14th Australia New Zealand Workshop in Experimental Economics€¦ · Effects of Individuals' Characteristics on Trading Strategies in an Asset Market Experiment Stephen L. Cheung

14th Australia New Zealand Workshop in Experimental Economics

PAPER ABSTRACTS

Day 1: Monday 25 November 2019 Session 1A – Market 1 Market Participation: Theory and Evidence Michael Shin University of Sydney Recent evidence suggests subjective returns are positively correlated with stock market participation. Furthermore, there is strong evidence that stock market experiences, i.e. realized returns, impact subjective returns. I bring a model into the laboratory and find that experience-based subjective returns can explain limited participation. Stock market participation is increasing in both subjective returns and past realized returns. I find micro-evidence that “learning from experience” generates heterogeneity in subjective returns, where subjects who experience low returns have lower subjective returns than subjects who experience high returns. In particular, subjects over-weigh the price trend when experiencing low returns and are more likely to believe that prices are trending downward when they receive a low return. Effects of Individuals' Characteristics on Trading Strategies in an Asset Market Experiment Stephen L. Cheung University of Sydney (Guanghan (Ryan) Zhang) We match subjects trading strategies in an asset market experiment to extensive data on their psychometric characteristics to investigate which characteristics are most important in explaining the (within-market) variation in the propensity to adopt different trading strategies. We find the effect of cognitive ability on trading strategy to be complex and multifaceted, and report first results for Locus of Control in an experimental finance setting. Of the “big five” personality traits, only neuroticism appears to be important. We also find little support for recent interest in Theory of Mind as measured by the “eyes test”, possibly because “non-analytical” forms of intelligence are better captured by other characteristics in our data. On the Efficiency and Risk Attitudes in Dynamic Portfolio Choices Olga Rud RMIT University (Jacopo Magnani, Jean Paul Rabanal and Yabin Wang) Using a within treatment design, we investigate the efficiency of subject investment decisions and risk attitudes across different environments. We apply the Payoff Distribution Pricing Model (PDPM) by Dybvig (1988) to study the efficiency of portfolio choices in a discrete dynamic portfolio environment. In our environment, the asset price follows a binomial process, and the subject payoff depends on the allocation of their wealth across the risky and/or the risk-free asset. These investment decisions can result in either 8 or 16 equally likely terminal wealth outcomes. We find that most subjects chose an efficient portfolio. Furthermore, our design

allows us to study risk-attitudes. Our results suggest that subjects may be more risk tolerant in the dynamic task. Incentive Schemes, Framing, and Market Behaviour: Evidence from an Asset-Market Experiment Nick Feltovich Monash University (Xuegang Cui and Kun Zhang) We investigate the effects of the form and framing of incentives on price and trading behaviour in experimental asset markets. Subjects buy and sell a high-risk asset, a low-risk asset, and riskless cash over 10 rounds. We vary, between-subjects, the incentive scheme (relative versus absolute performance), and how the variable portion of payment is framed (bonus versus penalty). Both relative-performance (tournament) incentives and penalty framing are associated with significant increases in the price of high-risk asset, relative to either its fundamental value or to the price of the low-risk asset. Additional analysis shows significant gender differences in trading behaviour and performance, and evidence that the two are connected. Session 1B – Social Preference 1 In Search of Competitive Givers Ronald Peeters University of Otago (David Fielding and Stephen Knowles) Much of the theoretical and experimental research on charitable giving allows for three main types of donor: pure altruists, impure altruists, and pure warm-glow givers. For none of these types should donations be increasing in the amount donated by others, and the fact that some experimental participants do behave in this way suggests a fourth type, the “competitive giver”. Our experimental results suggest that most (but not all) competitive giving is a result of uncertainty about the social norm: when information about the norm is revealed, the incidence of competitive giving is much lower. Altruism Begets Altruism: Nudging our Way to a More Virtuous Society Stephanie A. Heger University of Sydney (Robert Slonim) Economic research examining social preferences over the past several decades has increasingly focused on better understanding and teasing apart distinct motives. While the research has primarily focused on short term behavior, this paper goes beyond this literature to better understand how a short-term intervention, a nudge, can affect subsequent behavior. Using a popular policy nudge, the default option, we show that its effect on the choice to be more altruistic “today” causes an increase in altruism “tomorrow”. We rule out that the nudge has a direct inter-temporal effect and instead show that our findings are consistent with a model of habit formation and moral consistency; that is, altruism begets altruism. Our local average treatment effect indicates that the nudge-induced giving in Round 1 of the experiment causes

