Potential for Improved 2020 Election Forecasts with Quantum Computers

QxBranch publishes the world’s first US Presidential election forecast generated by a quantum computer in special issue of the Journal of the Physical Society of Japan. QxBranch Senior Data Scientist, Max Henderson, and Systems Engineer, Tristan Cook, along with former QxBranch Senior Data Scientist, John Novak, were published on March 1. The paper, “Leveraging Quantum Annealing for Election Forecasting,” demonstrates the potential of using quantum computers to generate election forecasts during the upcoming 2020 US presidential election.

After observing many high-profile election forecasts assign the highest probability of success to the Clinton campaign in the 2016 US presidential election, the QxBranch team identified a major issue present in most models – the absence of correlations between states.  That is, the likelihood that the result in one state would also be replicated in another state. For example, if Pennsylvania is forecasted to vote in a certain probability, what is the likelihood Ohio will also vote the same way? The reason for this omission in the modelling is understandable: simulating a highly correlated system using classical computers is computationally challenging as the system size increases. However, this presents an ideal opportunity to leverage quantum computing.

“The advantage of mapping the 2016 election onto a quantum computer is that the quantum computer is itself an intrinsically correlated system. By modelling forecast outcomes for each state and the correlations between those states on a quantum computer, our results ended up being a closer match to the to the final results than many other models,” explains Henderson.

By testing historical election and polling data on quantum training models, the QxBranch team as able to generate 2016 election forecasts similar to other professional forecasters, but with final outcomes that gave the Trump campaign a higher likelihood of victory.  Adding to the performance of the quantum model was more efficient sampling and the inclusion of additional error models and intrinsic, tunable state correlations.

Quantum computers are opening up a range of opportunities to provide alternative, novel solutions to computationally difficult problems, and forecasting elections is just one way in which quantum computing could help shape our decision making.

Click here to view and download the full article.

Subscribe to our newsletter

Recent Posts