Generalizability of categorical status quo bias : ordinary everyday goods experiment

(1)

Faculty of Economics and Business (FEB) Thesis proposal Master Economics: Behaviour Economics &

Game Theory 15 ECTS

Generalizability of categorical status quo bias: ordinary

everyday goods experiment

By Jens van de Pol, 10579060 Supervisor: mw. Giorgia Romangnoli PhD

Second reader: prof. dr. Joep Sonnemans

Abstract

Maltz developed a reference dependent model which explains and predicts status quo bias. His model predicts categorical status quo bias. This means that when an alternative of the same category as the endowment is presented no status quo bias will occur and when the alternative is of a different category status quo bias will occur (2016). Maltz and Romagnoli tested this categorical bias on the lottery market. They did this with the categories of risky and ambiguous lotteries. Their results are in line with Maltz’s

(2)

Statement of Originality

This document is written by Jens van de Pol who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

Table of contents 1. Introduction 4 2. Related literature 5 3. Methodology 8 3.1 Experimental design 8 3.2 Experimental procedure 8 3.3 Analysis 9

3.3.1 Status Quo Bias Index 9

3.3.2 Probit regression 11

4. Results 11

4.1 Participants 12

4.2 Status Quo Bias Index 12

4.2.1 Descriptive Statistics 12

4.2.2 Status Quo Bias Index Regression 17

4.3 Probit Regression 19

4.3.1 Descriptive Statistics 19

4.3.2 Probit Regression 21

4.4 Limitations 22

5. Conclusion and discussion 22

6. Appendixes

Appendix A Instructions part 1 treatment A 25

Appendix B Instructions part 1 treatment B 28

Appendix C Questions part 1 treatment A 31

Appendix D Questions part 1 treatment B 36

Appendix E Instructions part 2 treatment A 39

Appendix F Instructions part 2 treatment B 45

Appendix G Questions part 2 treatment A 49

Appendix H Questions part 2 treatment B 54

(4)

1 Introduction

“How do individuals make decisions? This question is of crucial interest to researchers in economics, political science, psychology, sociology, history and law” (Samuelson and Zeckhauser, p. 1. 1988). In their paper Samuelson and Zeckhauser report findings that, based on decision-making experiments to test for status quo effect, decision makers exhibit a significant status quo bias. Status quo bias is the term that describes the tendency to stick to the current state of affairs (1988). After these findings different field and lab experiments were performed and the status quo bias has been observed in many different markets. One of these is performed on everyday ordinary goods by Knetch, in which he observed status quo bias (1989).

Maltz and Romagnoli say that there are many theoretical frameworks trying to model the status quo bias. They can be generalized in three different approaches. The first approach subscribes the bias to incompleteness of preferences and predicts that status quo bias can only be found in the presence of ambiguity (Bewley, 1986; Ortoleva, 2010; Mihm, 2016). The second approach subscribes status quo bias due to loss aversion (Kahneman and Tversky, 1979; Koszegi and Rabin, 2006). In the third approach the decision maker is viewed 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 includes a psychological constraint set from which the agent chooses the best alternative according to his utility” (p. 21) (Masatlioglu and Ok, 2014; Maltz, 2016) (2017). In his approach Maltz predicts no status quo bias between options that belong to the same category. However, when options belong to different categories it does predict status quo bias (2016). Maltz and Romagnoli did an experiment with risky and ambiguous lotteries, which are two different categories. In this experiment they found indeed that status quo bias only occurred in asymmetric choice options (2017), which evidence is in support of Maltz his model. This paper investigates if this categorical approach to status quo bias is more general and does so in the market for everyday ordinary goods. The research was done by performing an incentivized experiment where participants had to make the same series of choices in two different frames. The first was a neutral frame where they had to choose their preferred good out of the two options and the second is a status quo frame they were endowed with a fixed good and had to choose to keep the endowment or switch it for the alternative. The alternatives were sometimes of the same category as the endowment and sometimes of a different category.

(5)

The results of both analysis are the same and conclude that when the alternative good is of the same category as the endowment no status quo bias is exhibited and when the alternative is of a different category status quo bias is exhibited. This is in line with the expectations based on Maltz’s model.

The remainder of this paper is as follows. Section 2 discusses related literature and in section 3 the methodology of the hypothesis based on the methods is discussed. Section 4 provides the main results and some limitations and sections 5 states the conclusion and discussion.

2 Related literature

This part regarding the related literature begins with research about status quo bias in the everyday ordinary goods market. Next the model which is a possible explanation for status quo bias is discussed and lastly the paper which tests this model is discussed.

One of the best-known studies in the field of status quo bias on everyday ordinary goods is done by Knetch (1989). In this paper Knetch describes his experiment consisting of 3 groups. One group was endowed with a mug, another group was endowed with candy and the last group wasn’t endowed at all. In every group the subjects were asked to choose between the mug and the candy. However contrary to the expectations the results showed that subjects who were endowed, chose their endowment significantly more compared to subjects that weren’t endowed.

The paper further discusses an experiment about the subject’s willingness to accept (WTA) money to give up their endowment, which were tickets or a candy bar and their willingness to pay (WTP) for the same good. Subjects WTA was significantly higher than their WTP for the same good, which again indicates a strong aversion to give up their endowment.

The paper also discusses a survey response where respondents were called and asked about changing their chance of having an accident with 0.5 percent or changing their income with $700. Half of the respondents were asked if they would accept $700 to have a 0.5 percent chance increase of having an accident. And the other group was asked if they would reduce their chance of having an accident by 0.5 percent by giving up $700. The same was done in a

(6)

In his paper Knetch used two different real exchange experiments with student participants and a survey with hypothetical questions with non-student participants. The results were all the same. Therefore, he provided proof of the existence of status quo bias.

This paper discusses a model which possibly explains the status quo bias which Knetch has proven to exist. It does so by letting subjects make a series of choices between everyday ordinary goods, which are sometimes within the same category of goods and sometimes in different categories, while in Knetch’s experiments and surveys subjects could only choose once between different categories.

