New jobless claims continue near historical lows; plus the “quick and dirty” forecast, and an important comparison vs. long term unemployment

 

 - by New Deal democrat

Let’s take our regular weekly look at jobless claims. Plus, this week I want to make a statement about jobless claims vs. long term unemployment.

First, initial jobless claims declined -2,000 to 197,000. Aside from 2 weeks in 2018, the only time prior to the pandemic that initial claims were this low was all the way back in the 1960s. The four week moving average declined -2,500 to 198,000, also the lowest since the 1960s. And with the typical one week delay, continuing claims rose 17,000 to 1.716 million, in the lowest range since early 2023:

In the YoY% comparisons more important for forecasting purposes, initial claims were lower by 15.5%, the four week moving average lower by 13.0%, and continuing claims lower by 11.0%:

These YoY comparisons keep getting better, even in the face of very good numbers one year ago. This is very positive for the economy over the next few months.

Not only are jobless claims very positive, but the stock market (blue in the graph below), concentrated on AI-related gains, is higher by about 15% YoY. Thus the “quick and dirty” forecast model, which also included the four week average of jobless claims (inverted, +10%), remains very positive for the economy as well:

It’s early enough in the month that I can eschew a comparison with the unemployment rate, but last Friday I read a prominent reporter touting how long term unemployment was recessionary. This is simply not the case. So let’s compare YoY jobless claims with long term unemployment by 27 weeks or more (blue in the graph below):

There are two important things to notice about the above graph. The first is that, if new jobless claims lead the unemployment rate, they lead long term unemployment even more. Secondly, note that long term unemployment continues to rise even after a recession is over; i.e., early during the ensuing economic expansion. In other words, a rising long term unemployment number is consistent with both a recession and also a renewed upturn in the economy.

Another way to look at this is the ratio of long term unemployment to initial claims. Here’s the historical look:

Again, note two things about the above graph. First, just like the number of long term unemployed, the above ratio also continues to rise into the ensuing expansion for months after the recession is over. Secondly, note that there has been a long term secular trend of a higher ratio of long term unemployed vs. newly unemployed both at the troughs and the peaks of the comparison. This has been particularly true since the 1980s. In part this is likely due to the relative strength of employers vs. workers and the decline of unions since the Reagan era. It may also be due to workers adapting to the availability of long term unemployment benefits, especially as more and more spouses also work; i.e., the unemployed worker can afford to be choosier about what next offered position of employment to accept.

Now here is the post-pandemic view:

Just as with prior periods, the ratio peaked in the summer of 2021, more than a year into the post-pandemic Boom. There are ambiguous signs that it may be peaking again now, going on a year after the late 2025 “mini-recession.”

In short, whatever long term unemployment may suggest about the economy, new jobless claims get there first.

Last week’s major revisions to personal income: comparing the source data

 

 - by New Deal democrat

While we are waiting for new data releases to start back up, let me follow up on the surprise revisioins last week to personal income, that took real personal income for declining for over a year to an almost relentless rise. So shocking was this reversal that for the first time I wondered if the T—-p Administration had been successful interfering with economic reporting.  To give you an idea of just how large the revisions were. here is the YoY% change as originally reported (orange) vs. as revised (blue) of nominal personal income:

Once adjusted for inflation, the entire trend in *real* personal income changed from negative to positive:

Ultimately an answer came back from someone with impeccable credentials who pointed out that, so long as the revisions were in accord with the source data series giving rise to the revisions, they should be accepted.

So what do those data sources say? Let’s start with the Census Bureau’s note expalining where the revisions came from:

While I haven’t been able to track down all of these sources, there are two - the QCEW and income tax withholding receipts - which are readily available.

The QCEW is what the late financial analyst Jeff Miller called “the gold standard” of employment data, as it is the actual reporting by roughly 97% of all employers. Its biggest drawback is that it is not reported until almost 6 months after the close of any given Quarter, and it is subject to revisions until March of the ensuing year. I’ve reported on the employment side of this release for many years, but haven’t paid particular attention to the income side. 

