The Big Idea
In search of a recession predictor
Stephen Stanley | August 23, 2024
This document is intended for institutional investors and is not subject to all of the independence and disclosure standards applicable to debt research reports prepared for retail investors. This material does not constitute research.
As the talk of recession heats up, financial market participants are desperately searching for a foolproof predictor of an economic downturn. Unfortunately, economic cycles are not that simple as each recession is different, sparked by different factors and occurring at different speeds. Nonetheless, initial jobless claims is as good a candidate as any, a timely weekly indicator with virtually no reporting lag, tied to a key part of the economy—the labor market—and even included in the Conference Board index of leading indicators. Its historical forecasting record, however, is a mixed bag.
Ancient history
The weekly jobless claims data extend back to 1967. The period from 1967 to 2024 encompasses eight recessions, allowing for a number of instances to examine how the number of new filers evolves around the beginning of a recession.
The first of those recessions began in 1969. It helps to look at the initial claims readings from six months prior to the official recession date set by the National Bureau of Economic Research to six months after, so there are 53 weekly figures (Exhibit 1). The zero point marks the week closest to the middle of the designated cycle month.
Exhibit 1: Initial claims before and after the December 1969 recession date
Source: Labor Department, NBER.
In the six months preceding the 1969 recession date, the number of new filers trended upward modestly, from around 200,000 to about 230,000. Layoffs surged much more steeply after the recession began.
The next recession, starting in November 1973, was caused by the first big oil shock, when Arab countries imposed an oil export embargo against the U.S. for supporting Israel in the 1973 Arab-Israeli war in October of that year (Exhibit 2). So, it is not surprising that initial jobless claims did not show a sharp increase earlier in the year.
Exhibit 2: Initial claims before and after the November 1973 recession date
Source: Labor Department, NBER.
The next recession came in January 1980. This was also a unique cycle, as inflation was crushing the economy. In a futile attempt to tamp down price hikes, the Carter Administration imposed credit controls on the economy in March 1980, exacerbating a downturn that was already in its early stages. These credit controls arguably worked too well, creating—or exacerbating, depending on your perspective—a deep contraction. By July, the Carter Administration lifted the credit controls, and the recession promptly ended.
Again for the January 1980 recession, it is not terribly surprising that the rise in initial claims leading up to the recession was modest (Exhibit 3). The trend moved from around 390,000 six months prior to the cycle date to around 420,000 at the official cycle date.
Exhibit 3: Initial claims before and after the January 1980 recession date
Source: Labor Department, NBER.
The fourth recession to examine is the 1981-82 downturn, which was at the time the worst since the Great Depression, with the unemployment rate soaring to almost 11% in 1982. This time, initial jobless claims were helpful in predicting a recession, as they crept up from around 400,000 six months ahead of the official cycle date to around 450,000 by July (Exhibit 4).
Exhibit 4: Initial claims before and after the July 1981 recession date
Source: Labor Department, NBER.
Modern times
Perhaps it is a telltale sign of my advancing age that I would peg the 1990-91 recession as the first of the “modern” era. This downturn had elements of several of the prior cycles. The Fed was forced to impose relatively tight monetary policy to tamp down accelerating inflation and was trying to navigate an elusive soft landing. However, echoing 1973, Iraq invaded Kuwait in August 1990, leading to a spike in oil prices and sealing the deal on a recession. Initial claims were running at an elevated level throughout this period, around 350,000 to 375,000, but they did not increase sharply until after the oil shock began to weigh on economic growth and the labor market (Exhibit 5).
Exhibit 5: Initial claims before and after the July 1990 recession date
Source: Labor Department, NBER.
The next recession began in March 2001. This is arguably the first of the cycles that we have examined in which initial unemployment claims offered a clear advanced warning, as the number of new filers moved up by nearly 100K during the six months prior to the official recession date (Exhibit 6). Interestingly, also in contrast to prior cycles, initial claims failed to rise sharply further after the economy officially entered recession (the six months after the recession date ended just before the 9/11 terrorist attacks that led to a further severe contraction in the labor market). Thus, one might argue that initial claims were an effective leading indicator leading up to the March 2001 downturn, but a skeptic might suggest that the labor market was simply a little ahead of GDP in terms of the timing of the contraction.
Exhibit 6: Initial claims before and after the March 2001 recession date
Source: Labor Department, NBER.
The next recession began in December 2007. Of course, the height of the Global Financial Crisis did not hit until the fall of 2008, but the economy was already in gentle decline in the first half of 2008. The unemployment rate backed up in the first half of 2008 from around 5% to about 5.5%, before surging to nearly 10% by the middle of 2009. Once again, there is a clear uptrend in the six months leading up to the cycle date, in this case from around 310,000 to about 350,000 (Exhibit 7).
Exhibit 7: Initial claims before and after the December 2007 recession date
Source: Labor Department, NBER.
The eighth and final cycle is the Covid recession. Since this downturn came virtually out of nowhere, it is no surprise that initial claims were quite low and steady in the months prior to the recession, though it is impossible to tell from the chart since the scale is so large to reflect the massive spike in filings during the early weeks of the pandemic (Exhibit 8).
Exhibit 8: Initial claims before and after the February 2020 recession date
Source: Labor Department, NBER.
Conclusion
Based on the historical record, I would argue that the jury is out on whether initial jobless claims represent a reliable predictor of recession. In the 1960s,1970s and 1980s, the answer would generally be no, but in the 2001 and 2007-09 downturns, they were helpful. In the current instance, one could point out that we have seen a bit of an uptrend over the past six months (Exhibit 9). Still, as I have detailed in the weekly recaps, the elevated readings in June and July appear to have been mostly 1-off factors. The last couple of weekly readings have settled back to around the 230,000 level, only modestly higher than the prevailing range of around 210,000 six months ago.
Exhibit 9: Initial claims over the past six months
Source: Labor Department.
More importantly, both the change and the level of initial claims is relevant to assessing the state of the labor market. And while the change piece is at least somewhat debatable, the level of initial claims is indisputably historically low. The rule of thumb that describes the past several decades is that the 300,000 level is consistent with a strong labor market, while a 400,000 level signals a recessionary labor situation. In that context, moving from 210,000 to 230,000 should barely move the needle.
The remarkably low level of initial jobless claims underscores a key, and somewhat unique, element of the labor market currently. Coming out of the severe labor shortage of 2021 and 2022, firms struggled to staff up and have reported that they are unusually hesitant to let good people go, for fear that they will be unable to grab them back if demand picks up. Thus, the cooling in the labor market over the past year or so has almost entirely reflected a slowdown in the pace of new hiring, as firms have largely filled the backlog of openings left over from that labor shortage period. In contrast, layoffs remain unusually low, with the JOLTS measure of layoffs corroborating the still historically modest pace of unemployment insurance claims. A sharp pickup in layoffs would undoubtedly offer a key signal of a possible economic downturn, but, for now, in my view, the data point to, as Fed officials have argued, a normalization of labor market conditions rather than a shift to outright weakness.