January 2017 Examination ANSWERS with marks

EC402
Econometrics

2016/2017 syllabus –not for resit candidates

Instructions to candidates

Time allowed: 2 hours

This paper contains TWO sections. Answer FOUR questions from SECTION A and THREE
questions from SECTION B. Sections A carries 40% of the overall mark, while Section B carries
60% of the overall mark.

You are supplied with: Dickey-Fuller Statistical Tables

Murdoch & Barnes Statistical Tables
Table A5 Durbin-Watson d-statistic

Calculators are NOT allowed in this examination.

c LSE LT 2017/EC402 Page 1 of 17

SECTION A
This section carries 40% of the total mark. Answer FOUR questions from this section.

1. [10 marks] Consider the classic linear regression model with dependent S 1 vector y and
regressor S k matrix X, assumed to satisfy:
A1 : rank(X) = k
true true true
A2 : y = X + with E =0
The regressors X are assumed to be unrelated to the error term true in one of two alternative
ways:
Strongly exogenous de…ned as: A3Rmi : E( true jX) =E true

W eakly exogenous de…ned as: A3Rsru : E( true x ) = 0 for every regressor variable
s sj

j = 1; ; k and every sample observation s = 1; ; S.

True, False, or Uncertain: “If we have a large enough sample, Strong Exogeneity is not a
particularly useful property. Whether our sample is a cross-section, panel, or time-series data
set, as long as S is very large, the critical property for the regressors is weak exogeneity.”
Carefully explain your answer.
ANSWER:
The two types of exogeneity as de…ned above di¤er in two fundamental ways:
(a) WE uses only the notion of Zero Correlation/Zero Covariance/Zero Cross Expectation
between error and regressor instead of the Stronger Mean Independence condition of SE. And
(b) WE is a Same Row (s for both the error and the regressor) condition, as opposed to the
Any Rows (s for the error and q for the regressor) condition, which of course is weaker.
It is true that for Asymptotics, the Zero Correlation/etc condition is the necessary one for the
results to follow through, since we are checking whether Sample Moment expressions like
S S
1X true 1X
xsj s and xsj xsl
S S
s=1 s=1

converge by a suitable LLN to the corresponding true Moment expections; and whether an
appropriate CLT applies to give the asymptotic normality of expressions like
S
1 X true
p xsj s
S s=1

But the statement is FALSE that Strong Exogeneity is not useful when we have large sam-
ples/Asymptotics because the (b) part of the di¤erence above (Same Row vs. Any Rows) is
critical in many contexts, even then: the unit of observation being s is applicable to whichever
type of data set we might have, be it Cross-Section, Time-Series or Panel/Longitudinal. Hence

c LSE LT 2017/EC402 Page 2 of 17

 the case of any rows, not just the same, s 6= q is very important depending on the type of the
data set:
(i) if Cross Section, s 6= q means di¤erent individuals, or …rms, or countries etc.
(ii) if Time Series, s 6= q means di¤erent time-periods. And
(iii) if Panel Data, s 6= q may mean either Di¤erent individuals (and same or di¤erent time
periods) OR Same Individual and di¤erent time periods.
So irrespective of the sample size, Strong Exogeneity may be necessary for good properties of
the estimators if we have data sets (ii) or (iii) — and even in (i) if we have correlations across
a Cross Section because of Spatial reasons etc.

c LSE LT 2017/EC402 Page 3 of 17

2. [10 marks] A researcher considers …tting a linear regression model to a data set (y; X) with
sample size S = 75 and k = 7 regressor variables. His aim is to test formally the set of
hypotheses:

