Equity Markets and the Economic Cycle: Stocks Call Recoveries Better Than Recessions

A share price discounts future profits, while the economy is measured on flows that have already happened. Equities therefore move before activity, but they do not lead it with the same reliability on the way into a recession and on the way out.

Nine US recessions since 1960 make the gap measurable. The lead is real, its direction never changes at the exit, and it slips away at the entry.

What an index actually anticipates

A share price summarizes what investors expect from a company’s future earnings, brought back to today with a discount rate. When those expectations or that rate move, the price moves the same day. Activity data record instead what has already been produced, sold or hired, and publish it weeks later.

That difference in nature sits at the heart of the efficient-markets hypothesis Eugene Fama set out in the Journal of Finance in May 1970: a price reflects the information available. It says nothing about the quality of that information. A market can react early to a turning point that will come, and just as early to one that will not.

Hence a common reading that is half right: “the stock market leads the economy by about six months.” The phrase assumes a symmetric lead, as reliable at the top as at the bottom. The data contradict it on that specific point.

Stocks call recoveries better than recessions.

Nine recessions, two turning points

To measure it, we used the total-return index of the US market published in Kenneth French’s data library: every listed stock in the CRSP database, dividends reinvested, monthly through August 2026. We set it against the peak and trough dates of the National Bureau of Economic Research (NBER) and nonfarm payrolls from the Bureau of Labor Statistics. For each recession, the market trough is the index low between the cycle peak and twelve months after its end, and the market top is the high of the three years before that trough.

Going in, the market top led the cycle peak by 0 to 13 months, with a median of 4. The spread is wide: 13 months in 1969 and 11 in 1973, but 2 months in 1990 and 2007, 1 month in 2020, and none in 1980, when the index peaked in January, the very month of the cycle peak.

Coming out, the order never changes. The market trough came before the payrolls trough in all nine recessions, by 1 to 12 months, median 5. It also came before the official end of the recession eight times out of nine. The one exception is 2001: the index bottomed in September 2002, ten months after the trough dated by the NBER, yet eleven months before jobs did.

That exception has an identifiable cause. In December 1999, Robert Shiller’s CAPE stood at 44.2, the highest reading in a series that starts in 1881. By September 2002 it was back to 22.4. Prices were no longer just tracking profits: they were working off a valuation excess, and that process ran on after the recession ended.

The tool above places each turning point on a monthly timeline. In “entry” mode, the origin is the market top and the comparison point is the cycle peak. In “exit” mode, the origin is the market trough, against which the end of the recession and the payrolls trough are plotted. The three declines without a recession appear for what they were: tops with nothing behind them.

On the way in, a signal that fails to fire or misfires

The first flaw of the entry signal is that it is often missing. Of the nine recessions since 1960, four were neither preceded nor accompanied by a 20% market decline: 1960 (−8.0% from top to trough in monthly data), 1980 (−12.0%), 1981 (−16.3%) and 1990 (−16.9%). Recessions have started without the 20% decline that usually defines a bear market.

The second flaw is the false start. Since 1950 the index has lost more than 20% eight times on monthly data. Five of those declines came with a recession. Three were followed by none: 1962 (−23.1%), 1987 (−29.9%) and 2021-22 (−24.8% between December 2021 and September 2022).

Behind the asymmetry lies a mechanism. A market top can come from a higher discount rate with no change in expected profits, as in 2022, when the Federal Reserve raised its target range by 4.25 percentage points. It can also come from a pure valuation shock, as in October 1987. In both cases prices fall while activity carries on. A trough arrives when expected profits stop deteriorating and monetary policy has already turned, two conditions that come before hiring picks up.

Monetary policy shows up in the numbers. In eight of the nine recessions, the federal funds rate had already peaked when the market bottomed, 2 to 26 months earlier, with a median of 11 months (monthly averages of the Federal Reserve’s FEDFUNDS series). In 1980 both turned in the same month, March, in the middle of Paul Volcker’s tightening. The market recovers on the rate cuts to come, payrolls on the recovery in sales, which comes later.

On the way out, a market ahead of jobs

The exit is where the market’s lead is easiest to see, because payroll data are then at their weakest. Between the index trough and the payrolls trough, the US market gained a median 28.2% across the nine episodes. The range runs from +13.6% in 2020, when jobs bottomed in April, one month after the market, to +55.6% between February 2009 and February 2010.

