E703: Advanced Econometrics I Exercise Questions - Serial Correlation Please attemp all questions 1. If {yt } is the covariance-stationary solution to the AR(1) equation yt = c + φyt−1 + εt or yt − φyt−1 = c + εt or (1 − φL)yt = c + εt , where {εt } is white noise with |φ| < 1, what is the least square (linear) projetion? b ∗ (yt |1, yt−1 ) E b ∗ [yt |xt ] = x0t β ∗ , where Hint: The least square (liner) projection of y on x is E 0 β ∗ = E[xt xt ]−1 E[xt yt ] (see Hayashi, pp. 139-140). b ∗ (yt |1, yt−1 ) = E(yt−1 εt ) = 0. Do the projection coefficients depend on t? Is E E(yt |yt−1 )? Hint: E(yt−1 εt ) = 0 but does it mean E(εt |yt−1 ) = 0? What is Ê ∗ (yt |1, yt−1 , yt−2 )? Now suppose that {yt } is covariance-stationary solution to the AR(1) equation when |φ| > 1. Is it true that E(yt−1 εt ) = 0? [Answer: b ∗ (yt |1, yt−1 ) = c + φyt−1 as was the case when |φ| < 1? No.] Does E 2. Suppose that the money supply process has the form mt = m+ρmt−1 +εt , where m is constant and 0 < ρ < 1. a. Show that it is possible to express mt+n in terms of the known value mt and the sequence {εt+1 , εt+2 , . . . , εt+n }. b. Suppose that all values of εt+i for i > 0 have mean value of zero (i.e. E[εt+i |mt ] = 0 ∀i > 0). Explain how you could use your result in part (a) to forecast the money supply n periods into the future. 3. (Proof of Proposition 6.1, Hayashi, p. 367) A well-known result about mean square convergence is i. zn converges in mean square if and only if E[(zm − zn )2 ] → 0 as m, n → ∞. ii. If xn →m.s. x and zn →m.s. z, then lim E(xn ) = E(x) and n→∞ lim E(xn zn ) = E(xz). n→∞ Take this result for granted in answering the following. P a. Prove that yt = µ + ∞ j=0 ψj εt−j converges in mean square as n → ∞ under the hypothesis of Proposition 6.1. Hint: Let yt,n = µ + n X ψj εt−j . j=0 What needs to be proved is that {yt,n } convergence in mean square as n → ∞ for each t. Given (i) above, it is sufficient to prove that (assuming m > n without loss of generality) # " m m X X 2 2 ψj2 → 0 as m, n → ∞. E ( ψj εt−j ) = σ j=n+1 j=n+1 Use the fact that an absolutely summable sequence is square summable, i.e., ∞ ∞ X X |ψj | < ∞ ⇒ ψj2 < ∞, j=0 j=0 and the fact that a sequence of real numbers {αn } converges (to a finite limit) if and only if αn is a Cauchy sequence: αn → α ⇔ |αm − αn | → 0 as m, n → ∞. Set αn = Pn j=0 ψj2 . b. Prove that E(yt ) = µ. Hint: Let yt,n = µ + was shown in (a) that yt,n →m.s. yt . Pn j=0 ψj εt−j as in (a). It c. (Proof of part(b) of Proposition 6.1) Show that E[(yt − µ)(yt−j − µ)] = lim E[(yt,n − µ)(yt−j,n − µ)]. n→∞ Have we shown that {yt } is covariance-stationary? [Answer:Yes.] d. Prove part (c) of Propostion 6.1, taking the follwoing facts from calculus for granted. (a) If j } is absolutely summable, then {aj } is summable (i.e. −∞ < P{a ∞ j=0 aj < ∞) and ∞ ∞ X X aj ≤ |aj | . j=0 j=0 (b) Consider P a sequence with two subscripts, {ajk }(j,P k = 0, 1, 2, . . .). ∞ Suppose ∞ |a | < ∞ for each k and let s ≡ jk k j=0 j=0 |ajk |. Suppose also that {sk } is summable. Then ! ! ! ∞ ∞ ∞ ∞ ∞ ∞ X X X X X X ajk < ∞ and ajk = ajk < ∞. j=0 j=0 k=0 k=0 k=0 j=0 Hint: Derive ∞ X j=0 |γj | ≤ σ 2 ∞ ∞ X X j=0 k=0 ! |ψj+k | · |ψk | < ∞.