Are random effects justified if there is no variability between groups?

Are random effects justified if there is no variability between groups?

I'm looking for more of a substantive ( but also a statistical ) answer.
If in the empty model (just intercept) there is no variability at level-2,
should HLM still be applied just because the data is theoretically nested
(employees in teams) ? Will a linear model suffice in this case?
If you have an answer for me I would appreciate a bibliographical
reference as well.