In recent years, the labor economy has seen great change due to the introduction of online platforms that facilitate gig work. The gig economy includes these jobs, like ridesharing, but more broadly all sorts of less-structured work arrangements, ranging from babysitting to independent contracting. This study seeks to measure the prevalence of gig work and its impacts on workers’ wellbeing. This is a difficult task given the complexity of categorizing gig workers and the need for a data source including information about an individual’s wellbeing in addition to their work arrangements. This project takes advantage of the University of Michigan’s own Panel Study of Income Dynamics (PSID), which is uniquely qualified for the task given its longitudinal nature and depth of information. The project involves first hand-sorting a small subset of the restricted narratives of work arrangements into a variety of employment roles. Themes identified in this round of coding are used to create a complete coding schema that is then used to code a much larger subset of data. These coded data are then pre-processed and used to train candidate machine learning models. These models are tested and the approach (by hand, a model, or some combination) that is best able to classify workers is chosen to classify the rest of the data. The full set of classified data are then combined with publicly-available PSID data to identify differences in outcomes, particularly those measuring wellbeing, across employment roles. The project will produce novel data measuring gig work, and the analysis will provide a new measure of the number of gig workers and estimates of the impact of gig work on worker wellbeing. The results have the potential to impact individuals’ decisions regarding entrance into the gig economy, inform policy decisions regulating it, and improve existing tools measuring gig work.