notes on

UNIVERSIDAD CARLOS III DE MADRID
ECONOMETRICS
Academic year 2009/10
FINAL EXAM (2nd Call)
June, 25, 2010
Very important: Take into account that:
1. Each question, unless otherwise stated, requires a complete analysis of all the outputs shown in the
corresponding problem.
For example, to answer those questions referring to “appropriate estimates”, or “given the estimates”or
“given the problem conditions”, the results based on the consistent and more e¢ cient among outputs,
must be used.
2. The questions have been randomly ordered.
3. Each output includes all the explanatory variables used in the corresponding estimation.
4. Some results in the output shown may have been omitted.
5. The dependent variable can vary among outputs within the same problem.
6. For the sake of brevity, we will say that a model is well speci…ed if it is linear in the conditioning
variables (as they appear in the model) and its error term is mean-independent of such variables.
7. OLS, and 2SLS or TSLS, are the corresponding abreviations of ordinary least squares and two stage
least squares, respectively.
8. Statistical tables are included at the end of this document.
Problem: Determinants of fertility.
We would like to study the determinants of the total woman’s number of children (KIDS). We are
interested in knowing if fertility rates (meaning the average number of children per woman) have changed
over time. We have a sample of 476 women from the General Social Survey of the National Opinion Research
Center (NORC) in the USA for the years 1972, 1978 and 1984.
The characteristics of the woman we are interested in are EDU C (Years of education), AGE (Age, in
years), AGE 2 (Age squared), BLACK (Binary variable that takes the value of 1 if the woman is black and
0 otherwise).
Moreover, in order to consider the possibility that fertility rates change over time, we have the variables
Y EAR (The year corresponding to the observation; this variable takes three possible values: 72, 78 or
84); Y 72 (Binary variable that takes the value of 1 if the observation corresponds to the year 1972 and 0
otherwise); Y 78 (Binary variable that takes the value of 1 if the observation corresponds to the year 1978
and 0 otherwise); Y 84 (Binary variable that takes the value of 1 if the observation corresponds to the year
1984 and 0 otherwise).
Also, we take into account the possibility of interacting Y EAR with education, (Y EAR EDU C).
The following models are considered to analyze the determinants of the number of children women have:
2
KIDS = 0 + 1 AGE + 2 AGE + 3 BLACK + 4 EDU C + 5Y EAR + "1 (I)
2
KIDS = 0 + 1 AGE + 2 AGE + 3 BLACK + 4 EDU C + 5Y 78 + 6Y 84 + "2 (II)
KIDS = 0+ 1 AGE + 2 AGE 2 + 3 BLACK + 4 EDU C + 5Y EAR (III)

+ 6 (Y EAR EDU C) + "3
We also observe two further binary variables, RU RAL (that takes the value of 1 if the woman lived in a
rural area in her teens and 0 otherwise) and LP OP (that takes the value of 1 if the woman lived in a highly
populated area in her teens and 0 otherwise). Of course, interactions of Y EAR with these two variables,
(Y EAR RU RAL) y (Y EAR LP OP ), can also be considered.
The results of the various estimations are presented below:
Output 1: OLS, using observations 1–476

Dependent variable: KIDS
Coe¢ cient Std. Error t-ratio p-value
const 2:1966 5.0370 0:4361 0.6630
AGE 0:4788 0.2178 2:1982 0.0284
AGE 2 0:0054 0.0025 2:1862 0.0293
BLACK 0:3640 0.2929 1:2429 0.2145
EDU C 0:1381 0.0298 4:6403 0.0000
Y EAR 0:0489 0.0152 3:2135 0.0014
Mean dependent var 2.67 S.D. dependent var 1.67
Sum squared resid 1197.9 S.E. of regression 1.60
R2 0.0993 Adjusted R2 0.0897
F (5; 470) 10.36 P-value(F ) 1.9e–09
2
const 6:0500 4.8054 1:2590 0.2087
AGE 0:4908 0.2179 2:2518 0.0248
AGE 2 0:0055 0.0025 2:2398 0.0256
BLACK 0:3814 0.2931 1:3014 0.1938
EDU C 0:1374 0.0298 4:6184 0.0000
Y 78 0:1001 0.1871 0:5351 0.5929
Y 84 0:5794 0.1827 3:1706 0.0016
R2 0.1020 Adjusted R2 0.0905
F (6; 469) 8.87 P-value(F ) 3.5e–09
Coe¢ cient covariance matrix (Output 2)

AGE AGE 2 BLACK EDU C Y 78 Y 84
? 0:0005 0:0013 0:0007 0:0034 0:0036 AGE
? 1:4 10 5 7:4 10 6 3:6 10 5 3:6 10 5 AGE 2
? 0 0:0030 0:0012 BLACK
? 0:0003 0:0008 EDU C
? 0:0180 Y 78
? Y 84

const 15:5816 7.4233 2:0990 0.0363
AGE 0:4401 0.2172 2:0261 0.0433
AGE 2 0:0050 0.0025 2:0148 0.0445
BLACK 0:3984 0.2917 1:3660 0.1726
EDU C 0:9904 0.4627 2:1403 0.0328
Y EAR 0:1321 0.0756 1:7473 0.0812
(Y EAR EDU C) 0:0143 0.0059 2:4438 0.0149

R2 0.1106 Adjusted R2 0.0992
F (6; 469) 9.7 P-value(F ) 4.2e–10

AGE AGE 2 BLACK EDUC YEAR (YEAR EDUC)
? 0:0005 0:0009 0:0066 0:0009 9:3 10 5 AGE
6 5 5
? 9:3 10 7:5 10 1 10 1 10 6 AGE 2
? 0:0069884 0:0011365 8:3 10 5 BLACK
? 0:034144 0:0027 EDUC
? 0:0004 YEAR
? (YEAR EDUC)
3
Output 4: TSLS, using observations 1–476
Instrumented: EDU C
Instruments: const AGE AGE 2 BLACK Y EAR RU RAL LP OP (Y EAR RU RAL) (Y EAR LP OP )
Coe¢ cient Std. Error z-stat p-value
const 43:3158 34.9236 1:2403 0.2149
AGE 0:6207 0.2728 2:2749 0.0229
AGE 2 0:0069 0.0031 2:2416 0.0250
BLACK 0:6224 0.3591 1:7331 0.0831
EDU C 3:0089 2.8246 1:0652 0.2868
Y EAR 0:3817 0.4369 0:8736 0.3824
(Y EAR EDU C) 0:0360 0.0350 1:0299 0.3031

