Statistically it's doable, still six years is relatively a short  period of time to be qualified as long term. So to answer your question, sure you can run a cointegration test as you have specified but your results won't be that accurate. 

at least you have to 15 points of time for running any panel co-integration and westerlung (2007) error-correction–based test need at least 25 observation on time series.

long and short terms are relative concepts, they depend of what kind of data we are  dealing with, from daily to annual data we can't keep the same reasoning, a single year with daily data can be qualified as long term for sure, but a year with monthly data approximately can't.  So the minimum of numbers isn't as relevant as common sens, which means that numbers must have a significant meaning. 

with your 180 obs you won't have problems with running cointegration test, still with annual data, six years aren't enough to specify a long term equation from a common sens point of view. 

A sample of six is too small for meaningful statistical inference. The more the sample the more reliable your results. To me the minimum should be 30.

i thought so too that 6 years is still not good enough.. I think i will change the model then. 

Thank You all for your response,,