This paper investigates the existence of club convergence on the NUTS (Nomenclature des Unités Territoriales Statistiques) 3 level in Serbia. While a common approach in investigating convergence is based on dividing units of observation a priori into individual groups based on some of their particular characteristics, we use a method developed by Phillips and Sul (2007) that allows identification of clusters of convergence on the basis of an algorithm that is data-driven and thereby avoids a priori classification of the data into subgroups. We use data on GVA per capita for the NUTS 3 level in Serbia for the period 2001-2017. Our results show that there are two convergence clubs in Serbia, while the Belgrade district shows no signs of convergence with any of the other clubs.
