The recession is right on track, according to today's employment number. Employment is still following the 7-recession average, dated from the time of the first 10y-3m spread inversion in each episode. Peak would come in November.
The Kantro indicator is saying a recession already started, the Sahm indicator says it's at least close. Neither has missed a recession since 1970. Important to keep track of the NBER 6 data as they are released.
After 7 months of yield curve inversions,
#3
/3: give it time. The last 8 recessions have started on average 12 months after the first yield curve inversion. In one case (2008), it took 17 months.
The Sahm and Kantro recession indicators are closely related but they process the data in somewhat different ways. Neither has missed a recession since 1953 and they have not given any false positives since 1970.
BTW, you may recall there's another recession indicator with a perfect record (at least so far) since 1970. This signal comes on average 15 months before Sahm and Kantro.
Market commentators are taking solace in the fact that initial claims released today are the lowest since Sep 2022. As you celebrate, bear in mind that this is a lagging indicator that peaks at the ends of recessions and is typically low before they start.
2008-09 was the worst recession I forecasted in real time. Again, cries of “What recession?” until it was well underway. As late as June 2008, most of the FOMC thought the economy would “skirt recession” even though it had started 6 months before. Go figure.
Historically, the yield curve inverts 6 to 17 months BEFORE a recession starts and the Kantro signal comes 0 to 7 months AFTER it starts. Having both now, the recession is indicated to have started in the intersection of the shaded ranges. TBH I'd like to see a bit more data.
Did a
#recession
start in December 2023? Exhibit 1: Employment (household survey) has fallen 3 straight months since November. This happens very rarely unless it’s connected with a recession as in 1980, 1982, 1990, 2001, 2008 (2020 recession was too short).
With today's employment report, the Kantro recession signal from April is confirmed, while the Sahm indicator moved sideways and has not yet presented a signal. Recall that both are lagging indicators of the start of a recession.
MONTHLY RECESSION WATCH: For August 2023, the 10yr-3mo Treasury spread averaged -1.28%, inverting for the 10th straight month. History suggests that an impending recession will last at least that long and that the probability that it will start within the next 12 months is 99.4%.
The stock market tends to reach an all-time high shortly before the start of recessions. That's not the only time it happens, but setting a record is not necessarily a sign of good things to come in the economy.
MONTHLY RECESSION WATCH: For January 2024, the 10yr-3mo Treasury spread averaged -1.30 for a 99.5% probability that a recession will start by January 2025. I know some of you are impatient but the 2008 Great Recession came 17 months after the first inversion.
Did a
#recession
start in December 2023? Exhibit 2: Monthly employment (household survey) peaked in November 2023 and its recent track resembles its behavior around the last 6 NBER peaks, which marked the starts of recessions.
What does it take to be a yield curve denier? (Plenty around lately.) It seems that first and foremost you need to be unwilling to take a look at even the simplest data (eg, below) or to be willing to ask others if they’re going to believe their lying eyes.
This week is the second anniversary of my first tweet. I joined because I thought the Fed would tighten considerably, causing a yield curve inversion and later a recession, and I wanted to share my tracking in real time. Tightening and inversion happened. Waiting for
#3
.
The unemployment rate may take some time to react after a recession starts, which may explain why it’s not in the NBER 6. At this point, there seems to be a bit of a pickup in the rate since the yield curve first inverted, somewhat ahead of the timing of the 2008 recession.
Is the Sahm rule ready to pop? The unemployment-based indicator rises above .5% within 7 months AFTER the start of every recession. With April at .37%, what are the chances of a signal? After rising to the .3-.4% range 3 times in the past, it has always breached .5%.
With the 10yr rate climbing fast, can a yield curve inversion end because the 10yr overtakes the 3mo? Yes. It happened in 3 of the last 8 inversions, 1990, 2007 and 2019. Did that forestall a recession? No. Recessions followed all 8 times. If past is prologue, the die is cast.
In a couple of days, we're going to see 20 consecutive months of yield curve inversion for the first time in the data sample going back to 1962. Of the NBER 6 real-economy indicators used to call recessions, only payroll employment hasn't pulled back so far.
Why is it better to look at the 10yr-3mo yield curve spread than at the 10yr-2yr? Inversions of the 2yr spread are shallower before recessions, even nonexistent before 2020. For the last 6 recessions, the 3mo signal was clearer (only 1990 was a tossup). Statistical tests confirm.
What do soft landings look like through the lens of the Sahm recession indicator? In both 1986 and 1995, the indicator went about halfway to the .5% critical signal level but then backed down. In 2024, it’s higher and getting closer to a signal. Will it back down?
