Creativity and Cooperation in Standard Public Goods Experiment

Creativity and Cooperation in Standard PublicGoods Experiment∗

Giorgi Jvarsheishvili †

February 17, 2016

AbstractNumber of researchers observed effects of earning endowments on chang-

ing behaviour in various experiments. Findings also suggest that levels ofcooperation within groups may depend on whether subjects were grantedor have earned their endowments. This paper expands this literature byanalysing how creativity affects cooperation in Public Goods experiment.More specifically, it is explored whether effort level and effort type haveaffect on cooperative behaviour. In addition, individual characteristicssuch as big five personality traits and social value orientation of groupmembers have been elicited to check if these factors influence coopera-tion. According to the results, exerted effort does not explain differencesin cooperation and that personal characteristics are better explanatoryvariables for individual cooperative decisions.

1 IntroductionGroup work has become increasingly common in modern organizations. In newtechnological era creator groups often come together to share their expertise andwith this, expand common stock of knowledge. For example often coders andprogram developers work in small groups, cross checking each others work andgenerating new software solutions whilst cooperation. In general, knowledgeflows and accumulation are inherently problems involving multiple actors, asare the range of approaches of organizing innovation through some form ofcompetition or collaboration (Boudreau and Lakhani, 2015).

Under these circumstances it can be very useful to know if groups com-pleted of creators cooperate in a different way compared to those who work on

∗I am thankful to Tobias Regner, Alexia Gaudeul, participants of the IMPRS BBS and theparticipants of 9th IMPRS Summer Academy for the valuable discussions. I thank ClarissaZimmermann for assistance. I am deeply grateful to Friedrich Schiller University of Jena forthe financial support.

†Friedrich-Schiller-University Jena, International Max Planck Research School on AdaptingBehavior in a Fundamentally Uncertain World. [email protected]

1

a non-creative, a routine task. Moreover, whether effort levels, which repre-sents behavioural sunk costs, alter intra-group decisions. Even if sharing ownknowledge can be mutually beneficial for all group members within the samecompany, it may carry costs for some individuals which may yield at aggregatelevel. Such conditions correspond to a cooperation dilemma, similar to publicgoods dilemma (Cabrera and Cabrera, 2002).

Researchers from the field of behavioural economics have addressed socialdilemmas, especially those that frequently come up in public goods games(PGG). In these games defection is considered to be a dominant strategy; how-ever, experimental evidence shows that individuals still contribute significantamounts. Scientists started looking for factors responsible for this result. Oneof explanations is that experimenters usually granted subjects with their en-dowments. Although, in general, levels of effort exerted for earning endowmentsshould not matter for making a decision on cooperation, research suggests thatpeople change their behaviour when behavioural sunk costs are present (Zee-lenberg and van Dijk, 1997).

To shed more light on this issue, scholars (e.g. Cherry et al. 2005; Spraggonand Oxoby 2009; Muehlbacher and Kirchler 2009) analysed how earning one’sendowment modifies the decision to cooperate in public goods games. Number ofother researchers have applied slightly different experimental designs to addressthe same question and their results are rather mixed. The current projectoffers a cleaner design to reveal whether cooperation changes if endowments areearned not only by different effort levels but also with different effort types. Theexpected link between effort and cooperation draws on literature on endowmenteffect (e.g. Loewenstein and Issacharoff 1994), that has shown that performanceheightens the value attributed to earnings. Assuming a straightforward relation,strengthened endowment effects should cause comparatively less cooperation.

Extending that, the literature on valuation anomalies such as creativity ef-fect, indicates that certain circumstances amplify such endowment effect (seee.g. Buccafusco and Sprigman 2010, 2011; Norton et al. 2012). Namely, self-construction, self-building and self-creation lead to liking, attraction and causeexaggerated valuation of endowments. To put it simply, creators are prone toovervalue their creation and are more hesitant to give up their earnings accruedfrom their creative endeavour.

Hence, Buccafusco and Sprigman (2011) suggest that the creativity effectis a cause for a type of inefficiency specific in markets for the licensing andtransfer of IP. Related to that, this paper reveals whether effort level in generaland the creativity effect in particular could be a reason for reduced cooperationin academia as well as in industry. The current basic research, to the best ofmy knowledge, is the first attempt to explore this question. It may be useful foreveryone to consider if decisions about cooperation are affected by endowmentand creativity effects.

The paper is organized as follows: First, I overview related literature, next Iset the research questions and hypotheses. As a next step I will provide detailedexplanation of the experimental design. Finally, I present results, discussion ofthe findings and suggestions for the future research.

2

2 Literature Review2.1 Origin of endowments and cooperationFrom rationality perspective the source of money should not matter while mak-ing a decision on how to spend it. Each unit of money has the same valuewhether it was earned through hard work or found in a street. In fact Zeelen-berg and van Dijk (1997) showed that this is not the case and individuals actdepending on how high are their behavioural sunk costs. Scientists from areas ofsocial psychology and behavioural economics have observed changing behaviourdepending on whether the participants of a laboratory experiment earned theirresources or not (e.g. Cherry et al. 2002; Franco-Watkins and Acuff 2013).Findings are mixed: some authors found that some subjects are not giving uptheir wealth easily when they have been granted it and are hesitant to invest orput their earnings under risk after working for it. On the opposite, some resultsshowed that experimental subjects are more generous and eager to give up theirearnings easily when they have exerted certain amount of efforts for it (Sprag-gon and Oxoby, 2009). While there is sufficiently large body of research in thisdirection, in what follows I overview some of the most interesting findings.

Cherry et al. (2002) designed a dictator game, which controls self-interestedstrategic behaviour by giving a person complete control over the distribution ofwealth. Compared to previous such games (e.g. Hoffman et al. 1996) experimentparticipants played with earned wealth rather than unearned wealth grantedby the experimenter. The main argument for this is that the earnings haveto be legitimate to produce rational behaviour. The key discovery was thatthe other-regarding behaviour is greatly diminished when bargaining involvesearned wealth, and this behaviour is nearly eliminated when earned stakes arecombined with anonymity. In other words, earning endowment led decisionmakers to dramatically reduce game theoretic off-equilibrium behaviour (i.e.contribution positive amount of money). As a result, the authors concludethat the high cooperation levels and lack of free riding could also be caused bywindfall endowments of the players.

Despite this conjecture, Cherry et al. (2005) could not find supporting ev-idence that the origin of endowments effects group members’ contributions topublic goods. They designed classical linear public goods game, with two stages:earnings and public goods contributions. In earning treatment (T1) participantshad to solve Graduate Management Admission Test (GMAT)1 to earn resources.In the second treatment (T2) endowments were windfall. The game was playedonly once to avoid reciprocal behaviour by other group members in future peri-ods, triggered by generosity in the first periods. The generated data indicatedno significant difference for subjects who earned their endowment relative tothose with windfall endowments. Thus, although the experiment protocol ofearning endowments was similar to that by Cherry et al. (2002), results dif-fer, which as authors themselves presume could be because of different context.

1For more information on GMAT test see www.mba.com

3

That is, public goods game represents a more complex task demanding greatercognitive effort and involving simultaneous decisions by other contributors.

Surprisingly, Spraggon and Oxoby (2009) found opposite effect of earningmoney. In their two person public good game some participants earned theirendowments through the GMAT questions, while the others were granted. Theyobserved that those who earned their endowments contribute more and thosewho were given their endowments contribute less. As a possible explanation theysuggest that such behaviour is due to “anticipatory reciprocity” (also discussedby Cherry et al. 2005; Kroll et al. 2007). That means earners expect thatnon-earners will cooperate and therefore they contribute significantly more.

