Characterizing the Luxembourg financial cycle: alternatives to statistical filters

This paper studies the cyclical properties of bank loans to non-financial corporations, bank loans to households, bank loans to the non-financial private sector and house prices in Luxembourg, by applying two methodologies to decompose and characterize the cycles. First, we use an unobserved components model (UCM) to extract classical cycles. We find evidence of medium-term cycles in the loan series over the sample 1980 Q1 to 2019 Q1. The nonfinancial corporation credit cycle is the most volatile and of larger duration compared to the other credit cycles. The length of the house price cycle in Luxembourg is estimated at 13.8 years. A dynamic synchronicity measure between households’ credit cycle and the house price cycle reveals that these cycles are sometimes synchronous. However, the early warning properties of the univariate unobserved components model are limited despite its good pseudo real-time estimates. A wavelet analysis complements the findings of the univariate UCM by providing growth cycles of financial variables. We show that these growth cycles could be useful as complementary information in financial cycle analysis. A coherence wavelet analysis is also conducted. The main findings show that the credit growth cycles of non-financial corporations and of the non-financial private sector are highly coherent with respect to households’ credit growth cycle. The household credit cycle often precedes house price growth. From a macroprudential policy perspective, these results support the use of complementary methodologies for assessing the financial cycle and confirm earlier studies on the limited role of the household credit cycle in the evolution of the house price cycle.

Keywords: credit cycle, wavelet analysis, unobserved components model, house prices

JEL codes: C30, E32, E51