UC Riverside

In any month, administrative data collected by government agencies contain a fraction of the politys adults, namely those who have interacted with government agencies in that month. For researchers and policymakers who want to evaluate questions that require a spatial location of the whole population of adults at a given time (e.g. job-residence spatial mismatch, impacts of local policies), these fragmentary records are insufficient. Combining administrative data from several agencies in the State of Washington, United States (US), we impute residential histories by parameterizing the decay in maintenance of an observed address. This process yields an imputed population whose demography and geographic distribution matches well with survey estimates. This work uses driving licence, voter, social services, and birth records to append address locations to Unemployment Insurance data, a process that could be replicated with administrative records in other US states and countries with sporadic address data from various agencies.