ASSET PRICING

MODELS
TOPIC # 7
Ajay Pandey
IIM Ahmedabad
ASSET PRICING MODELS:
MOTIVATION
 In the topic, portfolio analysis, we focused on how investors may choose to invest given the
opportunities available in financial markets. In that discussion, an investor was assumed to be
able to form estimates for all securities of E (ȓi)’s, σi’s and ρij’s, i.e., expected returns and
variance-covariance matrix. There was no discussion on the price at which assets were
available.
 From the first principles, i.e., the topics on time value of money and on equity valuation; we
know that the price of an asset is based on the cash flow from the asset, the probability
distribution of cash flow, and the discount rate used to discount the resultant expected cash
flow.
 Asset pricing models provide the analytical framework for figuring out as to how financial
markets may arrive at the discount rate while valuing risky cash flows from a financial asset,
whose payoffs/cash flows are risky or uncertain.
PRELIMINARIES
 Recall that any risk-averse investor, concerned only with expected return and variance of
returns, invests in a diversified portfolio of risky asset and risk-free asset. The risky asset
portfolio is based on the estimates of E (ȓi)’s, σi’s and ρij’s by the investor. In general, these
estimates could be different across investors as these estimates are forward-looking and there
is no way of perfectly forecasting the future.
 Notwithstanding the lack of ability to forecast perfectly, rational pricing by investors requires
that they are sensitive to violation of “ law of one price” and exploit any “arbitrage
opportunity” offered by the market.
 Further, like any other market, the market for any financial asset also clears through the price
of the asset such that demand is equal to the supply of the asset.
‘CAPM’ OR CAPITAL ASSET
PRICING MODEL
 Lets assume that all investors somehow estimate identical E (ȓi)’s, σi’s and ρij’s for all risky
assets given their prices, Pi. Essentially, there is no information asymmetry and all
expectations are homogenous.
 Further, assume that they are interested in investing for a single period. They all can lend and
borrow at risk-free rate, rf.
 There are no transaction costs and assets can be bought and sold in any quantity at the given
prices.
 If all the above assumptions hold, then all investors will hold some fraction of their investment
in the risk-free asset and the remaining fraction in identical risky asset portfolio, which is
tangency portfolio having highest Sharpe ratio.
 Now, if all the investors hold same efficient portfolio of risky assets, it has to be ‘market
portfolio’ logically.
CAPM AND CAPITAL MARKET
LINE
CAPM AND THE EXPECTED
RETURN OF AN ASSET
 While the CML specifies highest possible Sharpe ratio, how does one get to the object of interest-
expected return, E (ȓi), for any arbitrary risky asset in the CAPM world.
 Recall that the market portfolio has the highest Sharpe ratio. So if we combine any asset i with
weight wi and balance (1-wi) in the market portfolio, M, then the Sharpe ratio will attain maxima
when wi = 0. We know how to write expression for expected return and variance for such a
portfolio and Sharpe ratio requires just that.
 If we take maxima of Sharpe ratio as a function of wi at wi = 0, we get the following

E (ȓi) = rf + [E (ȓm) - rf ] *[ρim *σi * σm / σ2m]