Earlier experiment by Loewenstein and Issacharoff (1994) showed that stu-dents, after performing a task, exhibited a stronger endowment effect towardsa mug which they received as a compensation, compared to those who simplyreceived the same mug by chance. To test whether the level of effort reallyimpacts contribution in public good setting Muehlbacher and Kirchler (2009)designed experiment where participants had to earn their endowments thougheasy or effortful task. They hypothesize that those who exerted more effort toearn money should value their endowments more and contribute less. Importantnovelty in their design is that subjects within a group were not informed thattheir group members performed different effort tasks. Therefore, they claimto measure "pure impact of effort on cooperation". Their experiment was im-plemented as follows: In a low-effort condition participants had to watch aTV cartoon series and were questioned about the episode simultaneously. Inhigh-effort condition the same episode was shown but they had to listen toa soundtrack from a different episode. Results of debriefing proved that thismanipulation was successful and the participants contributed significantly lesswhen they were assigned to high-effort treatment. The approach is somewhatquestionable. Participants could suffer from irritation or distraction if they hadto listen to sounds which are not in unison with what they saw. Hence, lesscontributions could result not necessarily from higher effort, rather from otherunobserved factors, such as changes in mood. Moreover, in their experimentincentives were very low: participants earned 50 ECU, while the exchange ratewas 42 ECU = 1 EUR. Therefore, to validate the results of the study furtherreplication and modification of the design are needed.

The notion of "House money effects" was proposed by Thaler and Johnson(1990) and it describes increased risk seeking propensity under certain condi-tions. They discovered that individuals choose risky option more frequently intwo stage game, which allowed certain gains in the first stage and risky optionin second stage. While the distribution of gains in two lottery options was thesame, players were more risk seeking in a situation when they had some earningguaranteed in the first stage. Clark (2002) investigated “house money effects” inpublic goods game. He assumed that the decisions about contributions involverisky behaviour as the other group members can always free ride and enjoy ben-efits on the other’s expense. In other words, he expected decreased cooperationin a public goods game, in a situation when participants brought their ownmoney in order to take part in the experiment. Thus, the experiment did not

4

involve earning resources through real effort task. The author could not findsignificant difference between the average contribution levels in own money andhouse money (here it means windfall) treatments.

Harrison (2007) reconsidered the evidence from the Clark (2002)’s experi-ment. After using appropriate statistical methods for the data analysis he foundthat the “house money” does have an effect in standard public goods game. Itis estimated to reduce the probability of providing something by 8.2 percent-age points and this effect is statistically different from zero (p− value < 0.01).Despite final confirmation of house money effects in public goods setting themethod of asking participants to bring their money to a laboratory is unusualand somewhat problematic. An experimenter cannot control how much effortdid participants gave up to get the money. Therefore, although the authors areoften referred in the literature explaining cooperation in PGGs, no implicationsabout effort size and cooperation levels can be made.

Marginally relevant to the topic, since the authors take a look at bargainingrather than cooperation, is a paper by Franco-Watkins and Acuff (2013). Theyexamined how effort affects perceptions of fairness and allocation of resourcesduring bargaining situations. In previous literature scientists relied on one shotbargaining, or ultimatum game (e.g. Güth et al. (1982)) with windfall money,as a source of bargaining resource. The proposers usually found equal split50-50 as the fairest. To test how real effort influences valuation and bargainingFranco-Watkins and Acuff (2013) demanded that experimental subjects performself-arbitrary task, requiring clicking mouse 100 times and self-performance task– answering at least 70% of 20 trivia questions. The results indicated that thosewho completed performance task demanded significantly more money when theother party did not have to endure the same task. That means, willingness toaccept increased when others performed no task and at the same time willingnessto pay was decreased when others performed no task. Thus, even if the effortwas not directly linked to second phase it had small but significant effect onbargaining.

To summarize, results are mixed: some authors found no impact of earn-ing endowments on cooperative decisions in public goods games, while othersdid observe influence of effort level on cooperation. In some cases higher wasthe stated efforts for earning, the less cooperative subjects became. However,even if it seems to be quite intuitive that individuals get more endowed towardssomething they worked for, there are some exceptions when no such effect oreven the opposite effects where found. While the rational argumentation sug-gests that efforts should not matter at all. Thus, the topic still can be exploredfurther to set clear links between origins of wealth and the decision making oncooperation in a public goods setting.

2.2 Creativity and endowment valuationAll progress and innovation depend on ability to change existing thinking pat-terns, break with the present and build something new. Therefore, creativethinking is extraordinary capacity of human mind and has recently become

5

matter of central importance for interdisciplinary research (Dietrich and Kanso,2010). Much attention has been paid to organizational and managerial issuespertaining creativity, as it represents basic and critical element for innovationprocess (Udwadia, 1990). Since groups are often built of creative individuals itis important to know how creativeness impacts individual behaviour. This mayhelp to analyse group members’ decision making and overall group performance.Thus, to link cooperation and creativity, in the next paragraphs some of ratherfew scientific papers on creativity in experiments will be reviewed.

To understand how creativity affects decision making Buccafusco and Sprig-man (2010, 2011) made two experiments. In the first one, some of the partici-pants wrote three-line poems, some were told that they owned the poems andthe others were potential buyers. One of ten such poems had a chance to win aprize. Then the creators and owners were asked to name the price for what theywere willing to sell their poem. The values of their poems should theoreticallybe the value of prize divided by ten. However, subjects demanded significantlymore than expected. More specifically, money willing to accept by the authorswas more than four times to what rational choice theory would predict (i.e.probability of winning multiplied by value of prize). The willing to accept bycreators was two times to what bidders were willing to pay. Thus, owners feltmore endowed towards the poems that they possessed and for the authors ofthe poems, endowment effect was even greater. This amplified endowment ef-fect caused by being an author oneself is what Buccafusco and Sprigman namecreativity effect.

The endowment effect describes situation when the owners’ willingness toaccept the minimum price for their possession is greater than what potentialpurchasers are willing to pay. This phenomenon can be a source of substan-tial inefficiencies on markets. Thus, to examine strengthened endowment ef-fect, Buccafusco and Sprigman (2011), in their follow up paper, exclusivelyfocused on valuation anomaly that is related to creation of new works. Whatthey define as creativity effect may be not less important and significant thanendowment effect. It could be a reason for inefficient allocation of propertyrights or the lack of cooperation between creator groups. The second experi-ment was similar to the first, authors asked painter students from the Schoolof the Art Institute of Chicago to paint paintings and play the game identicalto previous one. New data showed that the valuation gap between creators (ofpaintings) and potential buyers was 4-to-1. This suggests that higher effortsincorporated into creation leads to further divergence in valuations.

Attraction and over valuation of self-constructed products have been ob-served by other authors as well. This phenomenon is regarded as the "IKEAeffect". Norton et al. (2012) have conducted simple experiment and showedthat experiment participants valued origami paper cranes significantly morewhen they built it themselves. The same pattern was found by Franke andSchreier (2010). In their experiment participants who designed scarves valuedit more than the same type of scarves designed by others. Finally, Dohle et al.(2014) found what they call "I cooked it myself" - effect. Their experimentparticipants liked and consumed self-prepared drink significantly more than the

6

same drinks prepared with exactly the same ingredients and receipt by others.Research results on how creativity can be incentivized are not unequivocal.

Eckartz et al. (2012) found in controlled experiment that monetary incentivesdo not enhance creativity. In an experiment, where participant were supposedto build words out of random set of letters, subjects’ creative output was equallywell whether they were paid with flat rate or according to their performance.

Erat and Gneezy (2015) added another dimension: incentivization by com-petition. In their experiment subjects had to create rebuses (puzzles made withwords and/or pictures with a hidden and non-obvious solution). the resultsshowed that monetary inducements do not improve creativity and competitiveincentives can be even detrimental for creative performance.

Although these findings on incentives and creative performance are not di-rectly linked to current project, they indicate that creative effort is not easyto manipulate, it is often driven by intrinsic motivation and therefore influenceof different levels of creative effort on prosocial behaviour may not be straight-forward. In what follows, I set research questions and hypotheses based onprovided theoretical discussion.

