Three experiments demonstrated that the experience of power leads to an illusion of personal control. Regardless of whether power was experientially primed (Experiments 1 and 3) or manipulated through roles (manager vs. subordinate; Experiment 2), it led to perceived control over outcomes that were beyond the reach of the power holder. Furthermore, this illusory control mediated the influence of power on several self-enhancement and approach-related outcomes reported in the power literature, including optimism (Experiment 2), self-esteem (Experiment 3), and action orientation (Experiment 3). These results demonstrate the theoretical importance of perceived control as a generative cause of and driving force behind many of power's far-reaching effects. A fourth experiment ruled out an alternative explanation: that positive mood, rather than illusory control, is at the root of power's effects.Here is a bit more detail from the article (slightly edited) on the procedures that produced the predicted results:
Experiment 1 manipulated power by asking participants to recall an experience with high power, an experience with low power, or an event unrelated to power. Illusory control was measured using a classic die-rolling paradigm in which participants are offered a reward for predicting the outcome of a roll and are given a choice of rolling the die themselves or having another person roll the die for them. Choosing to roll the die reflects an illusory sense of control; it indicates that the actor believes he or she can personally influence the outcome of the random roll and, thus, increase the odds of obtaining the reward. We predicted that participants in the high-power condition would be more likely than those in the other two conditions to choose to roll the die.
Experiment 2 tested whether illusory control mediates the established relationship between power and optimism. We manipulated power by instructing participants that they would be matched with a partner and play the role of either a manager or a worker. Before completing any tasks associated with their roles, participants were asked to complete a separate study that was unrelated to their power role and assessed perceived control and optimism.
Experiment 3 tested whether illusory control mediates power's effects on self-esteem and action orientation. We used a context for action—voting in a national election—that is particularly important given that democracies are based on active citizen involvement and are largely shaped by voter mobilization and turnout. Power was manipulated with the same experiential prime used in Experiment 1. We compared the high-power condition with a baseline condition in order to demonstrate that the effects observed in Study 2 were driven by the experience of power and not by powerlessness. After the power manipulation, participants completed measures of sense of control, self-esteem, and action orientation. We predicted that those imbued with a sense of power would demonstrate illusory control, which would mediate power-induced increases in self-esteem and action orientation.
In summary...power led to perceived control over outcomes that were uncontrollable or unrelated to the power. Power predicted perceived control over a chance event (Experiment 1), over outcomes in domains that were unrelated to the source of power (Experiment 2), and over future outcomes that were virtually impossible for any one individual to control (e.g., performance of the national economy, national election results; Experiment 3). Furthermore, this inflated sense of control mediated power's positive effects on optimism (Experiment 2), self-esteem (Experiment 3), and action orientation (Experiment 3).