Skip to main content
EMH vs. Behavioural Finance

Arbitrage and Its Real-World Limits: Why Mispricings Persist

Pomegra Learn

Why Can't Arbitrage Fix All Market Mispricings?

Arbitrage is the theoretical mechanism that should eliminate all mispricings: if a security is overpriced, sell it; if underpriced, buy it; repeat until prices converge to fundamentals. In classical finance, the threat of arbitrage is supposed to keep prices efficient. Yet mispricings visibly persist. Stocks trading at absurd valuations, bonds mispriced relative to credit spreads, commodities trading at nonsensical backwardations. If arbitrage works, these shouldn't exist. They do exist because arbitrage faces real-world limits. Capital for arbitrage is finite and costly. Arbitrageurs face leverage constraints, funding costs, and the risk that their correct analysis won't be vindicated before capital runs out. Some mispricings are too small to justify arbitrage costs. Some are too risky—betting against a bubble requires withstanding months or years of further misprice before reversal, during which you lose money and capital. These limits are not theoretical; they're binding constraints on what arbitrage can accomplish. Understanding the limits of arbitrage is essential for understanding why behavioral biases create lasting market effects rather than being instantly corrected.

The classical efficient market hypothesis depends on arbitrage working perfectly. If anyone identifies a mispricing, they buy the underpriced security and short the overpriced one, locking in risk-free profit. The act of arbitrage pushes prices together, correcting the mispricing. If this arbitrage were fast and free, all mispricings would be immediately corrected and all profitable opportunities would instantly disappear. Markets would be perfectly efficient. But arbitrage is neither fast nor free. Capital that can perform arbitrage is limited. Shorting restrictions, funding costs, and the need to hold positions until convergence all constrain arbitrage. Additionally, arbitrage is risky—the bet that prices will converge isn't risk-free; it's just lower-risk than directional bets. This risk creates the possibility of losses during the arbitrage holding period, which constrains how much capital arbitrageurs will deploy.

Quick definition: Limits of arbitrage are the real-world constraints—capital availability, funding costs, leverage restrictions, convergence risk, and holding period risk—that prevent perfect arbitrage from correcting all mispricings. These limits allow inefficiencies to persist and create opportunities for investors who understand the constraints.

Key takeaways​

  • Pure arbitrage (risk-free profit) is nearly impossible; practical arbitrage (low-risk profit) faces real constraints that limit capital deployment
  • Funding costs, leverage limits, and capital availability restrict how much arbitrage capital can address mispricings
  • Convergence risk: even correct analysis that a mispricing exists doesn't guarantee quick convergence, forcing arbitrageurs to hold losing positions
  • Limits to arbitrage create a margin of opportunity—inefficiencies can persist at magnitudes that violate classical efficiency but don't violate real-world constraints
  • Behavioral biases create mispricings that trigger arbitrage, but arbitrage's limits prevent complete price correction, allowing behavioral effects to persist
  • Understanding these limits explains why apparently obvious mispricings don't instantly correct and why behavioral finance insights remain profitable

The Theory of Perfect Arbitrage​

In theory, arbitrage is straightforward. Suppose Stock A is trading at $100 and Stock B is identical but trading at $98. Buy B and short A. Collect $100 from the short, spend $98 on the long. Lock in a $2 profit regardless of how prices move relative to each other. The profit is independent of the market direction; it's pure arbitrage.

In practice, this description masks reality. To execute the trade, you need:

  1. Capital to buy Stock B ($98)
  2. Ability to short Stock A (not all securities can be easily shorted)
  3. A broker willing to finance the position (funds not available instantly; they cost money)
  4. Confidence that the two stocks will remain identical (they might not perfectly correlate)
  5. Confidence that prices will converge before your capital is recalled

Each of these introduces a real-world friction that the classical theory ignores. Capital is limited and expensive. Shorting availability is restricted. Financing costs money. Correlations can break. Capital can be recalled. These frictions mean that even the most obvious arbitrage is not truly risk-free, and therefore not infinitely profitable.

