Table 3
Probit Regression: Marginal Probability Effects
(1)
(2)
SQ × T
R-R
0.03
0.04
(.264)
(.24)
SQ × T
R-A
0.09
∗∗∗
0.12
∗∗∗
(.003)
(.004)
SQ × T
A-R
0.07
∗∗∗
0.10
∗∗∗
(.01)
(.009)
SQ × T
A-A
0.01
0.01
(.761)
(.755)
White prize
−0.17
∗∗∗
(.000)
Black prize
−0.31
∗∗∗
(.000)
No. obs.
4287
4287
NOTES: Both specifications include treatment dummies.
P-values in
parenthesis. Standard errors are clustered at the subject level. *** Sig-
nificant at the 1% level.
16
Table 4
Mean and Median of the Status Quo Bias Index
R-R
R-A
A-R
A-A
Mean
0.4
1.2
∗∗∗
1.0
∗∗
0.1
(.277)
(.005)
(.015)
(.764)
Median
0
1
1
0
No. obs
33
43
31
36
NOTES: P-values of a two sided t-test in parenthesis. *** Significant at 1%
level. ** Significant at 5% level.
it serves as the status quo option have a positive index while those exhibiting the opposite
behavior have a negative index. Subjects who make the exact same choices in both parts
of the experiment receive an index of 0.
Figure 1
DISTRIBUTION OF THE STATUS QUO BIAS INDEX
0
.1
.2
.3
−9 −8 −7 −6 −5 −4 −3 −2 −1 0
1
2
3
4
5
6
7
8
9
R − R
0
.1
.2
.3
−9 −8 −7 −6 −5 −4 −3 −2 −1 0
1
2
3
4
5
6
7
8
9
R − A
0
.1
.2
.3
−9 −8 −7 −6 −5 −4 −3 −2 −1 0
1
2
3
4
5
6
7
8
9
A − R
0
.1
.2
.3
−9 −8 −7 −6 −5 −4 −3 −2 −1 0
1
2
3
4
5
6
7
8
9
A − A
17
In Table 4 we report the mean and median level of the status quo bias index across treat-
ments. The mean is not significantly different from 0 in the symmetric treatments. The
median in these treatments is also 0. We find a positive and significant mean index in the
two asymmetric treatments. The median in these treatments is 1. In Figure 1 we report
the full distribution of the status quo bias index across treatments. Consistently with the
aggregate statistics, we observe that the distribution of the index is centered around 0 in
the symmetric treatments while it is shifted to the right in the two asymmetric treatments.
The status quo bias can also be quantified in terms of forgone payoffs. The average
premium required to give up the (10, 4) gamble after receiving it as an endowment is 3.5%
of expected payoffs in treatment A-R and 4% of expected payoffs in treatment R-A. This
means that if a subject is indifferent between options x and y absent an endowment, then,
if x becomes his status quo, in order to restore indifference, the payoff of y has to increase
by 3.5% and 4% respectively.
17
4
Discussion of the Theoretical Literature
One approach to status quo bias associates the emergence of the bias to the incompleteness
of the preference relation (Bewley, 1986) and predicts that the status quo bias may emerge
only in the presence of ambiguity. As stated in the introduction, our study has been
motivated by this approach and more in general by the intriguing potential relationship
17
The quantification is derived from the status quo bias index as follows. In treatment A-R subjects
choose the gamble (10, 4) in one more question on average under the status quo frame compared to their
choice in the neutral frame. Since the subjects are presented with two sets of ordered alternatives they
typically exhibit two switching points. Hence the premium required by the average subject is equal to half
the difference in expected payoff of two subsequent gambles. There is a $1 difference in the prize associated
with a white chip between two successive gambles, or in other words, $0.5 in expectation (assuming a
uniform prior over states for the ambiguous gambles). As our average gamble in the experiment pays $7 in
expectation we roughly estimate the premium to be
0.5
2
7
≈ 0.035 in treatment A-R. In treatment R-A the
average status quo bias index is equal to 1.2 which, following a similar calculation, leads to an estimate for
the bias equal to 4% of expected payoffs.
18
between ambiguity and status quo maintenance. In the setting of Bewley (1986), the agent
faces outcomes which depend on states of the world, the probability of which may not be
objectively defined. The agent acts as if he has a set of priors over the possible states of
the world. Facing alternatives which do not dominate the status quo option for every prior
in that set, the agent exhibits inertia, that is, he sticks to the status quo option.
The decision maker may be thought of as acting “cautiously”: When the ranking of
alternatives is ambiguous (i.e., it changes depending on the prior under consideration),
he keeps his endowment to avoid making a mistake. Ortoleva (2010) provides axiomatic
foundations for such behavior by imposing behavioral postulates on the preferences of the
agent (this model is discussed formally in Appendix B).
The common theme of these models is that inertia may arise only when the ranking
of alternatives depends on the prior under which they are evaluated. Hence, the decision
maker may exhibit inertia only in the presence of ambiguous prospects. Consequently,
these models correctly predict no bias in the R-R treatment.
In the two asymmetric
treatments involving ambiguity there is scope for the emergence of the bias, as confirmed
by our findings. However the models fail to predict the absence of status quo bias in the
A-A treatment. Using the representation proposed by Ortoleva (2010) and under a natural
assumption on the set of priors that the agent holds, treatment A-A is, in fact, the most
prone to status quo biased behavior (we show this in Appendix B). More precisely, if status
quo bias is observed in the asymmetric treatments, then according to this model, it should
also be present in the A-A treatment, contrary to our findings.
In a similar spirit, Mihm (2016) models a decision maker whose ambiguity attitude is
affected by his reference point. While able to explain the findings of Roca et al. (2006),
this model predicts the absence of status quo bias in all cases where the endowment is non-
ambiguous, hence it is at odds with the presence of status quo bias in the R-A treatment,
19