The unemployment rate is a lagging indicator because it tends to increase only after an economic peak, ie, once a recession starts. Unemployment just waffled before the last 6 recessions (like the last 7 months) and only clearly rose after the cyclical peak, with varying lags.
A wise man once warned us that it’s tough to make predictions, especially about the future. In the last few months before a recession starts, the unemployment rate usually looks pretty good.
Nonfarm employment since the yield curve inversion last November has been tracking the average of previous inversion episodes very closely. The current July number is a bit lower than the average of 7 inversion periods, all of which led to recessions.
Fed staff is apparently no longer forecasting a recession. That would be great. It would be historically unprecedented, though, given 9 months of the deepest yield curve inversion since 1981. Let's hope they're right. Time will tell.
Employment growth (establishment survey) has remained solid so far after the yield curve first inverted. And yet, it still trails 2 of 8 previous pre-recession periods and is around the point where it peaked in 3 of the 8.
MONTHLY RECESSION WATCH: For July 2023, the 10yr-3mo Treasury spread averaged -1.49%, the 9th straight negative monthly reading. The probability that a recession will start within the next 12 months is 99.8%.
It's safe to say that the average yield curve spread for January will be negative. That makes the inversion 15 months long, which is the second longest since 1968. Inversions of 10+ months have preceded the 4 longest recessions in the period.
Total nonfarm employment tends to track the start of recessions more closely than other variables such as industrial production. This time, employment is right on track for a recession whereas IP is below average.
MONTHLY RECESSION WATCH: For February 2024, the 10yr-3mo Treasury spread averaged -1.15%, for a 98.9% probability that a recession will start by February 2025.
March will be the 17th month of the current yield curve inversion, now in clear second place for the longest since 1968. Implications? For one, long inversions have historically been followed by long recessions. Do you feel lucky?
I showed this before for employment and results are similar for real GDP. GDP keeps rising on average for 3 quarters after yield curve first inverts, then recession starts on the 4th. The number this morning is just the 2nd, so we have 2 more to go.
The yield curve inversion continues to be very impressive by recent standards. A soft landing is not impossible, but it would be very surprising given the last 56 years of history. In any case, we're destined to live in interesting times.
The BLS household employment survey counts workers, the payroll survey counts jobs. For those with more than 1 job, the total number of jobs is included in payroll. This double/multiple counting explains some of the recent drop in the difference between the 2 survey results.
This time, employment is following the normal lags but inflation did not get the memo. The sharp drop in inflation is coming way too early to be attributable to monetary tightening. Policymakers need to look elsewhere for the cause, which should influence their decisions.
Big surprise in nonfarm payroll employment. In terms of historical recession tracking, though, not as much. The level continues to track well within the middle of the distribution, measured from the first month of yield curve inversion. More tracking to do...
What’s the first thing you need to check if the yield curve signal was correct this time? Patience! We know that each of the last 8 recessions has started 6-17 months after the first inversion. So when people say “It’s been 8 months and nothing yet,” you know what to do.
The last time the 10-year Treasury rate was as high as today's close was 10/15/2007. At that time, a yield curve inversion had recently ended and the Great Recession was about to start. At present, the inversion continues (-81 bps). Probably not a good sign.
MONTHLY RECESSION WATCH: For December, the 10yr-3mo Treasury spread continued to fall once more and averaged -1.31%, consistent with a 99.5% probability that a recession will start in 2024.
October nonfarm payroll employment continues to follow the pre-recession yield curve inversion average very closely. We're 2 months away from where the average turns down and 7 months away from where the longest increase turned down (Feb 2008).
I know some are thinking, what’s with all the high probabilities and still no recession? If the yield curve hadn’t been right 4 straight times in real time, I might be tempted to agree. In 2006, I had to wait 17 months for the Great Recession, even longer for the NBER to call it.
If we ignore history, are we condemned to repeat it? September employment was up 336k. Two months before 2020 recession, it was up 334k. Two months before 1973 recession, up 331k with a much smaller labor force.
In fact, the number of multiple job holders in the BLS employment survey is at a record high since they started collecting this statistic in 1994. Multiple job holding is probably not a good sign for the state of the economy.
How will we know when the recession starts? The NBER lists a bunch of variables they track and they’ll make an announcement well after it starts. Till then, one variable I’ll keep in the crosshairs is total nonfarm employment, which generally peaks around the start.
Nonfarm employment from the payroll survey continues to astound. While every other one of the NBER 6 recession indicators is showing signs of weakness, this one just keeps going up. Needs a closer look.
FOMC M.O. around the last 3 recessions: Tighten gradually until the recession is baked in, start easing before it sets in, sharp easing once the recession starts. None of the 3 avoided, 2 severe. This time: tightening not so gradual, no easing yet. Outcome TBD.
