The purpose of this paper is to deal with questions of instability and economic crises, deriving theoretical arguments from Schumpeter's works and presenting relevant empirical evidence for the case of the US manufacturing sector in the time period 1958-2006, just before the first signs of the global recession made their appearance. More precisely, we use a wide dataset that contains 473 manufacturing industries, that are clustered based on their annual change of hourly earnings per worker and we make an attempt to interpret the economic fluctuations in the clusters formed. Meanwhile, we study the causal relationships between the crucial variables dictated by Schumpeterian theory. In this context, a number of relevant techniques have been used, such as hierarchical clustering, canonical discriminant analysis, cointegration analysis, periodograms and Granger causality tests. Our findings seem to give credit to certain aspects of the Schumpeterian theory of business cycles. The results are discussed in a broader context, related to the US economy. (C) 2017 Elsevier B.V. All rights reserved.
