VAR estimation issues

[diego_wanted]

Hi Guys,

I'm having a nightmare with my dissertation data analysis and my supervisor is one of those very busy ones who don't reply to e-mail, so I thought I'd come here for help.
Basically, I'm looking at the economics effects of trade openness on poverty reductions in Brazil. The variables I have are Poverty, Openness, Inflation, FDI and Government Consumption.
I have logged all variables apart from inflation ( as it's measured in rates ) and ran a unit root test to check the order of the variables and they're all non-stationary but become stationary at first order. I then differentiated all variables and saved the changes. I checked for co-integration using the Johansen test and the results showed that they aren't co-integrated. I then decided to use a VAR model to estimate my regression, but my results are still very weird..
None of the variables are significant and the R^2 is very low! I read on a few books that this is normal for VAR models and that people don't usually use their coefficient for interpretation.. is this correct? Also, I put Poverty and Openness as my endogenous variables and the rest of them as exogenous as the purpose of the paper is to see the effects of openness on poverty. Would this be correct or should I include all variables as endogenous?
I have attached here a print screen of all the process I went through, please if you can spot any mistakes let me know as I've spent 3 weeks trying to work this out without any success and I'm feeling very lost atm.

Thank you

Diego

[diego_wanted]

And this is the Johansen test results:

[trubador]

I really do not understand (nor appreciate) those supervisors who never find spare time for their students to provide them a proper guidance.
Although yours may not be an econometrics subject, you still need to spend enough time understanding the fundamentals of VAR or VEC analysis. I am sure you have access to plenty of good econometrics textbooks.
As for the empirical aspect, I might say a few things:
1) Number of observations are not enough. If you cannot extend the analysis period, then you might use quarterly figures instead.
2) I am not a fan of differencing variables to ensure stationarity, but in this analysis you really do not have to do so. You can safely work in levels with VEC analysis.
3) ARDL modeling can also be a useful alternative here. For details please see: http://davegiles.blogspot.com.tr/2015/0 ... ews-9.html
4) You should refer to papers/articles with similar subjects.

[diego_wanted]

Thank you so much for helping! I really appreciate it. I'm not very good with econometrics and as you see, my supervisor should have warned me about the data size and methods but he didn't. The problem I am facing is that I wasn't able to find appropriate quarterly data for the indicators I need. I can extend the year back to 1980 but before that the vast majority of the data is missing. What would you advise me to do in this case?
I will try and apply the ARDL model as you proposed, the only problem I have is that I have to learn it all as I've never heard of it, but I still have two weeks to go so hopefully that still be time.