2
+ 3
+ 4
= 1 and
( 5 2)4 + ( 6 7)
2
= 0

The researcher argues that given the relatively small sample size, he prefers to avoid using the
Wald testing approach in favour of the Likelihood Ratio approach. He explains that the Wald
approach would require the nonlinear “Delta Method” to implement, and hence be unreliable
with the rather small sample size. Discuss critically his arguments.
ANSWER:
The researcher’s argument is FALSE for at least two reasons:
(a) Even the LR approach of the two conditions exactly as stated, would still involve Nonlinear
estimation, since the LR relies on the maximized LLF for the Unrestricted as well as the Re-
stricted models and compares the two. The “R”model would involve the second (apparently)
non-linear condition as well as the …rst linear condition, and hence would need asymptotic
arguments to derive the estimator and test properties
(b) The second condition is not really non-linear, but it corresponds to the two *linear* con-
ditions 5 2 and 6 = 7 . Therefore these conditions amount to 3 purely linear conditions
that can be tested in …nite samples exactly by de…ning

true
H0 : R =q

where in this case R is the r =3 k matrix and q the 3 1 vector:

0 1 0 1
0 1 1 1 0 0 0 0 1
B C B C
R=@ 0 0 0 0 1 0 0 0 A and q = @ 0 A
0 0 0 0 0 0 1 1 0

c LSE LT 2017/EC402 Page 4 of 17

3. [10 marks] Consider a data set (y; X) with dependent S 1 vector y and regressor S k
matrix X, assumed to satisfy:
A1 : rank(X) = k
Consider two S S matrices W and Z, and de…ne the new data premultiplied by the two
matrices respectively:

y+ W y X+ WX
y~ Zy ~
X ZX
The Frisch-Waugh-Lovell (FWL) theorem proposes to calculate the Ordinary Least Squares
(OLS) coe¢ cients obtained from regressing y + on X + , whereas the Generalized Gauss-Markov
(GGM) theorem proposes to calculate the OLS coe¢ cients from regressing y~ on X. ~

(a) (5 marks)
Explain the form of the W matrix and what the FWL theorem achieves in this case.
What are the properties of the OLS coe¢ cients from regressing y + on X + ?
ANSWER:
.
The FWL theorem partitions the regressor matrix into two parts, X = XA ..XB where
XA contains kA regressors and XB the other k kA = kB regressors. Then the OLS
vector of coe¢ ents can be correspondingly partitioned into the estimated coe¢ ents for
the A part, ^ A , and the regressors for the B part, ^ B through the simple formula:

^ = (X +0 X + ) 1
X +0 y +
A

where
y+ W y X+ WX

with W = MB = IS XB (XB 0 X ) 1 X 0 is the projection matrix constructed using the

B B
XB regressors, which turns a vector or matrix premultiplied by it into OLS residuals from
a regression with XB as regressors.
So regressing y + on X + gives exactly the same OLS ^ as those extracted from the full
A
(X 0 X) 1 X 0 y formula.
(b) (5 marks)
Explain the form of the Z matrix in the GGM theorem. Do you need to add any further
assumptions to A1 for the GGM theorem to apply? What are the properties of the OLS
~ in this case?
coe¢ cients from regressing y~ on X
ANSWER:
The GGM applies to the case where the plain OLS estimator is not BLUE because the
VCov(.) matrix of the true error vector is not a scalar matrix but

true
A4 : V Cov( jX) = c2

c LSE LT 2017/EC402 Page 5 of 17

 where is S S symmetric, positive de…nite matrix with all elements real and …nite.
Therefore, 1 will also be pos.def. and hence it will have a matrix square root de…ned
by:
1 1=2 1=20
= = Z 0Z

(obtained either through the Similarity representation using eigenvalues and eigenvectors
or through the LU factorization of Gauss). If the data are transformed by premultiplying
by Z 1=20 to give the tilde quantities,

y~ ~
Zy X ZX

the true error term of the transformed model will have its distribution “rotated” so that
it will satisfy the A4GM assumption:

true
V Cov(Z jX) = Z Z 0 = 1=20 1=2
=I

~ by OLS will give the BLUE since the regular GM theorem applies.
Thus, regressing y~ on X
But of course this OLS estimator corresponds to the IGLS estimator since

~ 0 X)
(X ~ 1 ~ 0 y~ = (X 0 Z 0 ZX)
X 1
X 0 Z 0 Zy = (X 0 1
X) 1
X0 1
y

c LSE LT 2017/EC402 Page 6 of 17

 4. [10 marks] Consider the regression model with a single non-constant regressor xs and four
unknown coe¢ cients , , , and .