3 Research questions and hypothesesThe experimental literature, as discussed above, demonstrated that exertion ofefforts to acquire endowments ends up in different behaviours by subjects. Inmost of the cases working causes higher endowment effect towards earnings.That means participants value their resources, earned through hard work moreand become more hesitant to give them up. Therefore, the first research questionis:

1)How does earning endowments through different real effort tasks with dif-ferent levels of effort affect cooperation in PGG?

Since contribution to public good contains some risks, higher valuation ofendowments could cause less contributions by group members. In other words,they face uncertainty about other’s decisions and show more risk-averse be-haviour when they have earned their endowments through real effort task.Working causes higher valuation of endowments and therefore experiment sub-jects might cooperate less when they have earned their resources through dif-ficult (high effort) task. Thus, in line with Muehlbacher and Kirchler (2009),hypothesis I can be formulated as follows:

• Contributions to PG after earning through high effort task will be lowerthan contributions after earning through low effort task.

In addition, creation can lead to amplification of endowment effect (so calledcreativity effect) and emergence of valuation gap (Buccafusco and Sprigman,2010, 2011). Creators get attached to their creation and overestimate its im-portance. Therefore, it can be presumed that creativity effect may influenceindividuals’ behaviour, when they have to contribute to public good earningsfrom their creative endeavour. Thus the second research question is:

7

2) Do subjects change their cooperation behaviour if they have exerted cre-ative effort?

Although creators are often too attached to their earnings, sometimes cre-ators are eager to share their creation with others as well. For example as itis in the case of Free/Open Source Software (FOSS) (Crosetto, 2010). Suchattitude may even have counter effect on contributions and actually increasethem. However, the case seems to be more an exception rather than commonpattern. Generally, when there is no standard of common sharing, inventionsare assessed by creators themselves and moreover, often overvalued. Therefore,under such circumstances free riding incentives will be higher and cooperationlevel is more likely to be lessened. Thus, hypothesis II can be formulated asfollows:

• Contribution to PG after earning through creative effort task will be lowerthan contributions after earning through routine effort task.

In most social dilemma situations, people cannot communicate with oneanother and therefore they are uncertain about the decisions of their fellow groupmembers. To deal with this kind of environmental uncertainty, participants usetheir own social value orientations and personality traits as a guideline for choicebehaviour (de Kwaadsteniet et al., 2006). Thus, the third research question is:

3) How do individual personality traits and Social Value Orientation affectcooperation?

Social value orientation (SVO) is defined as a personality variable that in-dicates how people evaluate outcomes for themselves and others (Messick andMcClintock, 1968; Van Lange and Liebrand, 1991; de Kwaadsteniet et al., 2006).For example, Van Lange (1999) categorized SVO as (a) cooperation, i.e., thepreference to maximize joint outcomes and establish an equal distribution, (b)individualism, i.e., the preference to maximize own outcomes, and (c) com-petition, i.e., the preference to maximize relative advantage. Cooperators areusually considered as prosocials, individualists and competitors – proselfs. De-pending on this definition and findings by Murphy et al. (2011), hypothesisIII (a), can be formulated as follows:

• a) SVO will be predictive for cooperation levels: Prosocials will contributemore to PG than proselfs.

Last two decades’ research on personality proved that the personality mea-sures, may explain human behaviour in different dimensions, such as cooperationwithin groups (e.g. volk2011personality). Researchers investigating relationshipbetween Big Five personality traits and social behaviour (e.g. Volk et al. 2011)found that in PGG setting Agreeableness and prosocial values were indicative ofindividual preferences for cooperation. In addition, laboratory study by LeP-ine and Van Dyne (2001) observed positive relation between cooperation andpersonality traits: agreeableness, emotional stability and extraversion. Thus,hypothesis III (b) can be formulated as follows:

• b) Big Five dimensions of agreeableness, emotional stability and extraver-sion will be positively related to cooperation.

8

4 Experimental DesignProcedure and design. The experiment had two, high and low effort treat-ment, with two types of tasks, routine and creative. Hence, there were fourconditions (see table 2). The experiment was conducted in two phases. That is,for example, in the low effort treatment participants earned their endowmentsthrough a routing task, made their contributions in standard PGG, then theyearned through a creative effort task and again made their contributions tostandard PGG without getting feedback in between. To eliminate order effectssequence of tasks were switched in different sessions. Similar was the high efforttask, the only difference was the amount of effort needed to solve the tasks.Number of observations in different treatments and different sequences of thetasks were balanced. LTR1 and HTR1 means low and high effort treatmentswith the task sequence: routine task - creative task. LTR2 and HTR2 denotethe opposite sequence: creative task - routine task (see table 1).

LTR1 HTR1 LTR2 HTR2N of obs. 27 36 27 33

Table 1: Number of Observations in different sessions with varied sequence oftasks

After reading instructions on how the experiment is organized and how theyare supposed play the game, subjects had opportunity to observe how hypothet-ical contributions can be redistributed using PGG calculator (see instructions inappendix A.1.1). There were three person groups. Participants were informedthat they would earn twice and contribute to PGG twice, however only one oftheir decisions would be pay-off relevant. Any positive contribution was doubledand redistributed among the three group members.

πi = pi + 1/2∑N

i=1(ji)Where,

• πi - Earnings of i

• pi - allocation to personal account

• ji - allocation to joint account

• N - number of group members

Thus the marginal per capital return (MPCR) was 0.66. Which is higherthan usual 0.4 -0.5 in PG games, the reason is that research goal was to observeprobable decline of contributions across the treatments. In this case, it wassocially optimal if everyone contributes everything, individually the dominantstrategy was zero contribution in all of the treatments (Thaler and Johnson,1990). After getting to know how the PGG works, subjects answered controlquestions. In all treatments earning per subject was 50 ECU, i.e. there was

9

no inequality between group members and across groups. After the controlquestions they solved real effort task.

Real-effort task N of obs.Treatment "R.L.eff" Counting a letter 54Treatment "R.H.eff" Counting two letters 69Treatment "C.L.eff" Creating words 54Treatment "C.H.eff" Creating more words 69

Table 2: Experimental Conditions

Real effort tasks. Examples from the previous literature comparing be-haviour in PGG setting after effort exertion include: Graduate ManagementAdmission Tests (GMAT) (Cherry et al., 2005; Spraggon and Oxoby, 2009);and answering questions about plot and visual images from watched 6 minuteepisode of TV cartoon (Muehlbacher and Kirchler, 2009).

In this experiment in routine effort treatments participants had to count oneor two letters within given strings of letters2. For example, count how manytimes a letter “a” appears in characters strings with different lengths, the longerthe letter strings were, the higher were the points subjects would receive. Aftersurpassing certain threshold of points they would automatically proceed to thenext stage (See figure in appendix A.2).

In psychological creativity research literature, authors often ask subjects tothink of creative stories or creative solutions to open questions. Which are mod-ifications of the Torrence Test (Torrance, 1968). Other examples of creative orinnovative tasks include: Writing three line poems; Paint paintings (Buccafuscoand Sprigman, 2010); Designing an Auto-mobile (Cantner et al., 2009); Wordcreation task (Eckartz et al., 2012) and word extension (modified scrabble)tasks (Crosetto, 2010).

In this experiment I applied a word creation task 3 similar to one, used byEckartz et al. (2012). The main reasons behind are that it has many aspects ofa creative task and it mimics quite well a creative innovation. Also, it takes lesstime than other options, it is easy to program, objectively assesses creativityand makes possible to count efforts exerted by players (see figure in appendixA.2).

The longer words the participants built, the higher the points were granted(see table 3). After getting more than a threshold level 4 students would earn 50ECUs and proceed to the next stage. The difference between low and high effortroutine tasks was that in high effort treatment subjects had to count two letters

2The task was inspired by the real effort task designed by Rey-Biel et al. (2011)3I am thankful to Diego d’Andria and Igor Asanov for providing a code, implemented in

programming languages Ruby and R, to generate the letter string and the list of possiblewords for the creative effort task

4Students could collect up to 600 points with a given letter string. However, thresholdswere 18 points in low effort treatment and 27 points in high effort treatment to keep the tasksufficiently effortful, but not too difficult.

