tests of static asset pricing models

Upload: alexandra-hsiajsnaks

Post on 02-Jun-2018

216 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/11/2019 Tests of Static Asset Pricing Models

    1/47

    Tests of Static Asset Pricing

    Models

  • 8/11/2019 Tests of Static Asset Pricing Models

    2/47

    Tests of Static Asset Pricing Models

    In general asset pricing models quantify thetradeoff between risk and expected return.

    Need to both measure risk and relate it to theexpected return on a risky asset.

    The most commonly used models are:

    CAPM

    APT

    FF three factor model

  • 8/11/2019 Tests of Static Asset Pricing Models

    3/47

    Testable Implications

    These models have testable implications.For the CAPM, for example:

    Expected excess return of a risky asset isproportional to the covariance of its return andthat of the market portfolio.

    Note, this tells us the measure of risk used and its

    relation to expected return.There are other restrictions that depend upon

    whether there exists a riskless asset.

  • 8/11/2019 Tests of Static Asset Pricing Models

    4/47

    Testable Implications

    For the APT,

    The expected excess return on a risky asset is

    linearly related to the covariance of its returnwith various riskfactors.

    These risk factors are left unspecified by the

    theory and have been:

    Derived from the data (CR (1983), CK)

    Exogenously imposed (CRR (1985))

  • 8/11/2019 Tests of Static Asset Pricing Models

    5/47

    Plan

    Review the basic econometric methodology wewill use to test these models.

    Review the CAPM. Test the CAPM.

    Traditional tests (FM (1972), BJS (1972), Ferson andHarvey)

    ML tests (Gibbons (1982), GRS (1989)) GMM tests

    Factor models: APT and FF

    Curve fitting vs. ad-hoc theorizing

  • 8/11/2019 Tests of Static Asset Pricing Models

    6/47

    Econometric Methodology Review

    Maximum Likelihood Estimation

    The Wald Test

    The F Test

    The LM Test

    A specialization to linear models and linearrestrictions

    A comparison of test statistics

  • 8/11/2019 Tests of Static Asset Pricing Models

    7/47

    Review of Maximum Likelihood

    Estimation Let {x1, xT} be a sample of T, i.i.d.

    random variables.

    Call that vector x.

    Let xbe continuously distributed with density

    f(x|).

    Where, is the unknown parameter vector thatdetermines the distribution.

  • 8/11/2019 Tests of Static Asset Pricing Models

    8/47

    The Likelihood Function

    The joint density for the independent randomvariables is given by:

    f(x1|

    )f(x2|

    )f(x3|

    )f(xT|

    ) This joint density is known as the likelihood

    function,L(x|)

    L(x|)= f(x1|)f(x2|)f(x3|)f(xT|)

    Can you write the joint density andL(x|) thisway when dealing with time-dependentobservations?

  • 8/11/2019 Tests of Static Asset Pricing Models

    9/47

  • 8/11/2019 Tests of Static Asset Pricing Models

    10/47

    Idea Behind Maximum Likelihood

    Estimation Pick the parameter vector estimate, , that

    maximizes the likelihood,L(x|), of

    observing the particular vector ofrealizations, x.

  • 8/11/2019 Tests of Static Asset Pricing Models

    11/47

    MLE Plusses and Minuses

    Plusses: Efficient estimation in terms of

    picking the estimator with the smallest

    covariance matrix.Question: are ML estimators necessarily

    unbiased?

    Minuses: Strong distributional assumptionsmake robustness a problem.

  • 8/11/2019 Tests of Static Asset Pricing Models

    12/47

    MLE Example: Normal Distributions where

    OLS assumptions are satisfied

    Sample yof size T is normally distributed

    with mean xwhere

    Xis a T x K matrix of explanatory variables

    is a K x 1 vector of parameters

    The variance-covariance matrix of the errors

    from the true regression is 2

    I, whereIis a T x T identity matrix

  • 8/11/2019 Tests of Static Asset Pricing Models

    13/47

    The Likelihood Function

    The likelihood function for the linear model

    with independent normally distributed

    errors is:

  • 8/11/2019 Tests of Static Asset Pricing Models

    14/47

    The Log-Likelihood Function

    With independent draws, it is easier to maximize

    the log-likelihood function, because products are

    replaced by sums. The log-likelihood is given by:

  • 8/11/2019 Tests of Static Asset Pricing Models

    15/47

  • 8/11/2019 Tests of Static Asset Pricing Models

    16/47

  • 8/11/2019 Tests of Static Asset Pricing Models

    17/47

  • 8/11/2019 Tests of Static Asset Pricing Models

    18/47

    The Information MatrixCont

    The MLE achieves the Cramer-Rao lower bound, whichmeans that the variance of the estimators equals the inverseof the information matrix:

    Now,

    note, the off diagonal elements are zero.