ys = + xs + (xs ) + s

The sample consists of s = 1; : : : ; S observations with the sample size S being very large. The
sigmoid function ( ) is the cumulative distribution function of a standard N (0; 1) random
variable respectively, with ( 1) = 0, (0) = 0:5, and (+1) = 1. The error disturbance
s is believed to be fully independent from the xs for any observation s, and is distributed
independently and identically over s with mean zero and variance 2 < 1. A researcher, who
is only interested in parameters and , proposes two estimation strategies:

(a) (4 points)
Apply Ordinary Least Squares (OLS) to a regression of ys on a constant and xs .
(b) (6 points)
Ignore the ( ) term and …nd an instrument variable zs for the xs regressor. Then apply
Instrumental Variable Estimation (IVE) to a regression of ys on a constant and xs only,
using zs and the constant as instruments.

Explain carefully why both proposed strategies would be inappropriate, and which estimation
method you would propose instead.
ANSWER:

The problem with approaches (a) and (b) is that they correspond to the regression model

ys = + xs + us

where us = (xs ) + s
The regressor xs is thus endogenous w.r.t. to the composite error us and so approach (a) [OLS
of ys on a constant and xs ] will be biased and inconsistent.
The IVE approach of (b) would be consistent *provided* an instrument variable zs can be found
that is (i) a *valid* instrument in the sense that it is uncorrelated from us and (ii) a *relevant*
instrument in the sense that it is highly correlated with the endogenous regressor xs . Even if the
regressor xs is exogenous w.r.t. the original error term s , it is practically impossible to think of a
variable zs that is both valid and relevant. This is because the endogeneity is due to the (xs )
term inside the composite error us , which is of course correlated with xs .

c LSE LT 2017/EC402 Page 7 of 17

5. [10 marks] Consider a regression model where a dependent variable Y is regressed on four
regressors, x1, X2, X3, and x4. The sample is indexed by i = 1; ; N . Regressor x1 is
a constant vector of 1s, and regressor x4 is fully observed. The dependent variable Y and
regressors X2 and X3, however, are only observed through additive measurement errors:

Yi = yi + wi
Xi2 = xi2 + ui
Xi3 = xi3 + vi

The measurement errors w, u, and v are fully independent from the true variables
y, x2 , x3 , and x4 . Assume that the true regression function is given by:

A2 : yi = 1 + 2 xi2 + 3 ln(xi3 + 4 xi4 ) + i; E i =0

The true regressor variables x2 , x3 , and x4 are fully independent from the original regression
error term . Note that the variables x3 and x4 enter non-linearly in the model.
Given the imperfect observations on y, x2 , and x3 , the researcher is forced to work with the
equation:
A20 : Yi = 1 + 2 Xi2 + 3 ln(Xi3 + 4 xi4 ) + i

(a) (3 points)
Derive the properties of the error term i.
ANSWER:
Starting from A2 and substituting in the additive measurement error expressions, we
obtain

A20 : Yi = 1 + 2 Xi2 + 3 ln(Xi3 + 4 xi4 )+ i +wi 2 ui + 3 ln(xi3 + 4 xi4 ) 3 ln(Xi3 + 4 xi4 )

Hence i is de…ned by:

xi3 + 4 xi4
i = i +wi 2 ui + 3 ln(xi3 + 4 xi4 ) 3 ln(Xi3 + 4 xi4 ) = i +wi 2 ui + 3 ln
Xi3 + 4 xi4

(b) (4 points)
Explain whether or not i has zero expectation and whether or not it is uncorrelated from
the regressor variables.
ANSWER:
n o
xi3 + 4 xi4
E i is not zero since E i = Ewi = Eui = 0 but E ln X i3 + 4 xi4
6= 0 because of the
nonlinearity. The fact that Xi3 = xi3 + vi implies that the ln() term will not have zero
expectation.
The regressor variables in A20 are: Xi2 which appears linearly; and Xi3 and xi4 which
both appear nonlinearly.