As mentioned in the introduction Maltz and Romagnoli identify three different approaches to model the status quo bias. The model from Maltz tested in this paper belongs to the category that views the decision maker as a constrained maximizer (2016).

In his paper Maltz developed a reference dependent model with an initial endowment, where alternatives are divided into categories.

The model creates a link between agent’s exogenous endowment and her endogenous reference point. The actual reference point is the best feasible alternative, according to the agent’s preferences, which belongs to her endowment’s category. This endogenous reference point induces a constraint set from which the final choice is made according to utility maximization. (p. 1)

According to the model a decision maker follows the following procedure: First, identify the best alternative which is of the same category as the endowment. That alternative then serves as the new reference point. Next, this reference point causes a constraint set from which the alternative that maximizes the utility is chosen.

Further the paper shows that if the decision maker follows the procedure above the decision is consistent with the Weak Axiom of Revealed Preferences (WARP), the Weak Status Quo Bias (WSQB), the Categorical Referential Equivalence (CRE) and the Categorical Status Quo Irrelevance (CSQI). WARP is a consistency requirement imposed on the decision maker’s behavior, which states for any given endowment utility maximization takes place. WSQB states that an endowment can only increase the possibility of being chosen over an alternative. CRE states that when a decision maker is looking for a new good it would not have an effect on the final choice whether this person is endowed with an older or newer model of this specific good.

(7)

means this model predicts rational behavior when the alternatives are of the same category (2016).

Maltz also discusses the difference between some related works, of which Masatlioglu and Ok (2014) is the most related. In this paper the differences, with respect to the assumptions of the model, the predictions and the procedure the decision maker follows are discussed (2016). To summarize, the model predicts no status quo bias between options that belong to the same category, however when options belong to different categories it does predict status quo bias.

As mentioned before this paper discusses if this model’s predictions can be found in the everyday ordinary goods market and to test the generalizability of the model. The most related paper to this one is that of Maltz and Romagnoli (2017), because in their experiment Maltz and Romagnoli researched the effect of dissimilarity on status quo bias. In their paper Maltz and Romagnoli explore what role different types of uncertainty (risk and ambiguity) play on status quo bias. They do this by performing an incentivized experiment consisting of four treatments and every treatment itself consists out of two parts.

The four treatments consist of all four possible combinations of endowment and alternative between the categories risk and ambiguity lotteries. These are risk-ambiguity (R-A), risk-risk (R-R), ambiguity-ambiguity (A-A) and ambiguity-risk (A-R). Except for the difference in choice options the treatments are the same.

The first part is performed in a neutral frame, where subjects had to make a series of choices between the two options which are dependent of the treatment the subject is in. The second part is performed in a status quo frame, where subjects had a fixed endowment and subjects had to make a series of choices between keeping the endowment and switching to the alternative.

They found no status quo bias when the endowment and alternative were of the same category but did find a bias when the endowment and alternative were of different categories. They state that these findings are not consistent with existing models on loss aversion or incomplete preferences and suggest Maltz’s model, for the reasons mentioned earlier, as an alternative explanation. In their paper Maltz and Romagnoli use two different kinds of analyze methods and both of them give rise to the same result (2017).

(8)

3 Methodology

This part consists of three subparts. First the experimental design is described, then the experimental procedure is stated and finally the analysis which explains the analysis methods and hypothesis is discussed.

3.1 Experimental design

The experiment consisted of two treatments. The difference between the treatments is the series of options and the category of the good the subject was endowed with. In each treatment a within-subject design is adopted which consists of three parts. In the first part subjects were asked to make a series of choices in a neutral frame between a fixed good and different alternative. This means that in this neutral frame, subjects were presented two goods and asked to choose their preferred good. Also, in part one, five intermission questions were added where the two goods did not belong to either category. The purpose of these question was to make sure that subjects didn’t create a sense of attachment to the goods. Apart from that these questions didn’t serve any other purpose and were not included in the analysis.

In the second part, the status quo frame, subjects were endowed with the fixed good of part one and had to make the decision to keep the endowment or switch to the alternative. Status quo bias is observed when subjects choose the endowed option more often in part two, where they own the good, than in part one. The categories used in this experiment are pens and notebooks. These categories have been chosen because they approximately have the same value for the average participant. Each category has five different goods which means both treatments have a total of 23 preference questions. All goods are approximately of the same financial value so subjects preferences could not be influenced by the financial value of the goods.

Part three is a small survey with demographic questions. The experiment was pencil-paper based and involved 60 subjects. At the end all subjects received one of the goods they chose. This choice is randomly selected by throwing a 10 sided-dice.

3.2 Experimental procedure

In this paragraph the experimental procedure of the experiment is outlined. Except for the difference in the series of options the subject had and the category of the endowment within the procedure, the two treatments are the same.

(9)

9 are between a fixed good and 4 alternatives of the same category and 5 of the alternatives of the different category (Appendix C and D). The other 5 questions in the envelope are the 5 intermission questions. The fixed good in the pen treatment is a black pen and in the notebook treatment a checkered size A5 notebook.

In part 2 the subjects receive the instructions of Part 2 (Appendix E and F), the endowment and an envelope with written “Part 2” on it. Depending on the treatment this endowment is the black pen or the A5 checkered notebook. In the envelope are 9 questions (Appendix G and H), in which the subjects are asked if they want to keep their endowment or want to switch to the alternative good shown on the question. Again, of these 9 questions 4 are with an alternative of the same category and 5 with a different category. So, the subjects make the same series of choices twice.

In part 3 the subjects are provided with a small survey with 4 questions (Appendix I). After the survey a 10 sided-dice is thrown to randomly select the subject’s payment. First the part is selected; when an uneven number is thrown part 1 is selected and part 2 is selected when an even number is thrown. Then a question from the 9 possible options is selected from the selected part by also throwing a 10 sided-dice, where the numbers 1 until 9 represent the questions and when 0 is thrown the subject has to throw again. Then the subjects receive the good they had chosen on the selected question.