10

in the same letter strings and collect twice more points. In high effort creativetask subjects had to create words from the same letter string, however they hadto collect one and half times more points than in low creative task. It has tobe emphasized that effortfulness of routine and creative tasks were maximallyapproximated: In low effort routine task solving the longest letter string wouldsuffice to proceed. Likewise, creating the longest words (which students couldusually come up with) was enough to collect necessary points. While in higheffort treatment students had to solve minimum of three letter strings and findabout three words to proceed.

Words from the letter set " a c c d e e e g i n s t "ad 1 + 2 = 3 pointsand 1 + 2 + 3 = 6 pointscats 1 + 2 + 3 + 4 = 10 points... ...teasing 1 + 2 + 3 + 4 + 5 + 6 + 7 = 28 points

Table 3: Measuring creativity: Longer words generate more points (Eckartzet al., 2012)

Finally, after making second contribution decision, experiment participantswere asked manipulation-check questions. Then, they were informed which oftheir contribution decisions was pay-off relevant, amounts of contributions bytheir group members and their total profit. As a last step, they answered SVO(Murphy et al., 2011) and personality (Gosling et al., 2003) questionnaires.They were remunerated and released.

The experiment was conducted in Z-tree (Fischbacher, 2007). Partici-pants were recruited by Online Recruitment System for Economic Experiments(ORSEE) (Greiner, 2015).

11

5 ResultsFor statistical analysis and graphs I used R software (R Core Team, 2013).

Experiment was conducted at Friedrich Schiller University Jena, duringNovember 2015. In total 123 subjects have participated in the experiment,majority of them were students at the university. Average earning was 8 euros(with a range from 3.8 to 11.2 euros), for the 45 minute experiment, which is25 % higher than minimum student assistant salary.

Manipulation check. To enable comparison of effort levels exerted, per-ceived difficulty of the tasks and exerted creative effort, the following manipula-tion questions where asked to the subjects: 1. In comparison to the first task didyou feel that the second task required higher degree of effort? (1= same degreeof effort; 8= too high effort) 2. In comparison to the first task, did you feel thatthe second task was as easy as the first task ? (1= as easy as the first task; 8=too difficult) 3. In comparison to the first, did you have feeling that the secondtask required creative effort? (1= same degree of creative effort; 8= too highcreative effort)5. Results (see table 4) of manipulation check questions showedthat the participants considered creative task to require high level of creativeeffort. While the the effort level as well as perceived difficulty was about thesame for the two different types of tasks. Difference between the perceived effortlevel and perceived creative effort was significant at p < 0.01 level.

Effortfulness difficulty Creative effort EFF_ManipulationMedian :4.000 Median :3.000 Median :5.000 LowMean :4.074 Mean :3.426 Mean :4.926 LowMedian :3.000 Median :3.000 Median :5.000 HighMean :3.464 Mean :3.275 Mean :4.841 High

Table 4: Manipulation check

In figure 1 contribution frequencies are presented across different treatments.Unlike previous research by Muehlbacher and Kirchler (2009) (where there wasonly one free rider among all participants) in this experiment, portion of absolutefree riders, those who contributed zero is substantial. The reason behind isadequate incentivization, which heightens internal validity of the experiment.

In the figure 2 mean contributions across treatments are displayed. As itcan be seen, there are no significant differences across treatments, which meansthat the type of task and the effort levels are not indicative for the chosen levelof cooperation. (See Anova and Boxplot for the contributions across treatmentsin appendix A.3)

I the table 5 results from the paired t-tests for the same levels of effortsand results from the Welch two sample t-tests for different levels of effort areprovided. The examination of the means confirms that there is no significantstatistical difference between the means.

5In different sequences of the tasks, manipulation-check questions were reformulated sothat subjects always compared creative task to routine task.

12

Figure 1: Frequency of contributions across different treatments

Figure 2: Mean Contributions and Standard deviations across treatments

13

R.Low.eff - C.Low.eff Results1 Paired t-test: t(53) = 1.29, p = .201, d = 0.25

R.High.eff - C.High.eff Results2 Paired t-test: t(68) = -1.29, p = .200, d = -0.22

R.Low.eff - R.High.eff Results3 Welch Two Sample t-test: t(117.78) = 1.12, p = .264, d = NA

C.Low.eff - C.High.eff Results4 Welch Two Sample t-test: t(112.95) = -0.05, p = .963, d = NA

Table 5: T-tests

To identify effect of effort type on contribution size first I pooled all contri-butions and regressed 6 them on effort type (see table 5). Performing low effortroutine task seems to lead to higher cooperation in comparison to creative higheffort task, however the difference is not significant (p− value > 0.1). Hence,hypotheses I and II are not supported.

Dependent variable:Contributions Contributions

OLS/Censored OLS Tobit(1) (2) (3)

R.L.eff 2.189 0.736 3.261(3.024) (2.215) (4.256)

R.H.eff -0.144 0.753 -0.542(3.024) (2.314) (4.278)

C.L.eff -1.072 1.159 -2.025(2.834) (2.127) (4.000)

Constant 24.014∗∗∗ 20.362∗∗∗ 25.019∗∗∗

(2.004) (1.512) (2.831)Observations 246 171 246R2 0.005 0.002Adjusted R2 -0.007 -0.016Log Likelihood -860.309Residual Std. Error 16.646 (df = 242) 10.365 (df = 167)F Statistic 0.402 (df = 3; 242) 0.102 (df = 3; 167)Wald Test 1.591 (df = 3)

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01

Table 6: Cooperation and types of efforts

Executing Tobit regressions (see table 7) with dummy variables for the tasktype and effort level shows that routine effort positively affects contributions.While high effort decreases cooperation for routine tasks. However, the coef-ficients are not significant (p− values > 0.1). Therefore, the influence of tasktype and effort level are not proved.

In appendix A.6 I have dichotomized dependent variable, distinguishing be-tween zero and positive contributions. For the routine effort task, effort level

6For the regression tables Stargazer package for programming language R was applied(Hlavac, 2014)

14

Dependent variable:All Contributions Contribution R. task Contribution C. task

(1) (2) (3)R. Eff. dummy 0.548

(3.001)EFF_level dummy -5.109 0.581

(3.981) (4.577)Constant 24.785∗∗∗ 28.093∗∗∗ 24.542∗∗∗

(2.130) (2.980) (3.434)Observations 246 123 123Log Likelihood -861.089 -436.946 -422.745Wald Test (df = 1) 0.033 1.647 0.016

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01

Table 7: Dummies for Task type and Effort level

dummy is marginally significant for model (3), indicating that higher effortwould cause less positive contributions.

Interestingly, number of zero contributions to public good after creative efforttask was 33% higher than number of no contributions after routine task. Despitefinding no difference in mean contributions across treatments, this finding relatesto hypothesis II, which predicted that creative effort should lessen contributions.

In order to analyse affects of Social Value Orientation, I have applied method-ology developed by Murphy et al. (2011). Personality traits of the experi-ment participants were measured by Ten-Item Personality Inventory (TIPI) byGosling et al. (2003). Correlation among Big Five measures and SVO is pro-vided in appendix A.4. According to the results, SVO is positively related tocontributions (p− value < 0.05), extraversion is also positively related to coop-eration, however only after performing routine effort task (p− value < 0.05),for creative effort task significance was marginal (p− value < 0.1). Surpris-ingly, agreeableness is negatively linked to contributions (p − value < 0.05).(All regression results are reported in the appendix A.5 7).

6 DiscussionIn the current paper possible effects of earning endowments on cooperation deci-sions have been investigated. Previous research has identified that behaviouralsunk costs, i.e. exerting effort to acquire endowments affected subject decisionin different domains. Muehlbacher and Kirchler (2009) found that effort levelcould determine the willingness to cooperate within a group. However, the find-ing contradicts other researchers’ results. For example, Spraggon and Oxoby(2009) found that real efforts increased contribution size compared to those whodid not perform any task at all. Whereas, Cherry et al. (2005) did not observe

7Note: In all regressions second model includes reversed measures for big five personalitytraits.

