Replication Data for: A Long-Range State-Level Forecast of the 2020 Presidential Election

Most U.S. presidential election forecast models produce forecasts three to five months in advance of the election, and often only generate a prediction of the national popular vote. This paper presents a model that generates predictions of state-level outcomes a year in advance of the election, which can then be extrapolated up to national popular vote and Electoral College projections. This model suggests a high probability of Biden victory in 2020.当前绝大多数美国总统选举预测模型均会在选举前3至5个月开展预测工作，且通常仅能输出全国普选票的预测结果。本研究提出的模型则可在选举前一年即生成各州选举结果的预测，并可据此外推得到全国普选票及选举人团（Electoral College）的预测结果。该模型预测拜登在2020年美国总统选举中拥有较高的获胜概率。