In this paper nonstationary stochastic seasonality is investigated to model
the German M2 money demand function using quarterly data for the period fr
om 1960 to 1996. Empirical evidence is found in favour of a stable long-run
M2 money demand function relying on seasonally unadjusted data while it is
not possible to establish a stable long-run relationship using seasonally
adjusted times series. The seasonal error correction model exhibits satisfa
ctory properties and fits the data quite well. However, it shows parameter
nonconstancy, especially in the middle of the 1970s and at the beginning of
the 1990's when major monetary regime changes took place. (C) 2000 Elsevie
r Science B.V. All rights reserved.