Two things, one anecdotal, one scientific:
(1)
When I was in 6th grade, they had some "kid executives" thing where we got to spend a day shadowing a local "businessperson." I got to follow around a senior VP of Bank of America. Not exactly the bank with the best reputation. But while I was there, when I was away from the SVP, a junior level guy walk talking with me, and I must have asked something about needing to be a jerk to succeed in business, because he told me something I never forgot: "you know everyone says you need to be a jerk to succeed, but this guy is the clearest example that that's not true." Anecdotal, yeah, but it stuck with me even as I grew up (read: became slightly more cynical). I've since met far wealthier, or far more recognized individuals. Maybe hearing that once when I was young enough to embrace it encouraged me not to write off all powerful people because of the actions of a notorious few.
(2)
I think the study has some flaws. When I was an undergraduate, I worked in a psychology lab that did studies like this. They don't bring in a bunch of "powerful people" or anything like that. They have a bunch of other undergraduates play the Prisoner's Dilemma with 5 bucks. Here's the methods for one of the experiments[1]:
Participants and Procedure. Sixty-six Stanford University students (62.9% female; age: M = 21.2, SD = 3.1) were recruited from a large subject-pool. They arrived at the laboratory in groups of 8-12 participants and were each seated in a private cubicle. Participants read the instructions, made their decisions and responded to the post-decision questionnaire using the computer.
The Intergroup Prisoner’s Dilemma (IPD) game. Participants learned that they had been randomly assigned to a four-person group, and that their group was randomly yoked to another four-person group. They did not know who was in their group or in the other group and could not communicate with any of the other participants in their session. Each participant was endowed with 10 game chips and had to decide how to allocate them. Each chip that was kept paid $2 to the individual; each chip that was contributed to the group pool added $1 to each in- group member including the contributor; in addition, it subtracted $1 from each out-group member. Thus, the same action simultaneously benefitted in-group members and harmed out- group members.[sic]
My problem with studies like this, while commendable for trying to experimentally control possible factors, is that the context is super important. An analogy might be measuring the speed of light in the same gas several times to control for external effects, but not realizing that the speed of light through a gas changes based on the gas. One such context, as the David Hornik pointed out, is a temporal one: people who screw everyone over once might seem aggressive and powerful, and are signalling that they're independent, whatever. People who screw everyone over for years might get left behind or ousted.
To be fair, the authors of the study acknowledge (but do not address) the temporal issue in their conclusion. And model systems like the Prisoner's Dilemma are important in any field. But twitchy when lab results like this are applied willy nilly into complex social interactions. Nice guys, bad guys, both evolved under intense evolutionary pressure for a reason.
[1] http://www.gsb.stanford.edu/news/research/halevy_nice_2011.h... (link to study on right)