In any event, as per usual the data lags by half a year. Additionally it is not seassonally adjusted, so the only valid measure is YoY. To cut to the chase, here is the post-pandemic YoY quarterly record of aggregate wages:

While I can’t show you a graph, I also went back and calculated the full quarterly change in withholding income taxes paid YoY from the Treasury’s “Daily Treasury Statement.” There are very volatile and noisy on a daily, weekly, and even to some extent monthly basis, but by the time you take the entire 13 weeks that constitute a Quarter, the vast majority of the noise has been ironed out.

With that explanation done, below is a list showing the revised YoY% change in nominal personal income beginning with the 4th Quarter of last year, and ending with the first two months of Q3 of this year from the personal income releases vs. the entire third quarter for withholding tax payments. Note that since the QCEW has only been reported through Q1 of this year, there are no subsequent comparisons for that metric:

Quarter  //  P. I. (O) //  P. I. (R)  //  QCEW  //  D.T.S.

Q4 ‘25   //  +4.4%  //   +4.8%      //  +4.2%  //  +4.6%

Q1  ‘26  //  +3.6%   //   +4.6%      //   +3.9% //  +5.0%

Q2  ‘26  //  +3.5%  //   +4.4%      //  n/a        //  +1.9%

Q3  ‘26  //  +3.8%*  //   +4.4%*    //  n/a      //  +5.6%

(O) = original; (R) = revised

*average of first two months of quarter

For the entire period through August, the originally personal income averaged +3.82%. As revised it averaged +4.55%. For the entire fiscal year ending in September, YoY withholding tax payments averaged 4.3%, much closer to the revised income data than the original. for the first two quarters of the fiscal year, however, the QCEW is much closer to the original income data than the revisions (although note that the QCEW explicitly only covers wages, whereas the D.T.S. includes estimated tax payments which of course cover all sources of income, not just wages).

The QCEW of the 2nd quarter of this year should be reported next month. At that point we will have a better idea of how well that data tracks the revisions to personal income.

In any event, given the major revisions to the personal income data, going forward I will not just pay attention to the employment data in the QCEW, but also the wage data; and I will also report on the 13 week/quarterly average of withholding payments.

 < sigh > more work. For some reason I keep thinking of this old “Far Side” cartoon:

September vehicle sales send mixed message

 

 - by New Deal democrat

One piece of economic data that got reported late last week that is worth following up on is the monthly tally of heavy truck (red) and car sales (blue, right scale):

Both of these are leading indicators, but heavy truck sales are less noisy, and generally signal first (although, like the unemployment rate, they lag significantly coming out of recessions).

In any event, in September car sales were 16.0 million annualized, while heavy truck sales declined sharply in the month to 396,000. Car sales were therefore right in the middle of their range over the last three years, while truck sales are back down into recessionary territory (more than -10% from their recent peak) for the second month in a row. Here’s the zoomed in post-pandemic view that shows the situation a little more clearly:

Without confirmation from car sales, truck sales are not forecasting an imminent recession, but count this as a negative indicator.

Leading indicators from the Setpember employment report continue positive trends

 

 - by New Deal democrat

We’ve entered our post-employment report lull in new economic data, so let’s take a further look at the leading indicators contained in that report. Generally speaking, this includes the more cyclical sectors like construction and manufacturing, and goods production in general; real aggregate nonsupervisory payrolls; and, during expansions leading up to recessions, the unemployment rate (vs. immediately after recessions, when they lag).

Let me start with the last metric first. Almost every week I write that initial jobless claims lead the unemployment rate. Recently I’ve been writing that new jobless claims forecast further downward pressure on the unemployment rate, to the point of declining below 4.0%. Well, in Friday’s report the rate rose 0.1% to 4.2%. So is the forecast busted? 

Hardly. That’s because when we take the raw numbers going into that rate, and extend them another decimal point, they only went up 0.03% from 4.14% to 4.17%, as shown in dark red in the graph below. The lighter, thin orange line is the “official” employment rate, vs. the four week average of initial claims (purple, right scale):

The unemployment rate has drifted higher from its low in July, but the downward trend is intact. There is every reason to expect the unemployment rate to decline further over the next several months.

