Rethinking the long leading indicators

 - by New Deal democrat

I’ve been writing about the economy, and employing forecasting models, for over 20 years. For the entire duration of the period, there have been portents of DOOOM written about by many others. I have been much more cautious, going on “Recession Watch” only twice since 2008: in 2019 and late 2022. I have never gone on “Recession Warning.” 

Which is broadly concurrent with the economy since then. Since June 2009, the economy has been in recession for only the 2 month period immediately after COVID hit and brought society to a near standstill. For the other 200+ months, it has been expanding.

We’ll never know if, absent COVID, there would have been a recession in 2020, although at the time I believed we were going to narrowly miss it. But in 2022, the broad warning signs were manifest - and yet no recession occurred anyway.

So, after 20 years, I think it’s time to examine whether the broad range of long leading indicators hold up. I make use of the 4 identified by Prof. Geoffrey Moore in the 1980s (and subequently used by ECRI), plus common measures of the yield curve, and also real retail sales per capita.

Let me start with Prof. Moore’s four components [Note: all of the below graphs use YoY% comparisons for easier viewing; although the models generally use the absolute measures]. Here are corporate bonds, corporate profits deflated by labor costs, real money supply, and housing permits, covering the periods of 1960-94, 1995-2026:

Although the data is very noisy, when we look for periods when *all 4* components were at or below 0, the only such times were roughly 1 year before the onset of each recession, plus 1966, as well as the 2019 and 2022 periods. Plus last year, as shown in this close-up

Between December 2024 and May 2025 all four were negative. Again, although we came close last year, no recession has occurred as of June 2026.

Next, let me compare corporate bond yields as above with the 10 year Treasury minus 2 year yield, calculated as the YoY change (thus, for example, if the spread declined form +0.40% to +0.20%, this shows up as a -0.20% change). Here’s the entire period from the mid-1970s to the present:

While both measures are negative before the onset of recessions - frequently reverting to positive immediately beforehand as the Fed lowers interest rates - there are a number of false positives as well, in 1984, 1994, 2016, and very much so in 2022. The message I take away from this is that interest rate changes are probative, but they give too many false positive recession signals. Meanwhile, housing measures, while also probative, focus on too narrow a slice of the economy.

So let’s turn to the broad “real world” long leading indicators, comparing corporate profits and real retail sales per capita. Here’s what they look like from 1953-2000, 2001-2019, and 2020-present:

I’m much more satisfied with this measure. The only times both measures (using a 3 month moving average for real retail sales per capita) have been simultaneously negative has been shortly before recessions have begun, with the false positive of 1966 and one quarter in 2022.

And those two misses have something in common: massive fiscal stimulus. In 1966 LBJ’s “guns and butter” budget poured massive amounts of spending into both military spending and domestic social program spending. In 2021, the COVID stimulus had led to an abrupt 15% increase in retail spending, but with an absolute lower level of employment. Profits boomed, then paused, but clearly could - and did - boom further as businesses amped up production and hired more employees as the COVID supply bottlenecks unspooled.

This is an intellectual work in progress, so there is much more to think about. But preliminarily, my takeaway is that while interest rates and real money supply are important background conditions, that is all they are. The “real world” early indicators from housing, corporate profits, and real consumer spending per capita are necessary for confirmation. And I need to take a more detailed look at fiscal conditions and price level shocks, which often seem to be precipitating elements for recessions.

Preliminary evidence that both business expansion - and widespread inflation - have continued in June

 - by New Deal democrat

 Yesterday, in addition to the retail sales report, general business sales and inventory were reported - but unfortunately only through April. Nominally, sales (red in the graph below) increased 1.2% while inventories increased 0.5%:

Here is the longer term look at total business sales and inventories:

In case it isn’t obvious from the above graph, sales turn both higher and lower before inventories, and the signature of an oncoming recession is sales having turned down while inventories are still increasing. Since sales were up through April - and would be even if we adjusted for inflation using either the PPI or CPI - that confirms that the economy was expanding - 2 months ago.

But the early indications from the New York and Philadelphia Fed manufacturing surveys are that manufacturing activity has continued to expand through June, but inflationary pressures are still elevated as well.

Here are the averages of the two manufacturing indexes for general activity (blue) and new orders (red):

Both showed expansion. To cut down on noise vs. signal, I recommend using the three month average - which for both metrics remained stable at a moderate expansionary pace.