Given the strong jobs reports today, is a recession off the table? Don't give up on it just yet. It took 17 months from first inversion for the 2008 recession to start, which corresponds to April. Indicators are still within bounds from the last 8 recessions. Keep on tracking!
Household survey employment is weaker than payroll. It peaked last November and is more in the ballpark of the previous 8 recessions. Why so different?
If you expect stock prices to drop in anticipation of the recession, don't hold your breath. It happened before the 2001 recession but in most cases the drop comes during the recession itself.
How do recessions start? The last 8 have started with monetary tightening that inverts the yield curve. Sure, economic and financial conditions were different for each one, but this one factor was a constant every single time.
This time is different. It always is! Last 8 recessions: 2020 covid, 2008 subprime + cds, 2001 wtc, 1990 gulf war + s&l, 1980-1981 stagflation, 1973 opec, 1969 post-vn fiscal. Only one constant: prior yield curve inversion.
MONTHLY RECESSION WATCH: For November, the 10yr-3mo Treasury spread averaged -.96%, which is consistent with a 97.3% probability that a recession will start by November 2024. Of the previous 8 inversions, 4 lasted longer than this one before a recession started.
Today’s total nonfarm employment figure continues to track just below the historical pre-recession average. Among the NBER 6 monthly indicators, this is the single variable that best coincides with NBER recession dating.
Even with the upward blip in July CPI 12-month inflation, it's still running well below experience in every yield curve inversion episode 1968-2018. Monetary policy is not driving disinflation, at least not yet.
The 1990-91 recession was the first one we forecasted in real time using the yield curve. We heard “What recession?” so much until it was well underway...
Is possible that there will be no recession within the next 12 months? I can only estimate probabilities, but it would be the first time in 55 years that a yield curve inversion is not followed by a recession within that time horizon. Do you feel lucky?
What does it take to be a yield curve denier? (Plenty around lately.) It seems that first and foremost you need to be unwilling to take a look at even the simplest data (eg, below) or to be willing to ask others if they’re going to believe their lying eyes.
After forecasting the last 4 recessions in real time, I came back each time and did a post-mortem review of the data. Deniers always seem uninterested in that exercise. Granted, it can take a while to gather the data and you might not like the findings.
Real personal income was down in February. Looking at earlier pre-recession periods, it's too early to tell if this means that it peaked, but it could be significant. The level relative to first inversion now matches the 2008 recession.
The household employment survey shows a very different track from the establishment figures. February is the third straight month with a decrease and the series is trending toward the middle of the pack among pre-recession yield curve inversion periods.
The 2008 recession took long to develop. In Nov 2005, chair Greenspan told Congress my model wouldn’t work. In a Mar 2006 speech, new chair Bernanke explained why it wouldn’t. In Aug 2006, the yield curve first inverted. Recession started Jan 2008. NBER called it in Dec 2008.
No recession yet. I used the 6 monthly indicators listed by the NBER to construct an index that matches pretty well with the last 2 recessions. It’s normally between 0 and 1 std dev from 0, where it sits for July, but then plunges below 0 as a recession starts.
MONTHLY RECESSION WATCH: For March 2024, the 10yr-3mo Treasury spread averaged -1.17%, for a 97.3% probability that a recession will start by March 2025.
November employment is still inconclusive but it's following the 1990-91 recession pretty closely and it's below 3 others. As I've said before, it takes a while to know for sure.
Thanks
@LisaBeilfuss
for pointing out this indicator. With only 1 false positive in 1969, it's risen past 10% once during every recession since 1953 with lags of 0-7 months from the recession start. It just gave a signal in April, suggesting a recession may have started by then.
Concerns that initial unemployment claims (SA) have been suspiciously constant, ie, the same for 5 of 6 weeks ending 4/15. Going back to 1968, the number has repeated 89 times on consecutive weeks (3%), as in the last 2. Three consecutive equal numbers first happened last month!
The 10y-3m spread predicts recessions better than 10y-2y and Fed staff's 18m-3m. In graph, R-squared measures predictive fit with 1=perfect and 0=useless. 10y-3m is best for all horizons, 10y-2y is the worst at 4Q. More info at .
Of the previous 8 yield curve inversions, 7 ended after the Fed started cutting the funds rate. The only exception was 2007, when the T-bill market got tired of waiting for the FOMC to move and took the lead. Needless to say, recessions followed all 8 cases.
Industrial production continues its lackluster performance since the yield curve first inverted in Nov 2022, underperforming most of the earlier inversion periods.
