Loading field map
Asset pricing theory is the branch of financial economics that studies how the prices of financial assets—stocks, bonds, derivatives, real estate, and other claims on future cash flows—are determined. Its central task is to explain why a given asset commands a particular price today, given the uncertainty about what it will pay in the future. The field seeks to answer two interconnected questions. First, what is the relationship between an asset's risk and its expected return? Second, how should investors value a stream of uncertain future payments? The stakes are practical as well as intellectual: asset prices allocate capital across firms and industries, determine the cost of capital for investment decisions, and shape the wealth of households and institutions. A theory of asset pricing is therefore also a theory of how risk is measured, priced, and shared in an economy.
Every financial asset can be viewed as a promise to deliver cash amounts at future dates. If those cash amounts were known with certainty, the asset's price would simply be the sum of those future payments, discounted at a risk-free interest rate reflecting the time value of money. The complication, and the heart of asset pricing theory, is that future cash flows are uncertain. An investor holding a share of stock does not know what dividends it will pay; a bondholder faces the possibility of default; a derivative's payoff depends on the evolution of some underlying price. The question is how to adjust the present value of an uncertain cash flow to reflect that uncertainty.
The field's fundamental insight, expressed in what is often called the fundamental theorem of asset pricing, is that an asset's price equals the expected value of its future payoffs, discounted by a stochastic discount factor. This factor, often denoted m, varies across states of the world: it is high in bad economic states, when consumption is low and the marginal utility of an extra dollar is high, and low in good states, when consumption is abundant. The price of any asset is then:
Price = E[m × payoff]
This single equation frames the entire field. Every asset pricing model is, in effect, a specification of the stochastic discount factor. The expected return on an asset is determined not by its total risk but by its covariance with this discount factor—that is, by how well the asset pays off in states where investors most need wealth. An asset that pays well in bad times is insurance and commands a lower expected return; an asset that fails precisely when the economy is weak must offer a higher expected return to attract investors. This covariance-based logic is the unifying thread that ties together the seemingly diverse models in the field.
The earliest modern asset pricing model, the Capital Asset Pricing Model (CAPM), emerged in the 1960s from work by William Sharpe, John Lintner, and Jan Mossin, building on Harry Markowitz's portfolio theory. The CAPM specifies the stochastic discount factor as a linear function of the return on the aggregate wealth portfolio—the "market portfolio." The model's claim is that the only risk that matters for an asset's expected return is its beta, the sensitivity of the asset's return to the market return. Expected return on an asset equals the risk-free rate plus beta times the market risk premium. The CAPM was revolutionary because it turned a portfolio construction technique into a full equilibrium theory of asset prices, and it remains the intellectual baseline against which later models are measured.
The empirical record of the CAPM is poor. Assets with high market beta do not reliably earn higher returns than low-beta assets, and other observable characteristics—firm size, the ratio of book value to market value, momentum—predict returns in ways the CAPM cannot explain. In response, researchers developed multifactor models. The most prominent is the Fama–French three-factor model, which adds size and value factors to the market factor, and its later extension to five factors which adds profitability and investment factors. These models are best understood not as deep theories but as empirical descriptions: they assert that the stochastic discount factor is a linear function of a small set of traded portfolios, and they document that these factors capture much of the cross-section of average stock returns.
The interpretation of these empirical factors is one of the most active disputes in the field. One view, associated with a research programme sometimes called "factor zoo" empiricism, holds that the factors proxy for different sources of risk that investors care about, and that the models are approximate descriptions of a multi-dimensional risk premium. An opposing view argues that the anomalies reflect behavioral biases—overreaction, underreaction, or investor sentiment—rather than rational compensation for risk. This rational-versus-behavioral debate has not been resolved, and the durable contribution of factor models may be less their specific factors than the methodology: the systematic search for characteristics that predict returns, and the construction of portfolios that capture those predictable patterns.
A different line of research, developed from the late 1970s onward, grounds the stochastic discount factor in the consumption decisions of a representative household. The consumption-based capital asset pricing model (CCAPM) specifies m as the marginal rate of substitution between consumption today and consumption in each possible future state. Under standard assumptions of power utility, an asset's expected return depends on its covariance with aggregate consumption growth. The appeal of this approach is that it explains why risk is priced at all: assets are risky because they expose investors to fluctuations in their standard of living, and the price of that risk is determined by how much investors dislike consumption volatility.
The empirical failure of the CCAPM is severe and instructive. The model requires an implausibly high degree of risk aversion to explain the observed equity premium—the gap between average stock returns and the risk-free rate. This difficulty, known as the equity premium puzzle, became a central problem for the field. subsequent work has attempted to repair the model by enriching assumptions: habits, where utility depends on consumption relative to a recent standard; recursive preferences, which separate risk aversion from the willingness to substitute consumption over time; long-run risk, where consumption growth has a small persistent component; or rare disasters, where investors fear infrequent but catastrophic consumption declines. These models share a common ambition: they attempt to generate the observed asset prices from a plausible specification of household preferences and the stochastic process for consumption, without appealing to frictions or mispricing.
These consumption-based models rarely deliver sharp closed-form predictions about the cross-section of returns, and they are difficult to test because aggregate consumption is measured imprecisely and slowly. Yet they matter because they provide the economic discipline that purely empirical factor models lack. A factor that cannot be traced back to underlying risk or preference is, under this research programme, an unexplained anomaly rather than a contribution to understanding.