Finite Capital and Arbitrage Constraints​

The amount of capital available to perform arbitrage is limited. There are only so many active arbitrageurs managing capital. They have position limits, risk limits, and leverage limits. When a large mispricing appears (like during the Long-Term Capital Management crisis of 1998, when Russian government bonds diverged massively from other emerging market credits), arbitrageurs quickly deploy their capital. Once deployed, capital is exhausted—there's no more available.

This creates a peculiar dynamic: the most profitable arbitrage opportunities (the largest mispricings) exhaust available capital fastest. A 1% mispricing might attract modest capital; a 10% mispricing attracts everyone's capital. But if everyone's capital is exhausted on the largest mispricings, smaller mispricings remain uncorrected. The result is a distribution of mispricings where the largest ones attract exhaustive arbitrage (capital-limited), and smaller ones persist due to capital constraints.

During the 2008 financial crisis, this constraint was painfully visible. Credit spreads widened to absurd levels—investment-grade corporate bonds traded at spreads to Treasuries exceeding 600 basis points, levels that violated centuries of credit history. According to classical arbitrage logic, massive capital should have flooded in to buy these bonds and short duration (sell Treasuries). Capital did flow in, but not enough—spreads didn't compress immediately. Why? Because capital for relative-value arbitrage had been exhausted on other mispricings (equity prices, mortgage spreads, currency basis). There was plenty of "obvious" arbitrage, but finite capital meant not all could be addressed simultaneously.

The Federal Reserve's emergency lending facilities during the crisis were explicitly designed to inject capital into arbitrage to address the most problematic mispricings. Without that capital injection, mispricings would have persisted far longer, highlighting that private arbitrage capital alone was exhausted.

Funding Costs and Leverage Constraints​

Arbitrage requires capital, and capital is not free. If you need to borrow $100 million to perform an arbitrage earning $500,000 annually, the funding costs matter enormously. If overnight rates are 0.1%, funding costs are minimal ($10,000 annually, 2% of profit). If rates are 5%, funding costs are $5 million (10x the profit), rendering the arbitrage unprofitable.

This explains why arbitrage opportunities increase in low-rate environments and decrease in high-rate environments. In 2010-2020, near-zero rates made even small-profit arbitrage worthwhile. After 2022, when rates rose above 4%, the same arbitrage opportunities became unprofitable. The mispricings didn't change; the funding costs did. This is a limit to arbitrage: rates rising makes capital more expensive, constraining arbitrage capacity.

Additionally, most arbitrage requires leverage. An arbitrageur with $100 million might lever it to $500 million to execute more arbitrage. Leverage multiplies returns—a 0.5% arbitrage profit becomes a 2.5% return on capital. But leverage also multiplies losses and increases margin call risk. Regulatory changes and financial stability concerns have repeatedly tightened leverage constraints, effectively reducing arbitrage capital.

The Dodd-Frank Act (post-2008) increased capital and leverage requirements for banks and trading firms, reducing the amount of leverage they could employ. This directly reduced the capital available for arbitrage. Securities lending became less available (funding through lending collapsed during the crisis and never fully recovered), making shorting more difficult and expensive. These constraints are not theoretical; they're binding limits on arbitrage capacity.

Convergence Risk and the Timing Problem​

Here's where arbitrage becomes genuinely risky: the bet that prices will converge is not risk-free because convergence might not happen on your timescale. An arbitrageur might be correct that two prices should converge, but if convergence takes longer than expected (or doesn't happen before their capital is recalled), losses mount.

Consider a classic example: Long-Term Capital Management in 1998. LTCM analyzed emerging market bonds and identified that Russian government bonds traded at much wider spreads to other emerging market credit than fundamentals justified. LTCM bought Russian bonds and shorted other emerging credits, betting they would converge. They were probably correct in their analysis—the spreads were indeed too wide. But in August 1998, Russia defaulted on its bonds. Prices didn't converge to LTCM's fundamental prediction; they converged to a new, much worse state.

