For me the best way would be to estimate how much extra the development will cost if you have to do it on 2003.
Developers time is expensive. To work out how expensive typically double the average salary of the developers (this is a rough measure of all the additional costs of having that person around - office space, relevant taxes paid by the company, training, other benefits) and then divide by 220 (approximate number of working days in a year (260) after holidays (assume 20), public holidays (assumed 10) and average sick (assumed 10)) to establish the daily cost - so a developer earning $50,000 a year costs about $450 a day.
Then very roughly estimate the difference between doing it one way and the other in days (don't forget on-going support), multiply up by the daily rate and that's the cost.
That should have either eliminated or significantly cut into the cost of the 2008 licenses.
The other thing to do is look at how you can minimise the cost of the licenses - usually by cutting down on the number of servers (this will be over-speced a lot of the time) but also by looking at things such as MSDN which may allow you to get the development licenses for MS products on a better deal.
I'm not knocking things like the on-going support of the OS and the like as arguments, just my experience is that if someone is concerned about money, talk to them about money.
Update: As you're an internal department it may be that there is pressure that can be bought to bear as part of an overall company strategy. The advantage you have here is that in theory you share certain long term goals (as opposed to if you were a consultant or contractor you'd be a gun for hire doing what you were told). It's worth speaking to others within the IT department to see if there are strategic guidelines or standards that help you make your case (perhaps in the most extreme example no new apps running on 2003 servers).