Value of information and trade-off analysis for rural road investment frameworks: new insights from MCA sensitivity, waste per road, and cluster stability

Abstract

Governments in low-income countries often spend millions upgrading rural roads that yield negative economic returns, yet simpler decision tools risk leaving isolated communities without access. This paper extends our previous comparison of four decision methods for 166 rural roads in Sri Lanka with five new analyses. Value of Information (VoI) measures the extra economic benefit of using full integer linear programming (ILP) instead of PCA-clustering. The gain is largest at high budgets (Rs.400M at 100%) but only exceeds a hypothetical full evaluation cost (Rs.332M) at the 100% budget. Multi-criteria analysis (MCA) weight sensitivity shows that giving 50% weight to the benefit-cost ratio achieves the highest accuracy (92.8% at 25% budget), while condition-heavy weighting (IRI = 50%) performs poorly (38.3%). Waste per negative-NPV road is zero for PCA-clustering until the 55% budget, then rises to Rs.5.9–9.2M, whereas IRI ranking wastes Rs.19–69M per bad road. Cluster stability reveals a clear split: roads are funded either at 40% or only at 70% budget, confirming that PCA-clustering separates investable from non-investable roads. The results support a pragmatic rule: for budgets up to 40%, PCA-clustering or economic-heavy MCA deliver near-optimal efficiency and higher population reach; full ILP is justified only when economic data are already available, or the budget is very high (85%).