a 40 percentage point (or 200%) increase in the propensity to give in Round 2. Our findings suggest a way forward in promoting a more virtuous society. Partial Promises and Strategic Moves – An Experimental Study Rochelle Wilson RMIT University A promise is a common and time-honoured form of communication. Traditionally, communication has been viewed in economics as cheap talk. However, recent studies have demonstrated that individuals consider promises as credible signals of intended behaviour. Promises enable a promisor to send signals that generate trust and increase levels of cooperation during strategic interactions. Studies provide evidence that pre-play messages in the form of promises considerably enhance trust and subsequent levels of cooperation. It is demonstrated that a promisee’s expectations are raised by a promise, and aversion to disappointing these expectations leads the promisor to behave in a trustworthy manner. However, overpromising has become common. Individuals are more likely to break, and less likely to make or believe promises. Trust therefore, is fragile. Yet, oftentimes we make promises that we either have no intention of keeping, or little probability of fulfilling. Other times, unforeseen circumstances may arise that raises the costs of making good on a promise. Third parties can observe our track record of making good on our promises, which undermines or enhances our credibility in future interactions. For this reason, there is a cost to breaking a promise in terms of being perceived as untrustworthy in the future. In addition, there are psychological costs of breaking promises like negative emotions of guilt. Given the downsides of breaking promises, this research examines a new type of promise i.e., partial promise. Here the promisor signals an intention to carry out an action, albeit with less than certainty. They are considered as escape routes that provide wriggle room for individuals who are doubtful about fulfilling their promise or have no intentions of doing so. Partial promise might be an effective way of promising responsibly. This behavioural and experimental economics project uses two player sequential binary trust games to provide insight into promises. It consists of three studies, starting with a feasibility analysis followed by two economic experiments. Since partial promise is a relatively new concept we begin with a feasibility analysis that uses an anonymous online survey to gauge the opinions, perceptions and elicit norms of regular individuals regarding partial promises. The second study is an online experiment conducted using the strategy technique where individuals play the trust game. Finally, the third study is a controlled laboratory environment using two player binary sequential trust games with communication and a random device to demonstrate if partial promises can achieve better payoffs for people individually and together. We look to provide an efficient alternative to promising that provide better payoffs and lower punishments to all parties involved. The overall research aim is to find out if partial promises can help build, sustain, and promote cooperation, while explicitly managing expectations of individuals during strategic games when other mechanisms like full promises fail. The Role of Expectations in Repeated Gift Exchange Interactions Vera te Velde University of Queensland (Rosario Macera)

In Macera and te Velde (working paper) we develop a theory of reference-dependent reciprocity in the context of gift exchange. This theory builds on two well-established psychological insights---the importance of expectations in shaping behavior, and the adaptation to constant stimuli---to deliver four main messages about the efficacy and efficiency of gift exchange: (1) gifts are more powerful when they surprise workers; (2) the power of surprising gifts, however, wanes over time; (3) gifts are cursed, as once a firm grants one, it should grant it forever; and therefore (4) gifts are most likely profitable in short-term interactions. While these predictions explain some puzzling findings in existing field experimental studies of gift exchange, such as the waning of reciprocity over time after the gift is granted and the benefit of granting gifts surprisingly, competing models could potentially explain the same phenomena. However, our model makes clear predictions about repeated interactions between employer and employee that have not been tested. These primarily pertain to the long-run effects of granting an initial gift: if a gift reveals the employers ability and willingness to grant gifts, the worker will infer the likelihood of future gifts, which then undermines the ability to grant profitable gifts by taking away the surprising nature. And because of this difficulty, gifts are more effective in short-term interactions between employer and employee. We test these predictoins in a new laboratory gift exchange experiment in which we allow repeated interactions between the same worker and employer (for 25 rounds) without introducing repuational concerns that provide well-understood motivations for reciprocity. We do this by having employers pre-commit to a series of 25 wages, which they choose from a limited menu that we provide. Workers then learn their wages on a period-by-period basis and choose an effort level in response. By comparing effort across time in situations with no gift exchange (i.e. a constant wage), a single instance of gift exchange, a permanent wage raise part way through the 25 periods, or with a high or low probability of receiving a gift in each period, we are able to examine the long-run effect of gifts on expectations and on later effort. We find that while effort certainly correlates with wages, gift exchange exhibits decreasing power over time. This confirms the key predictions of the model. Session 1C – Political Economy Legislative Bargaining: An Experimental Study David (Enseen) Tang University of Technology Sydney I experimentally study the asymmetric discounting and recognition probability of the players in legislative bargaining model, according to Copiˇc’s (2016). In their paper, there is an unique equilibrium ˇ prediction about the division in which the average discount factor is determined by the average interest rate. However, the asymmetric recognition probability is shown to have no effect. Therefore, our first treatment would be lowering one of the three player’s discount factors. The second treatment would be making the asymmetric player’s discount factor even lower so that it violates the equilibrium condition. The last treatment would be modifying the uniform recognition and make one player more likely to become the proposer. Political Gaming: Using Economic Games to Study the Foundations of Political Ideology Scott Claessens University of Auckland

