Evidence-based election forecasting since 2004
- 20 years of evidence-based election forecasting
- Based on established forecasting principles
- Highly accurate election forecasts
- Widely published in the scientific literature
In 2004, PollyVote was initiated as a lasting initiative, aiming to showcase the effectiveness of evidence-based forecasting to a broad audience by applying forecasting principles to the prominent context of election forecasting.
Evolution of the PollyVote
PollyVote was founded by forecasting expert J. Scott Armstrong and political science professors Alfred Cuzán and Randall Jones.
PollyVote is founded on the principle of combining forecasts, a key discovery derived from over half a century of forecasting research, which has long been successfully applied in various fields, including economics, meteorology, and sports.
Since 2004, PollyVote has combined forecasts from various validated component methods. During the U.S. presidential elections in 2004 and 2008, PollyVote averaged forecasts within and across four component methods: polls, prediction markets, expert judgment, and econometric models. Over time, the original specification evolved, with two new component methods added before the 2012 and 2016 elections – prospective models and citizen forecasts. This adaptation involved averaging forecasts within and across six component methods to predict the popular two-party vote in the 2016 U.S. presidential election.
After the 2016 election, the PollyVote underwent crucial revisions to align with evidence-based findings from forecasting research and enhance forecast communication. Notably:
- Naïve forecasts were introduced to better acknowledge forecast uncertainty and adhere to the Golden Rule of Forecasting
- Creating a new method component named expectations, which combines betting markets, expert judgment, and citizen forecasts into a single category, avoiding bias from overweighting expectation-based forecasts.
- Distinguishing model-based forecasts contingent on the information they integrate, specifically whether they depend on retrospective information, prospective information, or a combination of both.
In addition to enhancing the accuracy of the PollyVote forecast, these revisions have streamlined the logic of PollyVote’s structure and reduced complexity, thereby offering users a clearer understanding of the forecasting model and its components.
Publications
Since 2004, the PollyVote has been used to forecast elections in the U.S., Germany and France. This work has been published in leading scientific journals.
U.S.
Five presidential elections,
2004 to 2020
12 publications
Germany
Three federal elections,
2013 to 2021
4 publications
France
Presidential election,
2022
1 publication
Graefe, A. (2023). Embrace the differences: Revisiting the PollyVote method of combining forecasts for US presidential elections (2004 to 2020). International Journal of Forecasting, 39(1), 170-177.
Graefe, A. (2022). Combining forecasts for the 2022 French presidential election: The PollyVote. PS: Political Science & Politics, 55(4), 726-729.
Graefe, A. (2022). Combining Forecasts for the 2021 German Federal Election: The PollyVote. PS: Political Science & Politics, 55(1), 69-72.
Armstrong, J. S., & Graefe, A. (2021). The PollyVote Popular Vote Forecast for the 2020 US Presidential Election. PS: Political Science & Politics, 54(1), 96-98.
Graefe, A. (2019). Accuracy of German federal election forecasts, 2013 & 2017. International Journal of Forecasting, 35(3), 868-877.
Campbell, J. E., Norpoth, H., Abramowitz, A. I., Lewis-Beck, M. S., Tien, C., Erikson, R. S., Wlezien, C., Lockerbie, B., Holbrook, T. M., Jerôme, B., Jerôme-Speziari, V., Graefe, A., Armstrong, J. S., Jones, R. J. J. & Cuzán, A. G. (2017). A Recap of the 2016 Election Forecasts. PS: Political Science & Politics, 50(2), 331-338.
Graefe, A. (2017). The PollyVote’s long-term forecast for the 2017 German federal election. PS: Political Science & Politics, 50(3), 693-696.
Graefe, A., Jones, R. J. J., Armstrong, J. S. & Cuzán, A. G. (2016). The PollyVote forecast for the 2016 American Presidential Election. PS: Political Science & Politics, 49(4), 687-690.
Graefe, A., Armstrong, J. S., Jones, R. J. J. & Cuzán, A. G. (2017). Assessing the 2016 U.S. presidential election popular vote forecasts. In A. Cavari, R. Powell & K. Mayer (Eds.), The 2016 Presidential Election: The Causes and Consequences of a Political Earthquake. Lanham, MD: Lexington Books, pp. 137-158.
Graefe, A. (2015). Accuracy gains of adding vote expectation surveys to a combined forecast of US presidential election outcomes. Research & Politics, 2(1), 1-5.
Graefe, A., Armstrong, J. S., Jones, R. J. Jr. & Cuzán, A. G. (2013). Combined Forecasts of the 2012 Election: The PollyVote, Foresight – The International Journal of Applied Forecasting, Issue 28, 50-51.
Graefe, A., Armstrong, J. S., Cuzán, A. G. & Jones, R. J. Jr. (2014). Accuracy of combined forecasts for the 2012 presidential elections: The PollyVote, PS: Political Science & Politics, 47(2), 427-431.
Graefe, A. (2015). German election forecasting: Comparing and combining methods for 2013. German Politics, 24(2), 195-204.
Graefe, A., Armstrong, J. S., Jones, R. J. Jr. & Cuzán, A. G. (2014). Combining forecasts: An application to elections, International Journal of Forecasting, 30(1), 43-54.
Graefe, A., Jones, R. J. Jr., Armstrong, J. S. & Cuzán, A. G. (2012). The PollyVote’s year-ahead forecast of the 2012 U.S. presidential election, Foresight – The International Journal of Applied Forecasting, Issue 24, 13-14.
Graefe, A., Armstrong, J. S., Jones, R. J. Jr. & Cuzán, A. G. (2009). Combined Forecasts of the 2008 Election: The PollyVote, Foresight – The International Journal of Applied Forecasting, Issue 12, 41-42.
Cuzán, A. G., Armstrong, J. S. & Jones, R. J. Jr. (2005). How we computed the PollyVote. Foresight: The International Journal of Applied Forecasting, Issue 1, 51-52.