COMPUTER EXERCISES 3 Import file „Economic indicators.wf1“ to EViews (download the file from the www.efzg.hr... and save it on your desktop). This workfile contains time series data from period 2000Q1 to 2014Q3 (59 quarters) with respect to following economic indicators of China: exchange rate (CNY/USD), fdi_capital (FDI, capital used in millions of CNY), gdp_current_price (GDP in current prices, billions of CNY), gdp_growth (in %), ind_production (volume index on industrial production) and m2 (money supply M2 in billions of CNY). Observations are dated as regular frequency data using commands „Proc -> Structure/Resize Current Page... -> Dated-regulary frequency -> Frequency: Quarterly -> Start date: 2000Q1 -> End date: 2016Q4 -> OK -> OK“. Range of observations is extended for the out of sample observations (observations in the sample + observations out of sample). For extended range of observations create new variables: 3 seasonal dummy variables (dummy variable for the second quarter is generated using „Quick -> Generate series… -> Enter equation: d2=@seas(2) -> OK“… repeat commands for the third and the fourth quarter ->…. d3=@seas(3) ->… d4=@seas(2). Note: seasonal dummy variable for the first quarter is omitted). variable „time“ (use „Quick -> Generate series… -> Enter equation: t=@trend -> OK“). variable „time squared” (use „Quick -> Generate series… -> Enter equation: t2=t*t -> OK“). “time dummy” variable (use the Genrate series dialog box and enter equation: „d1=@recode(t>21,1,0)“ -> this means that the value will be equal to 0 for the first 22 observations up to 2005Q3 and for all other observations the value will be equal to 1) 1. Estimate two following econometric models: a) ŷt ˆ 0 ˆ1t b) ŷt ˆ 0 ˆ1d1 ˆ 2 d1 t where y exchange rate (CNY/USD), t variable time ( t 0, 1, 2, 3, ...,58 ) and d1 0 before 2005Q3 and 1 after 2005Q3 (click on the Estimate button from Equation toolbar and specify equation by list of variables for model a): exchange c t and for model b): exchange c d1 d1*t). Compare the actual values and fitted values on the same graph within two estimated models (click on the Resids button from Equation toolbar). Which model fits data better according to R square? Explain the meaning of estimated slope coefficient after 2005Q3. Is slope coefficient after 2005Q3 statistically significant? 1 (perform t-test and F-test using linear restriction C(2)+C(3)=0 -> click on the View button from Equation toolbar and choose Coefficient Tests - > Wald – Coefficient Restrictions…) Model a) ŷt ________________________ R 2 ______ Model b) ŷt ________________________ R 2 ______ Slope coefficient after 2005Q3 equals to ___________ H 0 ... H1 ... t ___________ F _____________ p value : ____________ . 2. Estimate a model with a constant term (intercept), power trend and seasonal dummy variables to describe the behavior of the GDP in China, current prices (from Quick menu choose Estimate Eqution dialog box and specify equation by list of variables: gdp_current_price c t t2 d2 d3 d4). ŷt _______________________________________________________ Explain the meaning of estimated coefficients for dummy variables! Are dummy variables statistically significant? Compare the actual values and fitted values on the same graph (click on the Resids button from Equation toolbar). Calculate forecast values based on estimated model in range from 2000Q1 to 2016Q4 (click on the Forecast button and rename forecast values as gdp_forecast). Compare actual and forecast value of GDP on the same graph (select both variables and right click -> Open as Group -> View -> Graph… -> Single graph -> OK). 3. Estimate a same model again with a constant term, power trend and seasonal dummy variables and perform following diagnostic tests to check if: a) Residuals are normally distributed (View -> Residuals Diagnostics -> Histogram-Normality test) Write down the null and the alternative hypothesis, the value of JB test and the p-value. According to which distribution the test statistic is distributed? H 0 ... H1 ... JB ______ 6 24 2 3 _______ p value __________ 2 b) Residuals are mutually independent (View -> Residuals Diagnostics -> Serial Correlation LM test-> Lags to include: 2 -> OK). Write down the estimated test equation, the value of LM test and the p-value. t ____________________________________________________________ LM ______ _______ ______________ ; p value _______________ H 0 ... H1 ... If the null hypothesis is true then there is no serial correlation up to order two (number of lags is two by default). Can we reject the null hypothesis? Calculate the approximate value of the first order autocorrelation coefficient of residuals using DurbinWatson statistic? What can you conclude based on the value of the DW statistic in this example? DW _____________ ; ˆ 1 1 2 ________________ c) Residuals have constant variance (View -> Residuals Diagnostics -> Heteroskedasticity test-> Test type: BPG -> OK). Repeat the same test using White specification without cross terms (View -> Residuals Diagnostics -> Heteroskedasticity test-> Test type: White -> do not include white cross terms -> OK). Write down the estimated test equation, the value of LM test and the p-value. t2 ____________________________________________________________ LM ______ _______ ______________ ; p value _______________ H 0 ... H1 ... If the null hypothesis is true then there is no heteroskedasticity of residuls. Can we reject the null hypothesis? d) If heteroskedasticity problem exist estimate the same model using robust standard errors, i.e. from Estimate Eqution dialog box click Options and choose HAC (Newey-West) consistent standard errors). 3