618B pset

Economics 618B: Time Series Analysis Department of Economics State University of New York at Binghamton Problem Set #1 1. Generate n = 500 random numbers from both the uniform 1 b − a ( U [0 , 1] , uniform be- tween zero and one) and exponential λ exp ( − λx ) (set λ = 2 and let x ∼ U [0 , 1] ) distributions. Plot the histograms of each of the variables. What are the true and estimated means and variances? 2. Generate a random variable with n = 500 from the mixed normal distribution with P [ x ∼ N ( − 3 , 1)] = P [ x ∼ N (3 , 1)] = 0 . 5 . Plot the histogram. Note that this should be a bimodal distribution. Do not add the two variables together. This will lead to a unimodal density with mean near zero. 3. Generate a random variable with n = 500 from the Student-t distribution with 5 degrees of freedom, t 5 . Plot the histogram. 4. Generate a random variable with n = 500 from the χ 2 distribution with 1 degree of freedom. Plot the histogram. 5. Generate a random variable with n = 500 from the Normal distribution with mean 1 and variance 0.5. Plot the histogram. 6. This problem will show you how the central limit theorem works. Using a programming software show how the CLT works for the sample mean with the fi ve separate distrib- utions. Use n = 10 , 50 , and 100 with m = 100 , 200 , and 500 Monte Carlo replications. Plot histograms of the sample means for each distribution and each sample size. (a) Standard normal, N (0 , 1) (b) Uniform, U [0 , 1] (c) Mixed Normal, P [ x ∼ N ( − 3 , 1)] = P [ x ∼ N (3 , 1)] = 0 . 5 (d) Exponential, λ exp ( − λx ) (set λ = 2 and let x ∼ U [0 , 1] ) (e) Student- t, t 5 7. This problem is to conduct a Monte Carlo experiment to examine the fi nite sample per- formance of a test in size ( P ( reject H 0 : H 0 is true ) ) and power ( P ( reject H 0 : H 0 is false ) ). (a) Size: Generate a random sample x 1 , x 2 , . . . , x n of size n = 50 , with μ = 0 and σ 2 = 1 from two distributions, (i) normal and (ii) uniform. Pretend you do not know the mean and variance. Then test the null hypothesis that μ = 0 against the alternative that it is not equal to zero. Use Monte Carlo experiments with 500 replications to evaluate the size of the test statistic t. Use the asymptotic critical value at the 5% level (1.96). Repeat with n = 100 and 200 . What do you fi nd? (b) Power: Repeat part (a) with μ = 0 . 1 , 0 . 3 , − 0 . 1 , and − 0 . 3 . Plot the power function (include μ = 0 in the plot). 1

Economics 618B: Time Series Analysis Department of Economics State University of New York at Binghamton Problem Set #2 1. The fi le entitled SIM_2.XLS contains simulated data sets. The fi rst series, denoted Y1, contains 100 values of a simulated AR(1) process. Use this series to perform the following tasks. (a) Plot the sequence against time. Does the series appear to be stationary? (b) Plot the ACF and PACF. (c) Estimate the AR(1), AR(2), ARMA(1,1), ARMA(1,(1,4)), ARMA(2,1) and ARMA(2,(0,4)) models. (d) Estimate the series as both an AR(2) and ARMA(1,1) process without an inter- cept. (e) Use MSE, AIC and SBC to choose the best model from parts (c) and (d). (f) Obtain the one-step-ahead forecast and one-step-ahead forecast error from model. Which model performs the best? Is this surprising? (g) The second series in SIM_2.XLS, denoted Y2, contains 100 values of a simulated ARMA(1,1) process. Repeat steps (a-f) with the series Y2. (h) The third series in SIM_2.XLS, denoted Y3, contains 100 values of a simulated AR(2) process. Repeat steps (a-f) with the series Y3. 2. The fi le QUARTERLY.XLS contains demand deposits reported by commercial banks ( DDNSA ). The series is not seasonally adjusted. The series is quarterly averages over the period 1960:Q1 to 2002:Q1. Be sure to brie fl y explain the relevance of each step. (a) Plot the DDNSA sequence against time. Does the series appear to be stationary? (b) Plot the ACF and PACF of DDNSA . (c) Create the growth rate series log ( DDNSA t /DDNSA t − 1 ) and plot this series against time. Does the series appear to be stationary? (d) Plot the ACF and PACF of log ( DDNSA t /DDNSA t − 1 ) . (e) Seasonally di ff erence demand deposits using log ( DDNSA t /DDNSA t − 4 ) . Does this series appear to be stationary? (f) Plot the ACF and PACF of log ( DDNSA t /DDNSA t − 4 ) 2

