Ivan Puga-Gonzalez, Rachel Bacon, David Voas, George Hodulik, Wesley Wildman and I have just had a paper titled “Adapting Cohort-Component Methods to a Microsimulation: A Case Study” published in Social Science Computer Review. Available for download here.
This is the first publication (of no doubt many!) from the Modeling Religious Change project at the Center for Mind and Culture in Boston.
Abstract: “Social scientists generally take United Nations (UN) population projections as the baseline when considering the potential impact of any changes that could affect fertility, mortality or migration, and the UN typically does projections using the cohort-component method (CCM). The CCM technique is computationally simple and familiar to demographers. However, in order to avoid the exponential expansion of projections as new dimensions of individual difference are added, or to understand the sensitivity of projections to specific conditions, agent-based microsimulations are a better option. CCMs can mask hidden assumptions that are surfaced by the construction of microsimulations, and varying such assumptions can lead to quite different projections. Models based on the CCM are naturally the strongest form of validation for population-projection microsimulations but there are many complexities and difficulties associated with matching microsimulation projections and CCM projections. Here we describe our efforts to tackle these challenges as we validated a microsimulation for Norway by replicating a UN CCM projection. This provides guidance for other simulationists who seek to use CCMs to validate microsimulations. More importantly, it demonstrates the value of microsimulations for surfacing assumptions that frequently lie hidden, and thus unevaluated, within CCM projections.”