In the lead-up to the 2026 primary season, prediction markets such as Kalshi and Polymarket gained significant attention for their political forecasts, though their accuracy has been subject to scrutiny. Notably, a few days before Los Angeles’s mayoral primary, these platforms assigned a roughly 75 percent chance that reality TV star Spencer Pratt would advance to the general election. However, Pratt ultimately finished third, trailing behind Nithya Raman. This unexpected result prompted some supporters to allege voter fraud, despite polling data that had suggested a closer race.

Pratt's outcome was not isolated. Similar instances occurred in key races such as the Georgia gubernatorial primary, where a Trump-endorsed candidate was defeated by a healthcare executive, and in Kentucky, where incumbent Rep. Thomas Massie lost to a Trump-backed challenger. Despite these high-profile misses, experts emphasize that such results do not constitute outright failures of prediction markets but rather reflect their function as probabilistic tools.

A recent analysis examined 268 candidates who, in the 60 days prior to their primary, were given a 70 to 80 percent chance of winning on Kalshi or Polymarket. The findings revealed that in most cases, candidates favored by the markets did secure victory, with about a 3-to-1 ratio of correct predictions to misses. The broader data set, covering over 1,200 markets for House, Senate, and gubernatorial primaries across 2026, demonstrated a general alignment between market probabilities and actual outcomes, supporting the reliability of these platforms as forecasting instruments.

Despite their growing popularity—especially as nearly one-third of Polymarket bets since mid-2024 have been political—prediction markets have not replaced traditional polling or journalistic coverage. Experts caution that these markets reflect “quantifications of conventional wisdom” rather than independent predictive power. Polling data and news reports supply the foundational information on which bettors base their trades, meaning that prediction markets aggregate and synthesize multiple inputs rather than generate original forecasts.

Interpreting probabilities in political prediction markets can be challenging. A 75 percent chance of winning, for example, indicates a three-out-of-four likelihood, not a guarantee or a direct measure of vote share, a distinction that can be misunderstood by observers. This probabilistic nature accounts for outcomes where the favored candidate falls short, which is consistent with statistical expectation rather than failure.

Political analysts note that prediction markets do not explain the motivations or dynamics behind voter behavior, instead focusing solely on likely outcomes. In that respect, polls remain essential for gauging voter sentiment and understanding election drivers, albeit complemented by the real-time, market-based information these platforms provide.

While prediction markets have historically delivered useful forecasts—even predating modern polling over a century ago—their precision varies, and experts suggest caution in interpreting exact probabilities. Both Kalshi and Polymarket maintain that their markets are well-calibrated and valuable forecasting tools, having accurately predicted major 2024 political events and ongoing election outcomes.

Overall, prediction markets serve as an increasingly important component of political forecasting by consolidating diverse sources of information and trader strategies. However, they are best viewed as one element within a broader analytic ecosystem that includes traditional polling and journalistic evaluation.