment insurance numbers tell the weekly story:
Like everything else, the normally highly reliable UI tables have gone all wonky on us. The difficulty states are having with their ancient IT systems has been well-documented. But additionally, the substantial Pandemic Unemployment Assistance (PUA) is counted in the “All Programs” table, not with the normal UI. Many recipients are overlapping, but there is no way to shake that out. Right now, there are almost 11 million PUA recipients, and 1.5 million recipients of other smaller programs, mostly the Pandemic Emergency Unemployment Compensation, part of the same legislation.
These programs have now ended, but that won’t show up in the data for a couple of weeks.
So given that, we see three things
- The continued claims number has come off its downward trajectory in July and flattened out, though it may be coming down again in August.
- The all programs number has remained incredibly sticky at around 30 million.
- The net is that the difference of 12-15 million people remains incredibly sticky since mid June.
So given all the caveats, it’s hard to say precisely what is happening, except that we see flat lines in July. The BLS jobs surveys were conducted the week ending July 18, so it is capturing some, but not all of what is happening.
I find it helpful to compare everything we are going through with the GFC, because most everyone reading this remembers it well. Here’s that comparison:
So we see that last uptick is substantially smaller than the first two in May and June. Also, we have a long, long way to go, just to get to the GFC bottom.
Unlike most analysts, I have been using the not seasonally adjusted numbers since this all started. Why?
Under normal circumstances, the adjustments are important, because without them, every January and July would look like recessions.
I pulled out the recent action so you can see what a more normal 10-year period of NSA payrolls looks like.
But never has seasonality been less relevant. The adjustments are multiplicative, not additive, so they can be very large when we are dealing with large numbers like we are now. The adjustments in July are large, and inflate payrolls quite a bit in contrast to the three previous months, which are smaller and understate payrolls.
But most importantly, I want the answer to the question, “How many Americans were employed in mid-July?” The not seasonally adjusted numbers answer that question, or at least try to. It’s not without issues, as we will see, especially with education employment in July.
Problem for a later date: the seasonal adjustments are based off previous years’ NSA numbers. They are going to have to come up with some way of adjusting the adjustments for all the highly unusual data coming through in 2020. If they don’t, Q2 2021 will look like the greatest economic miracle of all time, only to be crushed hard in Q3 and Q4.
The Household Survey: In Which I Just Give Up
The household survey, from where the headline unemployment rate comes, has been a mess since April, and there are three sources here
- Surveys normally done through in-person interviews are happening on the phone, which is less than ideal. There is also limited staff at the offices.
- Historically low response rates.
- Coding error by the Census Bureau interviewers, now four months running. It was around 5 pp of unemployment in April, but now down to around 1 pp.
Low response rates are the curse of any survey, because it’s hard to get a random sample that way. The response rate is up to 67% from a low of 65% in June, but this is well off the normal 83% rate.
But given the numbers we have, especially combining with other data sources, the unemployment rate is most certainly higher than reported. By how much is hard to know.
This is an attempt to get at the real rate, adding back in the error and accounting for the substantial churn in the participation rate, but these are still based off numbers I consider unreliable. If I had a gun to my head and had to guess, I would put the rate at around 13% right now, which would be the high previous to 2020.
We’ll look at some alternative data, but it doesn’t make the picture much clearer.
The Employer Survey
The employer survey is also not without problems, even before any of this. The employer survey undergoes annual revisions, and in recent years, it has been overcounting by about half a million people every report.
But the response rates are back to normal, though procedures are not. Some of the data is being collected via online surveys, and these are less preferable to the usual phone interviews. But it’s the best we have.
So starting from the top.
42% of lost jobs have returned, which is another way of saying 58% have not. That translates to 8% of February jobs lost. That is a bit misleading, since the NSA numbers are counting lots of education workers still out of work, many of whom would not be counted as employed right now anyway. It’s still unclear how many will be returning to work in the fall.
But there is a tremendous amount of variation hiding underneath that. The diffusion index is a way of measuring how widespread job losses or gains are, not their size. Zero means that all industries are seeing job losses, 100 means all industries are seeing job gains, and 50 means split down the middle.
