Midterm market accuracy improved late after trailing poll-based forecasts

Prediction markets showed no consistent accuracy advantage over election forecasts for most of the midterm race, according to Outside the Beltway’s account of a New York Times analysis. Their stronger performance emerged in the final days rather than across the broader campaign period.

A late shift in performance

Over the three months before the midterms, market probabilities expressed less confidence and were frequently less accurate than forecasts based on polling, according to the analysis described by Outside the Beltway.

That comparison changed near the election. Markets performed better in the final days as polling and forecasts became stale. The findings therefore distinguish between markets’ performance over the preceding months and their relative advantage at the end.

Shared information, different timing

Forecaster Nate Silver attributed the markets’ accuracy directly to the work of pollsters and forecasters, according to Outside the Beltway. He said bettors relied mostly on information those researchers produced.

That assessment complicates a simple contest between markets and forecasting models: the market prices being evaluated drew heavily on the same information ecosystem as the forecasts they were measured against.

What it means for prediction markets

• The comparison window matters: Markets’ better performance in the final days did not extend across the three months before the midterms.
• Freshness matters alongside accuracy: The late improvement came as polling and forecasts became stale, making timing central to interpreting the results.
• Market odds were not an independent information source: Silver’s assessment points to pollsters and forecasters as key inputs into bettors’ judgments.