Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: THEend8_
ECONM1022 ECONOMETRICS Time allowed: SEVEN DAYS Answer ALL questions Candidates must show the basis of all calculations. Part A [25 marks] Consider the linear regression model y = β0 + β1x + β2w + u, where y, x and w are observed variables, u is an error term and β0, β1 and β2 are parameters. We denote β = (β0, β1, β2) ′. We assume exogeneity of the regressors E(u|1, x, w) = 0. We have an iid sample with n observations: {yi, xi, wi}ni=1. Regressing y on a constant, x and w by OLS yields the estimate β̂ = ( β̂0, β̂1, β̂2 )′ 1. (a) Write down the first-order conditions characterizing the OLS estimator β̂. [3 marks] (b) Derive the formal expression of β̂ as a function of data matrices that you need to specify. [3 marks] (c) Show that β̂ is an unbiased estimator of β. [3 marks] 2. (a) Describe a test for the hypothesis β1 = β2 against β1 6= β2 [3 marks] (b) Assume, for this question only, that the test does not reject β1 = β2 and consider an OLS regression of y on a constant and the regressor x−w. How would the coefficient of determination from this regression compare with the one from the OLS regression of y on a constant, x and w? [2 marks] (c) Consider the following test of the hypothesis β1 = β2. We regress y on a constant x, w and x − w by OLS and then test whether the coefficients of x and w are jointly equal to 0. Would you recommend this approach? [2 marks] 3. (a) Show that the first-order conditions characterizing β̂ (derived in 1.a) can be rewritten as a system of three equations involving only β̂0, β̂1, β̂2 and sample variances and covariances of y, x and w. (Hint: recall that the sample variance is the sample mean of the square minus the square of the sample mean and that the empirical covariance is the sample mean of the product minus the product of the sample means) [4 marks] (b) Show that if the covariance of x and w were equal to 0 then regressing y on a constant and x by OLS would yield a consistent estimator of β1. [5 marks] Part B [25 marks] Consider the linear regression model y = β0 + β1x + β2w + u, where y, x and w are observed scalar variables, u is an error term and β0, β1 and β2 are parameters. We denote β = (β0, β1, β2) ′. We assume that E(u|1, w) = 0 but for the moment we do not know whether x is exogenous or not. We also assume that there is an observed scalar variable z that is uncorrelated with the error term u. We have an iid sample with n observations: {yi, xi, wi, zi}ni=1. Lastly, we assume that the matrices E[(1, w, x)′(1, w, x)] and E[(1, w, z)′(1, w, z)] are full column rank. 1. In this first section, we will use z (together with the constant and w) as instrumen- tal variables. (a) Write down the IV identifying assumptions. [3 marks] (b) Describe the two steps of the 2SLS estimation of β and derive the expres- sion of the estimator, denoted as β̂2sls [4 marks] (c) Which validity condition can we test? How would this test work? [4 marks] (d) The 2SLS estimation gives β̂2sls1 = 1.24 and V̂ ( β̂2sls1 ) = 0.05. Is β1 different from 1 at the 5% significance level? [3 marks] 2. From now on, in this second section, we will assume that x is uncorrelated with u and we will use x (together with the constant and w) as instrumental variables. We will not use the variable z. (a) Are the IV identifying assumptions verified? [2 marks] (b) Write down the moment conditions following from the exogeneity of the three instruments (the constant, w and x). [3 marks] (c) Show that the IV estimator is then equal to the estimator obtained by re- gressing y on a constant, x and w by OLS. [4 marks] (d) Would you rather use this IV estimator (using x, w and the constant as instruments) or the 2SLS estimator from the first section (derived in question 1.b and using z, w and the constant as instruments)? [2 marks] Part C [25 marks] We have an iid sample {Yit, X ′it}N,Ti=1,t=2000 of N = 417 wines bottled from 1980 to 1995, where Yit is the log price for wine i during the year t ∈ {2000, 2005, 2010, 2015}, and Xit = (aromait, f lavorit, intenit) ′ is a vector of covariates measuring aromatic -aromait-, gustatory -flavorit-, and color intensity -intenit- characteristics. We posit the linear specification: Yit = X ′ itβ + ηi + ξit, where ηi is the terroir of wine i, and ξit is an idiosyncratic disturbance term. 1. Derive the First-Difference (FD) Estimator βˆfd. [5 marks] 2. Below is a tentative proof of consistency of the FD estimator. It contains mistakes. Identify and correct those mistakes. [6 marks] The starting point is βˆfd = ( T∑ t=2 N∑ i=1 ∆Xit∆X ′ it )−1 T∑ t=2 N∑ i=1 ∆Xit∆Yit. Replacing ∆Yit = ∆X ′itβ + ∆ξit, βˆfd − β = ( T∑ t=2 N∑ i=1 ∆Xit∆X ′ it )−1 T∑ t=2 N∑ i=1 ∆X ′it∆ξit. Multiplying by N/N , βˆfd − β = ( T∑ t=2 N−1 N∑ i=1 ∆Xit∆X ′ it )−1 T∑ t=2 N∑ i=1 N−1∆X ′it∆ξit. Since the data are iid, on the proviso that ∑T t=2 E(∆X ′it∆Xit) is invertible, the Law of Large Numbers and the Continuous Mapping Theorem imply( T∑ t=2 N−1 N∑ i=1 ∆Xit∆X ′ it )−1 → P T∑ t=2 E(∆X ′it∆Xit). (A) Since the data are iid, the central limit theorem and the restriction E(ξit|Xit) = 0 imply T∑ t=2 N∑ i=1 N−1∆X ′it∆ξit = E(∆Xit∆ξit) = 0. (B) Combining (A) and (B) using the Slutzky Lemma, one has that βˆfd converges in prob- ability to β. 3. What test would you use to differentiate between the random-effect and the fixed- effect models? Write down the null and alternative hypotheses, the test statistic and its asymptotic distribution under the null, and the decision rule. [6 marks] 4. Abraham, a wine expert, argues that aromatic and gustatory characteristics of a wine are both related to its price. Write down a test based on the FD estimator (null and alternative hypotheses, test statistic and its asymptotic distribution under the null, critical values, and the decision rule) that could be employed to confirm Abraham’s ar- gument. [4 marks] 5. Abraham also argues that the information on the wine bottle label is related with the average price of the wine. Could we use an Asymptotic T-test based on the FD estimator to confirm this argument? Justify. [4 marks]