110 likes | 230 Views
This comprehensive guide delves into the world of Vector Autoregressions (VARs) and Vector Error Correction Models (VECMs). We cover the entire process from estimation and forecasting to conducting impulse-response functions using quarterly data from January 1995 to December 2004. Utilizing STATA, we explore generating real variables, testing for stability, Granger causality, cointegration, and long-run relationships among key economic indicators such as GDP, consumption, and the GDP deflator. Perfect for anyone looking to enhance their macroeconomic analysis skills!
E N D
MACROECONOMETRICS LAB 4 – VARs and VECMs
ROADMAP • Everything is by definition linked with everything? • Estimating VARs • Estimation and forecasting • IRF • VECM • All about STATA
Dataset • Timing • Quarterly data, Jan. 1995 – Dec. 2004 • Data • GDP • Consumption • GDP deflator • M3 • Quarter dummies
STATA (1) • set memory 99m • set matsize 800 • Open downloaded file • Generate real variables: • Real GDP: gen rGDP=GDP/def_CPI • Real consumption • Generate logs (same way) • Log (realGDP) • Log (realconsumption) • Log (def_CPI) • Log (m3)
STATA (2) • Graphing • twoway (line lrpkb lrconsumption if date<q(2004.4)) (or alternatively, using the menu) • twoway (line lm3 ldef_CPI if date<q(2004.4)) (or alternatively, using the menu) • VARs • var lrpkb ldef_CPI lrconsumption lm3, exog(_q*)
STATA (3) • Diagnosis • Testing for number of lags • varsocvariables, maxlag(3) • Testing for VAR stability • varstable • Testing for Granger causality • vargranger • Testing for autocorrelation • varlmar
STATA (4) • Impulse-response functions • Starting • cap irf drop • Creating (Option #1) • varbasic lrpkb lrconsumption ldef_CPI lm3 if date<=q(2004q4), lags(1/3) • Creating (Option #2) • varirf create irf, set(irf, replace) • varirf graph irf, i(lrpkb) r(lrconsumption) / UNIQUE SHOCK/ AND • varirf graph oirf, i(lrpkb) r(lrconsumption) /ORTHOGONAL SHOCK/
STATA (5) • Forecasting • Computing • varfcast compute, step(8) dynamic(q(2005.1)) • Graphing • varfcast graph lrconsumption ldef_CPI lm3 lrpkb
STATA (6) • No. of cointegrating relations (Johansen trace test for the rank) • vecrank lrconsumption ldef_CPI lrpkb lm3, sindicators(_q*) lags(3)
STATA (7) • VECM • vec lrconsumption ldef_CPI lrpkb lm3, rank(3) sindicators(_q*) lags(2) REMEMBER: Cointegrating vectors not interpretable • Testing for stability • vecstable • Testing for autocorrelation • veclmar
STATA (8) • Testing for long run relationships (examples the same as in the lecture) • constraint 1 [_ce1]lrconsumption=-1 • constraint 2 [_ce1]ldef_CPI • constraint 3 [_ce1]lm3 • constraint 4 [_ce2]lm3=-1 • constraint 5 [_ce2]lrconsumption • constraint 6 [_ce2]ldef_CPI=1 • constraint 7 [_ce2]lrpkb=1 • vec lrconsumption ldef_CPI lrpkb lm3, rank(2) sindicators(_q*) lags(3) dforce bconstraints(1/7)