where the status quo is a risky lottery.
A second strand of the literature views the status quo bias as stemming from loss
aversion (Kahneman and Tversky, 1979). The two features which generate the status
quo bias as a prediction of the loss aversion model are: (i) The decision maker evaluates
outcomes in terms of gains and losses with respect to a reference point (typically chosen
as the status quo option if it exists, or set to 0 if it is absent); and (ii) losses loom larger
than gains. A recent non-parametric method by Abdellaoui et al. (2016) estimates the
parameters of this model and finds that the utility and the loss aversion parameter are
the same under risk and ambiguity. Thus, given our controlled design across treatments
in terms of odds and payoffs, loss aversion predicts a positive status quo bias in all of our
treatments. Hence, this approach does not explain the heterogeneity of our findings, i.e.,
why status quo bias is observed only in the asymmetric treatments.
18
One concern is that the reason for the absence of the bias in the symmetric treatments
is the fact that we can only observe choices over a finite set of alternatives. If the loss
aversion parameter is small, setting the difference in expected value between two consec-
utive gambles at 50 cents may be too large an interval and impair our ability to detect a
preference reversal across the two frames of choice. We address this point in Appendix A
using the version of the loss aversion model suggested by Koszegi and Rabin (2006) and
with a standard loss aversion parameter found in Sprenger (2015) which is also in line
with previous studies (Kahneman and Tversky, 1992; Pope and Schweitzer, 2011; Gill and
Prowse, 2012).
19
We show that our experimental setup is well calibrated: The adopted
18
Some theoretical papers have already raised the concern that loss aversion may be at odds with the
status quo bias phenomenon. Masatlioglu and Ok (2014) show that loss aversion may actually lead to
anti-status quo bias predictions. They illustrate that an agent abiding to loss aversion may choose x over y
in the absence of an endowment, but reverse his choices, and choose y over x when endowed with x. Their
example is a development of comments regarding the loss aversion model made even earlier, by Munro and
Sugden (2003) and Sagi (2006).
19
The Koszegi and Rabin (2006) model of reference dependence is well suited when gambles serve as
reference points as in our experiment.
20
grid of alternatives is fine enough to detect a bias if one is present.
A third modeling approach to the status quo bias has been proposed by Masatlioglu
and Ok (2014). Their decision maker acts as a “constrained maximizer”. In the absence
of a status quo option he is a standard rational agent. The presence of a status quo option
induces a psychological constraint set from which the agent chooses the best alternative
according to his utility. The status quo bias is captured by the fact that some alternatives
which are utility improving compared to some alternative x, may be excluded from x’s
constraint set and therefore will not be chosen when x serves as the endowment. While the
model accommodates status quo bias in general, it does not provide ex-ante predictions as
to which treatments in our experiment should give rise to it. As the constraint set has no
specific structure, the model is compatible with our findings but would also be compatible
with the opposite findings.
Our results show a novel pattern of choice where status quo bias appears only in choices
between a risky and an ambiguous gamble. This pattern highlights the dissimilarity be-
tween the status quo and the alternative option as a possible determinant for the status
quo bias. There are two reasons why this may happen. First, the status quo may shift pref-
erences towards objects with similar characteristics. In the realm of uncertainty, evidence
of such preference shifts has been documented by Dean (2008) and Dean et al. (2017) who
find that a risky status quo increases the chances that a risky alternative is chosen over a
certain monetary payoff. More evidence can be found in Sprenger (2015) who reports an
endowment effect for risk, i.e., the higher tendency to take up risk when the endowment
is risky. A preference shift towards the endowment’s type of uncertainty may explain why
we find status quo bias in choices between a risky and an ambiguous option, but not when
all options are risky or all are ambiguous.
The second reason supporting the idea of dissimilarity-based status quo bias is that
21
similar options may be easier to compare.
20
This seems very plausible in the context of
our experiment where all risky alternatives share the same risk profile and the ambiguous
gambles share a similar source of ambiguity. When options are easily compared, it stands
to reason that the agent will have a clear ranking of alternatives and choose according to
it, and irrespectively of the frame of choice. On the other hand, sticking to the status
quo may be appealing when the ranking is more difficult or impossible to determine. This
hypothesis in fact echoes the original idea of Bewley, namely, that inertia is a behavioral
phenomenon that helps resolve indecisiveness in favor of the status quo when preferences
are incomplete.
A recent paper by Maltz (2016) formalizes the idea of a dissimilarity-based status quo
bias. Adopting the Masatlioglu and Ok (2014) framework, his set up specifies a partition
structure on the space of alternatives, interpreted as categories of goods similar to each
other (such as risky or ambiguous gambles in our experiment). In his representation, the
decision maker has an (endowment-free) utility function and he settles his choice problem
according to the following procedure: First, he identifies the best alternative which is
similar to his endowment, i.e., in the same category as the endowment, and that alternative
serves as his reference point. Next, this reference point induces a constraint set from which
the decision maker chooses the best feasible alternative according to his utility.
An important feature of this model is that it predicts rational choice in the presence of
alternatives which belong to a single category. This is due to the first stage, in which the
agent identifies the best alternative in his endowment’s category to serve as the reference
point. If no goods from other categories are available for choice, this stage, in fact, identifies
the best alternative overall. In other words, in these contexts, the constraint set plays no
20
As noted by Schwartz (2000): “Within each category, it may be relatively easy to express preferences.
Between categories, however, expressing preferences is more problematic.” Schwartz uses the term categories
to refer to goods of similar nature (or, “similar kind of things” using Schwartz’s words).
22
Dostları ilə paylaş: |