R2 0.0061 Adjusted R2 0.0066
F (6; 469) 4.28 P-value(F ) 0.00033

AGE AGE 2 BLACK EDUC YEAR (YEAR EDUC)
? 0:0008 0:0058 0:1016 0:0195 0:0015 AGE
? 5:7 10 5 0:0012 0:0002 1:8 10 5 AGE 2
? 0:3550 0:0521 0:0042 BLACK
? 1:2291 0:0986 EDUC
? 0:0153 YEAR
? (YEAR EDUC)
Output 5A: OLS, using observations 1–476

Dependent variable: EDU C
const 15:7871 8.2430 1:9152 0.0561
AGE 0:6551 0.3326 1:9693 0.0495
AGE 2 0:0071 0.0038 1:8905 0.0593
BLACK 0:3120 0.4519 0:6904 0.4903
Y EAR 0:1492 0.0358 4:1732 0.0000
RU RAL 11:1516 4.1572 2:6825 0.0076
LP OP 5:0678 3.6998 1:3698 0.1714
(Y EAR RU RAL) 0:1509 0.0530 2:8455 0.0046
(Y EAR LP OP ) 0:0586 0.0473 1:2389 0.2160
R2 0.0974 Adjusted R2 0.0819
F (8; 467) 6.3 P-value(F ) 9.2e–08
Output 5B: OLS, using observations 1–476

Dependent variable: EDU C
const 24:0167 7.7205 3:1108 0.0020
AGE 0:7663 0.3354 2:2849 0.0228
AGE 2 0:0083 0.0038 2:1778 0.0299
BLACK 0:5512 0.4528 1:2174 0.2241
Y EAR 0:0779 0.0233 3:3433 0.0009
4
R2 0.0530 Adjusted R2 0.0449
F (4; 471) 6.6 P-value(F ) 0.00004
Output 5C: OLS, using observations 1–476

Dependent variable: (Y EAR EDU C)
const 322:8543 649.5532 0:4970 0.6194
AGE 54:4209 26.2127 2:0761 0.0384
AGE 2 0:5941 0.2974 1:9977 0.0463
BLACK 22:7724 35.6073 0:6395 0.5228
Y EAR 24:2667 2.8177 8:6124 0.0000
RU RAL 907:9992 327.5882 2:7718 0.0058
LP OP 355:6616 291.5432 1:2199 0.2231
(Y EAR RU RAL) 12:3139 4.1784 2:9471 0.0034
(Y EAR LP OP ) 4:0834 3.7263 1:0958 0.2737
Sum squared resid 17033368 S.E. of regression 190.98
R2 0.2610 Adjusted R2 0.2484
F (8; 467) 20.6 P-value(F ) 8.4e–27
Output 5D: OLS, using observations 1–476

Dependent variable: (Y EAR EDU C)
const 957:0200 609.1059 1:5712 0.1168
AGE 63:0297 26.4602 2:3821 0.0176
AGE 2 0:6831 0.3003 2:2748 0.0234
BLACK 40:9939 35.7203 1:1476 0.2517
Y EAR 18:7665 1.8383 10:2087 0.0000
Sum squared resid 17914806 S.E. of regression 195.03
R2 0.2228 Adjusted R2 0.2162
F (4; 471) 33.7 P-value(F ) 8.8e–25

const 43:3158 30.9547 1:3993 0.1624
AGE 0:6207 0.2418 2:5665 0.0106
AGE 2 0:0069 0.0027 2:5291 0.0118
BLACK 0:6224 0.3183 1:9553 0.0511
EDU C 3:0089 2.5036 1:2018 0.2300
Y EAR 0:3817 0.3873 0:9856 0.3249
(Y EAR EDU C) 0:0360 0.0310 1:1619 0.2459
RES5A 2:0124 2.5476 0:7899 0.4300
RES5C 0:0214 0.0316 0:6790 0.4975
NOTA: RES5A y RES5C son los respectivos residuos de las Output 5A y 5C.
5
R2 0.1250 Adjusted R2

Dependent variable: RES4
const 0:9642 6.0759 0:1587 0.8740
AGE 0:0135 0.2452 0:0551 0.9561
AGE 2 0:0001 0.0028 0:0536 0.9572
BLACK 0:0349 0.3331 0:1047 0.9166
Y EAR 0:0098 0.0264 0:3716 0.7104
RU RAL 0:8046 3.0643 0:2626 0.7930
LP OP 2:3062 2.7271 0:8456 0.3982
(Y EAR RU RAL) 0:0082 0.0391 0:2105 0.8334
(Y EAR LP OP ) 0:0315 0.0349 0:9027 0.3671
NOTA: RES4 son los residuos de la Salida 4.
R2 0.0046 Adjusted R2 0.0125
F (8; 467) 0.2689 P-value(F ) 0.9757
6
Universidad Carlos III de Madrid
ECONOMETRIC
June, 25, 2010
Tipo de examen: 1
DURATION: 125 minutes
Directions:
• BEFORE YOU START TO ANSWER THE EXAM:
– Fill in your personal data in the optical reading form, which will be the only valid answering
document. Remember that you must complete all your identifying information (name and
surname(s), and NIU, which has 9 digit and always begins by 1000) both in letters and in the
corresponding optical reading boxes.
– Fill in, both in letters and in the corresponding optical reading boxes, the course code and your
group.
• Check that this document contains 48 questions sequentially numbered.
• Check that the exam type that appears in the questionnaire matches the one indicated in the optical
reading form.
• Read the problem text and the questions carefully.
• For each row regarding the number of each question, fill the box which corresponds with your answer
in the optical reading form (A, B, or C).
• Each question has one correct answer.

Incorrect answers will be graded with zero points.
Questions with more than one answer will be considered incorrect and its score will be zero.
• To obtain a grade of 5 over 10 you must correctly answer 28 questions.
• If you wish, you may use the answer table as a draft, although such table does not have any official
validity.
• You can use the back side of the problem text as a draft (no additional sheets will be handed out).
• Any student who were found talking or sharing any sort of material during the exam
will be expelled out immediately and his/her overall score will be zero, independently
of any other measure that could be undertaken.
• Dates of grades publication: Monday, June, 28.
• Date and place of exam revision: Wednesday, June, 30, at 14:30 h. Rooms 15.1.39, 15.1.41 and
15.1.43.
• Rules for exam revision:
– Its only purpose will be that each student:

∗ check the number of correct answers;
∗ handout in writing, if (s)he wants, the possible claims about the problem text and the
questions, that will be attended by writing in the next 10 days since the revision date.
– To be entitled for revision, the student should bring a printed copy of the exam solutions, which
will be available in Aula Global from the date of grade publication.
Answer DRAFT
(a) (b) (c) (a) (b) (c) (a) (b) (c) (a) (b) (c)
1. 13. 25. 37.
2. 14. 26. 38.
3. 15. 27. 39.
4. 16. 28. 40.
5. 17. 29. 41.
6. 18. 30. 42.
7. 19. 31. 43.
8. 20. 32. 44.
9. 21. 33. 45.
10. 22. 34. 46.
11. 23. 35. 47.
12. 24. 36. 48.
Exam type: 1 página 2