A third tradition, which began in the early 1970s with the work of Fischer Black, Myron Scholes, and Robert Merton on option pricing, approaches asset pricing from a different direction. Rather than specifying preferences or an equilibrium, it asks: what can be said about the price of one asset given the prices of other assets? The key notion is arbitrage—the possibility of constructing a portfolio that costs nothing today, has no chance of losing money, and has a positive chance of making money. If markets are frictionless and such arbitrage is absent, then the prices of related assets must satisfy precise relationships.
The no-arbitrage approach reached its definitive form in the theory of complete markets and equivalent martingale measures, developed by J. Michael Harrison and Stanley Pliska. In this framework, the absence of arbitrage is not just a useful assumption but a necessary feature of any equilibrium. The stochastic discount factor exists precisely when there is no arbitrage, and it is unique when markets are complete—when every possible payoff can be replicated by trading in existing assets. For derivatives, this insight led to a remarkable practical result: the price of an option can be computed without knowing investors' risk preferences, because the option can be replicated by a dynamic strategy of trading the underlying asset and a risk-free bond. The replicating strategy pins down the price, and the price is consistent with any utility function as long as no arbitrage opportunities remain.
The no-arbitrage approach is the foundation of the modern derivatives industry. It does not explain why the prices of the underlying assets are what they are—it takes those prices as inputs—so it cannot on its own answer the field's central question about risk and return. But it provides the professionally indispensable toolkit for pricing options, swaps, and exotic contracts, and it supplies the conceptual bridge between observed prices and the models used to interpret them.
The four research programmes described above—the CAPM and its factor descendants, consumption-based equilibrium models, and no-arbitrage pricing—share a common simplifying assumption: markets are frictionless. Investors trade without transaction costs, cannot influence prices, have no taxes, can borrow and lend at the same rate, and face no constraints on short selling. Each programme relaxed some of these assumptions in specific directions, but a substantial body of work treats frictions as the central object of study.
This work examines how asset prices change in the presence of trading costs, liquidity risk, margin constraints, incomplete markets where some risks cannot be insured, heterogeneous investors who disagree or face different constraints, and limits to arbitrage—the possibility that mispricing persists because the investors who could correct it are unable to do so. Models in this tradition show that the law of one price can fail, that a security's liquidity can itself be a priced characteristic, and that an asset's price may depend on who holds it and on the structure of who can trade with whom. This literature bridges asset pricing with market microstructure, the study of how trading mechanisms affect prices.
These frictions are not mere refinements. They can overturn conclusions from frictionless models. For example, an asset that is otherwise identical to another but harder to trade can trade at a persistently lower price, exactly because the difficulty of trading is itself a risk that holders must bear. Such results matter for understanding the cross-section of returns across, say, government bonds with similar cash flows but different liquidity, or across closed-end funds and their underlying holdings.
As the field stands today, the different approaches have not merged into a single accepted theory. The empirical factor models are widely used in practice—for measuring fund performance, estimating the cost of capital, and in portfolio construction—but the lack of a clear economic interpretation of the factors remains unresolved. Consumption-based models have made progress in fitting aggregate stock market data, but they remain difficult to use for cross-sectional predictions and rely on preference assumptions that are hard to verify. No-arbitrage methods provide the workhorse models for derivatives but are silent on the bigger question of why the underlying assets are priced as they are. Frictions are increasingly recognized as central, but models with frictions are often harder to solve and test, and there is no unifying framework for which frictions matter most.
A notable recent development is the integration of these traditions. Many contemporary papers specify a consumption-based model with frictions—such as habits plus leverage constraints, or recursive preferences plus rare disasters—and then use the model's stochastic discount factor to price both aggregate returns and the cross-section. Another strand combines no-arbitrage constraints with empirical factors by asking what arbitrage relations imply about the joint dynamics of option prices and their underlying assets. Machine learning methods have entered the empirical side of the field, used to search for return predictors in large datasets, though whether they have changed the field's conceptual structure or merely its computational toolkit is an open question.
Perhaps the deepest unresolved issue in asset pricing is the equity premium puzzle and its relatives. The historical average return on stocks has been far higher than any standard consumption-based model can justify, and this fact has motivated a search for explanations—behavioral, institutional, or preference-based—that has not yet settled. Relatedly, the low volatility of interest rates, the high volatility of stock prices relative to fundamentals, and the predictability of returns over long horizons all remain only partially understood. The field is characterized less by consensus than by a set of competing frameworks, each with domains of success and failure, and the practitioners who use these models for valuation, risk management, and policy understand that the models are approximations whose validity depends on context.
For the educated newcomer, the durable structure of asset pricing theory is best grasped not as a chronological story of replacement but as a set of interrelated answers to one question: how do risk and time get translated into price? The CAPM and its descendants map risk onto a small set of observable portfolios. Consumption-based models ground that mapping in fundamental preferences. No-arbitrage theory derives what must be true of prices if markets are rational in the minimal sense of excluding free money. And frictions ask what happens when the idealized conditions break down. Each approach has a distinctive strength—the first is empirically operational, the second is economically disciplined, the third is mathematically rigorous, the fourth is institutionally realistic—and each has a corresponding limitation that the others attempt to address. The field's progress has been the gradual recognition that no single one of these perspectives is sufficient, and that the full pricing of risk requires elements of all of them.