Scribd
Upload
31 views

Homework 1

This document outlines 6 problems for an econometrics homework assignment. Problem 1 involves modeling a time series using a Markov-switching model. Problem 2 examines the effects of temporal aggregation on ARMA processes. Problem 3 defines and explores properties of affine models of order one. Problem 4 involves simulating AR and fractional white noise processes. Problem 5 examines descriptive statistics of macroeconomic time series data. Problem 6 tasks finding the best ARMA model to fit VIX data.

Uploaded by

wise1smile
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
31 views

Homework 1

This document outlines 6 problems for an econometrics homework assignment. Problem 1 involves modeling a time series using a Markov-switching model. Problem 2 examines the effects of temporal aggregation on ARMA processes. Problem 3 defines and explores properties of affine models of order one. Problem 4 involves simulating AR and fractional white noise processes. Problem 5 examines descriptive statistics of macroeconomic time series data. Problem 6 tasks finding the best ARMA model to fit VIX data.

Uploaded by

wise1smile
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

MASTER 2, Econometrics II

Homework # 1
Due date: February 14, 2020 at 17.00 (hard copy)

You can do the work in a team of two persons.

1. Markov-Switching model.

Let St be a two-state Markov chain, St ∈ {0, 1}. Define yt as

yt = (µ1 St + µ0 (1 − St )) + (σ1 St + σ0 (1 − St ))ut

where ut is assumed to be i.i.d., independent with St , and follows a


N (0, 1), with µ1 6= µ2 and σ1 6= σ2 .

(a) Prove that yt is second order stationary.


(b) Compute the mean, variance, skewness and kurtosis of yt .
(c) Compute the covariance structure of yt . Comment the results.
(d) Prove that yt2 is second order stationary.
(e) Compute the covariance structure of yt2 . Comment the results.
(f) Compute the spectral density functions of St , yt and yt2 .
(g) What happens if the two processes {St } and {ut } are dependent?
You can take an example.

2. Temporal aggregation of time series. Let {xt , t ∈ Z} be an


ARMA(1,1) process.

(a) Assume that one observes the process {y2t , t ∈ Z} with y2t = x2t .
Characterize the ARMA structure of the process {y2t , t ∈ Z}
(including the coefficients of the ARMA process as functions of
those of {xt }).
(b) Assume that one observes the process {z2t , t ∈ Z} with z2t =
x2t + x2t−1 . Characterize the ARMA structure of the process
{z2t , t ∈ Z}.

1
3. Affine Models. An affine model of order one is defined by

φt (u) = E[exp(iuxt+1 ) | xτ , τ ≤ t] = exp(a(u) + b(u)xt ),

where u is a real number, i2 = −1 and a(·) and b(·) are some complex
functions.

(a) What is the name of φt (u)? Give the properties of this function
without any proof.

(b) What is the name of the function ψt (·) defined by ψt (u) =


log(φt (u))? Give the properties of this function without any
proof.

(c) What are the connection between the moments and the cumulants
of a random variable?

(d) Prove that an AR(1) process with an i.i.d. innovation process is


an affine process of order one.

(e) Is an affine model of order one necessary an AR(1) process with


an i.i.d. innovation process? If no, provide an example.

(f) Compute recursively (when k increases) the functions


φt,k (u1 , u2 , · · · , uk ) given by
   
k
X
φt,k (u1 , u2 , · · · , uk ) = E exp i uj xt+j  | xτ , τ ≤ t .
j=1

(g) What can you do with φt,k (u1 , u2 , · · · , uk )?

4. Simulations.

(a) Simulate an AR(1) process with stable errors, including a Cauchy


distribution. Take the auto-regressive coefficient ρ ∈ {0, .5, .9},
the index of the stable distribution α ∈ {.5, 1, 1.5, 1.9}, the
sample size T ∈ {100, 250, 500, 1000}, and for each design do
500 replications. Compute the descriptive statistics of sample
moments and ACFs. Comment the results.

2
(b) Simulate a stationary fractional white noise with the long-
memory parameter d ∈ {.1, .25, .45}, the sample size T ∈
{100, 250, 500, 1000}, and for each design do 500 replications.
Compute the descriptive statistics of sample moments and ACFs.
Comment the results.

5. Descriptive statistics of macroeconomic and financial data.


Find the following data of a developed country: GDP growth, inflation,
short term interest rate, stock market index, and the exchange rate
against the currency of another developed country.

(a) Provide the following sample statistics: mean, variance, skewness,


kurtosis, ACF at several lags of the three series. Comment the
results.

(b) Provide graphs of the rolling means, variances, skewness, kurtosis.


Consider several windows and provide the best graphs (the
window should be the same for a given data). Comment the
results. A rolling window is a moving average sum of the current
and past observations with equal weights whose sum equals one.

(c) Compare with exponential smoothing.

6. ARMA. From the website of FRED (https://fred.stlouisfed.org/),


download the VIX data. Find the best ARMA model that fits this
variable.

You might also like