FINA 5250, 2022 Final Exam Practice Answers

Problem I. [Multiple Choices (single answer)]

(1) Which of the following plots is not a valid density function?

(a) plot 1 (b) plot 2 (c)plot 3 (d) plot 4

Answer: D. The density is: (1) always non-negative; (2) integrating to 1. The
plot 4 is not integrating to 1, but only to 0.5. (Lecture 1, p34)
(2) Which dataset represented by the following time series plot has the heaviest tail?
3

6
4
2

2
1

0
0

−2
−1

−4
−2

−6
−3

0 200 400 600 800 1000 0 200 400 600 800 1000

(a) (b)
3

5
2
1

0
0
−1
−2

−5
−3

0 200 400 600 800 1000 0 200 400 600 800 1000

(c) (d)

(a) (b) (c) (d)

1
2

Answer: D. The plot d presents more and larger extreme values on both sides
(positive/negative) than the others. (Lecture 2, p33-34)
(3-4). The following Q-Q plots compare the daily returns on S&P500 index with the normal
distribution and Cauchy distribution.

(3) The plots suggest that

(a) The returns have lighter tails than both normal and Cauchy.
(b) The returns have lighter tails than normal and heavier tails than Cauchy.
(c) The returns have heavier tails than normal and lighter tails than Cauchy.
(d) The returns have heavier tails than both normal and Cauchy.
Answer: C. Refer to Assignment 2, II; Lecture 2, p19-20.
(4) The 1% VaR and expected shortfall (ES) computed based on the normal assumption
are v and s respectively. Which of the following is most likely the case for the true
1% VaR and ES?
(a) VaR0.01 > v, ES0.01 < s (b) VaR0.01 < v, ES0.01 < s
(c) VaR0.01 < v, ES0.01 > s (d) VaR0.01 > v, ES0.01 > s
Answer: D. Given that the returns have heavier tails than normal, both measures
based on normal distribution are more likely to be underestimated. (Lecture 2, p24).
(5) The correlation between X and Y in the plot below is closest to
 3

(a) 1 (b) 0.7 (c) -0.7 (d) 0.05

Answer: C. We can clearly see that x negatively co-moves with y, their correlation
would be negative. (Lecture 3, p7-8)
(6) A random vector (X, Y ) follows a bivariate normal distribution with:

1
   
2 2
µ= , and Σ = ,
0 2 4

what is the correlation between X and Y ?

(a)0.707 (b) 0.000 (c) 0.800 (d) 0.250

p √
Answer: A. ρ(X, Y ) = cov(X, Y )/ V ar(X)V ar(Y ) = 2/ 2 ∗ 4 = 0.707. (Lec-
ture 3, p7-8)
(7) A dataset has 100 observations and the smallest 20 of them are:
-3.51 -2.17 -1.64 -1.63 -1.59 -1.59 -1.55 -1.54 -1.50 -1.47
-1.41 -1.38 -1.31 -1.28 -1.27 -1.21 -1.15 -1.11 -1.09 -1.05
which value below is/is closed to the empirical 5%-quantile of this dataset?
(a) -1.05 (b) 1.59 (c) -1.59 (d) -1.47

Answer: C. (Lecture 2, p3)

(8) Bank A reports that 1% VaR for monthly return is 2%. What is the proper interpre-
tation of this?
(a) If the bank invests $ 100 million for a month, we will always see losses larger
than $ 2 million.
(b) There is a 1% probability that the bank will gain less than 2% over a month.
(c) There is a 1% probability that the bank will lose more than 2% over a month.
(d) There is a 1% probability that the bank will lose less than 2% over a month.
Answer: C. Equivalently, the V aRq is the threshold value such that there is (1−q)
probability that the loss will not exceed this value. (Lecture 2, p23)
(9) You are considering three investment plans for the next 36 months. Plan X, plan Y,
and plan Z which is a mixture of X and Y that requires rebalancing at the beginning
of each month. Which of the following is correct about plan Z?
(a) Plan Z must be worse than the better one between plans X and Y.
(b) The performance of plan Z can be slightly better than both plans X and Y but
can not be substantially better.
(c) The performance of plan Z must be between the performance of plans X and Y.
(d) The performance of plan Z can be substantially better than both plans X and Y.
Answer: D (Lect5 Game).
(10) For a 5 million investment, it is known that P rob{loss > $30, 000} = 0.05 for tomor-
row. Which of the following is most likely the 1% $V aR for tomorrow?
(a) $10, 000
(b) $100, 000
4

(c) -$100, 000

(d) -$10, 000
Answer: B. $30,000 is the 5% $V aR, 1% $V aR must be greater than 5% $V aR.
(Lecture 2, p23)

Problem II. [True or False]

(1) T F Expected shortfall measures the average loss.