Dating lags widen the gap. The NBER only dates a turning point after the fact, once the data have been revised. The June 2009 trough was announced on September 20, 2010, fifteen months later. The April 2020 trough was announced on July 19, 2021. The December 2007 peak was only made official on December 1, 2008, by which time the index had already lost more than 40% from its October 2007 high, according to the announcements of the NBER dating committee.

In real time the market therefore has two leads: one from expectations, one from dating. Employment adjusts late for reasons specific to firms, such as the cost of hiring and doubts about the recovery, a pattern covered in why employment lags the business cycle.

Why the index and the economy measure different things

Even without a time lag, a stock index and GDP do not tell the same story. The first weighs profits, the second value added. Corporate profits amplify the cycle in both directions, because fixed costs do not shrink as fast as sales.

The amplitude is not comparable. Across the nine recessions since 1960, US corporate profits fell a median 14.9% from peak to trough in current dollars, while real GDP lost 2.2% (Eco3min calculation on the BEA’s CPROFIT and GDPC1 series). In 2007-08 the gap reached −35.4% against −3.8%. Profits are in nominal terms and GDP in volume: inflation softens the measured fall in profits, which makes the difference in amplitude starker still.

Geography adds a further wedge. Profits that US corporations receive from the rest of the world were equivalent to 26.4% of their total profits in 2025, against 35.8% in 2019 and 32.7% in 2010 (Eco3min calculation on national accounts from the Bureau of Economic Analysis). Part of the earnings the market discounts depends on demand outside the United States, which is one reason GDP growth is a poor guide to equity returns.

The level of valuation, finally, shows how far ahead the market is looking. Robert Shiller’s CAPE, which divides price by average real earnings over the past ten years, stood at 40.6 in September 2026, against a median of 16.6 over 1881-2025. At that level, prices discount profit growth far into the future and become more sensitive to the discount rate than to the current quarter. That channel is set out in how real rates drive equity valuations, and the measure itself in the Shiller CAPE and its long-run reach.

2022-2026: a bear market without a recession

The latest episode shows both sides of the signal. Between December 2021 and September 2022, the total-return index fell 24.8%. The 10-year minus 2-year Treasury spread turned negative on July 6, 2022, and its last negative session came on September 5, 2024: 792 days, the longest inversion since 1976. Both classic recession warnings were flashing.

No recession followed. The NBER has dated no peak since February 2020, and the USREC series that carries its dates on FRED stays at zero through September 2026. Industrial production did contract, but its low came in January 2024, fifteen months after the market’s. Unemployment, at 3.4% in April 2023, had only risen to 4.2% by September 2026.

From the end of September 2022 to the end of August 2026, the index returned 123.7% including dividends. As a recession signal, the December 2021 top was wrong. It did flag a rise in the cost of capital, which higher policy rates then delivered. How markets can rise in that setting is examined in why bad economic news can lift stocks.

What weakens this reading

Nine recessions are a small sample, and results depend on method. With end-of-month data, a top or trough lasting a few days can shift by a month. A shorter or longer search window moves a few episodes by a month or two. Across the two window definitions tested, the direction of the results does not change.

The US market is not the world economy either. A recession in Europe or China can weigh on the index without showing up in NBER dates, and the growing weight of large technology stocks changes how sensitive the index is to the domestic cycle. Finally, the lead at the exit owes something to central banks, which cut rates during recessions. In a downturn where they could not, for lack of room or because of inflation, the market trough could come later.

What the equity cycle can tell you

The most robust reading concerns the order of troughs. When the index has stopped falling while payrolls are still shrinking, the history of these nine recessions says the market has moved ahead, not that it is wrong. That gap feeds the familiar sense that bad economic news can be good news for stocks.

On the way in, the data warn against reading an index decline as a recession call. A sharp fall in the index measures a repricing of risk, which growth may or may not confirm. It still needs to be read alongside activity indicators such as jobless claims, the Sahm rule and credit, and five recession rules tested on 75 years of data give their hit rates. One limit weighs on any real-time use: a market trough can only be identified after the fact, and the index has often rebounded and fallen again before its true low, as in 2001-02 and 2008. The historical finding is about the order of events, not a counter to track: in the nine episodes, payrolls bottomed no more than twelve months after the market.

Last updated — 8 October 2026