The principle of parsimony suggests that as general rule, a regression model as simple should be kept as minimalistic as possible. If a substantial amount of the variation in the independent variable Y can be explained by a few variables, then it is not necessary to add variables as a matter of course. The error term will account for these minimal exclusions.
The logic is to start at a core set of explanatory variables and only add to this selection if a substantial amount of variation has not been explained by this collection of regressors.
Of course, if there is a variable which theoretically seems likely to exert a significant impact on the independent variable, Y, then of course this should be included.
Find us on facebook:
https://www.facebook.com/corkschoolofeconomics
Check out our World-Class Econometrics courses here: