| Summary: | This dissertation investigates the implications of different methods of identifying exogenous shocks to defense purchases for estimating the macroeconomic effects of shocks to government purchases. Four identification schemes are examined: the narrative approach of Ramey and Shapiro (1997), a more comprehensive narrative approach that tries to isolate exogenous reductions in defense purchases as well as exogenous increases in defense purchases, the Choleski decomposition, and long-run restrictions.
The effects of alternative methods of identifying policy shocks are examined within two common VAR systems estimated over specific sample periods. The benchmark model includes four lagged values of defense purchases, real GDP, the three month interest rate on Treasury bills, and the GDP deflator. The alternative model includes six lagged values of the same variables and a "Perron-type" time trend.
The effects of a shock to defense purchases reported for the narrative approach of Ramey and Shapiro (1997) and the more comprehensive narrative approach are similar in terms of patterns; however, they differ in terms of magnitudes. Furthermore, results presented using the narrative approaches are consistent with the shocks identified being exogenous. Finally, some of the key results reported using the narrative approaches, in particular for interest rates and non-residential investment, are sensitive to small perturbations in the dates of some of the shocks identified.
The IRFs reported for a shock to defense purchases using the Choleski decomposition and long-run restrictions are problematic. However, when new orders of defense products are substituted for defense purchases in the models estimated, the IRFs reported using the Choleski decomposition and long-run restrictions appear better than the IRFs reported for a shock to defense purchases.
The IRFs presented using the narrative approaches, the Choleski decomposition, and long-run restrictions differ greatly. Furthermore, the IRFs reported using the different identification schemes are not sensitive to perturbations in the list of variables used; however, they are sensitive to perturbations in the sample period used. As a result, in spite of methodological advances in estimating the effects of exogenous shocks to government purchases, there is still no consensus on the effects of these shocks.
|
|---|
