An in-depth look at the leading indicators from the employment report

 

 - by New Deal democrat

First things first: there’s almost no significant economic news at all this week until Thursday, so don’t be surprised if I play hooky for a day or two.

The coincident headline news out of last Friday’s employment report was very positive, so most all observers heaved a sigh of relief. Of course, precisely *because* it is coincident, it could all be reversed next month, or by next month’s revisions to Friday’s data.

But since I am all about leading indicators and forecasting, let’s take a deeper look at those indicators from Friday’s report.

First, a little perspective. Recall that last week I was writing about manufacturing and construction. The former has been showing at least mild contraction for many months according to most measures, while the latter has continued to grow. I pointed out that for an economic downturn, I’d be looking for both to contract in tandem. The service sector for many decades has tended to expand in all but the deepest recessions. So it’s only when the goods sector as a whole turns down that there is enough downward pressure to pull the economy generally down with it. Which is why, when we had the very good ISM services report on Thursday, which showed that the economically weighted average of manufacturing and non-manufacturing remained expansionary, I was relieved.

So, with that background let’s look at the manufacturing, construction, and other leading indicators from the jobs report.

Turning to manufacturing first, here are all employees in the sector (blue, left scale, normed to 100 as of their recent peak) vs. the average manufacturing workweek (red, right scale), and manufacturing production (gold, left scale, also normed to 100 as of their recent peak) from the industrial production report:

Manufacturing hours have been part of the official Index of Leading Indicators for many decades. That’s because factories cut back hours before they actually lay off employees, so they are the proverbial “canary in the coal mine.” Hours declined sharply in 2022, but have stabilized in the past 18 months except for a downturn last Holiday season. 

Production peaked later in 2022, but has also stabilized in the past 18 months. The number of manufacturing sector employees, meanwhile, continued to very gradually increase until peaking this past January. It has declined -0.4% since.

A historical look shows that while this is consistent with weakness, it is not recessionary:

Manufacturing employment in the past 40 years has typically declined by -5.0% or more before a recession has begun, and hours have declined sharply below an average of 40.5 per week. This year hours have stabilized at about 40.7.

Turning to construction, both total jobs in that sector and the even more leading housing construction jobs continued to increase to new post-pandemic highs in Friday’s report. In the case of the latter, it was also a 15+ year high; for the latter it was also an all-time high:

Note that housing construction jobs have always turned down first, and many months before the last three recessions before the pandemic. Total construction jobs also peaked before two of the three pre-pandemic recessions, although somewhat later.

Needless to say, this is very positive.

As I wrote last week, housing construction employment (red) tends to peak after housing units under construction (dark blue, right scale) does, sometimes by many, many months. It also has only peaked after housing units actually completed (light blue, left, /2.5 for scale) has peaked as well:

While the sharp decline in housing units under construction in the past few months remains a considerable concern, housing units completed has continued to increase, just as sharply. So it appears we have at least a few more months to go before residential building employment might turn down.

Several of the other leading indicators in employment are not faring so well.

Temporary help was an excellent leading indicator before each of the last three pre-pandemic recessions. Post-pandemic it peaked in early 2022 and has been declining sharply for 2.5 years ever since. Trucking employment, meanwhile, flattened out and sometimes declined before the previous recessions. It has also been declining for two years (the sharp decline one year ago was caused by the bankruptcy of one major trucking firm):

I suspect there are unique factors having to do with the severe “overshoot” in temporary employment right after the pandemic, and employers’ hoarding existing help thereafter, which make this series unreliable this time around (remember, no indicator is perfect!). Trucking is of somewhat more concern, and seems to mirror the downturn in manufacturing production we saw in the first graph above. 

Another such indicator is short term unemployment (less than 5 weeks). This series is similar to, but predated the tabulation of initial jobless claims by several decades. Its drawback is that it is much noisier, meaning there are many more false positives or false negatives, as shown in the historical graph below:

Here’s the post-pandemic view, with each series normed to 100 as of their lowest levels (in the case of unemployment, the lowest three month average):

As you can see, the four week average of initial jobless claims gives a much cleaner and less noisy signal, although as I point out weekly, there appears to be some residual post-pandemic seasonality.

Finally, let’s look at goods employment as a whole. The long term historical look shows that, with the exception of 1974 (the first oil shock) and 1982 (caused by quick sharp Fed rate hikes), goods employment has always peaked at least several months before a recession:

Goods employment also continued to increase to a new post-pandemic high last month, although the gains are decelerating, currently to +0.9% YoY. So let’s look at the YoY% situation, and subtract -0.9%, so that the current level shows at the zero line:

Before the 1980s, a sharp decline to this level of YoY growth almost always meant an imminent recession. But in the past 40 years since the early 1980s, this has been consistent with at least weak economic growth, and for much of the period it has been average.