= risk-free + market risk-premium* beta of the stock/asset
 For step-wise derivation of CAPM, please refer to:
http://people.duke.edu/~charvey/classes/ba350_1997/capm/capm.htm#:~:text=The%20above
%20derivation%20of%20the,hold%20mean%20variance%20efficient%20portfolios.
SECURITY MARKET LINE
(SML)
 The expected return of a stock is a linear function of its beta and hence can be graphically
represented as-
CAPM: IMPLICATIONS
 CAPM implies that the expected return of any asset or portfolio is risk-free return plus risk-premium.
 The risk-premium required by the market for any asset or portfolio depends upon the covariance of its
return with market relative to variance of market returns. This parameter, beta, of any asset or
portfolio is the sensitivity of its return relative to market portfolio.
 Beta or the sensitivity parameter multiplied with market risk premium determines the asset’s or
portfolio’s risk-premium. Since market risk-premium is a invariant across risky assets, the risk-
premium is a linear function in beta.
 Since beta is a measure of co-movement or sensitivity w.r.t. market portfolio, it is a measure of
systematic risk. And only systematic risk is relevant for risk-premium.
 Out of the total risk or variance of any asset returns, some part is caused by the systematic risk, beta,
and the remaining part is idiosyncratic risk, as we shall see later. Idiosyncratic risk can be diversified,
as we saw earlier in portfolio analysis, but the systematic risk due to covariance cannot be diversified
away and hence requires risk-premium for bearing that risk.
UNDERSTANDING BETA:
MARKET MODEL
 To understand the idea of how beta represents systematic risk, let us look at the return of an asset and that of
market portfolio, both of which are risky, and are denoted by ȓi and ȓm respectively. Because payoffs and
therefore returns on both are based on real economy, we acknowledge that in general they would be having
some covariance/correlation.
 Let us posit that the return on an asset, ȓi, is a linear function of return on the market portfolio, ȓm . In that case,
we can write-
ȓi = αi + βi * ȓm + i
 Based on the above, the variance equation will be-

σ2i = β2i σ2m + σ2εi

 Which of the two elements from can be diversified away if we assume Cov( i , j )=0 for all i and j? In other
words, all co-movements across assets are due to their covariance with market portfolio.
 In such a scenario, the variance can be decomposed in two parts such that the Total Risk= Systematic Risk +
Idiosyncratic Risk.
EXTENDING THE MARKET
MODEL TO A PORTFOLIO
 Let us extend the market model to all securities. In such case, what would be the variance of a portfolio having w i weight in ith
stock/security?
 Variance of a portfolio’s return can be expressed as-

N N
Var [ ∑ wi ȓi] = Var [ ∑ wi ( αi + βi * ȓm + i )]
i=1 i=1
N N
= ∑ (wi βi)2*Var (ȓm) + ∑ ….. Cov (ȓm, i) {which is zero}
i=1 i=1

