Correlation Arbitrage

Identifying mispricings between related markets that must obey logical constraints

What is correlation arbitrage?

Correlation arbitrage exploits pricing inconsistencies between related markets. Unlike pure arbitrage (which is risk-free by definition), correlation arbitrage involves positions in markets that are logically or statistically related, profiting when the relationship between their prices is mispriced.

In traditional finance, this includes pairs trading (long one stock, short a correlated peer), index arbitrage (the index vs. its components), and convertible bond arbitrage. In prediction markets, the structure of related contracts creates analogous opportunities.

Logical constraints on Kalshi

Many Kalshi markets have built-in mathematical relationships that must hold. When prices violate these relationships, there is a structural mispricing:

Bracket monotonicity

Consider weather temperature brackets for the same city and date:

• "High above 60°F" priced at 92 cents
• "High above 65°F" priced at 78 cents
• "High above 70°F" priced at 55 cents

These must be monotonically decreasing — the probability of exceeding a higher threshold cannot be greater than exceeding a lower one. If "above 70" were priced at 82 cents while "above 65" was at 78 cents, that would be a logical impossibility. Any realization of "above 70" necessarily means "above 65" also occurred.

Complementary markets

Some Kalshi markets are logical complements. If "Fed cuts rates in June" is at 65 cents and "Fed holds rates in June" is at 40 cents, the implied sum is 105 cents — which exceeds 100%. If these are the only two outcomes (ignoring rate hikes), this overpricing means at least one side is too expensive.

In practice, there may be more than two outcomes (cut, hold, hike), so the analysis requires enumerating all mutually exclusive possibilities.

Temporal consistency

Markets on the same outcome at different time horizons must be logically consistent. If "Bitcoin above $80K by June 30" is at 50 cents, then "Bitcoin above $80K by December 31" must be at least 50 cents (since June is a subset of the full-year window). Violations of temporal monotonicity create arbitrage conditions.

Statistical correlation

Beyond logical constraints, some markets are statistically correlated without being deterministically linked:

• Related economic indicators — CPI and PCE inflation measures are highly correlated. If the market for CPI reprices sharply on new data but the PCE market has not adjusted, the PCE market may be temporarily mispriced.
• Regional weather — Temperature in New York and Boston on the same day are highly correlated. If the NYC "above 75" contract reprices due to a forecast update, the corresponding Boston contract may lag.
• Cross-asset — Bitcoin and Ethereum prices are correlated. A sharp move in Bitcoin brackets may not immediately reflect in Ethereum brackets.

Statistical correlation arbitrage is riskier than logical arbitrage because correlations can break down, especially during unusual market conditions.

Execution challenges

• Simultaneous execution — True arbitrage requires both legs to execute simultaneously. On Kalshi, there is no mechanism for atomic multi-contract orders. One leg may fill while the other does not, leaving the trader with unintended directional exposure.
• Fee drag — Arbitrage profits from pricing inconsistencies are often small (a few cents per contract). After exchange fees on both legs, the net profit may be zero or negative.
• Liquidity asymmetry — The "cheap" leg of an arbitrage may be in a liquid market while the "expensive" leg is illiquid, making it hard to fill the complete position.
• Speed — Pricing inconsistencies are often corrected quickly by other participants. The window to execute may be seconds or minutes.

Historical context

Arbitrage in prediction markets has been studied since the early days of online platforms. Wolfers and Zitzewitz (2004) documented pricing discrepancies between InTrade and the Iowa Electronic Markets on the same events. Pennock et al. (2001) explored how combinatorial markets can maintain internal consistency and what happens when they don't.

On modern CFTC-regulated platforms like Kalshi, the existence of exchange fees and the lack of atomic multi-leg orders mean that small pricing inconsistencies can persist because they are not profitable to arbitrage after costs.

Key considerations

• Logical arbitrage vs. statistical arbitrage — Logical violations (bracket monotonicity) are risk-free in theory. Statistical correlation trades carry model risk — correlations are estimated, not guaranteed.
• Position limits — Kalshi may impose per-market position limits that prevent taking large enough positions to exploit small mispricings.
• Capital requirements — Both legs of an arbitrage tie up capital. If the capital cost exceeds the expected arbitrage profit, the trade is not worthwhile.
• Monitoring burden — Identifying pricing inconsistencies across hundreds of related markets requires systematic monitoring, not manual browsing.