1. Assume that model (III) verifies the assumptions of the classical regression model. If we
wanted to test that the mean number of children does not depend on the year, the null
hypothesis is:
(a) H0 : γ5 = γ6 .
(b) H0 : γ5 = γ6 = 0.
(c) H0 : γ5 = 0.
2. Assume that model (II) verifies the assumptions of the classical regression model. If the race
(BLACK) was an irrelevant variable, the variance of the estimators of the coefficients of the
relevant variables will be higher:
(a) The higher the proportion of black women in the sample.
(b) The higher the correlation of BLACK with the relevant variables.
(c) The lower the correlation of BLACK with the relevant variables.
3. Assume that the error of model (II) satisfies E (ε2 | AGE,BLACK,EDU C,Y 78,Y 84) 6= 0 for
any combination of values of the conditioning variables, while the homoskedasticity does not
hold. Additionally, we have four additional variables Z1 , Z2 , Z3 , Z4 , non included in the
model and uncorrelated with ε2 . Then, in any case:
(a) If we estimated the model (II) by OLS including Z1 , Z2 , Z3 , Z4 as additional variables,
the estimators obtained for the coefficients δ1 , δ2 , δ3 , δ4 , would be consistent.
(b) If we estimated the model (II) by 2SLS using Z1 , Z2 , Z3 , Z4 as instruments, the
estimators obtained for the coefficients δ1 , δ2 , δ3 , δ4 , would be consistent.
(c) If we estimated the model (II) by OLS, the estimators obtained for the coefficients δ1 ,
δ2 , δ3 , δ4 , would be inconsistent.
4. If a relevant variable were omitted in model (I), estimates in Output 1 will be:
(a) None of the other statements is true.
(b) Always inconsistent.
(c) Consistent, provided that the omitted variable is uncorrelated with the remaining
explanatory variables in the model.
5. Suppose that we can ensure that AGE, BLACK and, of course, Y EAR, are uncorrelated with
ε2 . Also, assume that RU RAL and LP OP are uncorrelated with ε2 . If we had estimated
model (II) by 2SLS but using RU RAL as the only instrument for EDU C, the estimators
obtained for model (II) parameters:
(a) The Gretl program would indicate us that we do not have enough instruments.
(b) Would be inconsistent.
(c) Would be less efficient than the ones by 2SLS estimator but using both RU RAL and
LP OP as instruments.
6. Suppose that we are interested in model (III). The information provided by Output 6 allows
to assess whether:
(a) RU RAL is a better instrument than LP OP .
(b) We can reject the null hypothesis about the exogeneity of education.
(c) We can reject the null hypothesis about instruments’ validity.

7. In model (III), suppose that education is endogenous. Given the instruments used in Output
4, we can assert that the model is:
(a) Over-identified, the number of over-identifying restrictions being equal to 2.
(b) Exactly identified.
(c) Over-identified, the number of over-identifying restrictions being equal to 1.
8. Assume that model (II) verifies the assumptions of the classical regression model. If we wanted
to test that the causal effect of age on the number of children is constant, the null hypothesis
would be:
(a) H0 : δ1 − δ2 = 0.
(b) H0 : δ1 = δ2 = 0.
(c) H0 : δ2 = 0.
9. Comparing models (I), (II) and (III):
(a) Models (I) and (II) are not comparable, because none of them is a particular case of
the other.
(b) Model (I) is the most restrictive.
(c) Model (III) is the least restrictive.
10. Comparing models (I) and (II), model (II) can be expressed as model (I) with the following
restriction:
(a) δ6 = 6δ5 .
(b) δ6 = δ5 .
(c) δ6 = 2δ5 .
11. Assume that model (I) verifies the assumptions of the classical regression model. If the variable
EDU C was measured with error (and such error had strictly positive variance), estimates in
Output 1 will be:
(a) Consistent, but less efficient than if the variable were measured without error.
(c) Inconsistent, only if the measurement error is correlated with the error term of the
model.
12. Assume that model (I) verifies the assumptions of the classical regression model. Consider
two white women in the same year and with the same age, but the first one having 7 years
of education less than the second one. The first woman, on average, will have approximately
(rounding to the closest integer number):
(a) 1 child more than the second one.
(b) 1 child less than the second one.
(c) The same number of children than the second one.

13. If education was an endogenous variable:
(b) The estimated coefficients in Output 1 would be inconsistent for model (I), but those
of Output 2 would be consistent for model (II).
(c) The estimated coefficients in Output 2 would be inconsistent for model (II), but those
of Output 3 would not be inconsistent for model (III).
14. If RU RAL and LP OP were uncorrelated with ε3 , and we wanted to test that the variables
RU RAL and LP OP are valid instruments for EDU C, the test statistic would approximately
be:
(a) 7.8.
(b) 51.4.
(c) 23.4.
15. Assume that model (I) verifies the assumptions of the classical regression model. An appro-
priate estimate of V (KIDS| AGE, BLACK,EDU C,Y EAR) (rounded to 1 decimal), is:
(a) 1.6.
(b) 2.8.
(c) 2.6.
16. Assume that model (III) verifies the assumptions of the classical regression model. Under the
evidence of Output 3, and taking into account only women younger than 40 years of age:
(a) The causal effect of education is the same for all women considered.
(b) More educated women have, on average, more children.
(c) Older women have, on average, more children.
17. Assume that model (I) verifies the assumptions of the classical regression model. Under the
evidence of Output 1, for a given age, race and education level (rounding to 1 decimal):
(a) A woman in 1978 had on average 0.3 children less than a woman in 1984.
(b) A woman in 1984 had on average 0.6 children less than a woman in 1972.
(c) A woman in 1978 had on average 0.3 children more than a woman in 1972.
wanted to test that education does not affect fertility, the null hypothesis is:
(a) H0 : γ4 = 0.

