Updating the long leading indicators: interest rate levels and duration spreads

 

 - by New Deal democrat

As I noted earlier this week, I haven’t updated my suite of long leading indicators in awhile, mainly because they have been overwhelmed by the economic chaos emanating from Washington, and partly also because several important ones are in terra incognito. On Monday and Wednesday I updated the non-financial elements of the long leading indicators having to do with housing and consumer spending. Today let me focus on one of the financial elements, that has been most puzzling: interest rates.

The interest rate elements of the long leading indicators come in two varieties: the *level* of rates, and the *duration spread* of rates. If we start from the idea that interest rates are “the cost of renting money,” then the former simply means that renting money at a lower rate of interest is always better than renting it at a higher rate of interest. The latter means that normally we would expect to pay a higher rate of interest the more we risk dealing with unforeseen issues further out into the future. If the immediate future is deemed more risky, something is wrong.

The level of rates has typically been measured by corporate bond yields, and there are systems that rely on such rates going back 100 years. In fact we can take the relationship all the way back to the 1850s if we measure by short term corporate paper. So let me start by directing your attention to the red line in the two below graphs, which shows the interest rate on BAA rated corporate bonds, historically from the 1980s through the present:

While it is absolutely true to say that not every significant increase in corporate bond yields presaged a recession, it is true that all recessions (except for the brief pandemic lockdown) were presaged by an increase in those rates.

Now let me explain the blue line. It is a similar measure, but for Treasurys, using the average of 3 month, 2 year, 5 year, and 10 year notes, which captures short, medium, and longer term rates. Like corporate bonds, there were significant increases in the average Treasury interest rates before each recession - although note the sharp increases in 1994-95 did not presage a recession.

Now let’s focus on the last few years of the same data:

After the very sharp interest rates increases of 2022 into 2023 - which did not presage a recession (at least not in the ensuing 3 years!), both corporates and Treasurys have meandered within relatively narrow ranges — the former between 5.4% and 6.4%, and the latter between 3.6% and 4.6%. Corporate bonds are currently almost right in the middle of that range, while Treasury rates at the short end have contributed to a downtrend in those rates which has not really been broken by the increase that started with the onset of the Iran war.

In short, the *level* of bond rates is not sending a negative signal at this time, but rather “neutral.”

Now let’s turn to the duration spread or yield curve measure.

Here is the historical graph from the 1980s until the pandemic of the 10 year (blue), 2 year (gold), and 3 month (red) Treasurys, together with the Fed funds rate (purple):

In general, throughout most of these economic expansions, the longer the duration of the government note, the higher the interest rate it commanded. Further, not every significant increase in the Fed funds rate (1984, 1994) gave rise to an inversion. But in 1989, 2000, and 2006 there were across the board inversions, where in response to the Fed hiking rates, the shorter the duration of the note, the higher the interest rate was demanded. And sure enough, recessions followed. Further, those recessions ended about 4 to 8 months after rates were fully normalized.

Now let’s turn to the post-pandemic record:

We had a nearly full inversion of the curve in 2023-24, with only the 3 month note paying slightly more than the Fed funds rate. Both widespread models for recession forecasting, featuring the 10 year minus 2 year, and 10 year minus 3 month spreads signaled “recession ahead.” 

But the curve started normalizing in late 2024 (as to the 10 year minus 2 year spread), and then in 2025 (as to the 10 year minus 3 month spread). The normalization was completed earlier this year when both the 2 year and 3 month spread turned higher than the Fed funds rate. 

Based on past experience, that is a positive signal for roughly year end 2026 and beyond.

But there is one wrinkle, because as I pointed out about a month ago, this is the first time in history that a yield curve has normalized by Treasury note rates moving *higher*, rather than following a decline in the Fed funds rate *lower.* This is terra incognito.

There is one somewhat similar episode from the past which ought to indicate caution is still very much warranted. Here is the smae graph focusing in on the Great Recession:

Note that the yeild curve started to normalize in late 2007, and was almost completely normalized (with only the Fed funds vs. 3 month spread remaining inverted) by spring 2008. Further - and very similarly to right now - part of that normalization was the increase in interest rates in the 3 months and 2 year notes. And then the investment banks imploded.

Just like now, in spring 2008 there was a gas price spike, from roughly $3 to $4.25/gallon:

It was this inflationary pulse which caused interest rates to increase. That finally broke the proverbial camel’s back.

While the current gas price spike is not as severe either in terms of GDP or average wages as the one in 2008, there is nevertheless an inflationary pulse in the economy. So while the duration spread indicator is positive, the uniqueness of its regularization warrants a big fat asterisk.

Jobless claims: still the most positive datapoint in the entire US economy

 

 - by New Deal democrat

Let’s take our weekly look at jobless claims, recently perhaps the most positive datapoint in the entire economy.

