uncertainty structure. Notice however, that their finding in isolation is also consistent with
both the loss aversion approach and Bewley (1986). Varying the type of uncertainty of the
status quo and the alternative, we are able to pinpoint precisely where the main existing
theories fall short in explaining the data.
Our study is also related to the literature on the endowment effect which refers to
individuals’ higher valuation of a good when they own it compared to when they do not
(Thaler, 1980) and can be interpreted as status quo bias in choices where one of the two
options is money. Several studies provide evidence for the endowment effect for both risky
and ambiguous gambles.
9
However, as one would expect, no effect shows up when trade
involves only monetary payoffs (Kahneman et al., 1991). Combining these findings with
ours, we find that status quo bias is absent in choices among monetary payoffs, risky
lotteries and ambiguous gambles, but it does emerge in choices across the aforementioned
categories. Thus, the findings of the endowment effect in the realm of uncertainty further
strengthen the pattern of our results, emphasizing the role of dissimilarity among options
in the choice set as a potential determinant of status quo bias (see Section 5 for a more
elaborate discussion). The exploration of such hypothesis seems a worthwhile pursuit for
future work, both within the domain of uncertainty and beyond it.
The paper is organized as follows: Section 2 describes the experimental design and in
section 3 we highlight the main results. Section 4 discusses different theoretical models
in light of our findings while Section 5 explores the findings alongside evidence from the
endowment effect literature. Section 6 concludes.
9
See for example Knetsch and Sinden (1984) and Eisenberger and Weber (1995).
7
2
Experimental Design
Experiment overview. We run four treatments corresponding to all combinations of
a risky or ambiguous status quo option with risky or ambiguous alternatives. In each
treatment we adopt a within-subject design where subjects make the same choices under
two different frames. In the first part of the experiment subjects make a series of choices
between a fixed gamble and different alternative gambles under a neutral frame: They are
sequentially presented with pairs of gambles and for each pair are asked to choose their
preferred option. In the second part, subjects receive the gamble which was fixed in the
first part as their endowment and face the same comparisons. This time around, the ques-
tions are presented in a status quo frame, i.e., as a decision between keeping the endowment
or switching to the alternative. In each treatment, status quo bias is observed if subjects
choose the endowment more often in the second part, when they own it, compared to the
first part when they do not. As elaborated later in this section, special aspects of the
design ensure that the fixed alternative does not play the unintended role of a reference
when choices are taken under the neutral frame. The experiment was pencil-paper based
and conducted at the CESS lab in NYU. It involved 143 subjects among the NYU under-
graduate population. All subjects received a show-up fee of $8 and earned on average a
total of $16. The duration of the experiment was approximately 30 minutes.
Experimental procedure: The Risky-Ambiguous (R-A) Treatment. We outline
the experimental procedure for treatment R-A, where the status quo option is risky and
the alternative is ambiguous. Following this outline, we explain how the design is modified
in the other three treatments. The instructions can be found in the online Appendix.
At the beginning of the experiment subjects are introduced with two bags. First, the
8
known bag which consists of a known composition of poker chips - 50 white and 50 black.
The second bag is the unknown bag and remains empty until all tasks have been completed.
Subjects are told that at the payment stage this bag will be filled with X white chips and
100 − X black chips, where X equals the two decimals of the Dow Jones Industrial Average
Index at the time of payment.
10
This mechanism ensures that subjects’ beliefs regarding
the composition of the unknown bag remain fixed throughout the experiment as we further
explain later in this section.
Following this introduction, participants proceed to the first part of the experiment
which entails 15 pairwise choices between a fixed gamble from the known bag and different
gambles from the unknown bag. The fixed gamble from the known bag pays $10 if a white
chip is drawn and $4 if the chip is black (henceforth the (10, 4) gamble). Thus, a typical
question compares the (10, 4) gamble on the known bag, with a (w, b) gamble from the
unknown bag, which pays $w if a white chip is drawn from that bag and $b if the chip is
black. Table 2(b) lists the prizes of the 15 alternative gambles in treatment R-A (the prizes
for the other treatments are shown in 2(a) and 2(c)). In the paper we will sometimes refer
to these gambles as the alternative set. Nine additional questions not involving gamble
(10, 4) are interspersed in between the 15 questions listed in Table 2(b) (for a total of 24
questions) and are intended to reduce the salience of gamble (10, 4).
11
Upon completion
of part 1, subjects participate in a non-incentivized intermission in which they answer
8 questions presented in the same format as in part 1 but involving larger stakes. The
intermission serves the purpose of separating the two main parts of the experiment and is
discussed in the online Appendix.
10
The value of the index was verified online with a volunteer subject at the payment stage.
11
A typical additional question is a choice between a (p, q) gamble on the known bag (where p = 10, q = 4)
and a gamble performed on the unknown bag. As the aim of part 1 is to elicit preferences in a neutral
frame of choice, the 9 additional questions prevent the gamble (10, 4) from accidentally playing the role of
a reference which repeatedly appears in all questions. The full list of questions is available in the online
Appendix.
9