γ4 − γ6 = 0
(b) H0 : .
γ6 = 0
(c) H0 : γ4 = γ6 .
19. Suppose that we are interested in model (II). Consider the following statement: “In 1972, the
fertility rate of a 30 year old black woman with 10 years of education is equal to the fertility
rate of black woman with the same educational level but 1 year younger”. Given the results
(rounding to 2 decimals):
(a) The test statistic is approximately t = 1.64.
(b) The test statistic is approximately t = 0.06.
(c) The test statistic is approximately t = 0.63.
20. The 2SLS estimated coefficients in Output 4 could have been obtained in an equivalent way:
(a) Estimating by OLS a model with KIDS as the dependent variable and including as
regressors all the exogenous explanatory variables and the corresponding predictions of
education and its interaction with Y EAR, by using Outputs 5A and 5C, respectively.
(b) Estimating by OLS a model with KIDS as the dependent variable and including as
education and its interaction with Y EAR, by using Outputs 5B and 5D, respectively.
(c) Estimating by OLS a model with KIDS as the dependent variable and including as
regressors all the exogenous explanatory variables, as well as the variables RU RAL
and LP OP and their corresponding interactions with Y EAR.
4, if we wanted to perform a test of over-identifying restrictions, the appropriate test statistic
would be:
(a) The R2 from Output 4 multiplied by the number of observations.
(b) The R2 from Output 7 multiplied by the number of observations.
(c) The R2 from Output 5A multiplied by the number of observations.
wanted to test whether the causal effect of education for women observed in 1978 is the same
as on women observed in 1984, the null hypothesis would be:
(a) γ4 = γ6 = 0.
(b) γ6 = 0.
(c) γ4 = γ5 = γ6 .
wanted to test that the mean number of children does not depend on the year:
(a) The question cannot be answered with the provided information.
(b) We do not reject it, because the p-value of the corresponding statistic is equal to 0.
(c) We reject it, given the the magnitude of the corresponding statistic obtained when
comparing the unrestricted model and the model that imposes such restriction.
24. Comparing models (I) and (II):
(a) Models (I) and (II) are not comparable, since model (I) includes variables that model
(II) does not, and vice versa.
(b) Model (I) is more restrictive, as it imposes that the effect of education on the number
of children in 1972 is null.
(c) Model (II) is less restrictive, as it allows, for a given race, age, and education level, the
fertility rate to change differently over time.

(b) Estimating by 2SLS a model with KIDS as the dependent variable and including as
regressors all the exogenous explanatory variables, the education and its interaction
with Y EAR, using as instruments the predictions based on Outputs 5B and 5D.
(c) Estimating by 2SLS a model with KIDS as the dependent variable and including as
with Y EAR, using as instruments the predictions based on Outputs 5A and 5C.
wanted to test whether the mean number of children in 1972 is the same for a black woman
with 10 years of education than the one for a black woman with the same age but with 12
years of education, the null hypothesis would be:
(a) H0 : γ4 = −144γ6 .
(b) H0 : γ4 + 72γ6 = 0.
(c) H0 : 2γ4 + 72γ6 = 0.
rate of black woman with the same educational level but 1 year younger”. Given the results:
(a) We cannot reject such assertion at the 5% significance level.
(b) At the 1% significance level, we can reject such assertion.
(c) At the 5% significance level, we can reject such assertion.
28. In model (III), suppose that education is endogenous. Results in Output 7 allow us to test
whether:
(a) Education is not correlated with the instruments used in Output 4.
(b) None of the instruments used in Output 4 is correlated with ε3 .
(c) Education is not correlated with ε3 .
29. Focusing on models (I) and (II):
(a) None of the other statements are correct.
(b) Model (II) is misspecified, since it omits the variable Y 72.
(c) Model (I) is a particular case of model (II).
30. Assume that model (I) verifies the assumptions of the classical regression model. Given the
estimated coefficients, for a given age, educational level and race:
(a) For every 50 women, there are approximately 2.4 more children in 1972 than in 1984.
(b) For every 50 women, there are approximately 29.3 more children in 1972 than in 1984.
(c) A woman in 1978 has approximately 29.3% less children than a woman in 1972.

evidence of Output 1, the effect of education on the number of children is:
(b) Negative (on average) for all the women in the sample.
(c) Positive (on average) for all the black women in the sample, as the coefficient of
BLACK is higher in absolute value than the coefficient of EDU C.
32. Suppose that we are interested in model (III). Given the results:
(a) We do not reject that the correlation of the instruments with the error in model (III)
is equal to zero.
(b) We reject that EDU C is exogenous.
(c) We do not reject that the correlation of the instruments with EDU C is equal to zero.
33. Assume that model (III) verifies the assumptions of the classical regression model. Given the
estimated coefficients (and rounding to two decimals), for a 20 year old white woman with 10
years of education, in 1972, the mean number of children is approximately:
(a) 0.98.
(b) 0.34.
(c) 1.12.
34. We want to obtain the causal effect of education on the number of children for model (III). If
education was an endogenous variable, given the available information:
(a) None of the other statements are correct.
(b) We would need at least one variable, not included in the model and uncorrelated with
ε3 , to get consistent estimates of the parameters of interest by 2SLS.
(c) We would need at least two different variables, not included in the model and uncorre-
lated with ε3 , to get consistent estimates of the parameters of interest by 2SLS, since
model (III) includes the interaction of education with the variable Y EAR.
35. Suppose that we are interested in model (II) and its relation with model (I). Consider the
following conjecture: “For a given age, race, and educational level, the decrease in the fertility
rate is constant over time”. If such conjecture was true, it must occur that:
(a) δ5 = δ6 .
(b) The constant terms of both models are equal, β0 = δ0 .
(c) 6β5 = δ6 − δ5 .
4, and given the results provided:
(a) There is not information to assess whether the instruments are valid or not.
(b) We do not reject the instruments’ validity.
(c) We reject the instruments’ validity.

37. Assume that model (III) verifies the assumptions of the classical regression model. Given
the estimated coefficients, for a 20 year old black woman with 5 years of education the mean
number of children has decreased between 1978 and 1984 (rounding to 1 decimal) by:
(a) −0.8.
(b) −0.7.
(c) −0.4.
38. Using KIDS as dependent variable, consider models that include a constant, AGE, AGE 2 ,
BLACK and EDU C. Then:
(a) If we also included Y EAR, Y 78 and Y 84 as explanatory variables, such model would
be more general than model (I) or model (II).
(b) If we also included Y EAR and Y 78 as explanatory variables and estimate by OLS,
the R2 would be higher than the one in Output 2.
(c) If we also included Y EAR and Y 84 as explanatory variables and estimate by OLS,
the estimated coefficients of AGE, AGE 2 , BLACK and EDU C would be the same
than those in Output 2.
as on women observed in 1984, we can conclude that:
(a) We reject the null hypothesis at the 1% significance level.
(b) We do not reject the null hypothesis at the 5% significance level.
(c) We reject the null hypothesis at the 5%, but not at the 1% significance level.
40. Assuming that models (I) and (II) verify the assumptions of the classical regression model, if
race (BLACK) was an irrelevant variable:
(a) Output 1 and Output 2 will provide consistent estimates for the parameters of models
(I) and (II), respectively.
(b) Output 1 will provide inconsistent estimates for model (I) parameters.
(c) Output 2 will provide consistent estimates for model (II) parameters only if we dropped
the variable BLACK from the explanatory variables.
41. If education was an endogenous variable, in order to test whether both RU RAL and LP OP
are valid instruments, we would have to:
(a) In a regression of EDU C on the exogenous variables of the model and both instru-
ments, test whether such instruments are individually significant.
(b) In a regression of EDU C on the exogenous variables of the model and both instruments
and on their corresponding interactions with the variable Y EAR, test whether such
instruments and their corresponding interactions are jointly significant.
(c) Test the hypothesis that the residual of the reduced form (linear projection of EDU C
on the exogenous variables of the model and both instruments) has a significant effect
on education.

