How does election forecasting work and how will it account for societal changes like the rise of social media? Ross Burkhart, professor in the School of Public Service, shared expertise on the current processes and potential changes in election forecasting as our world changes.
How forecasting works
There are many different theories that election forecasters run their predictions through, but for Burkhart’s research, he uses one commonly known as the “reward and punishment” theory, centering around the idea that voters are like “gods” who reward performance and punish non-performance. After choosing a theory, forecasters then test the theory with a variety of independent variables, such as the state of the economy or public opinion polling. Through regression analysis, forecasters can find out how these independent variables affect the dependent variable being tested (typically an incumbent candidate or political party).
Forecasting the 2024 U.S. election
To conduct a forecast for this year’s presidential election, Burkhart selected two different moments in time – one in July just before current President Joe Biden stepped out of the race, and one in early August towards the beginning of Kamala Harris’ campaign for presidency. He used Biden’s favorability ratings from both of these dates – 36% in July and 43% in August – and found that the total number of electoral votes were 262 and 306, respectively. With a total of 270 electoral votes needed to win, Burkhart’s forecast predicts former President Donald Trump winning as of July and the win of Vice President Kamala Harris as of August. Burkhart shared that there are some complications with these results given the quick turnaround of the Democratic party’s candidates for president. “Some of this modeling hinges upon whose approval are we really focusing on…the Biden presidency or the Harris nomination,” he said. “The race is relatively close, and then there’s also some room for interpretation here. It suggests that this election is going to be heavily contested.”
Challenges
There are a few complications that can present barriers to election forecasting. Firstly, the sample size for forecasting is fairly small. Burkhart is only able to use the last 19 presidential elections, and much of the forecasting model relies on prior election results for prediction. Forecasting also largely relies on the idea that voting behavior wouldn’t change much between elections. Another complicating factor is that in order for this type of structural forecasting model to be accurate, it is conducted at a single point in time, usually about 2-3 months before election day, which makes it difficult to account for influencing factors that may happen in between the forecast and the election. There are other models of forecasting where the predictions change as polls change, but Burkhart finds the structural forecasting model to be more accurate over time.
The future of forecasting
It’ll be interesting to see how election forecasting changes as we go forward, Burkhart concluded. With the rise of social media and digital methods for political campaigning, it’s hard to tell just yet how forecasts will be affected. “We need to build up the data…for this type of forecasting to work, you need variables that are consistently measured across time,” Burkhart shared. “Social media presence really did not start until 2008, but it already has been a major catalyst of political movements. I think it’s possible that if we gather more information about public opinion through social media in future election cycles, we might be able to use it in these models. We need to be sure that we’re using variables that can consistently be measured to be reliable.”
Learn more about Ross Burkhart