Government of Nepal provides quotas and reservations for women, indigenous nationalities,
Madhesi, untouchables, disables and people of backward areas. These statuses are not
homogenous in economic sense. We proposed few other decision trees (rules) that can predict
household poverty in Nepal based on 14,907 household observations employing classification
and regression tree (CART) approach. These decision rules were based on few practically
answerable questions (for respondents) and can be cross checked easily by the enumerators. We
modeled 5 different scenarios that respondents were likely to answer the asked questions. These
decision rules were 94% to in worst-case scenario 70% accurate in out-of-sample dataset. These
proposed meaningful decision rules can be helpful on policy making and implementation that
relate to positively discriminate (quota and reservation) for those who lie below poverty line.