The Take-the-best heuristic in election forecasts
Denise and Thomas were sat having their lunch in the staff room. As always, their discussion turned to politics, they talked about how much the politics of United States and Europe have changed in the last year. Like many people Denise and Thomas did not expect the dramatic change in leadership. Denise remarked that there was not way that anyone could have predicted the political shift.
The forecasting of political votes (i.e., elections etc) is important as many businesses and governments make long-term plans, an incorrect prediction can cost a lot of money. Forecasts often predict shifts in political leadership that represent the political views of the majority of voters. The forecaster’s models for predicting the outcome of elections must therefore take a lot of factors into account that include public opinion polls, the candidate’s issues and the candidate’s popularity. Many of the forecasting models are complex but what if there was a quick and easy way to predict the outcome of an election?
The take-the-best heuristic is one of the decision-making short-cuts (heuristics) that can be taken from the cognitive psychology literature and adapted to political forecasting. The take-the-best heuristic enables users of the heuristic to choose between two competing alternatives with ease by comparing the values, or attributes, that they both share (Gigerenzer, 1999). When making a simple decision about which of two products to buy you can compare the products on attributes that are important, for example … how much memory does a laptop have? how many USB ports does it have? The take-the-best heuristic states that the laptop with the best (or largest) amount of memory and number of laptops will be chosen.
The recent U.S. Presidential Election in 2016 saw Donald Trump beat Hilary Clinton to become the President to the surprise of many observers. The forecasting models predicted a close election but ultimately a win for Hilary Clinton. Clinton won the popular vote but not the electoral college vote. In the popular vote Clinton had almost 3 million more votes than Trump with 51.1% to Trump’s 48.9%. Importantly in the U.S. it is the electoral college votes (of which Trump had more) which decide the presidential election, not the popular vote. Historically, there’s very little variance between the popular vote and electoral vote (Graefe et al., 2017)
Before the election of Trump to the Presidential office in 2016 researchers at the Karlsrube Institute of Technology in Germany and the University of Pennsylvania (in the U.S.) began to investigate the accuracy of political forecasting models (Graefe et al., 2012; 2013; 2014). The researchers compared the performance of forecasting models to simple heuristics such as the take-the-best heuristic. They developed a simple model that uses a variation of the take-the-best heuristic called the Big Issue voting model (BI-H model). The BI-H model identifies the issue that is most important for voters (e.g., higher wages, health care insurance) and predicts the candidate that is most likely to win based on the public support for this issue. When tested against other forecasting models the BI-H outperformed the more complex models.
Since the average person has little interest in politics, models that use simple heuristics like the take-the-best heuristic can be used to accurately predict the outcome of elections. There is one caveat though that two-step electoral systems (i.e., popular vote & electoral college votes) can make predicting the outcome of elections even harder. Nonetheless, when data from the popular vote is feed into the BI-H model the heuristics can easily outperform complex models. It clear that a great deal more work needs to be done in political forecasting but heuristics-and-biases can help in the development of forecasting models.