THE CONCERN
Recent shifts in voter behavior and evolving campaign strategies have fueled doubts about traditional structural forecasting models. These models rely on fundamental variables like demographics, historical voting patterns, and socioeconomic factors to predict election outcomes months ahead of the actual vote.
Our Study
We tested a structural forecasting model across six established democracies: France, Germany, Italy, Spain, UK, US. We compared its predictive accuracy over time against sophisticated campaign data, focusing on whether voter engagement with campaigns reduces prediction reliability.
The Surprising Result
Contrary to expectations that elections are now more unpredictable due to campaign influence, our findings show no decline in the model's predictive power. Fundamental variables continue to yield accurate predictions even as campaign sophistication increases.
This Stability Matters
Our results challenge the assumption that campaigns have fundamentally altered election predictability. They suggest long-term structural factors remain central to electoral forecasting despite the increased emphasis on short-term campaign effects.
| Are Election Results More Unpredictable? A Forecasting Test was authored by Richard Nadeau, Ruth Dassonneville, Michael S. Lewis-Beck and Philippe Mongrain. It was published by Cambridge in PSR&M in 2020. |