(Chris Sibley, Ananish Chaudhuri and Quentin Atkinson) Politics is the process of navigating the conflicts of interest inherent to human group living. Issues like taxation, welfare, environmentalism, and criminal justice centre on the social dilemmas that naturally arise from group living, such as resource distribution, inequality, and social sanctioning. These parallel the basic social preferences studied in behavioural economics. Using data from the longitudinal New Zealand Attitudes and Values Study (n=926), we explored the relationships between political ideology, policy views, and social preferences across a comprehensive battery of incentivised one-shot economic games measuring fairness, altruism, cooperation, coordination, second-party punishment and third-party punishment. Replicating previous work, we found that social preferences in these games could be explained by two factors: cooperativeness and punitiveness. The cooperativeness factor was negatively related to economic conservatism, and positively related to a host of economically liberal views, including support for income redistribution, support for unemployment and sole parent benefits, support for taxes on the rich, and willingness to make sacrifices for the environment. In contrast, punitiveness did not relate to political ideology or policy views. Overall, these findings highlight the importance of basic social preferences in understanding the political disagreements that arise from group living. Who Wants to Be a Politician? Pushkar Maitra Monash University (Ananish Chaudhuri, Vegard Iversen and Francesca R. Jensenius) The attitudes of the individuals who enter political office is central to what policies are made and how these policies are implemented. This study is motivated by two primary questions. First, what are the attributes of those who choose to seek political office as opposed to regular members of the citizenry? Second, does length of tenure in political office tend to make people more dishonest? The negative selection hypothesis would suggest that it is only those that are inherently dishonest seek to enter political office. Here, socialization within political institutions plays a smaller role. Alternatively, the positive selection hypothesis would posit that those who enter into politics are more idealistic than ordinary citizens and genuinely wish to make a difference. But over time, and with greater exposure, to politics, they become more cynical or self-serving. We use survey and experimental data from politicians (newly elected and experienced) and regular citizens from rural West Bengal in India to address these questions. We find evidence in favour of the positive selection hypothesis. Compared to regular citizens, neophyte politicians are more honest, trusting, fair-minded and express much higher faith in political institutions and processes. We also find evidence that some of this idealism wears off with time in office so that experienced politicians are more cynical and less honest than incoming ones. Does Contact Help to Re-establish Cooperation and Trust after Violent Conflict? Testing Different Types of Contact in a Lab-in-the-Field Experiment in Indonesia Swee Hoon Chuah University of Tasmania (Katharina Werner) After violent conflicts, encounter programs are a commonly used intervention, assuming that contact helps to re-establish cooperation across groups by fostering forgiveness, mutual understanding and trust. Such encounters often combine different types of contact between

