Forecasting elections with agent-based modeling: Two live experiments.

Election forecasting has been traditionally dominated by subjective surveys and polls or methods centered upon them. We have developed a novel platform for forecasting elections based on agent-based modeling (ABM), which is entirely independent from surveys and polls. The platform uses statistical results from objective data along with simulation models to capture how voters have voted in past elections and how they are likely to vote in an upcoming election. We screen for models that can reproduce results that are very close to the actual results of historical elections and then deploy these selected models to forecast an upcoming election with simulations by combining extrapolated data from historical demographic record and more updated data on economic growth, employment, shock events, and other factors. Here, we report the results of two recent experiments of real-time election forecasting: the 2020 general election in Taiwan and six states in the 2020 general election in the United States. Our mostly objective method using ABM may transform how elections are forecasted and studied.

Abstract licence: CC BY

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.