c LSE LT 2017/EC402 Page 8 of 17

 (i) Xi2 is exogenous to i and to wi by the original assumptions; it is *not* exogenous
w.r.t. to ui since it contains ui ; and it is not exogenous w.r.t. to the ln() part of the
composite error because presumably x2 and x3 are correlated.
(ii) Xi3 is exogenous to i and to wi by the original assumptions; it is also exogenous w.r.t.
the ui part; but of course it is not exogenous w.r.t. to the ln() term for two reasons: the
ln() term contains xi3 , Xi3 , and xi4 .
(c) (3 points)
Consider two estimation strategies for this model: (i) Non-linear Least Squares (NLLS)
and (ii) Generalized Method of Moments (GMM). Can either method provide consistent
parameter estimates for the parameters? Why or why not?
ANSWER:
Either method proposed will be inconsistent.
(i) For NNLS to be consistent, all the regressors appearing on the RHS must be exogenous
w.r.t. to the true error term of the NLLS regression. As explained above, this is in general
impossible.
(ii) GMM would require valid Instrument Variables. But in view of this speci…cation, it
is highly unlikely that such IVs can be found.

c LSE LT 2017/EC402 Page 9 of 17

 SECTION B
Section carries 60% of the total mark. Answer THREE questions from this section.

6. [20 marks] For the classic linear regression model estimated with data (y; X) with
k regressors and S sample points, consider the Gauss-Markov (GM) theorem:
GM theorem: Assume the following four conditions hold:
A1 : rank(X) = k < S
true true true
A2 : y = X + where E =0
A3Rmi : E( true jX) =E true

A4GM : V Cov( true jX) = 2I

S

Consider the Ordinary Least Squares (OLS) estimator ^ ols = (X 0 X) 1 X 0 y and an arbitrary
Linear Unbiased (LU) estimator ^ LU = BX y, where the k S matrix BX satis…es
true
BX X = Ik . Then the OLS estimator is BLUE for in the sense that their respective
variance-covariance matrices satisfy:
V Cov( ^ lu jX) V Cov( ^ ols jX) V lu V ols is a positive semi-de…nite matrix for any LU
estimator.

(a) (6 points)
Consider the following …ve statements:
i. ^ 2ols + ^ 3ols + ^ 4ols is BLUE for true
2 + true
3 + true
4
ii. ^ 2ols is BLUE for true 2
iii. ^ ^ is BLUE for true true
4ols 5ols 4 5
iv. ^ 2lad + ^ 3lad + ^ 4lad is BLUE for true
2 + true
3 + true
4
v. ^ ^ is BLUE for true true
4lad 5lad 4 5

Explain how the GM theorem result about the positive de…niteness of the matrix di¤erence
V lu V ols above can be readily used to prove the …rst two statements, but not the last
three statements.
Hint: For (i) and (ii), consider how you would calculate var( ^ 2ols + ^ 3ols + ^ 4ols ) and
var( ^ ) from V ols , and how you would calculate var( ^ + ^ + ^ ) and var( ^ )
2ols 2lu 3lu 4lu 2lu
from V lu .
ANSWER:
Consider a nonzero k-dimensional vector c and de…ne the random variable zols = c0 ^ ols .
Use the same vector to de…ne zlu = c0 ^ lu . The variances of the two r.v.s will be respec-
tively:
var(zols ) = c0 V Cov( ^ jX)c var(zlu ) = c0 V Cov( ^ jX)c
ols lu

So var(zlu ) = Cov( ^ lu jX)c var(zols ) = c0 V Cov( ^ ols jX)c if and only if c0 V Cov( ^ lu jX)c
c0 V
c0 V Cov( ^ ols jX)c, OR if and only if c0 [V Cov( ^ lu jX) V Cov( ^ ols jX)]c 0. Since c is