evidence of Output 3:
(a) The causal effect of education is more negative in 1978 than in 1984.
(b) The causal effect of education is positive.
(c) The causal effect of education is more negative in 1978 than in 1972.
evidence of Output 1, we can assert that the mean fertility rates:
(a) We do not have conclusive evidence.
(b) Have remained constant over time.
(c) Have decreased over time.
wanted to test that the causal effect of education on the number of children does not depend
on the year, the null hypothesis is:
(a) H0 : γ5 = γ6 = 0.
(b) H0 : γ6 = 0.
(c) H0 : γ4 = γ6 = 0.
as on women observed in 1984, the test statistic, in absolute value (rounding to 1 decimal)
would be:
(a) 14.7.
(b) 1.7.
(c) 2.4.
(a) Model (I) imposes the constraint that the coefficient of Y 78 is exactly half of the
coefficient of Y 84.
(b) Models (I) and (II) are different models, since none of them is a particular case of the
other one.
(c) Model (I) imposes the constraint that the coefficients of Y 78 and Y 84 were equals.
estimated coefficients, in 1972, the mean difference in the number of children between a black
and a white woman with the same age but the second one with 5 years of education less is
(rounding to 1 decimal):
(a) 4.5 less children.
(b) 0.2 more children.
(c) 5.3 more children.

estimated coefficients, the mean difference in the number of children between two women with
the same characteristics, but one in 1972 and the other in 1978, is:
(a) The question cannot be answered with the provided information.
(b) Significantly different from zero.
(c) Statistically equal to zero.

ECONOMETRIC
June, 25, 2010
Tipo de examen: 2
Directions:
group.
reading form.

validity.
15.1.43.

Answer DRAFT
(a) (b) (c) (a) (b) (c) (a) (b) (c) (a) (b) (c)
1. 13. 25. 37.
2. 14. 26. 38.
3. 15. 27. 39.
4. 16. 28. 40.
5. 17. 29. 41.
6. 18. 30. 42.
7. 19. 31. 43.
8. 20. 32. 44.
9. 21. 33. 45.
10. 22. 34. 46.
11. 23. 35. 47.
12. 24. 36. 48.

(a) We reject it, given the the magnitude of the corresponding statistic obtained when
(b) We do not reject it, because the p-value of the corresponding statistic is equal to 0.
(c) The question cannot be answered with the provided information.
(a) The same number of children than the second one.
(b) 1 child less than the second one.
(c) 1 child more than the second one.
be:
(a) 23.4.
(b) 51.4.
(c) 7.8.
would be:
(a) H0 : δ2 = 0.
(b) H0 : δ1 = δ2 = 0.
(c) H0 : δ1 − δ2 = 0.
(a) H0 : 2γ4 + 72γ6 = 0.
(b) H0 : γ4 + 72γ6 = 0.
(c) H0 : γ4 = −144γ6 .

(a) H0 : γ4 = γ6 = 0.
(b) H0 : γ6 = 0.
(c) H0 : γ5 = γ6 = 0.
would be:
(a) 2.4.
(b) 1.7.
(c) 14.7.
(a) If we estimated the model (II) by OLS, the estimators obtained for the coefficients δ1 ,
(b) If we estimated the model (II) by 2SLS using Z1 , Z2 , Z3 , Z4 as instruments, the
(c) If we estimated the model (II) by OLS including Z1 , Z2 , Z3 , Z4 as additional variables,
(a) Have decreased over time.
(b) Have remained constant over time.
(c) We do not have conclusive evidence.
(a) At the 5% significance level, we can reject such assertion.
(b) At the 1% significance level, we can reject such assertion.
(c) We cannot reject such assertion at the 5% significance level.
whether:
(a) Education is not correlated with ε3 .
(b) None of the instruments used in Output 4 is correlated with ε3 .
(c) Education is not correlated with the instruments used in Output 4.

(a) Older women have, on average, more children.
(b) More educated women have, on average, more children.
(c) The causal effect of education is the same for all women considered.
(a) Statistically equal to zero.
(b) Significantly different from zero.
(c) The question cannot be answered with the provided information.
(a) Model (II) is less restrictive, as it allows, for a given race, age, and education level, the
(b) Model (I) is more restrictive, as it imposes that the effect of education on the number
(c) Models (I) and (II) are not comparable, since model (I) includes variables that model
Output 1 will be:
(a) Inconsistent, only if the measurement error is correlated with the error term of the
model.
(c) Consistent, but less efficient than if the variable were measured without error.
(a) Consistent, provided that the omitted variable is uncorrelated with the remaining
(c) None of the other statements is true.
(a) Model (I) imposes the constraint that the coefficients of Y 78 and Y 84 were equals.
(b) Models (I) and (II) are different models, since none of them is a particular case of the
other one.
(c) Model (I) imposes the constraint that the coefficient of Y 78 is exactly half of the

(a) Would be less efficient than the ones by 2SLS estimator but using both RU RAL and
(b) Would be inconsistent.
(c) The Gretl program would indicate us that we do not have enough instruments.
(a) Model (I) is a particular case of model (II).
(b) Model (II) is misspecified, since it omits the variable Y 72.
(c) None of the other statements are correct.
(a) 2.6.
(b) 2.8.
(c) 1.6.
(a) Test the hypothesis that the residual of the reduced form (linear projection of EDU C
on education.
(b) In a regression of EDU C on the exogenous variables of the model and both instruments
(c) In a regression of EDU C on the exogenous variables of the model and both instru-

(a) Positive (on average) for all the black women in the sample, as the coefficient of
(b) Negative (on average) for all the women in the sample.
(a) Estimating by 2SLS a model with KIDS as the dependent variable and including as
(b) Estimating by 2SLS a model with KIDS as the dependent variable and including as
(b) The causal effect of education is positive.
(a) The lower the correlation of BLACK with the relevant variables.
(b) The higher the correlation of BLACK with the relevant variables.
(c) The higher the proportion of black women in the sample.
(a) We reject the instruments’ validity.
(b) We do not reject the instruments’ validity.
(c) There is not information to assess whether the instruments are valid or not.
(a) −0.4.
(b) −0.7.
(c) −0.8.

