Economy & Markets
Five factors we use to track recession risk, and what they say now
Macro shocks have repeatedly upended markets in recent years. These have included the COVID-19 pandemic, with its disruptions and business closures, the subsequent inflation shock, aggressive policy rate hikes by the Federal Reserve (Fed) and other central banks, and a mini-meltdown in the U.S. regional banking sector.
These events contributed to the consensus in 2023 that a recession was around the corner. When that did not happen, a similar view re-emerged in 2024, when higher unemployment triggered the Sahm rule1—which had been a strong indicator of impending recession in decades past.
But recession did not ensue in 2024 either. In fact, for two years in a row, Wall Street economists were too pessimistic in their expectations for U.S. real GDP growth. This is shown below. We’ve added the consensus outlook for 2025 to the projections for 2023 and 2024.
Real GDP keeps beating expectations
Real GDP growth rate, %
Sources: Bureau of Economic Analysis, Wolters Kluwer/Haver Analytics. Data as of December 2024.
Why has U.S. growth defied so many economists’ projections? We see three primary reasons:
- Exceptionally strong corporate and household balance sheet fundamentals
- The 2020–22 inflation shock proved temporary, as we’ve described in depth previously
- The strong momentum of the U.S. economy, which was growing 5% to 6% annually when the Fed began hiking interest rates in 2022, meant the economy could withstand historically steep rate increases without going into recession
Despite the prevalence of hard-to-predict macro shocks, we think investors can still benefit from a systematic approach to modelling the business cycle and recession risk. We write this note to flesh out our baseline view that the current cycle will not end in 2025, and to explain why we forecast only moderate recession risk in the next 12 months.
Our current subjective view is that there is a 20% chance of recession. This is 10 percentage points above the unconditional 12-month probability of a recession occurring at any given time.2
Beneath these odds lies a range of models we build and consider. We’ll walk through several of the most important of these, what they’re telling us today and why they matter.
The parts of our model
1) The yield curve: We start modeling the probability of a recession by considering the shape of the yield curve. Prior to COVID, an inverted or flat curve had been a reliable predictor of recession within 12 months. Investors found this signal useful because they believed that, at turning points in the economy, the Fed tends to overemphasize lagging inflation data compared to real-time trends of demand weakness. Thus, market participants could snuff this out and price in an aggressive rates easing cycle before it began.
The current shape of the Treasury yield curve continues to suggest a high likelihood of recession in the next 12 months, as the chart shows. However, the yield curve has recently been a misleading variable due to the unusual way inflation has normalized since 2023: Normalization is coming more from supply expansion and productivity increases than from demand destruction. We continue to heavily discount the signal being sent by the yield curve, though as always, we track it systematically.
The yield curve’s views on recession
Probability of a recession in 12 months (U.S. Treasury yield curve), %
Source: Haver Analytics. Data as of December 2024. Probit model output based on three iterations of the yield curve: the 2s10s spread, the 5s30s spread and the 3m3m18mfwd spread.
2) Private sector debt service: Next, we consider the wise advice from Menzie Chinn, a professor of economics at the University of Wisconsin–Madison, which is to augment the Treasury curve recession model with the total private sector debt service ratio.3 U.S. recessions are typically preceded by a rising debt service ratio for the private sector because it tends to signal impending demand weakness: If companies and households are devoting more income to servicing debt, less will be available for current spending. At the moment, the debt service ratio remains low and has been rising only gradually, suggesting low odds of a recession.4
Little threat from debt
Total private sector debt service ratio, %
Sources: BIS, Fed, Haver Analytics. Data as of December 2024.
3) Economic momentum: We also analyze economic momentum, which is important because decreasing momentum can signal that recession risks are increasing. To be clear, however, we do not subscribe to the notion that economies can “run out of steam,” and we don’t regard a loss of momentum as a key causal feature of a recession.
Rather, we believe recessions are primarily caused by shocks that abruptly change the trajectory of aggregate demand or supply. 5 However, a loss of momentum makes an economy less able to weather a shock—this is what we mean when we say a loss of economic momentum increases the odds of a recession.
To measure momentum, we combine a measure of private final demand in the economy with the Philadelphia Fed’s data on coincident economic activity at the state level.6 The charts show that final demand has not slowed, but state-level momentum has deteriorated recently, reflecting divergent trends in state and national economic performance.
U.S. demand is strong, but state data shows a divergence
Left: Year-over-year change in real final sales to private domestic purchasers, %; Right: 3-month diffusion index of coincident economic activity, %
Sources: Bureau of Economic Analysis, Haver Analytics. Data as of September 30, 2024.
Sources: Federal Reserve Bank of Philadelphia, Haver Analytics. Data as of September 30, 2024.