c LSE LT 2017/EC402 Page 10 of 17

 a non-zero vector, this is exactly the de…nition of the matrix di¤erence [V Cov( ^ lu jX)
V Cov( ^ ols jX)] be positive semi-de…nite.
Letting c0 = (0; 1; 1; 1; 0; :::; 0) gives result (i), while letting c0 = (0; 1; 0; 0; :::; 0) gives result
(ii).
(iii) the OLS estimators may be BLUE, but the product of two of them is a nonlinear
function and hence it can never be shown to be BLUE for the true product.
(iv) the LAD is a nonlinear and biased estimator in …nite samples, so it can never be
BLUE.
(v) the LAD is nonlinear and biased as in (iv), but in addition now we consider the
*product* of two estimators which is a nonlinear function. Hence no BLUE results can
obtain on it.
(b) (6 points)
Now consider the above …ve statements with “BLUE”changed for “CUAN”(which stands
for Consistent, Uniformly Asymptotically Normal). What assumptions about the true
regression model would be required to guarantee all …ve statements?
ANSWER:
The OLS estimators will be CUAN provided A1, A2, A3Rsru or stronger, either A4GM
or A4 , and S ! 1. Then (i) and (ii) will hold automatically and (iii) will also hold by
Slutsky’s Continuous Mapping Theorem, since the product is a continuous function.
For (iv) we require the LAD estimator to be CUAN and for (v) for it to be CUAN
plus Slutsky’s Continuous Mapping Theorem. For the LAD to be CUAN, we require
stronger conditions than those for OLS to be CUAN, e.g., A3Rf i (fully independence of
the regressors from the errors for all rows, not just the same rows); A4GM iid and not
just A4GM or A4 ; and more restrictive A5 to guarantee that the CUAN results follow.
(c) (8 points)
Finally, change BLUE to "Best CUAN". What assumptions about the true regression
model would guarantee statements (i)-(iii), and what assumptions would guarantee state-
ments (iv)-(v)? Can all …ve statements hold at the same time?
Hint: Think in terms of Maximum Likelihood Estimation (MLE).
ANSWER:
The Best CUAN estimator for a “regular”estimation problem will be the MLE. The OLS
estimator will be the MLE under assumptions:
A1, A2, A3Rmi or stronger; A4GM (iid), and A5G : s jX N (:; :). This would guarantee
results (i)-(iii).
The LAD estimator will be the MLE under assumptions:
A1, A2, A3Rmi or stronger; A4GM iid, and A5DE : s jX DoubleExponential/Laplace(:; :)
with pdf / exp( c j s j). In such case, results (iv)-(v) will hold.
Clearly, the two sets of conditions cannot both be true at the same time.

c LSE LT 2017/EC402 Page 11 of 17

 7. [20 points] A researcher considers analyzing production function models using a
cross-sectional data set indexed by i = 1; ; N . Her data consists of observations on output
of …rm i indexed by Yi and the …rm’s two input variables, Capital Ki and Labour Li .
The researcher considers two alternative production function regression models, the
“Cobb-Douglas”(CD) production function and the “Constant Elasticity of Substitution”(CES)
production function. She models these two production functions through the relations

2
CD : Yi = 1 Ki Li 3 i

and
4 4 1= 4
CES : Yi = 1 2 Ki + 3 Li + ui

(a) (4 points)
Show how the researcher can transform the CD production function to yield a linear
regression model. Is a similar approach possible for the CES case?
ANSWER:
Taking logs gives

CD : ln Yi = ln 1 + 2 ln Ki + 3 ln Li + ln i

which is a linear regression model in the logged inputs and an additive logged error term.
There does not exist a transformation that will similarly operate on the CES function to
render it a linear regression, nor any other method exists to achive this. Hence the CES
must be analyzed as an intrinsically nonlinear regression model.
(b) (6 points)
How should the researcher estimate the unknown parameters 1 , 2 , and 3 for the CD
production function, and the 1 , 2 , 3 , and 4 parameters for the CES function?
ANSWER:
The logged CD model can be estimated by OLS to give BLUE estimators for ln 1 , 2 , and
3 under GM conditions. The estimators will also be BUE if the logged disturbance ln i
satis…es A5Gaussian, meaning that the original i was lognormally distributed. Note
that we cannot get BLUE or BUE for 1 , only for ln 1 .
Under A1, A2 : CES, A3Rsru, and A4GM iid, Nonlinear Least Squares (NLLS) will
provide CUAN estimates for the 1 , 2 , 3 , and 4 parameters for the CES function, by
solving:
XS n o2
4 4 1= 4
min ys 1 K
2 i + L
3 i
s=1

