'An Experimental Test of the Effectiveness of Terms & Conditions' (Northwestern Law & Economics Research Paper No. 13-32) by Zev J. Eigen
comments that
Requiring individuals to consent to “terms & conditions” is the overwhelmingly dominant strategy used to try to curb unauthorized use of products like motion pictures and music. This study is the first to employ a randomized controlled behavioral experiment testing whether this strategy is as effective as other means of achieving this goal. Individuals randomly assigned to either a “terms and condition” (“T&C”) frame or alternative frames (promise-keeping, trust, threat, naked request, and a control) were presented an opportunity to take an online presidential election poll more than once (and receive additional remuneration each time they did), even though they were made aware that they were not authorized to do so. The T&C frame was the least effective at keeping subjects from taking the poll more than once. Asking individuals to promise not to behave in the undesirable way, or signaling trust that they would not behave in the undesirable way were the best frames for curbing unauthorized multiple poll-taking.
Eigen reports that
The findings suggest that terms and conditions may be a suboptimal means of stymying
undesirable behaviors like taking a survey more than once or the comparable real world
behavior of unauthorized use or distribution of music or other media. Only 65% of subjects
randomly assigned to the terms and conditions frame took the survey once and then stopped.
Compared to the higher rates of other frames—especially the promise frame (77%), and the
trust frame (80%)—the findings suggest that the behavioral justification for using terms and
conditions deserves reconsideration. It is less clear which of the alternatives to terms and
conditions should be used. Some evidence from this study suggests that further research is
necessary to understand what variation exists across contexts that makes the frame appropriate
to the quality of the unwanted behavior being curbed. In the instant study, one finding in
particular makes this question salient. 27% of the trust frame multiple survey-takers were onetime
recidivists, as compared to 49% of the T&C multiple survey-takers. This may be an odd
result—why would subjects who reject the trust plea behave less compliantly than those who
reject the terms and conditions means of attempting to restrict behavior? This research does
not offer an answer to this question. One may speculate that this is a function of reactance
theory (Brehm 1966) associated with the rejection of the plea. The more negatively one feels as
a result of the decision to reject the frame, the greater the subjects’ need to self-compensate in the way of taking the survey as many times as possible. However, the instant study does not address this suggested mechanism.
While the promise frame generated a higher compliance rate than the T&C frame –
77.4%, as did the trust frame (80%), it is not clear from the results of this study that promise or
trust worked better than other options for preventing the unwanted behavior—specifically, the
naked request frame. This could be a function of the experiment being underpowered. Further
empirical work is necessary to test whether this is the case. It could also be that a naked request
not to take the survey more than once is inherently interpretable as the researchers placing
their trust in the subjects not to take the survey more than once. There is less of an explanation
for the relative success of the fraud frame.
It also seems that political orientation, choice of president, and gender are measures not
associated with compliant behavior. However, age and disagreement with the need for voter
identification laws are associated with greater compliance. Both of these associations present
key bases for future research that would directly impact the form-contracting arena. In terms of
the voter identification question, this question could be regarded as a proxy for measuring the
extent to which subjects believe that others cheat, or that cheating of a more serious kind than
is being measured is socially acceptable. This may offer a useful measure for future research
because it is significantly easier to measure the degree to which people think that others engage
in the kind of dishonest behavior of interest to researchers. It also harmonizes with existing
research that suggests that people are more likely to behave dishonestly when they perceive
that behavior as being socially acceptable (McCabe and Trevino 1993). In the context of
downloading media illegally, this finding could not be more salient.
Lastly, there appears to be some preliminary evidence to support the existence of cohort
effects in the results of this study. How much faith one puts in the specific age related findings notwithstanding, it is worth noting that subjects respond to the frames differently by age cohorts. Terms and conditions seemed to motivate subjects over 37 differently from subjects
under 30, especially those 18-21. Further research would help illuminate whether and to what
extent the preliminary findings are replicable by age cohort. This could be extremely important
as our population ages, and more individuals become normalized and perhaps desensitized to
terms and conditions. Contracts are social artifacts (Suchman 2003). They likely carry different
meanings for individuals who have more experience with negotiated instruments than those
who are more accustomed to non-negotiated instruments as the normal way of governing
exchanges.