3.3 Analysis

This part discusses the two different types of analysis performed that provide an answer to the research question, the Status Quo Bias Index analyses and the Probit Regression analyses. First the Status Quo Bias Index analysis and the formulation of the hypothesis based on the analysis are discussed. Secondly the Probit Regression analysis and the corresponding hypothesis are discussed.

3.3.1 Status Quo Bias Index

Like Maltz and Romagnoli (2017), the within subject design of this experiment allows a status quo bias index for each participant by comparing their choices under the two frames. This status quo bias index consists of two biases, the SQS, which is the status quo bias index for a

(10)

symmetrical and asymmetrical frames. When the index is negative the participant exhibits a negative status quo bias. When the index is positive, a positive status quo bias is exhibited. And when the index is 0, participants made the exact same choice in both parts. Now that the SQS and SQD are explained, the hypothesis can be constructed. The expectations are based on Maltz’s model which is explained in the related literature. This model predicts status quo bias in the asymmetrical choice frames and no bias in the symmetrical choice frames. Which gives rise to the following two hypotheses:

𝐻": 𝑆𝑄𝑆 = 0 𝑎𝑛𝑑 𝑆𝑄𝐷 > 0

𝐻_.: 𝑆𝑄𝑆 ≠ 0 𝑎𝑛𝑑, 𝑜𝑟 𝑆𝑄𝐷 ≤ 0

when SQS is not significantly different from 0 it shows that the symmetrical index is equal to 0 and that participants made the same choices in both parts of the experiment. When SQD is significantly greater then 0 it shows that the participants chose the fixed good more often in part 2 then in part 1 in the asymmetrical frame and thus exhibit status quo bias. When both are true, the null hypothesis cannot be rejected and therefore not reject that there is no categorical status quo bias in the everyday ordinary goods.

Additional research is done to test if the asymmetrical choice frame has an effect on the status quo bias. The method is provided and discussed here. Per participants two dependent observations are collected SQS and the SQD, which together form the status quo bias index. To test if an asymmetrical choice frame gives rise to the status quo bias, a linear regression with robust standard error clustering on ID is performed which has the following form:

𝑆𝑄_4,5 = 𝛽_"+ 𝛽_.𝐴𝑆𝑌𝑀_4,5+ 𝛽_;𝐴𝑆𝑌𝑀_4,5∙ 𝐶𝐴𝑇₄ + 𝛿𝑉₄ + 𝜀_4,5

where 𝑆𝑄_4,5 is the status quo bias index of person i and f is S or D which corresponds to the SQS or SQD index of which the status quo index is constructed. 𝛽_" is a constant. 𝐴𝑆𝑌𝑀_4,5 is a dummy variable of the asymmetrical frame for person i and takes the value 1 when f is D, which is actually the mean regressor in this regression. 𝐶𝐴𝑇4 is a dummy variable that indicates which

treatment person i made and takes the value 1 if it is treatment A, therefore this is also an indicator for which category of good the participant is endowed with. So, with this dummy variable we can see if the category of the fixed good has an influence on the status quo bias index. 𝑉4 is a vector of control variables of person i and 𝜀4,5 is the error term.

To test if the asymmetrical choice frame has an effect on the status quo bias the following hypothesis is made:

(11)

when 𝛽_; is significantly greater than 0 the null hypothesis cannot be rejected and therefore not reject that an asymmetrical choice frame has a significant effect on the rise of a status quo bias.

3.3.2 Probit regression

With a probit regression it is possible to test if the symmetrical questions in a status quo frame has no effect on the likelihood that the status quo option is chosen, and asymmetrical questions in status quo frame have an effect on the likelihood. This gives rise to the following probit regression:

𝑆𝑄𝑂_4,C = 𝛽_"+ 𝛽_.𝑆𝑌𝑀_C∙ 𝑆𝑄_C+ 𝛽_;𝐴𝑆𝑌𝑀_C∙ 𝑆𝑄_C+ 𝛿𝑉_4,C+ 𝜀_4,C

where 𝜙(𝑆𝑄𝑂_4,C) is the cumulative distribution function of the standard normal distribution and 𝑆𝑄𝑂4,C is the dependent dummy variable that takes the value 1, if participant i in question q

choose the status quo option. 𝑆𝑌𝑀 ∙ 𝑆𝑄 is an interaction term between the dummy variables SYM which takes the value 1, when question q is a symmetrical choice frame. And SQ which takes the value 1 when question q belongs to the status quo frame of the experiment. 𝐴𝑆𝑌𝑀 ∙ 𝑆𝑄 is an interaction term between the dummy variables ASYM which takes the value 1 when question q is an asymmetrical choice frame and SQ which is already explained. Further is 𝑉₄ a vector of control variables of person i in question q and 𝜀_4,C is the error term.

To test whether the symmetrical questions in the status quo frame have no effect on the likelihood that the status quo option is chosen and the asymmetrical questions in the status quo frame do have an effect, the following hypothesis is constructed:

𝐻": 𝛽. = 0 𝑎𝑛𝑑 𝛽; > 0

𝐻_.: 𝛽_. ≠ 0 𝑎𝑛𝑑, 𝑜𝑟 𝛽_; ≤ 0

when 𝛽_.is equal to 0 and 𝛽_; is significantly greater then 0 it is not possible to reject the null hypothesis, which states that symmetrical questions in the status quo frame have no significant effect on the likelihood that the status quo option will be chosen, and asymmetrical questions have an effect.

4 Results

(12)

4.1 Participants

In this part demographic statistics of the participants are provided, as illustrated in table 1 and 2. First, the gender of the participants, 36 of them are male and 24 female participants, the male are equal to 60 percent of the participants. Furthermore, is the average age 24 years old. In the experiment 52 participants which is equal to 86.67 percent are Dutch. Other participants come from Argentina, Belarus, China, France, Great Britain, India and Russia (see Table 2). Finally, 33 participants, 55 percent, had no strong preference (NSP) for a category of goods between notebook and pen.

