How to Handle Outliers and Influential Cases

• In Event: Political Forecasting Methodology

Abstract

The 2020 economy and election was turned upside down by the pandemic. Forecasting models have historically relied upon a variety of economic indicators (e.g., GNP, GDP, leading economic indicators, disposable income, and unemployment) in their predictions. The pandemic and the global and national economic disruption it caused resulted in most economic indicators fluctuating wildly during the first two quarters of 2020, exactly the period most forecasting models use in their models. As a result, the economy in 2020 was an outlier in forecasting data going back more than half a century. In this paper we present strategies for how to forecast in an environment with outlying data and influential cases. We place our discussion in the larger context of measurement error issues and their consequences for structural forecasting models.

Authors

• Michael S. Lewis-Beck, University of Iowa

• Charles P. Tien, CUNY, Hunter College & The Graduate Center