A recession is a substantial, widespread contraction in economic activity. It is the falling phase of the business cycle. For quant work, the practical distinction is between an observable warning signal and a recession label assigned after the event: confusing them can make a strategy appear to recognise downturns before the necessary information existed.
There is no single definition used everywhere. The Reserve Bank of Australia explains several approaches, including the familiar technical recession shortcut: two consecutive quarters of negative growth in real GDP. That shortcut has a precise meaning, but it is not a universal dating standard.
Two falling quarters: what the arithmetic establishes
Consider a fictional, seasonally adjusted quarterly real-output index, with the initial quarter set to 100. It falls from 100 to 99 to 98. These are index levels from one assumed vintage, not actual economic releases.
| Observation | Output index | Change from previous quarter, not annualised |
|---|---|---|
| Initial quarter | 100 | — |
| Following quarter | 99 | −1.00% |
| Next quarter | 98 | approximately −1.01% |
For the second decline,
There are two consecutive quarterly declines. Across the full interval, output changes by
The percentages do not add exactly because each quarterly rate uses a different starting level. Crucially, this calculation supplies no employment, income or sectoral evidence. Even if the fictional index were real GDP, those three readings alone would not establish a US NBER recession determination.
Fictional seasonally adjusted quarterly real-output index, initial quarter = 100. Growth rates are not annualised. Lines join observations; the chart is not an official recession chronology.
Why US dating uses more evidence
The NBER, an independent research organisation, maintains the widely used US chronology. It evaluates the depth of the fall, its diffusion across the economy and its duration. These dimensions involve judgment: an exceptionally deep and broad contraction can qualify despite being brief. They are not three fixed numerical cutoffs.
The NBER procedure FAQ describes monthly evidence such as real income excluding transfers, payroll and household employment, real consumption, industrial production and inflation-adjusted sales. GDP and gross domestic income also inform quarterly dating. There is no fixed weighted formula; two falling GDP quarters are neither required nor sufficient for its determination.
NBER dates are principally monthly. Under its counting convention, the recession starts in the month after the peak and includes the trough month. The subsequent expansion can begin before activity recovers its previous peak level. A low level of output, a falling level and slow positive growth describe different conditions.
Similarly, a warning about growth relative to trend need not mean outright contraction. The OECD's leading-indicator methodology targets growth cycles around trend and explicitly distinguishes a slowdown from a recession.
A recession date is not its announcement date
The NBER waits for sufficient evidence, including revisions, before announcing turning points. That makes its historical chronology useful for analysing outcomes, but unavailable as a contemporaneous trading signal before announcement.
In a backtest, keep the observation period, publication time and data vintage separate. Using a later recession announcement or revised output history at an earlier decision date introduces look-ahead bias. Record the jurisdiction, series coverage, frequency, seasonal adjustment, real or nominal basis and growth-rate convention alongside each signal. Evaluate a model's real-time recession estimate separately from the eventual chronology.
The Philadelphia Fed's real-time dataset preserves historical snapshots for this purpose. For an advanced owned application, the MV real-time macro forecasting capstone studies release dates and vintages for US industrial-production forecasting. Membership and sign-in are required. It teaches the information discipline; its target is not an official recession classification.
Sources
- Reserve Bank of Australia — Recession: alternative definitions and limitations of the two-quarter shortcut.
- NBER — Business Cycle Dating: US chronology, phase convention and depth, diffusion and duration.
- NBER — Business Cycle Dating Procedure: Frequently Asked Questions: indicators, GDP shortcut and announcement timing.
- OECD — Composite Leading Indicators: Methodology clarification distinguishing growth-cycle slowdowns from recessions.
- Federal Reserve Bank of Philadelphia — Real-Time Data Set for Macroeconomists: historical vintages and release-value datasets.