(c) (6 points)

c LSE LT 2017/EC402 Page 12 of 17

 The hypothesis of “Constant Returns to Scale”(CRS) corresponds to 2 + 3 = 1 for the
CD case and to 2 + 3 = 1 for CES. How should the researcher test formally CRS in the
two cases?
ANSWER:
ols ols
Use the usual t-ratio based on the OLS estimates ^ 2 + ^ 3 1 for the CD case and based
nlls nlls
on the NLLS estimates ^ 2 + ^ 3 1. OLS will be BLUE or BUE in …nite samples; and
if BUE when A5G, we’ll have the NLRM (normal linear regression model) with the exact
t-distribution with S k degrees of freedom.
In the CES, the NLLS will be CUAN so the tratio in the gammas will be approximately
t-distributed in very large samples.
(d) (4 points)
Explain how she should modify her estimation strategies in case it was believed that
Labour input Li is endogenous.
ANSWER:
For the CD, OLS estimation should be replaced by IVE after …nding a suitable (valid and
relevant) instrument variable for Li .
For the CES case, NLLS estimation should be replaced by GMM (or NLIVE) after …nding
suitable IV as for CD.

c LSE LT 2017/EC402 Page 13 of 17

 8. [20 points] Consider a cross-sectional data set of S households indexed by s = 1; ; S. We
wish to study a linear regression model with average consumption per family member as the
dependent variable yi , as a function of household characteristics like household income (xs2 ),
years of schooling of the household head (xs3 ), gender of the household head (xs4 ), and other
relevant regressor variables. We use k regressor variables in total. Hence we postulate:

ys = x0s + s

with xs and being k-dimensional column vectors, i.e., in vector/matrix notation:

y=X +

where y and are S-dimensional column vectors and X an S k matrix.

In view of the fact that yi is the average consumption per family member and assuming that
the error term does not exhibit autocorrelation, a sensible model for the variance/covariance
properties of the unobservable error is:
2 2 2
V Cov( jX) = diag( =n1 ; ; =ns ; ; =nS )

where ns denotes the number of members in household s.

(a) (7 points)
Assuming that the number of members in each household (ns ) are known, explain which
would be the Best Linear Unbiased estimator (BLUE) for the unknown coe¢ cients . Are
any further assumptions necessary to guarantee that the estimator you propose is BLUE?
What if you wished to derive the Best Unbiased Estimator (BUE) for ?
ANSWER:
Given that the true in the VCov(.) of the error term is known in this case, the Ideal
Generalized Least Squares estimator, given by

(X 0 1
X) 1
X0 1
y

will be the BLUE by the Generalized Gauss Markov theorem, where = diag(1=n1 ; ; 1=ns ; ; 1=n
If we want the BUE, we would need to assume also that A5Gaussian holds for the true
disturbance, since then the IGLS estimator coincides with MLE, and hence it is BUE
because it is unbiased (under A3Rmi or stronger).
(b) (8 points)
Now suppose that the number of members in each household (ns ) are not known. A
colleague suggests that the variance/covariance of the error be modelled instead by:
2 2 2
V Cov( jX) = diag( 1 + 2 x12 ; ; 1 + 2 xs2 ; ; 1 + 2 xS2 )

In other words, the variance of error s is postulated to be a linear function of the square
of that household’s income, where 1 and 2 are unknown parameters. Which estimation

c LSE LT 2017/EC402 Page 14 of 17

 approach would you now propose for the unknown coe¢ cients? What properties would
this estimator have?
ANSWER:
Since now depends on the unknown gammas, IGLS is infeasible, and so the FeasibleGLS
must be used instead whereby …rst-step gamma estimates are obtained, and then in a
second step ^ using the estimated gammas is plugged into the GLS formula.
Alternatively, if we assume also A5Gaussian, then full MLE could be applied that esti-
mates simultaneously the beta and gamma parameters.
The two methods are asymptotically equivalent — and are both Best CUAN for very large
sample sizes.
(c) (5 points)
Explain how you can test formally whether or not the error term in part (b) is homoskedas-
tic.
ANSWER:
Within the maintained hypothesis that the errors follow the diagonal given above,
homoskedasticity corresponds to the null hypothesis with the *single* restriction: H0 :
2 = 0.
The Wald approach for testing this would correspond to a simple t-ratio test on the
FGLS/MLE estimate for gamma2.

