Weigh(t)ing the dimensions of social vulnerability based on a regression analysis of disaster damages

Abstract. Social vulnerability determines how natural hazards can turn into social disasters. However, large uncertainties remain on how to quantify social vulnerability in a compact and comprehensive way. A principle component analysis (PCA) combines many social indicators, such as population density or the level of education, into a single vulnerability index. Whether the individual indicators increase or decrease vulnerability is usually based on consequential reasoning or derived from previous studies. These assumptions are rarely tested for their applicability to the study area. In a case study for the Austrian federal state of Styria we observe no correlation between disaster damages and an initial vulnerability index which was based on literature-derived assumptions about the influence of each indicator and following best practices for the PCA. We show that a regression analysis of past damages can improve the interpretation of social indicators by better representing the local situation. With the results from the regression-based analysis we can improve the PCA approach and only the updated vulnerability index correlates with observed damages. This indicates that the use of a PCA-based vulnerability index requires a thorough understanding about the regionally specific influence of each indicator on social vulnerability to explain differences in disaster damages.