Time Series: Assignment 3

Hang Nguyen

March 19, 2026

1. Multiple-choice questions

1. Consider the AR(1) process yt = a1 yt−1 + εt with |a1 | < 1. The unconditional mean
is:

(a) 0
a1
(b) 1−a1

1
(c) 1−a1

(d) Undefined

2. Which condition guarantees stationarity of an AR(2) process?

(a) a1 + a2 < 1

(b) All roots of the characteristic equation lie outside the unit circle

(c) |a1 | < 1

(d) σ 2 < 1

3. For an MA(2) process, the ACF:

(a) Decays geometrically

1
 (b) Cuts off after lag 2

(c) Is always zero

(d) Diverges

4. In a random walk process yt = yt−1 + εt , the variance of yt :

(a) Is constant

(b) Grows over time

(c) Is zero

(d) Depends only on y0

5. Which model captures leverage effects in volatility?

(a) ARCH

(b) GARCH

(c) EGARCH

(d) ARMA

6. For a GARCH(1,1) model, covariance stationarity requires:

(a) α1 > 1

(b) β1 > 1

(c) α1 + β1 < 1

(d) α0 = 0

7. The PACF of an AR(p) process:

(a) Cuts off after lag p

(b) Decays slowly

(c) Is always zero

2
 (d) Explodes

8. In a VAR(1) model, stability requires:

(a) Determinant of A1 = 0

(b) Eigenvalues of A1 lie inside unit circle

(c) Eigenvalues of A1 lie outside unit circle

(d) Trace of A1 = 1

9. Impulse Response Functions measure:

(a) Long-run variance

(b) Response of variables to shocks over time

(c) Mean of the process

(d) Stationarity

10. Forecast Error Variance Decomposition tells us:

(a) Mean forecast error

(b) Contribution of each shock to forecast error variance

(c) Stationarity condition

(d) Lag order selection

11. Historical decomposition is used to:

(a) Estimate parameters

(b) Decompose observed series into contributions of shocks

(c) Forecast future values

(d) Test stationarity

12. In Cholesky decomposition, ordering of variables:

3
 (a) Does not matter

(b) Affects IRF and FEVD

(c) Affects only mean

(d) Affects only variance

13. The Augmented Dickey–Fuller test is used to:

(a) Test for ARCH effects

(b) Test for unit roots

(c) Estimate ARMA models

(d) Estimate VAR models

14. If a series is difference-stationary, then:

(a) It is stationary in levels

(b) It becomes stationary after differencing

(c) It is always explosive

(d) It has constant variance

15. In an ARIMA(p, d, q) model, d represents:

(a) Number of AR terms

(b) Number of differences

(c) Number of MA terms

(d) Number of observations

16. Which of the following indicates overfitting?

(a) Low AIC

(b) High number of parameters with little improvement in fit

4
 (c) White noise residuals

(d) Stationarity

17. In ARCH models, large shocks tend to:

(a) Reduce variance

(b) Increase future variance

(c) Have no effect

(d) Make the series stationary

18. A unit root process implies:

(a) Mean reversion

(b) Permanent shocks

(c) Constant variance

(d) White noise

19. The characteristic equation of an AR(1) process is:

(a) r − a1 = 0

(b) 1 − a1 r = 0

(c) r2 − a1 = 0

(d) 1 − r = 0

20. Ljung–Box test is used to detect:

(a) Unit root

(b) Serial correlation

(c) Heteroskedasticity

(d) Structural break

5
2. Exercises

1. Given an initial condition for y0 , find and interpret the forecast function for each of
the following models:

1. yt = yt−1 + εt + 0.5εt−1

2. yt = 1.1yt−1 + εt

3. yt = yt−1 + 1 + εt

4. yt = yt−1 + t + εt

5. yt = µt + ηt + 0.5ηt−1 , where µt = µt−1 + εt

6. yt = µt + ηt + 0.5ηt−1 , where µt = 0.5 + µt−1 + εt

How can you make the models of parts 2 and 4 stationary? Does model 5 have an
ARIMA(p, 1, q) representation?

2. Consider the ARCH(2) process Eε2t = α0 + α1 ε2t−1 + α2 ε2t−2 .

(a) Suppose that yt = a0 + a1 yt−1 + εt . Find the conditional and unconditional

(b) Suppose that {yt } is an ARCH-M process such that the level of yt is positively
related to its own conditional variance. For simplicity, let

yt = α0 + α1 ε2t−1 + α2 ε2t−2 + εt .

Trace out the impulse response function of {yt } to a {εt } shock. You may
assume that the system has been in long-run equilibrium (ε−2 = ε−1 = 0) but
now ε1 = 1. Thus, the issue is to find the values of y1 , y2 , y3 , and y4 given that
ε2 = ε3 = · · · = 0.

(c) Use your answer to part B to explain the following result. A student estimated
{yt } as an MA(2) process and found the residuals to be white noise. A second

6
 student estimated the same series as the ARCH-M process

yt = α0 + α1 ε2t−1 + α2 ε2t−2 + εt .

Why might both estimates appear reasonable? How would you decide which is
the better model?

(d) In general, explain why an ARCH-M model might appear to be a moving average
process.

3. Given the initial condition y0 , find the general solution and forecast function for the
following variants of the trend plus irregular model:

(a) yt = µt + vt , where µt = µt−1 + εt , vt = (1 + β1 L)ηt , E(εt ηt ) = 0.

(b) yt = µt + vt , where µt = µt−1 + εt , vt = (1 + β1 L)ηt , and the correlation

(c) Find the ARIMA representation of each model.