Characterizing the impact of discrete indicators to correct for endogeneity in discrete choice models