Most previous attempts to test the adage that “power corrupts” have been unconvincing either because the participants were responding to hypothetical scenarios, or when the situation was real, the incentives for behaving badly were too low to tempt people to corrupt behavior. Two new studies by John Antonakis and his colleagues at the University of Lausanne (Switzerland) were intended to overcome these problems.
The authors define power as “having the discretion and the means to assymetrically enforce one’s will over others.” In the study, they manipulated power by varying the number of others who were dependent on the leader, and discretion, the number of choices available to the leader. Corruption occurs when leaders make decisions that benefit themselves personally, but which harm the common good by reducing the outcomes of their followers. The authors specify that to qualify as corrupt, the behavior must (1) violate social norms, and (2) “destroy public wealth”—that is, the amount gained by the leader must be less than the amount lost by the followers.
Antonakis has prepared this video to explain his two studies.
In the first experiment, participants played the dictator game, in which leaders completely controlled their own and their followers’ outcomes. Power was manipulated by varying the number of followers the leaders had (either one or three), and discretion, the number of options they had (either three, one of which harmed the followers, or four, two of which harmed the followers). A separate study with different participants showed that both of these manipulations made people feel more powerful.
As predicted, both number of followers and number of options increased corruption, as measured by the percentage of leaders who made corrupt choices. Converted to dollars, the participants came away with between $12.30 and $21.24 for a study that lasted less than 15 minutes.
I have a reservation about the manipulation of discretion. Giving some leaders three options and others four complicates the data analysis. The authors deal with this by collapsing the two antisocial choices in the four option condition into a single category. That’s not entirely satisfactory. If you assume the participants responded randomly, 50% of them would make an antisocial choice in the four option condition, while 33% would make an antisocial choice with three options. Their choices can be compared to chance, but not directly to one another. Fortunately, the results are so robust that this problem does not appear to compromise the study.
The second experiment was more ambitious, introducing several new variables not present in the first:
- The definition of corruption states that corrupt behavior violates social norms. After the dictator game was explained to the participants, but before leaders were assigned, participants were asked what they thought a responsible leader should do. Eighty-one percent chose the default option. The poll results were announced before the game began, making it clear to all participants that when leaders made antisocial choices, they were violating a group norm.
- The phrase “power corrupts” is sometimes taken to imply that corruption increases over time. In this experiment participants played 15 trials of the game. Some of the leaders received more power (an additional follower and an additional option) over time. They increased their corruption, while those whose power remained low throughout did not. Payoffs in study 2 varied between $35.47 and $98.06. Unfortunately, since the amount of power possessed by the leaders increased over time, it’s not clear whether their increased corruption was due to the passage of time or their increased power.
- It might be argued that these studies merely show that people are greedy. To check this, the researchers collected a separate behavioral measure of greed (a prisoner’s dilemma game). Their results showed that high power participants became more corrupt even when individual differences in greed were statistically controlled.
- The authors thought that individual differences in honesty and testosterone level might modify the effects of power on corruption. Honesty was measured with a paper-and-pencil test and testosterone from a saliva sample. These measures were collected at a separate session not connected to the main study. The honesty measure added little of interest; it predicted initial level of corruption, but did not interact with power. Testosterone, which was analyzed separately for men and women, did serve to amplify the effect of the power manipulation. The highest level of corruption occurred when high testosterone was combined with high power.
What can be done to reduce corruption? The testosterone results suggest that we should try to place more women in positions of power, but other than that, they have no practical implications. Since announcing the social norm didn’t help, simply shaming powerful people who behave badly probably won’t help. Antonakis suggests that we need greater sanctions to deter corruption, but research is needed to address the practical problems of what the punishments should be and how they can be enforced.
You may also be interested in reading: