Cross Cultural Differences in Decisions from Experience: Evidence from Denmark, Israel, and Taiwan

Running head: CULTURAL DIFFERENCES IN DECISIONS FROM EXPERIENCE 1

Cross Cultural Differences in Decisions from Experience:

Evidence from Denmark, Israel, and Taiwan

Sibilla Di Guidaa, Ido Erevb, Davide Marchioric

a Corresponding author, Department of Business and Economics, COHERE, University of

Southern Denmark, Campusvej 55, 5230 Odense M, Denmark;. Tel: +45 6550 7295, Email:

[email protected]

b Faculty of Industrial Engineering and Management, Technion, Technion City, Haifa 32000,

Israel, [email protected]

c Strategic Organization Design Unit, Department of Marketing and Management,

University of Southern Denmark, Campusvej 55, 5230 Odense M, Denmark,

[email protected]

CULTURAL DIFFERENCES IN DECISIONS FROM EXPERIENCE 2

Cross Cultural Differences in Decisions from Experience:

Evidence from Denmark, Israel, and Taiwan

Abstract

Three studies are presented that compare decisions from experience in Denmark, Israel, and

Taiwan. They focus on two change-related cultural differences suggested by previous

research on dialectical vs. analytic approach to thinking. The first implies that East Asians

are more likely to change their behavior over time (i.e., are less consistent), the second that

they expect more changes in the environment. The results show that the “less consistency in

the East” hypothesis has a high predictive value. This hypothesis accurately predicts a

behavioral pattern that was documented in all three studies, as well as a non-trivial effect of

limited feedback in Study 3: When feedback was limited to the obtained payoff, the

participants from Taiwan exhibited less risk aversion than the Israeli. Analysis of the

“expecting more changes in the East” hypothesis reveals mixed results. This hypothesis was

supported in Study 2, which examined relatively complex multi-alternative multi-outcome

tasks, but not in Studies 1 and 3, which examined simple two-alternative two-outcome choice

tasks. A possible explanation for the different predictive value of the two examined

hypotheses is discussed.

Key words: The hot stove effect; underweighting of rare events; decisions from

experience; clicking paradigm; recency effect; dialectical thinking.

JEL classification: C91; D81; D83.

PsycINFO classification: 2343; 2930.

Word count: 9,799


1. Introduction

Previous studies of judgment and decision making in different cultures reveal two interesting

change-related differences between individuals from East Asian and Western cultures. The

first is that Easterners tend to change their opinions over time more often than Westerners.

For example, Yates et al. (1998) and Wallsten and Gu (2003) show that this tendency can

explain why Chinese people appear to be more overconfident than North Americans (see

Yates et al., 1989). The second is that Easterners expect more changes in the environment

than individuals from Western cultures (Ji et al., 2001, and 2008; Spencer-Rodgers et al.,

2010). For example, when presented with a graph summarizing the trend of the economic

growth rate or of the worldwide death rate for cancer, participants from the Peoples’ Republic

of China (PRC) were about twice as likely as Americans to predict a reversal of the trend in

the next period (Ji et al.’s Study 2, 2001).

Both change-related cultural differences can be explained by the higher propensity of

East Asians to tolerate contradictions (Peng & Nisbett, 1999). This tendency is commonly

referred to as dialecticism (see discussions in Wallsten & Gu, 2003, and in Ji et al., 2001).

On the one hand, dialecticism reduces the effort to be consistent,1 thus resulting in a higher

rate of changes in opinion; on the other, it induces a vision of an environment in constant

mutation, thus leading Easterners to expect more changes in the environment.

Previous attempts to examine the effect of cultural differences on decision making

have mainly focused on risk preferences and risk perception in decisions from description (à

la Kahneman & Tversy, 1979), i.e., in situations in which decision makers are provided with

1 Typically, when facing seeming contradictions, people of Chinese culture “retain basic elements of opposing

perspectives by seeking a ‘middle way,’” adopting a compromise approach (Peng & Nisbett, 1999:741). In

contrast, Westerners are more likely to emphasize contradictions in the attempt to logically evaluate alternative

perspectives, as heritage of the Aristotelian logic and, in particular, of its principle of tertium non datur.


full and accurate description of the incentive structure. These studies have shown, for

example, that the Chinese from the PRC are remarkably more risk seeking than North

Americans in decisions from description with hypothetical stakes, although this difference is

significant only in the financial/economic domain (Weber & Hsee, 1998; Hsee & Weber,

1999).2

The current paper explores the implications of the two mentioned change-related

cultural differences in the context of repeated decisions with feedback. Specifically, we

focus on pure decisions from experience, i.e., situations in which decision makers are not

provided with prior information about the incentive structure, but can only rely on the

outcomes of their past decisions. Our interest in exploring cultural differences in the context

of decisions from experience has two main motivations. First, recent research has shown that

choice behavior can greatly differ depending on whether decision makers can rely on the full

description of the problem at hand, or only on the observed outcomes from each of the

possible alternatives (cf. Section 1.2; Barron & Erev, 2003; Erev & Haruvy, 2014). This

result is commonly referred to as the experience-description gap (Hertwig & Erev, 2009).

Second, most of the previous studies on experiential decision-making were run with

“WEIRD” participants (i.e., from Western, Educated, Industrialized, Rich, and Democratic

Countries; see Henrich et al., 2010).

Therefore, the present investigation has two main goals. The first is to understand the

effects of the highlighted change-related cultural differences on choice behavior in decisions

2 Weber et al. (1998) hypothesize that the higher rate of risk seeking in the PRC can be explained by the

collectivist nature of that society: Compared to an individualistic society such as that of the USA, people in

mainland China are more likely to receive financial help if they are in need, and, consequently, are less risk

averse. Weber et al. call this the cushion hypothesis.


from experience, the second to evaluate the generality across cultures of the main behavioral

regularities observed in decisions from experience.

1.1 Baseline Experimental Design and Hypotheses

To pursue our two goals, we designed and ran three experimental studies, which are variants

to the basic “clicking paradigm” described in Figure 1. In this basic design, participants are

just told that the experiment will include many trials, and that their task at each trial is to

select (click on) one of the two unmarked keys that are presented on the screen. Each

selection is followed by the presentation of the outcomes from both keys. Importantly,

participants are not given any information about the payoff rule that generates the (stochastic)

outcomes from each option (key). Therefore, participants can assess the value of each option

only based on the observation of past outcomes. Although extremely simple, this paradigm

reliably replicates the main properties of operant conditioning and experiential learning (Erev

& Haruvy, 2014, provide a comprehensive review of the empirical studies adopting this

methodology).

The two change-related cultural differences mentioned in the introduction lead to two

hypotheses about behavior in the decisions from experience setting.

The first hypothesis is that of “less consistency in the East.” This hypothesis predicts

more changes in the preferred option (key) by Easterners, as a consequence of the more

frequent changes in opinion typical of people from dialectical cultures. Therefore, under this

hypothesis, the probability of repeating the previous choice (tendency referred to as inertia) is

predicted to be lower for East Asians.

The second hypothesis is that of “expecting more changes in the East.” As another

possible effect of dialecticism, Easterners are predicted to be less likely to choose the option

(key) that has yielded the best payoff in the previous trial. This is because East Asians expect


more changes in the environment, and, for this reason, are less likely to believe that the

option that has yielded the best outcome in the previous trial will do so also in the future.