15

any difference in contribution sizes among those who did perform a task andthose who did not. The main difference between Muehlbacher and Kirchler’sexperimental design and that of previous researchers was that in other exper-iments participants were informed about the asymmetry of origins of endow-ments. Providing such information could have led to "anticipated reciprocity",expectation that non-earners would contribute more. In order to measure pureeffort affect, Muehlbacher and Kirchler (2009) did not inform the subjects aboutthe diverse origins of group members’ endowments. Such approach successfullydealt with part of the problems, however, the design and the implementationof the experiment suffered from several technical problems: Effort was manipu-lated by watching a cartoon video under various soundtracks, which could havecaused distraction, annoyance and irritation. Thus, the mood effects could haveplayed a role rather than effort. Moreover, most likely incentives were not highenough to trigger self-interested decision making: endowments 50 ECU = 1.2EUR, which could have amounted to maximum earning of less than 3 EUR forthe group member who had fully defected, while all other group members hadfully cooperated. Such poor incentive could be reason why there was only onedefector (contributing zero ECUs).

To fix the above mentioned problems, firstly adequate incentives where given:50 ECU = 5 EUR, and maximum possible earning was 11.6 EUR. As a results12.2 % of all 246 contribution decision were zeros. Besides, in this research, aletter counting task was introduced (inspired by the task from Rey-Biel et al.2011), which could not have led to any kind of mood effects and required simpleroutine effort from the students.

The experiment, also advanced the state of research further. To the best ofmy knowledge, this is the first experiment to analyse effects of creative effort oncooperation. While the previous literature examined influence of different levelsof efforts on cooperation, in the present study, types of efforts were also ad-dressed by employing conceptually different, word creation task (Eckartz et al.,2012). Understanding whether the effort type effects cooperation is very impor-tant as in modern knowledge based economies, types of tasks differ in multipledimensions. Demand for highly creative workers, such as programmers, codersand data scientists are several times higher than available specialists. In addi-tion, such experts often work in small groups. They usually have earned theirknowledge and wealth with their past, individual creative endeavour. Thus, itis important to know how they possibly feel about cooperating with their peers.Therefore, current experiment can be considered as opening of a new horizonfor the future research in this part of behavioural science.

In line with previous literature it was expected that the higher level of effortcould have led to lower cooperation. Earning an endowments could have causedreversed sunk cost effect, an increase in risk aversive choices, which in this casewas contribution to the public good. Thus, through the "behavioural sunk cost"(Zeelenberg and van Dijk, 1997), subjects could have gotten more endowed totheir earnings and hesitate to cooperate. The reason behind non-observed dif-ferences could be complexity of the Public Goods Game: even though it waswell explained and PGG calculator (an instrument to ease understanding of the

16

character of the game) was extensively used, participants might have failed toinclude the endowment effect in a more complex task, demanding greater cogni-tive effort and involving simultaneous decisions by other contributors (Cherryet al., 2005). Alternative explanation for non-observed difference in contributionmeans is amount of effort exerted by experimental subjects. Since in laboratoryexperiments take on average one hour, it is difficult to imitate the effort levelsexerted in real life situations.

In addition, to no significant effects of effort level, no affects of effort typewas identified. This could be related to the following factors:

1) The literature which has observed over-valuation of creations - the creativ-ity effect, has applied tasks related to arts: writing poems or painting paintings( Buccafusco and Sprigman 2010, 2011), which are more personal and could havecaused more attachment to the artefact. In psychological creativity literaturecommonly tasks are open questions, related to diverse topics. The responsesare usually analysed by several independent jurors. This study applied differ-ent, word creation task for the following reasons: First, it is relatively easyto program and execute in computer laboratory; Second, no evaluator jurorsare needed, which is highly time consuming and despite analysis of correlationamong jurors’ decisions, estimation of creative performance can still be con-sidered to be subjective. In word creation task, effort levels could be easilymanipulated by increasing threshold of needed points and moreover, it was sim-ilar to routine effort task in many ways: Both tasks were to be done on letters,in both cases participants could have chosen alternative strategies: solve shortletter strings/find short words and collect needed points in small incrementsor solve longer letter strings/find long words and collect needed threshold withrelatively bigger increments. One of important issues could be, that letting theparticipants know what the exact value (50 ECU = 5 EUR ) of their effort was,probably limited the individual interpretation and valuation of their creativity.Eventually there was no room left for creativity effect to emerge.

2) While in current experiment participants had to contribute earnings fromcreative task, in previous research subjects had to make decisions on the cre-ations. For example, sell paintings drawn by themselves, sell self-designed scarf,or the self-built origami paper cranes.

3) Eckartz et al. (2012) found that although the word creation task is consid-ered to demand creative effort, it was also enjoyable, entertaining and thereforeincentives could not increase creative performance on this task. Thus, the en-joyability of the task could have caused additional utility to participants, thatin the end counterbalanced the unpleasantness of effort exerted while workingon it.

4)It is generally acknowledged that for subjects environment in a laboratoryduring experiments is often comparable to that during examinations. There-fore, it is logical to assume that participants get certain level of stress whileperforming real effort tasks. Research by von Dawans et al. (2012) showedthat experiment participants who experienced acute social stress, induced bya standardized laboratory stressor, engaged in substantially more prosocial be-haviour (trust, trustworthiness, and sharing) compared with participants in a

17

control condition. Thus, engaging in prosocial behaviour in response to stress,functions in an opposite direction compared to endowment or creativity effect,which initially were proposed to lead to less social behaviour. As a result, theabove mentioned stress response could have eliminated the manipulation effect.

5) Another critics to the current project is the manipulation of effort levels.High effort could also be interpreted as solving a more difficult task withinthe same time limit as easy task. For example, adding single figured numbersmultiple times or multiplying two, three-digit numbers once, by hand. Althoughsuch approach is interesting it is relatively complicated to administer, as itmay lead to losing experimental control as experimenter cannot be sure thatparticipants will manage to solve the task in a given time-frame at all. Solutionto that could be to invite only specialists of a given tasks. Therefore, futureresearch can also be focused on investigating how effort levels affect cooperativedecisions of individuals specialized in certain fields.

According to the results SVO was positively related to contributions to PGG.This is one more proof of an accuracy of SVO measure for prosocial behaviour.Big Five personality domain of Extraversion was positively related to coopera-tion level (p− value < 0.05). However, unexpectedly Agreeableness was nega-tively related to contribution size (p− value < 0.05). To measure Agreeablenessparticipants were asked how sympathetic and warm did the participants con-sidered themselves on a scale 1 to 7. Apparently, those who regard themselvesas warm and sympathetic are in fact less cooperative. This could be explainedthe social context that was created in experiment. Not only environmentalinfluences change person’s behaviour, but behaviour is also affected by indi-vidual cognition, i.e. how they perceive something evoked by different context(Hennessey, 2003; Hoff et al., 2012). In this case, once defectors made low con-tributions, they might have been inclined to regard themselves as open personsin Big Five questionnaire, comparatively more open than those who were actu-ally more cooperative. With this they might have tried to present themselvesas open individuals despite their non-cooperative behaviour in the experiment.

To summarize, it can be said that exerted effort levels, for earning endow-ments, are not determinants for the chosen cooperation level in small groups.Moreover, creative type of effort does not changed preferred cooperative be-haviour. The research has several limitations: the first is that effort manipula-tion took only few minutes, while in real life individuals invest a lot more effortto earn their wealth. The experiment tested only for the effects of specific typeof creativity, namely verbal creativity. Therefore, further research is needed toincrease the effort level needed to earn money, and to check for other types ofcreative efforts to conclude that creative and innovative effort in general is notan influential factor for cooperation.

