Controlling unobserved determinants in time series regression

Question

I'm trying to estimate what effect the tariff rate in a specific country has on evasion of tariffs (or rather the underreporting of the value of goods in customs). My dependent variable is a proxy for evasion and the explanatory one is tariffs. I have data spanning continuously over about 15 years.

Now I want to control for other possible unobservable determinants of evasion (such as an increase in customs enforcement or some technological improvement) as robustness and am not sure what would be appropriate (pretty new to econometrics). Would year dummies do the trick (and is this equivalent with "time/year fixed effects"?)? And finally, what would be the difference with estimating a model in first difference? (Can you do this for multiple years as in my dataset?)

Any input would be much appreciated!

Thanks

Oscar

Answer 1

You might consider constructing what is known as an ARMAX Model or a Transfer Function Model in general. Do not assume any differencing or lag structure in your known causals or output series unless you are in solid theoretical grounds. Do perform Intervention Detection to control for unobservable or perhaps simply unkmown deterministic structure. Do allow for parameters to change over tome and do allow for variance changes to be detected and incorporated. Do allow for ARIMA structure to control for unknown ir unspecified stochastic input series. You might review some of the questions/answers , both mine and a few others, to previous time series questions for more enlightenment and support.