From SimulationCraft 4.1 scale factor results:
Balance Druids: Hit to cap, then Haste >> Mastery > Crit
Arcane Mage: Hit to cap, Mastery > Crit > Haste (but all are somewhat close)
Fire Mage: Hit to cap, then Haste = Crit > Mastery
Frost Mage: Hit to cap, then Haste > Mastery > Crit
Frostfire Fire Mage: Hit to cap, then Crit > Haste > Mastery
Frostfire Frost Mage: Hit to cap, then Haste > Mastery > Crit
Shadow Priest: Haste > Hit to cap == Mastery == Crit
Elemental Shaman: Hit to cap > Haste > Mastery > Crit
Drain Life Affliction Warlock: Hit to cap > Haste > Crit > Mastery
Affliction Warlock: Hit to cap > Haste > Crit > Mastery
Demonology Warlock: Hit to cap > Mastery > Crit > Haste
Destruction Warlock: Hit to cap > Haste > Crit > Mastery
Number of specs with Haste as #1 non-Hit stat (or equal #1):
9
Number of specs with Crit as #1 non-Hit stat (or equal #1):
2
Number of specs with Mastery as #1 non-Hit stat (or equal #1):
2
Normalized by class:
Haste:
-------
Druid: 1
Mage: 0.6
Priest: 1
Shaman: 1
Warlock: 0.75
== 4.35 / 5 = 0.87
Crit:
------
Mage: 0.4
== 0.4 / 5 == 0.08
Mastery:
---------
Mage: 0.2
Warlock: 0.25
== 0.45 / 5 = 0.09
Thus Haste is by *far* the best caster secondary (non-Hit) stat for the majority of casters.
If you normalize by picking the highest single-target DPS spec for each class:
Mage: Arcane
Warlock: Affliction Drain Life
Haste:
--------
Priest + Shaman + Druid + Warlock
Crit:
------
None
Mastery:
-----------
Mage
So 4 out of 5 caster classes prefer Haste over Crit or Mastery.
Only 1 prefers Mastery over the others.
Even if you make the assumption that there are 3 times as many Warlocks and Mages as each hybrid caster that gives:
Haste:
--------
Priest + Druid + Shaman + 3 * Warlocks == 6
Mastery:
----------
3* Mage == 3
Or 67% prefer Haste.
That and haste for arcane is greatly devalued due to the 4set bonus, you can expect arcane mages want to get more haste when they break it.