And it continued to be so this week. New jobless claims rose 6,000 to 214,000, while the four week moving average rose 750 to 210,750. With the typical one week lag, continuing claims rose 12,000 to 1.821 million:

As a refresher, here is the entire 50+ year series of both new and continuing claims until the pandemic, with current values normed at the 0 line to show just how historically low the current levels are - and that doesn’t even take into account that the US population has doubled during that time:

Initial claims in particular were never this low during the entire period from the early 1970s until just before the pandemic. Basically, right now if you have a job, there is very little chance of your being laid off.

As per usual, the YoY levels are more important for forecasting purposes. As a quick refresher, that’s beause recessions have started and ended at very different absolute levels of jobless claims, whether or not adjusted for population. But they have a relatively consistent pattern of being likely once initial claims have risen more than 12.5% YoY and stayed elevated above that level for 2 months:

As of this morning’s report, initial claims are lower than -4.5% YoY, the four week average is lower by -4.6%, and continuing claims are lower by -1.0%:

Jobless claims are very much forecasting continued economic growth over the next several months.

Finally, I almost always point out that jobless claims lead the unemployment rate. Here is the historical YoY look for the 50+ years before the pandemic:

And here is the YoY comparison of both initial and the smoother but less leading initial+continuing claims with the YoY% change (note that is a % of a %) in the unemployment rate for the past several years:

Jobless claims are forecasting continued downward YoY pressure on the unemployment rate, which was 4.2%-4.3% last spring. IN other words, jobless claims are forecasting the unemployment rate to decline below 4.2% to 4.1% or even 4.0% in the next several months.

So we leave off where we began: jobless claims are currently the most positive of any US economic data being released.

Updating the long leading indicators: corporate profits and housing

 

 - by New Deal democrat

This week I’m spending some time updating the non-financial long leading indicators. As I wrote on Monday, I haven’t updated these in a while in large part because the chaos coming out of Washington has blindsided much of the economic data. There’s not a big point in obsessing over what the yield curve of US Treasurys in the bond market might mean for the economy in 2027 when the Commander in Chief barges in and says, “Hey, let’s see what happens when Iran shuts the Strait of Hormuz!”

On Monday I looked at real spending and consumption per capita. Basically, with some noise these have been trending sideways since last summer, and are hovering around 0% gains YoY. Yesterday we saw that gas prices had caused consumers to spend significantly more in the aggregate than they had in the past months, mainly to fill up their gas tanks. Excluding gas, “real” sales were still below last August’s peak.

There are two more nonfinancial long leading indicators, one on the production side and one that is an important interface between production and consumption.

The long leading indicator on the production side is corporate profits deflated by unit labor casts. There have had a long history of turning down one year or more before the onset of recessions. Below I show them both including (dark blue) and excluding (light blue) inventory and capital spending adjustments. I also show proprietors’ income (red), which is almost as leading but has the virtue of being reported at least one month before corporate profits as part of GDP:

As you can see, after virtually stalling during 2024, and then being whipsawed by the front-running of tariffs and the subsequent payback, in the last two quarters of 2025 both measures of deflated corporate profits rose sharply. Proprietors’ income did turn down slightly, and although that bears watching, it is not as important. 

A more up-to-date measures is the weekly update of reported and estimated corporate profits as they will be reported to Wall Street, from Earnings Insight. Here is the most recent update from last week:

Q1 profits have just begun to be reported. Typically companies massage their earnings expectations so that they “beat” them in the report, so I anticipate the current anticipated Q1 level will improve by at least several points. In any event, the weekly updates confirm that Q3 and Q4 saw stout growth in corporate profits but a possible pullback in Q1.

Nevertheless, deflated corporate profits are a long leading positive as of the end of 2025.

The other nonfinancial long leading indicator, which captures an important producer and consumer interface, is housing permits. Below I show total permits (light blue) along with the much less noisy single family permits (dark blue), and also mortgage rates (red), which largely - but not exclusively - drive much of the changes in homebuilding:

In general we can see that for the past two years housing permits have gradually trended downward as mortgage rates oscillated between 6% and 7%.

The relationship is best shown in the YoY% changes of the same data, below (with mortgage rates inverted better to show the relationship):

Mortgage rates have been close to stable for most of the past year with a few positive oscillations (before the March spike secondary to the Treasury market’s reaction to the oil price spike). Building permits on a YoY basis have continued to fade slightly.

As I have written many times in the past year, the downturn in the housing market at this point is frankly recessionary.

To sum up, of the three non-financial long leading indicators I have looked at this week, one - corporate profits - is positive. One - building permits - is negative. And one - real spending per capita - is close to being flat. My suspicion is that if the oil price spike persists for another month or two, real spending per capita is going to turn down. At that point we will see just how much of a burden the AI data center construction driven improvement in corporate profits can carry all by itself.

Housing permits will next be updated on April 29th, and proprietors’ income on April 30 as part of the first estimate of Q1 GDP.

