Factor Models, Arbitrage Pricing Theory, and Arbitrage Mechanics

Factor Models and Arbitrage Pricing Theory

  • Factor models describe asset returns as a function of systematic risk factors rather than idiosyncratic security-specific characteristics.

  • The Capital Asset Pricing Model (CAPM, referenced phonetically as CAPAB) is a single-factor model that imposes a strict general equilibrium condition:

    • In equilibrium, the market portfolio represents the optimal tangency portfolio.
    • Every investor holds the market portfolio according to market capitalization weights, scaling risk exposure through the risk-free asset.
  • The Arbitrage Pricing Theory (APT) provides an alternative framework to the CAPM:

    • APT models returns as a linear function of multiple systematic factors.
    • Unlike the CAPM, APT does not rely on strict equilibrium conditions or the assumption that all investors hold the market portfolio.
    • APT is derived from the law of one price and the absence of arbitrage opportunities in well-diversified portfolios.
  • Hedge funds utilize the APT framework to identify factor exposures, detect relative mispricings, and construct market-neutral trading strategies.

Systematic Risk Compensation and Arbitrage Fundamentals

  • In a factor model, asset returns represent compensation for bearing exposure to systematic sources of risk:

    • Market risk compensation (βmarket\beta_{\text{market}}).
    • Momentum risk compensation (βmomentum\beta_{\text{momentum}}).
    • Additional macroeconomic and fundamental factor compensations.
  • Theoretical definition of pure arbitrage requires three simultaneous conditions:

    • Zero net investment today (Cost=0\text{Cost} = 0): The portfolio is self-financing and requires no initial capital outlay.
    • Risk-free profile (Risk=0\text{Risk} = 0): The position carries zero variance and no exposure to unexpected economic states.
    • Certain positive profit in the future (Profit>0\text{Profit} > 0): The position produces a strictly non-negative return across all possible future states, with a strictly positive payoff in at least one state.
  • Violations of the Law of One Price create arbitrage opportunities:

    • The Law of One Price dictates that if two assets generate identical cash flows in all future states of nature, they must sell for the exact same price today.
    • Discrepancies between prices of identical cash flows allow arbitrageurs to short the overvalued asset and purchase the undervalued asset to lock in riskless profit.

Real-World Arbitrage and Market Frictions

  • Case study on cross-market mispricing (SKH):

    • A cross-listed company (SKH) registered shares in the United States while trading domestically in South Korea.
    • A substantial price divergence emerged between the domestic shares and the US-registered shares for the exact same underlying equity.
    • Arbitrageurs entered the market by buying the cheaper shares and selling the more expensive shares with the intention of converting domestic shares into US-listed stock.
    • Regulatory intervention: The Korean regulator intervened due to market disruption, placing a binding quota/cap on the conversion of domestic stock into American shares.
    • This regulatory barrier broke the conversion mechanism and prevented price convergence, causing arbitrage trades to fail.
  • Treasury cash-futures basis:

    • Empirical market pricing frequently reveals a basis or pricing gap between the physical US Treasury bond and the US Treasury futures contract.
    • Differentiation between price risk and liquidity risk:
    • Price risk at maturity is zero because a US Treasury bond pays a fixed par value of exactly 100100 dollars at maturity, leaving no cash flow uncertainty.
    • Liquidity risk remains present prior to maturity, exposing arbitrageurs to balance sheet constraints, margin calls, and execution frictions.

State-Space Arbitrage Construction

  • State-space scenario setup:

    • Three securities: Stock A, Stock B, and Stock C.
    • Current price for each stock is identical today: P0(A)=8P_0(A) = 8 dollars, P0(B)=8P_0(B) = 8 dollars, and P0(C)=8P_0(C) = 8 dollars.
    • The economy evaluates three discrete future states.
  • Synthetic replication:

    • Constructing a portfolio with a weight of 0.50.5 in Stock A and 0.50.5 in Stock B yields a guaranteed payoff of 1010 dollars in each of the three states.
    • Because the combined payout across all states is identically 1010 dollars, the synthetic asset is entirely free of uncertainty.
    • Stock C yields a payoff of only 99 dollars across the states for the same purchase price of 88 dollars today.
  • Executing the arbitrage:

    • Form a long position in the synthetic combination of A and B while shorting the underperforming counterpart C.
    • Condition enforcement: The strategy requires zero net capital outlay at t=0t = 0 and delivers positive payoff with zero variance across all future states.

