Financial Econometrics

HT Week 1 Assignment Answers

February 2021

Exercise 4.5

Write the AR(1) Yt = φ0 + φ1Yt−1 + εt as an MA(∞) assuming |φ1 | < 1.

Yt = φ0 + φ1Yt−1 + εt
using Yt−1 = φ0 + φ1Yt−2 + εt−1
Yt = φ0 + φ1 (φ0 + φ1Yt−2 + εt−1 ) + εt
= φ0 + φ1 φ0 + φ12Yt−2 + φ1 εt−1 + εt
using Yt−2 = φ0 + φ1Yt−3 + εt−2
= φ0 + φ1 φ0 + φ12 (φ0 + φ1Yt−3 + εt−2 ) + φ1 εt−1 + εt
= φ0 + φ1 φ0 + φ12 φ0 + φ13Yt−3 + φ12 εt−2 + φ1 εt−1 + εt
= φ10 φ0 + φ11 φ0 + φ12 φ0 + φ13Yt−3 + φ12 εt−2 + φ11 εt−1 + φ10 εt−0
X∞ X∞
= i
φ1 φ0 + φ1j εt− j
i=0 j=0
∞
φ0 X
= + φ1j εt− j
1 − φ1
j=0

1
Exercise 4.12
Answer the following questions:

1. Under what conditions on the parameters and errors are the following processes covariance station-
ary?

(a) Yt = φ0 + εt
If εt is a white noise.
(b) Yt = φ0 + φ1Yt−1 + εt
If εt is a white noise and |φ1 | < 1 .
(c) Yt = φ0 + θ1 εt−1 + εt
If εt is a white noise.
(d) Yt = φ0 + φ1Yt−1 + φ2Yt−2 + εt
If εt is a white noise and φ1 and φ2 lie within a triangular region bound by (-2,-1),(0,1) and (2,-1) where
the coordinates are (φ1 , φ2 ). This can be easily observed using the quadratic formula.
(e) Yt = φ0 + φ2Yt−2 + εt
If εt is a white noise and |φ2 | < 1 is finite. Note that technicall we need to check the roots of z2 − φ2 but
√
these are just c = ± φ2 so that these are less than 1 in absolute value if |φ2 | < 1.
(f) Yt = φ0 + φ1Yt−1 + θ1 εt−1 + εt
If εt is a white noise and |φ1 | < 1.

2
Exercise 4.19
Suppose you observe the three sets of ACF/PACF in the figure below. What ARMA specification would you
expect in each case. Note: Dashed line indicates the 95% confidence interval for a test that the autocorrela-
tion or partial autocorrelation is 0.

1. The first appears consistent with an MA(2). The key features are that the ACF cuts off sharply after 2 lags
while the PACF appears to be more complex.

2. The second matches the shape that one would expect if the data was generated by an AR(1) with a positive
AR coefficient. The ACF decays exponentially while the PACF cuts off sharply after one lag.

3. The third appers are be consistent with a White Noise process. None of the ACF or PACF values are outside
of the CI, and so there appears to be no dependence in the data.

3
 Autocorrelation and Partial Autocorrelation function
ACF PACF
(a)
0.50 0.5

0.25
0.0
0.00

1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12

(b)

0.5 0.5

0.0 0.0

1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12

(c)
0.2 0.2

0.0 0.0

−0.2 −0.2
1 2 3 4 5 6 7 8 9 10 11 12 1 2 3 4 5 6 7 8 9 10 11 12

Figure 1: The ACF and PACF of three stochastic processes. Use these to answer the question .