DP18960 Accounting for uncertainty and bias in archaeological and historical evidence on wealth inequality
We provide methods and code that transform raw wealth data into probability distributions representing likely levels of inequality at a particular site and date that (insofar as possible) are representative of the underlying population (e.g. rectifying biases arising from small or non-random samples including the frequent absence of data on those entirely without wealth) and comparable across differing asset types (e.g. burial goods, dwelling area, storage area, land cultivated), ownership units (individual or household), and scale (from villages to nations). These distributions are based on the raw data along with independent information on how it came to be available. Because we account for the propagation of uncertainty through every stage of the data generation and estimation process, the degree of uncertainty indicated by most of these (non-Gaussian) probability distributions is substantially greater than what one would infer from conventional bootstrapped estimates.
