This paper discusses how CBO constructs “synthetic firms”—businesses composed of artificial groups of workers—used in the agency’s health insurance simulation model, or HISIM2, which underlies projections of insurance coverage.
Summary
By Lucas Goodman, Ben Hopkins, Alex Minicozzi, and Eamon Molloy.
This paper discusses how the Congressional Budget Office constructs “synthetic firms”—businesses composed of artificial groups of workers—used in the agency’s health insurance simulation model, or HISIM2, which underlies projections of insurance coverage. Detailed information on a worker’s firm and coworkers is important for modelling whether that worker is offered employment-based insurance, because a worker’s access to such insurance depends on the collective effect of how much employees at the firm demand insurance and how much they will cost to insure. Unfortunately, no complete data source has both rich information on the entire nonelderly population and detailed data on those who are employed. Consequently, CBO uses nationally representative data from the Current Population Survey and supplements those data by building synthetic firms that mimic the composition of actual firms in other data sets. This working paper expands on a slide deck released in April 2020, which detailed both how the agency constructs synthetic firms and how HISIM2 models a firm’s decision about offering health insurance. Specifically, the paper describes how the agency constructs synthetic firms, which are the model’s decisionmaking units for whether or not to offer insurance. Since the publication of that slide deck, CBO has improved its approach, adding data on marital status and health care spending to its method of selecting coworkers for firms. The new data and methods better capture the observed heterogeneity in workforces’ demand for employment-based insurance and thus, in theory, more accurately allow for modelling the insurance coverage effects of a wide variety of policies affecting employers’ offer decisions. The paper includes summary statistics from the newly created data sources that should help others create synthetic firms that are representative of actual firms.