Financial Econometrics - Homework 3

Nguyen Minh Quan - MAMAIU19036

Lam Hue Dung - MAMAIU18060

Le Nguyen Dang Khoa - MAMAIU19008

May 9, 2022

Problem 3. Consider the following three models that a researcher suggests might be a reasonable
model of stock market prices.
yt = yt−1 + ut ;
yt = 0.5yt−1 + ut ;
yt = 0.8ut−1 + ut .

(a) What classes of models are these examples of?

(b) What would the autocorrelation (ACF/PACF) function for each of these processes look like?
(do not need to calculate, simply consider what shape it might have given the class of model
from which it is drawn)

(c) Which model is more likely to represent stock market prices from a theoretical perspective, and
why? If any of the three models truly represented the way stock market prices move, which
could potentially be used to make money by forecasting future values of the series?

Answer. blank

(a) The first two models are AR(1), while the last is an MA(1).

(b) Since the theoretical ACF of an MA(q) process will be 0 after q lags, the ACF of the MA(1)
will be 0 at all lags after 1. For an AR(p) process, the ACF dies away gradually. It will die
away fairly quickly for model 2, with each successive autocorrelation coefficient taking on a value
equal to half that of the previous lag. For the first model, however, the ACF will never die away,
and in theory will always take on a value of 1 at all lags.
The PACF for the first two models would have a large positive spike at lag 1, and no statistically
significant spike at other lags. The PACF for model 3 will decline geometrically.

(c) The first model is more likely to represent stock prices in practice. The discounted dividend
model of share prices states that the current value of a share will be simply the discounted
sum of all expected future dividends. If we assume that investors form their expectations about
dividend payments rationally, then the current share price should embody all information that
is known about the future of dividend payments, and hence today’s price should only differ from
yesterday’s by the amount of unexpected news which influences dividend payments. Therefore,
stock prices should follow a random walk.
If the stock market really followed the process described by models 2 or 3, then we could
potentially make useful forecasts of the series using our model. In the latter case of the MA(1),
we could only make one-step ahead forecasts since the ’memory’ of the model is only that length.
In the case of model 2, we could potentially make a lot of money by forming multiple step ahead

1
 forecasts and trading on the basis of these. However after 1 period, it is likely that other
investors would spot this potential opportunity and hence these models would no longer be a
useful description of the data.

Problem 5. You obtain the following estimates for an AR(2) model of some returns data:

yt = 0.803yt−1 + 0.682yt−2 + ut

where ut is a white noise error process. By examining the characteristic equation, check the estimated
model for stationarity.

Answer. Rewrite the model as

yt (1 − 0.803L − 0.682L2 ) = ut
and note that the lag polynomial 1 − 0.803L − 0.682L2 has roots 0.758 and 1.934 which are not ALL
greater than 1, hence the estimated model is not stationary.

Problem 9. You obtain the following sample autocorrelations and partial autocorrelations for a
sample of 100 observations from actual data:
Lag 1 2 3 4 5 6 7 8
ACF 0.420 0.104 0.032 −0.206 −0.138 0.042 −0.018 0.074
PACF 0.632 0.381 0.268 0.199 0.205 0.101 0.096 0.082

(a) Can you identify the most appropriate time series process for this data?

(b) Use the Ljung−Box Q∗ test to determine whether the first three autocorrelation coefficients
taken together are jointly significantly different from zero.

Answer. blank
√
(a) We class an ACF/PACF coefficient as significant if it exceeds ±1.96/ T = ±0.196. Hence the
sample ACF at lags 1 and 4 are significant, and the sample PACF at lags 1−5 are all significant.
This clearly looks like the data are consistent with an MA(1) since all but the first ACF are not
significant (the significant lag 4 ACF is a typical wrinkle that one might expect with real data
and should probably be ignored), and the PACF has a slowly declining structure.

(b) The test statistics is given by

m
X τk2
Q∗ = T (T + 2)
T −k
k=1
0.4202 0.1042 0.0322
 
= 100 · 102 · + + = 19.41 > 11.3 = χ23,0.01
100 − 1 100 − 2 100 − 3

hence we would reject the null hypothesis, i.e. the first three autocorrelation coefficients are
jointly not significantly different from zero.