Home Page > > Details

Advanced Econometrics 2020 - Homework 2

Deadline: 23:59:59, Friday November 27, 2020

Instructions1

:

• Form groups of three yourself.

• As a solution, provide one Jupyter Notebook with R source-code. Code should be properly

commented, interpretations of results as well as theoretical derivations2

should be written

in markdown cells. This is the only file you need to send.

• Use “set.seed()“ function, so I can replicate your results.

• Be concise (no lengthy essays please). Although, be sure to include all important things as

I cannot second-guess your work.

• The problem set is due on 27th November. Late submission automatically means

0 points.

• If you have any questions concerning the homework, do contact me by mail. Do it rather

sooner than later.

1The contact person for this homework is Matěj Nevrla, the same mail as for submission of

homeworks.

2

If you prefer not to write formulas in LATEX, you can send PDF with your derivations and

interpretations in additional file and R code in Jupyter Notebook.

1

Problem 1 Bootstrap

(1.5 points)

Consider dataset city from the package boot. Data consists of random sample

of size n = 10 from the population of big cities in the US and each observation

corresponds to the number of inhabitants of a given city (in thousands) in year 1920

(variable u) and number of inhabitants of a given city in year 1930 (variable x).

Let (U, X) be a random vector corresponding to the number of inhabitants in the

years 1920 and 1930, and (ui

, xi), i = 1, . . . , 10, be our random sample.

Consider the following ratio

R ≡

E(X)

E(U)

.

a) Draw the dependence between xi and ui using the dataset.

b) Estimate the value of R from the random sample.

c) Compute the bootstrap standard error and the bootstrap bias of the estimate of

R. Use both ”brute force” bootstrapping and function boot from the respective

package.

d) Compute the 95% confidence intervals using both percentile and normal approximation

method. Argue whether the normal approximation is valid in this case.

e) Test (without refinement) H0 that R = 1 against the alternative HA : R ̸= 1.

Calculate bootstrap t-statistic and compare it with critical values of the standard

normal distribution. Use also the percentile method for testing H0 and compare

the results.

2

Problem 2 Endogeneity

(2 points)

Let us follow the idea of the first exercise from Seminar 6 but for now we create

another artificial dataset containing 300 observations (note that although variance

of RVs is specified below, R commands often require to specify standard deviation

instead):

z1 ∼ N(2, 3

2

), z2 ∼ N(2, 1.5

2

), z3 ∼ N(0, 2

2

), z4 ∼ N(1.8, 2.5

2

)

ϵ1 ∼ N(0, 1.5

2

), ϵ2 ∼ N(0, 1.5

2

), ϵ3 ∼ N(0, 1.5

2

)

x1 = 0.3z1 − z2 + 0.9z4 + 0.75ϵ1

x2 = 0.75z2 + 0.75ϵ2

x3 ∼ N(0, 1)

y = 1 + 2.5x1 − x2 + 0.45x3 + ϵ3

On the dataset, we should estimate the following model:

y = const + β1x1 + β3x3 + ϵ

a) Discuss the nature of the endogeneity problem in the system above. You might

check important correlations and you should explain the difference between x1

and x3. Do you expect to observe any bias within the OLS estimation? Explain

why.

b) Estimate the model by OLS and interpret.

c) The data set includes some potential IV candidates: z1; z2; z3; z4. What assumptions

need to be satisfied in order to have a ‘good’ instrument? Which of these

candidates seem to be ‘good’ instruments and why? Test their relevancy statistically.

Is there any invalid, irrelevant, or weak instrument?

d) Based on section d), choose the best instrument and run the IV regression. Run

also 2SLS regression using all ‘good’ instruments. Compare coefficient estimates

and standard errors

e) Finally, test for the endogeneity using the Hausman test. Report and interpret

the results.

f) Using extended dataset (simulate more data from the data generating process),

show that OLS is not consistent estimator of β1 and β3. Show that 2SLS provides

consistent results.

3

Problem 3 GMM

(1.5 points)

In the dataset hw_data.csv, you have a time series which comes from Moving Average

process with q lags - MA(q) process. You will need libraries gmm and tseries

to answer the following questions. Note that if you do not answer correctly point a),

all consecutive questions will be wrong, hence no points can be earned.

a) How would you identify the lag q? What is the lag?

b) Derive the moment conditions function for the process you identified previously

and write a corresponding function in R. Use more moment conditions than is

the number of coefficients that you want to estimate.

c) Estimate the model using gmm and both identity and optimal weighting matrix.

