This definitely represents a prime example of extreme manipulation. It may, however,
have resulted in variable values outside the “relevant range.” BHS’s theory relates
to uncertain events, while there is little or no uncertainty left in the manipulation phrasing.
There is no question they are varying likelihood, which is their theoretical construct,
and arguably there is still some uncertainty involved. Hence, their test is valid.
Likely, however, these descriptions resulted in participants accounting for events that, in
their minds, were certain (not) to occur and hence completely (un)foreseeable. As a consequence,
the adjustment decision is easy and somewhat devoid of tension: clearly, one
should not anticipate an extremely unlikely event by exerting costly adaptive effort,
hence a discretionary bonus adjustment is appropriate should the event occur against all
odds, and vice versa. This raises the question what would happen in cases where event
likelihood moves away from such extreme values. Especially the “extremely likely to
occur” condition does not seem very realistic, and a high likelihood may in practice
usually amount to a 5, 10 or 20 percent likelihood, as opposed to virtually 100 percent.
In such cases, superiors should more carefully consider that the subordinate faces a
trade-off between costly adaptive effort exertion, lowering his/her but also company performance
on average, versus the potential impact of the event with a relatively high
(but still very modest) likelihood. Some amount of business risk is likely considered
acceptable, and costly innovative effort is unwarranted, even when adaptation is possible.
This trade-off is essentially muted in BHS’s experimental design, given that costs
(adaptive effort) and benefits are “certain.” Arguably, this makes for a weaker test and
explains the huge difference in discretionary adjustments between low and high likelihood
conditions (absent compensation interdependence). Given the main focus on the
interaction with compensation interdependence, however, this design choice is not illogical
and in no way invalidates the empirical test. The effect of event likelihood should
be seen as a baseline from which to study the interactive effect of compensation interdependence.
Nevertheless, it limits the contribution of the effect of event likelihood in the
absence of compensation interdependence. Hence, the main effect of event likelihood
should not be interpreted in isolation, and the sensitivity of BHS’s results to the parameters
set for likelihood warrants further investigation.