Here are the same averages for prices paid (blue) and received (red):

Both of these continue at more widespread levels than during the pre-pandemic expansion, although not as widespread as during the immediate post-pandemic inflationary period. But perhaps most importantly, both prices paid and received continued to show widespread increases this month, indicating that the inflationary shock from the Iran war has not ended.

More evidence for the re-emergence of residual seasonality in jobless claims, but still very positive

 - by New Deal democrat

 From 2023 through midyear 2025, there was a distinct pattern of unresolved post-pandemic seasonality to jobless claims, which rose in the first half of the year, and then declined in the second half. Beginning at the end of June last year, though, there was a “change of regime in jobless claims numbers,” probably related to the collapse of immigration and/or fear in immigrant communities, resulting in significantly lower claims on a YoY basis. In the last few weeks there have been signs that the post-pandemic seasonality may be reasserting itself, so as I wrote last wek, “it will be interesting to see if the negative YoY comparisons continue, or if they fade away. If the change of regime was a one-time thing, driven mainly by immigrant worker issues, then these good YoY comparisons will fade between now and the end of July.”

To cut to the chase, the good YoY comparisons have not started to fade yet.

On a weekly basis, initial claims declined -4,000 to 226,000, while the four week moving average rose 4,000 to 223,250. With the typical one week delay, continuing claims rose 24,000 to 1.810 million. Here is the look since the beginning of 2023:

The close-up since the beginning of 2025 better shows how there has been a significant increase in claims in the past three weeks, to levels equivalent to those seen last July through December:

So the question is: do claims stabilize here, or start to decline again in July?

As usual, the YoY% comparison is more important for forecasting purposes, and so measured, initial claims were down -7.0%, the four week average down -7.8%, and continuing claims down -6.5%:

These continue to be very positive signs for the economy.

We’re far enough along in the months that we can extend our comparison with the unemployment rate:

Despite the increase in jobless claims in the last few weeks, there is every reason to believe that the unemployment rate, which follows claims with a lag, will decline further in the next several months.

Even adjusting for gas prices, consumers went on a (wealth effect- generated?) spending spree in May

 

 - by New Deal democrat

Consumer spending is about 70% of the economy, and retail sales is our first wide measure of that spending. And since consumption leads employment, it is an important real world measure. In May, after two months of being dominated by gas prices, it was even more decisively driven by a likely splurge tied to the wealth effect from Booming stock prices.

Nominally, total retail sales rose 0.9% in May, but after taking the monthly 0.5% increase in consumer prices into account, real sales rose 0.4% (blue):

Since gas prices have been a major driver of inflation in the past few months, here’s a look at the monthly % changes in nominal retail sales excluding gas stations (orange) vs. total retail sales (blue). Retail sales excluding-gas increased 0.7%:

Since CPI increased 0.5%, and excluding energy rose only 0.2% in May, both of these are positive in “real” terms as well.

On a YoY basis, nominal total retail sales were up 6.9%, but in real terms were only up 2.6%. Excluding gasoline, nominal sales YoY were up 5.4%, and deflating by using CPI excluding energy were up 2.5%:

In both the March and April personal income and spending reports, we saw that consumes handled gas prices increases by essentially just putting the extra spending on their credit cards. Last month I suggested that consumers might have switched to a “wait and see” mode, but this month’s report indicates that consumers have been spending with wild abandon - as has been indicated for weeks by the Redbook sales report, the four week average of which as of this week is up 9.1% YoY nominally!:

This isn’t because wages have been increasing sharply. Quite the contrary, as we saw with the employment report. Rather, it is *very* likely that this is “wealth effect” spending by upper income consumers triggered by the near 20% rise in the stock market since the end of March.

Finally, since consumption leads employment, here is the update of YoY real retail sales (/2 for scale) together with employment (red):

This suggests that on a YoY basis the rebound we have seen in the last three jobs reports is likely to continue in the next several months. Which is all good news, provided the stock market wealth that is likely driving consumption reflects a Boom rather than a bubble.

Housing continues in the doldrums, but its decline failed to give rise to a post-pandemic recession

 

 - by New Deal democrat

When it comes to housing, the post-pandemic economy has been odd. That’s because, as of last year by all accounts housing had deteriorated  sufficiently that a recession should already have begun. Not only did that not (quite) happen, but the current situation would be more congruent with such a recession ending than ongoing. In general, that’s because even though job creation stagnated, real incomes did not, and real consumer spending was sufficient to keep the economy in expansion.

In the past few months, the situation has reversed, with real incomes declining, but manufacturing and job creation rebounding. And housing is playing a role in that conundrum as well. Let’s take a look.