    ).,(

    21

    I

  • 8/11/2019 Tests of Static Asset Pricing Models

    19/47

    The Information MatrixCont

    The negative of the expectation is:

    The inverse of this is:

  • 8/11/2019 Tests of Static Asset Pricing Models

    20/47

    Another way of Writing I(,2)

    For a vector, , of parameters, I(), theinformation matrix, can be written in a secondway:

    This second form is more convenient forestimation, because it does not require estimatingsecond derivatives.

  • 8/11/2019 Tests of Static Asset Pricing Models

    21/47

    Estimation

    The Likelihood Ratio Test

    Let be a vector of parameters to be estimated.

    Let H0be a set of restrictions on theseparameters.

    These restrictions could be linear or non-linear.

    Let be the MLE of estimated withoutregard to constraints (the unrestricted model).

    Let be the constrained MLE.

    U

    U

    R

  • 8/11/2019 Tests of Static Asset Pricing Models

    22/47

    The Likelihood Ratio Test Statistic

    If and are the likelihoodfunctions evaluated at these two estimates,

    the likelihood ratio is given by:

    Then, -2ln() = -2(ln( )ln( ) ~2with degrees of freedom equal to thenumber of restrictions imposed.

    )( UUL )( RRL

    )

    (

    UUL )

    (

    RRL

    )(

    )(

    UU

    RR

    L

    L

  • 8/11/2019 Tests of Static Asset Pricing Models

    23/47

    Another Look at the LR Test

    Concentrated Log-Likelihood: Many problemscan be formulated in terms of partitioning a

    parameter vector, into {1, 2} such that thesolution to the optimization problem, can bewritten as a function of , e.g.:

    Then, we can concentrate the log-likelihoodfunction as: F*(1, 2) = F(1, t(1)) Fc().

    2

    1

    ).( 12 t

  • 8/11/2019 Tests of Static Asset Pricing Models

    24/47

    Why Do This?

    The unrestricted solution to

    then provides the full solution to the

    optimization problem, since t is known.

    We now use this technique to find estimates

    for the classical linear regression model.

    )( 11 cFMax

  • 8/11/2019 Tests of Static Asset Pricing Models

    25/47

  • 8/11/2019 Tests of Static Asset Pricing Models

    26/47

    Ex: Concentrating the Likelihood Function

    Inserting this back into the log-likelihood yields:

    Because (y- X)(y- X) is just the sum of

    squared residuals from the regression (ee) we can

    rewrite ln(Lc) as:

  • 8/11/2019 Tests of Static Asset Pricing Models

    27/47

    Ex: Concentrating the Likelihood Function

    For the restricted model we obtain the restricted concentrated log-likelihood:

    So, plugging in these concentrated log-likelihoods into our definitionof the LR test, we obtain:

    Or, T times the log of the ratio of the restricted SSR and theunrestricted SSR, a nice intuition.

    )(

    1ln)2ln(1

    2

    )ln( ' RRcR ee

    T

    TL

    ee

    eeTLR RR

    'ln

    '

  • 8/11/2019 Tests of Static Asset Pricing Models

    28/47

  • 8/11/2019 Tests of Static Asset Pricing Models

    29/47

    ExampleCont

    The first-order conditions for the estimates and

    simply reduce to the OLS normal equations:

  • 8/11/2019 Tests of Static Asset Pricing Models

    30/47

    ExampleCont

    Solving

    Substituting into the FOC for yields:

    xy

    T

    t t

    T

    t tt

    xxyyxx

    1

    2

    1

    )()))(((

  • 8/11/2019 Tests of Static Asset Pricing Models

    31/47

    ExampleCont

    Solve for as before:2

    2

    1

    2 )(1

    T

    t

    tt xyT

  • 8/11/2019 Tests of Static Asset Pricing Models

    32/47

    ExampleCont

    The restricted model is exactly the same, except that is constrained

    to be one, so that the normal equation reduces to:

    and

    One can then plug in to obtain and form the likelihood ratio, which

    is distributed 2(1).

    2R

  • 8/11/2019 Tests of Static Asset Pricing Models

    33/47

    The Wald Test

    The problem with LR test: Need both restricted

    and unrestricted model estimates.