Table 1 Descriptive demographic statistics participants

Variable Mean Standard Deviation Minimum Maximum

MALE 36 0.4940 0 1 AGE 24.3667 5.0387 18 58 DUTCH 52 0.3428 0 1 NSP 33 0.5017 0 1 N = 60 Table 2 Nationalities

Countries Number of participants Argentina 1 Belarus 1 China 1 France 2 Great Britain 1 India 1 Netherlands 52 Russia 1

4.2 Status Quo Bias Index

This part discusses is the descriptive statistics discussed and the performed regression analyses.

(13)

3. First is the variable Fixed Good Chosen in Part 1 with the same categorical alternatives (FGC1S), which has a mean of 2.3333 and a standard deviation of 1.3105. This means it is significantly different from 0 and this indicates that the fixed good is on average chosen more than half of the time. Second is the variable Fixed Good Chosen in Part 1 with different categorical alternatives (FGC1D), which has a mean of 2.3667 and a standard deviation of 1.7942. This means it is also significantly different from 0 and this indicates that the fixed good is on average chosen more than half of the time. The participants were given a series of 9 questions per part, 4 of those were with the same categorical alternatives and 5 were with different alternatives. Therefore, it is not surprising that FGD1S has a higher mean. Third is the variable Fixed Good Chosen in Part 2 with the same categorical alternatives (FGC2S), which has a mean of 2.3167 and a standard deviation of 1.2281. Again, this is significantly different from 0 and the fixed good is chosen more than half of the time. Finally, is the variable Fixed Good Chosen in Part 2 with different categorical alternatives (FGC2D), which has a mean of 3.15 and a standard deviation of 2.0238. Once again this is significantly different from 0 and chosen more than half of the time.

The variables discussed are imported to this research, because when FGC1S is subtracted from FGC2S and FGC1D from FGC2D the variables give rise to the two variables SQS and SQD, which will be used as dependent variables in the Status Quo Bias Index analyses part. Because there were 27 participants with a strong categorical preference the statistics of the status quo bias index are provided two times, one time with all participants (table 4) and one time without the participants with strong preference (table 5).

The status quo bias index with all participants, the SQS has a mean of -0.0167 and a standard deviation of 1.0655. Second is the SQD, which has a mean of 0.7833 and a standard deviation of 1.3789. Furthermore, the status quo bias index with only the participants who had no strong preference. The SQS variable has a mean of 0.0909 and a standard deviation of 0. The SQD variable has a mean of 1.303 and a standard deviation of 1.5709.

Table 3 Descriptive statistics fixed good chosen

(14)

Table 6 and figure 1 showc that with all 60 participants the mean of the SQS is not significantly different from 0 and the mean of the SQD is significantly different from 0 with a P-value smaller than 0.000. When taking an additional look at figure 2 and 3, it shows that the distribution of the symmetrical frame (SQS) is centered around 0, while it shifted to the right in the asymmetrical frame (SQD). This is all in line with the null hypothesis.

Next, when looking at figure 4 and table 7, this shows the statistical results of the 33 participants with no strong preference. Here, agian the results show the mean of the SQS is not significantly different from 0 and the mean of the SQD is significantly different from 0 with a P-value smaller than 0.000. When taking a closer look at the distribution of the index, figure 5 and 6, this shows again that the distribution of the symmetrical frame is centered around 0, while the asymmetrical frame shifted to the right. Only now there is no negative index observed in the asymmetrical frame. Therefore, the results are in line with the null hypothesis and cannot reject the presence of categorical status quo bias.

Table 4 Descriptive statistics status quo bias index

SQS -0.0167 1.0655 -3 3

SQD 0.7833 1.3789 -2 5

N=60

Table 5 Descriptive statistics status quo bias index

SQS 0.0909 0.8427 -2 2

SQD 1.3030 1.5709 0 5

(15)

Mean values of the SQS and SQD. -0,1000 0,0000 0,1000 0,2000 0,3000 0,4000 0,5000 0,6000 0,7000 0,8000 0,9000 SQS SQD st at us qo u bi as Figure 1

Status Quo Bias Index all participants Table 6 T-test statistics status quo bias index

Variable Mean Standard Error T Value P Value 95% Conf. Interval

SQS -0.0167 0.1376 -0.1212 0.9040 -0.2919 0.2586 SQD 0.7833 0.1780 4.4003 0.0000 0.4271 1.1395 N=60 0 .2 .4 .6 D ist ri b u ti o n -4 -2 0 2 4 SQS

Figure 2 Distribution SQS Index all participants

0 .1 .2 .3 .4 .5 D ist ri b u ti o n -2 0 2 4 SQD

(16)

Mean values SQS and SQD 0 0,2 0,4 0,6 0,8 1 1,2 1,4 SQS SQD st at us qo u bi as

Figure 4 Status Quo Bias Index NSP participants Table 7 T-test statistics status quo bias index

Variable Mean Standard Error T Value P Value 95% Conf. Interval

SQS 0.0909 0.1467 0.6197 0.5399 -0.2079 0.3897 SQD 1.3030 0.2735 4.4003 0.0000 0.7460 1.8601 N=33 0 .2 .4 .6 .8 D ist ri b u ti o n -2 -1 0 1 2 SQS

Figure 5 Distribution SQS Index NSP participants

0 .1 .2 .3 .4 D ist ri b u ti o n 0 1 2 3 4 5 SQD

(17)

4.2.2 Status Quo Bias Regression

The descriptive statistics in the previous part show that it is not possible to reject the categorical status quo bias. Additional analyses are performed to see if asymmetrical choice frames give rise to the status quo bias. The analyses performed here is, as mentioned earlier, a linear regression with robust standard errors clustering on ID. The results of the performed analyses are stated in table 8. As earlier indicated the variable under interest is ASYM.

Regression 1 shows a significant positive effect at a level of 1 percent for the variable ASYM, which indicates that the asymmetrical frame indeed plays a role in the rise of a status quo bias. However, when controlling for the participants with no strong preference (NSP) (regression 2) in asymmetrical frame, it shows that the significant positive effect of ASYM disappears. Regression 2 also shows that the variable ASYM*NSP has a 1% significant effect, which as expected, indicates that the subjects with no strong preference (NSP) in an asymmetrical choice frame, are playing a crucial role. This is as expected, because participants with strong preferences for a category should not exhibit categorical status quo bias. This is the reason why additional regressions have been performed with only participants that showed no strong preference.