Next, let’s look at the leading employment sectors. These include manufacturing (gold), residential construction (red, right scale), and goods production as a whole (blue). Also included is nonresidential construction, which includes construction on AI data centers (purple, right scale). All four lines are nomred to 100 as of just before the pandemic:

With the exception of residential construction, the trend line in all of these sectors is up. Even residential construction employment appears to have made a temporary bottom at least in July.

Another leading subsector is trucking employment. This too appears to have bottomed at the beginning of this year and has trended higher since:

Average hours worked in manufacturing is also an “official” leading indicator from the employment report. That increased 0.2 hours in the report to 42.0 hours, not just the highest number since the pandemic, but one of the highest numbers in the entire past 80 years:

Only one month in the 1990s was equal to this number, and most of 2014 and 2018 had higher numbers. In other words, the manufacturing sector is running “hot,” likely in main part due to AI data center related employment. I should point out that the secular increase over the decades probably has much to do with the tilting of the balance of power towards employers vs. employees as part of the general decline of unions that began with Ronald Regan’s presidency. In other words, rather than hire more employees, employers simply work their existing employees harder.

Finally, let’s take a look at real aggregate nonsupervisory payrolls. 

To begin with, average hourly earnings on a YoY basis are only equal to the inflation rate:

If workers are taking home more money in real terms, it would have to be via an increase in hours. So here is the long term historical view of aggregate nonsupervisory hours:

Note this is in log scale to better show the earlier historical data; and also note that with population increase we would expect hours to increase as well. Anyway, what I wanted to point out is that this is generally a coincident indicator, flattening out and peaking within several months before or after the onset of recessions.

Now here is the same data from the past 3.5 years:

Note that since May the number of hours worked has been all but flat. This indicates profound weakness, but such pauses have also occurred in economic slowdowns that have not resulted in recessions, such as in 1966, 1995, and 2016. 

Total hours * wages = aggregate payrolls. Then we adjust for inflation for the “real” number. Here is what they look like for the past 3.5 years, also normed to 100 as of just before the pandemic (ending with August, because September inflation hasn’t been reported yet):

As of August, real nonsupervisory payrolls were up only 0.1% in the 9 months since last November.

Further, the Cleveland Fed estimates that September inflation is likely to come in at +0.5%. So here are the last 12 months of the monthly changes in nominal aggregate nonsupervisory payrolls (blue) and inflation (red):

Last September, both payrolls and inflation rose 0.3%. With payrolls only rising 0.2% in September, if the Cleveland Fed’s estimate proves correct, real nonsupervisory payrolls will decline -0.3%. If that happens, real nonsupervisory payrolls will only be up about 0.65% YoY, the lowest since the pandemic, as shown in the below graph which subtracts -0.65% to show the likely September value at the “0” line:

Not only that, but in the past 60 years YoY increases of 0.65% or less have only occurred in 4 months without having indicated an oncoming or existing recession: once in 1968, once in 1995, and twice in 2011:

We’ll have problems calculating this over the next several months because of the blank October value last year due to the government shutdown, but since CPI only increased 0.3% last November over last September, that averages only 0.15% each month. Should the oil shock from the Iran war continue, it is at least possible that there could be *no* YoY increase in real aggregate nonsupervisory payrolls by December, which would be a strong coincident indicator of recession.

With one important caveat: pace last week’s major positive revisions to personal income, keep in mind that employees’ earnings will be subject to benchmark revisions next February based on the QCEW. And one very good proxy we have for that is withholding tax payments. While those are extremely noisy on a weekly or even monthly basis, once we use the entire past 3 months, they become much less noisy. And for the three month period of July through September, withholding tax payments were higher 5.6% YoY (not shown). vs. 4.2% for aggregate payrolls (blue in the graph below):

This suggests that the last few months, at least, of aggregate nonsupervisory payrolls could be revised higher in the benchmarking process early next year.