March retail spending: including or excluding gasoline makes all the difference

 

 - by New Deal democrat

Consumer spending is about 70% of the economy, and retail sales is our first wide measure of that spending. This morning’s update for March (still after all these months delayed about a week compared with the schedule before last autumn’s government shutdown) was unsurprisingly dominated by what happened at gas stations. 

Nominally, total retail sales rose 1.7% in March, and February was revised slightly higher to up 0.7%. After taking the monthly 0.9% increase in consumer prices into account, real sales increased 0.6% (blue). Because of the outsized impact of gas prices this month, the below graph also shows real sales ex-gasoline, which *declined* -0.3% (gold):

As is obvious, including vs. excluding gasoline sales makes a big difference. Including such sales, real retail sales were the highest since the end of 2022. Excluding them, however, they remain below their peak from last August. 

Interestingly, the same phenomenon occurred in the 20006-07 period before the Great Recession, where total real retail sales trended slightly higher, but excluding gasoline trended lower:

This can (and I think, should) be read as consumers cutting back on other purchases in order to buy gas for their vehicles. While March is only one month, if this continues several more months it would be sending a significant warning signal.

On a YoY basis, nominal total retail sales were up 4.0%, and in real terms up 0.7%. Excluding gasoline, YoY were up 2.9%, and in real terms *down* -0.4%. Further, if you believe, as I do, that the shutdown shelter kludge removed about 0.2% from consumer inflation during the September-October period, which is still affecting the YoY calculations, then the situation is similarly weaker:

Following up on yesterday’s post, per capita real retail sales have been a very good but not perfect long leading indicator. Since the populaiton estimates have not been updated through March yet, I can’t show you an updated graph, but in the past few months population gains have been just barely above 0, so the population-adjusted changes for March would be almost identical to the graphs above.

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

This suggests that on a YoY basis employment is unlikely to deteriorate further over the next several months. Since there were positive numbers last year from February through May, this suggests there will also be gains - if small ones - in the next several months. 

To sum up, March’s spike in gas prices did cause consumers to reallocate spending slightly, but not to cut back retail spending in the aggregate.

Updating the long leading indicators: per capita real retail spending and real spending on goods

 

 - by New Deal democrat

So much of the changes to the economy has been dictated by the whims of the White House that it has not made much sense to throughly update the suite of long leading indicators in many months. After all, why chart their incipient effects when they are consistently being overtaken by new lurches in diktats?

With a few empty days of new economic data this week, I thought I would take a look at a few of them; in particular, those most closely tied to the “real” vs. financial, economy. Today, let’s revisit and update real sales per capita.

Real retail sales per capita have long been an element of my list of long leading indicators, because they typically turn down about a year before the actual onset of recessions. In the past several years I have added the very similar indicator of real consumption expenditiures on goods from the personal income and spending report. Like the less broad retail sales measure, their growth either turns down or at least stalls out many months before the onset of most recessions.

Since the data on both these metrics goes back many decades, I have split the data by historical sections. First I look at it in absolute terms, and then we’ll look at it YoY, which better shows the turndowns.

First, here are real retail sales per capita have (red) and real spending on goods per capita (blue) from 1959-82, 1983-2008, and 2009-present:

While both metrics are somewhat noisy month over month, it is obvious that both measures of consumer spending increase almost continually throughout expansions, but either turn down (real retail sales) or at least stall (real spending on goods) typically on the order of a year before the onset of recessions in the past 60+ years. There have only been two exceptions: 1966 for real retail sales, and 2022-23 for both metrics. 

Why did the indicators fail in 2022-23? Mainly as an artifact of comparison. The stimulus payments to consumers in 2020 and 2021 during the pandemic led to a lot of durable goods purchases in particular, often goods delivered to homes. Simply put, for some time consumer demand for goods, especially durable goods, was largely satiated. Secondarily, the big decline in the price of gas from $5 to $3 a gallon beginning in July 2022 freed up lots of consumer cash to be spent elsewhere, thus avoiding a consumer-centric economic downturn. 

Note that in the past year both metrics have essentially stalled. And insttead of a big decline in gas prices, consumers have been faced over the past six weeks with a big incrrease.

Now let’s look at the same historical series YoY:

Again, note how positive YoY improvements in real consumer retail spending, and spending on goods, have always correlated with continued economic expansions. With the exception of 1966, the only negative comparisons which did not correspond with the start of or oncoming recessions were sporadic months in 1987, 1994, 1996, and the near “double-dip” of 2002-03.

Now let’s look at the post-pandemic YoY record:

Again, we see the poor 2022-23 performance, which was a major false negative resulting in large part from poor comparisons compared to 2021. While the poor performance of real retail sales continued into 2024, it was not confirmed by the broader measure of real spending on goods. Both decelerated sharply in 2025, and both were negative during December. Which means they aren’t signaling an oncoming recession now, but the trend is not good. And the YoY comparisons will be challenging for the next several months: for retail sales, February was already -0.5% below last March, and -0.2% below last April.

Real retail sales will be updated for March tomorrow, and real spending on goods will be updated on the 30th.