We studied Semeru, East Java, Indonesia, to show the population exposure to volcanic threats from its persistent, daily eruptive activity which endangers at least 50,000 of the 950,000 inhabitants living on the East, South and South-East slopes and ring plain. Surveys, mapping and statistical investigation enabled us to assess the extent of exposure of 145 neighbourhoods (termed blocks) and characterize hazards and response to eruptions in 15 rural villages and small towns. Statistical analyses of datasets of 23 variables (11 of exposure, 7 of hazards, and 5 of response) and their attributes involved three operations: 1. Univariate and bivariate analyses enabled us to explore data and characterize the relation- ships between 11 variables to compute a multi-component exposure index. 2. Polytomous Logistic Regression (PLR) models selected six optimal exposure variables, suggesting that logistic regression can predict the exposure index for blocks outside the survey area and potentially on any active volcano. 3. Multivariate analyses and Hierarchical Agglomerative Clustering (HAC) distinguished four groups of blocks based on attributes of all variables correlated with the exposure index score. To contribute to disaster risk reduction, the distance/time criterion was applied to access ways and response facilities to highlight remote or blocked blocks in danger of imminent eruption including evacuation. Statistical analysis of optimal variables from local scale surveys can help identify neighbourhoods where disaster risk mitigation requires improvement.