The cross sectional data collected from time to time had been used to analyze and model the determinants of poverty but no attempt have yet been made to analyze it on the provincial level in Pakistan. MICS (Multiple Indicator Cluster Survey of Punjab) 2003-04 with more than 30,000 households provides this opportunity and this paper tries to model the various demographic and socio-economic determinants of poverty. A logistic regression analysis has been carried out taking two dependent variables of Per capita monthly expenditures and per capita monthly calorie intake. Different households are classified either poor or non poor on the basis of a threshold monthly per capita expenditures of PKR 848.798 and a daily calorie intake of 2350 calories. The results show that age, education and gender of the household head significantly explain the variations in the likelihood of being poor. Moreover, households receiving remittances and holding agriculture land are more likely to exit from the poverty trap. The dependency ratio and larger family size positively affect the possibility of entering the poor household group. The employment sector also significantly explains for the cross regional and geographical differences in the poverty determinants. The empirical results for the three mutually exclusive regions of the Rural, Other Urban and the Major Cities suggest considerations for the policy makers and provide poverty dynamics over these regionally differentiated localities.