LTCM lost nearly all its capital despite being "correct" in a fundamental sense. They identified a real mispricing but failed to anticipate the mechanism that would change prices—default, not spread compression. This is convergence risk: the bet that prices will move toward fair value might be correct in direction but wrong in timing or mechanism.

Convergence risk prevents arbitrageurs from holding indefinitely. If the convergence takes 10 years and arbitrage capital is recalled in 1 year, the arbitrageur loses money despite being fundamentally correct. This forces arbitrageurs to limit the positions they take in favor of trades they believe will converge quickly. The result is that slow-moving mispricings persist because they're not worth the risk of holding against convergence uncertainty.

Short Sales Constraints and Directional Bias​

Many of the most obvious mispricings require short selling to exploit. If Stock A is wildly overvalued, the arbitrage is to short it. But short-selling faces real constraints. Some stocks are difficult or impossible to borrow for shorting. Some markets (like certain emerging markets) restrict short selling entirely. Short sellers face indefinite holding requirements—there's no guaranteed buyback, so they might hold short positions for years.

These constraints create a "short-sale bias" in markets: overpriced securities are less easily corrected (shorting is hard) than underpriced securities (buying is easy). This explains why extreme overvaluation can persist longer than extreme undervaluation. Tech bubble stocks in 2000 traded at absurd valuations for months despite being clearly overpriced, partly because widespread shorting was impossible (shares unavailable to borrow, short-squeezed prices). By contrast, when stocks trade cheaply, arbitrage can buy them immediately.

The 2021 GameStop situation exemplifies short-sale constraints. The stock was by virtually any fundamental analysis wildly overvalued. Arbitrage should have corrected this through massive shorting. But short-sale restrictions (shares unavailable to borrow, regulatory concerns about short squeezes) limited shorting capacity. The mispricing persisted for months despite being obvious. This directly demonstrates limits to arbitrage—the corrective mechanism was constrained.

Regulatory restrictions on short selling (enacted after the 2008 crisis for financial stocks) explicitly acknowledged that short-sale limits can create overvaluation. If short selling is restricted, prices can diverge upward from fundamentals without the corrective pressure of arbitrage. This is a policy-driven limit to arbitrage.

Liquidity Risk and Flash Crashes​

Arbitrage requires liquidity—you need to be able to sell the overpriced asset and buy the underpriced one at quoted prices. But liquidity is not guaranteed. During stress, liquidity evaporates. Bid-ask spreads widen, forcing arbitrageurs to take losses on execution. This execution risk limits how much arbitrage capital will deploy when stress is increasing.

The 2010 Flash Crash exemplified this dynamic. Liquidity evaporated in seconds, and prices executed at absurd levels ($1 stocks trading at $0.01, etc.). Arbitrageurs who tried to exploit the dislocation couldn't execute at quoted prices; they faced worse prices due to lack of liquidity. The mispricings were real but uncorrectable due to liquidity evaporation.

More broadly, illiquid securities can trade far from fundamental value because arbitrage is impossible when you can't find a buyer. Securities with low trading volume might be mispriced for extended periods. It's not that arbitrage isn't willing to correct them; it's that arbitrage can't execute due to lack of counterparties.

This liquidity constraint is particularly binding during crisis. The securities most likely to be mispriced during crisis (illiquid corporate bonds, emerging market sovereigns, structured securities) are precisely those where arbitrage is most difficult because liquidity has evaporated. The corrections are delayed until liquidity returns and arbitrage becomes feasible.

Hedging Costs and the Cross-Security Problem​

Some mispricings involve complex relationships across securities. A corporate bond might be mispriced relative to equity, require hedging the equity move through options (which are expensive), and the hedging cost exceeds the profit. Or a pair-trade might require shorting one illiquid security and buying another, facing execution costs on both sides.