The diffusion index hit 4 in April, the lowest ever, beating the 2009 record of 17. Let’s zoom in:
This indicates that April job losses were incredibly widespread, in almost every industry. May and June saw winners beating losers, but not like losers beat winners in April. July was a retreat from June, but still with winners beating losers.
So this frames what we see in the employer survey’s detailed splits: some jobs are coming back at a pretty fast rate. But there is a group of industries, which we also saw were the hardest hit in the Q2 GDP report. These are services that have high fixed costs and make up for it with density and volume – passenger transportation, leisure, hospitality, travel, residential medical, child day care, video production, and personal care.
Together, these were 19% of all jobs in February, but only 17% by July with employment still down 19% from February levels. They account for 39% of all remaining job losses, and 55% of remaining ex-education losses.
What jumps out immediately is restaurants and bars. This is the largest group of job losses, returners and remaining job losses. June consumption in the recent Q2 GDP report saw fast food return to February levels with the rest lagging badly.
For reference, that fast food chart is what an actual V-shaped recovery looks like.
The detailed splits in the employer survey are a month behind, so the data only goes through June. Looking at non-supervisory workers at restaurants – servers, kitchen staff, fast food workers – these were 23% of all job losses through April, and 19% of all remaining ex-education job losses in June. But there is a big divergence between fast food and full service places, as you might imagine.
In April, 69% of full service restaurant workers were unemployed, but only 23% of fast food workers. In June, full service employment was still down 31%, and fast food only 8%. 64% of lost fast food jobs returned, but only 58% of full service.
So even with fast food wages rising slightly, this has been offset by falling hourly wages for tip-reliant workers in full service places, down 3.3% from February to June. With reduced weekly hours at full service, and increased weekly hours at fast food, weekly wages for those still employed are down 2.1% overall. But this is hiding the splits, where fast food workers are taking home 5.5% more in weekly pay, but full service workers getting 4.4% less.
So like in many places in the economy, we are seeing rapid changes in consumer behavior – in this case eating out less, and shifting to fast food when they do.
Next up is leisure, called recreation services over at the GDP report. This is where we see the hardest hit categories by percentage, with movie theaters and live sports still down 100% in June.
Overall, employment is still down 20% from February to July, but that was 50% in April. Again, the detailed splits only take us through June, but the numbers in many subcategories were still very large:
Fitness centers are by far the largest group here, so that’s a key thing to keep an eye on. Golf courses are the only winners, and part of that is seasonal. The other part is social distancing is possible on a golf course, and it’s one of the few paid recreational activities that remain safe.
In any event, it’s just really hard to imagine full recoveries here absent a safe, effective vaccine.
Next up is accommodations, with employment still down 28% from February to July. Looking at the consumption numbers through August, we see that recovery has been very slow:
Breaking it down, we see the larger hotel categories are still way down through June, but outdoor accommodations have seen increased employment.
Again, part of this is seasonal, but it’s also much safer to go camping.
Personal care and clothing services employment is still down 20% from February to July. In the detailed splits through June, hair and nail salons were still down 34%.
Passenger transportation is one of the hardest hit categories, with employment continuing to decrease after April, and representing almost 6% or remaining lost jobs.
In the consumption data through June, we see very small recoveries outside vehicle services.
Employment is reflecting that, though airlines are holding up a little better due to $24 billion in Federal relief for them in the last package.
Once again, there is a long road back here.
Finally, let’s break out the services I lumped into “Other”:
- Travel agencies, including online ones, are of course dependent on the transportation and accommodations categories.
- Day care will not recover until there are fewer parents stuck at home.
- Residential medical has a host of issues with COVID-19, and we’ve seen numerous outbreaks at nursing homes. Nursing home consumption is down 15% from February to June, and employment is starting to follow, down slightly every month since March.
- Film and TV sets are crowded and chaotic places where social distancing is next to impossible. On top of that, highly-paid actors and directors, the lynchpin of the whole operation, can afford to take a private plane to their ranch in Montana to ride this out in comfort, surrounded by bison and no humans.
So overall we are seeing a group of services, about a fifth of pre-pandemic employment, where recovery thus far has been weak, and there is not a lot of hope absent a safe, effective vaccine.