group members, ranging from mere meetings over talking about the conflict to joint projects with a common goal. Although the programs require a lot of resources, the efficacy of contact after conflict lacks empirical support. We exogenously induce contact between Muslim and Christian students who otherwise have little contact, both in a city with a history of violent religious conflict and a peaceful region in Indonesia. In different treatments, we vary the type of contact, before measuring its effect on cooperation and trust across groups in a Multilevel Public Goods Game and a trust elicitation task. The results suggest that contact is more efficient in the peaceful region and that the efficacy of the different interventions depends on participants’ individual conflict exposure. Free communication and a joint task enhance interreligious cooperation for certain participants, but only free communication increases trust towards other out-group members. Talking about the history of conflict can, in some cases, even be detrimental to cooperation. Different forms of contact affect attitudes like animosity, forgiveness and empathy differently. Session 1D – Risk 1 Sports Betting and Overconfidence: Evidence from a Natural Experiment in Kenya Matthew Olckers Monash University (Joshua E. Blumenstock) Sports betting is one of the fastest-growing forms of gambling in the world. In Kenya, the largest sports betting operator—SportPesa—is the most popular internet search query in the country. We study if gamblers participate rationally or if participation is driven by a behavioral bias. We use betting transaction data from more than 6 000 individuals to argue that sports betting is driven by overconfidence. If a gambler treats some “near misses" as wins, he can become overconfident in his ability to predict the outcome of matches correctly. A rational gambler should respond to both negative and positive signals of his ability whereas an overconfident gambler will be less responsive to negative signals relative to positive signals. SportPesa’s weekend jackpot provides a natural experiment to test our hypothesis. Each week 17 football matches are selected by SportPesa for the jackpot. A gambler only wins a prize if he predicts 12 or more matches correctly. Variation in the number of correct predictions below 12 provides a signal of ability without any income effect. We find that gamblers who predict a high number of matches correctly are more likely to bet the following week whereas gamblers who predict a low number of matches correctly do not display statistically different behavior from the base category. Socially Embedded Risk Taking Daniel John Zizzo University of Queensland (Alexandros Karakostas and Giles Morgan) This paper reports the results of an experiment on how the social environment affects individual risk taking. Subjects repeatedly play an individual investment task by which they choose how much to invest in a more than actuarially fair lottery. Specifically, there is a 50% chance the amount a participant invests will be tripled, or a 50% chance the investment will be lost. We

use a 2 x 2 factorial design by which we vary: (i) whether the subjects initially observed high or low investment social prompts, based on previous investment decisions by a different sample of subjects; (ii) whether social group feedback is provided after each round. Risk taking is a function of social group feedback. Initial social prompts shape initial investment levels, leading to mean investment levels that are double in size in the high prompt treatment relative to the low prompt treatment. If and only if ongoing social feedback is provided, social learning takes place leading the low prompt subjects to tend to converge with time to the mean level of investment of high prompt subjects. We find however significant group heterogeneity in mean investment levels. Investment tends to be clustered by social group. Equilibrium Play in Experimental Parimutuel Betting Markets Joshua Miller University of Melbourne (Martin Dufwenberg) Parimutuel wagering markets have attractive features, both as test beds for financial market theory, and as mechanisms for aggregating information. We design a laboratory experiment to test theoretical predictions of behavior in parimutuel betting markets. Our results are largely consistent with Bayesian Nash equilibrium play and shed light on the question of origin of the well-known favorite-longshot bias. Our results suggest that mechanisms which lead to heterogeneous private beliefs are behind the favorite-longshot bias, rather than the presence of probability distortions or risk-loving preferences. Inequality Aversion or Risk Attitude? Experimental results and a New Model David Butler Griffith University We develop a new experimental design to identify the distribution of inequality-aversion (IA) and risk-aversion parameters. Using four sets of popular 2x2 games, we use a choice list task to elicit preferences first when only own-payoffs are visible (to estimate risk attitude (RA) over own-payoffs) and then with both player’s payoffs visible (to estimate IA parameters). The distribution of IA parameters assuming risk neutrality, as in the IA model, shows every subject to be averse to disadvantageous inequality. When we include measured RA however, we find every subject favours disadvantageous inequality in these games. We then consider other structural models that allow for non-risk neutrality over others' payoffs and discuss the implications for future models. Session 2A – Field Experiments Payments for Multiple Ecosystems Services: A Framed Field Experiment in Oaxaca, Mexico Pallab Mozumder Florida International University (Eric Van Vleet and David Bray) Mexico has one of the most comprehensive PES (Payment for Ecosystem Services) programs globally. It is trying to reduce deforestation by placing forests on communal lands under PES contracts and has experimented with a carbon forestry program. Mexico is also working closely

