Modelling moving feasts determined by the Islamic calendar

National and religious events always influence economic activity. Islamic events also influence production and consumption. Moreover, these Islamic feasts move over time, depending on the Hegirian calendar, which is based on the lunar cycles, even though, some Islamic countries use officially the Gregorian calendar. The lunar calendar is shorter than the Gregorian calendar, which is based on the cycles of the Earth revolution. Consequently, every year the dates of religious events change in the official calendar creating moving events. Tunisia offers a good example of this phenomenon. Twelve relevant series are analysed and five feasts are considered in our work. Modelling the effect of moving holidays improves the quality of the final adjustment. Removing Islamic feasts from time series is crucial to have better forecasting and comparison results. We adopt an approach initially developed by Bell and Hillmer (1983) to analyse the Easter effect. Since the effect is not the same, we consider three regressors for before, during, and after the holiday for each feast. For model selection and determining the number of regressors and their interval length, two methods are used: the F-adjusted Akaike's information criterion and a criterion based on forecast errors. The empirical results confirm our model selection for all the macroeconomic time series considered except for the exports and the broad money which are not affected by the religious feasts.