This paper aims at investigating the primary fiscal balance of Brazilian municipalities from 1999-2007 by employing the bias selection methodology for in-panel data proposed by Wooldridge (1995). More specifically, the study assumes that fiscal surpluses and deficits should not be taken as purely random events; alternatively, it assumes the existence of a first stage that determines the existence of a fiscal surplus or deficit. The results, controlled for the presence of sampling bias, show that fiscal surpluses and deficits are partially influenced by the fiscal characteristics and demographic structure of the population. However, political traits do not exert any significant influence on the fiscal balances of Brazilian municipalities.