+ ∑ ∑ …….. Cov (i , j )
 If Cov (i , j ) =0 for all i’s and j’s, then –

σ2 p = β p 2 * σ2 m
 If the market model is able to explain all co-movement between securities, then only beta of the portfolio determines the risk
of the portfolio returns. Conversely, if the co-movements are not completely due to market return (or market risk) then the
portfolio risk will be more than what can be explained by just the beta of the portfolio. This has implications for multi-factor
models as we will see.
OPERATIONALIZING CAPM…
 Operationalizing the CAPM requires- (a) risk-free rate, (b) risk-premium for ‘market’ portfolio,
and (c) beta of the security.
 What is ‘market’ portfolio? By construction, it consists of all the risky assets in the economy be
they real or financial. Can everyone invest in such a portfolio? In fact, it includes future earnings
of individuals (labor- unskilled, skilled and professional) in the economy!
 At best, listed and traded financial assets are available for investment and hence a proxy for
market such as S&P 500 or equivalent is used. Why?
 Risk-premium is estimated using large enough sample assuming that realized excess returns (over
risk-free rate) in the past provide unbiased estimate for risk-premium expected going forward.
 Risk-free rate used is yield or return on short-term T-bills as CAPM is a single-period model.
 Beta is estimated using index model as discussed earlier, i.e., by regressing stock return on the
corresponding market returns.
CAPM: EMPIRICAL EVIDENCE
 Using high market capitalization index as a proxy for market portfolio, the empirical evidence
suggests that historically low (high) beta stock have generated returns higher (lower) than
what is expected, if the CAPM is valid.
 This means that there might be more than one-risk factor, which may be taken into account by
the market. It could also mean that the proxy for market portfolio is not specified correctly or
that the proxy (stock index) is not mean-variance efficient at least, ex post.
 There are anomalies such as ‘size’ (small cap stocks relative to large cap), ‘value’ (low price-
to-book stocks relative to high), ‘momentum’ (the stocks which have done well in the recent
past relative to the ones which have performed poorly) which generate higher return after
adjusting for beta. Similarly, ‘illiquid’ stocks also generate higher returns. All the above are
risky strategies in the sense that they do not generate higher return consistently but do that
over longer horizon and risks assumed in following them cannot be diversified away.
CAPM: WHAT COULD BE THE
LIMITATIONS?
 To start with, the CAPM assumes that only mean (expected return) and variance of return
matters. The return distribution may create risks which is not fully accounted for by the
variance.
 Further, the investor may not just be interested in the likely returns and the associated
probabilities but might also be interested in under what condition those returns are generated.
Somewhat loosely speaking, even if two different securities/portfolios may have identical
expected return and variance yet they might be seen not as equally desirable if one pays off
well in good times and the other one in bad times.
 Investors may not be able to invest in all real (propertied and embodied) and financial assets.
In such a case, the co-movement between their financial and other real assets will impact the
risk assumed and associated risk-premium.
 Of course, expectations may not be homogenous and finally, market/investors may be
irrational (but not in a perfectly predictable and consistent manner! Why?).
MULTI-FACTOR MODELS: APT
 The Arbitrage Pricing Theory (APT) generalizes the idea behind CAPM by making no explicit assumption about
either the preferences of investor or by restricting their choice based on expected return and variance of return.
 It is based on the idea from portfolio theory that security-specific idiosyncratic risk can be diversified away but not the
risks due to sensitivity of the securities to systematic sources of risk (beta risk). Given the assumptions of CAPM, there
is only one systematic risk factor (market risk), whereas in APT, there could be K factors (say, macroeconomic factors)
which are priced by the market (with K<<<<N, why?). In such a case, the expected return of i th security is given by-
​ E(ȓi ​)=rf​+β1​f1​+β2​f2​+…+βK​fK​
where:
E(ȓi ​)= Expected return; rf​= Risk-free return;
βK​= Sensitivity to the factor of k; fK​= Kth factor risk-premium​
 One of the K factors could be the market factor.
​
MULTI-FACTOR MODELS:
FAMA-FRENCH
 Recall that empirically there are anomalies such as size, and value whereby small stocks relative
to large and value stocks relative to growth seem to have performed better. These have been
incorporated in Fama-French three-factor model, which is-
E(ȓi ​) = rf​+ β1​i [E (ȓm) - rf ] + β2i​SMB​+β3i​HML​

where, SMB is the risk premium for size factor (small minus big) and
HML is the risk-premium for value factor (high minus low)
 Extending it is four-factor Carhart model, which is-

E(ȓi ​) = rf​+ β1​i [E (ȓm) - rf ] + β2i​SMB​+β3i​HML​+β4iWML

where, WML is the risk-premium for momentum factor (winner minus losers)
MULTI-FACTOR MODELS FOR
RISK: BARRA
 While the APT and Fama-French models are more focused on which systematic risks are
priced (or, require risk-premium) in the market; the other class of multi-factor models are used
to understand the risks a portfolio might be exposed to. This is based on the fact that the
covariance between securities need not be only due to the risk factors priced in the market.
 One such popular model is that of Barra (now part of Morgan Stanley). It uses over 40 data
metrics, including earnings growth, share turnover, and senior debt rating in addition to yield,
earnings growth, volatility, liquidity, momentum, size, price-earnings ratio, leverage, and
growth.
 Analogous to market model, each securities return sensitivity is measured with respect to
multiple factors. The idea is that the residuals across securities will be uncorrelated after
throwing large number of factors.
SUMMARY
 The basic motivation of asset pricing models is to figure out how market discounts future cash
flows or payoffs to price any risky asset.
 All of the models recognize that some of the risks are diversifiable and hence may not require
risk-premium.
 The expected return or discount rate therefore, may depend upon the systematic risk of an
asset.
 The models vary in terms of how many risk factors might be priced and which ones? They
also vary in terms of how many factors may explain co-movement of securities return. The
latter is particularly important in risk management.
Any Questions?