Abstract Tax competition may be different in ‘new economic geography settings’ compared to standard tax competition models. If the mobile factor is completely agglomerated in one region, it earns an agglomeration rent which can be taxed. Closer integration first results in a ‘race to the top’ in taxes before leading to a ‘race to the bottom’. We reexamine these issues in a model that produces stable equilibria with partial agglomeration in addition to the core–periphery equilibria. A bell-shaped tax differential also arises in our model. Therefore, the ‘race to the top’ result generalises to a framework with partial agglomeration.