User specified shocks within SVAR models
Posted: Fri Jul 26, 2013 3:12 am
I'm trying to replicate Blanchard and Perotti (2002, http://www.nber.org/papers/w7269.pdf?new_window=1) and estimate fiscal multipliers for a given economy from structural VAR impulse response functions.
For those unfamiliar with the paper, the basic VAR structure is simple enough. It's a system with logarithms of three endogenous variables (quarterly GDP, taxes and gov. expenditure) and four lags, while the structural identification assumes some contemporaneous elasticities between output and government finances which is crucial for the correct estimation of impulse responses (see paper for details).
Now, when feeding shocks into the system using the Structural Decomposition option in EViews, I get very convincingly shaped IRFs, but I'm having trouble interpreting their numerical values.
Obviously, fiscal multipliers are given as the %-response of output to a 1% innovation in taxation or expenditure, but seeing as how the size of the structural shock is one standard deviation by default, I'm guessing I can't interpret the IRFs as being indicative of the fiscal multipliers in question.
I've tried defining impulse vectors that correspond to revenue/expenditure shocks amounting to 1% GDP and applying them to the VAR using the User specified option, but I seem to lose the contemporaneous relations inherent in the structural factorization in the process.
I've seen a couple of other people raise similar questions, but I haven't seen any solutions suggested that I could apply.
Any help or comments would be greatly appreciated.
For those unfamiliar with the paper, the basic VAR structure is simple enough. It's a system with logarithms of three endogenous variables (quarterly GDP, taxes and gov. expenditure) and four lags, while the structural identification assumes some contemporaneous elasticities between output and government finances which is crucial for the correct estimation of impulse responses (see paper for details).
Now, when feeding shocks into the system using the Structural Decomposition option in EViews, I get very convincingly shaped IRFs, but I'm having trouble interpreting their numerical values.
Obviously, fiscal multipliers are given as the %-response of output to a 1% innovation in taxation or expenditure, but seeing as how the size of the structural shock is one standard deviation by default, I'm guessing I can't interpret the IRFs as being indicative of the fiscal multipliers in question.
I've tried defining impulse vectors that correspond to revenue/expenditure shocks amounting to 1% GDP and applying them to the VAR using the User specified option, but I seem to lose the contemporaneous relations inherent in the structural factorization in the process.
I've seen a couple of other people raise similar questions, but I haven't seen any solutions suggested that I could apply.
Any help or comments would be greatly appreciated.