Provide the output and interpret the coefficient significance and the J-test statistics.

4

Problem 4 CAPM beta of Apple Inc.

(3 points)

You are going to estimate CAPM betas for Apple Inc. on two different time periods

and compare the results. Inference on the estimated parameters will be performed

using bootstrap.

CAPM beta for company i can be estimated by regressing its returns on market

returns using simple OLS

is risk-free interest rate, all at time t.

1. Download adjusted daily prices of Apple Inc. (ticker 'aapl') for year 2008

and 2017, separately. Download daily data for Nasdaq Composite Index (ticker

'^ixic'), which will be used as a proxy variable for market. Chicago Board

Options Exchange (CBOE) 10y interest rate T-note (ticker ’^tnx’) will be used

as a risk-free rate.

2. Compute log-returns for all series.

3. Draw histogram and kernel density approximation of Apple log returns. Compare

it with normal distribution with the same mean and standard deviation.

Does samples considerably deviate from normality? Which noticeable features

do you observe in the data? Perform this whole analysis on both time periods

separately.

4. Compute CAPM beta for Apple using simple OLS on each time period separately.

Compare the obtained results.

5. Use the nonparametric bootstrap to compute bootstrap standard error of CAPM

beta estimate based on 1000 bootstrap replications and bootstrap sample size

equal to the size of the original sample. Use boot command. Perform this

whole analysis on both time periods separately.

5

6. Graphically compare histogram and kernel density approximation (use gaussian

kernel only) of the boot-based set of 1000 bootstrapped CAPM betas with normal

distribution. Compute 95% percent confidence intervals for CAPM beta

for both time periods using percentile method and method based on normal

approximation. Which method is in this case probably more suitable. Based

on the estimated confidence intervals, do the estimates of CAPM beta significantly

differ over these 2 periods?

7. Another approach how to use bootstrap is instead of resampling the observations

is to resample the estimated residuals. Estimate CAPM beta by OLS,

save the estimated residuals, and perform bootstrap by resampling the residuals

and generating new values of dependent variables as

are resampled from empirical distribution function. In each bootstrap

replication, calculate values of dependent variable for all values of independent

variable. The rest of the procedure is the same as in the case of resampling the

observations. Perform this analysis on one time period of your choice. Compare

the results obtained from this approach and from the traditional approach from

above. Why this method may not be the optimal in this case?

Contact Us(Ghostwriter Service)

- QQ：99515681
- WeChat：codinghelp2
- Email：99515681@qq.com
- Work Time：8:00-23:00

- Cs2461-10 Programminghelp With ,Ghostw... 2021-03-02
- Ghostwriter Program Programming,Help W... 2021-03-02
- Programming Coursehelp With ,Ghostwrit... 2021-03-02
- Ghostwriter Csc1-Ua Programming,Help W... 2021-03-02
- Help With Program Programming,Ghostwri... 2021-03-02
- Ghostwriter Data Programming,Help With... 2021-03-02
- Cse 13S Programminghelp With ,Ghostwri... 2021-03-02
- Mat136h5 Programminghelp With ,C/C++ P... 2021-03-01
- Ghostwriter Ee425x Programming,Help Wi... 2021-03-01
- Cscc11 Programming Coursehelp With ,Gh... 2021-03-01
- Ghostwriter Program Programming,Python... 2021-03-01
- Help With R Programming|Help With Data... 2021-03-01
- Data Structuresghostwriter ,Help With ... 2021-03-01
- Help With Data Programming,C++，Python... 2021-03-01
- Ghostwriter Aps 105 Programming,C/C++ ... 2021-03-01
- Fre6831 Computational Finance 2021-02-28
- Sta141b Assignment 5 Interactive Visu... 2021-02-28
- Eecs2011a-F20 2021-02-28
- Comp-251 Final Asssessment 2021-02-28
- Ghostwriter Cs1027 Course Programming,... 2021-02-28

Contact Us - Email：99515681@qq.com WeChat：codinghelp

© 2014 www.asgnhelp.com

Programming Assignment Help！