Economics 618B: Time Series Analysis Department of Economics State University of New York at Binghamton Problem Set #4 1. The fi le QUARTERLY.XLS contains the real GDP data ( GDP ). The series is mea- sured over the period 1960:Q1 to 2002:Q1. (a) Plot the GDP sequence against time. Does the series appear to be stationary? (b) Regress the series on an intercept and a third-order polynomial of time ( t, t 2 , and t 3 ). (c) Plot the ACF and PACF of the residuals from the model estimated in (b). What can be said about the residuals? (d) Perform the Dickey-Fuller test on the GDP sequence. (e) Perform the Augmented Dickey-Fuller test on the GDP sequence. (f) Construct the rate of growth of GDP as dlrgdp t = log ( GDP t /GDP t − 1 ) . (g) Plot the dlrgdp sequence against time. Does the series appear to be stationary? (h) Model dlrgdp as an AR(2) process. (i) Plot the ACF and PACF of the residuals from the model estimated in (h). What can be said about the residuals? (j) Perform the Dickey-Fuller test on the dlrgdp sequence. (k) Perform the Augmented Dickey-Fuller test on the dlrgdp sequence. (l) It is often argued that the oil price shock in 1973 reduced the trend growth rate of real U.S. GDP. Perform the Perron test to determine whether the series is trend stationary with a break occurring in mid-1973. (m) Decompose the real GDP series into the temporary and permanent components using the HP fi lter. Plot the transitory component that you obtain from the HP fi lter. (n) Suppose that real GDP is trend stationary with a break occurring in mid-1973. Let the deviations from trend constitute the transitory component of the series. How does this transitory component compare with your answer found in part (m)? 4

Economics 618B: Time Series Analysis Department of Economics State University of New York at Binghamton Problem Set #5 1. The fi le COINT_PPP.XLS contains quarterly values of Canadian, German and Japanese wholesales prices and bilateral exchange rates with the United States. The fi le also contains the U.S. wholesale price level. The names of the individual series should be self-evident. For example, p _ US is the U.S. price level and ex _ g is the German ex- change rate with the United States. All variables except the mark/dollar exchange rates run from 1973:Q4 to 2001:Q4 and all have been normalized to equal 100 in 1973:Q4. (a) Form the log of each variable. Estimate the long-run relationship between Canada and the United States log ( ex _ ca ) = α + β 1 log ( p _ ca ) + β 2 log ( p _ us ) + ε (b) Do the point estimates of the slope coe ffi cients seem to be consistent with long-run PPP? (c) Let b ε denote the residuals from the long-run relationship. Use these residuals to perform the Engle-Granger test for cointegration. Use three lagged changes. (d) Perform a t -test on the coe ffi cient for ∆ b ε t − 1 . Explain how a rejection of this null of signi fi cance indicates that long-run PPP fails. (e) Repeat parts (a-d) with Germany. (f) Repeat parts (a-d) with Japan. 2. The fi le INT_RATES.XLS contain interest rates paid on U.S. 3-month, 3-year and 10- year U.S. government securities. The data run from 1954:7 to 2002:12. These columns are labeled TBILL, R3 and R10, respectively. (a) Pretest the variables to show that the rates all act as unit processes. Speci fi - cally, perform augmented Dicky-Fuller tests using the lag length selected by AIC. Include and intercept, but no time trend. (b) Estimate the cointegrating relationships using the Engle-Grange procedure. Per- form augmented Dickey-Fuller tests on the residuals. Use T-Bill as the left-hand- side variable with an intercept and the remaining two securities on the right-hand- side. (c) Repeat part (b) using R10 as the left-hand-side variable. (d) Repeat part (b) using R3 as the left-hand-side variable. (e) Estimate an error-correction model using 12 lags of each variable. Use the resid- uals from part (b) as the error-correction term and do not include a separate intercept. (f) Do the residuals from part (e) appear to be white noise? 5