2.7 Measuring Returns, Volatility, and Correlation
Question 1
The GARCH \text{GARCH} GARCH model is useful for simulating asset returns. Which of the following statements about this model is FALSE?
A. The Exponentially Weighted Moving Average (
EWMA
\text{EWMA}
EWMA) approach of RiskMetrics is a particular case of a
GARCH
\text{GARCH}
GARCH process.
B. The
GARCH
\text{GARCH}
GARCH allows for time-varying volatility.
C. The
GARCH
\text{GARCH}
GARCH can produce fat tails in the return distribution.
D. The
GARCH
\text{GARCH}
GARCH imposes a positive conditional mean return.
Answer: D
The
GARCH
\text{GARCH}
GARCH model allows for time-varying volatility by describing the conditional variance as a function of the previous period’s volatility and the most recent variance estimate:
σ t 2 = α 0 + α 1 r t − 1 2 + β σ t − 1 2 \sigma^2_t=\alpha_0+\alpha_1r^2_{t-1}+\beta\sigma^2_{t-1} σt2=α0+α1rt−12+βσt−12, where
- α 0 = γ V L \alpha_0=\gamma V_L α0=γVL,
- V L = α 0 1 − α 1 − β V_L=\cfrac{\alpha_0}{1-\alpha_1-\beta} VL=1−α1−βα0,
- α 1 + β + γ = 1 \alpha_1+\beta+\gamma=1 α1+β+γ=1,
- α + β < 1 \alpha+\beta<1 α+β<1
It is useful in simulating leptokurtic return distributions with fat tails.
The EWMA \text{EWMA} EWMA is a special case of the GARCH \text{GARCH} GARCH model with γ + 1 \gamma+1 γ+1, α 1 = 1 − γ \alpha_1=1-\gamma α1=1−γ, β = γ \beta=\gamma β=γ.
The model does not impose the requirement of a positive conditional mean return.
Question 2
Which of the following GARCH \text{GARCH} GARCH models will take the shortest time to revert to its mean?
A.
h
t
=
0.05
+
0.03
r
t
−
1
2
+
0.96
h
t
−
1
h_t= 0.05 +0.03r_{t-1}^2 + 0.96h_{t-1}
ht=0.05+0.03rt−12+0.96ht−1
B.
h
t
=
0.03
+
0.02
r
t
−
1
2
+
0.95
h
t
−
1
h_t = 0.03 + 0.02r_{t-1}^2 +0.95h_{t-1}
ht=0.03+0.02rt−12+0.95ht−1
C.
h
t
=
0.02
+
0.01
r
t
−
1
2
+
0.97
h
t
−
1
h_t = 0.02 + 0.01r_{t-1}^2 + 0.97h_{t-1}
ht=0.02+0.01rt−12+0.97ht−1
D.
h
t
=
0.01
+
0.01
r
t
−
1
2
+
0.98
h
t
−
1
h_t = 0.01+ 0.01r_{t-1}^2 + 0.98h_{t-1}
ht=0.01+0.01rt−12+0.98ht−1
Answer: B
The model that will take the shortest time to revert to its mean is the model with the lowest persistence defined by
α
+
β
\alpha+\beta
α+β. So B is the right answer with
α
+
β
=
0.97
\alpha+\beta= 0.97
α+β=0.97.
Question 3
The current estimate of daily volatility is 1.5 % 1.5\% 1.5%. The closing price of an asset yesterday was USD 30.00 30.00 30.00. The closing price of the asset today is USD 30.50 30.50 30.50. Using the EWMA \text{EWMA} EWMA (Exponentially Weighted Moving Average) model (with λ = 0.94 \lambda = 0.94 λ=0.94), the updated estimate of volatility is:
A.
1.5096
%
1.5096\%
1.5096%
B.
1.5085
%
1.5085\%
1.5085%
C.
1.5092
%
1.5092\%
1.5092%
D.
1.5083
%
1.5083\%
1.5083%
Answer: A
h
t
=
λ
h
t
−
1
+
(
1
−
λ
)
r
t
−
1
2
h_t=\lambda h_{t-1}+(1-\lambda)r_{t-1}^2
ht=λht−1+(1−λ)rt−12
σ = 0.94 × 0.001 5 2 + ( 1 − 0.94 ) [ In ( 30.5 30.0 ) ] 2 = 1.5096 % \sigma =\sqrt{0.94\times0.0015^2 +(1-0.94)\left[\text{In}(\cfrac{30.5}{30.0})\right]^2} =1.5096\% σ=0.94×0.00152+(1−0.94)[In(30.030.5)]2=1.5096%
Question 4
Given λ \lambda λ of 0.94 0.94 0.94, under an infinite series, what is the weight assigned to the seventh prior daily squared return?