Answer: F. Expected shortfall measures the conditional average loss given that
the loss exceeds V aRq . (Lecture 2, p29)

(2) T F A√random variable X ∼ N (0, 1). It√is known that the density of N (0, 1)
at 0 equals 1/ 2π, and so P (X = 0) equals 1/ 2π.
RH
Answer: F. Density and probability is not the same, P (H ≥ X ≥ L) = L f (x)dx.
Note that P (X = a) is zero in a continuous distribution. (Lecture 1, p37)

(3) T F All the normal distributions form a location-scale family, and if Y ∼

N (µ, σ 2 ), then Y can be written as µ + σZ for Z ∼ N (0, 1).
Answer: T. (Lecture 2, p11)

(4) T F In kernel density estimation, the choice of kernel is crucial.

Answer: F. In kernel density estimation, the choice of bandwidth is crucial, not
the choice of kernel. (Lecture 1, p55-56)

(5) T F It is possible that X and Y are independent but ρ(X, Y ) 6= 0.

Answer: F. ρ(X, Y ) = 0 if the random variables X and Y are independent. (Lec-
ture 3, p7)

(6) T F When comparing two models, the model that has a higher R2 is better.
Answer: F. If two models are nested, the larger model would always have a
higher R2 . An F-test can tell in-the-sample whether the larger model is significantly
better. An out-of-sample comparison based on prediction accuracy would be even
more informative. (Lect4, p5-6)

(7) T F The standard deviation of returns is usually smaller than the mean of
returns.
Answer: F. The standard deviation of returns completely dominates the mean of
returns at short horizons such as daily. (Lecture 6, p8)

(8) T F The unconditional distribution of daily returns usually have fatter tails
than the normal distribution.
Answer: T. (Lecture 6, p6)
(9) T F σt from a GARCH(1,1) process is determined by the information before
time t.
Answer: T. (Lecture 7, p4)

(10) T F The market related variance and non-market related variance can be
diversified away by holding a large enough portfolio.
 5

Answer: F. Only the non-market related variance can be diversified away by

holding a large enough portfolio. (Lecture 5, p26)

Problem III. Suppose we have the following market models for assets A and B
RA,t = −0.000002 + 0.5RM,t + A,t

RB,t = 0.00001 + 1.5RM,t + B,t

The market volatility for tomorrow is σM,t+1 = 2%; the standard deviations for A,t and B,t
are τA = 0.001 and τB = 0.0015 respectively.
1. Find the conditional covariance matrix Σt+1 for the returns of the two assets.
2
Answer:(Lecture 5, p20) σA,t+1 = 0.52 ∗ 0.022 + 0.0012 = 0.000101
2
σB,t+1 = 1.52 ∗ 0.022 + 0.00152 = 0.00090225
σAB,t+1 = 0.5 ∗ 1.5 ∗ 0.022 = 3 ∗ 10−4

2. For a portfolio with 30% A and 70% B, calculate the portfolio αP , βP (the intercept
and slope of the market model for the portfolio) and the conditional variance of the portfolio
2
return σP,t+1 = Vart (RP,t+1 ).
Answer:(Lecture 5, p24-25) αP = 0.3 ∗ (−0.000002) + 0.7 ∗ 0.00001 = 6.4e − 06
βP = 0.3 ∗ 0.5 + 0.7 ∗ 1.5 = 1.2
2
σP,t+1 = 1.22 ∗ 0.022 + 0.32 ∗ 0.0012 + 0.72 ∗ 0.00152 = 0.0005771295

3. Assume normality. For a portfolio with 0% A and 100% B, find the 1% VaR for
the one-day ahead return. Is this higher or lower than the one-day 1% VaR of the portfolio
considered in question 2 above?
2
Answer:(Assignment 4) from Q1 above, σB,t+1 = 0.00090225
0.01
√ √
V aRt+1 = 0.00090225 ∗ 2.33 − 0.00001 ≈ 0.00090225 ∗ 2.33 = 0.07
0.01
This is higher than V aRt+1 for the portfolio considered in 2, since the variance of B is larger
than that of the portfolio. √
0.01
(note that in Q2, V aRt+1 = 0.0005771295 ∗ 2.33 − 6.4e − 06 = 0.056)

Problem IV. Excess return "rex " of a stock has been fitted into a single factor model
with predictor market excess return "rMex " using LS regression. Here is the summary of the
fit:
lm(formula = r_ex ~ rM_ex)
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 0.123617 0.007175 17.23 <2e-16 ***
rM_ex 0.997074 0.007662 130.14 <2e-16 ***
---
Signif. codes: 0 ’***’ 0.001 ’**’ 0.01 ’*’ 0.05 ’.’ 0.1 ’ ’ 1
Residual standard error: 0.1139 on 250 degrees of freedom
Multiple R-squared: 0.9855, Adjusted R-squared: 0.9854
F-statistic: 1.694e+04 on 1 and 250 DF, p-value: < 2.2e-16

1. Based on the fitted result, is the predictor "rMex " useful in explaining the variation in
the return?
Answer: (Lecture 3, p34-35) The p-value of rM _ex is 2e − 16 which is less than 5%, so
rM _ex is useful in explaining the variation in the return. See also Hwk 3.
6

2. Is α significantly different from 0? If so, which direction?

Answer: (Lecture 3, p34-35)p-value for α is 2e-16, so α is significant different from zero
at 1% significance level. And it is significantly higher than zero.

