Living near dumpsites: A machine learning and econometric assessment of how environmental conditions shape rental prices

Abstract

This research examines the impact of environmental (dis)amenities on residential rental values in the urban areas of Rawalpindi and Islamabad, Pakistan. Using a unique dataset of 849 households and geospatial data on 35 irregular dumpsites, we quantify how proximity to environmental disamenities depresses rental prices. Specifically, results confirm that irregular dumpsites significantly depress rental values, especially for properties situated near the closest distance rings. The analysis employs a hedonic pricing model, complemented by instrumental variable (IV) mediation analysis and machine learning (ML) classification models, such as Naïve Bayes, k-nearest neighbours (k-NN) and classification trees, to explore both causal relationships and predictive patterns. The IV mediation approach confirms that the presence of odorous sewers significantly mediates the negative effect of dumpsites on rent. ML models, particularly k-NN, demonstrate high predictive accuracy (>90%) in identifying high-rent properties based solely on environmental variables. These findings emphasise the economic cost of environmental degradation in urban housing markets and highlight the necessity of stricter waste management policies and improved sanitation infrastructure to drive sustainable urban development.