These complexities create mispricings that are real but uneconomical to arbitrage. A $10 million bond mispricing might exist but cost $15 million in hedging and execution to exploit, making it unprofitable despite the existence of the opportunity. These uneconomical mispricings persist not because arbitrage failed but because arbitrage was never economically viable.

The complexity of modern finance creates many such situations. Structured products are often mispriced by small amounts that are real but don't justify the cost of understanding them well enough to arbitrage. Credit derivatives might offer small pickups over cash bonds but require modeling and execution through complex transactions. The result is persistent mispricings in complex securities that simpler, more liquid securities don't exhibit.

Real-World Examples​

The Closed-End Fund Discount (1930-Present): Closed-end funds often trade at discounts to their net asset value—the fund is worth less than the value of its holdings. This should be instantly arbitraged: buy the fund at discount, simultaneously sell its holdings at fair value. Yet the discount persists, sometimes for decades. Why? Because (1) transaction costs of unwinding holdings exceed the discount, (2) tax efficiency and legal constraints limit arbitrage, (3) retail investor behavior drives the discount (they sell the fund even though fair value is higher). This is a textbook example where a clear mispricing persists because arbitrage constraints are binding.

The 3-Month LIBOR-OIS Spread Crisis (2008): During the financial crisis, the spread between LIBOR and OIS (overnight index swaps) widened to extreme levels, indicating a severe credit stress signal. Arbitrage should have corrected this spread—they're both interbank reference rates that should be nearly identical. But they diverged sharply because (1) capital for arbitrage was exhausted elsewhere, (2) funding became available only at punitive rates, (3) counterparty credit risk made arbitrage itself risky. The spread didn't correct until central banks injected liquidity and credit conditions improved. This was a real mispricing (spreading had no fundamental justification), but arbitrage was constrained.

Tesla Valuation (2020-2023): Tesla traded at valuations (price-to-sales, price-to-book) far exceeding historical automotive company norms, suggesting overvaluation. Shorting it directly faced (1) unavailability of shares to borrow, (2) short squeezes where other shorts covered, (3) fundamental uncertainty about whether EV transition would render traditional comparison metrics obsolete. These constraints prevented arbitrage from correcting the overvaluation for years, even as skeptics were right that the valuation was extreme.

Common Mistakes in Analyzing Arbitrage Limits​

Mistake 1: Assuming that if arbitrage is constrained, all mispricings should be assumed permanent. Constraints limit arbitrage capital but don't eliminate it. Over time, constraints relax (rates fall, capital becomes available, leverage limits rise), and delayed arbitrage occurs. A mispricing that persisted for years might correct in a month when conditions change.

Mistake 2: Treating arbitrage capital as truly "risk-free." Even pure arbitrage positions carry risks—counterparty risk, liquidity risk, execution risk, timing risk. These risks are smaller than directional risks, but they're real. A model that treats arbitrage as completely riskless underestimates how much capital is actually required to hold arbitrage positions through stress scenarios.

Mistake 3: Ignoring that arbitrageurs have other opportunities. An arbitrageur deciding whether to deploy capital to correct mispricing A must compare to the expected return from mispricing B, C, etc. Capital allocation across multiple mispricings means that the largest opportunities attract capital (correct), but smaller opportunities remain uncorrected (not a market failure, just rational capital allocation).

Mistake 4: Assuming convergence speed is predictable. Arbitrageurs might know that a mispricing will eventually correct but have no ability to predict when. Holding a position for 1 month at 0.5% monthly returns looks attractive; holding for 12 months looks less attractive. Uncertainty about convergence timing constrains capital deployment.

Mistake 5: Conflating limits to arbitrage with evidence of irrationality. A persistent mispricing doesn't prove that market participants are irrational; it might prove that arbitrage constraints prevent correction. The market could be rationally pricing in the constraint (e.g., "this bond is worth 95, but it will trade at 100 due to constraints, so the price reflects expected value").