Here’s the conclusion: with one important exception - real aggregate nonsupervisory payrolls - all of the other leading indicators contained in the employment report were either positive (e.g., manufacturing and construction indicators) or at least consistent with an ongoing positive trend (the unemployment rate). 

Additionally: as I cautioned last month, leave your ideological priors at the door before evaluating the data. The US economy is extremely resilient, and so far it has found work-arounds for the incompetence and irrationality coming out of Washington.

Economically weighted ISM indexes for October show a broader, but even more inflationary, expansion

 

 - by New Deal democrat

I think if I ever decided to cut back my online writing about the economy to the bare bones, it would be two posts a month. The first of the two posts would be this one, because the economically weighted ISM manufacturing + services indexes continue to be the best timely snapshot of the US economy, and the former in particular has an 80 year history (the latter about 25) of being accurate. The second post would be at the end of the month to see how well the other economic data confirmed this average - and while it would be noisy, I would expect it to be largely on target, especially over a three month period.

With that introduction, last week the manufacturing sector for October showed expansion across the board, with a heavy dose of inflation. This morning the services index was updated to the same effect with the slight exception of employment. A reminder that the weighting, based on their impact on the economy, is 25% manufacturing and 75% services. Further, to cut down on monthly noise, I particularly look at the three month averages.

To the numbers: the headline services index declined -0.5 to 54.9 [recall that any number over 50 indicates expansion]. The three month average was 54.8. Since the three month average for manufacturing was 54.9, the economically weighted average was 54.8, a declne of -0.2 from last month [note: in all of the graphs below, the manufacturing number is blue, and services gray]:

New orders declined -1.1 to a still strong 59.8. The three month average was 59.3. The three month average for manufacturing was 55.2, so the economically weighted of this most forward looking component was 58.3, an increase of 1.1 from September:

The relatively punk metric in these indexes has been employment, which has been in contraction for over a year. This month it improved all the way to slightly above neutral, as the services employment subindex rose +2.3 to 50.1. The three month remained contractionary, however, averaging 48.4. Since the manufacturing employment subindex average was 52.2, the economically weighted average was (slightly) below 50 for the third month in a row as well, unchanged at 49.4:

Importantly, although it has been better than summer of last year, the ISM weighted average has only shown expansion in two months this year - February and June - in contrast with the generally positive monthly jobs reports. It is possible that this difference is because these are diffusion indexes, so if slightly less than 50% of industries are hiring, the ISM average would be negative, while if the net actual hiring was more focused on the expanding industries, the nonfarm payrolls number would be positive. 

Finally, widespread price increases not only continue to be a problem, but they appear to be worsening, as the prices paid index for services rose +1.4 to 74.0, its highest number in over two years, with the three month average rising 2.1 to 72.3. The three month average for manufacturing rose further this month to 73.4 so the economically weighted average increased 0.2 to 72.6:

This has become almost as bad as during the post-pandemic inflation, with the worst reading since mid-year 2022 (note that unlike the other three graphs, this one shows the last five years for better comparison).

In sum, the economically weighted ISM averages indicate that as of now, the economy remains in reasonably strong expansion. Further, the very positive new orders indexes suggest it might get even stronger. This appears to be bringing along employment, which is essentially neutral and on the cusp of turning (slightly) positive. But along with the increased demand, inflationary pressures are also becoming even stronger; and if inflation is becoming even more widespread at the producer level, can consumer inflation be far behind? Maybe I should call this an even more inflationary even broader expansion.

Weekly Indicators for September 28 - October 2

 

 - by New Deal democrat

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

When looking at the economy, you have to put your ideological priors aside. Most of us feel that T—-p is inevitably going to crash the economy, in between his stupendous incompetence and mafia bust-out tactics. But when the data say that isn’t the case, at least not at present, you must listen to the data.

And the data says that the economy is expanding, rebounding from its brush with recession last year. With the very major qualification that the expansion is also quite inflationary (a consequence of the bust-out fiscal policies).

By clicking over and reading, you can bring yourself thoroughly up to date as to all of the gory details, and reward me with a penny or two for collecting and collating the data for you.