Measuring Business Cycles Beyond Borders: Lessons from Sweden

Mar 24

How constructing leading and coincident indicators for a small, open economy reveals universal principles of business cycle measurement — and where the standard playbook breaks down.

Martin Harrison · Archimedes Research Group · March 2026

Why Study Sweden?

In 2008, as the Global Financial Crisis swept across Europe, Sweden’s GDP contracted by more than 5% — one of the sharpest declines among advanced economies. The country had only been measuring quarterly GDP since 1993, and its modern macroeconomic toolkit was barely fifteen years old. Yet the speed and depth of the downturn posed a question that resonates well beyond Scandinavia: would the same indicators that reliably signal recessions in the United States have provided warning in Sweden?

Most business cycle research centers on the United States. That makes sense — the U.S. produces the deepest set of high-frequency economic data in the world, and the NBER’s recession dating methodology has set the standard for how economists think about cyclical turning points. But anchoring exclusively to one economy creates blind spots. Variables that are powerful predictors in the U.S. may behave differently in economies with different structures, trade exposures, and institutional histories.

Sweden offers an instructive contrast. It is a small, open, export-driven economy where exports account for roughly 47% of GDP — compared to about 12% in the United States. It has a highly egalitarian income distribution (Gini index of 29.25) and a services-dominated GDP composition typical of advanced economies: 66% services, 21% industry, and 11% agriculture. Its population of 10.4 million and GDP of approximately $635 billion put it in a category that is economically sophisticated but structurally distinct from the large, consumption-driven U.S. economy.

Critically, Sweden’s modern macroeconomic measurement infrastructure is relatively young. Quarterly GDP was not measured until 1993, following a nationwide financial crisis from 1990 to 1994 that was triggered by a housing bubble, a credit crunch, and widespread banking insolvency. That crisis served as the catalyst for developing more mature economic measurement tools — a pattern that repeats across economic history. Crises drive measurement. Since 1993, Sweden has experienced four technical recessions measured on a quarterly basis, providing a meaningful sample for testing indicator behavior across multiple cycles.

Building the Indicator Framework

The goal of this analysis was to construct both a Leading Economic Indicator (LEI) and a Coincident Economic Indicator (CEI) for the Swedish economy — not by importing assumptions from the U.S. framework, but by letting the data determine which variables lead, coincide with, or lag GDP in Sweden specifically.

The process follows a disciplined methodology. First, assemble a set of candidate variables that plausibly relate to economic activity: confidence surveys, equity markets, bond yields, construction permits, industrial production, retail sales, and employment. Second, graph each variable alongside GDP and visually assess the relationship. Third, use cross-correlation analysis — essentially measuring how strongly each variable tracks GDP at different time offsets — to statistically determine whether a given variable leads, coincides with, or lags economic output. Fourth, select the variables with the strongest leading or coincident properties. Fifth, construct composite indices by calculating monthly percentage changes in each selected variable and weighting them by their relative explanatory power.

This is the same essential methodology that underpins the Conference Board’s indicators in the United States, the OECD’s Composite Leading Indicators, and — in a more refined form — the indicator suite we are building at Archimedes Research Group. The framework is universal. The specific answers it produces are not.

What the Data Revealed

The cross-correlation analysis produced a clear classification of each variable’s relationship to Swedish GDP — and the results were not what a U.S.-trained economist would necessarily expect.

Leading Indicators:

• Employment Rate (ACF peak: 2 quarters) — GDP lags employment changes, the opposite of the U.S. pattern where employment is typically lagging

• Short-Term Bond Yields, 3-month inverse (ACF peak: 3–4 quarters) — The strongest leading signal, consistent with monetary policy transmission

Coincident Indicators:

• Stock Market Index (ACF peak: 0 quarters) — Moves in real-time with GDP

• Retail Sales Index (ACF peak: 0 quarters) — Near-perfect coincident behavior

• Industrial Production (ACF peak: 0 quarters) — Strong positive correlation

• Building Permits (ACF peak: 0 quarters) — Weaker raw correlation due to pronounced seasonality

• Long-Term Bond Yields (ACF peak: 0 quarters) — Coincident in Sweden despite being considered leading in many other countries

Lagging Indicator:

• Economic Tendency Indicator (ACF peak: 2 quarters) — Confidence surveys reflect rather than predict Swedish conditions

The most consequential finding was the behavior of employment. In the United States, the unemployment rate is a classic lagging indicator — it peaks well after a recession has ended. In Sweden, employment exhibited leading properties, with GDP changes following employment changes by approximately two quarters. This likely reflects Sweden’s labor market structure: strong union protections, active labor market policies, and an export-driven economy where hiring adjustments propagate more quickly.

Equally notable was the behavior of long-term government bond yields. In the United States, the yield curve is among the most reliable recession predictors. In Sweden, long-term yields behaved as a coincident indicator. The short end of the curve, however, did provide a genuine leading signal of three to four quarters, consistent with the transmission lag of Riksbank monetary policy.

Constructing the Composite Indicators

Based on the cross-correlation results, two composite indices were constructed.

The Leading Economic Indicator combined three variables: the stock market index, the employment rate, and inverse short-term (3-month) bond yields. Each variable was standardized using inverse standard deviation weighting. The employment rate received the highest weight (0.86), followed by the inverse short-term yield (0.10) and the stock index (0.04).

The Coincident Economic Indicator combined four variables: building permits, the Economic Tendency Indicator, industrial production output, and retail sales. Industrial production received the dominant weight (0.52), followed by the Economic Tendency Indicator (0.40), retail sales (0.08), and building permits (0.01).

How Well Did They Work?

The CEI tracked real GDP growth with strong fidelity across the sample period. It captured the depth and duration of each of Sweden’s four post-1993 recessions, including the sharp 2009 contraction and the brief but severe COVID-19 downturn in early 2020.

The LEI told a more complex story. In earlier periods, it provided a clear leading signal. However, during the era of negative interest rates in Sweden from roughly 2015 to 2019, the bond yield component introduced significant volatility. When yields approach zero or turn negative, the inverse transformation amplifies small movements into outsized swings, distorting the composite signal.

This is not a problem unique to Sweden — it is a structural challenge that any yield-curve-based indicator faces in an era of unconventional monetary policy.

The Broader Lesson

The Sweden case study illustrates a principle central to how we build indicators at Archimedes Research Group: the relationship between economic variables and the business cycle is not fixed. It varies across countries, across time periods, and across monetary policy regimes.

This has direct implications for U.S. indicator construction. The yield curve inversion of 2022–2024 was the longest and deepest since the 1980s, yet the widely anticipated recession did not materialize. The same structural issue that corrupted our Swedish LEI in a negative-rate environment is present, in milder form, in the United States today.

The solution is not to abandon any single variable but to build composite indicators that are robust to structural shifts — drawing on multiple data sources, validating against multiple cycles, and regularly re-evaluating the signal properties of each component. That is the approach we take with the ARG Recession Probability Model and the broader suite of indicators we publish.

Business cycles are fundamentally similar across advanced economies. But the specific way those drivers manifest in observable data is highly context-dependent. The only way to build indicators that actually work is to let the data tell you what leads, what follows, and what moves in real time. Assumptions are the enemy of accuracy.

Sources: Statistics Sweden (SCB), OECD, Riksbank, Federal Reserve Bank of St. Louis (FRED), The Conference Board.

Marty Harrison https://www.archimedesresearchgroup.com