Reading the Business Cycle: Choosing the Right Time Horizon

Most cycle-reading errors are not data errors. They are horizon errors, a window too short to filter noise, applied to a phenomenon that unfolds over years.

Cycle analysis is biased less by the data themselves than by the window through which the data are read. An isolated quarter does not produce the same diagnosis as a sequence of three to five years, and the same release will support opposite conclusions depending on the horizon retained. Filtering tools, moving averages, underlying trends, cumulative deviations, exist precisely to separate signal from noise. Their absence is the structural source of most recurring diagnostic reversals: economies declared in recession that keep growing, recoveries called too early, slowdowns missed until they are confirmed by the labor market.

The angle that escapes the news flow is precisely the choice of observation window. Most cyclical commentary anchors itself on the latest published quarter or the most recent monthly print. The focal length is too narrow for the phenomenon being read, and it produces an interpretive volatility that the underlying economy does not have.

The isolated quarter is not a signal

A single quarter of growth does not establish an expansion, and a single contraction does not establish a recession. The Bureau of Economic Analysis measured a 0.6% annualized contraction in US GDP in the first quarter of 2025, immediately followed by a 3.8% rebound in the second and 4.4% in the third: read one at a time, those three quarters tell three opposite stories. The four-quarter moving average stayed between 2.0% and 2.4% throughout, signaling an unchanged trajectory. The real business cycle reads on multi-quarter sequences, not on isolated prints.

Statistical revisions compound the problem. US Q1 2025 GDP was revised three times between the first estimate of 30 April 2025 and the 25 September release: -0.3%, then -0.2%, then -0.5%, then -0.6%, a net gap of 0.3 point with a change of direction along the way. A diagnosis anchored on the first estimate produced different conclusions than the same diagnosis run on the final revision, a structural feature of national-accounts production, not a quality problem with the BEA. The recurring pitfalls of short-window economic reading include treating a single release as informative independently of its revision history. It rarely is.

Match the horizon to the question

Three horizons coexist in cycle analysis, and they answer different questions. The short term, one to three quarters, captures cyclical fluctuations: inventory swings, one-off shocks, base effects. It is the relevant window for traders pricing the next release. The medium term, three to five years, reveals investment, credit and productivity dynamics, which is where the actual business cycle plays out. This is the window covered in our analysis of the real business cycle beyond the quarterly prints, which details expansion and turning-point phases against the noise of the high-frequency calendar. The long term, ten to thirty years, surfaces industrial cycles and structural transformations: demographics, technology adoption, financial regulation regimes. Each horizon answers a different question, and conflating them produces incoherent diagnoses. A related answer: the lead-lag between credit and the real economy.

The divergence between the real economy and financial markets illustrates the same point from the market side. Markets price expectations on a 6 to 18-month horizon while the real economy unfolds over years. An equity rally does not mechanically signal a macro recovery, and an index drawdown does not necessarily foreshadow a recession. The two most discussed episodes of the century point in opposite directions: the S&P 500 fell 25.4% between 3 January and 12 October 2022 with no recession following, and it peaked in the autumn of 2007, weeks before the cycle peak the NBER dates to December 2007. An indicator that errs in both directions is a reminder that the asset-market window and the real-economy window do not synchronize.

Higher frequency does not solve the horizon problem

The proliferation of high-frequency data, card spending, air traffic, mobility indices, electricity demand, satellite indicators, promised to reduce dependence on quarterly releases. These sources do offer unprecedented granularity and very short publication lags, and they have proved useful as cross-checks during structural breaks such as the 2020 shock. They have not, however, replaced traditional cycle analysis. Coverage is partial, aggregation across sources is methodologically delicate, and the very granularity that makes them informative also makes them noisier on a per-print basis.

The lesson is not that the new data are useless, they are valuable as confirming or contradicting evidence, but that data frequency is not a substitute for horizon discipline. The analytical frameworks of the business cycle point to a regularity: the quality of a diagnosis depends less on how often the data are released than on whether the horizon chosen to read them matches the horizon of the phenomenon being measured. Cycle questions deserve cycle-length answers, regardless of how frequently the underlying series prints.

Last updated — 22 September 2026