The increasing availability of long economic time series poses a challenge for seasonal and calendar adjustment in practice as seasonal and calendar movements now may exhibit changes which are barely identifiable over shorter time periods. The popular X-11 and ARIMA model-based seasonal adjustment methods are capable of dealing with a fair amount of moving seasonality. The respective pretreatment regression models, however, are based on the assumption of constant calendar effects and, thus, do not allow for a direct estimation of ``moving'' calendar effects. As a compromise, official statistics usually follow a rather pragmatic approach based on dividing the entire observation span into several (potentially overlapping) sub-spans and performing separate seasonal and calendar adjustments on each of those. Still, the key question remains: can calendar effects be generally assumed to stay constant over at least several years? Applying structural time series models, this paper exemplarily studies monthly currency in circulation for Germany as of January 1980 up to February 2018. Since data is reported at the last banking day of a month, this series is likely to be affected by the particular weekday which the last day of a given month falls onto. We refer to this effect as the end-of-month effect and find both smooth transitions and sudden changes in these effects over time, providing empirical evidence against the assumption of constant end-of-month effects.