engleski

Calendar adjustments of retail time-series

Fluctuations in economic activity are often influenced by calendar-based various factors. Such factors are non-working (non-trading) days, leap years, public holidays and the like. Most economic series are observed on a monthly or quarterly basis, but months (aggregated into quarters) are not comparable due to the different number of working and non-working days (different number of Mondays, Tuesdays, etc.). If the calendar effects are not properly adjusted, the identification of the ARIMA model for a given time series might not be correct, and the quality of seasonal adjustment is poor. An inappropriate calendar adjustment can generate false signals and negatively affect interpretation of adjusted data, which is particularly important for time series of retail sales and industrial productions. However, there is no general or unique procedure for correcting calendar effects in a pre-adjustment process of a time series. Therefore, this paper compares various regression models using alternative explanatory variables that take into account calendar effects and applied them to the time series of real retail trade turnover (RRT), i.e. monthly data observed from January 2000 to December 2019. The paper seeks to define a new explanatory variable (a regressor with time varying ratio between the average number of working days and the average number of non- working days) providing the most accurate correction of a RRT time series influenced by calendar effects.

Time-series ; calendar adjustments ; retail trade

nije evidentirano