(a) If we also included Y EAR and Y 84 as explanatory variables and estimate by OLS,
(b) If we also included Y EAR and Y 78 as explanatory variables and estimate by OLS,
(c) If we also included Y EAR, Y 78 and Y 84 as explanatory variables, such model would
(b) Exactly identified.
hypothesis is:
(a) H0 : γ5 = 0.
(b) H0 : γ5 = γ6 = 0.
(c) H0 : γ5 = γ6 .
(a) We would need at least two different variables, not included in the model and uncorre-
(b) We would need at least one variable, not included in the model and uncorrelated with
(c) None of the other statements are correct.
(a) Output 2 will provide consistent estimates for model (II) parameters only if we dropped
(b) Output 1 will provide inconsistent estimates for model (I) parameters.
(c) Output 1 and Output 2 will provide consistent estimates for the parameters of models
(a) A woman in 1978 has approximately 29.3% less children than a woman in 1972.
(c) For every 50 women, there are approximately 2.4 more children in 1972 than in 1984.

(a) 1.12.
(b) 0.34.
(c) 0.98.
(a) We do not reject that the correlation of the instruments with EDU C is equal to zero.
(b) We reject that EDU C is exogenous.
(c) We do not reject that the correlation of the instruments with the error in model (III)
is equal to zero.
(a) 5.3 more children.
(b) 0.2 more children.
(c) 4.5 less children.
(a) 6β5 = δ6 − δ5 .
(b) The constant terms of both models are equal, β0 = δ0 .
(c) δ5 = δ6 .
restriction:
(a) δ6 = 2δ5 .
(b) δ6 = δ5 .
(c) δ6 = 6δ5 .
(a) The estimated coefficients in Output 2 would be inconsistent for model (II), but those
(b) The estimated coefficients in Output 1 would be inconsistent for model (I), but those
(a) A woman in 1978 had on average 0.3 children more than a woman in 1972.
(c) A woman in 1978 had on average 0.3 children less than a woman in 1984.
would be:
(a) The R2 from Output 5A multiplied by the number of observations.
(c) The R2 from Output 4 multiplied by the number of observations.
(a) H0 : γ4 = γ6 .

γ4 − γ6 = 0
(b) H0 : .
γ6 = 0
(c) H0 : γ4 = 0.
(a) We reject the null hypothesis at the 5%, but not at the 1% significance level.
(b) We do not reject the null hypothesis at the 5% significance level.
(c) We reject the null hypothesis at the 1% significance level.
(a) Model (III) is the least restrictive.
(b) Model (I) is the most restrictive.
(c) Models (I) and (II) are not comparable, because none of them is a particular case of
the other.
(a) γ4 = γ5 = γ6 .
(b) γ6 = 0.
(c) γ4 = γ6 = 0.
to assess whether:
(a) We can reject the null hypothesis about instruments’ validity.
(b) We can reject the null hypothesis about the exogeneity of education.
(c) RU RAL is a better instrument than LP OP .

ECONOMETRIC
June, 25, 2010
Tipo de examen: 3
Directions:
group.
reading form.

validity.
15.1.43.

Answer DRAFT
(a) (b) (c) (a) (b) (c) (a) (b) (c) (a) (b) (c)
1. 13. 25. 37.
2. 14. 26. 38.
3. 15. 27. 39.
4. 16. 28. 40.
5. 17. 29. 41.
6. 18. 30. 42.
7. 19. 31. 43.
8. 20. 32. 44.
9. 21. 33. 45.
10. 22. 34. 46.
11. 23. 35. 47.
12. 24. 36. 48.

(a) Model (II) is misspecified, since it omits the variable Y 72.
(b) None of the other statements are correct.
(c) Model (I) is a particular case of model (II).
(a) The higher the correlation of BLACK with the relevant variables.
(b) The higher the proportion of black women in the sample.
(c) The lower the correlation of BLACK with the relevant variables.
(a) In a regression of EDU C on the exogenous variables of the model and both instruments
(b) In a regression of EDU C on the exogenous variables of the model and both instru-
(c) Test the hypothesis that the residual of the reduced form (linear projection of EDU C
on education.
(a) More educated women have, on average, more children.
(b) The causal effect of education is the same for all women considered.
(c) Older women have, on average, more children.
would be:
(a) 1.7.
(b) 14.7.
(c) 2.4.
(a) H0 : γ6 = 0.
(b) H0 : γ5 = γ6 = 0.
(c) H0 : γ4 = γ6 = 0.

(a) Exactly identified.
(b) Over-identified, the number of over-identifying restrictions being equal to 2.
restriction:
(a) δ6 = δ5 .
(b) δ6 = 6δ5 .
(c) δ6 = 2δ5 .
(a) 1 child less than the second one.
(b) 1 child more than the second one.
(c) The same number of children than the second one.
(a) We do not reject the instruments’ validity.
(b) There is not information to assess whether the instruments are valid or not.
(c) We reject the instruments’ validity.
(b) None of the other statements is true.
(a) γ6 = 0.
(b) γ4 = γ6 = 0.
(c) γ4 = γ5 = γ6 .

(a) We reject that EDU C is exogenous.
(b) We do not reject that the correlation of the instruments with the error in model (III)
is equal to zero.
(c) We do not reject that the correlation of the instruments with EDU C is equal to zero.
(a) −0.7.
(b) −0.8.
(c) −0.4.

γ4 − γ6 = 0
(a) H0 : .
γ6 = 0
(b) H0 : γ4 = 0.
(c) H0 : γ4 = γ6 .
hypothesis is:
(a) H0 : γ5 = γ6 = 0.
(b) H0 : γ5 = γ6 .
(c) H0 : γ5 = 0.
(a) The estimated coefficients in Output 1 would be inconsistent for model (I), but those
(c) The estimated coefficients in Output 2 would be inconsistent for model (II), but those
(a) H0 : γ4 + 72γ6 = 0.
(b) H0 : γ4 = −144γ6 .
(c) H0 : 2γ4 + 72γ6 = 0.

(b) We cannot reject such assertion at the 5% significance level.
(a) For every 50 women, there are approximately 29.3 more children in 1972 than in 1984.
(c) A woman in 1978 has approximately 29.3% less children than a woman in 1972.
would be:
(a) H0 : δ1 = δ2 = 0.
(b) H0 : δ1 − δ2 = 0.
(c) H0 : δ2 = 0.
Output 1 will be:
(a) Always inconsistent.
(b) Consistent, but less efficient than if the variable were measured without error.
(c) Inconsistent, only if the measurement error is correlated with the error term of the
model.
(a) We do not reject it, because the p-value of the corresponding statistic is equal to 0.
(b) The question cannot be answered with the provided information.
(c) We reject it, given the the magnitude of the corresponding statistic obtained when

be:
(a) 51.4.
(b) 7.8.
(c) 23.4.
(b) 4.5 less children.
(a) Model (I) is the most restrictive.
(b) Models (I) and (II) are not comparable, because none of them is a particular case of
the other.
(c) Model (III) is the least restrictive.
(a) Model (I) is more restrictive, as it imposes that the effect of education on the number
(b) Models (I) and (II) are not comparable, since model (I) includes variables that model
(c) Model (II) is less restrictive, as it allows, for a given race, age, and education level, the
(a) Significantly different from zero.
(c) Statistically equal to zero.

