#QF604 MCQ Practice Test 4

QF604 MCQ Practice Test 4
Please tick the most suitable answer to each multiple-choice question.
Q1. What is X 2 + Y 2 + Z 2 where X, Y, Z are independently distributed as N (0, 2)?
(A) 3 × [N (0, 2)]2

(B) 2 × χ23
(C) 3 × χ22
(D) None of the above
Q2. Suppose we run an Ordinary Least Square regression of Yi = a + bXi + ei where ei is a white noise that
is independent of Xi . Given sample averages for Y and X are 300 and 250, respectively, and b̂ = 1.0,
what is â?
(A) 1.20
(B) 1.19
(C) 50.0
(D) Indeterminate from the given information.
Q3. For the general covariance-stationary processes such as ARMA(p, q) (p, q being any reasonably finite
integers)
(A) conditional mean and conditional variance are constant at each t

(B) conditional mean and conditional variance change at each t
(C) conditional mean changes and conditional variance is constant at each t.
Q4. If a stochastic trend exists in a price process Zt with i.i.d. increments, then this is not likely to show
up as
(A) mean reversion

(B) a possibly changing conditional mean term
(C) a decreasing variance as time decreases
(D) a correlogram (graph of correlation function against time lags) that decays very slowly.
Q5. Suppose we are testing if Pt is a unit root or I(1) process, and we perform the following OLS regression
∆Pt = δ + θPt−1 + et , where et is a stationary random variable. Suppose the critical ADF statistic for
this case at 1% significance level is −3.44, at 5% significance level is −2.86, and the computed θ̂ < 0
is -3.965, then you
(A) reject null of unit root at 1%

(B) cannot reject null of unit root at 1%
(C) cannot reject null of no unit root at 1%
1
(D) cannot reject null of no unit root at 5%
Q6. If Qt = e−rt Pt , and Qt follows a martingale, then
(A) Et (Pt+1 ) = Pt
(B) Et (Pt+1 ) = er Pt
(C) Et (Pt+1 ) = ert Pt
(D) Et (Pt+1 ) = er(t+1) Pt
Q7. The reason why stock returns may be predictable is associated with
(A) daily time trend of the stock return

(B) long-term stock autocorrelations
(C) weekly stock price variations
(D) momentum of stock movements
Q8. A stock beta is obtained by running
(A) Cross-sectional regression

(B) Time series regression
(C) Cross-sectional regression with constraint on loadings
(D) Time series regression with constraint on residual error
Q9. An irrelevant variable was included in a mulitple linear regression,. Which of the following is most
reasonable?
(A) the estimated coefficients will always be biased

(B) the estimated coefficients will always be biased downward
(C) the estimated coefficient standard errors are typically larger
(D) the estimated coefficient standard errors are typically smaller
Q10. The Fama-McBeth procedure is
(A) First running cross-sectional regression, then running time series regression
(B) First running cross-sectional regression, then repeat after sorting
(C) First running time series regression, then running cross-sectional regression
Q11. Suppose Yi = c0 + c1 Xi + ei , i = 1, 2, . . . , N . Xi and zero mean ei are stochastically independent,

and ei is heteroskedastic, then the OLS estimator is
(A) BLUE
(B) biased but consistent
2
(C) unbiased but not efficient
Q12. Suppose Yt = c0 + c1 Xt + c2 Zt + et , t = 1, 2, . . . , T and et satisfies the classical conditions. However,

in a regression, Zt was omitted. If Zt = ρZt−1 + ut where ρ < 0, and ut is i.i.d., the D-W statistic in
the above is likely to be
(A) less than 2

(B) about 2
(C) more than 2
(D) cannot be computed.
Q13. In a regression of future excess return on current dividend/price variable, suppose the errors or dis-
turbances are not contemporaneously correlated with dividend yields, but are serially correlated, then
the OLS estimates will be
(A) unbiased but inconsistent

(B) unbiased and consistent
(C) biased and consistent
(D) wrong because of wrong t-statistic.
Q14. If null hypothesis is that abnormal return ARit ∼ N (0, σ 2 ), what is the null of the distribution of
cumulative abnormal return at the end of a window of 5 days?
1 2
(A) 5 N (0, σ )
√
(B) 5N (0, σ 2 )
(C) 5N (0, σ 2 )
(D) N (0, σ 2 )
Q15. There are typically many specifications that are consistent with the unbiased expectations hypothesis.
The following is one.
St = c0 + c1 Ft−k,t + et , k > 0
What restrictions on the regression coefficients and disturbance are implied by the UEH?
(A) c0 × c1 = 0, E(et | Ft−k,t ) = 0