---

Figure 1

---

1.2 Behavioral Regularities in Decisions from Experience

The current analysis examines the impact of cultural differences on five of the most important

regularities documented in previous studies of decisions from experience. In order to clarify

these regularities and their measurement, we focused on the six problems described in Table

1. The right hand columns and the lower rows of Table 1 summarize the results of two

experimental studies of these problems (Di Guida et al., 2012, and Erev & Haruvy, 2013) that

were conducted at the Technion (Haifa, Israel) using the clicking paradigm.

Notice that all six problems involve choice between the Status Quo (i.e., an option

that yields a payoff of zero with certainty), and an action that changes the status quo (which

we refer to as “Action option”). The participants faced each problem for at least 100 trials,

were not given prior information about the payoff rule behind each alternative, received

feedback about the payoff from each alternative after each choice, and were paid a show up

fee of 25 Israeli Shekels (1 Shekel ≈ $0.25) plus the payoff (in Shekels) from one randomly

selected trial.

1.2.1 Underweighting of rare events

The best known property of decisions from experience is the tendency to underweight rare

events (Barron & Erev, 2003; Hertwig et al., 2004; Rakow & Newell, 2010; Barron &

Ursino, 2013). Repeated experience with the choice task increases the tendency to select the

option that has yielded the best payoff most of the time, even when this choice behavior


impairs expected return. This tendency was documented in distinct experimental paradigms

(see Shafir et al., 2008; Ungemach et al., 2009; Danziger et al., 2014) and can be explained as

a product of reliance on small samples3.

The current analysis of this tendency focuses on the comparison of behavior in

Problems 1 and 2 (Table 1). Notice that in both problems, the action option alternative to the

status quo involves an extreme and rare (10%) outcome. The experimental results reveal

deviation from maximization in both problems: In Problem 1, most participants preferred the

Status Quo over an Action option with positive expected value (i.e., the prospect “+10 with p

= 0.1; -1 otherwise”): Only 27% (SD = 0.24) preferred the Action option; on the contrary, in

Problem 2, most participants (57%, SD = 0.29) preferred the Action, with negative expected

value (the prospect “-10 with p = 0.1; -1 otherwise”), over the status quo. Both deviations

can be explained by the hypothesis that subjects favor the option that lead to the best payoff

in 90% of the trials, and do not pay enough attention to the rare (10%), but extreme,

outcomes. That is, individuals behave as if they underweight rare events.

The underweighting score in the lower section of Table 1 presents the common

quantification of this effect. More in detail, this score is derived considering behavior in

Problems 1 and 2: Preferring the “Status Quo” option in Problem 1 and the “Action” option

in Problem 2 is an indication of insufficient sensitivity to the (rare) event occurring 10% of

the time. Therefore, the average of the difference between 0.5 (rate that indicates

indifference between the two options) and the rate of Action choices in Problem 1, and of the

difference between the rate of Action choices and 0.5 in Problem 2 can be used to assess this

3 Reliance on small sample implies a tendency to underweight rare events because rare events are likely to be

underrepresented in small samples. For example, the probability that a 10% event will be represented in sample

of 10 is only 0.38. The tendency to rely on small samples, in turn, can be the product of cognitive limitations

and/or a sophisticated attempt to response to patterns (Hertwig & Erev, 2009).


tendency. In particular, a significantly positive score can be interpreted as an overall

indication of underweighting of rare events in the two problems.

1.2.2 The payoff variability effect

Another robust behavioral regularity emerging in decisions from experience is the payoff

variability effect (Myers & Sadler, 1960; Busemeyer & Townsend, 1993): An increase in the

variability of payoffs moves decisions toward random choice, even when this bias impairs

expected returns and increases the probability of losses.

Problems 3, 4, 5, and 6 (Table 1) were designed to test for this effect. More in detail,

Problems 5 and 6 involve a choice between the Status Quo and an Action option that yields a

certain payoff of, respectively, 1 and -1. The results show that adding variability to the

payoff from the Action option, without changing its expected value (as in Problem 3 and 4),

reduces the tendency to choose it (see results reported in Table 1). Therefore, comparison of

the Action choice rate in Problems 5 and 3 and in Problems 6 and 4 suggests that the addition

of variability moves choice behavior toward random choice. Notice that these results cannot

be accounted for by any pattern of choice behavior that assumes stable aversion to risk and/or

losses.

The payoff variability score reported in the lower section of Table 1 presents the

common quantification of this effect. Specifically, this score is obtained by averaging the

difference between the maximization rates in Problem 5 and 3 (i.e., the difference between

the Action choice rates in the two problems), and the difference between the maximization

rates in Problem 6 and 4 (i.e., the difference between the Status Quo choice rates in the two

problems). A significantly positive score can be interpreted as a decrease in the

maximization rate in correspondence to an increase in the variability of payoffs, thus

providing indication of the payoff variability effect.


1.2.3 Strong inertia

Analysis of decisions from experience reveals high inertia rates, i.e., a strong tendency by

participants to repeat their last choice. This pattern is demonstrated by the inertia column in

Table 1 that presents the proportion of trials in which subjects repeated their last choice. The

mean values over the four problems with payoff variability (lower panel in Table 1) show

that the inertia rate is .82, (SD = 0.14).

1.2.4 Best reply

Most models of learning (e.g., Erev & Roth, 1998; Camerer & Ho, 1999; Marchiori &

Warglien, 2008) assume a positive recency effect. That is, people are expected to prefer the

alternative that has led to the best payoff in the most recent trial. Experimental results

support this assumption: The best reply column of Table 1 presents the proportion of trials in

which subjects selected the option that led to the best payoff in the very recent trial. The

mean over the four problems with variability is 0.62 (SD = 0.15), and this value is

significantly larger than 0.5 (the rate expected under no best reply, t[177] = 10.90, p < .001),

and significantly lower than the inertia score (t[177] = 11.98, p < .001).

1.2.5 Risk aversion

Studies of decisions from experience document large differences in risk attitudes in different

problems. For example, the results discussed above suggest risk aversion in Problems 1 and

3, and risk seeking in Problems 2 and 4. In addition, the results show that the aggregated

behavior over the four problems reflect a slight bias toward risk aversion. The risk-rate over

the four problems (lower part of Table 1) is 0.43.


---

Table 1

---

2. Study 1

Study 1 replicated the experiments summarized in Table 1, and was run in three different

locations: At the Lab@SDU (University of Southern Denmark, Odense, Denmark), at the

Interdisciplinary Center (IDC) Herzliya (Israel), and at the AI-Econ Lab of the National

Chengchi University (Taipei, Taiwan). In these three locations we could recruit participants

with different cultural backgrounds, in particular for what concerns the propensity to

dialectical thinking. Therefore, the replication of this experimental study in the three

mentioned locations allowed us to check for the robustness to cultural background of the five

properties of decisions from experience summarized in Table 1, and to test our hypotheses

about the two change-related cultural differences described earlier.

2.1 Design

Participants played for 201 times (200 decisions with feedback) each of the six binary

problems presented in Table 1. The order of the problems was independently randomized for

each participant.4 Participants were told that the experiment would include many trials and

that their task would be to choose between two unmarked keys that appear on the screen.

