Question 1. What are the pricing factors smb and hml and how have they been computed? (hint: read the Fama and French (1996) paper – its on iLearn) (2 marks)

Question 2. Open the excel data file in EViews (File/Open/Foreign data as workfile). a) Create a new variable (mkt_rf) for excess returns of the market over risk free, also known as daily market risk premium, i.e. mkt_rf = market – rf. (hint: If you do this correctly the first value of mkt_rf will be -0.710.) Next, create a new variable for excess returns on the SMALL_HiBM portfolio, i.e. SH_rf = SMALL_HiBM – rf. (hint: if you do this correctly the first value of SH_rf will be -0.531.) (1 mark) b) Provide a graph of rf and comment on its behaviour before and after the global financial crisis in 2008. (1 mark)

Question 3. Provide a graph and descriptive statistics for both the SH_rf and mkt_rf returns, and compare them. Highlight any important differences between them. (3 marks)

Question 4. Repeat the exercise from Question 3 for a different sample period: 3 January 2000 – 31 December 2007. You do not need to repeat the preliminary steps. Use the “sample” tab in Eviews to set the new sample period. Comment on the performance of the SMALL_HiBM portfolio relative to the market portfolio during this period. (2 marks)

For the following questions restore the sample period to the full period in EViews for the remainder of the assignment, unless a question specifically asks you to do otherwise.

Question 5. Estimate the following model for the full sample period:

0 1 2 3 __ t t t t t SH rf mkt rf smb hml      = + + + + (1)

Present the fitted equation showing coefficient estimates, standard errors and t-statistics (type the results in your assignment, do not simply copy and paste the EViews output). (4 marks)

Question 6. Is the estimate of 0 
significant at the 5% level? Explain your finding, do not just answer yes or no. How do you interpret this result? (2 marks)

Question 7. a) Explain whether we should expect the estimates of 12 , 
and 3 
to be positive or
negative. (2 marks) b) Are the signs of your actual estimates the same as what you expected? If not, how do they differ? (1 mark)

Question 8. a) Conduct a hypothesis test to determine whether 𝛽0, 𝛽2 and 𝛽3 are jointly significantly different to zero, i.e. test the following null hypothesis 𝐻0: 𝛽0=0 and 𝛽2=0 and 𝛽3= 0. Set out all of the steps for a formal hypothesis test and state the conclusion. Use a 5% significance level. (hint: in Eviews click View Coefficient diagnostics/Wald test) (2 marks) b) What does your test result imply in relation to the validity of the CAPM? Why? (1 mark)

Question 9. Conduct the basic diagnostic tests on the estimated model, i.e. autocorrelation (use 5 lags of residuals), heteroskedasticity (White’s test with no cross product), non-normality, misspecification of functional form (only one fitted term, quadratic). You do not need to write out all of the steps of the hypothesis tests and you may copy the EViews output of the tests into your
assignment. However you must clearly write out the null and alternative hypotheses in each case, and clearly state the conclusion of each test. Use a 5% significance level. (4 marks)

Part B Total number of marks: 25

Question 1. Conduct ADF and KPSS unit-root tests on the market series for the full sample period (conduct the tests in levels, not differences, with an intercept and no time trend, and use default values for the remaining settings). Be careful to properly state the null and alternative hypotheses for the two tests. You may copy in the relevant parts of the EViews output. Comment on your findings. (4 marks)

Question 2. Plot the ACF and PACF functions for market (include 12 lags). Comment on the magnitude and significance of the correlations. What optimal ARMA(p,q) model would you choose based on these graphs? Why? (4 marks)

Question 3. Select an optimal ARMA(p,q) model for the returns based on an information criterion (see below). Select from the set of models up to and including the largest model of ARMA(3,3). You may use the automatic procedure (set no transformation, max.-difference=0 and max SAR=0) or you may undertake this task manually. (hint: Refer to examples in the Week 7 tutorial) a) Present a single table of the criterion values for AIC, SBIC and HQ over all combinations of p and q. What is the preferred model on the basis of the AIC criterion? (2 marks) b) What is the preferred model on the basis of SBIC? (1 mark) c) Do both information criteria select the same model? Explain why the two criteria may select different models. (1 mark)

Question 4. a) Estimate the ARMA(1,1) model for the market series. Report the fitted equation and comment on the significance of the parameter estimates. (3 marks). b) Present the ACF and PACF graphs and statistics for the residuals (use 12 lags) and comment on them. (2 marks)

Question 5. a) Perform a test for fifth order ARCH effects in the estimated residuals of the model in Question 4. (hint: After you estimate the model for Question 4, click on View from the
Equation Window and select Residual Diagnostics and then Heteroscesdasticity tests. In the ‘Test type’ box, choose ARCH and the number of lags to include is 5). Write out the null and alternative hypotheses for the test. Explain your conclusion from the test. (3 marks) b) Estimate the ARMA(1,1)‐ARCH(5) model for the market series (hint: select Quick, Estimate Equation, under method select ARCH. In the mean equation box type market c ar(1) ma(1) and in the variance part, specify the order of ARCH as 5 and the order of GARCH as 0). Report the fitted equation. Do all of the parameter estimates satisfy the ARCH restrictions? (2 marks) c) Graph the conditional variance from part (b) and comment on its behaviour. Explain why a GARCH(1,1) specification might be more effective for modelling persistent volatility clustering than the ARCH(5) specification (you are not required to estimate the GARCH model). (3 marks)