Over the last 25 years neighbourhood economic outcomes have become increasingly polarised in Australia. The growing spatial dimensions of this inequality have generated discussion about the existence of ‘neighbourhood effects’, localised externalities and other endogenous processes, leading to underinvestment in education, lower levels of job-creation and economic activity, than might be expected in disadvantaged neighbourhoods. Recently there has been renewed interest in economic models incorporating social interactions, dependence among economic agents and spatial spillovers. In failing to acknowledge the interdependence between neighbouring regions and individuals, traditional economic models ignore an important aspect of space. A range of spatial regression techniques have been developed to measure latent forces of interaction and handle data that violate statistical assumptions of independence. This paper explores the drivers of changing labour market outcomes within Metropolitan Sydney in 1996 and 2001 (POA level). Moran statistics will be used to determine whether these represent statistically significant clusters of employment and other demographic variables, known as ‘hot spots’ (high unemployment) and ‘cold spots’ (low unemployment). Spatial econometric techniques are them employed to ascertain the key drivers of these ‘hot’ and ‘cold’ spots; and the extent to which, independent of traditional explanatory factors, location-specific interactions within or between regions might be responsible for any observed polarisation.