c LSE LT 2017/EC402 Page 15 of 17

 9. [20 points] Consider a linear regression model with data (y; X) satisfying the following as-
sumptions:
A1 : rank(X) = k < S
A2 : y=X + with E = 0
A3Rmi : E( jX) = E
A4 : E( 0 jX) = c2
A5G : s jX N (0; 2 )
where S is the sample size and k is the number of regressors in X.
Consider the following situations, and explain which assumption or assumptions would be
modi…ed in each case, and what the implications for estimation would be:

(a) (4 points)
The second regressor is the …rst lag of the dependent variable, i.e., xs2 = ys 1 , and the
symmetric, positive de…nite matrix with all elements being real and …nite, in fact equals
the identity matrix of order S, i.e., = IS .
ANSWER:
A3 needs to be modi…ed to the Weak Exogeneity assumption A3Rsru since the fact that
the lag of the dep.variable is a regressor means that we can no longer condition on the
full X matrix. In view of A4GM now holding, no complication arises and so OLS will
be CUAN. (It will *not* be unbiased since A3Rmi no longer holds.)
(b) (4 points)
The second regressor is the …rst lag of the dependent variable, i.e., xs2 = ys 1 , and the
symmetric, positive de…nite matrix with all elements being real and …nite, has every
o¤-diagonal element di¤erent from zero.
ANSWER:
Now the lagged-y regressor xs2 is *endogenous* w.r.t. to the autocorrelated disturbance
since xs2 depends on the lagged disturbance which is correlated with today’s disturbance
because of the autocorrelation. So OLS will be *inconsistent* in this case.
(c) (4 points)
(
1 if observation s is male
The second regressor is a male dummy, i.e., xs2 = , while
0 otherwise
(
1 if observation s is female
the third regressor is a female dummy xs3 = .
0 otherwise
ANSWER:
Given this de…nition, there is a perfect linear relationship between regressor 2 and regressor
3. since they add up to 1 for any s. This is not necessarily a problem *unless* the RHS
also contains a *constant* regressor — that would lead to perfect multicollinearity among
the regressors and hence A1 will be violated. Without A1, the OLS estimator cannot be

c LSE LT 2017/EC402 Page 16 of 17

 computed. In sum, the regression can contain only *two* of the regressors xs2 , xs3 , and
an intercept/constant regressor.
(Recall: if A4GM , work with OLS estimator. If A4 work with suitable GLS etc.)
(d) (4 points)
The disturbance s follows the “Cauchy” distribution, which means that every moment
of s (E s , E 2s , E 3s , etc.) are all not …nite.
ANSWER:
This will mean that A2 cannot hold since that requires E s = 0; that A4 cannot hold
because all versions of A4 require that process possesses …nite variances and covariances;
and A3Rsru cannot hold because corr( s ; xsj ) cannot be zero since it is not …nite in the
Cauchy case.
When A4GM iid, CUAN (and in fact Best CUAN) estimation is available in terms of full
MLE using the density of the Cauchy distribution to de…ne the Likelihood function. But
OLS will be poorly behaved, even asymptotically: in the simplest regression case with
just an intercept to estimate, OLS is the sample mean of the dependent variable — but
this remains Cauchy distributed for *any* sample size, even in…nite!
(e) (4 points)
The disturbance s follows the “Uniform” distribution with support [ q; q] where q > 0.
ANSWER:
OLS would be BLUE and CUAN, but not BUE and not Best CUAN.
If we have A4GM iid, the Best CUAN will be full MLE using the A5U nif orm assump-
tion to derive the Likelihood function (by multiplying together the densities in view of
A4GM iid).

c LSE LT 2017/EC402 Page 17 of 17