    One or the other could be hard to compute. The Wald test is an alternative that requires

    estimating the unrestricted model only.

    Suppose y~ N(X, ), with a sample size of T,

    then:21 ~)()'( TXyXy

  • 8/11/2019 Tests of Static Asset Pricing Models

    34/47

    The Wald TestCont

    Under the null hypothesis that E(y) = X, the

    quadratic form above has a 2distribution. If the

    hypothesis is false, the quadratic form will belarger, on average, than it would be if the null

    were true.

    In particular, it will be a non-central 2with the

    same degrees of freedom, which looks like acentral 2, but lies to the right.

    This is the basis for the test.

  • 8/11/2019 Tests of Static Asset Pricing Models

    35/47

    The Restricted Model

    Now, step back from the normal and let be the

    parameter estimates from the unrestricted model.

    Let restrictions be given byH0:f() = 0.

    If the restrictions are valid, then should satisfy

    them.

    If not, should be farther from zero than

    would be explained by sampling error alone.

    )(f

  • 8/11/2019 Tests of Static Asset Pricing Models

    36/47

    Formalism

    The Wald statistic is

    Under H0in large samples, W ~ 2with d.f. equal to the

    number of restrictions. See Greene ch.9 for details.

    Lastly, to use the Wald test, we need to compute the

    variance term:

    )()])([()'( 1 ffVarfW

  • 8/11/2019 Tests of Static Asset Pricing Models

    37/47

    Restrictions on Slope Coefficients

    If the restrictions are on slope coefficients of a linear regression, then:

    where

    and K is the number of regressors.

    Then, we can write the Wald Statistic:

    where J is the number of restrictions.

    12 )'(][][ XXsVarVar

    22 ' KT

    TKTees

    ][)())'(])'()[(()'( 2112 JfGXXsGfW

  • 8/11/2019 Tests of Static Asset Pricing Models

    38/47

    Linear Restrictions

    H0: R- q= 0

    For example, suppose there were three betas, 1,

    2, and 3. Lets look at three tests.(1) 1= 0,

    (2) 1= 2,

    (3) 1= 0 and 2= 2. Each row of Ris a single linear restriction on the

    coefficient vector.

  • 8/11/2019 Tests of Static Asset Pricing Models

    39/47

    Writing R

    Case 1:

    Case 2:

    Case3:

  • 8/11/2019 Tests of Static Asset Pricing Models

    40/47

    The Wald Statistic

    In general, the Wald statistic with J linear

    restrictions reduces to:

    with J d.f.

    We will use these tests extensively in our

    discussion of Chapters5 and 6 of CLM.

    ][]')'([]'[ 112 qRRXXRsqRW

  • 8/11/2019 Tests of Static Asset Pricing Models

    41/47

  • 8/11/2019 Tests of Static Asset Pricing Models

    42/47

    Why Do We Care?

    We care because in a linear model withnormally distributed disturbances under the

    null, the test statistic derived above is exact.This will be important later because undernormality, some of our cross-sectional CAPMtests will be of this form and,

    A sufficient condition for the (static) CAPM tobe correct is for asset returns to be normallydistributed.

  • 8/11/2019 Tests of Static Asset Pricing Models

    43/47

    The LM Test

    This is a test that involves computing only

    the restricted estimator.

    If the hypothesis is valid, at the value of therestricted estimator, the derivative of the log-

    likelihood function should be close to zero.

    We will next form the LM test with the J

    restrictions f() = 0.

  • 8/11/2019 Tests of Static Asset Pricing Models

    44/47

    The LM TestCont

    This is maximized by choice of and

    )(')]()'[(

    )2(

    1)ln(

    2

    )2ln(

    2

    )ln(2

    2

    FXyXyTT

    LLM

    .2

  • 8/11/2019 Tests of Static Asset Pricing Models

    45/47

    First-order Conditions

    and

  • 8/11/2019 Tests of Static Asset Pricing Models

    46/47

    The LM TestCont

    The test then, is whether the Lagrange

    multipliers equal zero. When the

    restrictions are linear, the test statisticbecomes (see Greene, chapter 7):

    where J is the number of restrictions.

    ][]')'([]'[ 112 qRRXXRsqRLM R

  • 8/11/2019 Tests of Static Asset Pricing Models

    47/47

    W, LR, LM, and F

    We compare them forJlinear restrictions in thelinear model with K regressors. It can be shownthat:

    and that W > LR > LM.

    ,FJKT

    TW

    ,1

    1ln

    FJ

    KTTLR

    ,]))/(1(1)[(FJ

    FJKTKTTLM