4) Profits: Next, we consider corporate profit margins. Deteriorating corporate profit margins tend to be a leading recession indicator, as declining profitability makes it more likely firms will shed labor. This makes businesses a key vector in propagating negative demand shocks throughout the economy, as they can contribute to a negative feedback loop of layoffs and reduced consumer spending.7
Here, we note that corporate profit margins have been resilient. Indeed, they have been rising of late, suggesting there is little cause for concern about layoffs and consumer spending.
More generally, we believe U.S. corporation and household balance sheet fundamentals point to low odds of a recession. Further, if a cyclical downturn were to occur unexpectedly, these positive balance sheet fundamentals would likely reduce its severity. (A digression: The flipside of improving household and corporate balance sheets has been a deteriorating government balance sheet, but in a recent article, we explained in depth why we are not losing very much sleep over it.8)
Corporate profit margins and household balance sheets are solid
Left: Profit margins of nonfinancial corporate businesses, %; Right: Total liabilities as a percentage of net worth (households), %
Source: Haver Analytics. Data as of September 30, 2024.
Sources: Federal Reserve Board, Haver Analytics. Data as of September 30, 2024.
5) Financial markets: Lastly, we look at financial markets. Markets move in advance of the economy, so the signal they send can be helpful for modelling recession odds—even though the signal is usually noisy. To incorporate the signal sent by markets, we consider the prices of risk assets—which have a different investor base from Treasury markets—as well as measures of sentiment.
The news here is also comforting. This chart shows that markets priced in an elevated recession risk in 2022, but since then, they have priced out that risk. Markets are currently pricing minimal recession risk for the year ahead.9
The news here is also comforting. This chart shows that markets priced in an elevated recession risk in 2022, but since then, they have priced out that risk. Markets are currently pricing minimal recession risk for the year ahead.9
Markets see little recession risk
Market recession odds, %
Sources: S&P, Federal Reserve, Haver Analytics. Data as of December 2024.
Assembling the pieces
We build recession odds models for each of these vectors and then average the outputs.10 The resulting output shows that the likelihood of recession in the next year is just under 30%. We think this is too high because (as we’ve discussed) we believe the yield curve signal is currently distorted.
The chart below—our model’s full output—is illuminating because it produces the same result even if the yield curve signal is excluded: It is striking how similar the two lines have been for the past three decades—and how much they diverge in the last two years. This underscores our reasons for downplaying the yield curve signal today.
We think the signal that excludes the yield curve is most informative: Rather than a 30% chance of a recession in 2025, it projects a 12% chance. As mentioned, our subjective view is that the chances of a recession are 20%, which reflects the possibility of a policy mistake made by the incoming Trump administration with respect to tariffs and the risk its policies could incite a global trade war.
The gap between the models
Probability of recession (combined models vs. combined ex yield curve), %
Source: J.P. Morgan Private Bank. Data as of December 2024.
Why this isn’t an academic exercise
When discussing recession odds with investors, we are often asked whether these odds even matter. The particular critique is that by the time our model shows that recession odds have increased, markets will already have priced in the change. In other words, clients wonder if these modeled odds are predictive of markets or merely coincide with evidence as it emerges.
Our work shows that models such as these can indeed lead markets. The below table shows the timing of three initial equity market sell-offs before three previous recessions11 and how much our recession odds signal moved up before the sell-off. In all three cases, our indicator had risen substantially prior to the sell-off.
How our model behaves ahead of equity sell-offs
Initial equity market sell-off and increases in recession odds, %
Source: J.P. Morgan Private Bank. Data as of December 2024.
An important related point: We find evidence that recession odds tend to rise nonlinearly before a recession begins. All else equal, we find that a one-percentage point rise in recession odds is associated with a 70-basis-point decline in stock prices. But if recession odds rise nonlinearly, those initial steps up in recession odds can serve as an early warning signal for investors that a sharper market downturn could be in store in the coming quarters.
Rest assured, this early warning signal is not currently flashing, as our estimate of recession odds (excluding the yield curve) has been falling since early 2023.
How stocks react as recession odds rise
Average percentage change in stock prices and increases in recession odds, %
Source: J.P. Morgan Private Bank. Data as of December 2024. Periods considered are the 12 quarters leading up to the peak rise in recession odds during the early 90s’ recession, the dot.com recession and the Global Financial Crisis. The fitted blue line is a polynomial of order 3 fit, which produced an R-square of 0.92 compared to a linear fit of only 0.46.
To review our recession model components, only one is showing elevated risk—the yield curve signal—while the economic momentum signal shows moderate risk, and the debt serve ratio, corporate profitability and risky market signals all show low recession risk in the next 12 months.
Results summary table
Degree of recession risk (various indicators), qualitative assessment
Source: J.P. Morgan Private Bank. Data as of December 2024.
Final words
There are two caveats to this approach to modeling recession odds.
