QuantRocket logo

© Copyright Quantopian Inc.
© Modifications Copyright QuantRocket LLC
Licensed under the Creative Commons Attribution 4.0.

Disclaimer

The Capital Asset Pricing Model and Arbitrage Pricing Theory

by Beha Abasi, Maxwell Margenot, and Delaney Granizo-Mackenzie

The Capital Asset Pricing Model (CAPM) is a classic measure of the cost of capital. It is used often in finance to evaluate the price of assets and to assess the impact of the risk premium from the market at large. In this lecture, we discuss the CAPM and the more general Arbitrage Pricing Theory (APT) to form a basis for evaluating the risk associated with various factors.

Idiosyncratic and Systematic Risk

In general, portfolios and assets can face two types of risk: idiosyncratic and systematic risk. Idiosyncratic risk refers to risks that are firm-specific and can be diversified away, such as a management change or a faulty production, while systematic risk is market-wide and affects all market participants. An example could be a slowing of the economy or a change in the interest rate. Because all firms are exposed to systematic risk, it cannot be diversified away.

Risk Premia

As the number of assets in a portfolio increases, many of the idiosyncratic risks cancel out and are diversified away. This is the key reason why we want to avoid position concentration risk. As your portfolio grows larger and makes more independent bets through diversification, the variance of the portfolio declines until only the systematic risk remains. As we cannot remove systematic risk, investors must be given a risk premium above the rate of risk-free return to compensate them for the risk they take on by investing in this portfolio. The individual firm-level risks in this portfolio do not have associated premia as this would create arbitrage opportunities. Shareholders could collect the risk premium while diversifying away the risk associated with them. That would mean additional profit without any additional exposure. This is the definition of an arbitrage opportunity!

From this reasoning we can conclude that the premium on an asset should have no relation to its idiosyncratic risk, but should instead rely solely on the level of systematic risk it carries. In order to accurately compute the risk premium of an asset, and consequently our expected return, we need to find a measure of systematic risk. If we have that, then we can theoretically define the return of an asset in the following way:

$$E[\mbox{Return}] = \mbox{Risk-Free Rate of Return} + \mbox{Risk Premium}$$

One way to do this is to estimate how changes in the excess return of an asset are related to changes in the excess return of the market. Expressing this as a linear regression gives us the relationship as the change in expected return of an asset for each 1% change in the return of the market portfolio.

In theory, this market portfolio should have no diversifiable risk left and would therefore only fluctuate with systematic shocks. In practice, we use a market index such as the S&P500 as a proxy for the market portfolio. The beta that we get from regressing an asset's returns on the returns of the market will be our measure of systematic risk. This beta represents the sensitivity of an asset's return stream to market-wide shocks.

Given this beta, the risk premium of asset $i$ is defined as:

$$\mbox{Risk Premium of Asset}_i = \beta (\mbox{Market Risk Premium})$$

We call this simplistic model the Capital Asset Pricing Model (CAPM).

Capital Asset Pricing Theory

We can express the CAPM more clearly like so:

$$E[R_i] = R_F + \beta(E[R_M] - R_F)$$

where $R_i$ is the return of asset $i$, $R_F$ is the risk-free rate, and $R_M$ is the return of the market. The CAPM is one of the most basic measures of the cost of capital. It determines the minimum return required to entice investors to hold a certain asset.

To put it another way, CAPM says that the return of an asset should be the risk-free rate, which is what we would demand to account for inflation and the time value of money, as well as something extra to compensate us for the amount of systematic risk we are exposed to.

We can then use our calculated beta exposure to make predictions of returns.

CAPM Assumptions

In our derivation of the CAPM, we made two main assumptions:

The following is a programmatic derivation of the efficient frontier for portfolios of four assets.

This uses the cvxopt module which is not installed by default. To install it, execute the following cell:

Each blue dot represents a different portfolio, while the red line skimming the outside of the cloud is the efficient frontier. The efficient frontier contains all portfolios that are the best for a given level of risk.

