Business Cycle Clustering and Asset Returns
18 Pages Posted: 28 Aug 2018
Date Written: August 17, 2018
Business cycles are a central element of the economy. Over the last decades, more and more sophisticated methods and data have become available for business cycle classification. This paper contributes to the field of business cycle and recession estimation by applying a fuzzy c-means clustering algorithm on a large macroeconomic dataset from the United States. The clustering algorithm separates the data into two clusters which represent economic upswings and downturns. The clustering score not only identifies recessions, it also has strong predictive power on the returns of financial asset classes. Equities perform 6.3% better when in upswing than in downturn. On the other side, government bonds perform 5.8% better when in downturn than in upswing. Applying the clustering score directly to an asset allocation decision, buying more equities in upswings and more bonds in downturns, results in significantly better risk-adjusted returns compared to the asset classes individually. The clustering score therefore offers significant added-value in risk-on/risk-off portfolio decisions.
Keywords: business cycle, clustering, macroeconomic data, recessions, asset allocation timing
JEL Classification: E30, E32, E37, C32, C38, C55
Suggested Citation: Suggested Citation