3. Test if β is significantly different from 1.

Answer: (Lecture 3, p34-35) To test if β is significantly different from 1,
β̂−1 0.997−1
Set t-statistics : t = sd( β̂)
= 0.007662 = −0.3819.
The t-statistics is smaller than 2 in absolute value, so we don’t reject the null hypotheis.
I.e., no strong evidence against the null hypothesis that the β is equal to 1.
Note also that (not required) the p-value for t=-0.3819 with df=257 is P (> |t|) = 0.7. So
the value β is not significantly away from 1.

Suppose the excess return of stock is then fitted into Fama-French three factor model using
LS regression. Here is the summary of the fit:
lm(formula = r_ex ~ rM_ex + rSmB + rHmL)
Coefficients:
(Intercept) 0.087626 0.015218 5.758 2.51e-08 ***
rM_ex 0.999595 0.007262 137.638 < 2e-16 ***
rSmB 1.139450 0.217624 5.236 3.50e-07 ***
rHmL 1.349625 0.688321 1.961 0.051 .
---
Signif. codes: 0 ’***’ 0.001 ’**’ 0.01 ’*’ 0.05 ’.’ 0.1 ’ ’ 1
Residual standard error: 0.1076 on 248 degrees of freedom
Multiple R-squared: 0.9871, Adjusted R-squared: 0.987
F-statistic: 6329 on 3 and 248 DF, p-value: < 2.2e-16

4. Based on the fitted result, are the three factors as a whole statistically significant for
FF-3 factor model? Is any single one of the factors statistically significant for FF-3 factor
model? (Set significance level to be 5%) .
Answer: (Lecture 4, p10-17) p-value for the F-test is < 2.2e-16. The p-value indicates
that the three factors as a whole is significant.
Checking p-values for each factors in FF-3:
p-value for rM_ex is < 2e-16 ***, indicating rM_ex is statistically significant at 5% level.
p-value for rSmB is 3.50e-07, indicating rSmB is statistically significant at 5% level.
p-value for rHmL is 0.051, indicating rHmL is not statistically significant at 5% level.

5. Based on all information given in this problem, how much can the single factor explain
the variation in the returns? How much can the three factors explain the variation in the
returns?
Answer: (Lecture 3, p38) For single factor model: R2 is 0.9855, so single factor can
explain 98.5% of the variation in the stock returns.
For FF-3: R2 is 0.9871 , so the three factors explain 98.7% of the variation in the asset
returns.

6. Suppose we conduct a partial-F test to check if the 3-factor model explian statistically
significantly more variation in the response than the single factor model. Here is the summary
of the partial-F test:
Analysis of Variance Table
Model 1: r_ex ~ rM_ex
 7

Model 2: r_ex ~ rM_ex + rSmB + rHmL

Res.Df RSS Df Sum of Sq F Pr(>F)
1 250 3.2422
2 248 2.8737 2 0.36847 15.899 3.186e-07 ***
---
Signif. codes: 0 ’***’ 0.001 ’**’ 0.01 ’*’ 0.05 ’.’ 0.1 ’ ’ 1
What can you conclude from this test result?
Answer: (Lecture 4 p10-11) The p-value for the partial F test is 3.186e-07 ***, if we
consider 5% or even 1% statistical significance level, then the 3-factor model does explain
statistically significantly more variation in the response than the single factor model.
Problem V. The variance of a stock return data (n=500) is fitted by GARCH(1,1) model:
2
σt+1 = 0.000002 + 0.1 ∗ Xt2 + 0.78 ∗ σt2 , and the most recent 3 observations are listed below:

day-498 day-499 day-500

Return(Xt ) 0.0012 0.0009 -0.0001
σ̂t 0.0001
Assume returns Xt = σt Wt , where Wt ∼i.i.d. N (0, 1).
1. Compute the fitted volatilities√σ̂499 and σ̂500 .
2 2
Answer:√ (Lecture 7, p12) σ̂499 = 0.000002 + 0.1 ∗ 0.0012 + 0.78 ∗ 0.0001 = 0.0015;
2 2
σ̂500 = 0.000002 + 0.1 ∗ 0.0009 + 0.78 ∗ 0.0015 = 0.0020

2. Forecast the (conditional standard) deviation for day-501 (given information up to day
500). p
Answer: (Lecture 7, p12) σ̂501 = 0.000002 + 0.1 ∗ (−0.0001)2 + 0.78 ∗ 0.00202 = 0.0023

3. What is the distribution of the return of day-501 given all the information up to
day-500? Construct a 95% prediction interval for the return of day-501.
Answer: (Lecture 7, p4) N (0, 0.00232 ); 95% prediction interval: 0 ± 1.96 ∗ 0.0023