(a) Would be inconsistent.
(b) The Gretl program would indicate us that we do not have enough instruments.
(c) Would be less efficient than the ones by 2SLS estimator but using both RU RAL and
(a) We do not reject the null hypothesis at the 5% significance level.
(b) We reject the null hypothesis at the 1% significance level.
(c) We reject the null hypothesis at the 5%, but not at the 1% significance level.
(a) Output 1 will provide inconsistent estimates for model (I) parameters.
(b) Output 1 and Output 2 will provide consistent estimates for the parameters of models
(c) Output 2 will provide consistent estimates for model (II) parameters only if we dropped
whether:
(a) None of the instruments used in Output 4 is correlated with ε3 .
(b) Education is not correlated with the instruments used in Output 4.
(c) Education is not correlated with ε3 .
(a) 0.34.
(b) 0.98.
(c) 1.12.
(a) Negative (on average) for all the women in the sample.
(c) Positive (on average) for all the black women in the sample, as the coefficient of

(a) 2.8.
(b) 1.6.
(c) 2.6.
(b) If we also included Y EAR, Y 78 and Y 84 as explanatory variables, such model would
(a) Have remained constant over time.
(b) We do not have conclusive evidence.
(c) Have decreased over time.
(a) The constant terms of both models are equal, β0 = δ0 .
(b) δ5 = δ6 .
(c) 6β5 = δ6 − δ5 .
would be:
(a) The R2 from Output 7 multiplied by the number of observations.
(c) The R2 from Output 5A multiplied by the number of observations.
(a) We would need at least one variable, not included in the model and uncorrelated with
(c) We would need at least two different variables, not included in the model and uncorre-

to assess whether:
(a) We can reject the null hypothesis about the exogeneity of education.
(b) RU RAL is a better instrument than LP OP .
(c) We can reject the null hypothesis about instruments’ validity.
(a) Models (I) and (II) are different models, since none of them is a particular case of the
other one.
(b) Model (I) imposes the constraint that the coefficient of Y 78 is exactly half of the
(c) Model (I) imposes the constraint that the coefficients of Y 78 and Y 84 were equals.
(a) If we estimated the model (II) by 2SLS using Z1 , Z2 , Z3 , Z4 as instruments, the
(b) If we estimated the model (II) by OLS including Z1 , Z2 , Z3 , Z4 as additional variables,
(c) If we estimated the model (II) by OLS, the estimators obtained for the coefficients δ1 ,
(a) A woman in 1984 had on average 0.6 children less than a woman in 1972.
(c) A woman in 1978 had on average 0.3 children more than a woman in 1972.
(a) Always inconsistent.
(c) Consistent, provided that the omitted variable is uncorrelated with the remaining
(a) The causal effect of education is positive.
(b) The causal effect of education is more negative in 1978 than in 1984.

ECONOMETRIC
June, 25, 2010
Tipo de examen: 4
Directions:
group.
reading form.

validity.
15.1.43.

Answer DRAFT
(a) (b) (c) (a) (b) (c) (a) (b) (c) (a) (b) (c)
1. 13. 25. 37.
2. 14. 26. 38.
3. 15. 27. 39.
4. 16. 28. 40.
5. 17. 29. 41.
6. 18. 30. 42.
7. 19. 31. 43.
8. 20. 32. 44.
9. 21. 33. 45.
10. 22. 34. 46.
11. 23. 35. 47.
12. 24. 36. 48.

(a) Model (III) is the least restrictive.
(b) Models (I) and (II) are not comparable, because none of them is a particular case of
the other.
(c) Model (I) is the most restrictive.
(a) γ4 = γ5 = γ6 .
(b) γ4 = γ6 = 0.
(c) γ6 = 0.
(a) We reject it, given the the magnitude of the corresponding statistic obtained when
(c) We do not reject it, because the p-value of the corresponding statistic is equal to 0.
(a) Consistent, provided that the omitted variable is uncorrelated with the remaining
(c) Always inconsistent.
(a) H0 : γ4 = γ6 = 0.
(b) H0 : γ5 = γ6 = 0.
(c) H0 : γ6 = 0.
(a) We reject the instruments’ validity.
(b) There is not information to assess whether the instruments are valid or not.
(c) We do not reject the instruments’ validity.
(a) The estimated coefficients in Output 2 would be inconsistent for model (II), but those
(c) The estimated coefficients in Output 1 would be inconsistent for model (I), but those
(a) Output 2 will provide consistent estimates for model (II) parameters only if we dropped
(b) Output 1 and Output 2 will provide consistent estimates for the parameters of models
(c) Output 1 will provide inconsistent estimates for model (I) parameters.
would be:
(a) 2.4.
(b) 14.7.
(c) 1.7.
whether:
(a) Education is not correlated with ε3 .
(b) Education is not correlated with the instruments used in Output 4.
(c) None of the instruments used in Output 4 is correlated with ε3 .
(a) We reject the null hypothesis at the 5%, but not at the 1% significance level.
(b) We reject the null hypothesis at the 1% significance level.
(c) We do not reject the null hypothesis at the 5% significance level.
(a) 6β5 = δ6 − δ5 .
(b) δ5 = δ6 .
(c) The constant terms of both models are equal, β0 = δ0 .

(a) We do not reject that the correlation of the instruments with EDU C is equal to zero.
(b) We do not reject that the correlation of the instruments with the error in model (III)
is equal to zero.
(c) We reject that EDU C is exogenous.
(b) We cannot reject such assertion at the 5% significance level.
(a) The same number of children than the second one.
(b) 1 child more than the second one.
(c) 1 child less than the second one.
(a) Model (I) is a particular case of model (II).
(c) Model (II) is misspecified, since it omits the variable Y 72.
(a) If we estimated the model (II) by OLS, the estimators obtained for the coefficients δ1 ,
(b) If we estimated the model (II) by OLS including Z1 , Z2 , Z3 , Z4 as additional variables,
(c) If we estimated the model (II) by 2SLS using Z1 , Z2 , Z3 , Z4 as instruments, the
Output 1 will be:
(a) Inconsistent, only if the measurement error is correlated with the error term of the
model.
(b) Consistent, but less efficient than if the variable were measured without error.
(c) Always inconsistent.
(b) Over-identified, the number of over-identifying restrictions being equal to 2.
(c) Exactly identified.
(a) A woman in 1978 had on average 0.3 children more than a woman in 1972.
(c) A woman in 1984 had on average 0.6 children less than a woman in 1972.
would be:
(a) H0 : δ2 = 0.
(b) H0 : δ1 − δ2 = 0.
(c) H0 : δ1 = δ2 = 0.
restriction:
(a) δ6 = 2δ5 .
(b) δ6 = 6δ5 .
(c) δ6 = δ5 .
to assess whether:
(a) We can reject the null hypothesis about instruments’ validity.
(b) RU RAL is a better instrument than LP OP .
(c) We can reject the null hypothesis about the exogeneity of education.
(a) Older women have, on average, more children.
(b) The causal effect of education is the same for all women considered.
(c) More educated women have, on average, more children.