In May, almost all of the headline housing data was negative. Starts (blue), which are noisier and slightly lag permits (gold), declined sharply, down -215,000 to 1.177 million units annualized. Permits declined -10,000 to 1.413 million. The one slightly bright spot was that the metric which is the least noisy as well as being most leading, single family permits (red, right scale), rose  5,000 to 886,000: 

Permits, including single family permits, were near five year lows, but close to average readings for the past year. But housing starts had their worst month in over 5 years! But because they are so noisy, I am discounting that. Still an “average” down in the doledrums month is definitely not positive.

Perhaps more interesting is that the recent uptick in mortgage rates (blue in the graph below) due to the Iran war has not really had an effect on the sideways trend:

Still, on a YOY% basis, starts are down -8.7%, while permits are down -0.2% and single family permits are down -7.1%:

This is a sector that, to repeat, is down in the doldrums, and shows no sign of improvement.

On the other hand, as I have noted for the past several months, the YoY downtrend has not been worsening for many months. Typically permits and starts have been down 20% or more at the onset of recessions in the past, although in the 1991 and 2001 recessions, they were only down about -10%; and there have been a number of times, for example 1966, 1987, and 1995, where construction has been down -10% or more YoY without a recession occurring:

On the contrary, the negative but relatively minor and stable negative YoY changes beginning last year have been just as consistent with mid-expansion slowdowns as with recessions, and stable if negative YoY changes have sometimes occurred during recessions a few months before recoveries.

But the statistic in this release that most closely aligns with the actual economic activity represented by the sector is the number of housing units under construction. In May they also declined, by -11,000 to 1.266 million annualized, only 4,000 above their post-pandemic low in March, down -7.1% YoY, and down -26.1% from their peak:

The historical version of this metric shows that more often than not in the past by the time a decline in units under construction had declined by this much, a recession had already begun:

Indeed, typically this metric does not turn upward until after the recession has been over - part of the current conundrum.

Finally, let’s update the graph that additionally shows the typical last shoes to drop before recessions, including houses for sale (gold) and residential construction employment (red, right scale), all normed to 100 as of their respective post-pandemic peaks:

Residential construction employment has been close to flat for the past nine months, and the number of housing units for sale actually rebounded slightly as of its last report for April.

It’s fair to say that the housing market did not follow its typical predictive pattern following the pandemic, and the big increase in interest rates that began in 2022. Although all of the relevant housing metrics fell by percentages from peaks that normally signaled recessions in the past, that did not happen this time. In large part that was because of the huge backlog in building that meant that units under construction did not significantly decline until the mortgage rate and manufacturing situations in the economy stabilized. The latter in particular began to improve substantially beginning in the latter part of 2024, despite the subsequent headwinds of new tariff costs imposed last year. And the continued increases in real incomes and payrolls through 2025 meant that the 75% or so of the economy that is the service sector continued to be fueled by consumer spending. 

But housing is also not contributing positively to the economy, either. That means that if the recent downturn in real incomes continues, the rest of the economy may yet go down.

Goods production sector of the economy remained in expansion in May

 

 - by New Deal democrat

Late last year I noticed that the regional Fed manufacturing indicators were improving - despite the “Liberation Day” tariffs and the general chaos coming out of Washington. It appeared that manufacturers had found a modus vivendi and had adapted to the new environment. That subsequently showed up in a number of manufacturing related statistics, like durable and capital goods orders, and even in manufacturing employment.

It also showed up - as did the Boom in AI data center construction - in industrial production, which was updated this morning for May. I used to call Industrial Production “the King of Coincident Indicators,” but with the shrinking of manufacturing as a share of the US economy to about 25%, that is no longer the case. Nevertheless, it is an important coincident indicator, with the emphasis on *coincident.* It doesn’t tell us where we are going, but is an important signpost about where we *are.*

In April manufacturing (red in the graph below) broke out of that range to the upside, increasing 0.6% for the month, while the more noisy utility production (gold, right scale) increased 1.9%, driving total production (blue) to a 0.7% increase and a new post-pandemic high. May total prodcution and manufacturing continued to be positive, although much more subdued. Total production increased 0.1%, manufacturing production less than 0.1% rounded to 0. Meanwhile utilities declined -0.4% [Note: all values are normed to 100 as of September 2018, the peak month for production in the last expansion]:

On a YoY basis, manufacturing production was up 1.7%, manufacturing up 1.4%, and utilities up 3.1%:

There is no sign yet of any break in the recent trend in any of these indicators. Simply put, the goods producing sector of the economy remained in expansion in May.