with the World Bank’s Forest Carbon Partnership Facility (FCPF) to take advantage of REDD+ initiative. During the formation of Mexico’s PES program, planners debated whether or not to select participants and communities through conservation auctions or flat-rate payment schemes. Given that the success of PES programs critically relies on the underlying institutions and incentive structures, in this study we report findings from a field experiment conducted in four communities in the Sierra Norte of Oaxaca, Mexico to examine the potential efficacy of conservation auction mechanism for four different PES programs (Mexico’s existing payment for hydrological services, Nicaragua’s silvopastoral program, Indonesia’s program for erosion control on coffee farms and a voluntary carbon forestry program similar to those in Chiapas or Oaxaca in Mexico). We randomly selected participants and allowed them to communicate with peers in some sessions (and not in some other sessions) for each auction. All participants stated their willingness to accept (WTA) bid between 100-1600 pesos for a range of PES programs. Our findings suggest that on average participants demand a higher willingness to accept (WTA) than the current payment amounts offered in the Mexico’s PES program. Results also suggest that participants understood the different PES program requirements and significant differences in WTA bid exist across different PES programs (e.g. erosion control, silvopastoral and carbon forestry practices). We also find that communication matters in shaping WTA bid values in these conservation auctions. Furthermore, under conservation auction mechanisms, Mexico’s existing PES program will be more expensive compared to its current flat-rate payment regime. Results also suggest that communities are interested in a more diverse portfolio of PES programs and that forest conservation policy could respond to broadening the number and type of PES programs offered to these communities. Identifying the Actual Willingness to Reduce Overfishing with a Fishnet Auction: Evidence from Artisanal Fishing Villages in Brazil Carina Cavalcanti Griffith University In this field study we implemented a fishnet exchange program to investigate the actual willingness of fishermen to reduce overfishing in Brazilian artisanal fishing villages. Fishermen who use fishnets with small mesh sizes were invited to take part in an auction to exchange their fishnet for a fishnet with a bigger mesh size. Interestingly, we observe that the majority of fishermen are willing to exchange their fishnets without further compensation. Importantly, we observe that environmental perceptions and experiences with fishnets help explaining the heterogeneity in bids. These findings provide useful information about limitations and possibilities of changing the behavior of individuals who strongly exploit resources. Cash in Hand and Savings Decisions Lisa Spantig University of Essex Cash is an important means of transaction, generally assumed to be fungible. However, behavioral economics and consumer research show that ‘cash in hand’, physically holding on to cash and then handing it away, affects purchasing decisions. I study how cash in hand influences decisions in a different but very important domain: savings. Savings accounts are a promising tool for reducing poverty, but the use of savings accounts is often puzzlingly low. Holding on to cash that needs to be physically deposited into a savings account may increase the psychological costs of saving. This study experimentally identifies the causal effect of cash

in hand on savings deposits of microfinance clients in the Philippines. In contrast to many laboratory and several field studies with similar interventions, I do not find reduced savings deposits due to cash in hand. I discuss reasons for and consequence of this surprising finding, in particular for developing economics where lots of transactions are still cash-based. A Nudge to Remember: Evidence from a Field Experiment with Pawn Shop Customers Kartika Sari Juniwaty Centre for International Forestry Research This study reports the results of a field experiment with the state-owned pawnshop (PT. Pegadaian) customers in Indonesia. Established in 1901, PT. Pegadaian is a well-established financial institution with 4,221 outlets across the country that provides fast microloans based on physical collateral. When a loan reaches the due date, customers have to redeem their collateral or pay the interest to extend the loan period; otherwise, after few days of grace period, the pawned items will be auctioned. The experiment was designed to evaluate the impact of a text message reminder to the settlement of customers’ pawned items before the due date. As part of the experiment, 1,020 customers from a branch in Jakarta were randomly assigned into either a treatment or a control group. Those who were assigned into the treatment group received a text message to remind them about the due date of the loan, while those who were part of the control group received no message. This study finds that receiving a text message reminder increases the proportion of borrowers settling their loan arrangement at the pawnshop before the due date by 6.9 percentage points. Furthermore, there is a gender difference in the effect of the reminder, with women being much more responsive to the text reminder than men. Session 2B – Gender Are Women Evaluated Differently for Equal Work? Role of Attribution Biases and Financial Sanctions Nisvan Erkal University of Melbourne (Lata Gangadharan and Boon Han Koh) Abstract: This paper studies how gender distorts perceptions of outcomes in risky environments. Decision makers make costly and unobserved effort choices. Outcomes are determined by a combination of their choices and luck. Using experimental methodology, we investigate how the evaluation of outcomes is affected by the gender of the decision makers and the gender of the evaluators. Are failures or successes of female decision makers evaluated differently from those of male decision makers? We find that the answer to this question depends on the incentives faced by the decision makers. We consider decision environments with sanctions, where evaluators have the power to adjust the earnings of the decision makers, and decision environments without sanctions, where they do not. The findings from our study contribute to our understanding of the factors that may be driving gender gaps in leadership or performance pay. Our results have implications for the structure of performance review teams. Gender Differences in Altruism and Reciprocity in a Developing Country Ananta Neelim RMIT University (Pushkar Maitra and Chau Tran)