(B) c0 × c1 = 0, E(et | Ft−k,t ) 6= 0
(C) c0 = 0, c1 = 1, E(et | Ft−k,t ) 6= 0
(D) c0 = 0, c1 = 1, E(et | Ft−k,t ) = 0
Q16. An ARMA(p,q) process, where p, q are finite, can be represented as an infinite AR process provided
(A) it is stationary
(B) it is invertible
3
(C) it does not have unit roots
(D) it is also autoregressive
1
Q17. In a GARCH model where return is rt = ht2 et , et is i.i.d. distributed as N (0, 1), and conditional
2
variance of et is ht = β0 + β1 rt−1 + β2 ht−1 e2t−1 , what is the most plausible set of estimates?
(A) β0 = 0.03, β1 = 0.7, β2 = −0.1

(B) β0 = −0.03, β1 = 0.7, β2 = −0.1
(C) β0 = 0.03, β1 = 0.4, β2 = 0.7
(D) β0 = −0.03, β1 = 0.4, β2 = 0.7
Q18. Which technique can you use to address the endogeneity bias problem in a linear regression?
(A) Fama-MacBeth’s grouping procedure

(B) White’s HCCME
(C) Instrumental variable method
(D) generalized least squares
Q19. Suppose a first linear regression is Y1t = a0 + a1 Xt + et for t = 1, 2, . . . , T , and the vector of residual
errors has T × T covariance matrix Σe . Suppose a second linear regression is Y2t = b0 + a1 Zt + vt
for t = 1, 2, . . . , T , and the vector of residual errors has T × T covariance matrix Σv . et and vt are
independent. If instead we form dependent variable vector M2T ×1 = (Y11 , . . . , Y1T , Y21 , . . . , Y2T )T , and
explanatory variable matrix S such that M = SB + E where B is vector of coefficients from the two
regressions to be estimated, and
Σe 0
cov(E) = ,
0 Σv 2T ×2T
what is the dimension of matrix S?
(A) 2T × 2
(B) 2T × 3
(C) 2T × 4
Q20. In a linear regression model Y = XB + e, YN ×1 takes only binary values of 1 or 0, X is a N × 3

explanatory variable matrix where the first column contains all ones, and e is a N vector of zero mean
1
i.i.d. residual errors. e is non-normal. Let the ith row of X be Xi . Suppose E(Yi |Xi ) = 1+e−X iB
using
logistic regression. If the estimates of B are (0.1, 0.1, 0.1), what is the estimated probability of Y = 1
when Xi = (1, 2, 3)?
(A) 0.62
(B) 0.65
(C) 0.68
Ans: Q1 (B), Q2 (C), Q3 (C), Q4 (A), Q5 (A), Q6 (B), Q7 (B), Q8 (B), Q9 (C), Q10 (C)
Q11 (C), Q12 (C), Q13 (B), Q14 (B), Q15 (D), Q16 (B), Q17 (A), Q18 (C), Q19 (B), Q20 (A)

#QF604 MCQ Practice Test 4

Uploaded by

Copyright:

Available Formats

#QF604 MCQ Practice Test 4

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

#QF604 MCQ Practice Test 4

Uploaded by

Copyright:

Available Formats

QF604 MCQ Practice Test 4

Please tick the most suitable answer to each multiple-choice question.

Q1. What is X 2 + Y 2 + Z 2 where X, Y, Z are independently distributed as N (0, 2)?

(A) 3 × [N (0, 2)]2

(A) conditional mean and conditional variance are constant at each t

(A) mean reversion

(A) reject null of unit root at 1%

Q6. If Qt = e−rt Pt , and Qt follows a martingale, then

(A) daily time trend of the stock return

Q8. A stock beta is obtained by running

(A) Cross-sectional regression

(A) the estimated coefficients will always be biased

Q10. The Fama-McBeth procedure is

Q11. Suppose Yi = c0 + c1 Xi + ei , i = 1, 2, . . . , N . Xi and zero mean ei are stochastically independent,

Q12. Suppose Yt = c0 + c1 Xt + c2 Zt + et , t = 1, 2, . . . , T and et satisfies the classical conditions. However,

(A) less than 2

(A) unbiased but inconsistent

(A) c0 × c1 = 0, E(et | Ft−k,t ) = 0

(A) β0 = 0.03, β1 = 0.7, β2 = −0.1

(A) Fama-MacBeth’s grouping procedure

Q20. In a linear regression model Y = XB + e, YN ×1 takes only binary values of 1 or 0, X is a N × 3

You might also like