Loading field map
Asset pricing theory asks a deceptively simple question: what determines the price of a financial asset? The answer has unfolded through a sequence of frameworks that disagree about which risks matter, whether markets are efficient, and how far arbitrage can correct mispricing. Two core tensions run through the entire story: risk-based explanations versus behavioral explanations, and equilibrium models versus no-arbitrage logic. Each framework emerged by addressing a gap left by its predecessors, and several now coexist, serving different roles in the field's current division of labor.
Harry Markowitz's 1952 paper "Portfolio Selection" shifted the focus from picking individual stocks to constructing portfolios. The key insight was that investors should care about the trade-off between expected return and variance of return, not just expected return alone. By formalizing diversification—showing how combining assets with imperfect correlations reduces portfolio risk—Markowitz gave investors a quantitative tool for optimization. The efficient frontier, the set of portfolios offering the highest expected return for each level of risk, became the benchmark for rational portfolio choice. Modern Portfolio Theory (MPT) did not yet produce an asset pricing model; it described how a single investor should behave, not how prices are set in equilibrium. But it supplied the foundation: if everyone optimizes mean-variance portfolios, what must prices look like?
William Sharpe, John Lintner, and Jan Mossin independently extended MPT into an equilibrium pricing model. The Capital Asset Pricing Model (CAPM) assumes that all investors hold the same beliefs and can borrow and lend at a risk-free rate. Under these conditions, every investor holds the same risky portfolio—the market portfolio—and the only risk that matters for pricing is the asset's covariance with that market portfolio, measured by beta. Expected return on any asset is the risk-free rate plus beta times the market risk premium. CAPM was a breakthrough: it turned portfolio theory into a testable prediction about the cross-section of returns. But its assumptions were strong, and empirical tests soon revealed anomalies—patterns in returns that beta alone could not explain, such as the size and value effects.
At roughly the same time, a different tradition emerged from the work of Kenneth Arrow and Gérard Debreu on general equilibrium under uncertainty. State-Price Asset Pricing does not start with investor preferences or a market portfolio. Instead, it prices assets by assigning a price to each possible future state of the world. The price of any asset is the sum of its payoffs in each state multiplied by the state price. This framework introduced the concept of a stochastic discount factor (SDF)—a random variable that adjusts payoffs for risk and time. The SDF framework is not a single model but an infrastructure: any asset pricing model can be expressed as a specification of the SDF. CAPM, for instance, implies a particular SDF that is linear in the market return. State-price thinking also underlies derivatives pricing, where the absence of arbitrage forces the existence of a positive SDF. This framework coexists with equilibrium models by providing a unifying language rather than competing directly.
CAPM was a single-period model. Robert Merton's Intertemporal CAPM (ICAPM, 1973) extended the logic to a dynamic setting where investors care about future investment opportunities. In ICAPM, risk factors include not just the market return but also state variables that predict changes in the investment opportunity set—such as interest rates or dividend yields. This preserved the equilibrium approach while allowing multiple sources of priced risk. A parallel development, the Consumption-Based Capital Asset Pricing Model (CCAPM), emerged from the work of Robert Lucas (1978) and Douglas Breeden (1979). CCAPM absorbs the intertemporal insight by linking asset prices directly to aggregate consumption growth: an asset is risky if its return is low when consumption growth is low. The SDF in CCAPM is the marginal rate of substitution between current and future consumption. This framework is elegant and tightly grounded in microeconomic theory, but it struggled empirically. The equity premium puzzle—the observation that stocks have offered far higher returns than bonds relative to their consumption risk—suggested that either investors are extremely risk-averse or the model misses something fundamental.
Stephen Ross's Arbitrage Pricing Theory (APT) took a different route. Instead of deriving prices from equilibrium, APT relies on the absence of arbitrage: if two portfolios have the same risk, they must have the same expected return. APT assumes that asset returns are driven by a small number of common factors, and that idiosyncratic risk can be diversified away. The expected return on any asset is a linear function of its sensitivities to those factors. APT does not specify what the factors are; it provides a theoretical justification for multi-factor pricing without requiring investors to hold the market portfolio. This made APT more flexible than CAPM but also less precise. The coexistence of CAPM and APT reflects a deeper divide: CAPM is an equilibrium model with strong assumptions and a single factor; APT is a no-arbitrage model that leaves factor identity open. Both remain active, but APT's structure proved especially fertile for empirical work.
