Shortly after any grant award come the part where you have to figure out how to do the work with the money they are actually going to give you, rather than what you budgeted. With the NIH modular budget, there's a bit of room for slop with the numbers. NSF holds you a little closer to the fire when it comes to the total. Therefore, that cushion you might be able to build in an NIH budget is like a wafer thin motel pillow in an NSF budget.
There are ways to make it work, however. Because you have to account for every dime in your budget it is important to add some flexibility without "padding" the numbers.
Every institution will have different resources that need to be considered, but I am often able to use grad student support as a place to cut. This is only true because my college has TA and RA support available for students, allowing me to employ the same number of students but have them paid internally. This also helps if tuition was budgeted into the grant, because internal support covers tuition as well.
Since NSF typically only covers 1 month of summer salary per year in each grant if you have a 9mo appointment, the potential of using overhead return to cover one or more of the months in the grant is something to explore. Up to you to determine whether your department or Dean might consider this option.
Another possibility involves postdocs. If you have a postdoc line budgeted and you know who you intend to hire, figuring out what kind of health insurance they need is useful. Will they need family insurance? This is one spot you want to be careful, however, because if their status changes you will need to adjust accordingly. Possibly playing with fire. **Update** This is NOT a suggestion to favor hiring single postdocs. At all.
Cutting in the equipment category is rarely worth it because there is no overhead associated with this category. Therefore, your cuts here don't affect your bottom line in the same way as cuts elsewhere.
I'm generally loathe to cut travel money. I realize that money can be moved between categories later, but travel is damn expensive and is pretty critical if you have a field component and want to attend meetings.
Here is the category that is typically most nebulous. Certainly you need to make sure you can get the work done, but if you've been smart you have found ways to ensure you can absorb some cut here. This is also the area with the most potential for cross-fertilization between projects. I've never seen a lab that marks certain pipette tips or gloves for certain experiments, depending on where the purchasing funds came from. That may be what the federal bean counters would like to see, but it's impossible in reality. Therefore, the general supplies money of different projects can be considered a pool, relieving the pressure on a single project.
One benefit of federal funding taking forever to become a reality these days is that some aspects of the project really do drop in cost over time. DNA sequencing is a prime example of a technology that gets cheaper every month. For instance, slight cuts in the year two sequencing budget will have little impact on the amount of data produced.
I typically budget money here to fund my broader impacts activities (remember, $$$ makes you accountable!). Once the proposal is funded the opportunity exists to leverage the funding for support within your institution for the activities you have planned. If you get a couple interested parties to chip in towards a workshop, for instance, you can reduce the load on the budget.
Depending on your institution and the support you have from the admin, there may be better places for you to cut your own budget, but this is what has worked for me. It's nearly impossible to get commitments prior to funding, but once you have money in hand, you might be amazed at what people are willing to do to accommodate you.