The conclusion of this analysis is based on regression 5. This regression shows a 10 percent positive significant effect of 0.42, with respect to variable ASYM. This indicates that the asymmetric choice frame in part 2 has a significant positive effect on status quo bias. Furthermore, interesting is the positive significant effect at a 1 percent level of 1.7943 in relation to the interaction term ASYM and CAT, whereas mentioned earlier CAT is a dummy variable for treatment A. This is an indication that being endowed with the notebook (in this research) raises the likelihood for a status quo bias to happen. Further, these results do not reject the null hypothesis that 𝛽_; > 0.

(18)

Table 8

Status Quo Bias Index regression

Status Quo Bias Index (SQS and SQD)

Variable (1) (2) (3) (4) (5) MALE 0.2846 0.3060 0.015 0.2309 0.2486 (0.2427) (0.230) (0.3454) (0.3204) (0.2914) AGE -0.0047 0.0020 -0.0059 -0.0084 -0.006 (0.0193) (0.0159) (0.0818) (0.0740) (0.068) DUTCH 0.0204 0.1340 0.2345 0.3399 0.3493 (0.2514) (0.2686) (0.3595) (0.3863) (0.4023) ASYM 0.7837*** 0.1962 1.1684*** 1.1831*** 0.42* (0.2260) (0.1962) (0.3126) (0.3028) (0.2353) ASYM*NSP 1.1457*** (0.3188) CAT 0.8216** -0.0906 (0.3353) (0.3319) ASYM*CAT 1.7943*** (0.6384) Constant 0.0887 -0.3621 0.0351 -0,4737 -0,1470 (0.5530) (0.5116) (2.0263) (1.8742) (1.6892) Regression

type linear linear linear linear linear

Observations 120 120 66 66 66

R-squared 0.1055 0.2032 0.1833 0.2633 0.3661

Notes: Each linear regression accounts for robust standard error cluster on ID. Each regression has at least 3 control variables.

Regressions 3, 4 and 5 are without participants with a strong categorical preference.

(19)

4.3 Probit

This part discusses the descriptive statistics discussed and the performed regression analyses.

4.3.1 Descriptive Statistics

In this part the outcome statistics of the experiment in relation to the probit regression analyses are provided. Like in the Status Quo Bias Index paragraph, also here are the descriptive statistics provided two times due to the 27 participants that showed a strong preference for a category.

In table 9 are the statistics of all the participants and in table 10 of the participants that showed no strong preference (NSP). Because a dependent dummy variable is used in the probit analyses with, value 1 if the participant chose the status quo option and each participant was provided with 18 preference questions, there are 18 ∙ 60 = 1080 observations with all participants and 18 ∙ 33 = 594 observation with the NSP participants.

First the descriptive statistics of the status quo option chosen for all participants are discussed. The variable Status Quo Option is chosen (SQO), which has a mean of 0.56667 and indicates that 612 times the status quo option has been chosen and the standard deviation is 0.4958. Because the SQO consists of the status quo options chosen in part 1 (SQO1) and in part 2 (SQO2). The variable SQO1 has a mean of 0.263, so the status quo option is 284 times been chosen and the standard deviation is 0.4404. And the variable SQO2 has a mean of 0.3037, so 328 times is the status quo option chosen and the standard deviation is 0.4601.

Now the status quo options chosen for the NSP participants are discussed. First the SQO, which has a mean of 0.5253, so 312 times is the status quo option chosen and the standard deviation is 0.4998. Second the SQO1, which has a mean of 0.2256, and 134 times the status quo option is chosen and the standard deviation is 0.4183. Third the SQO2, which has a mean of 0.2997, thus 178 times the status quo option is chosen and the standard deviation is 0.4585. When looking at table 11 and 12 it shows that in both the cases with all participants as well as with only the NSP participants, the mean of SQO1 is significantly smaller than the mean of SQO2. This is an indication of a on average status quo bias in both cases.

(20)

Table 9 Descriptive statistics status quo option chosen all participants Variable Amount Standard Deviation Minimum Maximum

SQO 612 0.4958 0 1

SQO1 284 0.4404 0 1

SQO2 328 0.4601 0 1

N=1080

Table 10 Descriptive statistics status quo option chosen NSP participants

Variable Amount Standard Deviation Minimum Maximum

SQO 312 0.4998 0 1

SQO1 134 0.4183 0 1

SQO2 178 0.4585 0 1

N=594

Table 11 Descriptive statistics status quo option chosen all participants Variable Mean Standard Error 95% Conf. Interval

SQO 0.56667 0.0151 0.5371 0.5963

SQO1 0.2630 0.0134 0.2367 0.2893

SQO2 0.3037 0.014 0.2762 0.3312

Table 12 Descriptive statistics status quo option chosen NSP participants Variable Mean Standard Error 95% Conf. Interval

SQO 0.5253 0.0205 0.4850 0.5655

SQO1 0.2256 0.0172 0.1919 0.2593

SQO2 0.2997 0.01881 0.2627 0.3366

N=594 N=1080

(21)

4.3.2 Probit Regression

The results of the regression analysis performed can be found in table 13. As earlier indicated the variables SYM*SQ and ASYM*SQ are under interest in this part.

In regression 1, without control variables and regression 2, with control variables it is observed that both variables have no significant effect. For the SYM*SQ this is according to the predictions, however it is against the predictions for the ASYM*SQ variable. Because these regressions also include the 27 participants that have a strong categorical preference, the same regressions were performed again, but now controlling for these 27 participants by dropping them from the data, these regressions are 3 and 4.

In regression 3, without control variables and regression 4, with control variables, it is observed there is no effect for the SYM*QB variable and a 5 percent significant effect for the ASYM*QB variable, which is in line with the predictions. According to regression 4, the ASYM*QB frame, raises the likelihood with 38.86 percent for the status quo bias to happen.

These results are in line with the null hypothesis, so also here it is not possible to reject there is no categorical status quo bias.