Participants were given no information about the problems’ incentive structure. After each

selection, the payoffs from the selected and the non-selected keys were presented on the

4 In spite of the apparently huge number of choice problems (clicking tasks) faced by each participant (i.e.,

1,206), the duration of the whole experiment was about 35 minutes. In addition, within each cultural group, we

did not observe any significant difference in choice behavior in the same problem when it was encountered at

the beginning (first 201 trials), or at the end of the experiment (trials 1,006 to 1,206).


screen (complete feedback). The Status Quo and Action options were randomly assigned to

keys, independently for each participant, and each option was maintained in the same

position in all trials of the same problem. Participants were told when each problem had

terminated.

2.2 Instructions and payment

In Denmark and in Israel, instructions were given in English, whereas in Taiwan both the

English and the Chinese versions were provided (the two-language method mimics the

original Technion study with instructions both in English and Hebrew). The translation to

Chinese Mandarin was made by a native speaker fluent in English. It is worth noting that in

the clicking paradigm, due to the simplicity of the task, instructions are minimal (see Figure

1), and their impact on participants’ performance is limited. The two buttons on the screen

were simply referred to as the “right” and “left” alternative. Instructions appeared on the

screen at the beginning of the experiment.

Participants were paid according to the amount of experimental points they earned

with their decisions during the experiment. Specifically, participants received an initial

endowment of 150 experimental points, plus the positive or negative outcome (in

experimental points) from six of their choices (one for each problem), randomly selected at

the end of the experiment.

The conversion rate between experimental points and the local currency was clearly

stated in the instructions. In order to make payments’ purchasing power comparable across

the three Countries, we rescaled payoffs according to the “The Economist’s Big Mac Index.”

Thus, one experimental point was worth 0.4 Danish Kroner, 0.2 Israeli Shekels, and 1

Taiwanese Dollar, so that the initial endowment of 150 points was equivalent to the price of

two Big Mac sandwiches.


2.3 Participants

Thirty-two Danish students (15 females, Mage = 22.7 years, age range: 19-28 years), thirty-

one students from IDC (16 females, Mage = 24.6 years, age range: 22-33 years),5 and forty-

eight Taiwanese students (34 females, 14 males, Mage = 21.7 years, age range: 19-26 years)

served as participants in the experimental sessions.

2.4 Results

Study 1 results are presented in Table 2, using the same format of Table 1. The first three

summary statistics reported in the lower section of Table 2 (see Tables A1 and A2 for

pairwise comparisons) show the robustness of the main properties of decisions from

experience.6 The difference between the risk rate in the three locations is not significant, F(2,

108) = 1.87, ns. The underweighting of rare events score is .46 (SD = 0.32) in Denmark, .43

(SD = 0.39) in Israel, and .53 (SD = 0.35) in Taiwan, and the group effect is not significant,

F(2, 108) = 0.87, ns. The payoff variability score is .32 (SD = 0.25) in Denmark, .38 (SD =

0.21) in Israel, and .37 (SD = 0.22) in Taiwan, and the group effect is not significant, F(2,

108) = 0.51, ns. The payoff variability and underweighting scores are significantly larger

than zero in all three groups (for the underweighting score, t[31] = 8.16, p < .001 in

Denmark; t[30] = 6.20, p < .001, in Israel; and t[47] = 10.57, p < .001, in Taiwan; for the

payoff variability score, t[31] = 7.43, p < .001 in Denmark; t[30] = 9.77, p < .001, in Israel;

and t[47] = 11.74, p < .001, in Taiwan), providing significant indication for these two effects.

Moreover, the underweighting and payoff variability scores as well as the risk rate observed

5 Twenty-one Israeli students and ten students from the IDC International Psychology and International

Business Programs (four from the US, three from Italy, two from France, and one from Venezuela). The other

two cultural groups included, respectively, only Danish and Taiwanese participants.

6 We did not find significant gender effects on choice behavior (cf. analysis reported in the Supplemental

Material).


in the two Western locations are not significantly different from those observed in the

previous Technion studies summarized in Table 1 (see the Supplemental Material for

pairwise comparisons).

Analysis of the inertia rates confirms the prediction of the “less consistency in the

East” hypothesis. It reveals a smaller inertia rate in Taiwan (M = .71, SD = 0.10) than in

Denmark (M = .77, SD = 0.08) and Israel (M = .75, SD = 0.12), F(2, 108) = 3.77, p = 0.03, η2

= .07. In particular, the Taiwanese are observed to be significantly less likely than the Danes

to repeat their previous choice (see Table A2 for detailed pairwise comparisons).

In contrast, our generalization of the “expecting more changes in the East” result was

not supported. Analysis of the best reply scores reveals higher best reply rates in Taiwan (M

= .79, SD = 0.18) than in Denmark (M = .68, SD = 0.12) and Israel (M = .72, SD = 0.18). The

group effect is significant (F[2, 108] = 4.26, p = .02, η2 = .07), although the only significant

pairwise difference is between Denmark and Taiwan (see Table A1). Therefore, contrary to

what hypothesized, the Taiwanese behaved as if they expected fewer changes, as they were

more likely to select the option that had yielded the highest payoff in the previous trial.

Comparison of the present best reply and inertia rates to those from the reviewed Technion

studies (Table 1) reveal similar patterns in the three groups with Western participants.

---

Table 2

---

Under one explanation for the inaccuracy of our “expecting more changes in the East”

hypothesis, the results suggest a qualitative difference between decisions from experience and

prediction tasks. That is, it is possible that Easterners expect more changes when are asked to

predict the future (as in Ji et al., 2001), but not when they decide based on past experience. A

second explanation assumes that the critical difference between Study 1 and Ji et al.’s results


involves the number of possible outcomes. It is possible that Easterners expect more changes

when the number of possible outcomes is large (as in Ji et al. tasks predicting continuous

rates tasks), but the opposite hold in Study 1’s two-outcome tasks. One justification to this

“number of outcomes” explanation rests on the following two assumptions: First, some

people may believe that, in the current state of nature, the payoff from the action option tends

to alternate between positive and negative outcomes (see Rapoport & Budescu, 1997). Given

this belief, expecting a continuation of the current state implies a low best reply rate (e.g.,

selecting the Status Quo option after a positive outcome from the action option), whereas

expecting a change (namely, a termination of the alternating pattern) implies a high best reply

rate. Second, people are more likely to expect more alternations in two-outcome choice tasks

(like those of Study 1) than in multi-outcome tasks (like in Ji et al.’s prediction tasks). Notice

that this latter explanation suggests that it is not easy to derive the predictions about behavior

of the “expecting more changes in the East” hypothesis: Such an exercise would require

additional assumptions concerning the participants’ belief about of current state of nature,

and these beliefs can be creative. Yet, the current explanation suggests that our original

prediction (i.e., lower best reply rate in the East, cf. Section 1.1) is more likely to hold when

the number of possible payoffs is large.

