Exploring Spatial Patterns and Determinants of Poverty: New Evidence from Pakistan

This study aims to explore two types of spatial determinants of district level poverty in Pakistan: factors that have direct effect, and indirect or spillover effect, on poverty levels of neighbouring districts. The Spatial Autoregressive (SAR) model has been applied to estimate previously mentioned objectives. Data of 148 districts were collected from the National Socio- Economic Registry (NSER), and provincial development statistics. The Small Area Estimation (SAE) technique provides district level poverty estimates. Empirical results reveal that spatial autocorrelation arises owing to the lag effect of outcome variables, and autocorrelation of error terms with neighbouring districts. Moreover, results are suggestive of factors that have direct influence on poverty levels of respective districts. These include urbanisation, population growth rate, average family size, education, road infrastructure as well as climatic factors (i.e. monthly temperature and rainfall). Apart from direct effects, some determinants of district level poverty have spillover or indirect impact on poverty levels of neighbouring districts. Such factors include level of employment, road length, literacy rate, and climatic factors. Poverty in one district itself has a spillover impact on determining poverty level of adjacent districts. The findings of this paper suggest that the government should enhance regional connectivity, which may be helpful in exploiting the spillover effect of road, health, and education infrastructure to reduce regional poverty levels in Pakistan.