This paper examines gender differences in reciprocity and altruism in a developing country. Using a gift exchange experiment, we generate a simulated workplace where managers offer workers a wage and workers choose a level of effort. We do not find any evidence of gender differences in wages offered by male and female workers and received by male and female workers. However, conditional on the wage received, female workers exert less effort than males. We also find that women are significantly less altruistic than men in a charitable dictator game. Our results that significant implications for labour market outcomes in developing countries. Fame, What’s Your Name? An Experiment on Gender Discrimination in Artwork Valuation Bronwyn Coate RMIT University (Robert Hoffmann) Gender discrimination has been widely explored within economics. Within this fruitful area of research, experimental methods have also been applied to show that discrimination exists (e.g. Booth and Leigh, 2010). Yet within the field of cultural economics scarce research exists that has specifically explored issues of gender discrimination, and of that which can be found empirical findings have relied upon market data based on art prices where evidence across different samples have is mixed (Cameron et al 2019, Bocart et al 2018, Adams et al 2017). We add to the literature by examining gender discrimination in the valuation of artworks using an experimental design. We report five experimental studies to examine both the extent and motivation of economic gender discrimination in artwork valuation. Specifically, we address whether people discriminate in their valuation of artworks on the basis of the gender associated with an artworks creator. We do this using a sample of 1,112 participants who chose between female and male-originated artworks with and without artist information based on the name of the artist. Gender-specific artist names do not affect aesthetic judgments but cause significant swings towards male-originated artworks when participants are incentivised to identify the more famous or the more expensive artwork. Artist fame also causes a male swing when participants guess others' aesthetic judgments. We conclude that gender discrimination in art is rooted in social and institutional rather than individual aesthetic processes. Cooperation, Punishment, and Endogenous Information Manipulation Christine Grimm Vienna University of Economics and Business A large stream of experimental literature on cooperation in public good games with punishment shows a positive effect of a punishment option on cooperation (CHAUDHURI 2011 for an overview). Several papers extending the original setup by FEHR and GÄCHTER (2000) identified the importance of various factors for punishment increasing social welfare. One is the information subjects receive. It has been shown to be very important for punishment to be effective. For example, CHEUNG (2014) points out that information on contributions significantly shapes punishment and contribution decisions. Similar, NIKIFORAKIS (2010) identifies a positive impact of contribution information on efficiency. Additionally, he shows that information on earnings lower the effectiveness of punishment. GRECHENIG, NICKLISCH and THÖNI (2010) and AMBRUS and GREINER (2012) demonstrate that adding some uncertainty to displayed contribution records can decrease welfare. Subjects, who