Table 13 Probit Regression SQO Variable (1) (2) (3) (4) SYM*SQ 0.0253 0.0341 0.2388 0.2386 (0.1413) (0.1395) (0.1586) (0.1540) ASYM*SQ 0.1464 0.1538 0.3871** 0.3886** (0.1248) (0.1231) (0.1662) (0.1597) CAT*SQ 0.2363 0.2242 0.129 0.1396 (0.1753) (0.1708) (0.2513) (0.2475) MALE 0.0871 0.0503 (0.1637) (0.2069) AGE 0.0253*** -0.0603 (0.0094) (0.0514) DUTCH -0.0756 -0.0648 (0.3618) Constant 0.065 0.6938 -0.1227 1.3317 (0.0854) (0.2666) (0.1105) (1.0300) Regression type probit probit probit probit

(22)

4.4 Limitations

In this paragraph the most important limitations of the experiment itself and the analyses are provided. First the limitations of the experiment are discussed and second the limitations of the analysis.

In this experiment the choice is made between the categories notebooks and pens. In this case out of the 60 participants only 30 exhibit a status quo bias, from which 27 participants have a strong preference for a category and of these 27 participants 21 strongly preferred the notebook and 6 the pen. This is an indication that these two goods are not the most optimal choice for this experiment. Due to this choice, some observations cannot be used and are dropped, which lowers the internal validity of the experiment.

Next to that this experiment has two treatments; the A treatment and B treatment. In the treatments the symmetrical and asymmetrical frames are presented. This could cause the high number of participants with a strong preference, because the participants are given an option to choose between both categories and this gives them the opportunity to create a strong preference for one of them.

The participants of the experiment are mainly students with the average age of around 24. This lowers the external validity of the results, because this group of participants is not a representation of the general population. Because the experiment is mainly performed on the UvA and the participants are mainly students with a high education, this also lowers the external validity and could potentially cause an omitted variable bias in the probit analyses as well as in the additional analysis regarding the status quo bias index part, with a variable educational level.

5 Conclusion and discussion

This research answers the research question; “Can categorical status quo bias be observed in everyday ordinary goods.” This is answered by performing an incentivized experiment with a within-subject design. The data from this experiment is analyzed in two different methods, the Status Quo Bias Index and the probit regression.

The status quo bias index consists out of two biases; the SQS and SQD. When the index is negative the participants exhibit a negative status quo bias, when it is positive it is a positive status quo bias and when it is a 0, the participants made the exact same choices in both parts.

(23)

rejected. Besides that, additional research has been done in this method. A linear regression tested the asymmetrical choice frame which has a positive effect on the status quo bias. The effect is significantly different from 0 on a 10 percent significance level, so the null hypothesis could also not be rejected. What caught the attention is the coefficient from the interaction term ASYM and CAT, which indicates that being endowed with the notebook raises the likelihood for a status quo bias to happen. However, it is unclear why this happens, and it is not possible to give an explanation for this effect. Therefore, further research to investigate this effect is recommended.

The probit regression tested if the symmetrical questions in a status quo frame has no effect on the likelihood that the status quo option is chosen and the asymmetrical questions in the status quo frame has an effect on the likelihood. This has been tested by performing a probit regression. The regressor for the symmetrical status quo frame did not significantly differ from 0, but the regressor for the asymmetrical status quo frame did on a 5 percent significance level. These results are in line with the expectations and therefore the null hypothesis could not be rejected.

The categories of goods used in this experiment to create the symmetrical and asymmetrical frames are notebooks and pens. These two good are, by assumption, representative for the everyday ordinary goods market. It could be the case that with other categories or other combinations of categories the results would differ.

The results from both analyses are in line with the expectations based on Maltz’s model. Which means that it is not possible to reject the hypothesis that categorical status quo bias can be observed in everyday ordinary goods. Next to that, the research does not allow to reject the hypothesis, that an asymmetrical status quo frame increases the probability for a status quo bias from happening.

The generalizability of the categorical status quo bias has important implications for decision makers and thus also for policy makers. The reason for these implications are that decisions makers can be nudged into choosing the status quo option even if this does not maximize their utility. Based on the results from this research and Maltz & Romagnoli’s research (2017) on categorical status quo bias, the conclusion could be made that decision makers can be nudged into remaining the status quo option or being influenced in the final

(24)

This research has some limitations which create suggestions for future research. First of all is about the internal validity of the experiment. Notebooks and pens are not the most optimal choice of categories for this kind of research, especially in combination with the two treatment designs. Because this could have caused the high number of participants with a strong preference. In order to raise the internal validity, the recommendation is made to work with other categories and with four treatments, of which the categories are divided in symmetrical and asymmetrical treatments.

The second remark regards the external validity. The participants are mostly students from the UvA around the age of 24, so this is not a representative group for the general population. Therefore, the recommendation is made to have a more diverse group of participants.

Third and last remark is about the omitted variable bias. The group of participants mostly consisted out of UvA students, which rises the potential omitted important regressor, which is the educational level. The recommendation here is to also have a more diverse group of participants and to include an educational level in the survey so this can be controlled.

(25)

Appendix A Instruction part 1 treatment A Dear participant,

Welcome to this experiment. Today you will earn a good which is depended on your choices and chance. If you have any questions about the instructions, please raise your hand and an experimenter will come to your desk to answer your question.

First some general instructions:

• Please don’t use your phone during the experiment.

• The experiment consists of two parts and a small survey. On your desk you have 2 envelops and the instruction sheet.

• You will receive a good according to your answers in part 1 or part 2. There is a 50% chance you receive a good chosen from part 1 and a 50% chance to receive a good chosen from part 2. At the end of the experiment the experimenter will throw a 10-sided dice, if an uneven number is thrown will receive a chosen good from part 1. If even is thrown, you will receive a chosen good from part 2.

• When the part has been selected the 10-sided dice will decide which good you will receive. In each part 9 questions can be selected for payment. So, when 0 is thrown, the dice will have to be thrown again.

• Which good you will receive will be determined and provided at the end of the experiment.

• There are no right or wrong answers in this experiment. This study is interested in your preferences.

