NBER Study: Prediction Markets Show No Systematic Partisan Bias
Understand this event
On October 9, a16z cited NBER Working Paper No. 35846 in “This Week in Charts,” using a Kalshi market chart for Brazil’s presidential election to illustrate how prediction markets reflected the outcome about three weeks in advance. On October 12, PPC Land reported that the paper examined prediction markets dating back to 1880 and found no evidence that they systematically overestimated left-wing or right-wing candidates. Across all markets, the partisan coefficient was +0.03 percentage points, with a standard error of 2.66; the confidence intervals for all four political regressions included zero. The study shows only that prediction errors did not consistently favor one party; it does not prove that markets make accurate predictions.
AI 根据报道生成 · 3 days ago更新
Impact and actions
AI compilation of public coverage, separating facts, inference and scenarios. Verify rules against the original sources.
Prediction market research found no significant partisan bias, which does not mean predictions are accurate
First update · Latest update (Beijing time)
Impact level
⚪ Background
⚠ Risk:The absence of significant partisan bias does not mean there is no pricing bias, and the findings cannot be generalized to sports contracts.
What happened
On October 11, PPC Land reported that an NBER working paper published in October 2026 analyzed returns from U.S. and non-U.S. election prediction markets from 1880 to 2025. The overall partisan coefficient was +0.03 percentage points, and the 95% confidence intervals for all four groups of political regressions included zero, indicating no statistically significant partisan bias was detected; this does not prove that predictions are accurate. The total sample size and regional composition were not provided. [1]
Seller impact
- InferenceThe research summary also reports that across most political markets, underdogs are overestimated and favorites are underestimated. Therefore, when using market probabilities in business decisions, the absence of partisan bias is not enough to show that the probabilities are reliable; the available information is also insufficient to compare the platforms’ current predictive performance. [1]
Suggested action
Before using election prediction probabilities in business decisions, check the original paper’s sample and the scope of the contracts it covers, then decide whether to treat them as supplementary information.
Evidence that would disprove the assessment
- Assessment 1：Recheck the original paper over the next two weeks. The assessment should be revised if a formal revision rejects the pricing bias described above or provides an accuracy comparison across platforms using consistent metrics.
Coverage timeline
Follow coverage to understand different perspectives.
- PPC Land advertising and platform updatesOriginal ↗NBER Paper: Prediction Markets Have Shown No Partisan Bias Since 1880
NBER Working Paper No. 35846 found that prediction markets did not systematically overestimate left-wing or right-wing candidates. Across all markets, the partisan coefficient was +0.03 percentage points, with a standard error of 2.66; the confidence intervals for all four political regressions included zero. On October 9, a16z cited the study in “This Week in Charts,” using its Kalshi market chart for Brazil’s presidential election to illustrate how prediction markets reflected the outcome about three weeks in advance. The study shows only that errors did not consistently favor one party; it does not prove that markets make accurate predictions.
Interest trend for this event
Current interest 3·Comparable scope peak 10（10/12 02:00）·Change over 24 hours within comparable scope -49%
Trends compare the same continuously observed entities, which may be fewer than the current interest count. Move the pointer or select the chart to see hourly values; use arrow keys to navigate.