To sum up, there are some legitimate areas of concern in the leading indicators from the jobs report, mainly trucking and short term unemployment. There are several other weakly negative or neutral leading indicators: manufacturing employment and hours. But the main tone remains positive, if weakly so, mainly due to the continued strength in the construciton sector, which is boosting goods production employment as a whole. 

So long as construction employment, and in particular residential construction employment, holds up, the economy will remain in decent shape.

Weekly Indicators for September 30 - October 4 at Seeking Alpha

 

 - by New Deal democrat

My “Weekly Indicators” post is up at Seeking Alpha.

There was a slight fading of several indicators in the short leading and coincident sphere, but overall the positive and improving trend continues.

As usual, clicking over and reading will bring you up to the virtual moment as to the state of the economy, and reward me a little bit for organizing and presenting it to you.

September: very much a “soft landing” jobs report. But will the Fed use this to fall behind the curve again?

 

 - by New Deal democrat



Especially in view of the relative weakness in the jobs report for the past few months, my focus continues to be on whether jobs gains are most consistent with a “soft landing,” i.e., no further deterioration, or whether there is further decline towards a recession. 

For this month at least, the verdict was clear: both the Establishment and Household Surveys pointed to “soft landing.”

Below is my in depth synopsis.

HEADLINES:

• 254,000 jobs added. Private sector jobs increased 223,000. Government jobs increased by 31,000. 
• For a change, there were *upward* revisions to the last two months. July was revised upward by 55,000, and August by 17,000, for a net increase of 72,000. This breaks with the pattern from nearly every month in the past 18 months of a steady drumbeat of downward net revisions.
• The alternate, and more volatile measure in the household report, showed an increase of 430,000 jobs. On a YoY basis, this series has only risen by 314,000 jobs, which remains consistent with recession, as it has for months. On the other hand, it is an improvement from last month, where there was an actual YoY decline.
• The U3 unemployment rate declined -0.1% for the second month in a row, to 4.1%, which also means the “Sahm rule” recession indicator, for the first time in three months, is no longer in effect.
• The U6 underemployment rate fell -0.2% to 7.7%, still 1.3% above its low of December 2022.
• Further out on the spectrum, those who are not in the labor force but want a job now rose another 60,000 to 5.97 million, vs. its post-pandemic low of 4.925 million in early 2023.

Leading employment indicators of a slowdown or recession

These are leading sectors for the economy overall, and help us gauge how much the post-pandemic employment boom is shading towards a downturn. This month they were mixed:

• the average manufacturing workweek, one of the 10 components of the Index of Leading Indicators,  was unchanged at 40.7 hours. This remains down -0.8 hours from its February 2022 peak of 41.5 hours, but on the other hand is only 0.1 hour below its 18 month high.

• Manufacturing jobs declined -7,000.
• Within that sector, motor vehicle manufacturing jobs declined -6,500. 
• Truck driving declilned -700.
• Construction jobs increased 25,000.
• Residential construction jobs, which are even more leading, rose by 2,000 to another new post-pandemic high.
• Goods producing jobs as a whole rose 21,000 to another new expansion high. These should decline before any recession occurs.

• Temporary jobs, which have generally been declining late 2022, fell by another -13,800, and are down about -500,000 since their peak in March 2022. This appears to be not just cyclical, but a secular change in trend.
• the number of people unemployed for 5 weeks or fewer declined -322,000 to 2,146,000.

Wages of non-managerial workers

• Average Hourly Earnings for Production and Nonsupervisory Personnel increased $.08, or +0.3%, to $30.33, for a YoY gain of +3.9%. This continues the trend of deceleration from their post pandemic peak of 7.0% in March 2022, and was less than 0.1% higher than its post-pandemic low set in July. Most importantly, though, this continues to be significantly higher than the 2.6% YoY inflation rate as of last month.

Aggregate hours and wages: 

• the index of aggregate hours worked for non-managerial workers increased 0.2%, and is up 1.2% YoY, in trend for the past 12+ months.

•  the index of aggregate payrolls for non-managerial workers was rose 0.4%, and is up 5.2% YoY. These have been slowly decelerating since the end of the pandemic lockdowns, and that trend continued this month, which was tied for the post-pandemic low. Nevertheless, with the latest YoY consumer inflation reading of 2.6%, this remains powerful evidence that average working families have continued to see gains in “real” spending money.