• Some notebooks don’t have the same color as the picture, but besides that they are identical to the picture and description.

• The questions that are labelled with a number and B in Part 1 are hypothetical questions, so will not be used for payment.

(26)

Part 1

The experiment

Please take the envelope labeled ‘Part 1’. Inside are 14 questions. Each question has a picture of two goods. For each pair, you will be asked to choose the good that you prefer among the two. There is no order in which you should answer, you are free to get any question.

EXAMPLE

CHECKERED notebook size A5 LINED notebook size A5

Marking the box below the good indicates that you prefer and choose that good over the other. For each question mark your answer with an X in the box below your preferred choice. Please make sure to mark only one of the two boxes and please make a choice – do not leave a question blank. If you want to change your choice you can use the eraser, but your preferred good should be clearly marked. After you marked your choice, place the question back in the envelope labeled ‘Part 1’and move on the answer the next question. Once you placed the question back in the envelope, please do not take it out again.

Payment for Part 1

Recall that with 50% probability you will receive a good according to your answers in this part of the experiment. If part 1 is randomly drawn the following procedure will take place at the end of the experiment:

(27)

Notice that you will receive no good if there is no mark or more than one mark observed on the question that was selected.

Final example payment

John is a participant in the experiment. He answers the 14 questions one by one and places them back in the envelope. At the end of the experiment, if Part 1 is randomly drawn, he throws the 10-sided dice and suppose it stops on the number 8. John will be instructed to take question 8 from the envelope. Suppose it is marked as follows:

CHECKERED notebook size A5 LINED notebook size A5

8

Given his choice, John will receive the good on the right side, the lined notebook. Please take a few minutes to read the instructions on your own and feel free to ask questions.

(28)

Appendix B Instruction part 1 treatment B Dear participant,

Welcome to this experiment. Today you will earn a good which is depended on your choices and chance. If you have any questions about the instructions, please raise your hand and an experimenter will come to your desk to answer your question.

First some general instructions:

• Please don’t use your phone during the experiment.

• The experiment consists of two parts and a small survey. On your desk you have 2 envelops and the instruction sheet.

• You will receive a good according to your answers in part 1 or part 2. There is a 50% chance you receive a good chosen from part 1 and a 50% chance to receive a good chosen from part 2. At the end of the experiment the experimenter will throw a 10-sided dice, if an uneven number is thrown will receive a chosen good from part 1. If even is thrown, you will receive a chosen good from part 2.

• When the part has been selected the 10-sided dice will decide which good you will receive. In each part 9 questions can be selected for payment. So, when 0 is thrown, the dice will have to be thrown again.

• Which good you will receive will be determined and provided at the end of the experiment.

• There are no right or wrong answers in this experiment. This study is interested in your preferences.

• Some notebooks don’t have the same color as the picture, but besides that they are identical to the picture and description.

• The questions that are labelled with a number and B in Part 1 are hypothetical questions, so will not be used for payment.

(29)

Part 1

The experiment

Please take the envelope labeled ‘Part 1’. Inside are 14 questions. Each question has a picture of two goods. For each pair, you will be asked to choose the good that you prefer among the two. There is no order in which you should answer, you are free to get any question.

EXAMPLE

RED BIC pen BLUE BIC pen

Marking the box below the good indicates that you prefer and choose that good over the other. For each question mark your answer with an X in the box below your preferred choice. Please make sure to mark only one of the two boxes and please make a choice – do not leave a question blank. If you want to change your choice you can use the eraser, but your preferred good should be clearly marked. After you marked your choice, place the question back in the envelope labeled ‘Part 1’and move on the answer the next question. Once you placed the question back in the envelope, please do not take it out again.

You will throw a 10-sided dice to decide which question from Part 1 will be selected and receive the good chosen in that question. Of the 14 questions there are already 9 selected and there are

(30)

John is a participant in the experiment. He answers the 14 questions one by one and places them back in the envelope. At the end of the experiment, if Part 1 is randomly drawn, he throws the 10-sided dice and suppose it stops on the number 8. John will be instructed to take question 8 from the envelope. Suppose it is marked as follows:

RED BIC pen BLUE BIC pen

8 Given his choice, John will receive the good left

Please take a few minutes to read the instructions on your own and feel free to ask questions.

(31)

Appendix C Questions part 1 treatment A

BLANC notebook size A5 CHECKERED notebook size A5

1A

LINED notebook size A5 CHECKERED notebook size A5

(32)

CHECKERED notebook size A6 CHECKERED notebook size A5

3A

LINED notebook size A5 CHECKERED notebook size A5

(33)

BLUE BIC pen CHECKERED notebook size A5

5A

(34)

RED BIC pen CHECKERED notebook size A5

7A

(35)

GREEN BIC pen CHECKERED notebook size A5

(36)

Appendix D Questions part 1 treatment B

BLUE BIC pen BLACK BIC pen

1A

RED BIC pen BLACK BIC pen

2A

(37)

BLUE BIC pen BLACK BIC pen

3A

GREEN BIC pen BLACK BIC pen

(38)

CHECKERED notebook size A5 BLACK BIC pen

5A

BLANC notebook size A5 BLACK BIC pen

6A

(39)

LINED notebook size A5 BLACK BIC pen 7A

(40)

LINED notebook size A5 BLACK BIC pen

9A

(41)

Appendix E Instructions part 2 treatment A

Part 2

The good you just received

CHECKERED notebook size A5

This is a picture of the notebook. For participating in this experiment, you have earned this good and it is now yours.

The experiment

As of now, you earned and own the one on your desk.

Please take the envelope labeled ‘Part 2’. Inside are 9 questions. The question shows a picture of a good. You are then asked whether:

• You would like to keep your own good (the one shown on your card/table), or • Give up your good and switch to the good shown in the question.

(42)

EXAMPLE

Do you exchange your notebook with the good shown underneath? BLANC notebook size A5

No, I choose to keep my notebook

Yes, I give up my notebook and switch to this notebook

Please mark the box indicating your preference. Remember that you still own your good. So, in each question you are comparing the good on that question with the good you own. Please mark one and only one box in each question. After you answer a question place it back in the envelope labeled ‘Part 2’. Once you placed the question back in the envelope, please do not take it out again. And again, there is no order in which you should answer, you are free to get any question.

