Electoral Predictions on Polymarket: A Quantitative Description of Trading, Commenting, and Reacting

Abstract

 Election betting platforms, which combine real money trading with public political discussion and social feedback, have seen explosive growth worldwide. Prediction market research emphasizes aggregate accuracy and price dynamics, while computational political communication research examines discourse in digital forums; however, these literatures remain analytically separate. Here, we provide a quantitative description of how financial, discursive, and social behaviors are jointly observable on a single platform. We analyze 778,634 accounts (unique wallet addresses) over 846 days, documenting 30,502,864 trades totaling $5.86 billion, 384,586 comments, and 571,523 reactions across three behavioral modalities (trading, commenting, reacting) on the Polymarket platform, spanning 959 election-related events. We find that within events, trading and commenting are largely decoupled and temporally ordered: among wallets exhibiting both behaviors, trading and commenting co-occur in only 21.8% of wallet-event combinations, are only moderately correlated within events, and 90.2% of joint pairs are trade-first. Participation is highly concentrated across all behaviors, with concentration varying systematically by behavior type: financial trading (Gini = 0.951; top 1% account for 73.7% of volume) concentrates most sharply, followed by social reactions (Gini = 0.845), then political discourse (Gini = 0.814; top 10% produce 76.4% of comments). This cross-behavior ordering persists when the analysis is restricted to higher minimum-activity thresholds. We also document the volume of single-action wallets: 12.6% of trading accounts execute exactly one trade, 41.1% of commenters post exactly one comment, and 34.3% of reactors leave exactly one reaction. While single-mode accounts (engaging in only one modality) dominate numerically (95.9% of accounts, predominantly Trader Only at 94.4%), a small multi-modal minority (engaging in two or more modalities; 4.1%) contributes disproportionately across financial, discursive, and social behaviors. These findings characterize observable wallet-level political activity when financial commitment is visible alongside discursive acts, informing comparative and explanatory research on hybrid platforms. More broadly, quantifying who participates, how activity concentrates, and when engagement occurs provides a foundation for studying participation across behavioral domains that have typically been analyzed separately.