Prediction Markets Are Now Financial Infrastructure
How a 19th-century mechanism got rebuilt with 21st-century plumbing — and why the volume charts matter less than the feedback loops.
Sixteen months ago, prediction markets were a curiosity that happened to call the 2024 election better than the pollsters. Today, Kalshi processes over $100 billion in annualized volume, Robinhood users trade event contracts alongside stocks, and the Federal Reserve publishes research validating prediction market prices as macroeconomic forecasting tools.
None of this is new. The mechanism is 150 years old. What's new is the infrastructure — and the feedback loops that infrastructure enables.
The Scale Shift
The numbers tell one story. The system dynamics tell a better one.
In 2024, prediction markets were a proof of concept. Polymarket called the election. Kalshi fought regulatory battles. Total industry volume measured in single-digit billions.
By December 2025, Kalshi announced a $1 billion Series E at an $11 billion valuation — an 8x volume increase from July alone. CNN and CNBC became broadcast partners. Robinhood integrated Kalshi contracts directly into its trading interface, putting event markets in front of millions of retail investors who'd never heard of a prediction market six months earlier.
February 2026 set records: Kalshi alone processed $9.8 billion in monthly volume. Industry-wide, the two major platforms combined for $17.9 billion in February notional volume.
But volume is an output, not a cause. The interesting question is what changed in the system to produce this output. Three feedback loops converged:
These loops explain why the growth curve is nonlinear. The system passed a threshold where the feedback became self-reinforcing.
The Long Road Here
The mechanism itself is old. Only the implementation changed.
The pattern is worth noting: prediction markets didn't fail in the 20th century because the mechanism was wrong. They failed because the infrastructure was suppressed — first by Progressive reformers, then by post-9/11 politics. What looks like innovation is actually restoration. The mechanism is 150 years old. The pipes are new.
The Design Constraint: Markets That Create What They Measure
Not every question should have a prediction market. Some markets would create the outcomes they're supposed to forecast. This is where the DARPA critics had a point.
The assassination market is the canonical example. A liquid market on "Will [leader] be killed before December?" creates a bounty for killing that leader. The more liquid the market, the larger the potential payout, the stronger the incentive.
The same logic extends through related categories:
Terrorism. A market on "Will there be a terrorist attack in [city] before [date]?" could provide operational funding for the attack itself.
Corporate sabotage. A market on "Will [company] suffer a data breach?" creates an incentive to cause the breach.
Self-fulfilling manipulation. Even non-violent markets face reflexivity problems. A market on "Will [stock] fall below $X?" could be manipulated by a trader with enough capital to cause the price drop.
The design principle that emerges: prediction markets work when they reveal information about events, not when they create incentives to cause events. The boundary isn't always obvious, but it's the core constraint separating useful infrastructure from dangerous instruments.
What You Can Trade Now
Kalshi currently lists over 770,000 active markets. The category expansion reveals the system's growth pattern:
The breadth matters. Prediction markets are no longer "election betting." They're becoming a parallel pricing layer for uncertainty across domains — anywhere the question is tractable and the manipulation risk is containable.
The Fed Validation
In January 2026, the Federal Reserve Board published a working paper titled "Kalshi and the Rise of Macro Markets." The findings formalized what the volume charts already suggested.
For fed funds rate forecasts 150 days out (about three FOMC meetings), Kalshi's mean absolute error matched professional forecaster surveys. But unlike surveys — which provide a snapshot every six weeks — Kalshi offers a continuously updating full distribution.
On the day before FOMC meetings, Kalshi's median and mode forecasts achieved perfect accuracy across the sample period. Every Fed decision was correctly priced the day before it happened.
The paper's conclusion: prediction markets provide "a high-frequency, continuously updated, distributionally rich benchmark that is valuable to both researchers and policymakers."
When the Fed validates a forecasting mechanism, that validation attracts more sophisticated participants. More sophisticated participants improve price discovery. Better price discovery produces more accurate forecasts. The loop reinforces itself.
The Favorite-Longshot Bias
No system is perfectly calibrated. A February 2026 CEPR analysis of over 300,000 Kalshi contracts found a persistent favorite-longshot bias: cheap contracts (low probability events) win less often than their price implies, while expensive contracts (high probability events) win slightly more often.
A 10-cent contract wins less than 10% of the time. This pattern — well-documented in sports betting for over a century — has now been confirmed in prediction markets across politics, entertainment, and economic data.
The mechanism behind the bias isn't fully understood, but several hypotheses exist:
Risk-seeking on longshots. Traders may overpay for low-probability events because the potential payout is exciting. The lottery effect.
Overconfidence on favorites. Traders may underpay for high-probability events because "95% likely" feels like certainty, leading to underpricing of the remaining risk.
What this means for users: don't treat market prices as perfectly calibrated probabilities. A 15% contract might represent a 12% actual probability; a 75% contract might represent 78%. The bias is small but consistent.
The Deeper Shift
Something structural changed. The volume charts are a symptom; the underlying mechanism shift is the cause.
For decades, forecasting operated as a serial process: expert analyzes data → expert forms opinion → expert publishes forecast → audience consumes passively. Information flowed in one direction. The feedback loop was weak and delayed.
Prediction markets are a parallel distributed system. Information enters from thousands of nodes simultaneously. Each node is weighted not by credentials but by stake size — essentially, confidence-weighted voting. Errors get arbitraged out in real-time. The price is the aggregation function, continuously updated.
This is Hayek's 1945 insight about prices, applied to forecasting: markets aggregate dispersed information that no central authority could collect. A derivatives professional notices something in positioning data. A campaign operative bets on internal polling. A journalist in a swing state senses enthusiasm gaps. None of them publishes; each of them trades. The price synthesizes signals that would otherwise never combine.
The implication for investors: a real-time probability dashboard for policy and economic uncertainty now exists. Fed expectations are visible as explicit probability distributions. CPI surprise risk is priced before the release. This information used to cost a quarter-million a year. Now it's a free tab in any browser.
The Civil War: Kalshi vs. Polymarket
The industry has consolidated around two platforms, and their competition is producing rapid iteration.
Polymarket — Crypto-native, global reach, minimal KYC friction, dominant in geopolitics and crypto contracts. The architecture optimizes for accessibility and international participation.
Kalshi — CFTC-regulated, US-focused, Robinhood-integrated, dominant in sports, domestic politics, and economic data. The architecture optimizes for regulatory legitimacy and retail distribution.
Both architectures have advantages. Polymarket's permissionless access attracts international traders with information about global events. Kalshi's regulatory approval enables integration with mainstream finance. The competition is producing better products — faster settlement, tighter spreads, broader contract coverage.
A wildcard: Coinbase is rumored to launch a native prediction market in late Q1 2026. If a third major platform enters with Coinbase's distribution and liquidity, the current duopoly restructures.
The Mechanism, Restored
The 19th-century bookmakers on the Curb Exchange would recognize what Kalshi built. The prices look different — digital instead of chalkboard — but the mechanism is the same: distributed bettors, weighted by conviction, aggregating information that no single expert could possess.
What's different is the infrastructure that allows the mechanism to scale: instant settlement, regulatory legitimacy, mobile access, integration with mainstream finance. The plumbing is new. The physics is old.
The price is the probability. It always was. The only question was whether the infrastructure existed to surface it at scale.
Now it does.
Capital Writ — An Equicurious Commentary Desk
Capital Writ covers the macro forces, monetary policy decisions, and structural plumbing that shape how capital moves through the system. Prediction markets are the newest piece of that plumbing — a 150-year-old mechanism finally connected to 21st-century pipes.