A.
4.68
%
4.68\%
4.68%
B.
4.40
%
4.40\%
4.40%
C.
4.14
%
4.14\%
4.14%
D.
3.89
%
3.89\%
3.89%
Answer: C
Weight =
0.94
×
(
1
−
0.94
)
=
4.14
%
0.94\times(1-0.94) = 4.14\%
0.94×(1−0.94)=4.14%
Question 5
PE2018Q45 / PE2019Q45
A risk analyst is estimating the variance of stock returns on day
n
n
n, given by
σ
n
2
\sigma_n^2
σn2, using the equation:
σ
n
2
=
γ
V
L
+
α
μ
n
−
1
2
+
β
σ
n
−
1
2
\sigma^2_n=\gamma V_L +\alpha\mu^2_{n-1}+\beta\sigma^2_{n-1}
σn2=γVL+αμn−12+βσn−12, where
μ
n
−
1
\mu_{n-1}
μn−1 and
σ
n
−
1
\sigma_{n-1}
σn−1 represent the return and volatility on day
n
−
1
n-1
n−1, respectively. If the values of
α
\alpha
α and
β
\beta
β are as indicated below, which combination of values indicates that the variance follows a stable
GARCH
(
1
,
1
)
\text{GARCH}(1,1)
GARCH(1,1) process?
A.
α
=
0.082427
\alpha = 0.082427
α=0.082427 and
β
=
0.925573
\beta = 0.925573
β=0.925573
B.
α
=
0.084427
\alpha = 0.084427
α=0.084427 and
β
=
0.905573
\beta = 0.905573
β=0.905573
C.
α
=
0.085427
\alpha = 0.085427
α=0.085427 and
β
=
0.925573
\beta = 0.925573
β=0.925573
D.
α
=
0.090927
\alpha = 0.090927
α=0.090927 and
β
=
0.925573
\beta = 0.925573
β=0.925573
Answer:B
Learning Objective: Calculate volatility using the
GARCH
(
1
,
1
)
\text{GARCH}(1,1)
GARCH(1,1) model
For a GARCH ( 1 , 1 ) \text{GARCH}(1,1) GARCH(1,1) process to be stable, the sum of parameters α \alpha α and β \beta β need to be below 1 1 1.
Question 6
PE2018Q70
Which of the following four statements on models for estimating volatility is INCORRECT?
A. In the exponentially weighted moving average (
EWMA
\text{EWMA}
EWMA) model, some positive weight is assigned to the long-run average variance.
B. In the
EWMA
\text{EWMA}
EWMA model, the weights assigned to observations decrease exponentially as the observations become older.
C. In the
GARCH
(
1
,
1
)
\text{GARCH}(1,1)
GARCH(1,1) model, a positive weight is estimated for the long-run average variance.
D. In the
GARCH
(
1
,
1
)
\text{GARCH}(1,1)
GARCH(1,1) model, the weights estimated for observations decrease exponentially as the observations become older.
Answer: A
Learning Objective: Explain the weights in the
EWMA
\text{EWMA}
EWMA and
GARCH
(
1
,
1
)
\text{GARCH}(1,1)
GARCH(1,1) models.
The EWMA \text{EWMA} EWMA model does not involve the long-run average variance in updating volatility, in other words, the weight assigned to the long-run average variance is zero. Only the current estimate of the variance is used. The other statements are all correct.
Question 7
PE2018P29 / PE2019Q29
A junior risk analyst is modeling the volatility of a certain market variable and is trying to decide between EWMA and
GARCH
(
1
,
1
)
\text{GARCH}(1,1)
GARCH(1,1) models. Which of the following statements about the two models is correct?
A. The EWMA model is a special case of the
GARCH
(
1
,
1
)
\text{GARCH}(1,1)
GARCH(1,1) model with the additional assumption that the long-run volatility is zero.
B. A variance estimate from the
GARCH
(
1
,
1
)
\text{GARCH}(1,1)
GARCH(1,1) model is always between the prior day’s estimated variance and the prior day’s squared return.
C. The
GARCH
(
1
,
1
)
\text{GARCH}(1,1)
GARCH(1,1) model always assigns less weight to the prior day’s estimated variance than the EWMA model.