7 ConclusionContribution of the paper to the literature is twofold: Firstly, endeavour toreplicate findings of already existing research in a cleaner experiment was made,

18

by employing a design free of technical problems, present in previous research.Past research found out that earning endowments has influence on contributiondecisions. In some cases exerted high effort discouraged prosocial behaviour, andin other cases encouraged it. The results of the present research do not supportthe findings of those authors ( Muehlbacher and Kirchler 2009; Spraggon andOxoby 2009)and favour the conclusions made by those who could not observeimpact of efforts on cooperation (Cherry et al., 2005). Secondly, influenceof conceptually different, creative type of effort on contributions was tested inpublic goods game. Based on analysis of the experiment data, it can be statedthat there was no creativity effect found in public goods setting.

In conclusion, further research is needed to account for the problems dis-cussed in the previous section. Implement longer experiment, invite actual cre-ators, such as artists, coders or programmers to the lab and let the experimentparticipants contribute self-created artefacts to the public good.

19

ReferencesBoudreau, K. J. and Lakhani, K. R. (2015). Innovation experiments: Research-ing technical advance, knowledge production and the design of supportinginstitutions. In Innovation Policy and the Economy, volume 16, edited byJosh Lerner and Scott Stern. Chicago: University of Chicago Press.

Buccafusco, C. J. and Sprigman, C. J. (2010). Valuing intellectual property:An experiment. Cornell Law Review, 96:1–46.

Buccafusco, C. J. and Sprigman, C. J. (2011). The creativity effect. Universityof Chicago Law Review, 78:31–52.

Cabrera, A. and Cabrera, E. F. (2002). Knowledge-sharing dilemmas. Organi-zation studies, 23(5):687–710.

Cantner, U., Güth, W., Nicklisch, A., and Weiland, T. (2009). Competition inproduct design: An experiment exploring innovation behavior. Metroeconom-ica, 60(4):724–752.

Cherry, T. L., Frykblom, P., and Shogren, J. F. (2002). Hardnose the dictator.The American Economic Review, 92(4):1218–1221.

Cherry, T. L., Kroll, S., and Shogren, J. F. (2005). The impact of endowmentheterogeneity and origin on public good contributions: evidence from the lab.Journal of Economic Behavior & Organization, 57(3):357 – 365.

Clark, J. (2002). House money effects in public good experiments. ExperimentalEconomics, 5(3):223–231.

Crosetto, P. (2010). To patent or not to patent: a pilot experiment on incentivesto copyright in a sequential innovation setting. In Open Source Software: NewHorizons, pages 53–72.

de Kwaadsteniet, E. W., van Dijk, E., Wit, A., and de Cremer, D. (2006). Socialdilemmas as strong versus weak situations: Social value orientations and tacitcoordination under resource size uncertainty. Journal of Experimental SocialPsychology, 42(4):509–516.

Dietrich, A. and Kanso, R. (2010). A review of eeg, erp, and neuroimagingstudies of creativity and insight. Psychological bulletin, 136(5):822–848.

Dohle, S., Rall, S., and Siegrist, M. (2014). I cooked it myself: Preparing foodincreases liking and consumption. Food Quality and Preference, 33:14–16.

Eckartz, K., Kirchkamp, O., and Schunk, D. (2012). How do incentives affectcreativity?

Erat, S. and Gneezy, U. (2015). Incentives for creativity. Experimental Eco-nomics, pages 1–12.

20

Fischbacher, U. (2007). z-tree: Zurich toolbox for ready-made economic exper-iments. Experimental Economics, 10(2):171–178.

Franco-Watkins, A. M., E. B. D. and Acuff, R. E. (2013). Effort and fairness inbargaining games. Journal of Behavioral Decision Making, 26:79–90.

Franke, N. and Schreier, M. (2010). Why customers value self-designed prod-ucts: The importance of process effort and enjoyment*. Journal of ProductInnovation Management, 27(7):1020–1031.

Gosling, S. D., Rentfrow, P. J., and Jr., W. B. S. (2003). A very brief mea-sure of the big-five personality domains. Journal of Research in Personality,37(6):504 – 528.

Greiner, B. (2015). Subject pool recruitment procedures: organizing experi-ments with orsee. Journal of the Economic Science Association, pages 1–12.

Güth, W., Schmittberger, R., and Schwarze, B. (1982). An experimental anal-ysis of ultimatum bargaining. Journal of economic behavior & organization,3(4):367–388.

Harrison, G. W. (2007). House money effects in public good experiments: Com-ment. Experimental Economics, 10(4):429–437.

Hennessey, B. A. (2003). The social psychology of creativity. ScandinavianJournal of Educational Research, 47(3):253–271.

Hlavac, M. (2014). stargazer: LaTeX code and ASCII text for well-formattedregression and summary statistics tables. R Foundation for Statistical Com-puting, R package version 5.1.

Hoff, E. V., Carlsson, I. M., and Smith, G. J. (2012). Chapter 11 - personality.In Mumford, M. D., editor, Handbook of Organizational Creativity, pages 241– 270. Academic Press, San Diego.

Hoffman, E., McCabe, K., and Smith, V. L. (1996). Social distance andother-regarding behavior in dictator games. The American Economic Review,86(3):653–660.

Kroll, S., Cherry, T., and Shogren, J. (2007). The impact of endowment het-erogeneity and origin on contributions in best-shot public good games. Ex-perimental Economics, 10(4):411–428.

LePine, J. A. and Van Dyne, L. (2001). Voice and cooperative behavior ascontrasting forms of contextual performance: evidence of differential relation-ships with big five personality characteristics and cognitive ability. Journalof Applied Psychology, 86(2):326–336.

Loewenstein, G. and Issacharoff, S. (1994). Source dependence in the valuationof objects. Journal of Behavioral Decision Making, 7:157–157.

21

Messick, D. M. and McClintock, C. G. (1968). Motivational bases of choice inexperimental games. Journal of experimental social psychology, 4(1):1–25.

Muehlbacher, S. and Kirchler, E. (2009). Origin of endowments in public goodgames: The impact of effort on contributions. Journal of Neuroscience, Psy-chology, and Economics, 2(1):1–12.

Murphy, R. O., Ackermann, K. A., and Handgraaf, M. (2011). Measuring socialvalue orientation. Judgment and Decision Making, 6(8):771–781.

Norton, M. I., Mochon, D., and Ariely, D. (2012). The {IKEA} effect: Whenlabor leads to love. Journal of Consumer Psychology, 22(3):453 – 460.

R Core Team (2013). R: A Language and Environment for Statistical Comput-ing. R Foundation for Statistical Computing, Vienna, Austria.

Rey-Biel, P., Sheremeta, R., and Uler, N. (2011). (bad) luck or (lack of) effort?:Comparing social sharing norms between us and europe. Technical report.

Spraggon, J. and Oxoby, R. J. (2009). An experimental investigation of en-dowment source heterogeneity in two-person public good games. EconomicsLetters, 104(2):102 – 105.

Thaler, R. H. and Johnson, E. J. (1990). Gambling with the house moneyand trying to break even: The effects of prior outcomes on risky choice.Management Science, 36(6):643–660.

Torrance, E. P. (1968). Torrance tests of creative thinking. Personnel Press,Incorporated.

Udwadia, F. E. (1990). Creativity and innovation in organizations: Two modelsand managerial implications. Technological Forecasting and Social Change,38(1):65–80.

Van Lange, P. A. (1999). The pursuit of joint outcomes and equality in outcomes:An integrative model of social value orientation. Journal of personality andsocial psychology, 77(2):337.

Van Lange, P. A. and Liebrand, W. B. (1991). Social value orientation andintelligence: A test of the goal prescribes rationality principle. Europeanjournal of social psychology, 21(4):273–292.

Volk, S., Thöni, C., and Ruigrok, W. (2011). Personality, personal values andcooperation preferences in public goods games: A longitudinal study. Per-sonality and individual Differences, 50(6):810–815.

von Dawans, B., Fischbacher, U., Kirschbaum, C., Fehr, E., and Heinrichs,M. (2012). The social dimension of stress reactivity acute stress increasesprosocial behavior in humans. Psychological science, 23(6):651–660.

