Referendum Model Forecasts: Trump and the 2022 Midterm Errors

Abstract

To forecast the 2022 congressional races, we returned to our structural model, which we have utilized to good effect since 2010. Our forecasts for 2022 were published before the election. The model rests on strong theory and expresses itself in a political economy equation. For the House, which we focus on in this brief essay, midterms are assumed to be referenda on the president and the incumbent party, where voters reward or punish according to key economic and political issues, as measured by aggregate indicators. These ex ante national forecasts we adjust a bit, to account for local conditions via expert judgement. In 2022, our Structure-X forecast for the House foresaw a Democratic loss of thirty-seven seats. This prediction is correct in that it foretells the ruling party would experience a net loss, so upholding the “iron law” of midterm incumbent performance. Moreover, this loss technically falls, just barely, within the 95% confidence interval for the structural OLS equation (i.e., 37 1/2 (1.96 # 18.82) 5 [.11 to 73.84].) Thus, when strictly judged as an outlier, it lands on the line. Nevertheless, there is no denying the error of twenty-eight seats looks large (given the actual Democratic loss of nine seats). Here we begin to assess the source of