Study 2 was designed to compare these explanations by examining decision problems

with multiple outcomes for each alternative, as well as problems with more than two possible

alternatives.7

3. Study 2

Study 2 replicated the second study in Ert and Erev (2007), and was run in Taiwan (National

Chengchi University, Taipei) and in Israel (Technion, Haifa). It used the clicking paradigm,

7 We decided to leave the study of cultural differences in pattern detection to future research.


described earlier, and focused on the six problems presented in Table 3. The six problems

included two baseline problems (7.1 and 8.1), and two variants to each of these problems.

The baseline problems 7.1 and 8.1 imply choice between a two-outcome risky alternative

(labeled with R), and a safer prospect S with five possible outcomes.

Compared to the six problems considered in Study 1, in problems 7.1 and 8.1 task

complexity is increased by adding some noise to the outcome from the safe prospect.

Problems 7.3, 8.3, 7.25, and 8.25 further increase task complexity by increasing the number

of available alternatives at each trial.

Problems 7.3 and 8.3 used the same payoff distributions as in the baseline problems

7.1 and 8.1, with the exception that two “replicas” were added to each alternative (cf. Figure

2). That is, at each trial, subjects saw a total of six keys on the screen: The outcomes from

three keys were independent draws from payoff distribution R, and the other three were

independent draws from payoff distribution S.

Problems 7.25 and 8.25 included 24 replicas of each alternative (thus, participants

could choose among a total of 50 keys at each trial). At each trial, the different replicas

yielded an independent draw from the reference distribution R or S. Figure 2 presents the

typical experimental screens under the three “replicas” conditions.

---

Figure 2

---

Notice that in the baseline problems, perfect best reply to recent payoffs implies a

frequency of R choices (henceforth, R-rate) of 10%: That is, extreme underweighting of the

rare event in Problem 8.1 (when the R option maximizes expected return), and strong risk

aversion. In contrast, in Problems 7.25 and 8.25 the probability that at least one of the 25

risky prospects will lead to a large gain is very high (.91 in 7.25, and .93 in 8.25). Thus,


perfect best reply implies overweighting of rare event in Problem 7.25 (when choosing R

impairs expected return), and high risk seeking rates.

3.1 Design

As in Study 1, at each trial, subjects had to select (click on) one of the unmarked keys

presented on the screen, and were given feedback about the outcomes from both the selected

and unselected keys (complete feedback).

Each subject faced each of the six problems for 50 trials. The order of the problems

was independently randomized for each subject, as well as the order of the keys presented on

the screen. Participants were not given any information about the exact length of the

experiment, but were just told that it would include six distinct phases. In addition,

participants were informed when each problem (phase) had terminated.

3.2 Instructions and payment

Instructions were translated in the local language by native speakers, and appeared on the

screen at the beginning of the experiment both in the local language and in English. The

conversion rate between experimental points and the local currency was communicated at the

beginning of the experiment, and was the same as in Study 1 (one experimental point was

worth 1 TWD, or 0.2 ILS).

Participants were paid according to the amount of experimental points gained during

the experiment. Specifically, participants received a show up fee of 24 Shekels or 120 TWD,

plus a bonus corresponding to the experimental points cumulated in one of the phases

(randomly selected at the end of the experiment).


3.3 Participants

One experimental session was run at the Technion (Haifa, Israel), whereas the other at the

AI-Econ Lab of the National Chengchi University (Taipei, Taiwan). Twenty Israeli students

(9 females, Mage = 24.2 years, age range: 20-28 years) and twenty-two Taiwanese students

(15 females, Mage = 22.5 years, age range: 20-26 years), who did not participate in Study 1,

served as participants in this study.

3.4 Results

Study 2 experimental results are presented in Table 3. Analysis of aggregate choice rates

shows no significant cultural differences. The data from Israel and Taiwan show very similar

risk rates (respectively, M = .34, SD = 0.20, and M = .33, SD = 0.17; F[1, 40] = 0.01, ns),

maximization rates (respectively, M = .52, SD = 0.08, and M = .51, SD = 0.07; F[1, 40] =

0.04 , ns), and underweighting of rare events scores (respectively, M = .15, SD = 0.23, and M

= .16, SD = 0.18; F[1, 40] = 0.02, ns).8 As already observed in Ert and Erev (2007) (see also

Grosskopf et al., 2006), the joint effect of an increase in the number of available alternatives

and of the tendency to select the alternative with the highest recent outcome dramatically

lowers the tendency to underweight rare positive outcomes. However, in both cultures, the

underweighting rates are significantly larger than zero (t[19] = 2.90, p = .01, in Israel; and

t[21] = 4.02, p = .001, in Taiwan).

Analysis of the inertia rates shows the robustness of the pattern documented in Study

1 and predicted by the “less consistency in the East” hypothesis. The inertia rate in Taiwan

8 We found no significant gender effects on risk rates (cf. analysis reported in the Supplemental Material).


(M = .43, SD = 0.19) is smaller than that in Israel (M = .55, SD = 0.19). In the current study,

this effect is significant at the 5% level in a one-tail test, t(40) = 2.01, p = .03.9

Analysis of the best reply rates reveals a reversal of the pattern documented in Study

1. The best reply rate averaged over all problems is .41 (SD = 0.13) in Taiwan, and .53 (SD =

0.21) in Israel. This difference is significant, F(1, 40) = 4.80, p = .03, η2 = .11, and robust:

The best reply rates in Taiwan are smaller than the rates in Israel in all six problems. Thus, in

the current context, our generalization of Ji et al. (2001) finding is accurate: The results

suggest that the subjects from Taiwan behaved as if they expected more changes in the

environment. The difference between the two locations is particularly clear in the multi-

alternative problems (7.24, and 8.24). The best reply rate in these problems is .31 (SD =

0.28) in Israel, and .17 (SD = 0.16) in Taiwan. This result suggests that the inconsistency

between Study 1 results and the “expecting more changes in the East” pattern documented in

Study 2 (as well as in Ji et al., 2001) is not a reflection of a qualitative difference between

prediction tasks and decisions from experience. Rather, the relationship between the

expectation of changes and the tendency to best reply to the most recent outcome is likely to

depend on subjects’ beliefs about the nature of the current state of nature, and these belief can

be creative. Thus, it is not easy to predict the impact of this hypothetical cultural difference

in the context of decisions from experience.

---

Table 3

---

9 The marginal significance of this result in a two-tail t test might be due to the limited number of subjects

involved in this study. However, we refer to this effect as “robust” in the sense that we observed consistently

lower inertia rates for the Taiwanese participants across all three studies.


4. Study 3

Study 3 was designed to test a non-obvious implication of the “less consistency in the East”

hypothesis on risk behavior when feedback information is limited to the obtained outcome.

Specifically, this third experimental study was identical to Study 1, but for the fact that, at

each trial, participants were given feedback only about the outcome from the selected key (as

in a 2-armed bandit problem).

Previous research has shown that when the feedback is limited to the obtained

outcome, learning increases risk aversion (Denrell & March, 2001; Denrell, 2007; Erev and

Haruvy, 2014). This phenomenon, known as the hot stove effect, is a consequence of the fact

that bad payoffs from a particular option reduce the probability of choosing this option in the

future and of collecting more information about its value. In contrast, good payoffs increase

the probability of collecting more information. The slower adaptation after bad payoffs

implies that the impact of bad payoffs is expected to last longer than the impact of good

outcomes. For example, a sequence of -9 outcomes from the action option in Problem 3 can

lead subjects to stop choosing this attractive action and believe that its expected value is

negative. When feedback is limited to the obtained payoff, this belief will not be revised

unless the subjects choose to explore this apparently unattractive option again.