(a) −0.4.
(b) −0.8.
(c) −0.7.
(a) Test the hypothesis that the residual of the reduced form (linear projection of EDU C
on education.
(b) In a regression of EDU C on the exogenous variables of the model and both instru-
(c) In a regression of EDU C on the exogenous variables of the model and both instruments
(b) 4.5 less children.
(a) H0 : 2γ4 + 72γ6 = 0.
(b) H0 : γ4 = −144γ6 .
(c) H0 : γ4 + 72γ6 = 0.
(a) H0 : γ4 = γ6 .
(b) H0 : γ4 = 0.

γ4 − γ6 = 0
(c) H0 : .
γ6 = 0
would be:
(a) The R2 from Output 5A multiplied by the number of observations.
(c) The R2 from Output 7 multiplied by the number of observations.
(a) Would be less efficient than the ones by 2SLS estimator but using both RU RAL and
(b) The Gretl program would indicate us that we do not have enough instruments.
(c) Would be inconsistent.
(a) Model (II) is less restrictive, as it allows, for a given race, age, and education level, the
(b) Models (I) and (II) are not comparable, since model (I) includes variables that model
(c) Model (I) is more restrictive, as it imposes that the effect of education on the number
(a) 1.12.
(b) 0.98.
(c) 0.34.
(a) We would need at least two different variables, not included in the model and uncorre-
(c) We would need at least one variable, not included in the model and uncorrelated with
(b) The causal effect of education is more negative in 1978 than in 1984.
(c) The causal effect of education is positive.
(a) 2.6.
(b) 1.6.
(c) 2.8.

hypothesis is:
(a) H0 : γ5 = 0.
(b) H0 : γ5 = γ6 .
(c) H0 : γ5 = γ6 = 0.
(b) If we also included Y EAR, Y 78 and Y 84 as explanatory variables, such model would
(a) Model (I) imposes the constraint that the coefficients of Y 78 and Y 84 were equals.
(b) Model (I) imposes the constraint that the coefficient of Y 78 is exactly half of the
(c) Models (I) and (II) are different models, since none of them is a particular case of the
other one.
(a) Have decreased over time.
(b) We do not have conclusive evidence.
(c) Have remained constant over time.
be:
(a) 23.4.
(b) 7.8.
(c) 51.4.

(a) Statistically equal to zero.
(c) Significantly different from zero.
(a) A woman in 1978 has approximately 29.3% less children than a woman in 1972.
(c) For every 50 women, there are approximately 29.3 more children in 1972 than in 1984.
(a) Positive (on average) for all the black women in the sample, as the coefficient of
(c) Negative (on average) for all the women in the sample.
(a) The lower the correlation of BLACK with the relevant variables.
(b) The higher the proportion of black women in the sample.
(c) The higher the correlation of BLACK with the relevant variables.

notes on

Uploaded by

notes on

Uploaded by

UNIVERSIDAD CARLOS III DE MADRID

June, 25, 2010

Very important: Take into account that:

KIDS = 0+ 1 AGE + 2 AGE 2 + 3 BLACK + 4 EDU C + 5Y EAR (III)

Output 1: OLS, using observations 1–476

Coe¢ cient covariance matrix (Output 2)

Output 3: OLS, using observations 1–476

Mean dependent var 2.67 S.D. dependent var 1.67

Coe¢ cient covariance matrix (Output 3)

Mean dependent var 2.67 S.D. dependent var 1.67

Coe¢ cient covariance matrix (Output 4)

Output 5A: OLS, using observations 1–476

Output 5B: OLS, using observations 1–476

Output 5C: OLS, using observations 1–476

Output 5D: OLS, using observations 1–476

Output 6: OLS, using observations 1–476

Output 7: OLS, using observations 1–476

• BEFORE YOU START TO ANSWER THE EXAM:

• Check that this document contains 48 questions sequentially numbered.

• Read the problem text and the questions carefully.

• Each question has one correct answer.

• To obtain a grade of 5 over 10 you must correctly answer 28 questions.

• Dates of grades publication: Monday, June, 28.

• Rules for exam revision:

– Its only purpose will be that each student:

Exam type: 1 página 2

Exam type: 1 página 3

Exam type: 1 página 4

Exam type: 1 página 6

Exam type: 1 página 7

Exam type: 1 página 8

Exam type: 1 página 9

Exam type: 1 página 10

Exam type: 1 página 11

• BEFORE YOU START TO ANSWER THE EXAM:

• Check that this document contains 48 questions sequentially numbered.

• Read the problem text and the questions carefully.

• Each question has one correct answer.

• To obtain a grade of 5 over 10 you must correctly answer 28 questions.

• Dates of grades publication: Monday, June, 28.

• Rules for exam revision:

– Its only purpose will be that each student:

Exam type: 2 página 2

Exam type: 2 página 3

Exam type: 2 página 4

Exam type: 2 página 5

Exam type: 2 página 6

Exam type: 2 página 7

Exam type: 2 página 8

Exam type: 2 página 10

• BEFORE YOU START TO ANSWER THE EXAM:

• Check that this document contains 48 questions sequentially numbered.

• Read the problem text and the questions carefully.

• Each question has one correct answer.

• To obtain a grade of 5 over 10 you must correctly answer 28 questions.

• Dates of grades publication: Monday, June, 28.

• Rules for exam revision:

– Its only purpose will be that each student:

Exam type: 3 página 2

Exam type: 3 página 3

Exam type: 3 página 4

Exam type: 3 página 5

Exam type: 3 página 6

Exam type: 3 página 7

Exam type: 3 página 8

Exam type: 3 página 9

Exam type: 3 página 10

• BEFORE YOU START TO ANSWER THE EXAM:

• Check that this document contains 48 questions sequentially numbered.

• Read the problem text and the questions carefully.

• Each question has one correct answer.

• To obtain a grade of 5 over 10 you must correctly answer 28 questions.

• Dates of grades publication: Monday, June, 28.

• Rules for exam revision:

– Its only purpose will be that each student:

Exam type: 4 página 2