D. A variance estimate from the EWMA model is always between the prior day’s estimated variance and the prior day’s squared return.
Answer: D
Learning Objective:
- Apply the exponentially weighted moving average (EWMA) model to estimate volatility.
- Describe the generalized autoregressive conditional heteroskedasticity ( GARCH ( p , q ) \text{GARCH}(p,q) GARCH(p,q)) model for estimating volatility and its properties.
The EWMA estimate of variance is a weighted average of the prior day’s variance and prior day squared return.
A is incorrect. EWMA is a particular case of GARCH ( 1 , 1 ) \text{GARCH}(1,1) GARCH(1,1) with the weight assigned to the long-run average variance rate as zero and the sum of the weights of the other two parameters equal to 1.
B is incorrect because there is also weight assigned to the long-run average variance rate.
C is incorrect because such a comparison can only be done under specific parameter configurations.
Question 8
In the EWMA model, the half-life is defined as the time, T T T, at which λ T = 1 / 2 \lambda ^T = 1/2 λT=1/2, where λ \lambda λ is the decay factor of the EWMA model. A risk analyst is using a specific EWMA model to calculate volatility and determines that the half-life of the model is 23 23 23 days. Based on the above information, which weight will be applied to the return that is five days old?
A.
0.026
0.026
0.026
B.
0.031
0.031
0.031
C.
0.781
0.781
0.781
D.
0.859
0.859
0.859
Question 9
A firm uses an EWMA model to estimate the daily volatility of the return of a security. The following table shows the beginning-of-day estimate of the daily volatility, the end-of-day closing price, and the daily return, for each day during the past week:
Day | Estimated Volatility ( % \% %) | Closing Price (USD) | Daily Return |
---|---|---|---|
Monday | 2.86 2.86 2.86 | 50.39 50.39 50.39 | − 0.50 -0.50 −0.50 |
Tuesday | 2.80 2.80 2.80 | 50.65 50.65 50.65 | 0.52 0.52 0.52 |
Wednesday | 2.75 2.75 2.75 | 50.89 50.89 50.89 | 0.47 0.47 0.47 |
Thursday | 2.70 2.70 2.70 | 50.72 50.72 50.72 | 0.33 0.33 0.33 |
Friday | 2.64 2.64 2.64 | 50.58 50.58 50.58 | 0.28 0.28 0.28 |
Assuming the mean daily return is zero and using the information above, what is the value of the smoothing parameter used by the firm in its EWMA \text{EWMA} EWMA model?
A.
0.93
0.93
0.93
B.
0.894
0.894
0.894
C.
0.96
0.96
0.96
D.
0.98
0.98
0.98
Question 10
PE2019Q70
Which of the following statements on models for estimating volatility is correct?
A. The
EWMA
\text{EWMA}
EWMA model assigns a positive weight to the long-run average variance rate.
B. In the
EWMA
\text{EWMA}
EWMA model, the weights assigned to observations decrease exponentially as the observations become older.
C. The
GARCH
(
1
,
1
)
\text{GARCH}(1,1)
GARCH(1,1) model is a particular case of the EWMA model if the weight assigned to the long-run variance rate is not zero.
D. In the
GARCH
(
1
,
1
)
\text{GARCH}(1,1)
GARCH(1,1) model, the weights estimated for observations increase exponentially as the observations become older.
Answer: B
Learning Objective: Explain the weights in the
EWMA
\text{EWMA}
EWMA and
GARCH
(
1
,
1
)
\text{GARCH}(1,1)
GARCH(1,1) models.
In the EWMA \text{EWMA} EWMA model, the weights assigned (to volatility and changes in the variable) decrease exponentially as one moves back through time. The EWMA \text{EWMA} EWMA model does not involve the long-run average variance in updating volatility, in other words, the weight assigned to the long-run average variance is zero. Only the current estimate of the variance is used. The EWMA \text{EWMA} EWMA model is a particular case of GARCH ( 1 , 1 ) \text{GARCH}(1,1) GARCH(1,1) when the weight assigned to the long-run variance rate is zero. The GARCH ( 1 , 1 ) \text{GARCH}(1,1) GARCH(1,1) model is the same as the EWMA \text{EWMA} EWMA model except that, in addition to assigning weights that decline exponentially to past observations, it also assigns some weight to the long-run average variance rate. For a GARCH ( 1 , 1 ) \text{GARCH}(1,1) GARCH(1,1) process to be stable, the sum of the parameters α \alpha α and β \beta β needs to be less than 1.0 1.0 1.0.