Other significant data:

• Professional and business employment rose 17,000. These tend to be well-paying jobs. This series had generally been declining since May 2023, but earlier this year had resumed increasing again. As of this month, they are only higher YoY by 0.5% - a very low increase that has *only* happened in the past 80+ years immediately before, during, or after recessions. On the other hand, there has been no meaningful further YoY deterioration in the past 12 months.
• The employment population ratio rose 0.2% to  60.2%, vs. 61.1% in February 2020.
• The Labor Force Participation Rate also remained steady at 62.7%, vs. 63.4% in February 2020. The prime 25-54 age  participation rate declined -0.1% to 83.8%, vs. 84.0% in July, which was the highest rate during the entire history of this series except for the late 1990s tech boom.

SUMMARY

Last month I wrote that “Along with the big downward revisions to the last several months, this month’s report is the first time that there is substantial evidence that the jobs market may have moved past a ‘soft landing’ into a hard slowdown that could easily tip over into outright declines by the end of the year.” 

This month’s report completely reversed that. Not only were the headline numbers all good, but there were positive revisions to the past several months. Average middle and working class workers continue to see good wage and hourly increases in pay, with real inflation-adjusted increases to buying power. And the headline Household survey employment number joined in the good news for a change.

Not everything was rosy. To reiterate what I wrote last month, “manufacturing seems at long last to have rolled over,” with the third decline in four months. The sector has shed -0.4% of its total so far this year. Trucking continues to slowly shed jobs as well. This isn’t recessionary compared with the last 30 years, but it is weak. At the further edge of the spectrum, those who haven’t looked actively but want a job now increased close to a 24 month high.

But the construction sector continues to do well, including - surprisingly - residential construction jobs, despite the downturn in housing construction metrics. Short duration unemployment declined, mirroring the recent downturn in initial claims.

To reiterate my opening statement, for this month at least the “soft landing” scenario is fully intact. If there is a concern, it is that the Fed, which has been behind the ball for about 3 years, will take this as a sign that it can maintain high interest rates, despite the fact that it has to make policy for the jobs sector one year from now, not for last month.

The ISM services index, measuring 75% of the economy, sounds an ‘all clear’ - for now, anyway

 

 - by New Deal democrat

Recently I have paid much more attention to the ISM services index. That’s because, since the turn of the Millennium, manufacturing’s share of the economy has contracted to the point where even a significant decline in that index has not translated into an economy-wide recession, as for example in 2015-16. 

When we use an economically weighted average of the non-manufacturing index (75%) with the manufacturing index (25%), it has been a much more reliable signal, particularly when we use the 3 month average, requiring it to be below 50. 

Once again this month the contraction shown in the manufacturing index has been more than counterbalanced by continued expansion in the services index, which was reported at 54.9. The more leading new orders subindex (not shown in the graph below) came in stronger, at 59.4:

Here are the last six months, including September, of both the manufacturing (left column) and non-manufacturing index (center column) numbers, and their monthly weighted average (right) :

APR 49.2  49.4.  49.3 

MAY 48.9. 53.8. 52.5

JUN 48.5. 48.8. 48.7

JUL. 46.8. 51.4. 50.2

AUG. 47.2. 51.5  50.4

SEP. 47.2. 54.9  53.0

And here is the same data for the new orders components:

APR 49.1. 52.2. 51.4

MAY 45.4. 54.1. 51.9

JUN. 49.3  47.3. 47.8 

JUL.  47.4. 52.4. 51.2

AUG. 44.6. 53.0. 50.9 

SEP.  46.1. 59.4. 56.1 

While the single month average for both the headline and new orders components showed contraction in June, but it did not trigger a signal based on the three month average. Last month the three month weighted average of the headline numbers was 49.7, and the new orders component 49.97, which rounded to 50.0.

With this month’s data, the three month headline average for manufacturing is 47.1, and for non-manufacturing is 52.6. That makes the economically weighted headline average 51.2. For new orders the weighted three month average is 52.7.

Essentially, with the poor June readings for the new orders subindex in particular out of the average, the economically weighted ISM indicator has returned to a moderately positive aggregate reading. This is good news.

Tomorrow I will be looking to see what happens with manufacturing and construction jobs in particular, and whether new entrants to the labor force continue to distort the unemployment rate higher. We’ll see then.

Jobless claims: not so good as the headline, but not so bad either

 

 - by New Deal democrat

Initial jobless claims will be up against some very challenging comparisons for the next 6 months or so, due to some unresolved post-COVID seasonality. Which means that the headline numbers this week, which look very benign at the surface, are not quite so good as they have been for the past year.

