Lies, Damned Lies, and University Overheads

For the uninitiated, one of my responsibilities as an academic research is to raise money. Mobile systems research like my group does requires a fairly large number of graduate students, and usually their tuition and very modest stipends are paid using external funds. For example, currently I’m working with eight Ph.D. students 2, five of whom are externally supported by three different active grants. Two others work for the department as teaching assistants, which is another way that Ph.D. students support themselves during their degree. The last one is self-supported, which is rare.

So when you do receive external awards, you want the money to go as far as possible.I don’t want to digress into the history and current challenges associated with research funding—​a topic for another post, or several—​but let’s just say that external funding is not easy to come by. I spend more of my time than I would prefer writing grant proposals, and most funding programs are extremely competitive. I actually don’t have too many complaints about the National Science Foundation grant application process, except maybe that it’s too slow. Fundamentally, there are just too many people trying to eat the same pie. Overall, the result is that at least at present external funding is hard to obtain. So when you do receive external awards, you want the money to go as far as possible.

That’s where overheads come in. Every grant request contains two components:

Unlike direct costs, which can be included as line items in the budget, it would be a nightmare and probably impossible to budget for indirect costs. So instead, they are typically assessed as a percentage of the direct costs included in the grant. Actually it’s way more complicated than that, but I’m writing a long blog post, not a David Foster Wallace novel, so I’ll spare you the gory details of how the "Modified Direct Cost" is calculated. Except to point out that this futzing has funny effects. For example, at some institutions researchers encourage their students to avoid purchasing equipment that is under a certain amount because it will incur overheads, while larger equipment purchases do not. So go ahead and add a multi-TB Flash RAID array to that new desktop!

So far, so good: I don’t think any academics have a problem with the idea of indirect costs. But here’s where things start to get problematic.

The first interesting observation is that overhead rates vary widely between institutions. Here are ten examples for research universities where I was able to find their overhead rates online 4:

Without more context, you might conclude that it costs 50% more in overheads to do the same research at Harvard compared to North Dakota State. But the second problem with overhead rates is that they are impossible to compare, because they don’t reflect many other differences between institutions, including how research is supported, other costs associated with research, and how overheads are used. All of these institution-specific policies end up either increasing or decreasing the negotiated overhead rate. (It’s kind of like loopholes in the U.S. tax code, except they seem to only ever have a downward effect.)

Examples of these differences include:

So how does this all work out for in my case? Well, UB has a 59.5% overhead rate (fairly high), returns no portion of the overheads to the PI (sad), charges full out-of-state tuition to grant-supported foreign Ph.D. students (and most of them are foreign), does not require me to budget academic year salary (although other departments do) but does require me to budget computer service fees. Putting everything in three big categories, here’s how it all works out:

So 29% of my grant budget goes to overheads. Maybe that’s not so bad. But it’s also almost $23K for a single year. That’s enough to hire half of a well-paid administrator, and I certainly don’t have anywhere near that much administrative support—​not total, much less per award. Alternatively, it’s almost enough to add another half-student to the award, and given how hard my students work that would make a huge difference. As a final comparison, at North Dakota State, North Dakota’s top-ranked research university, 6 the same project would cost $5.7K (or 7.3%) less. Why?

Plus the usual caveats apply. Although UB’s tuition rates are quite reasonable, the $17K that I pay for tuition is three times more than it would be for an in-state student. And the direct costs include $2K of computer service fees that as far as I can tell constitute extra overhead.

All of the inter-institution differences between research policies make it hard to have a conversation about whether your institution is doing a good job of providing a supportive research environment. And in my experience, that conversation is hard enough anyway. I’ve asked several university administrators at UB about why our overhead rates are so high. Here’s how that chat usually goes:

No: the line about snow removal is not made up. I have heard that used as an explanation before. But beyond it’s transparent silliness, it also points to another problem with overheads, which is that in a lot of cases the things people claim they are covering are actually mixed-use facilities that are probably actually being paid for by tuition dollars. As far as I know, UB has no special parking lots or spots for faculty researchers 7, and the lots need to be plowed when classes are in session. So that’s something that should be (and probably is) paid for with tuition revenue, not with research overheads.

I think that academic researchers suspect that lurking behind all of these differences are the real issues of institutional administrative competence, efficiency, and planning. Unfortunately, all of these differences make it hard to perform the apples-to-apples comparisons required to determine whether your own institution creates a competitive financial environment for research or not.

So let’s try an apples-to-apples comparison. If you’d like to participate, please complete this survey. It includes a variety of questions about your institution and shouldn’t take too long to complete. An optional (but extremely important) component asks you to create a sample one-year budget similar to mine above, and that should provide the best way to compare the impact of at least some of the differences I outlined above.

Assuming a decent number of responses, I’ll return to this topic in the spring and present an analysis of the survey responses. I look forward to your contribution. And if you have any suggestions are other comments, feel free to email me.