Notice that this logic implies that frequent changes in choice behavior are expected to

reduce the hot stove effect, as they would facilitate the collection of additional information

about the risky alternative. Therefore, when feedback is limited to the obtained payoff, the

frequent changes in the preferred alternative implied by the hypothesis of “less consistency in

the East” is expected to lead to a comparatively lower level of risk aversion in the East.

Study 3 tests this prediction.


4.1 Design, instructions, and payment

The design was identical to that of Study 1, with the only exception that feedback was limited

to the payoff from the selected key (partial feedback).

4.2 Participants

One experimental session was run at the Technion (Haifa, Israel), whereas the other at the

AI-Econ Lab of the National Chengchi University (Taipei, Taiwan). Twenty Israeli students

(5 females, Mage = 23.8 years, age range: 19-32 years) and forty-eight Taiwanese students (34

females, Mage = 21.9 years, age range: 19-32 years), who did not participate in either of the

previous experiments, served as participants in this study.

4.3 Results

The experimental data reported in Table 4 confirm the implication of the “less consistency in

the East” hypothesis mentioned earlier. That is, in comparison to the full feedback design

adopted in Study 1, limiting feedback information to the outcome from the selected key

reveals cultural differences in risk attitudes: In Study 3, we observe a significantly higher

propensity to take risk in Taiwan (M = .48, SD = 0.20) than in Israel (M = .35, SD = 0.23),

F(1, 66) = 5.01, p = .03, η2 = .07.10 In addition, comparison with the results from Study 1

suggests a stronger hot stove effect in Israel than in Taiwan, i.e., a stronger tendency by

Israeli participants to avoid the Action choice after a negative outcome. This effect can be

quantified, within each cultural group, by the difference between the risk rate under the

complete and partial feedback conditions. The hot stove score is 0.12 (SD = 0.20) in Israel,

and 0.002 in Taiwan (SD = 0.16). Although the observed differences in the risk rates support

10 We found no indication for gender effects on choice behavior (cf. analysis reported in the Supplemental

Material).


our conjecture, a two-way analysis of variance only indicates a marginal interaction effect of

the feedback and group factors, F(1, 143) = 3.64, p = 0.06, and the main effects of group and

feedback are not significant.

Confirming the pattern observed in the previous two studies, the inertia rate for the

Taiwanese participants (M = .76, SD = 0.15) is significantly lower than that observed for the

Israeli (M = .85, SD = 0.12), F(1, 66) = 5.39, p = .02, η2 = .08. The lower inertia rate

confirms that the higher risk rate observed for the Taiwanese participants can be explained by

a stronger propensity to explore more the two alternatives. This comparatively higher

propensity to exploration is compatible with the stronger propensity to dialectical thinking

documented in East Asian cultures.

Comparison of the best reply rates shows similar values in the two cultures. This rate

is .54 (SD = 0.08) in Israel, and .53 (SD = 0.10) in Taiwan. The difference across the two

groups is not significant, F(1, 66) = 0.41, ns. This similarity is consistent with the hypothesis

that, in simple two-outcome two-alternative choice tasks, Easterners do not expect more

changes in the trends than do Westerners.

In addition, limiting feedback information similarly increases the weighting of the rare

events in the two groups. A two-way analysis of variance of the underweighting scores in

Studies 1 and 3 shows a significant main effect of the feedback factor, F(1, 143) = 33.19, p <

.001, whereas the main effect of the group factor and the interaction effect of the group and

feedback factors are not significant.

---

Table 4

---


5. General Discussion

The current paper investigates the effect of two change-related cultural differences that were

documented in previous studies comparing judgments by individuals from East Asian and

Western cultures. The first implies that East Asians are more likely to change their behavior

over time, whereas the second implies that they expect more changes in the environment.

The two cultural differences are similar in the sense that they can be described as reflections

of the higher degree of tolerance to contradictory pieces of information, or dialectical

thinking, featuring East Asian cultures (see Peng & Nisbett, 1999, Wallsten & Gu, 2003, and

Ji et al., 2001). Yet, in the context of decisions from experience, they lead to different

predictions.

The present investigation shows that the “less consistency in the East” hypothesis has

a high predictive value. The pattern predicted by this hypothesis was documented in all three

studies, and in Study 3 it correctly predicts the non-trivial effect of limited feedback: When

the feedback is limited to the obtained payoff, the participants from Taiwan exhibit less risk

aversion than the Israeli.

Our analysis of the hypothesis “expecting more changes in the East” reveals mixed

results. The results of Study 1 reveal higher best reply rates in Taiwan than in Israel. Thus,

they seem to indicate expectation of fewer changes by Easterners. In contrast, Study 2 results

show lower best reply rates in Taiwan, whereas Study 3 results document similar best reply

rates in the two locations. These results suggest that it is not easy to derive the prediction of

the hypothesis of “expecting more changes in the East” in the context of decisions from

experience. The implications of “expecting more changes” depend on subjects’ beliefs

concerning the current state of nature, and it seems that these beliefs are creative and highly

sensitive to the small differences between our experiments.


We also find that the basic properties of decisions from experience (i.e., the

documented tendency to underweight rare events, the payoff variability effect, and the larger

tendency to inertia than to best reply in choice behavior) appear to be robust to the cultural

background. These properties of experiential decision-making, in particular the tendency to

underweight small probability events, have been shown to have important practical

implications for policy making and mechanism design in Western culture contexts (e.g., see

studies on enforcement of safety rules at the work-place or of safe medical procedures, as

pointed out in Erev & Roth, 2014).

The high predictive value of the “less consistency in the East” hypothesis prompts the

question of what are the psychological processes that underlie it. In particular, in what terms

is dialecticism likely to affect choice behavior in our experimental settings? Our favorite

answer rests on two observations. First, previous research show that the main properties of

decisions from experience can be captured with the assumption that people try to select the

option that led to the best payoff in similar situations in the past (see Gonzalez et al., 2003;

Erev & Haruvy, 2015). Second, studies of dialecticism suggest that this is a rather broad

concept that refers to a whole set of lay beliefs by people from Asian cultures, including not

only tolerance for contradiction and the expectation of change, but also holistic thinking and

cognition (Spencer-Rodgers et al., 2009 and 2010). Indeed, previous cross cultural studies

(e.g., Nisbett & Norenzayan, 2002, and Dong & Lee, 2008) have shown that dialecticism also

affects the way people collect and process visual information: Whereas East Asians tend to

pay more attention to the entire context/field and to its causal relations (holistic style of

cognition), Westerners tend to detach objects from their field and pay less attention to the

entire field (analytic style of cognition). Thus, it is possible that dialectical thinking reduces

consistency in decisions from experience, as it increases the variety of the criteria subjects

can use to judge similarity between present and past experiences.


Hsee and Weber (1999) explained the stronger risk seeking tendency in the PRC with

the cushion hypothesis. According to this hypothesis, compared to an individualistic society,

people in mainland China are more likely to receive financial help if they are in need, and, for

this reason, are less risk averse. Our results from Study 3 complement this observation by

highlighting another possible contributor, i.e., explorative behavior, for the lower risk

aversion observed in East Asia.

