2023-04-04 11:07:43 ET
Summary
- The National Bureau of Economic Research has indicated they particularly rely on 5 indicators: real GDP, nonfarm payrolls, industrial production, real sales, and real personal income less government transfers.
- Only one of these - industrial production - has turned down as of last report.
- But the values of 4 of the 5 are together in a configuration that in the past has typically occurred during or shortly before a recession.
- A recession may even have already begun, but we don’t know it yet because of reporting delays, noise, and revisions.
Introduction
The constellation of long leading indicators first turned down over 12 months ago, suggesting that a recession could occur any time beginning with Q1 this year. While they were heavily influenced by the price of oil, the short leading indicators also turned down by late last autumn, also suggesting a recession could occur by this spring.
But what of the most important coincident indicators; the ones that are relied upon by the National Bureau of Economic Research ("NBER") to determine if the economy is expanding or in recession? Conventional wisdom at this time is that they have not peaked, and a recession is at least some distance away, if one occurs at all.
But a closer look reveals that the indicators most highlighted by the NBER are in configurations that in the past have typically occurred near if not actually during recessions. There is a non-trivial possibility that the NBER will determine that a cycle peak already happened in January. What follows below discusses why.
Let me emphasize that in the below discussion I am not cherry-picking a set of "indicators du jour." All of the below are either the exact indicators that the NBER has indicated they follow, or their close correlates.
3 of 4 monthly indicators may have peaked
As everyone knows, nonfarm payroll growth has remained strong, averaging over 400,000 for the past two months. But as we'll see below, that does not foreclose the possibility that a recession has begun.
There are at least 3 other monthly metrics - industrial production, real manufacturing and trade sales, and real personal income less government transfers - that also are important to this determination. And they *may* have peaked.
To wit, here is a graph of industrial production (blue), real manufacturing and trade sales (gold), and real personal income less government transfers (red) normed to 100 as of last September, when production peaked:
Industrial production, real sales, and real personal income (FRED)
Industrial production is down more than -1% since then. Personal income less government transfers growth has slowed down consistently since then, up only 0.03% monthly in February, which rounds to unchanged. Only real manufacturing sales increased through its last, January, update.
A back of the envelope estimate using an average of industrial production and real retail sales suggests real manufacturing and trade sales probably declined in February as well:
The YoY trends in these three coincident indicators are also consistent with an imminent or already ongoing recession. As of their last report, industrial production had increased +0.32% YoY, real manufacturing and trade sales +1.0%, and real personal income less government transfers +1.55% (all three series normed to zero as of last report):
YeY industrial production, real sales, and real personal income (FRED)
Now here is a long-term look at all three. As in the graph above, below I have subtracted their current value over zero so that the same value in the past shows up exactly at the zero line:
YoY industrial production, real sales, and real personal income (FRED)
In the past 60 years, while any one might have been below its current value without being anywhere near a recession, when *all 3* were at their current values or lower, it was always after a recession had already started, or in three cases within 2-6 months of the start of the next recession (November 1979, January 1990, and December 2019).
There is also a 75 year history that, with the exceptions of 1951 and 1967, whenever the 3 month average of real retail sales (a large component of real manufacturing and trade sales) have turned negative YoY - as they did in Q4, where they were down -0.4% YoY - a recession has followed shortly or already begun:
In other words, 3 of the 4 monthly indicators relied upon by the NBER are currently in trends that suggest a cycle peak may just have occurred, or else is within a few months of occurring.
Real GDP and GDI
In addition to the monthly data discussed above, the NBER has indicated that it also places reliance upon quarterly real GDP, which as of its last revision, increased +2.6% on an annualized basis during Q4.
But, in the 12 recessions since the end of 1945, in the last full quarter before the recession began, real GDP averaged +1.7%, and the median was 2.3%, with a low of -2.9% and a high of +9.3%. Indeed, in the first quarter of the 12 recessions themselves since 1945, real GDP was still positive in real terms, with the average of +0.4% and a median of +0.5% (and a low of -4.6% and high of +4.9%):
Real GDP quarterly change (FRED)
When we take out inventories, and focus on real final sales and real final sales to domestic purchasers, which were +0.3% and less than +0.1% in Q4, respectively, in the past 60 years when both of those quarterly increases or worse occurred together, with the exceptions of 1979 and 1987, a recession was 2 quarters or less away.
Q/q% change in real final sales; real final sales to domestic purchasers (FRED)
Further, real Gross Domestic Income, the accounting "flip side" of GDP, declined -0.3%, or -1.1% annualized, in Q4, again typically in the past a number that either obtained during a recession, or within two quarters of its start (with the exceptions of 1993, 2012, and 2016):
Q/q% change real domestic income (FRED)
But what of payrolls?
Finally, let's look at the final monthly indicator relied upon by the NBER: nonfarm payrolls. As we all know, payrolls have remained strongly positive.
But again, let's look at the long term history. It's true that in the 25 years after the end of World War 2 (first graph below), and again for the past 40 years (second graph), no recession has begun without nonfarm payrolls turning down:
Nonfarm payrolls 1945-69 (FRED) Nonfarm payrolls 1983-2019 (FRED)
But there is an exception to that rule. In the 3 recessions that began with 1970, nonfarm payrolls continued to rise for several months - up to 8 months in the case of 1974 - into a recession:
Nonfarm payrolls 1969-82 (FRED)
In other words, so long as the other coincident indicators had turned down, the persistence of jobs growth did not negative the onset of a recession.
Conclusion
While 4 of the 5 series highlighted by the NBER have remained positive or unchanged as of their last reports, with the exception of payrolls they are in a configuration in that in the past has either been during a recession or in close proximity thereto. While this is only a nowcast, not a forecast, it is non-trivially possible that a recession has already begun and we just don't know it yet because of reporting delays, noise, and revisions.
For further details see:
Coincident Indicators Nowcast: Non-Trivial Chance The Cycle Peak Was January