No, Bayes does not like Mayor Pete. (Pitfalls of using implied betting market odds to estimate electability.)

2026-05-27 · Source: Statistical Modeling, Causal Inference, and Social Science · Field: Finance & Economics — Economic Analysis & Policy, Capital Markets & Investment Management · Depth: Intermediate, short

Summary

A 2019 analysis critically examines economist Greg Mankiw's method of using Predictit betting market odds and Bayes' Theorem to estimate 2020 US presidential candidate electability. Mankiw's initial calculation, based on April 27, 2019 data, indicated Mayor Pete Buttigieg had an 0.80 probability of winning the general election if nominated, with Joe Biden at 0.77. The critique highlights several issues with interpreting these odds as true probabilities. These include potential market bias, exemplified by Andrew Yang's 9% nomination contract price despite 0.8% polling. Significant volatility was also noted, with Biden's calculated electability shifting from 0.77 to 0.68 and Warren's from 0.44 to 0.68 between April 27 and June 12. Additionally, the sum of all candidate prices on Predictit exceeded 100%, suggesting market prices are not direct probabilities due to factors like the "vig." A re-evaluation using June 2019 data found Buttigieg's electability at 0.56, making him the least electable among the three, contradicting Mankiw's original finding.

Key takeaway

For data scientists or research scientists analyzing political prediction markets, you must critically evaluate raw odds. Do not directly interpret market prices as true probabilities without accounting for inherent biases, significant volatility, and market mechanics like the "vig" and rounding. Your models should incorporate adjustments for these factors, especially with early or thin market data, to avoid drawing misleading conclusions about candidate electability or other outcomes.

Key insights

Prediction market odds require careful scrutiny before direct interpretation as true probabilities.

Principles

• Betting market prices are not direct probabilities.
• Market thinness introduces significant noise.
• Early prediction market data is highly volatile.

In practice

• Scrutinize prediction market data for bias.
• Account for market volatility in analysis.

Topics

• Prediction Markets
• Bayesian Inference
• Election Forecasting
• Market Bias
• Data Volatility
• Predictit

Best for: Data Scientist, Research Scientist

Related on AIssential

• Introduction to quantitative finance Part 26: Whether to use forecasting methods, or to tell an LLM… — Forecasting in quantitative finance faces challenges with LLMs, necessitating memory-enhanced reward shaping for consistency.
• The Types Of Candidate You Find In The California Gubernatorial Race — The California gubernatorial race attracts a wide array of unconventional candidates, often driven by personal agendas, self-promotion, or unique beliefs.
• Cluely CEO Roy Lee admits to publicly lying about revenue numbers last year — Misrepresenting financial metrics and interview origins can damage credibility, even for viral startups.
• Sethi on “Betting on Geopolitical Violence” — Crypto-based prediction markets enable anonymous speculation on geopolitical violence, attracting insider trading and potential manipulation.
• Use AI This Election — AI can effectively personalize complex political research, significantly improving voter decision-making.
• Signal Or Noise? A Decision Tree For Evaluating Unusual Trading Activity — A structured five-step process enhances consistent and informed responses to unusual trading activity.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Statistical Modeling, Causal Inference, and Social Science.