without getting into too much detail
the additional cost of simultaneous localization is caused by churn and inconsistency. it's like trying to paint a picture of a movie. by the time you're finished, you've had to re-paint half of what you did up front. this is unavoidable. games in development are continually in flux.
if you want to minimize cost of localization, a stable, completed product will always provide the best foundation for minimum word count (translation is generally cost per word) and minimum time investment.
at that point you can spend as much or as little as you like on quality. localization generally consists of three steps (translation, editing, proofreading or TEP) - you can do just some of those steps, but you're going to have a shitty product if you skimp out. many games skimp out.
whether you bill the localization to Sega of America or Sega of Japan is completely irrelevant
whether you use in-house localization resources or outsource the translation is also irrelevant. though generally speaking, outsourced localization companies usually use freelancers so they're going to be significantly cheaper than in-house resources, though at the cost of quality/context