22

Zeelenberg, M. and van Dijk, E. (1997). A reverse sunk cost effect in riskydecision making: Sometimes we have too much invested to gamble. Journalof Economic Psychology, 18(6):677 – 691.

23

A AppendixA.1 General InstructionsWelcome, Thank you for participation in this experimental study!

This experiment will not be particularly difficult or involve trick questions.You will simply need to follow the instructions as they gradually appear onyour screen. The answers you will provide will be confidential. During theexperiment, you will be asked to make choices. It is therefore important forthe success of the experiment that you do not talk to each other and that youread the instructions very carefully. If you have questions during the experiment,please raise your hand. At the end of the experiment, you will receive a payment.The actual amount will depend partly on your choices and partly on the choicesof the other participants. If needed, you can use area below for calculations.

100 ECU = 10 EURGood Luck!

A.1.1 PGG Instructions

You will solve two different tasks; as a result you will receive 50 ECUs as acompensation for your efforts for the tasks. In the next step all participants willbe ordered as three member groups. Each player then makes a decision howmuch of own endowment (from 0 to 50) to contribute to common account. Youcan think of the contributions to the common account as an investment in acommon project. If any positive amount is contributed to the common accountit will be doubled and redistributed among group members. YOUWILL SOLVETWO TASKS; WILL RECEIVE 50 ECUs TWICE AND WILL PLAY IN THEGROUP TWICE, HOWEVER ONLY ONE OF THESE DECISIONS IS PAY-OFF RELEVANT. Information about the results both of these games will beprovided in the end of the second game. Your total pay-off will be calculatedas follows:

Total Profit = ( 50 - your contribution to the project ) + 2 * (Total contribution to the project ) /3

Below you can try out how the profit after hypothetical contributions canbe distributed.

Insert contributions in the gaps and click ’Aktualisieren’. If you have under-stood everything well, please click ’OK’ button.

24

Instructions and PGG calculator

A.2 Real effort tasks

Example of low routine effort tasks

25

Example of low creative effort tasks

A.3 Contributions: ANOVA and Boxplot

26

A.4 Correlation Big five and SVO

extra. agree.R consci. emot.R open. extra.R agree. consci.R emot. open.Rextra.

agree.R 0.02consci. 0.19* -0.09emot.R -0.30*** -0.06 -0.02open. 0.31*** -0.06 0.24** -0.15

extra.R -0.58*** -0.18* -0.02 0.26** -0.05agree. 0.22* -0.15 0.15 0.00 0.23* -0.06

consci.R -0.18* 0.13 -0.64*** 0.06 -0.08 0.12 -0.12emot. 0.34*** -0.11 0.21* -0.37*** 0.17 -0.11 0.21* -0.19*

open.R -0.14 -0.07 -0.02 0.02 -0.22* 0.04 -0.30*** 0.14 -0.08SVO -0.07 -0.10 -0.13 0.07 -0.02 0.14 0.08 0.14 -0.09 -0.01Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01

27

A.5 Regression Results: Tobit models

Dependent variable:Contribution R. task

(1) (2) (3)svo_angle 0.408∗∗ 0.363∗∗ 0.359∗∗

(0.163) (0.165) (0.165)extra1 2.843∗∗

(1.397)agree2 -3.316∗∗

(1.570)consci1 -1.181

(1.503)emot2 0.181

(1.423)open1 -0.124

(1.645)extra2 -0.666

(1.229)agree1 0.293

(1.298)consci2 1.472

(1.329)emot1 -1.573

(1.325)open2 -1.054

(1.279)edu 0.411 0.604 0.510

(0.621) (0.631) (0.634)age 0.896 0.684 0.659

(0.745) (0.773) (0.756)gender -1.484 -1.392 0.212

(3.992) (4.147) (3.919)prev.particip 3.483 4.379 3.398

(4.509) (4.688) (4.658)sequence -3.907 -4.451 -3.707

(3.811) (3.846) (3.864)EFF_level -1.394 -3.106 -4.070

(3.896) (3.900) (3.849)Constant 2.202 1.815 -2.357

(20.690) (20.053) (18.163)Observations 123 123 123Log Likelihood -428.701 -430.803 -432.709Wald Test 18.813∗ (df = 12) 14.306 (df = 12) 10.408 (df = 7)

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01

28

Dependent variable:Contribution C. task

(1) (2) (3)svo_angle 0.377∗∗ 0.295

(0.190) (0.192)extra1 2.767∗

(1.640)agree2 -4.550∗∗

(1.826)consci1 -0.342

(1.749)emot2 0.707

(1.662)open1 0.179

(1.921)extra2 0.154

(1.448)agree1 -1.017

(1.532)consci2 1.103

(1.563)emot1 -2.121

(1.573)open2 -1.945

(1.531)edu 0.176 0.572 0.376

(0.724) (0.744) (0.741)age 0.099 -0.189 -0.124

(0.864) (0.906) (0.879)gender -0.728 0.211 1.184

(4.669) (4.906) (4.580)prev.particip 7.894 9.754∗ 7.801

(5.256) (5.518) (5.431)sequence 0.495 1.226 0.880

(4.452) (4.502) (4.505)EFF_level 3.268 0.777 0.821

(4.562) (4.599) (4.494)Constant 12.759 29.028 9.580

(24.193) (23.282) (21.165)Observations 123 123 123Log Likelihood -415.706 -419.055 -419.853Wald Test 14.261 (df = 12) 7.354 (df = 11) 5.857 (df = 7)

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01

29

A.6 Regressions with dichotomized dependent variableIf contribution equals 0 then 0, else 1. Applied model: Logit.


(1) (2) (3)svo_angle 0.067∗∗ 0.072∗∗ 0.057∗∗

(0.030) (0.031) (0.027)extra1 0.344

(0.289)agree2 -0.345

(0.331)consci1 -0.438

(0.409)emot2 0.026

(0.244)open1 -0.310

(0.344)extra2 -0.160

(0.239)agree1 0.165

(0.237)consci2 0.242

(0.264)emot1 -0.226

(0.216)open2 -0.038

(0.231)edu 0.012 0.023 -0.007

(0.118) (0.109) (0.102)age -0.153 -0.203 -0.159

(0.147) (0.145) (0.133)gender -0.887 -0.866 -0.496

(0.782) (0.777) (0.686)prev.particip -0.397 -0.040 -0.062

(0.871) (0.861) (0.791)sequence 0.951 0.983 1.008

(0.770) (0.789) (0.735)EFF_level -1.210 -1.408 -1.571∗

(0.912) (0.879) (0.851)Constant 10.032∗∗ 6.663∗ 5.713∗

(4.802) (3.782) (3.268)Observations 123 123 123Log Likelihood -29.112 -30.410 -31.831Akaike Inf. Crit. 84.224 86.820 79.663

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01

30


(1) (2) (3)svo_angle 0.048∗ 0.040∗

(0.025) (0.023)extra1 0.325

(0.229)agree2 -0.318

(0.268)consci1 -0.234

(0.269)emot2 0.109

(0.210)open1 -0.140

(0.256)extra2 -0.141

(0.191)agree1 -0.138

(0.207)consci2 0.209

(0.214)emot1 -0.159

(0.197)open2 -0.270

(0.179)edu -0.118 -0.078 -0.103

(0.097) (0.093) (0.088)age -0.109 -0.134 -0.138

(0.119) (0.120) (0.114)gender -0.704 -0.684 -0.518

(0.625) (0.637) (0.577)prev.particip 0.175 0.509 0.248

(0.671) (0.687) (0.652)sequence 1.167∗ 1.216∗ 1.171∗

(0.677) (0.645) (0.642)EFF_level 0.859 0.421 0.345

(0.683) (0.617) (0.588)Constant 5.413 6.866∗∗ 4.390


Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01

31

A.7 Regressions with dichotomized dependent variableIf contribution is less than mean then, 0, else 1. Applied model: Logit.