JOLTS data are out today. The news release says that everything was "little changed" (or "changed little") but there appear to be some trends since monetary tightening started.
@MacroAlf
1) You're looking at the wrong yield curve spread. We've had 7 months of inversion of 10y-3m. 2) It takes time for the recession to start. Historically, it's 6-17 months after the first inversion (of the right spread). Give it time.
Who to believe? Two pairs of NBER 6 recession indicators that should give similar signals are diverging in recent months. Household employment peaked last November, real sales last December, but payroll employment and real consumption are still rising. Will the pairs converge?
The current yield curve inversion compared with the one before the Great Recession. This one will be longer. Also, uninversion does not eliminate chances of recession.
Other than the BLS payroll employment survey, the other NBER 6 recession indicators have already shown signs of weakness since the yield curve inverted. Given the double counting issue, performance of the payroll indicator should be interpreted cautiously.
The NBER looks at both the payroll and household employment surveys when dating recessions, and this time the signals are completely different. HH peaked last November while payroll keeps going straight up. The difference is related to double counting in payroll.
The dream of a soft landing is still alive. Q4 real GDP growth is the strongest after a yield curve inversion since 1968, though it continued to grow for a longer period before 3 earlier recessions. I’m impressed but this mid-Nov avg data does not rule out a 2024 recession.
Downward revision of 0.3% in Q2 real GDP growth leaves the outlook essentially unchanged. The path is well within the bounds of earlier pre-recession yield curve inversion periods. Again bear in mind that this is historical information for April-June.
For those who like to go beyond the average, here is real GDP from each first yield curve inversion into 7 individual recessions. The present case (red line labelled Q4 2022) is right in the middle.
MONTHLY RECESSION WATCH: For October, the 10yr-3mo Treasury spread averaged -.69%, which is consistent with a 92.5% probability that a recession will start by October 2024.
Graphs like this are so misleading. They show a growing economy but they don’t tell you that this is normal before a recession starts. They can’t show what’s coming next. I’ll give you a couple of examples...
Annual PCE inflation was essentially unchanged in February at 2.45%. The good news is that it's not getting any worse and that it seems already close to the Fed's target by historical standards. The bad news is that progress toward the target seems to be slowing down.
In the movie, Oppenheimer reassures Gen Groves by saying that the chances of destroying the world are near zero. Reminded me of my model where the chances of avoiding a recession starting between now and next June are .0013. What do you want from a model?
Real sales took a bit of a plunge in January. After topping all the pre-recession data in December, sales fell below the 1990 and 2008 recessions, relative to the first inversion month. As usual, release of this series lags the others in the NBER 6 by a month.
Atlanta Fed’s Q2 GDP nowcast took a big plunge this week from the mid-May estimate. It’s not always on target, but there was another big drop back in 2022 Q2 and the final number that quarter was negative.
The Taylor Principle says that to stabilize inflation, the interest rate should be raised eventually by more than the increase in inflation. That might explain why there’s still talk of further fed funds tightening.
Looking at the period since the yield curve inverted in 11/2022, real personal income flattened out from January to April. It still tracks below the inversion period that preceded the Great Recession of 2008.
This graph shows my usual pre-recession inversion analysis applied to bank business lending. The current inversion period is well off the distribution, on the negative side.
MONTHLY RECESSION WATCH: For June 2023, the 10yr-3mo Treasury spread averaged -1.55%, the 8th straight negative monthly reading. The probability that a recession will start within the next 12 months is 99.9%.
Inflation is still very low by yield curve inversion period standards. However, I'm afraid we can't give the Fed much credit yet because demand is only starting to feel the rates. Much of the drop is from receding supply pressure. When demand responds, it could go a lot lower.
The yield curve is going through another disinversion push, driven by a rise of 26 bps in the 10-year rate since the middle of the month. Still, it's going to take a big shift in expectations about monetary easing to fully uninvert.
How long will the recession be? Judging by the last 8, probably long. Inversion > 9 months leads to recession > 9 months. The only exception is the 1980 recession, which ended with the fed funds rate being cut by 12.31%. Not even possible now.
After 7 months of yield curve inversions,
#1
/3: prepare for recession, eg, don’t quit your job, build up savings, review risky investments (incl. through mutual fund, IRA, 401k).
The lags from the start of a recession to the Sahm and Kantro signals have been growing steadier and generally shorter since 1970. For the last recession, both signals came in 1 month after the recession started.
After breaking out strongly following the first month of yield curve inversion in Nov 2022, household survey employment fell in December and is close to the average over pre-recession inversions that lasted this long.
The household survey normally shows more employment than the payroll survey. Right now, household employment is very weak compared with payroll. The difference is as low as it's been other than during the last recession.