Natalie Jackson: “Polls and election ‘basics’ like those from FiveThirtyEight, The Economist, and Nate Silver are more exciting from an empirical perspective. The statistical manipulation is truly challenging and interesting. You can combine national and state polls, economic factors, incumbent factors, and voting history, and then spin all of that into state-by-state predictions that can be used to simulate presidential elections. We’re talking thousands of lines of code.”
“Sounds pretty empirical, right? In the sense that you collect a ton of data and build a huge model that outputs a ton of data. But every decision about what goes into the model is subjective. The modeler decides which polls to use, whether to adjust for the quality of the pollster or their past accuracy, how much individual polls can shift trends, which economic indicators to use, which political factors are important, and how all of this is coded. If you make a different decision at any stage, the model’s predictions change…”
“We can calculate probabilities all day long, but we have no idea how accurate they are. Polls have margins of error and additional margins of error. Models have their own margins of error. The judgments made by the people who build the models have margins of error. We don’t know how big that cumulative error is.”
Save to favorites
