Summary
- Most observers expect a recession this year, but there are an increasing number forecasting a “soft landing,” where inflation cools without a recession.
- The NBER will rely on monthly data, which is reported with a lag as much as two months, and is usually revised.
- But there are 3 proxies, all reported within a week, and that are never revised, that will give a much more timely signal if there is no soft landing.
Introduction
While a significant minority of observers believe the Fed will achieve a “soft landing,” in which inflation fades without a recession, the large majority (including me) believe a recession is all but unavoidable at this point. It’s hard not to come to that conclusion when one looks at the behavior of the Conference Board’s Index of Leading Indicators (via Advisor Perspectives):
Index of Leading Indicators (Conference Board via Advisor Perspectives)
The Index has declined for 9 straight months, and is now -5% below its recent peak. The index has *never* declined this much without a recession having occurred. In fact, its current decline is almost as much as, or even more than, 3 recessions in the past 60+ years (1960, 1970, 1982) and nearly 50% as deep as the maximum declines in 3 others (1980, 1990, 2001).
Are there data points that can help us reasonably conclude, on a timely basis, that a recession has begun, or has been avoided? As described below, there are at least three.
The 3 quickly reported, never revised, coincident indicators of recession
Recently, in almost every synopsis of my “Weekly Indicators” update, which features high-frequency, weekly (or even daily) economic indicators, I have emphasized that I am particularly paying attention to 3 metrics: Redbook consumer sales, The American Staffing Association’s temporary help Staffing Index, and the US Treasury’s report of tax withholding payments in its Daily Treasury Report. Let’s look at each in turn.
Redbook consumer spending
The NBER has indicated that among its “big 4” coincident recession indicators is “real manufacturing and trade sales.” The problem is, it is only reported with a 2 month delay. For example, we are almost at the end of January, and tomorrow it will be reported for November! (Red, YoY, in the graph below). Retail sales (Blue) make up about 1/2 of this metric, and they are reported more timely, but still in the following month:
Real manufacturing and Trade Sales and real retail sales ((FRED))
In contrast, there used to be as many as 4 weekly updates of consumer spending. The most well-known was the ICSC same store sales index, but there were also indexes by Shoppertrak and Gallup, in addition to Redbook. Unfortunately, one by one they were all discontinued until Redbook was the last one standing.
Redbook calculates the nominal YoY% change in retail sales. Here’s the long-term 20+ year graph:
Redbook consumer spending (Redbook (via Trading Economics))
Note the dip at the beginning of the 2008 recession, followed by the big plunge when the financial crisis hit. Note also the smaller dip (but still positive) in 2015-16, the plunge during the pandemic lockdowns, and the Boom thereafter.
Here is the comparable time period for YoY nominal retail sales:
The results are very similar. There is a very small dip at the beginning of the Great Recession and also 2015-16, and large ones during the two recessions, with a huge spike during the 2021 Boom.
This tells us that the two metrics follow the same trajectory. The differences are that Redbook is reporting the following week and is never revised, while the official retail sales metric is reported with a one month delay, and is significantly revised thereafter. In other words, Redbook gives us a much more timely, and reasonably accurate, read on YoY consumer spending (unadjusted for inflation).
Now here is the close-up of Redbook for the past year:
Redbook 2022 consumer spending (Redbook (via Trading Economics))
YoY sales were averaging 14% last January. They slowly decelerated to roughly 12% gains through last July, and then started to decline sharply. For the first three weeks of January through last week, they are only up 5% YoY.
With the exception of the March-May period (due to the comparison with stimulus spending in March-May 2021), nominal retail sales tell a similar story:
Now let’s add in the effects of inflation. Here are some notable monthly comparisons in the YoY number for the past year. The first number is Redbook, the second YoY CPI, and the third Redbook after adjusting for inflation:
Jan 2022: 14%-7.5% = +6.5%
July 2022: 13%-8.5% = +4.5%
August 2022: 12%-8.2% = +3.8%
September 2022: 11%-8.2% = +2.8%
December 2022: 7.5%-6.5% = +1.0%.
By my rule of thumb for YoY nominal data, which is that the peak occurs close in time to when the YoY comparison declines from 1/2 from peak, real consumer spending probably peaked in roughly September.
