VAR-sign restrictions-absurd results
Posted: Wed Apr 13, 2016 3:25 am
Dear Sir,
I am new to RATS and currently replicating some codes given in VAR Workbook - Chapter Sign restrictions (Page 79, Example 7.1), which details the Uhlig(2005) code. I am using this code to work on a 6 variable VAR with the identification - y(GDP),d(dividends),p(inflation),comm_e(commodity price index),lp(provisions) and sp (share price) applied to US data. I imposed the sign restriction on prices and GDP as well as my policy variable lp. However, I find that results do not observe the sign restrictions on prices at all which is contrary to both the literature where price puzzle is solved on addition of commodity price index as well as my own results from recursive VAR.
Further, even with the basic identification (y-d-p-comm_e-int_rate_sp), which produces reasonable results as in Gali(2013) - Effects of MOnetary Policy Shocks on Stock Market Bubbles', the percentage of draws accepted is very very low which is absurd as it is tried and tested on US data. I also find that if I extend the number of draws beyond 100,000 or increase number of variables, my RATS shut down abruptly.
How do I resolve these issues in my work. The code is attached along with the data for your kind reference. Appreciate any help.
Thanks.
I am new to RATS and currently replicating some codes given in VAR Workbook - Chapter Sign restrictions (Page 79, Example 7.1), which details the Uhlig(2005) code. I am using this code to work on a 6 variable VAR with the identification - y(GDP),d(dividends),p(inflation),comm_e(commodity price index),lp(provisions) and sp (share price) applied to US data. I imposed the sign restriction on prices and GDP as well as my policy variable lp. However, I find that results do not observe the sign restrictions on prices at all which is contrary to both the literature where price puzzle is solved on addition of commodity price index as well as my own results from recursive VAR.
Further, even with the basic identification (y-d-p-comm_e-int_rate_sp), which produces reasonable results as in Gali(2013) - Effects of MOnetary Policy Shocks on Stock Market Bubbles', the percentage of draws accepted is very very low which is absurd as it is tried and tested on US data. I also find that if I extend the number of draws beyond 100,000 or increase number of variables, my RATS shut down abruptly.
How do I resolve these issues in my work. The code is attached along with the data for your kind reference. Appreciate any help.
Thanks.