Factor-Based Arbitrage and Beta Neutrality

  • Identifying positive alpha:

    • Consider an asset with an abnormal positive expected alpha of 2%2\%.
    • An investor cannot simply buy the asset outright to capture alpha, as doing so introduces unhedged factor exposure (e.g., a factor beta of β=1.3\beta = 1.3).
  • Beta neutralization condition:

    • To transform a positive alpha into a risk-free arbitrage, the systematic exposure must be completely eliminated: βportfolio=0\beta_{\text{portfolio}} = 0.
    • The risk-free asset has a beta of zero (βf=0\beta_f = 0) and cannot be used to offset or neutralize a positive systematic factor beta.
    • To eliminate a factor exposure of β=1.3\beta = 1.3, an investor must short a portfolio PP carrying positive systematic factor exposure.
  • Mathematical conditions for an arbitrage portfolio with weights wiw_i:

    • Zero factor risk (Beta neutrality):     ∑i=1Nwiβi=0\sum_{i=1}^N w_i \beta_i = 0
    • Zero initial cost (Self-financing condition):     ∑i=1Nwi=0\sum_{i=1}^N w_i = 0
    • Strictly positive alpha payoff:     ∑i=1Nwiαi>0\sum_{i=1}^N w_i \alpha_i > 0
  • Equation constraints in portfolio formation:

    • With two unknown weights (wAw_A and wBw_B), solving for an arbitrage requires two independent linear equations: the beta neutrality equation and the zero-cost equation.
    • If a system has three unknown weights and only two equations, the system is underdetermined and cannot yield a unique solution without additional constraints.

Practical Implementation, Regulatory Risks, and APT Equilibrium

  • Structured factor arbitrage example:

    • Consider benchmark Portfolio A, a risk-free asset or portfolio F, and a mispriced Asset E.
    • Asset E has a factor beta of βE=23\beta_E = \frac{2}{3} and an expected return of E[RE]=9%E[R_E] = 9\%.
    • Evaluating Asset E relative to the benchmark identifies a positive alpha (αE>0\alpha_E > 0), indicating that the optimal arbitrage strategy must go long Asset E (wE>0w_E > 0).
    • To achieve zero net systematic factor risk, the long exposure in Asset E must be matched by a short position in Portfolio A:     βE−wAβA=0\beta_E - w_A \beta_A = 0
    • Setting βA=1\beta_A = 1, the investor must sell exactly 23\frac{2}{3} units of Portfolio A for every unit of Asset E held:     23−23(1)=0\frac{2}{3} - \frac{2}{3}(1) = 0
    • To fulfill the zero-cost condition (∑wi=0\sum w_i = 0), the net proceeds must be balanced through the zero-beta asset/portfolio F:     wE+wA+wF=0w_E + w_A + w_F = 0
  • Theoretical assumptions versus practical limits of arbitrage ("Patrick's risk"):

    • In theoretical models, arbitrage is assumed to be fully costless and riskless.
    • In reality, arbitrageurs face structural constraints, execution costs, and legal/regulatory interventions (referred to colloquially as practical risk or "Patrick's risk").
    • Unanticipated interventions, such as a regulator capping share conversions, truncate the arbitrage mechanism and expose positions to open-ended losses.
  • Arbitrage Pricing Theory equilibrium condition:

    • In theoretical APT equilibrium, the existence of competitive arbitrageurs rapidly eliminates any abnormal alpha opportunities.
    • Therefore, in expectation, all alphas across assets are driven to zero (α=0\alpha = 0):     E[Ri]=Rf+∑k=1Kβi,kλkE[R_i] = R_f + \sum_{k=1}^K \beta_{i,k} \lambda_k
    • Expected returns are entirely accounted for by the asset's linear sensitivities (β\beta) to systematic factor risk premia (λ\lambda).
  • Exam scope note:

    • All content discussed up through factor arbitrage and APT marks the conclusion of examinable material for the midterm exam.
    • Subsequent coursework covering the Information Ratio and advanced performance evaluation will not appear on the midterm.

Questions & Discussion

  • Question / Prompt: If an asset displays a positive outcome or alpha of 2%2\%, why can an investor not simply purchase the asset outright?

    • Response: Simply purchasing the asset leaves the position exposed to systematic factor risk (such as a beta of 1.31.3). To construct a genuine arbitrage, the factor risk must be eliminated. Because the risk-free asset has a beta of zero (β=0\beta = 0), adjusting cash weights cannot lower the factor beta; the investor must actively short an asset with positive factor beta to achieve beta neutrality.
  • Question / Prompt: Why must the portfolio weights sum to zero (∑wi=0\sum w_i = 0)?

    • Response: A pure arbitrage requires zero net capital expenditure at inception (Cost=0\text{Cost} = 0). To ensure the trade costs nothing to establish, the long positions must be exactly financed by short positions, forcing the sum of weights to equal zero.
  • Question / Prompt: When constructing the arbitrage for Asset E with β=23\beta = \frac{2}{3}, what position must be taken in Portfolio A?

    • Response: Portfolio A must be sold (shorted). Specifically, an investor must sell 23\frac{2}{3} units of Portfolio A to cancel out the 23\frac{2}{3} beta of Asset E, ensuring the net portfolio beta equals zero.
  • Question / Prompt: Is the pricing spread between Treasury futures and physical Treasuries completely free of risk?

    • Response: There is no price risk at maturity because the Treasury pays a certain par value of 100100 dollars. However, liquidity risk exists prior to maturity, meaning the trade is not free of intermediate holding and liquidity risks.