anticipate a weakening of the link between low contributions and received punishment, reduce their provisions to the public good. The existing experimental literature on uncertainty in information transmission in cooperative contexts (e.g. CASON KHAN 1999; DAL BÓ FRÉCHETTE 2018; AMBRUS GREINER 2012, GRECHENIG ET AL. 2010) mainly focuses on the reaction of information receivers on exogenously induced misinformation or uncertainty. Our study endogenizes the uncertainty by allowing subjects to manipulate contribution information. This captures the fact, that in many situations people who control the information about past behaviour have an incentive to manipulate this information in order to influence other subject’s behavior. Hence, we are interested in studying how allowing for strategic misinformation in a public good game with punishment changes punishment and contribution behaviour as well as the resulting welfare. In our experiments, participants are randomly assigned to groups of four. Each member is equipped with an endowment, which she can either keep or contribute to the public good. One group member is randomly assigned the role of a manipulator. After being informed about the contribution behaviour of all group members, the 1 Vienna University of Economics and Business, Institute for Markets and Strategy manipulator can freely alter the information about who contributed and who kept the endowment. The other group members who know that manipulation might have taken place then receive the (potentially altered) public record. Based on the provided information all players decide on who they want to punish or not. As the matching protocol is very important for the effectiveness of punishment (e.g. FEHR and GÄCHTER 2000), we employ both a stranger and partner matching. In order to be able to isolate the impact of the manipulation opportunity, we also run base treatments without manipulation options for both matching protocols. Preliminary results indicate a negligible impact of endogenous information manipulation for stranger treatments, whereas for partner treatments different patterns emerge. Some groups seem to refrain from manipulation and successfully foster efficiency. Others fail to establish an efficient norm when facing uncertainty. Even though, some manipulators make use of their option to fake a contribution, a major share can truthfully report cooperation. Session 2C – Market 2 Thar She Resurges: The Case of Assets that Lack Positive Fundamental Value Zhengyang Bao Monash University (Andreas Leibbrandt) This experimental study investigates the trading of assets that mimic the features of most cryptocurrencies. Groups of traders are randomized into asset markets where fundamental values are either positive, zero, or negative. Our findings indicate the presence of much larger bubbles in asset markets with non-positive fundamental values than in asset markets with positive fundamental values. In a further set of experiments, we observe assets with the same fundamental values but with variable dividends and potential losses. We find that bubbles are even larger in the presence of variable dividends compared to fixed dividends. We show that these findings are consistent with trader expectations but not with loss aversion and confusion. This study provides experimental evidence that supports the need for particular scrutiny of asset markets that lack positive fundamental value. The Impact of Identity Transparency on Informational Efficiency in Experimental Asset Markets

Regev Bar Deakin University In financial markets, identity, pre-trade and post-trade transparencies are said to impact informational efficiency, and various market quality measures. Prior studies report inconsistent empirical findings and theoretical predictions. Supporting these inconsistencies are the different transparency policies applied to asset markets across the world. One of the strong arguments for moving into opaque markets is the assumption that identity transparency can expose large (financially and transactionally) market participants to front-running activities, which in turn, may deter them from transacting in that market, and therefore reduce the overall market quality. To study these, we employ an experimental continuous double auction environment in the laboratory. Each group of participants consists of a fixed number of informed, uninformed, and liquidity traders. Groups are assigned to different transparency settings. Informed traders also receive higher endowments compared to uninformed participants, to simulate market conditions where wealthier participants are trying to transact in the market. In our preliminary analysis, we find that under equal endowments, informational efficiency is improved in opaque markets compared to transparent markets. However, transparent markets are more liquid. Unequal endowments do not appear to have a significant impact on informational efficiency. These findings suggest that the positive impact of opaque markets on informational efficiency comes at the expense of market liquidity. Price Bubbles and Profits in Experimental Markets: The Roles of Iterative Reasoning, Anticipatory Reasoning, and Analytic Ability Alexander Svenson University of Sydney (Pablo Guillen Alvarez and Stephen Cheung) The efficient markets model proposes that traders who are able to accurately calculate the price of assets come to dominate financial markets. Opposing views argue that human psychology exerts an influence in markets, leading to mispricing and financial bubbles. The latter views imply a profit opportunity for traders who better understand how others think. This study takes advantage of experimental market methodology to explore: i) whether traders’ use of iterative reasoning influences mispricing and bubbles, ii) the extent to which more accurately anticipating how others will think and act (anticipatory reasoning) increase market earnings, and iii) whether the association between earnings and anticipatory reasoning depends on the similarity between traders in a market. Performance on the p-beauty contest (Nagel, 1995) was used to measure iterative and anticipatory reasoning. Moreover, traders were sorted into homogenous or mixed markets based on their use of iterative reasoning. Neither in the p-beauty contest nor in markets was behaviours generally consistent with predictions reliant on iterative elimination of strategies. Markets with more iterative thinkers, or where traders were more similar, showed high mispricing. However, anticipatory reasoning did predicted higher earnings, and the association was stronger in markets where traders were more similar. Interestingly, analytic ability was associated with poorer market timing and negative interaction effects with anticipatory reasoning. Further analysis revealed a weak quadratic relationship for analytic ability on earnings. Overall, these findings point to a mechanism of selection where traders who appreciate psychological dynamics come to dominate the market, which due to greater homogeneity, is more likely to be mispriced. Our results suggest that policy tools which actively diversify market participation and expectations may support market efficiency.

