Financial Econometrics Open Book Final Examination WS 2020/21
Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit
Financial Econometrics
Open Book Final Examination WS 2020/21
Problem 1: Univariate Volatility Model
Consider the following conditional variance model for a daily log-return series s芒 of a financial asset:
s芒 = c + e芒
e芒 = 3芒尸7芒(2)
7芒(2) = u + ae芒(2)-〇 + ye芒(2)-〇 1(et − 1 <0} ,
where u > 0, a > 0, a + y > 0. 1(et − 1 <0} denotes the indicator function, which is equal to 1, if e芒-〇 < 0 and 0 otherwise.
a) What is the idea behind the model specified in (3)? How is it different from an ARCH(1) model? 2 P
b) Describe shortly the concept of the leverage effect in the context of condi- tional volatility modeling. 1 P
c) Is the model defined in (3) able to capture the leverage effect? Provide both a mathematical and an intuitive explanation. For the mathematical explanation derive cou(7芒(2)′〇, e芒 N~芒-〇 ), where ~芒 is the information set avail- able up to time o. Under which parameter restrictions is cou(7芒(2)′〇, e芒 N~芒-〇 )
negative? 6 P
Hint: Note that o 33 o(3)d3 = _尸 and +o 33 o(3)d3 = 尸 with o(.)
be the pdf of the standard normal distribution. Moreover, you can use the fact that cou[e芒(2) , 1(et<0}N~芒-〇 ] = 0 .
d) Derive the conditional variance forecast over the next two periods E[7芒:(2)芒′2N~芒].6 P
Problem 2: Best Volatility Model and Value-at-Risk
Assume you want to invest 10 000e into a long position of a financial asset with a log-return s芒 . Table 2.1 summarises the outputs from estimating various GARCH models for s芒 such that:
s芒 = c + e芒
e芒 = 3芒尸7芒(2)
3芒 r, (0, 1)
where 7芒(2) is modelled by ARCH(1), GARCH(1,1), TGARCH(1,1) and EGARCH(1,1).
Table 2.1: Estimated parameters, p-value of the ARCH LM test on standardized residuals and information criteria.
|
ARCH |
GARCH |
TGARCH |
EGARCH |
cˆ |
_0.02*** |
_0.03*** |
_0.03*** |
_0.03*** |
|
0.65*** |
0.07*** |
0.09*** |
0.03*** |
|
0.29*** |
0.12*** |
0.12*** |
0.24*** |
8ˆ |
|
0.87*** |
0.86*** |
0.97*** |
|
|
|
0.01 |
_0.01 |
AIC |
3.92 |
3.80 |
3.81 |
3.82 |
BIC |
3.93 |
3.82 |
3.83 |
3.84 |
p-value ARCH LM |
0.03 |
0.71 |
0.71 |
0.67 |
*** indicates significance of the parameter at the 1% level.
no star indicates no significance at 1,5 or 10% level.
a) Based on the results of Table 2.1, which model is the best choice for 7芒(2)? Explain your answer based on the results of the table. 3 P
b) Please write down the equation for 7芒(2) of the model you choose to be best at point a). Please define correctly the parameters and their restrictions. 2 P
c) Compute the Value-at-Risk at the p = 1% level for the next two periods VaR(p,2) for your initial investment of 10 000e and by using the best model chosen at point a). You have that eˆ芒 = _0.015 and 芒 = 0.05.
Hint: The 1% quantile of the standard normal distribution is -2.3263. 10 P
Problem 3: Transition-GARCH Processes
Consider the following process:
y芒 = c + e芒
e芒 = 尸7芒(2)3芒
7芒(2) = w + ae芒(2)-〇G(_e芒-〇 ) + ye芒(2)-〇G(e芒-〇 ) + 87芒(2)-〇
where Z芒 r(0, 1) and G(e芒 ) is a probability transition function of e芒 satisfying G(_e芒 ) = 1 _ G(e芒 ).
Assume that u, a, 8, y satisfy the necessary conditions for assuring that 7芒(2) is positive and covariance-stationary.
a) Show that E[G(e芒 )N~芒-〇 ] = E[G(_e芒 )N~芒-〇 ] = 1/2. 3 P Hint: Start writing E[G(e芒 )N~芒-〇 ] by additionally conditioning on positive and negative values of e芒 and account for the fact that e芒 N~芒-〇 ~r(0, 7芒(2)) .
b) Compare the Transition-GARCH(1,1) model with a simple GARCH(1,1) model. What are the similarities and what are the differences between the two approaches? 2 P
c) Show that, conditional on the information set ~芒-〇 , e芒(2) and G(e芒 ) are un- correlated. 2 P
d) Assume now that the transition function is given by:
G(e芒 ) = 1(et<0} (4)
where 1(.} denotes the indicator function.
1. Show that the transition function G(.) from Equation (4) satisfies the condition: G(_e芒 ) = 1 _ G(e芒 ). 1 P
2. Which inequality relation between a and y is reasonable under the assumption of a leverage effect. 2 P
3. Derive the News Impact Curve of the Transition-GARCH model with the transition function given in Equation (4). 1 P
e) Now let the transition function be the logistic function
G(e芒 ) = 1 + ezp(1)(_Ae芒) , A > 0 (5)
1. Show that the transition function G(.) from Equation (5) satisfies the condition: G(_e芒 ) = 1 _ G(e芒 ). 1 P
2. Calculate the s-step ahead forecast E[7芒(2)′5 N~芒] of the conditional vari- ance for t = 1, 2 for the transition function given in Equation (5). 2 P
3. Derive the News Impact Curve of the Transition-GARCH model with the transition function given in Equation (5). 1 P
2022-03-10