Innovative Approaches to Demographic Data Collection and Methods

Ruben Castro, University of Pennsylvania

"Many surveys include both spouses and/or more than one wave, so consistency measures can be computed from the data. But so far, literature about consistency focuses mainly on the joint distribution of consistency and salient variables of the population, raising for example the propensity of men to over-report economic accomplishments while women report as vulnerable. This paper gives such descriptions, especially in relation to sexual behavior, but it also deals with the alternatives to go further and actually include some consistency measures back on regular data analysis, with the goal of recovering better estimates from reported data. The article, though, relies on empirical data from the MDICIP project in Malawi; after some interpretation is made on each particular consistency measure, it analyzes the alternatives of using them as either measurement errors, importance weights as in sampling methodologies, latent variable in the sense of IRT models, or proxies for non-observables".