One attractive feature of the current results is that they clarify the relationship

between cultural differences in judgment confidence and in risk attitudes. Previous studies of

these issues reveal three phenomena that appear to be inconsistent. First, as noted above,

Yates et al (1989) found that Chinese subjects exhibit more overconfidence in probability

judgments. For example, when their Chinese subjects estimated that the probability that they

have provided the correct answer by 90%, they were only correct in 60% of the cases; the

American exhibited weaker overconfidence: Their 90% estimates were accurate in 75% of

the cases. Second, Hsee and Weber (1999) found higher tendency to take risk in China than

in the USA (although this cultural difference appears to hold only in the financial domain; see

Weber et al., 1998). Finally, Yates et al. (1997) found that the fact that Chinese state they are

highly confident in the correctness of their answers does not imply that they are willing to

pay more (than Americans) for gambles that yield a gain if they are in fact correct. Thus,

relative to their overconfidence, Chinese take less risk than Americans. The current analysis

highlights the same sufficient condition to all three phenomena. The hypothesis that

individuals from East Asian cultures are more likely to change their opinion is a sufficient

condition for overconfidence (see Wallsten and Gu, 2003), masks different risk preferences

across cultures (as shown in Study 3), and also implies less confidence in the accuracy of

previously expressed opinions.


In summary, the current analysis shows that the hypothesis that individuals from East

Asian cultures are likely to change their behavior over time has high predictive value.

Previous research shows that this hypothesis can explain, for example, why Chinese subjects

appear to be more overconfident than North Americans. The present results show the

robustness of the stronger tendency to change in the East in three decisions from experience

studies, and that this tendency correctly predicts less risk aversion by Asian participants when

feedback is limited to obtained payoffs.


Acknowledgements

This research was supported by the Max Wertheimer Minerva Center, a grant from the Israel

Science Foundation, a grant from the “Fonds de la Recherche Fondamentale Collective”

(grant no. 2.4614.12), and a grant from the National Science Council of Taiwan (grant no.

101-2410-H-004-002). The valuable assistance of Heng-Hui Lin in the collection of data is

gratefully acknowledged.


References

Barron, G., & Erev, I. (2003). Small feedback-based decisions and their limited

correspondence to description-based decisions. Journal of Behavioral Decision

Making, 16, 215-233

Barron, G., & Ursino, G. (2013). Underweighting rare events in experience based decisions:

Beyond sample error. Journal of Economic Psychology, 39, 278-286

Busemeyer, J. R., & Townsend, J. T. (1993). Decision field theory: A dynamic-cognitive

approach to decision making in an uncertain environment. Psychological Review, 100,

432-459

Camerer, C., & Hua Ho, T. (1999). Experience‐weighted Attraction Learning in Normal

Form Games. Econometrica, 67, 827-874

Danziger, S., Hadar, L., & Morwitz, V. G. (2014). Retailer pricing strategy and consumer

choice under price uncertainty. Journal of Consumer Research, 41, 761-774

Denrell, J. (2007). Adaptive Learning and Risk Taking. Psychological Review, 114, 177-187

Denrell, J., & March, J. G. (2001). Adaptation as information restriction: The hot stove effect.

Organization Science, 12, 523-538

Di Guida, S., Marchiori, D., & Erev, I. (2012). Decisions among defaults and the effect of the

option to do nothing. Economics Letters, 117, 790-793

Dong, Y., & Lee, K. P. (2008). A cross-cultural comparative study of users’ perceptions of a

webpage: With a focus on the cognitive styles of Chinese, Koreans and Americans.

International Journal of Design, 2, 19-30.

Erev, I., Ert, E., Roth, A. E., Haruvy, H., Herzog, S. M., Hau, R. , Hertwig, R., Stewart, T.,

West, R., & Lebiere, C. (2010). A choice Prediction Competition: Choices from

Experience and from Description. Journal of Behavioral Decision Making, 23, 15-47


Erev, I., Haruvy, E. (2014). Learning and the economics of small decisions. Invited chapter

submitted to Kagel, J.H. and Roth, A.E. (Eds.), The Handbook of Experimental

Economics. Princeton University Press.

http://www.utdallas.edu/~eeh017200/papers/LearningChapter.pdf

Erev, I., & Roth, A. E. (1998). Predicting how people play games: Reinforcement learning in

experimental games with unique, mixed strategy equilibria. American Economic

Review, 88, 848-881

Erev, I., & Roth, A. E. (2014). Maximization, Learning and Economic Behavior. PNAS, 111,

10818–10825.

Ert, E., & Erev, I. (2007). Replicated alternatives and the role of confusion, chasing, and

regret in decisions from experience. Journal of Behavioral Decision Making, 20, 305-

322

Grosskopf, B., Erev, I., & Yechiam, E. (2006). Foregone with the wind: Indirect payoff

information and its implications for choice. International Journal of Game Theory,

34, 285-302

Gonzalez, C., Lerch, J. F., & Lebiere, C. (2003). Instance-based learning in dynamic decision

making. Cognitive Science, 27, 591-635

Henrich, J., Heine, S. J., & Norenzayan, A. (2010). The weirdest people in the world?

Behavioral and Brain Sciences, 33, 61-83

Hertwig, R., Barron, G., Weber, E. U., & Erev, I. (2004). Decisions from experience and the

effect of rare events in risky choice. Psychological Science, 15, 534-539

Hertwig, R., & Erev, I. (2009). The description-experience gap in risky choice. Trends in

Cognitive Sciences, 13, 517-523

Hsee, C. K., & Weber, E. U. (1999). Cross-national differences in risk preferences and lay

predictions. Journal of Behavioral Decision Making, 12, 165-179


Ji, L.-J., Guo, T., Zhang, Z., & Messervey, D. (2009). Looking into the past: Cultural

differences in perception and representation of past information. Journal of

Personality and Social Psychology, 96, 761-769

Ji, L.-J., Nisbett, R E., & Su, Y. (2001). Culture, Change, and Prediction. Psychological

Science, 12, 450-456

Ji, L.-J., Zhang, Z., & Guo, T. (2008). To buy or to sell: Cultural differences in stock market

decisions based on price trends. Journal of Behavioral Decision Making, 21, 399-413

Marchiori, D., & Warglien, M. (2008). Predicting human interactive learning by regret-driven

neural networks. Science, 319, 1111-1113

Myers, J. L., & Sadler, E. (1960). Effects of range of payoffs as a variable in risk taking.

Journal of Experimental Psychology, 60, 306-309

Nisbett, R. E., & Norenzayan, A. (2002). Culture and cognition. In H. Pashler & D. L. Medin

(Eds.), Stevens Handbook of Experimental Psychology : Cognition (3d Ed., Vol. 2)

(pp. 561-597). New York: John Wiley & Sons.

Peng, K., & Nisbett, R. E. (1999). Culture, Dialectics, and Reasoning about Contradiction.

American Psychologist, 54, 741-754

Rakow, T., & Newell, B. R. (2010). Degrees of uncertainty: An overview and framework for

future research on experience‐based choice. Journal of Behavioral Decision Making,

23, 1-14

Rapoport, A., & Budescu, D. (1997). Randomization in individual choice behavior.