Further, inflation-adjusted consumer spending as measured by Redbook is on the verge of going negative YoY, unless the final week of January spending is better than the first 3, and/or inflation plunges sharply (unlikely as gas prices have risen in January). Should that happen, that would be a good marker that the “real sales” component of recession indicators used by the NBER has been triggered.
The ASA’s Staffing Index
Temporary help employment is generally viewed as a leading indicator for the job market as a whole. Here’s what that looks like for the past 25 years:
Temporary help employment did indeed turn down months in advance of all three recessions during that time.
Since the Staffing Index measures YoY temporary employment, it ought to be a good proxy for that metric. So first, let’s look at the official measure YoY:
YoY temporary employment ((FRED))
This has turned negative in advance of recessions, but also during several other periods of weakness: 2003, 2016, and 2019. It also just turned negative last month, in December.
The Staffing Index did not start until 2006, but it did also show weakness YoY in 2016 and 2019:
2015-16 ASA Staffing Index (American Staffing Association) 2018-19 ASA Staffing Index (American Staffing Association)
It has also been virtually unchanged YoY since the beginning of November:
2021-23 ASA Staffing Index (American Staffing Association)
So the ASA’s Staffing Index has a very good record of being a proxy for temporary help employment in the payrolls report. I would not expect temporary help in the payrolls report to remain negative unless this Index turns firmly negative as well.
Payroll tax withholding
Unlike the vast majority of other metrics, payroll tax withholding is not a survey or sample. It is the actual total amount of payroll withholding taxes paid into the U.S. Treasury as reported each and every workday. Although there are some differences - there is a maximum income limit for Social Security withholding, and stock option payments may be included - it ought to be a good proxy for aggregate nominal payroll growth YoY, particularly for nonsupervisory workers.
Matt Trivisonno calculates this for the entire 365 day rolling period YoY. Here’s what that looks like from 2000 until 90 days ago:
YoY tax withholding payments (DailyJobsUpdate.com)
Again, we can see that it tracks recessions and expansions very well, deteriorating sharply immediately advance of recessions, and turning negative during the recessions.
The YoY% change in withholding peaked at roughly 22% last January. Although not shown (since Trivisonno only posts the public data with a 3 month delay), it is easy to calculate that by December 31, it had declined to +7.0%. After adjusting for inflation, that’s a decline from +13.5% to only +0.5%!
Similarly, several times in the past year Liz Ann Sonders of Charles Schwab has posted the YoY% change in the 10-week rolling average of withholding payments. Her last such graph was at the end of August:
10 week YoY tax withholding (Liz Ann Sonders (Twitter))
The deterioration in the 10 week average has also continued. By 10 weeks into fiscal 2023 in December it was only +4.5%. As of last Friday it was 3.4%. Once we factor in inflation, this way of measuring had already turned negative by the end of July.
By way of comparison, here’s what the nominal YoY% change has been for total and nonsupervisory aggregate payrolls in the past 12 months:
The overall trend is similar to what is shown by tax withholding, but the latter has declined at a much steeper rate than the former (which is based on a survey of employers). I should mention one caution that has been posited by the California Treasury Department, for whom tax withholding includes stock option payments, which needless to say have declined sharply in the past year with the decline in the stock market.
But again, by my rule of thumb, hypothetically seasonally adjusted tax withholding has already peaked, probably in September, meaning so have aggregate payrolls. In the past, when YoY nominal withholding tax collections have turned negative, payrolls as measured by either the Establishment or Household reports have done so within 2 months thereafter. Should tax withholding continue on its current trajectory, it could turn negative YoY as early as in the next several weeks. That would be a clear marker that the jobs reports could give consistent negative prints within just a couple of months.
Conclusion
If Fed tightening is going to cause a recession, and not just a “soft landing,” then both employment and real sales should turn negative. Monthly measures will only give us this information with a significant delay, and initial reports will be revised. By contrast, the 3 indicators discussed above are updated with only a one week, or in the case of the U.S. Treasury, one day, delay; and they are never revised.
The first two are on the cusp of turning negative. The third, payroll withholding payments, may have already turned negative in real terms.
For further details see:
Soft Landing Or Recession? 3 Quickly Reported Coincident Indicators To Watch