PS 3 EF 5070: Financial Econometrics EF 5070: Financial Econometrics Problem Set 3 Due 5:00 pm, Nov 30th, 2020 Notes 1. Due Monday, 5:00pm, Nov 30th. 2. Please submit your problem set zip files which contains all related material into CANVAS by the deadline. Late submissions will not be accepted. 3. Hand in your problem set together with the i) R codes that you used to generate the results (print out your script file), ii) the associated R log file (print out your console window output), and iii) your written (typed) solution. 4. Each student needs to write his/her own solutions, even though discussions of the assignments between students are encouraged. 5. If not specifically specified, use 5% significance level (the associated critical value is 1.96 for standard normal distribution) to draw conclusions in this problem set. 6. For this problem set, you may use the following R packages: (See R demo codes provided in Chapter 4 from Canvas for details). library(’TSA’) library(’fGarch’) library(’parallel’) library(’rugarch’) 1. Consider the daily VIX index. VIX, calculated and published by the Chicago Board Options Exchange (CBOE), is widely used as a measure for market level uncertainty. (a) Please download the daily VIX index from January 1, 2006 to Nov 1, 2020 using the quantmod command in R. hint Page 1 PS 3 EF 5070: Financial Econometrics #To use a specific column of your dataset, say the 6th column in this question, and transform it into numeric format, consider the following command: vix<-as.numeric(VIX[,6]) (b) Use ARCH(q) model with the default Gaussian distribution as a baseline model to fit the VIX dataset. (1) Write down an equivalent AR representation for the ARCH(q) process. (2) Explain and show how the best order q can be determined. (3) Write down the fitted ARCH(q) and its equivalent AR representation. (c) Now, use GARCH(1,1) model with the default Gaussian distribution to fit the VIX data. (1) Write down an equivalent ARMA representation for the GARCH(m,s) process. (2) Write down the fitted GARCH(1,1) model.(3) Do you observe significant GARCH effect at 5% level? (d) Next, use GARCH(1,1) model with a student-t distribution to fit the VIX data. Write down the fitted model. Hint, consider the following R commands. model<-garchFit(~garch(m,s),data=,cond.dist=’std’,trace=F) (e) Fit the dataset with a ARMA(p,q)-GARCH(1,1) model. (1) Please explain and show how to choose the ARMA orders, (p,q). (2) Write down the fitted model. (3) Please briefly explain why we would prefer GARCH(1,1) over ARCH(q) when modeling the latent dynamic volatility process here. (f) Which model would you prefer to explain the evolutions of the market volatility, VIX dataset? Briefly explain. 2. Consider the daily returns of Pfizer stock from January 2, 2009 to Nov 1, 2020. Download the Starbucks data using the ’quantmod’ package in R. Using daily closing price to construct simple returns so as to form log returns. Multiple the log returns by 100 to obtain the percentage returns. Let rt be the percentage log returns. (a) Is the expected value of rt zero? Write down the null and alternative hypothesis and the test statistics. Write down your conclusion. Consider the following R command: t.test(rt) (b) Are there any serial correlations in rt and r 2 t ? Why Page 2 PS 3 EF 5070: Financial Econometrics (c) Fit a Gaussian ARMA(p,q)-GARCH(1,1) model to the rt series. Obtain the normal QQplot of the standardized residuals (hint: plot(model)), and write down the fitted model. Is the model adequate? Why? (d) Let zt = rt − r¯t, where r¯t = 1n ∑n i=1 rt is the sample mean of rt. Fit an IGARCH(1,1) model with a constant term to the at series zt. Write down the fitted model. (e) Let σt be the fitted volatility of the IGARCH(1,1) model. Define the standardized residuals as t = zt σt . Is there any serial correlation in the standardized residuals? Why? (Hint: consider the LB test). Consider the following R command: sresi=zt/model$volatility (f) Using the provided package (garchM.R), fit a GARCH-M model to rt. Write down the fitted mode. Do the mean evolutions of log returns statistically significantly depend on conditional volatility? Why? What is estimated level of risk premium? > source("garchM.R") > model=garchM(data) Page 3
欢迎咨询51作业君