Extended Data Fig. 5: Extended version of Fig. 4, showing the correlation of a larger number of indicators along their main principal component versus k block complexity.

Symbol size denotes population density (corresponding to more urban regions) and colours show the specific indicator variation along the PC1 variation with block complexity, k. We observe that measures of advantage, such as percent of women with secondary or higher education, of households with a refrigerator, of population with electricity or births delivered at a health facility are all anti-correlated with k and decline with each unit increment by −4.2, −1.9, −4.3 and −5.5%, respectively. Conversely, measures of disadvantage such as percent of population with natural floors, of population using open defecation,  < 5 child mortality rate, and women aged  > 6 with no education are all positively correlated with k, with each increment leading to a 3.5, 1.9, 7.5 and 3.5% increase in these metrics.