(a) Generate T = 25 observations on xt and yt as independent random walks with IIN(0, 1) disturbances. Run the regression yt = α + βxt + ut and test the null hypothesis Ho; β = 0 using the usual t-statistic at the 1%, 5% and 10% levels. Repeat this experiment 1000 times and report the frequency of rejections at each signi?cance level. What do you conclude?
(b) Repeat part (a) for T = 100 and T = 500.
(c) Repeat parts (a) and (b) generating xt and yt as independent random walks with drift as described in (14.13), using IIN(0, 1) disturbances. Let γ = 0.2 for both series.
(d) Repeat parts (a) and (b) generating xt and yt as independent trend stationary series as described in (14.11), using IIN(0, 1) disturbances. Let α = 1 and β = 0.04 for both series.
(e) Report the frequency distributions of the R2 statistics obtained in parts (a) through (d) for each sample size and method of generating the time-series. What do you conclude? Hint: See the Monte Carlo experiments in Granger and Newbold (1974), Davidson and MacKinnon (1993) and Banerjee, Dolado, Galbraith and Hendry (1993).