In his inaugural lecture, Professor Benjamin Lauderdale seeks to answer the following question: how can we predict election results?
In presenting his response, Lauderdale explains that the ability to forecast elections is constrained by several factors including the complexity of the electoral system, our everchanging interactions with the system, and our ability to measure the current state of the system. One of the most restrictive, yet unchangeable, constraints is the infrequency of elections. Lauderdale argues that the fact that the significant gaps of time in between elections makes it difficult to predict how people will vote considering that political stage is always evolving. To demonstrate this difficulty, Lauderdale compares the task of predicting elections to predicting the weather.
Meteorologists, he points out, can retrieve data about weather patterns constantly – allowing them to draw accurate conclusions about future patterns. Political scientists, however, do not have this luxury. The U.K. general election, for example, only happens (usually) every five years. This means that in predicting the outcome of the next general election, data from the previous one will be outdated, making the task of predicting elections trickier due to limited data.
This combined with the vast array of actors that influence political campaigns from parties to voters to candidates to lobbyists, leave us in a position in which we are “not great at measuring what actors want, think, or intend to do.” Thus, Lauderdale argues that long term predictions are bound to be imprecise, and if they aren’t, they are overconfident. Short term predictions, however, offer more plausible predictions.
Lauderdale explains that short term predictions require an understanding of 1) models of likely short-term changes and 2) how to measure the current state of the system. Drawing on his experience at YouGov – a British data analytics firm – Lauderdale outlines three areas he studies to make short-term election predictions.
1) Who can vote?
This area refers to population data, including census data, population surveys, and electoral registration. It is important to understand how people voted in the past using past election results and surveys.
2) Who will vote?
Information about past election turnout is helpful as most people who vote do so all the time. This section presents challenges, however, as people often lie about whether they voted in surveys.
3) Who will vote for whom?
This part is all about choice. Although some people do change their minds last minute, people generally tend to voter for the same parties they have in the past.
Despite the difficult situation Lauderdale is faced with in predicting elections, he has so far been quite successful. In 2017, while the world was expecting a conservative majority in the general election, Lauderdale and YouGov predicted a hung parliament. Despite his predictions being unpopular at the time, he was right – demonstrating that while accurately predicting election results is no easy feat, it is possible.
In closing his lecture, Lauderdale leaves the audience with an impactful statement: when governments and voters start to look at you, you must be aware. When Lauderdale predicted the hung parliament in 2017 the pound dropped significantly. Similarly, when he predicted the conservative majority in 2019, the point jumped. Thus, Lauderdale’s job of predicting elections is not only difficult, but incredibly impactful on society.
Professor Lauderdale brings an admirable combination of passion and pragmatism to his work in the field of political science. Many thanks to him for this insightful lecture.