Retail is another hugely important area that is also under a tremendous amount of stress. But again, there is a real split in what is happening beneath the topline numbers.
Retail employment was down 15% through April, 11% of all job losses, but has shown significant returns in some areas, but not in others. Like so much else, the split reflects changing consumption patterns.
In the first place, we’ve seen a shift to nonstore retail, mostly online these days, but there is still mail order catalogs. Nonstore was never down that badly in the first place and has recovered 89% of those lost jobs.
But brick-and-mortar is where it gets interesting.
In March and April, people were hoarding food and household supplies, so employment numbers were buoyed by supermarkets and warehouse stores. But then in May and June, the nest feathering began, as people looked for more comforts and diversions in their homes. Sales of some household durables have not only recovered, but are above February levels:
So if we group together retail for household goods, these workers have not done so poorly.
Losses were never as bad to begin with for this portion of the retail landscape, and they are more or less back to February levels by July. But even the rest are not doing as badly as I had feared. This is a giant employment sector, so it could have been much worse.
The Office Jobs That Have Disappeared
In contrast to retail which is better off than I expected, there are a group of jobs categorized under professional, business, and information services that are underperforming my expectations. Altogether, this was 22 million jobs in February, 16% of all jobs. These are largely office jobs that could be done from home, but the job losses in these categories have been very sticky, with only a third of lost jobs recovered. In July they accounted for 13% of the remaining lost jobs.
Information services is publishing, broadcasting, video production, telecommunications and data services. Every one of those categories is down, and the jobs losses persisted past April through July. Many of these are likely permanent job losses.
Professional and business services is a larger grouping, so it is more mixed, but many of the categories are still down substantially, some still reducing workforce since April.
I keep expecting these to turn around, but they haven’t really.
Manufacturing always gets an inordinate amount of attention, which is an artifact of an earlier time. In the first place, we have a mostly service-based economy now, so goods-making is not nearly as important as it once was to the economy or to employment.
As many people now work in retail as they do in manufacturing, and a lot more work in leisure and hospitality. But manufacturing is certainly not small. Even though employment is down 4.8%, much less than in the services categories we looked at, this is also 5.1% of remaining lost jobs, so we’re still talking about 600k jobs.
Generally speaking, if we pull out food and petroleum, nondurables are doing a lot worse than durables. Food manufacturing is one of the best categories, and petroleum one of the worst.
Surprisingly, those household durables industries we were discussing are still seeing large job losses in the high single digits through July. There are very low inventories right now, so I expect that to change as soon as the next report.
Alternative High Frequency Data
Things are changing so rapidly that we are looking to new data sources, especially ones with shorter than monthly periodicity. The jobs report surveys were done the week ending July 18, so it is a slice in time, not a summary of the whole month. Normally, that’s fine, but things are happening too rapidly now.
We start with the Dallas Fed’s Real Time Population Survey (RPS), a twice monthly, much less detailed version of the household survey. It is telling us a very different story:
As you can see, since May the RPS has been giving us a much higher number, and also shows the recovery reversing in late June. The rates are much higher, especially now with a 7 pp difference, which is about 11 million people. Which one is right? Neither? I can’t tell you.
Google and Apple have been releasing mobility data from their maps apps. Google’s is a little more detailed, and tells us generally, where people were headed on their trips. Helpfully, workplaces are one of the categories.
The recovery in workplace trips stalled in mid June nationally and has been going sideways more or less since. If we look at COVID-19 hotspots in the sunbelt, the change is more dramatic. I’ll zoom in on this one so you can see the recent action better.
These have all fallen much more than the national numbers since mid June. We may see activity coming back in August, which is consistent with other data.
As we saw, full-service restaurants are going to be key to the jobs recovery. OpenTable has been publishing data on YoY reservations. It’s unclear how representative a sample this is, but we can assume it mostly covers the top tier of restaurants. But these are also the highest paying restaurant jobs.
We begin here on May 1. Previous to that, there was a month and a half of -100%.
So you can see that August bounce I was talking about, but the rise is much slower than the one that stalled mid June.