The optimal, or most efficient, portfolio on this line is found by maximizing the Sharpe ratio, the ratio of excess return and volatility. We use this to determine the portfolio with the best risk-to-reward tradeoff.

The line that represents the different combinations of a risk-free asset with a portfolio of risky assets is known as the Capital Allocations Line (CAL). The slope of the CAL is the Sharpe ratio. To maximize the Sharpe ratio, we need to find the steepest CAL, which coincides with the CAL that is tangential to the efficient frontier. This is why the efficient portfolio is sometimes referred to as the tangent portfolio.

We can look at the returns and risk of the individual assets compared to the optimal portfolio we found to easily showcase the power of diversification.

Capital Market Line is CAL through market portfolio

Our optimal portfolio has a decently high return as well as less risk than any individual asset, as expected. Theoeretically, all investors should demand this optimal, tangent portfolio. If we accumulate the portfolios of all investors, we end up with the market portfolio, since all shares must be held by someone. This means that the tangency portfolio is the market portfolio, essentially saying that demand must equal supply.

When a risk-free asset is added to the portfolio, the Capital Asset Line turns into the Capital Market Line (CML). According to the CAPM, any stock or portfolio that lies to the right of CML would contain diversifiable risk and is therefore not efficient.

The mapping of each security's beta to its expected return results in the Security Markets Line. The difference between a security's return and the expected return as predicted by CAPM is known as the alpha.

For more details on the CAPM, check out the wikipedia page.

Arbitrage Pricing Theory

The CAPM, while widely used and studied, has many drawbacks. With strict, limiting assumptions, it does not hold up well in empirical tests. Arbitrage Pricing Theory (APT) aims to generalize the CAPM model, as assets may be exposed to classes of risks other than the market risk and investors may care about things other than just the mean and variance.

APT is a major asset pricing theory that relies on expressing the returns using a linear factor model:

$$R_i = a_i + b_{i1} F_1 + b_{i2} F_2 + \ldots + b_{iK} F_K + \epsilon_i$$

A factor is a return stream that is determined completely by some characteristic. For example, the CAPM has only one factor, market return. If we have modelled our rate of return as above, then the expected returns should take the form of:

$$ E(R_i) = R_F + b_{i1} \lambda_1 + b_{i2} \lambda_2 + \ldots + b_{iK} \lambda_K $$

where $R_F$ is the risk-free rate, and $\lambda_j$ is the risk premium - the return in excess of the risk-free rate - for factor $j$. This premium arises because investors require higher returns to compensate them for incurring higher risk.

We'll compute the risk premia for our factors with Fama-Macbeth regression. However, there are various ways to compute each $\lambda_j$!

Arbitrage

Now that we have a reasonably general way to compute expected return, we can discuss arbitrage more technically. There are generally many, many securities in our universe. If we use different ones to compute the $\{\lambda_i\}$, will our results be consistent? If our results are inconsistent, there is an arbitrage opportunity (in expectation), an operation that earns a profit without incurring risk and with no net investment of money. In this case, we mean that there is a risk-free operation with expected positive return that requires no net investment. It occurs when expectations of returns are inconsistent, i.e. risk is not priced consistently across securities.

Say that there is an asset with expected rate of return 0.2 for the next year and a $\beta$ of 1.2 with the market, while the market is expected to have a rate of return of 0.1, and the risk-free rate on 1-year bonds is 0.05. Then the APT model tells us that the expected rate of return on the asset should be

$$ R_F + \beta \lambda = 0.05 + 1.2 (0.1 - 0.05) = 0.11$$

This does not agree with the prediction that the asset will have a rate of return of 0.2. So, if we buy \$100 of our asset, short \\$120 of the market, and buy \$20 of bonds, we will have invested no net money and are not exposed to any systematic risk (we are market-neutral), but we expect to earn $0.2(100) - 0.1(120) + 0.05(20) = 9$ dollars at the end of the year.