You will throw a 10-sided dice to decide which question from Part 2 will be selected and receive the good chosen in that question.

(43)

10-sided dice and suppose it stops on the number 7. John will be instructed to take question 7 from the envelope. Suppose it is marked as follows:

John now owns the good that was handed to him earlier:

After all 9 questions are answered, the payment stage starts and assume that part 2 is randomly drawn. The dice rolls and stops, for example, on number 7. John takes question 7 out of the envelope. Suppose his question was marked as follows:

Yes, I give up my notebook and switch to this notebook 7

(44)

Now instead suppose that John’s choice of question 7 was marked as follows: Do you exchange your notebook with the good shown underneath?

BLANC notebook size A5

Yes, I give up my notebook and switch to this notebook 7

John has indicated that he wants to switch his good for the alternative good and will therefore receive the alternative good.

(45)

Appendix F Instructions part 2 treatment B

Part 2

The good you just received

This is a picture of the pen. For participating in this experiment, you have earned this good and it is now yours.

The experiment

As of now, you earned and own the on your desk.

Please take the envelope labeled ‘Part 2’. Inside are 9 questions.

The question shows a picture of a good. You are then asked whether:

• You would like to keep your own good (the one shown on your card/table), or • Give up your good and switch to the good shown in the question.

(46)

Do you exchange your pen with the good shown underneath? RED BIC pen

No, I choose to keep my pen

Yes, I give up my pen and switch to this pen

Please mark the box indicating your preference. Remember that you still own your good card. So, in each question you are comparing the good on that question with the good you own. Please mark one and only one box in each question. After you answer a question place it back in the envelope labeled ‘Part 2’. Once you placed the question back in the envelope, please do not take it out again. And again, there is no order in which you should answer, you are free to get any question.

You will throw a 10-sided dice to decide which question from Part 2 will be selected and receive the good chosen in that question.

John is a participant in the experiment. He answers the 9 questions one by one and places them back in the envelope. At the end of the experiment, if Part 2 is randomly drawn, he throws the

(47)

John now owns the good shown in the picture below that was handed to him earlier:

After all 9 questions are answered, the payment stage starts and assume that part 2 is randomly drawn. The dice rolls and stops, for example, on number 7. John takes question 7 out of the envelope. Suppose his question was marked as follows:

Yes, I give up my pen and switch to this pen 7

(48)

Now instead suppose that John’s choice of question 7 was marked as follows: Do you exchange your pen with the good shown underneath?

RED BIC pen

Yes, I give up my pen and switch to this pen 7

John has indicated that he wants to switch his good for the alternative good and will therefore receive the alternative good.

(49)

Appendix G Questions part 2 treatment A

Yes, I give up my notebook and switch to this notebook 1A

Do you exchange your notebook with the good shown underneath? LINED notebook size A5

(50)

Do you exchange your notebook with the good shown underneath? CHECKERED notebook size A6

Yes, I give up my notebook and switch to this notebook 3A

Do you exchange your notebook with the good shown underneath? LINED notebook size A5

(51)

BLUE BIC pen

Yes, I give up my notebook and switch to this pen 5A

Do you exchange your notebook with the good shown underneath? BLUE BIC pen

(52)

Do you exchange your notebook with the good shown underneath? RED BIC pen

Yes, I give up my notebook and switch to this pen 7A

Do you exchange your notebook with the good shown underneath? BLACK BIC pen

(53)

Do you exchange your notebook with the good shown underneath? GREEN BIC pen

(54)

Appendix H Questions part 2 treatment B

Do you exchange your pen with the good shown underneath? BLUE BIC pen

Yes, I give up my pen and switch to this pen 1A

(55)

Do you exchange your pen with the good shown underneath? BLUE BIC pen

Yes, I give up my pen and switch to this pen 3A

Do you exchange your pen with the good shown underneath? GREEN BIC pen

(56)

Do you exchange your pen with the good shown underneath? CHECKERED notebook size A5

Yes, I give up my pen and switch to this notebook 5A

Do you exchange your pen with the good shown underneath? BLANC notebook size A5

(57)

LINED notebook size A5

Yes, I give up my pen and switch to this notebook 7A

CHECKERED notebook size A6

(58)

Do you exchange your pen with the good shown underneath? LINED notebook size A5

(59)

Appendix I Survey

Survey

What is your gender?

………

What is your nationality?

………

What is your age?

………

Did you strongly prefer one category of goods above the other (pen, notebook)?

………

This was the experiment. Thank you for participating!

(60)

7 References

Bewley, T. F. (1986): “Knightian Decision Theory. Part I,” Cowles Foundation discussion paper, 807.

Kahneman, D. and A. Tversky (1979): “Prospect Theory: An Analysis of Decision Under Risk,” Econometrica, 47(2), 263–292.

Knetsch, J. L. (1989). The endowment effect and evidence of nonreversible indifference curves. The american Economic review, 79(5), 1277-1284.

Koszegi, B. and M. Rabin (2006): “A Model of Reference-Dependent Preferences,” The Quarterly Journal of Economics, 121(4), 1133–1165.

Maltz, A. (2016): “Exogenous Endowment - Endogenous Reference Point,” Working Paper.

Maltz, A., & Romagnoli, G. (2017). Status Quo Bias under Uncertainty: An Experimental

Study (No. WP2017/6).

Masatlioglu, Y. and E. A. Ok (2014): “A Canonical Model of Choice with Initial Endowments,” The Review of Economic Studies, 81(2), 851–883.

Mihm, M. (2016): “Reference dependent ambiguity,” Journal of Economic Theory, 163, 495– 524.

Ortoleva, P. (2010): “Status quo bias, multiple priors and uncertainty aversion,” Games and Economic Behavior, 69(2), 411–424.

Samuelson, W. and R. Zeckhauser (1988): “Status Quo Bias in Decision Making,” Journal of Risk and Uncertainty, 1(1), 7–59.