### SVAR IRFs, filtering and cointegration questions

Posted:

**Sun Aug 06, 2017 4:38 am**Hi,

So i'm trying to estimate the impact of both monetary and oil price shocks on various macroeconomic variables in Norway. I'm employing the use of a SvAR in order to carry out this analysis and I have a number of questions:

Firstly, my IRFs seem to come out mostly insignificant for oil price shocks. This is counter to what the literature and basic economic intuition say. So i'm wondering if there's something I haven't accounted for? However, once I use a HP filter on the macroeconomic variables (GDP, unemployment and inflation) the IRFs suddenly become similar to those found in other research, though other research doesn't transform the data beyond logging. I've attached a picture of my IRFs and my B matrix.

With regards to filtering, would using a HP filter on a few of the variables make sense? My reasoning is that I want to examine the impacts of shocks and how they influence the business cycle. If not, what other filters should I be looking at?

Since I will be testing for cointegration would it make sense to test for cointegration on the variables prior to filtering or only after filtering?

I also notice that a few of my variables are unit root processes, what would be best way to deal with this? Differencing the variables also leads to insignificant results.

Thanks in advance.

So i'm trying to estimate the impact of both monetary and oil price shocks on various macroeconomic variables in Norway. I'm employing the use of a SvAR in order to carry out this analysis and I have a number of questions:

Firstly, my IRFs seem to come out mostly insignificant for oil price shocks. This is counter to what the literature and basic economic intuition say. So i'm wondering if there's something I haven't accounted for? However, once I use a HP filter on the macroeconomic variables (GDP, unemployment and inflation) the IRFs suddenly become similar to those found in other research, though other research doesn't transform the data beyond logging. I've attached a picture of my IRFs and my B matrix.

With regards to filtering, would using a HP filter on a few of the variables make sense? My reasoning is that I want to examine the impacts of shocks and how they influence the business cycle. If not, what other filters should I be looking at?

Since I will be testing for cointegration would it make sense to test for cointegration on the variables prior to filtering or only after filtering?

I also notice that a few of my variables are unit root processes, what would be best way to deal with this? Differencing the variables also leads to insignificant results.

Thanks in advance.