The APT assumes that these opportunities will be taken advantage of until prices shift and the arbitrage opportunities disappear. That is, it assumes that there are arbitrageurs who have sufficient amounts of patience and capital. This provides a justification for the use of empirical factor models in pricing securities: if the model was inconsistent, there would be an arbitrage opportunity, and so the prices would adjust.

Goes Both Ways

Accurately knowing $E(R_i)$ is incredibly difficult, but this model tells us what the expected returns should be if the market is free of arbitrage. This lays the groundwork for strategies based on factor model ranking systems. If you have a model for the expected return of an asset, then you can rank those assets based on their expected performance and use this information to make trades. This creation of a ranking scheme is the hallmark of a long-short equity strategy.

Testing Arbitrage Pricing Theory

Most empirical tests of the APT are done in two steps: estimating the betas of individual factors, then comparing it to actual prices to see how predictions fared.

Here we will use the return streams from long-short equity strategies built from various microeconomic indicators as our factors. Then, we will use the Fama-Macbeth regression method to estimate our risk premia.

Now we use pipeline to get all of our data.

To get our factor return streams, we rank equities based on their profit margin, their R&D spending, and their cash flow. Then, for each indicator, we go long the assets in the top percentile and short the ones in the bottom.

Finally, we'll put everything together in our Fama-Macbeth regressions. This occurs in two steps.

First, for each asset we regress its returns on each factor return stream:

$$R_{1, t} = \alpha_1 + \beta_{1, F_1}F_{1, t} + \beta_{1, F_2}F_{2, t} + \dots + \beta_{1, F_m}F_{m, t} + \epsilon_{1, t} \\ R_{2, t} = \alpha_2 + \beta_{2, F_1}F_{1, t} + \beta_{2, F_2}F_{2, t} + \dots + \beta_{2, F_m}F_{m, t} + \epsilon_{2, t} \\ \vdots \\ R_{n, t} = \alpha_n + \beta_{n, F_1}F_{1, t} + \beta_{n, F_2}F_{2, t} + \dots + \beta_{n, F_m}F_{m, t} + \epsilon_{n, t}$$

Second, we take the beta estimates from the first step and use those as our exogenous variables in an estimate of the mean return of each asset. This step is the calculation of our risk premia, $\{\gamma_K\}$.

$$E(R_i) = \gamma_0 + \gamma_1 \hat{\beta}_{i, F_1} + \gamma_2 \hat{\beta}_{i, F_2} + \dots + \gamma_m \hat{\beta}_{i, F_m} + \epsilon_i$$

It is imperative that we not just use our model estimates at face value. A scan through the accompanying statistics can be highly insightful about the efficacy of our estimated model. For example, notice that although our individual factors are significant, we have a very low $R^2$. What this may suggest is that there is a real link between our factors and the returns of our assets, but that there still remains a lot of unexplained noise!

For a more in-depth look at choosing factors, check out the Factor Analysis with Alphalens lecture.

Now that we have estimates for our risk premia we can combine these with our beta estimates from our original regression to estimate asset returns.

Finally, as a rough comparison between APT and CAPM, we'll look at the returns from Long-Short strategies constructed using each model as the ranking scheme.

Next Lecture: Beta Hedging

Back to Introduction

This presentation is for informational purposes only and does not constitute an offer to sell, a solicitation to buy, or a recommendation for any security; nor does it constitute an offer to provide investment advisory or other services by Quantopian, Inc. ("Quantopian") or QuantRocket LLC ("QuantRocket"). Nothing contained herein constitutes investment advice or offers any opinion with respect to the suitability of any security, and any views expressed herein should not be taken as advice to buy, sell, or hold any security or as an endorsement of any security or company. In preparing the information contained herein, neither Quantopian nor QuantRocket has taken into account the investment needs, objectives, and financial circumstances of any particular investor. Any views expressed and data illustrated herein were prepared based upon information believed to be reliable at the time of publication. Neither Quantopian nor QuantRocket makes any guarantees as to their accuracy or completeness. All information is subject to change and may quickly become unreliable for various reasons, including changes in market conditions or economic circumstances.