Scribd Logo
Upload

Welcome to Scribd!

  • 30 views

    Ec401 Midterms

    Uploaded by

    Anuj Diwedi
    This document contains an answer key for a midterm exam in Econometric Methods. It provides answers to 4 questions: 1) Model specification affects efficiency, with omitted variables or irrelevant variables leading to inefficiency. 2) Methods for testing whether some observations are influential in a regression, including calculating residuals. 3) Explains that one coefficient may be biased while another is unbiased in a regression where the expectation of the error term is not zero. 4) Interprets a coefficient on education as the average return, and suggests using a test based on the delta method to see if returns vary between those with and without children.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    Ec401 Midterms

    Uploaded by

    Anuj Diwedi
    30 views11 pages
    This document contains an answer key for a midterm exam in Econometric Methods. It provides answers to 4 questions: 1) Model specification affects efficiency, with omitted variables or irrelevant variables leading to inefficiency. 2) Methods for testing whether some observations are influential in a regression, including calculating residuals. 3) Explains that one coefficient may be biased while another is unbiased in a regression where the expectation of the error term is not zero. 4) Interprets a coefficient on education as the average return, and suggests using a test based on the delta method to see if returns vary between those with and without children.

    Original Title

    Ec401_midterms

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    This document contains an answer key for a midterm exam in Econometric Methods. It provides answers to 4 questions: 1) Model specification affects efficiency, with omitted variables or irrelevant variables leading to inefficiency. 2) Methods for testing whether some observations are influential in a regression, including calculating residuals. 3) Explains that one coefficient may be biased while another is unbiased in a regression where the expectation of the error term is not zero. 4) Interprets a coefficient on education as the average return, and suggests using a test based on the delta method to see if returns vary between those with and without children.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    30 views11 pages

    Ec401 Midterms

    Uploaded by

    Anuj Diwedi
    This document contains an answer key for a midterm exam in Econometric Methods. It provides answers to 4 questions: 1) Model specification affects efficiency, with omitted variables or irrelevant variables leading to inefficiency. 2) Methods for testing whether some observations are influential in a regression, including calculating residuals. 3) Explains that one coefficient may be biased while another is unbiased in a regression where the expectation of the error term is not zero. 4) Interprets a coefficient on education as the average return, and suggests using a test based on the delta method to see if returns vary between those with and without children.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    You are on page 1of 11

    ANSWER KEY MIDTERM 1, Summer 2018

    Course 401: Econometric Methods

    1. Consider the two models (a) y = Xβ + u where X is n × K and (b) y = Zγ + w


    where Z is n × r. Under classical assumptions (and Z and X are non-stochastic) if
    model (a), that is y = Xβ + u is the true model, show that E(σ̂w2 ) ≥ σu2 and explain
    the implication of your result. 4 marks
    Answer : See problem set 1. The idea was that specification matters: the cases of
    omitted variables and inclusion of irrelevant variables are special cases. Bottom line:
    any specification error leads to inefficiency.
    2. Consider Model A: y = Xβ + ε with n observations, and E(εε0 ) = σ 2 I. Say there
    is reason to suspect that p of these observations may be influential, p << n. The
    following augmented regression (Model B) is run using OLS: y = Xβ + Zγ + u, with
    Z = M Dp where M = (I − X(X 0 X)−1 X 0 ), and Dp is a n × p matrix, each column
    of which takes on value 1 if the ith observation is suspected of being influential, and
    zero otherwise. I.e. each column of Dp contains only one element which takes value
    1, the remaining are zero.
    (a) Define ep = Dp0 e. Show that γ̂OLS from Model B is given by (Dp0 M Dp )−1 ep . 2
    marks
    (b) Show that the residual sum of squares in the augmented regression is given by
    û0 û = e0 e − e0p (Dp0 M Dp )−1 ep 4 marks
    (c) How would you test whether these p observations collectively are influential 2
    marks
    Answer : Once again, see problem set 1.
    3. Consider the following regression expressed in partitioned form as y = X1 β1 +X2 β2 +ε
    where X1 has K1 variables and X2 has K2 variables and K1 + K2 = K. Denote the
    corresponding OLS estimators by b1 and b2 respectively. All the classical assumptions
    hold, except that E(ε) 6= 0. Instead E(ε) = X1 Γ.
    Show that b1 is biased 3 marks
    Show that b2 is unbiased 5 marks
    Answer : Use the FWL theorem. Showing that b1 is biased is relatively straightfor-
    ward and follows directly from the theorem.
    As far as b2 is concerned, note that
    0 0
    b2 = (X2 M1 X2 )−1 X2 M1 y
    0 0
    = (X2 M1 X2 )−1 X2 M1 [X1 β1 + X2 β2 + ε]
    0 0 0 0 0 0
    = (X2 M1 X2 )−1 X2 M1 X1 β1 + (X2 M1 X2 )−1 X2 M1 X2 β2 + (X2 M1 X2 )−1 X2 M1 ε
    0 0
    = 0 + β2 + (X2 M1 X2 )−1 X2 M1 ε
    Where we use the fact that M1 X1 = 0.
    0 0
    ⇒ E[b2 |X] = β2 + (X2 M1 X2 )−1 X2 M1 X1 Γ = β2
    once again using M1 X1 = 0. Therefore unbiased. Many students assumed that X1
    and X2 are orthogonal. This was not specified in the question; therefore it was not
    accepted as a valid way to proceed.
    4. A researcher has estimated the following regression
    lnearnings = β1 + β2 kids + β3 education + β4 kids X education + ε
    Where kids is a dummy variable taking value 1 if the individual has children and zero
    if not, and education is the number of years of completed education. She obtains the
    following results:
    Variable Coefficient t-ratio
    Constant 3.24 1.99
    Kids -0.35 -2.83
    Education 0.07 2.67
    Kids X Education 0.20 0.93
    How should she interpret the estimated b3 ? 2 marks
    How should she test if returns to education vary for those with and without kids? 3
    marks
    You may ignore considerations of endogeneity and specification
    Answer : This was a modified version of example 6.1 on page 154 of Greene. The
    intent was to (a) recognize even this simple version of a non-linear relationship, which
    translates the dummy variable into a multiplicative (rather than an additive) shift (b)
    cast this in the form of a test based on the delta method, and (c) note the apparent
    insignificance of the b4 . Answers based on a Course 003 - level understanding of this
    as a semi-elasticity (or percent changes) without recognizing that this interpretation
    does not go through automatically for dummy variables were given minimal credit if
    any.

    You might also like