Thursday, August 15, 2013
University Presidents: Pay for Performance or Agency Problem?
Our story so far: nonprofit managers have reason & opportunity to manage their firm in a way that increases their personal rewards. Even if the dollar amounts are small, the managerial consequences could be significant. For instance, we saw last time evidence that university CEO pay, tuition, and expenditures rose sharply for most of the last decade while returns for faculty and for-profit CEOs were relatively flat on average. One can at least tell stories about why a president who wanted to be paid more would increase tuition or overall spending. How can we test those stories? It’s tricky. Universities are complicated places, and lots of factors could be at work. Maybe presidents were just doing a great job, and invested their extra resources wisely, making college a better deal and earning higher pay for themselves.
Our paper therefore tries to look at the extent to which CEO pay is correlated with measures of “agency costs.” That is, is pay lower at schools where presidents are monitored by other stakeholders more closely? If so, that would be evidence at least that presidents’ opportunism contributes to pay levels, though of course it wouldn’t rule out the “pay for performance” alternative I just mentioned--both could be contributing factors.
Long story short, we used a school’s dependence on donations to measure how closely its president was watched. Few donors really take much role in details of school governance; they are rationally ignorant and hope to free ride on watchdog efforts by others. But it turns out that if you make someone angry (or give them a really good feeling from participating) they don’t free ride as much -- their emotions motivate them to be more actively involved. We hypothesize that donors don’t like excessive pay, and that the threat of donor “outrage” would constrain pay levels.
And that’s what we find.
First, we find that donors don’t like learning about high president pay. In one set of regressions, we find that each dollar of reported pay reduces giving in subsequent years (controlling for all the other important observable facts about the university we could measure) by about $30. For instance, if Harvard could have been Harvard while paying its president zero dollars, our results imply it could have pulled in another $30-$40 million in donations.
In another (not on ssrn yet), we look at the effect of being singled out by the Chronicle of Higher Education’s list of the “Top 10 Most Highly Compensated” presidents. Presidents who make the list see about $6 million less in donations, even relative to presidents who just missed making it.
We then find that being more dependent on donations (i.e., getting a bigger fraction of revenue from gifts) does reduce presidents’ pay. Presidents who are more than one standard deviation above average (the top 1/6 or so) in dependence on donations get paid an average of about $110,000 less, all else equal. We also find that showing up in the Top 10 list one year tends to slow the rate of growth of a president’s pay, though obviously there are people like Gordon Gee who just show up every year.
It’s admittedly not immediately obvious how this relates back to the tuition story, and you are already bored, so I’ll fill in what we think the implications of our findings are next time.
TrackBack URL for this entry:
Listed below are links to weblogs that reference University Presidents: Pay for Performance or Agency Problem?:
Good idea for a paper. What's especially nice about this, which perhaps you mention in the paper itself, is that if presidents are paid for performance, then we'd expect a school with more donations to have HIGHER pay not just that it wouldn't matter. That's the sort of thing one often sees with private company executive pay, where the high pay is the result of bonuses based on stock price.
Perhaps it woudl be interesting to look at the corporate directors of the schools with the highest pay relative to size and other such characteristics. I'm not sure what variables you'd look at for self-dealing directors, though.
Posted by: Eric Rasmusen | Aug 19, 2013 4:26:09 PM