Unformatted text preview:

12 Week 5a Empirical methods overheads12.1 Motivation and overview1. Expected return, beta model, as in CAPM, Fama-French APT.1TS regression, define  : = + 0+  =1 2 .2Model: ()=0 (+)(+) because the point of the model is that  should = 0.2. How do we take this to the data? Objectives:(a) Estimate parameters. ˆˆˆ.(b) Standard errors of parameter estimates. How do ˆˆˆ vary if you draw new data andtry again?(c) Test the model. Does ()=0? Are the true  =0?Arethe ˆ that we see due tobad luck rather than real failure of the model?(d) Test one model vs. another. Can we drop a factor, e.g. size?12.2 Time-series regression1. Time Series Regression (Fama-French)(a) Method:i. Run= + 0+  =1 2 .foreachii. Interpret the regression as a description of the cross section, without running moreregressions³´= 0 ()+ =1 2i()eiERSlopei1()Ef201(b) Estimates:i. ˆˆ : OLS time-series regression.= + 0+  =1 2 for each .ii.ˆ : Mean of the factor,ˆ =1X=1=¯(c) Standard e rrors: Assume independent over time (but not across portfolios).i. OLS standard errors ˆˆ.ii.ˆ :(ˆ)=()√(d) Test  are jointly zero?i. Answer: look atˆ0(ˆ ˆ0)−1ˆPrecise forms,ˆ0(ˆ)−1ˆ = h1+¯0Σ−1¯i−1ˆ0Σ−1ˆ˜2 −  − h1+¯0ˆΣ−1¯i−1ˆ0ˆΣ−1ˆ˜− −whereˆ = estimated alphas¯ = ()Σ= (0)Σ = ( 0) = sample length =numberoffactors = number of assets2= chi-squared distribution = F distributionDo not memorize! Understand, look up.ii. Basic idea: like the square of a t testˆ(ˆ)˜iii. If  are also normal then a refinementisvalidinsmallsamples. −  − h1+¯0ˆΣ−1¯i−1ˆ0ˆΣ−1ˆ˜− −“GRS Test.”202iv. Distribution of ˆ0(ˆ ˆ0)−1ˆ — if the true alphas are zero, how often should wesee this number be one, two, etc.?0.5 1 1.5 2 2.5 3 3.5 4 4.5500.10.20.30.40.50.60.70.80.91χ2 distributionsχ21χ22v. Fama and French “GRS” quote p. 57. The huge rejection because Σ is so small. −  − h1+¯0ˆΣ−1¯i−1ˆ0ˆΣ−1ˆ˜− −12.3 Cross-sectional regression1. Another idea. The main point of the model is across assets,i()iERSlopei()=0 (+)  =1 2Why not fitthisasacross sectional regression?2032. Two step procedure(a) TS (over time for each asset) to get ,= + +  =1 2 for each .(b) Run CS (across assets) to get .()=()+ +  =1 23. Estimates:(a)ˆ from TS.(b)ˆ slope coefficient in CS.(c) ˆ from error in CS: ˆ =1³P=1´−ˆˆ ˆ 6=  isnottheinterceptfromthetimeseries regression any more.4. Standard errors.(a) (ˆ)fromTS,OLSformulas.(b) (ˆ). You can’t use OLS form ulas.(c) Answer: With no intercept in CS,2(ˆ)=1h¡0¢−10Σ(0)−1³1+0Σ−1´+ Σi(d) (ˆ)(ˆ)=1³ − (0)−10´Σ³ − (0)−10´³1+0Σ−1´(e) See notes for the form ula with an intercept in CS.5. Testˆ0(ˆ ˆ0)−1ˆ˜2−−16. General tool: how to correct OLS standard errors for correlated errors = + ; (0)=Ωˆ =(0)−10ˆ =(0)−10( + )(ˆ)=! (unbiased)(ˆ)=h(0)−100(0)−1i=(0)−10Ω(0)−1If Ω = 2then we’re back to our friend(ˆ)=2¡0¢−120412.4 Fama - MacBeth procedure.1. Run TS to get betas.= + 0+  =1 2 for each .2. Run a cross sectional regression at each time period,=()+0+  =1 2 for each 3. Estimates of   are the averages across timeˆ =1X=1ˆ;ˆ=1X=1ˆ4. Standard errors use our friend 2(¯)=2() but applied to the monthly cs regressioncoefficients!2(ˆ)=1(ˆ)=12X=1³ˆ−ˆ´2(ˆ)=1(ˆ)=12X=1(ˆ− ˆ)(ˆ− ˆ)This one main point. These standard errors are easy to calculate.5. Testˆ0(ˆ ˆ0)−1ˆ˜2−16. Fact: if the  are constant over time, the estimates are identical to those of cross-sectionalregressions! Standard errors are close.7. Other applications of Fama-MacBeth: any time you have a big cross-section, which may becorrelated with each other.(a)+1=  +  ln +  ln + +1; () 6=0This is an obvious candidate, since the errors are correlated across stocks but not acrosstime. FF used FMB regressions in the second paper we read.(b)investment=  +  × Book/Market+  × profits+ (c) This is a big deal! Either FMB or “Cluster.”20512.5 Testing one model vs. another1. Example. FF3F.()=+ + + Drop size?()=+ + No, because the  and  change.2. Fallacy: see if =0.3. Example 2: CAPM w orks well for size portfolios, but shows up in FF model()=0+← works, higher with higher ()()=0++ ← works too, all =1and 04. Solution:(a) You can drop smb if the other factors price smb.= + + + We can drop smb from the three factor model if and only is zero. .(b)


View Full Document

Chicago Booth BUSF 35150 - Empirical methods overheads

Documents in this Course
CLONES

CLONES

8 pages

Load more
Download Empirical methods overheads
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view Empirical methods overheads and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view Empirical methods overheads 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?