Principal component analysis applied to multidimensional social indicators longitudinal studies: limitations and possibilities

Abstract

Principal component analysis (PCA) is a popular technique for building social indicators in the field of spatial analysis. However, literature shows that there is no consensus on how to apply PCA to longitudinal studies, and researchers have done the analysis using different approaches, varying the way data are combined and the frequency in which the data are sampled. This research explores such approaches with two objectives: to draw attention to the limitations of using PCA in longitudinal analyses, and to show how to overcome these limitations. For this purpose, indicators of urban inequality of eight cities are compared in each approach. The results show that the use of PCA presents limitations for the longitudinal study of urban inequality either because the evolution of the phenomenon is not always captured, or a large part of the indicators does not explain the phenomenon properly, or yet when a change in the calculation of the indicator distorts and enhances the differences in urban inequality through the years. An analytical chart is proposed to guide researchers with explanations and justifications that should accompany the use of PCA in longitudinal analyses.