Looking at the COVID-19 hotspots, the pattern is even more dramatic:
Again, I’m not sure how representative the OpenTable data is, but it does generally comport with what we are seeing elsewhere.
The final look is possibly the most scary. Yelp just reported that 55% of the closings in their business listings are now permanent:
Breaking it into business categories, we see that restaurants are even higher, at 60%.
This is grim, especially for jobs. Something like half of all workers work at small businesses.
Too long? Gotcha covered:
- The household survey remains a statistical mess, but the actual unemployment rate is almost certainly higher than what is reported. By how much is up for grabs.
- The employer survey is doing better, and shows the jobs recovery slowing down considerably in July.
- While job losses in April were widespread across almost all industries, job returns have been more narrow.
- The biggest problem area continues to be high density services, which accounted for about a fifth of jobs pre-pandemic. Full service restaurants are particularly key here.
- Retail is a huge employment sector, so whatever happens there is crucial. We see a split between household goods retail, doing pretty well considering the environment, and the others not doing nearly as well.
- The stickiness of job losses in information and business services is troubling. These are mostly jobs that could be done from home.
- Manufacturing job losses right now are concentrated in the non-food nondurables categories, especially petroleum. I expect more of a recovery here, especially in household durables.
- The high frequency data confirms the slowing in July as the pandemic spread like wildfire. As it eased in late July and August, we may be on the upswing again, but at a much slower pace than before.
Stuck in Containment
I keep coming back to this since he said it, because I think it frames our situation very well. During their Q2 earnings call, Citigroup (NYSE:C) CEO Mike Corbat above finished up his scripted remarks with this:
We are in a completely unpredictable environment which no models, no cycles to point to. The pandemic has a grip on the economy and it doesn’t seem likely to loosen until vaccines are widely available.
When asked to elaborate later in the call:
I think of this going through four stages, containment, stabilization, normalization and ultimately a return to growth… I would describe right now that broadly in the world we are somewhere between containment and stabilization, right? Containment is that we can bend the curve in terms of the transmission of cases. Stabilization is that as we remove or start to take down some of the barriers or actions that were put in place…
And when you get to the third phase around normalization and simply put normalization to me is am I willing to get on the airliner? Am I willing to get in a subway? Am I willing to go into a crowded venue to watch a sporting event or a concert or what it may be? And I think realistically when we get to that third bucket, I just don’t see that coming and I would say many don’t see that coming until we feel like there’s an anti-virus vaccine that’s available for the mass population around that. And so I think one of the things that people struggle with today is the disconnect in some ways between where the market is in some ways and actually where we are in terms of this health pandemic… So I don’t want to be pessimistic in there. I want to be a realist and I just think that in order to truly normalize, that’s what’s necessary to do that. [emphasis added]
Let’s break down the key points:
- The pandemic caused the recession, not the shutdowns. This is at root a public health problem, not a political or economic one.
- We cannot get out of containment.
- We will not get to normalization until there is a safe, effective vaccine. The recession only begins to end then.
- There is a huge disconnect between the market and economy right now.
Since March, Treasury has increased the Federal debt by over $3 trillion, now at $26.5 trillion. Over half that $3 trillion is still sitting in Treasury’s checking account.
So for over a trillion dollars, and more to come, we bought some time, about 5 months of it. The point was to use that time to learn about the virus, set up a national system of best practices, and ramp up testing and tracing so that turnaround was better than this:
Demand for testing has soared in recent weeks. And we are providing testing results in about two days for the highest priority patients, and the average turnaround time for non-priority patients is at least seven days.
-Quest Diagnostics (DGX) CEO Steve Rusckowski on their earnings call July 23
In seven days of not knowing, an infected patient can transmit to dozens of other people if they don’t self-quarantine. This is not best practices.
So we already spent over a trillion dollars, with more at the ready, and we were never able to get to containment. That money and the time it bought was wasted.
So with the parties in DC still very far apart, and enhanced benefits expired, this slow trudge back looks like where we will remain until there is a safe, effective vaccine.
I’ll just leave you with this chart again:
It’s a slow ride.
Disclosure: I/we have no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.