Constructing a composite coincident indicator for a post-transition country

The aim of this article is to construct a monthly coincident indicator of real economic activity in Croatia. For that purpose, we use a database containing altogether 278 time series, ranging from January 1998 to December 2010. In step one we use correlation analysis, logit and Markov switching (MS) model in order to select time series that closely follow the overall business cycle and its turning points. The following four series have been detected as having the best coincident properties: industrial production, volume of retail sales, VAT revenues and total credit to households. In step two we apply dynamic factor model methodology to the aforementioned coincident series in order to estimate their common component, which is then used to construct a monthly coincident indicator of real economic activity.