Predicting the Presidency

Episode Summary

Executive Summary: The episode explains how Nate Silver uses Bayesian updating to predict elections, showing how his forecasts change as new polling data arrives rather than replacing prior beliefs. It also explores a listener question about the longevity of US presidential candidates’ mothers versus British political leaders’ mothers, finding the US group lives significantly longer, though some of the difference reflects broader cohort life expectancy patterns.

Main Topics: Nate Silver and election forecasting (Priority: 5/5): The segment profiles Nate Silver and his use of statistical prediction, especially in the 2008 and current US presidential races, where his forecasts drew attention for their accuracy and frequent updates. Bayes’ theorem as the basis for updating predictions (Priority: 5/5): The show explains Thomas Bayes’ key idea: start with prior beliefs and revise them systematically when new evidence arrives, which is central to Silver’s method. Polling, uncertainty, and the value of probabilities (Priority: 4/5): Silver discusses how he processes polls, emphasizes state-by-state modeling, and frames election outcomes as odds rather than certainties, similar to betting markets. Comparison of prediction models and bookmakers (Priority: 3/5): The episode compares FiveThirtyEight’s odds with betting markets, suggesting that bookmakers often provide a useful benchmark for political probabilities. Are politicians’ mothers unusually long-lived? (Priority: 4/5): A listener question leads to statistical analysis of the ages at death of mothers of post-war US presidential candidates and British political leaders, revealing a real difference in averages. Cohort effects and life expectancy (Priority: 4/5): The discussion notes that women born in later decades had higher life expectancy in the UK than in the US, and that both groups’ mothers outperform their general cohorts.

Key Arguments: Nate Silver’s forecasting method is Bayesian: he begins with a prior assessment and updates it when new polling data arrives. Individual polls should not be treated as decisive events; they should be incorporated systematically and sceptically into a broader model. Election prediction is best expressed as probabilities and odds, not absolute certainty. State-level modeling matters because the US presidential election is decided state by state, not by national popular vote alone. Betting markets can be a useful reality check because they convert vague notions like “too close to call” into precise odds. The apparent longevity advantage for mothers of US political leaders is statistically real, though it partly reflects broader historical cohort differences. Both US and UK political leaders’ mothers tend to live longer than comparable women in the general population.

Data Points: Nate Silver’s 2008 state predictions: 49 out of 50 states correct - His forecast of the 2008 US presidential election Nate Silver’s 2008 Senate predictions: 100% correct - He got every US Senate race right in 2008 Obama win probability on 538: 86% - Silver’s blog forecast after the first presidential debate, dated Thursday 4 October Romney odds: about 5 to 1 underdog - Current betting-market comparison mentioned in the interview US political mothers sample: 24 presidential candidates since the war - Listener’s question about mothers of Republican or Democrat candidates Very long-lived US mothers: 3 mothers lived to 100 - Among the 24 post-war US presidential candidates’ mothers US mothers reaching 90: 6 more mothers - Among the same US sample US mothers aged 80+: 5 more mothers - Among the same US sample Comparison sample size: about 40 people total, about 20 in each group - The Oxford gerontologist notes the statistical comparison is based on small numbers Cohort comparison: Every decade from the 1840s to the 1940s - For each birth decade, UK women’s life expectancy exceeded comparable US women’s life expectancy Age distribution pattern: Normal distribution - Expected and observed shape of the age-at-death data for both groups

Pivotal Quotes: "What Bayes' theorem is... it's just a very simple mathematical formula that tells you what do you do when you encounter new information." — Narrator: Introduction to Bayesian updating and why it matters for prediction "You use all the information but tend to use it more sceptically." — Nate Silver: Explaining how he incorporates polling data into his model "The US curve has moved further to the right... we've got more people in the distribution who are reaching 100 in their 90s in the US." — Dr George Leeson: Interpreting the longevity comparison between political mothers

Implications: Listeners get a clear example of how probabilistic forecasting works in practice and why updating models matters. The longevity segment shows how small-sample patterns can still be meaningful when checked against broader demographic context.