By the early 1980s, a growing list of anomalies—excess volatility, the equity premium, the size and value effects—challenged the risk-based frameworks. Robert Shiller's 1981 paper "Do Stock Prices Move Too Much to be Justified by Subsequent Changes in Dividends?" argued that stock price fluctuations are far larger than any rational model of future dividends could justify. This was not a minor tweak; it suggested that investor sentiment, not just fundamentals, drives prices. Behavioral Asset Pricing draws on psychological evidence about how investors actually form beliefs and make decisions. Models in this tradition incorporate overconfidence, loss aversion, and limited attention to explain patterns that risk-based models cannot. Prospect Theory, developed by Daniel Kahneman and Amos Tversky, provided a microfoundation for how people evaluate gambles differently from expected utility maximization. Behavioral models do not reject the idea that risk matters; they argue that mispricing from investor psychology can persist and affect prices alongside risk. This raised a natural question: if rational arbitrageurs exist, why don't they eliminate mispricing?
While APT provided the theoretical justification for multiple factors, it did not say which factors to use. Empirical Multi-Factor Asset Pricing filled that gap. The landmark 1992 paper by Eugene Fama and Kenneth French showed that two firm-level characteristics—size (market equity) and value (book-to-market ratio)—captured variation in average stock returns that beta alone missed. Their three-factor model added a size factor and a value factor to the market factor. This was not a purely empirical exercise; it was guided by the APT logic that any systematic source of risk should be priced. But the factors themselves were discovered empirically, not derived from theory. The Fama-French model became the workhorse for evaluating portfolio performance and testing new anomalies. Later work added factors for momentum, profitability, and investment. The explosion of factors—hundreds have been proposed—created a new tension: are these factors compensation for risk, or do they reflect persistent mispricing? Empirical multi-factor models are now the standard toolkit, but they do not settle the debate between risk-based and behavioral explanations.
If behavioral models identify mispricing, why does it persist? The Limits to Arbitrage framework, formalized by Andrei Shleifer and Robert Vishny in their 1997 paper "The Limits of Arbitrage," explains that real-world arbitrage is costly, risky, and constrained. Arbitrageurs typically manage other people's money and face the risk that mispricing worsens before it corrects, forcing them to liquidate at a loss. Noise trader risk—the possibility that sentiment becomes more extreme—can deter arbitrage even when a mispricing is obvious. This framework does not replace behavioral asset pricing; it complements it by explaining why anomalies are not quickly arbitraged away. Limits to Arbitrage also has broader implications for market efficiency: it suggests that markets can be efficient in normal times but become inefficient precisely when mispricing is largest, because that is when arbitrage is most constrained. The framework transformed the debate from "are markets efficient?" to "under what conditions is arbitrage effective?"
Today, no single framework dominates asset pricing. They coexist because they serve different purposes. CAPM remains a benchmark for introductory teaching and for estimating the cost of equity in practice, even though its empirical failures are well known. State-Price Asset Pricing provides the unifying SDF language that all other models can be expressed in. Intertemporal and Consumption-Based models continue to be refined, especially in macro-finance, where the equity premium puzzle and related puzzles drive research into habits, rare disasters, and long-run risks. APT and Empirical Multi-Factor models are the standard tools for empirical work: researchers use factor models to test new anomalies and to evaluate fund managers. Behavioral Asset Pricing and Limits to Arbitrage together offer a coherent alternative to the risk-based tradition, explaining anomalies as the product of psychological biases combined with constraints on arbitrage.
The central unresolved disagreement is whether the factors identified by empirical models—size, value, momentum, profitability—reflect compensation for systematic risk or persistent mispricing. Risk-based theorists argue that these factors capture dimensions of undiversifiable risk that matter to marginal investors. Behavioral theorists argue that the factors arise from investor overreaction, underreaction, and sentiment, and that limits to arbitrage prevent their elimination. Both sides can point to supporting evidence, and the debate remains active. What both sides agree on is that the single-factor CAPM is insufficient, that multiple factors are needed to describe the cross-section of returns, and that the SDF framework is the correct mathematical language for expressing any asset pricing model. The field's pluralism is not a sign of weakness; it reflects the complexity of financial markets, where risk, psychology, and institutional constraints all play a role.