We either use ACT1 or others (as above), or we spike in luciferase RNA early during the RNA prep. This is especially good for fractionated mRNAs (e.g. from gradients) where it is not clear how a 'reference' mRNA will behave across the gradient so a spike in allows normalisation in an unbiased way.
We used SCR1 - its a pol III transcript that is at high expression and very stable (used frequently in mRNA decay papers.) and from my experience does not change under many environmental conditions (e.g galactose, stationary phase, heat shock)
The downside is that you cannot used oligo-dT with this one.
Other common genes were mentioned earlier - ACT1, PMA1, rRNA (e.g. RDN18-1 mentioned above).
However, even ACT1 and PMA1 mRNA levels can change under different environmental conditions or stresses.
Spiking in luciferase RNA is a very good idea, but unfortunately not feasible for my experiences.
I will also have a look at TAF10 or SCR1, as e.g. PMA1 is one of many genes I cannot use.. I have checked the publication by Teste et al. but wanted to know further experiences.