I suggest you to use your data from 1971-2010 to create a "primary model", and thus, use it to generate the missing data of the variable using Bootstrap.

Efron, B., & Efron, B. (1982). The jackknife, the bootstrap and other resampling plans (Vol. 38). Philadelphia: Society for industrial and applied mathematics.

Dear Udari,

You can find a proxy variable for energy use for what the data from 1966 is available. By the way,  I agree with Dr.Fares to change your time to 1971 to 2010.

Yours.

Ehsan is correct and a covariate is another avenue. Nevertheless, your results for the short period must be consistent  with your full period, otherwise, no strict Econometrican 

will pay attention to your results. Also, be transparent and say that this what you did. Your work will be respected because of it. Good luck !

Better to do the study from 1971. Omit the variable and see the impact from 1966 to 2010. Try to use some other variable as a proxy and see the result.

Mousumi,

Covariates reflect use of a proxy variables

Thanks all for your valuable answers. Since this research is related to international trade I can consider the period starting from 1977, the year which trade liberalization policies introduce. What do you think about this option?

Why is 1977 important in your study? If it is to examine whether trade liberalization has been important; tell me. I will respond right away. Tell me about our model just a little bit. I will soon go to bed  

My research is on the nexus between Trade openness and CO2 emissions. Whether Trade promote or reduce CO2 emissions? 

I'm using ARDL bounds test for cointegration for analysis

I will be glad to  help you althoughI am going to sleep. If you want me to run it for you then send it and I have  my assistants do ARDL and forward to you. Then, you can ask me questions in this medium. Send it as a mail to me.

I forgot send us your email. My office email is [email protected]

Dear Udari Niranjala,

I would suggest you to go for analysis from 1971 to 2010. If you necessarily need to analyse from 1966, you have to generate the missing data as suggested by Ouzzani Fares... If you are using proxy variable for the shorter period, it must be justified by having an uniformity with that of 1971-2010.

Regarding your second question, you can consider your period starting from 1977, but you should try the above options and then compare it with 1977-2010. You may find something interesting!

I hope this helps.

All the best !

You may like to impute missing years using some time-series forecasting/predictive methods

You basically have three choices.  (1) purchase the newest versions of Evews which accounts for this problem, (2) change your sample to 1971-2010, or (3) drop the variable from the regression.

Dear

Can you please go through the text attached. Hope it work.

Good Luck

dear 

change your sample from 1971 to 2010

I see no problem to run a regression from 1971 to 2010.

But I would recommend to run a regression without energy use from 1966 to 2010, too. If energy use is unrelated to the other explaining variables (which can also be checked with the data 1971-2010), then the coefficients of these variables should not change much.

If you have time series of other energy variables (e.g. oil consumption, electricity production) longing back to years before 1971, you could, of course, try to run a regression between these variables and your energy use variable and calculate the missing data on the base of this regression. 

Best wishes

your dataset in larg

you can use 1971.2010 period

Dear all, 

Thank you for your comment. I have solved this question and now analyzing my results. Thank you again.