Pollsters Keep Putting Voters in Boxes. Voters Keep Breaking Them.

Pollsters have long crafted ways to categorize Americans, but recent political events have shown how quickly voter preferences can change within those categories. In Wisconsin's gubernatorial primary, David Crowley narrowly defeated Francesca Hong by just half a percent, despite polls showing Hong leading by around 20 points. Even earlier in Michigan, polls failed to capture Abdul El-Sayed's significant victory in the Senate primary.

This raises the question of whether polling methods have become less accurate, or if politics has simply exposed the limitations of human categorization.

California lawmakers are considering a bill that would require agencies to offer a separate category for Jewish ancestry or ethnicity, highlighting another attempt to refine demographic classifications. While better data can help measure important issues like discrimination or health disparities, it's crucial to distinguish between visibility in data and precision in predicting individual behavior. Voters' identities and preferences can shift rapidly, making it difficult for pollsters to accurately capture the electorate.

The problem lies in the fact that demographic categories imply a degree of political stability that may not exist. Race, age, and education change slowly, but candidate preferences can shift significantly in a short period. Even Crowley's own pollster underestimated his support in the final days of the election. Pollsters can refine their sampling, turnout models, and weighting of demographic groups to some extent, but they ultimately cannot turn a fluid electorate into a static one.

As political strategist David Axelrod put it, "It's hard to know who's going to show up." This limit cannot be eliminated by any amount of demographic refinement, as the question of whether a person will vote and how they will vote remains the ultimate challenge for pollsters.