Reproduction Analysis: A Reanalysis of a Social Vulnerability Index
Additive vs. Weighted Social Vulnerability Models
A reproduction and reanalysis of Spielman et al. (2020) seeks to use Spielman et al’s methodology to reproduce their application of the Social Vulnerability Index (SoVI). Subsequently, the reanalysis extends and evaluates the original SoVI model construction by adjusting Spielman’s original additive SoVI model to weight factor components by the proportion of variability each factor explains. Furthermore, this reanalysis tests the internal and theoretical consistency of a social vulnerability index model. If there is internal consistency in the model, the relative rank of SoVI scores by county should remain relatively consistent regardless of the geographic extents used as inputs for the model.
My main contribution to the reproduction reanalysis of Spielman et al. (2020) is the addition of an interactive map that depicts results of SoVI score outputs by county calculated at the national scale. The results show the least and most vulnerable counties. It is interesting to note the spatial clustering of lower SoVI values in metropolitan areas of the mid-Atlantic region and higher SoVI scores along the US-Mexico border and in Alaska. In addition, I created an output table for factor loadings of the national SoVI model, calculated the percent variance explained by each component of the national SoVI model, and contributed to the discussion of the analysis. The results of the reanalysis highlight differences in results based on geographic inputs and implementing additive versus weighted SoVI models. Furthermore, given the sensitivity of results based on model inputs and construction, policy makers and development practitioners should be aware of the inconsistency of findings when prioritizing vulnerable people and places to implement disaster risk reduction efforts.