FAQ​

If limits to arbitrage are binding, can behavioral biases create lasting market effects?​

Yes. If behavioral biases create mispricings and arbitrage is too constrained to correct them, the mispricings can persist and create real financial market dynamics. Bubbles are explained this way: behavioral bias creates overvaluation, arbitrage is constrained (especially short selling), overvaluation persists and grows, and the bubble eventually pops when constraints relax or sentiment shifts.

Do algorithmic traders eliminate limits to arbitrage?​

Partially. Algorithms can execute faster than humans and identify arbitrage more systematically. But they face constraints too: leverage limits, funding costs, execution slippage in less-liquid securities. Algorithms might reduce the magnitude of arbitrage-correctable mispricings, but they don't eliminate them. Flash crashes suggest that algorithmic trading can create new mispricings when algorithms respond similarly and liquidity evaporates.

How does the distinction between arbitrage and fundamental value matter?​

True arbitrage is risk-free (or nearly so) and should pay off regardless of market direction. Fundamental analysis tries to identify real value, which might be different from market price but isn't guaranteed to win. Arbitrage limits apply to true arbitrage; if you're making a directional bet on fundamental value, you face different constraints (market risk, volatility, timing of realization).

Why do market makers not arbitrage away all mispricings?​

Market makers perform arbitrage constantly, but they face the same constraints as other arbitrageurs: capital limits, funding costs, leverage limits. Additionally, market makers deliberately add a small markup (bid-ask spread) for their service. They arbitrage the most profitable opportunities but don't arbitrage everything. The small mispricings that remain are roughly equal to market maker spreads.

Can central bank intervention overcome arbitrage limits?​

Yes, explicitly. When the Federal Reserve injects liquidity or purchases securities, it's providing capital and reducing funding costs for arbitrage. This relaxes constraints and allows private arbitrage to correct mispricings that seemed stuck. The 2008 Fed interventions did exactly this—they solved the liquidity and capital shortage that was preventing arbitrage from correcting the most severe mispricings.

Does the existence of arbitrage limits mean that beating the market is easy?​

No. Even if arbitrage is limited, prices still adjust toward fundamental value over time. The opportunities created by limits are usually small (covered by transaction costs) and available only to those with capital and expertise. Most investors can't exploit them profitably.

Are limits to arbitrage static or changing?​

Changing constantly. Technology reduces execution costs and timing constraints. Regulation changes leverage limits. Interest rates change funding costs. These shifts alter which mispricings are economically correctable. A mispricing that was uncorrectable in a high-rate environment becomes correctable when rates fall.

Summary​

Limits to arbitrage explain why behavioral biases and mispricings persist in real markets despite the theoretical presence of arbitrage as a corrective mechanism. Perfect arbitrage is nearly impossible due to real-world constraints: capital for arbitrage is finite and expensive, funding costs reduce profits, leverage limits restrict position sizing, and convergence risk means that price correction might not happen on the arbitrageur's timescale. These constraints interact—the largest mispricings exhaust available capital fastest, leaving smaller mispricings uncorrected; funding cost changes alter which mispricings are economically correctable; liquidity risk prevents arbitrage during the periods when mispricings are largest. Short-sale constraints create a bias toward overvaluation persisting longer than undervaluation. The result is a market where mispricings are not instantly corrected but gradually arbitraged away as conditions permit. Understanding these limits explains why obviously wrong prices don't instantly correct, why behavioral finance insights remain profitable rather than immediately arbitraged away, and why crisis periods with extreme capital constraints create the largest mispricings. Arbitrage is not absent—it continuously works to correct prices—but it's constrained by real-world factors that prevent the instantaneous, perfect correction that classical finance assumes. These constraints create the gaps where behavioral investors can profitably exploit mispricings.

Next​

→ Noise Traders and Market Prices