Question 11
PE2020Q44 / PE2021Q44 / PE2022Q44
A financial analyst is concerned about the market risk of a stock. Based on the stock’s return data of the most recent
12
12
12 months, it has been estimated that the historical volatility of the monthly returns is
4.5
%
4.5\%
4.5%. Which of the following is most likely correct?
A. The implied volatility of the annual returns is
15.6
%
15.6\%
15.6%.
B. The implied volatility of the annual returns is
54.0
%
54.0\%
54.0%.
C. The volatility of the annual returns is
15.6
%
15.6\%
15.6%.
D. The volatility of the annual returns is
54.0
%
54.0\%
54.0%.
Answer: C
Learning Objective: Define and distinguish between volatility, variance rate and implied volatility.
This is 12 × 0.045 = 0.156 \sqrt{12}\times0.045=0.156 12×0.045=0.156.
A and B are incorrect. The implied volatility depends on the option price and it does not depend on the historical volatilities.
D is incorrect. This incorrectly scales the volatility linearly with time instead of by the square root of time, giving 12 × 0.045 = 0.54 12\times0.045=0.54 12×0.045=0.54
Question 12
PE2020Q45 / PE2021Q45 / PE2022Q45
A credit risk manager is in charge of credit risk analysis of large corporates at Bank XYZ. The manager is in possession of credit ratings provided by two rating agencies, X and Y, for
30
30
30 companies the manager oversees. The ratings are classified into four categories:
Rating categories | Description |
---|---|
1 | High investment grade |
2 | Mid investment grade |
3 | Low investment grade |
4 | Non-investment grade |
The manager plots the rating categories from the two agencies as shown below:
Which of the following statistical measures could best help the manager approximate the link. between rating categories from the two agencies?
A. Spearman correlation
B. Pearson correlation
C. Structured correlation matrix
D. Covariance
Answer: A
Learning Objective:
- Explain the relationship between the covariance and correlation of two random variables and how these are related to the independence of the two variables.
- Define correlation and covariance and differentiate between correlation and dependence.
Pearson correlation, correlation matrix and covariance are used to measure the degree of the relationship between linearly related variables. The credit ratings in this question are ordinal data and have nonlinear relationship as showed in the graph. So, the Spearman correlation is a better measure to indicate if one variable is monotonically related to the other variable.
Question 13
PE2022Q30
A risk analyst is assessing the correlation between the returns of two financial assets. The analyst wants to determine if the two sets of returns are dependent. Which of the following is correct regarding correlation and dependence?
A. Returns on financial assets tend to be independent.
B. Pearson’s correlation measures both linear and nonlinear dependence.
C. Correlation and the regression slope are closely related.
D. If the returns of the two assets are normally distributed, their rank correlation and Pearson’s correlation would not be equal.
Answer: C
Learning Objective: Define correlation and covariance and differentiate between correlation and
dependence.
C is correct. Correlation and the slope of the regression are intimately related, as regression explains the sense in which correlation measures linear dependence.
A is incorrect. Financial assets are highly dependent and exhibit both linear and nonlinear dependence.
B is incorrect. Pearson’s correlation only measures linear dependence.
D is incorrect. The rank correlation is virtually identical to the Pearson’s (also known as linear) correlation for normal random variables.
Question 14
PE2022Q46
An analyst is evaluating a dataset of annual returns for a financial asset. The analyst decides to use the Jarque-Bera test to determine if the returns of the asset are normally distributed. Which of the following is correct regarding the Jarque-Bera test?
A. The Jarque-Bera test statistic follows a Student’s t distribution.
B. The Jarque-Bera test only takes into account the skewness and kurtosis of a distribution.
C. The Jarque-Bera test requires that a Gaussian copula be applied to the returns data before conducting the test.
D. The Jarque-Bera test statistic does not depend on the sample size of the returns dataset.
Answer: B
Learning Objective: Explain how the Jarque-Bera test is used to determine whether returns are normally distributed.
B is correct. The Jarque-Bera test statistic is used to formally test whether the sample skewness and kurtosis are compatible with an assumption that the returns are normally distributed.
A is incorrect. The Jarque-Bera test assumes a normal distribution.
C is incorrect. There is no need for a Gaussian copula to be applied prior to testing.
D is incorrect. The test statistic is also a function of the sample size.