this link might help: https://www.jstor.org/stable/pdf/resrep14771.5.pdf

Why are you looking data for a long period of time? Nowdays, you can use data for a shorter period of time ( 2015-2021), and modelling your variables in Bayesian statistics. Markov chain can stimulate the data into larger samples from small samples. No need to use a long historical data, unless modelling is a problem for you

@ Mwoya Byaro thanks for the response sir, I really appreciate it.

But sir, the topic is given to me by my supervisor for my MSc research thesis with the scope clearly stated sir.

Thanks

@ Oluwaseyi, Thanks for the response but fuel subsidy data is not on WDI sir