(1) (2) (3)svo_angle 0.039∗∗ 0.031∗ 0.029∗

(0.018) (0.017) (0.017)extra1 0.182

(0.151)agree2 -0.332∗

(0.171)consci1 -0.006

(0.158)emot2 0.186

(0.156)open1 0.043

(0.179)extra2 -0.072

(0.126)agree1 0.029

(0.135)consci2 0.137

(0.138)emot1 -0.153

(0.140)open2 -0.212

(0.135)edu 0.044 0.080 0.062

(0.068) (0.066) (0.064)age 0.164∗∗ 0.148∗ 0.146∗

(0.082) (0.081) (0.078)gender -0.412 -0.306 -0.203

(0.427) (0.430) (0.391)prev.particip 0.090 0.249 0.094

(0.482) (0.490) (0.466)sequence -0.361 -0.347 -0.234

(0.408) (0.400) (0.386)EFF_level 0.562 0.384 0.323

(0.423) (0.403) (0.386)Constant -5.354∗∗ -4.077∗ -4.712∗∗


Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01

32


(1) (2) (3)svo_angle 0.039∗∗ 0.031∗ 0.029∗

(0.018) (0.017) (0.017)extra1 0.182

(0.151)agree2 -0.332∗

(0.171)consci1 -0.006

(0.158)emot2 0.186

(0.156)open1 0.043

(0.179)extra2 -0.072

(0.126)agree1 0.029

(0.135)consci2 0.137

(0.138)emot1 -0.153

(0.140)open2 -0.212

(0.135)edu 0.044 0.080 0.062

(0.068) (0.066) (0.064)age 0.164∗∗ 0.148∗ 0.146∗

(0.082) (0.081) (0.078)gender -0.412 -0.306 -0.203

(0.427) (0.430) (0.391)prev.particip 0.090 0.249 0.094

(0.482) (0.490) (0.466)sequence -0.361 -0.347 -0.234

(0.408) (0.400) (0.386)EFF_level 0.562 0.384 0.323

(0.423) (0.403) (0.386)Constant -5.354∗∗ -4.077∗ -4.712∗∗


Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01

33

A.8 Regressions with Effort level and Effort type dum-mies

Including only SVO in the main regression. Applied model: Tobit.

Dependent variable:Contributions Contribution R. task Contribution C. task

(1) (2) (3)C.eff.dummy -0.548

(3.001)EFF_level -4.450 1.111

(3.913) (4.509)svo_angle 0.356∗∗ 0.351∗

(0.164) (0.189)Constant 25.333∗∗∗ 17.773∗∗∗ 14.439∗∗

(2.118) (5.550) (6.396)Observations 246 123 123Log Likelihood -861.089 -434.619 -421.043Wald Test 0.033 (df = 1) 6.341∗∗ (df = 2) 3.449 (df = 2)

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01

If contribution equals 0 then 0, else 1. Applied model: Logit.


(1) (2) (3)C.eff.dummy -0.324

(0.405)EFF_type -1.407∗ 0.214

(0.812) (0.369)svo_angle 0.062∗∗ 0.025

(0.025) (0.015)Constant 2.225∗∗∗ 1.726∗ -0.810


Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01

34

If contribution is less than mean then, 0, else 1. Applied model: Logit.


(1) (2) (3)C.eff.dummy 0.000

(0.255)EFF_type 0.214 0.214

(0.369) (0.369)svo_angle 0.025 0.025

(0.015) (0.015)Constant 0.016 -0.810 -0.810


Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01

35

A.9 Regressions with Interactions:Tobit model


(1) (2) (3)svo_angle 0.224 0.018 0.110

(0.259) (0.273) (0.255)EFF_level 39.903 27.607 35.969

(41.382) (40.986) (36.628)extra1 4.069∗

(2.376)agree2 -5.624∗∗

(2.689)consci1 -0.741

(1.924)emot2 1.501

(2.575)open1 -2.316

(2.857)extra2 -0.157

(1.886)agree1 1.220

(2.099)consci2 2.765

(1.945)emot1 -3.690

(2.479)open2 -2.379

(1.991)edu 1.257 1.760∗ 1.743∗

(0.900) (0.922) (0.904)age 1.814∗ 1.236 1.321

(1.053) (1.126) (1.066)gender -3.277 -5.701 -2.278

(6.303) (6.670) (5.782)prev.particip 11.548∗ 14.749∗∗ 11.519∗

(6.723) (6.982) (6.850)sequence -0.611 -0.039 -0.549

(5.912) (6.042) (5.838)svo_angle:EFF_level 0.374 0.551 0.443

(0.330) (0.342) (0.331)EFF_level:extra1 -1.495

(3.120)EFF_level:agree2 3.096

(3.398)EFF_level:consci1 -1.473

(3.085)EFF_level:emot2 -1.123

(3.171)EFF_level:open1 2.807

(3.614)EFF_level:extra2 -0.496

(2.453)EFF_level:agree1 -1.384


(2.710)EFF_level:emot1 3.189

(2.936)EFF_level:open2 1.606

(2.598)EFF_level:edu -1.222 -1.919 -2.020

(1.265) (1.284) (1.271)EFF_level:age -1.741 -0.991 -1.139

(1.501) (1.584) (1.535)EFF_level:gender 2.827 7.226 5.093

(8.261) (8.562) (7.830)EFF_level:prev.particip -13.284 -16.516∗ -12.998

(9.059) (9.243) (9.213)EFF_level:sequence -9.306 -9.442 -7.876

(7.744) (7.948) (7.692)Constant -17.543 -16.296 -26.209

(29.139) (30.323) (26.903)Observations 123 123 123Log Likelihood -423.569 -424.926 -428.319Wald Test 30.042 (df = 23) 26.675 (df = 23) 19.578 (df = 13)

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01

36


(1) (2) (3)svo_angle 0.081 -0.302 -0.048

(0.295) (0.314) (0.294)EFF_level 41.697 -18.686 16.516

(46.695) (45.891) (41.705)extra1 5.602∗∗

(2.719)agree2 -6.187∗∗

(2.984)consci1 -1.406

(2.179)emot2 2.775

(2.890)open1 0.983

(3.226)extra2 1.475

(2.165)agree1 0.182

(2.364)consci2 3.089

(2.222)emot1 -6.340∗∗

(2.869)open2 -7.214∗∗∗

(2.398)edu 0.071 1.295 0.907

(1.020) (1.030) (1.033)age -0.034 -1.092 -0.440

(1.184) (1.249) (1.216)gender -2.855 0.134 -0.715

(7.168) (7.568) (6.645)prev.particip 14.829∗ 21.022∗∗∗ 15.781∗∗

(7.623) (7.927) (7.829)sequence 16.658∗∗ 15.368∗∗ 15.474∗∗

(6.732) (6.791) (6.698)svo_angle:EFF_level 0.465 0.799∗∗ 0.536

(0.372) (0.387) (0.377)EFF_level:extra1 -3.075

(3.538)EFF_level:agree2 2.689

(3.789)EFF_level:consci1 0.626

(3.454)EFF_level:emot2 -2.297

(3.563)EFF_level:open1 -1.253

(4.073)EFF_level:extra2 -2.111

(2.777)EFF_level:agree1 -0.540


(3.036)EFF_level:emot1 5.363

(3.358)EFF_level:open2 7.346∗∗

(3.034)EFF_level:edu 0.172 -1.050 -0.850

(1.426) (1.427) (1.446)EFF_level:age -0.561 0.514 -0.044

(1.683) (1.752) (1.742)EFF_level:gender 2.091 0.836 3.147

(9.345) (9.628) (8.949)EFF_level:prev.particip -14.931 -21.379∗∗ -15.869

(10.180) (10.328) (10.458)EFF_level:sequence -28.719∗∗∗ -26.491∗∗∗ -26.556∗∗∗

(8.788) (8.880) (8.800)Constant -0.835 51.965 10.642

(33.039) (34.089) (30.596)Observations 123 123 123Log Likelihood -407.196 -407.124 -413.723Wald Test 32.352∗ (df = 23) 31.526 (df = 23) 18.671 (df = 13)

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01

37

Documents

Creativity and Cooperation in Standard Public Goods Experiment