For the week, initial claims rose 6,000 to 225,000. The four week moving average declined -750 to 224,250. Continuing claims, with the typical one week lag, declined -1,000 to 1.826 million:

Again, as you can see from the above, all of these look pretty benign in absolute terms.

But for forecasting purposes, the YoY% change is more important, and viewed that way, the story is a little different. Initial claims were up 4.2%, the four week average up 3.6%, and continuing claims up 2.1%:

With the exception of a couple of weeks in January, both initial claims comparisons are the worst since last October. Continuing claims has trended slightly higher since their low YoY point several weeks ago.

This does *not* mean recession. For even a yellow caution flag, these would need to be higher YoY by 10%. But initial claims are no longer positive for the economy either. Rather they suggest an economy slowing to so-so expansion.

Finally, let’s take our last look at what claims suggest about the unemployment rate tomorrow and for the next few months.

First, here is the YoY% change look:

After trending higher one year ago, earlier this year both initial and continuing claims hewed close to unchanged for most of this year. The unemployment rate, which has been heavily affected by new (immigrant) entrants to the labor force, has not followed - a very rare event over the past 60 years.

Here is the absolute level view:

Based on historical experience, the unemployment rate should have trended down, towards 3.7%, not up, during this summer. Even with the effect of a sharp increase in the labor force, initial and continuing claims suggest that the unemployment rate will not tick higher, but remain at 4.2% +/-0.1%. We’ll see tomorrow.

Are manufacturing and construction in a synchronous downturn? If so, that’s Trouble

 

 - by New Deal democrat

I wanted to follow up on a point I made yesterday: although manufacturing is no longer a big enough slice of the US economy to bring about an economic downturn on its own - unless for some reason the manufacturing downturn were unusually severe - when it is paired with a downturn in construction, that historically has been a reliable (but of course not perfect!) harbinger of recession.

And while yesterday’s construction spending report was equivocal, there are other signs suggesting that just such a synchronous downturn may be occurring.

Let me start with a refresher on the historical value of the ISM manufacturing index, especially its new orders component. This survey has been in existence for over 75 years, since 1948. It has been recognized as a leading indicator almost since the start.

The ISM no longer allows FRED to update its readings, but fortunately about 10 years ago I pulled a graph of all of the new orders data, and here it is (blue, left scale):

With the exceptions of 1970, 1973, and 2008, it always turned down below 50 in advance of recessions. In those three cases it was coincident. In addition, however, there were a number of false positives; namely, in 1952, 1966, 1984, 1995, 2002, and 2012.

Here is the updated graph I ran yesterday:

Again, there were false positives in 2015-16 and 2022-23. Arguably 2019 was as well, but because the COVID pandemic intervened, we’ll never really know. This is what I have been highlighting in noting that a downturn in manufacturing no longer carries enough weight to be a leading indicator by itself.

Now let’s take a look at the second element, construction, as reflected in housing units under construction, which as I typically point out each month, reflects the actual full economic activity of the housing market, vs. sales, permits, starts, or completions.

The easiest way to visualize this is YoY, which I’ve done below. Because typically it takes a 10% or more decline to signal recession, I have added 10% to the values so that -10% YoY shows at the zero line (note: data only begins in 1971):

With the exception of 2001, a downturn of 10% YoY or more has always meant recession, with the exception of a single month in 1988.

Here is the same data presented as absolute numbers:

I am showing this last graph to compare downturns with the ISM new orders data since 1971. In each case of a “false positive” in the ISM new orders - 1984, 1995, 2002, 2012, 2015-16, and 2022-23 - housing construction was either increasing or at least not declining in any significant way, i.e., by less than 5%.

So the fact that as of this past month’s housing construction report, we now have a 10% downturn in units under construction, coinciding with a significant decline in manufacturing including its new orders component, is a real cause for concern.

A significant positive, meanwhile, is that real spending on goods has continued to increase:

And indeed is up 2.2% YoY on a per capita basis. In the past 50 years, with the exception of 2001, this has always turned negative coincident with or before the onset of a recession:

As I wrote yesterday, tomorrow we will get the ISM services report, which will help us see how much this downturn is showing up as a slowdown in that much larger sector of the economy as well. And in Friday’s employment report, we will get an updated look at the number of employees in manufacturing and construction. As of last month, the former had turned down, while the latter was still growing. If there is a significant synchronous downturn, both sectors of employment should turn down, which would be the final warning signal that the odds of a recession as opposed to a “soft landing” have increased greatly.