Why do people vote? What factors contribute to people’s eventual voting choice, and which factors that seem like they should be relevant are actually meaningless in the aggregate? Political scientists have long investigated questions of how people end up at their eventual voting choice, and the relative strength of the decisive factors are still debated in the field today.
Two foundational studies to voting choice in political science are the Michigan and Columbia Models. Named after the universities of their major developers, the two models vary in the weight they place on various factors that can explain electoral choice.
The Michigan Model was developed from data on the 1952 presidential election, and was presented in the seminal book The American Voter, written by Angus Campbell, Philip Converse, and colleagues. When it comes to voting for presidential candidates, party attachment is by far the most significant factor that affects electoral choice. Party attachment is a very stable characteristic of an individual over time, and is determined by social influences such as parents, friends, or colleagues.
The authors of The American Voter also established a causal model among factors that result in a person’s voting choice. They call this model the “funnel of causality,” where the funnel narrows until it reaches someone’s vote choice.
The broad end of the funnel includes broad characteristics that form the baseline factors that determined someone’s vote. These include sociological elements such as race and gender, social status characteristics such as what one’s occupation is, and the characteristics of one’s parents (specifically their partisanship). These three factors then determine party identification (Republican or Democrat, since this model was developed in US settings with the assumption of 2 parties).
Further down in the funnel, party identification is influenced in any given election by perceptions of the candidates involved and perceptions of the relevant issues in an election. The result of these interactions is the end of the funnel: the vote choice of the individual.
This model is useful in that it covers a lot of ground with respect to potential factors, and is also empirically testable through various equations. However, some criticisms of the model are that it only functions in one direction. For example, issue perceptions could not influence party identification. The role of political information in determining voting choice is very ambiguous in the model. Some scholars also question whether party identification is given too great an influence in the model.
The other major model for determining vote choice is the Columbia Model, which was developed in the 40s using panel data collected at 7 times from a group of 600 people in a Midwestern county. The authors of the study hypothesized that candidates were marketed like products to rational voters. However, this idea was not supported. Partisanship is a force that will keep people from being affected by short-term factors of any given campaign that might sway a neutral voter.
In their analyses, they found that a select few factors like socio-economic status, religion, and whether someone lived in an urban/rural area were highly effective at predicting the vote. They found that campaigns did not function how they expected – efforts by candidates to attract votes didn’t just transmit information equally to all voters. Instead, campaigns were most influential on those who are cross-pressured in some way. For instance, a cross-pressured voter could live in a Republican area but go to work at a largely Democratic workplace.
They found that most of the people in their sample were apathetic to politics, and that media campaigns didn’t really influence most people’s voting choice. When information does matter, it is because a highly attentive part of the polity takes in information through media and then transmits it to others in their social circle.
Some of the criticisms of the model are that the data was collected in 1940 and 1948, both atypical times in US history. One year was during the buildup to WWII, and the other was in the relieved aftermath of that war. These forces could have changed people’s interest in politics from some baseline. The data was also collected from a very limited geographic area and the authors did not make use of psychological concepts in their analysis like the Michigan Model does.
The Michigan Model also stands in contrast to political scientists who think that campaigns really do matter in influencing votes. Both the Columbia and Michigan models seem to dismiss the effects of campaigns on determining a person’s electoral choice. Future political scientists would also argue in the 60s that people vote retrospectively, meaning that they take a look at how a party has performed since it was elected and then vote accordingly depending on whether that performance was good or bad.
Many voting models have elements of truth to them. The Columbia and Michigan models had a point in downplaying the effects of campaigns – many people fall into echo chambers where the media they are exposed to only enforces their opinion. However, neither model fully appreciates how identity and the need to belong impacts a person’s partisanship, and how that partisanship in turn colours all a person’s political perceptions. A master theory to explain all voting behaviour would be very complex indeed.