Psychological Review, 104, 603-617

Shafir, S., Reich, T., Tsur, E., Erev, I., & Lotem, A. (2008). Perceptual accuracy and

conflicting effects of certainty on risk-taking behaviour. Nature, 453, 917-920


Spencer-Rodgers, J., Boucher, H. C., Mori, S. C., Wang, L., & Peng, K. (2009). The

dialectical self-concept: Contradiction, change, and holism in East Asian cultures.

Personality and Social Psychology Bulletin, 35, 29-44

Spencer-Rodgers, J., Williams, M. J., & Peng, K. (2010). Cultural Differences in

Expectations of Change and Tolerance for Contradiction: A Decade of Empirical

Research. Personality and Social Psychology Review, 14, 296-312

Ungemach, C., Chater, N., & Stewart, N. (2009). Are probabilities overweighted or

underweighted when rare outcomes are experienced (rarely)? Psychological Science,

20, 473-479

Wallsten, T. S., & Gu, H. (2003). Distinguishing choice and subjective probability estimation

processes: Implications for theories of judgment and for cross-cultural comparisons.

Organizational Behavior and Human Decision Processes, 90, 111-123

Weber, E. U., Hsee, C. K., & Sokolowska, J. (1998). What Folklore Tells Us about Risk and

Risk Taking: Cross-Cultural Comparisons of American, German, and Chinese

Proverbs. Organizational Behavior and Human Decision Processes, 75, 170-186

Yates, J. F., Lee, J.-W., & Bush, J. G. (1997). General knowledge overconfidence: Cross-

national variations, response style, and “reality”. Organizational Behavior and

Human Decision Processes, 70, 87-94

Yates, J. F., Lee, J.-W., Shinotsuka, H., Patalano, A. L., & Sieck, W. R. (1998). Cross-

Cultural Variations in Probability Judgment Accuracy: Beyond General Knowledge

Overconfidence? Organizational Behavior and Human Decision Processes, 74, 89-

117

Yates, J. F., Zhu, Y., Ronis, D. L., Wang, D. F., Shinotsuka, H., & Toda, M. (1989).

Probability judgment accuracy: China, Japan, and the United States. Organizational

Behavior and Human Decision Processes, 43, 145-171


The current experiment includes many trials. Your task, in each trial, is to click on one of the

two keys presented on the screen. Each click will be followed by the presentation of the

keys’ payoffs. Your payoff for the trial is the payoff of the selected key.

Figure 1. The typical instructions screen in studies of decisions from experience adopting the “clicking

paradigm”–the full-feedback paradigm in Hertwig and Erev’s (2009) classification. Participants do not receive

a description of the payoff distributions but receive feedback about their choices. The feedback information

disclosed after each choice includes the outcome from the payoff distributions associated to each key.


Figure 2. Sample screenshots from Experiment 2 after the first choice in the three experimental conditions

(with, respectively, 0, 2, and 24 replicas for each alternative). The selected key is highlighted with a white

background.


Table 1. Re-analysis of behavioral data from experiments that examined choice behavior in the six conditions

using the clicking paradigm described in Figure 1 (Erev & Haruvy, 2014, and Di Guida et al., 2012)

Technion Data

Problem Action option Expected Value Action

rate

Inertia

rate

Best Reply

rate

Number of

observations

1 (10, .1; -1) +0.1 .27 (.24)

.87 (.12)

.69 (.20)

158

2 (-10, .1; 1) -0.1 .57 (.29)

.82 (.16)

.56 (.24)

158

3 (11, .5; -9) +1 .50 (.31)

.76 (.18)

.52 (.14)

50

4 (9, .5; -11) -1 .36 (.27)

.77 (.19)

.60 (.13)

50

5 1 with certainty +1 .95 (.11)

.97 (.06)

.95 (.11)

50

6 -1 with certainty -1 .03 (.11)

.99 (.02)

.98 (.11)

30

Summary statistics Value

(sd) Lower Upper

Number of

observations

Risk rate .43 (.19)

.40 .46 178

Underweighting score .33 (.27)

.25 .40 50

Payoff Variability score .38 (.19)

.33 .43 50

Inertia rate .82 (.14)

.80 .84 178

Best Reply rate .62 (.15)

.60 .65 178

Note. All six problems involve repeated choices (100 to 200 trials) between the status quo (a prospect that

yields zero with certainty) and the payoff distribution presented in the second column (the “Action” option).

The distribution (x, p; y) implies x with probability p, y otherwise. By problem statistics: The “Action rate” is

the proportion of choices of the “Action” option over all trials; the “Inertia” rate is the proportion of times in

which the choice at t was repeated at t + 1; and the “Best Reply” rate indicates the proportion of choices at time

t that best respond to the lottery outcome at t - 1, over all trials. Summary statistics: The “Risk rate” is the

Action rate averaged over problems 1-4; the underweighting score is defined as (A2 - .5) + (.5 - A1), where Aj

is the action rate in Problem j, j = 1, 2, 3, and 4; the payoff variability score is defined as (A5 - A3 + A4 - A6) /

2, or as A5 - A3 for those participants who did not play Problem 6; finally, the Inertia and Best Reply rates,

defined as in the “by problem” statistics, are averaged over problems 1-4. The standard deviations of the rates

are reported between parentheses. The Lower and Upper columns in the lower part of the table report the

extremes of the 95% normal confidence interval for each effect.


Table 2. Summary of Study 1 results

West Europe (Denmark) West Asia (Israel) East Asia (Taiwan)

Prob. Action option Action

rate

Inertia

rate

Best

Reply

rate

Action

rate

Inertia

rate

Best

Reply

rate

Action

rate

Inertia

rate

Best

Reply

rate

1 (10, .1; -1) .21

(.18)

.90

(.06)

.76

(.17)

.27 (.29)

.87 (.14)

.73 (.26)

.22

(.20)

.86

(.09)

.80

(.20)

2 (-10, .1; 1) .67 (.23)

.82 (.14)

.67 (.21)

.70 (.28)

.82 (.13)

.72 (.27)

.75 (.24)

.82 (.14)

.80 (.24)

3 (11, .5; -9) .46 (.21)

.65 (.15)

.70 (.18)

.47 (.25)

.67 (.19)

.71 (.21)

.51 (.16)

.59 (.15)

.78 (.20)

4 (9, .5; -11) .34 (.19)

.70 (.17)

.62 (.14)

.45 (.22)

.64 (.18)

.73 (.21)

.44 (.13)

.57 (.14)

.80 (.19)

5 1 with

certainty .91 (.21)

.95 (.16)

.91 (.21)

.89 (.26)

.95 (.13)

.89 (.26)

.89 (.23)

.94 (.14)

.89 (.23)

6 -1 with

certainty .15 (.30)

.97 (.07)

.86 (.30)

.12 (.23)

.95 (.16)

.88 (.23)

.09 (.20)

.97 (.08)

.91 (.20)


(sd) Lower Upper

Value

(sd) Lower Upper

Value

(sd) Lower Upper

Risk rate .42 (.13)

.38 .46 .47 (.18)

.41 .54 .48 (.12)

.45 .51


.34 .56 .43 (.39)

.29 .57 .53 (.35)

.43 .63

Payoff variability score .32 (.25)

.24 .41 .38 (.21)

.30 .45 .37 (.22)

.31 .43

Inertia rate .77* (.08)

.74 .80 .75 (.12)

.66 .79 .71* (.10)

.68 .74

Best Reply rate .68* (.12)

.64 .73 .72 (.18)

.71 .79 .79* (.18)

.74 .84

Note. All six problems involve repeated choices (201 trials with feedback) between the status quo (a prospect

that yields zero with certainty) and the payoff distribution presented in the second column (the “Action” option).