The first is obvious: It will not capture exogenous shocks that come about suddenly and are not intrinsic to the U.S. business cycle. COVID-19 was an exogenous shock. Geopolitics has the potential to produce an exogenous shock that could have macroeconomic consequences in the coming years.12
The second caveat is that our analysis should not be interpreted as an argument that predicting recessions is all that is needed to form an effective investment strategy. Bear markets in stocks can occur without an economic recession. In fact, the frequency of these has been increasing. In late 2018 and in 2022, stocks entered a bear market triggered by central bank rate hikes, and the market’s fear that the central bank would “overdo” it, even though in both of those instances an economic recession didn’t follow.
Another significant point: Many investors are currently less concerned about recession risk and more focused on the possibility inflation could pick up again. We think the main inflation risk to markets today has less to do with the U.S. economy’s underlying trajectory—labor-driven inflation continues to normalize and the Fed is still running a restrictive approach to monetary policy—and more to do with fiscal policy and trade policy risks under the new Trump administration.
While exogenous or policy shocks are inherently more difficult to predict than underlying economic fundamentals, the models we have described throughout this article help us understand the state of the U.S. economy, and they show that it remains in a strong position. As we wrote in our annual outlook, we remain constructive on markets in 2025. However, we think returns for risk assets are likely to revert to trend-like rates because of challenges such as growing geopolitical conflict, increased fiscal and trade policy uncertainty and volatility, and elevated financial market valuations.
We can help
If you have questions about the possible outcomes we’re expecting for markets and for the broader economy, and how these may affect your investment objectives, contact your J.P. Morgan team.
1 The Sahm rule tracks the unemployment rate. It states that if the three-month moving average of the U.S. unemployment rises more than 50 basis points from its low point over the last year, the U.S. economy is already in recession. This indicator had a sterling track record for decades leading up to its “triggering” by rising unemployment in July 2024.
2 The unconditional odds are based on data from 1985 to the present.
3 Chinn, Minzie. “The Debt-Service Ratio and Estimated Recession Probability,” Econbrowser. September 22, 2023. https://econbrowser.com/archives/2023/09/the-debt-service-ratio-and-estimated-recession-probability
4 In large part, the slow rise in debt service of late has been due to the private sector’s shift away from floating-rate debt and toward long-duration, fixed-rate debt during the prior cycle of the 2010s and during the pandemic.
5 For example, Russia’s invasion of Ukraine in 2022 caused a large commodity price and uncertainty shock to hit the global economy. However, this shock did not push most national economies into recession because economic momentum was strong at the time of the invasion.
6 The Philly Fed State Coincident Indexes are economic indicators developed by the Federal Reserve Bank of Philadelphia. These indexes are updated monthly and provide a snapshot of economic activity in each of the 50 U.S. states. Each state’s coincident index combines four state-level indicators: nonfarm payroll employment, average hours worked in manufacturing, the unemployment rate, and wage and salary disbursements.
7 We discussed the negative feedback loop between layoffs and consumer spending more completely in an article we published last fall. Seydl, Joe, “Why this economic cycle is defying history—and breaking the rules,” J.P. Morgan Private Bank. September 20, 2024. https://privatebank.jpmorgan.com/nam/en/insights/markets-and-investing/why-this-economic-cycle-is-defying-history-and-breaking-the-rules
8 Seydl, Joe. “Is the deficit threat being overhyped?,” J.P. Morgan Private Bank. December 9, 2024. https://privatebank.jpmorgan.com/nam/en/insights/markets-and-investing/ideas-and-insights/is-the-deficit-threat-being-overhyped
9 To be sure, this risky asset vector is prone to rapid moves. However, we’ve constructed this model output to react only to large market movements—not typical market volatility. This model shows that absent other factors, the average time between market downturns is 22 quarters. The last market downturn was in 2022, about 10 quarters ago.
10 We pursue this approach, rather than putting all the variables into one regression equation, to avoid multicollinearity, especially given the sample size is low (only 150 data points). The trade-off will be that all the variables will be weighed equally in the averaging approach we pursue, but as we explain next, we toggle with including and excluding certain variables (e.g., the Treasury curve variables) as a robustness check.
11 Excluding the COVID-19 recession, as that was caused by an exogenous shock (a national health crisis), as opposed to a dynamic intrinsic to the business cycle.
12 Two potentialities we are monitoring closely are an escalation in tariffs and trade frictions under the new U.S. presidential administration, as this could inflame a global trade war, and we are watching the risk related to a military conflict involving Taiwan. On the latter, the St. Louis Fed recently published a paper on the consequences of a military conflict over Taiwan, coming to the conclusion that this would produce a severe economic shock to global trade with enormous human and physical capital costs. https://s3.amazonaws.com/real.stlouisfed.org/wp/2024/2024-034.pdf
We explain why creating economic models is useful even though markets have endured many external surprises in recent years. So what do the models say about 2025?
Joe Seydl
Senior Markets Economist
Published Jan 28, 2025
EXPERIENCE THE FULL POSSIBILITY OF YOUR WEALTH
We can help you navigate a complex financial landscape. Reach out today to learn how.
Contact us