2; finally, the Inertia and Best Reply rates, defined as in the “by problem” statistics, are averaged over problems

1-4. The standard deviations of the rates are reported between parentheses. A star (*) implies a significant

effect (p < .05) of the group factor for the statistics reported on the lower part of the table. The Lower and

Upper columns in the lower part of the table report the extremes of the 95% normal confidence interval for each

effect.



West Asia (Israel) East Asia (Taiwan)

Prob. Basic alternatives

Number

of

replicas

R rate Inertia

rate

Best

Reply

rate

R rate Inertia

rate

Best

Reply

rate

7.1 R: (32, .09; 0)

S: A draw from {1, 2, 3, 4, 5} 0

.20 (.24)

.90 (.09)

.79 (.22)

.29 (.27)

.80 (.14)

.68 (.24)

8.1 R: (32, .1; 0)

S: A draw from {1, 2, 3, 4, 5} 0

.20 (.24)

.89 (.10)

.79 (.24)

.33 (21)

.69 (.19)

.68 (.20)

7.3 R: (32, .09; 0)

S: A draw from {1, 2, 3, 4, 5} 2

.24 (.19)

.42 (.29)

.52 (26)

.26 (.19)

.32 (.29)

.38 (.16)

8.3 R: (32, .1; 0)

S: A draw from {1, 2, 3, 4, 5} 2

.32 (.28)

.50 (.35)

.45 (.29)

.30 (.24)

.29 (.25)

.37 (.17)

7.25 R: (32, .09; 0)

S: A draw from {1, 2, 3, 4, 5} 24

.52 (.32)

.32 (.34)

.30 (.30)

.40 (.32)

.25 (.29)

.17 (.18)

8.25 R: (32, .1; 0)

S: A draw from {1, 2, 3, 4, 5} 24

.54 (.32)

.25 (.25)

.32 (.32)

.40 (.31)

.22 (.31)

.18 (.15)


(sd) Lower Upper

Value

(sd) Lower Upper

Risk rate .34 (.20)

.25 .42 .33 (.17)

.26 .40

Maximization rate .52 (.08)

.48 .55 .51 (.07)

.49 .54

Underweighting score .15 (23)

.05 .25 .16 (.18)

.08 .23

Inertia rate .55 . (.19)

.46 .63 .43 . (.19)

.35 .51

Best Reply rate .53* (.21)

.44 .62 .41* (.13)

.36 .47

Note. The risk, inertia, and best reply rates are computed over all six replica treatments. By problem statistics:

The “R rate” is the proportion of choices of the “R” option over all trials; the “Inertia” rate is the proportion of

times in which the choice at t was repeated at t + 1, over all trials; and the “Best Reply” rate indicates the

proportion of choices at time t that best respond to the lottery outcome at t - 1, over all trials. Summary

statistics: The “Risk rate” is the R rate averaged over problems 1-6; the “Maximization rate” is the rate of

expected payoff maximizing choices (i.e., S in problems 7.* and R in problems 8.*) in problems 1-6; the

underweighting score is defined as 0.5 - P(R in problems 8.1, 8.3, and 8.25); finally, the Inertia and Best Reply

rates, defined as in the “by problem” statistics, are averaged over problems 1-6. For the four statistics reported

on the lower part of the table, a star (*) indicates that the group factor is significant at the 5% confidence level,

whereas a dot (.) indicates significance at the 10% confidence level. The Lower and Upper columns in the

lower part of the table report the extremes of the 95% normal confidence interval for each effect.



West Asia (Israel) East Asia (Taiwan)

Prob. Action option Action

rate

Inertia

rate

Best

Reply rate

Action

rate

Inertia

rate

Best

Reply rate

1 (10, .1; -1) .27 (.31)

.91 (.08)

.68 (.25)

.42

(.31)

.83

(.16)

.56

(.25)

2 (-10, .1; 1) .45 (.34)

.87 (.15)

.48 (.28)

.51 (.26)

.78 (.19)

.52 (.21)

3 (11, .5; -9) .36 (.31)

.81 (.18)

.50 (.04)

.51 (.27)

.71 (.25)

.51 (.07)

4 (9, .5; -11) .32 (.33)

.80 (.25)

.52 (.06)

.47 (.27)

.71 (.22)

.52 (.07)

5 1 with certainty .90 (.20)

.96 (.04)

.90 (.20)

.85 (.17)

.89 (.13)

.85 (.17)

6 -1 with certainty .13 (.17)

.94 (.04)

.87 (.17)

.23 (.22)

.90 (.09)

.77 (.22)


(sd) Lower Upper

Value

(sd) Lower Upper

Risk rate .35* (.23)

.25 .45 .48* (.20)

.42 .53


.01 .36 .10 (.42)

-.02 .21

Payoff Variability score .36 (.27)

.24 .48 .29 (.18)

.24 .34

Inertia rate .85* (.12)

.80 .90 .76* (.15)

.72 .80

Best Reply rate .54 (.08)

.51 .58 .53 (.10)

.50 .56

Note. All six problems involve repeated choices (201 trials with feedback) between the status quo (a prospect

that yields zero with certainty) and the payoff distribution presented in the second column (the “Action” option).







2; finally, the Inertia and Best Reply rates, defined as in the “by problem” statistics, are averaged over problems

1-4. The standard deviations of the rates are reported between parentheses. A star (*) implies a significant

effect (p < .05) of the group factor for the statistics reported on the lower part of the table. The Lower and

Upper columns in the lower part of the table report the extremes of the 95% normal confidence interval for each

effect.


APPENDIX

Table A1. Comparisons of average Best Reply rates across the three cultural groups in Study 1

Comparison Difference lower upper Adjusted p-value

IL-DK 0.040 -0.059 0.139 .61

TW-DK 0.107 0.017 0.197 .01

TW-IL 0.067 -0.023 0.158 .19

Note. Values of the 95% family-wise confidence level intervals, computed with Tukey’s ‘Honest Significant

Difference’ method. DK = Denmark, IL = Israel, and TW = Taiwan.

Table A2. Comparisons of average Inertia rates across the three cultural groups in Study 1

Comparison difference lower upper Adjusted p-value

IL-DK -0.017 -0.076 0.042 .77

TW-DK -0.058 -0.112 -0.005 .03

TW-IL -0.041 -0.095 0.012 .16

Note. Values of the 95% family-wise confidence level intervals, computed with Tukey’s ‘Honest Significant

Difference’ method. DK = Denmark, IL = Israel, and TW = Taiwan.

Cross Cultural Differences in Decisions from Experience: Evidence from Denmark, Israel, and Taiwan

Documents