Pricing Exchange Traded Funds in Laboratory Markets Jean Paul Rabanal, Monash University (John Duffy and Olga A. Rud) The impact of exchange traded funds (ETFs) for the pricing and volume of trade in the assets that underlie ETFs is not clear [Lettau and Madhavan (2018)]. We address the impact of ETFs on prices and trade in the underlying assets of the ETF using laboratory asset markets where agents can trade in multiple assets. We study two main environments, one with and one without ETFs. In the no ETF environment, agents can buy and sell two assets, A and B, that have different fundamental values and either zero or negatively correlated returns. Our ETF environment adds a third asset C (the ETF) that is a composite claim to 1 unit of asset A and 1 unit of Asset B. Agents can directly buy and sell all three assets A, B, and C in the ETF environment. They can observe prices on all three assets as well as the net asset value of the composite asset C. We find that when the returns of the underlying assets A and B are negatively correlated, the presence of the ETF asset C reduces mispricing and price volatility, relative to the environment without the ETF, and does not affect the volume of trade in these underlying assets. In the case of zero correlation in asset returns, the ETF has no impact on mispricing, price volatility or trade volume relative to the no ETF environment. Thus, on balance, we find that ETFs help or do no harm to price discovery and market liquidity. Session 2D – Risk 2 Behavioral Benefits of Credit Transfer: Evidence from Randomized Controlled Trials in Bangladesh Jinnat Ara Queensland University of Technology Risk attitude and time preference play significant roles in the economic decision-making process of nearly all individuals. A great deal of attention has been devoted to studying the effects of these preference attributes on economic outcomes (such as income, labor supply, savings, investments, borrowing), demographic and social identity outcomes (such as age, gender, race, education, cognition), major life events (such as the experience of war, job displacement, serious health condition, inflation), length of life, and planning for end of life care (Sutter et al., 2013; Huffman et al., 2017; Shapiro, 2010; Benjamin, 2010; Bommier, 2006; Sahm, 2012; Dohmen et al., 2011; Donkers et al., 2001; Tymula et al., 2012; Tanaka et al., 2010). However, most of these studies assume these preference parameters are exogenous to the outcome of interest (e.g., saving or borrowing decisions). In practice, income or wealth may alter time and risk preferences. Available evidence shows that endowment has a positive effect on the level of patience and risk aversion (Halvey, 2008; Guiso and Paiella, 2008). That is, these preference parameters are subject to income effects, and therefore endogenous to saving or borrowing decisions. Existing evidence hardly provides convincing evidence on causality between income on the one hand, and time and risk preferences on the other. This paper aims to address this gap in the literature. We use two rounds of panel data from a randomized experiment in Bangladesh conducted by BRAC's Targeting the Ultra Poor (TUP) program in 11 districts covering around 5000

households. The program provides a soft loan1 along with additional monetary transfers to the female heads of the selected poorer households to help them move out of poverty through the initiation or expansion of income-generating activities. The randomized credit shock, by design, provides an ideal opportunity to explore and quantify the causal effects of the accrual of income and wealth on time preference and risk attitudes in the monetary domain. We achieve this by embedding the experimental tools used by Bauer et al. (2012), and Sutter et al. (2013) in the household survey. Despite randomization, we estimate the effects of the intervention using difference-in-difference method. We also control for the respondents’ household characteristics to account for any compositional, income and wealth differences in the sampled households at the baseline. We find robust evidence that the program beneficiaries exhibit more patience, and risk tolerance than the non-beneficiaries. These are consistent across various tools of elicitation and measures of time and risk preferences. Beneficiaries are less likely to be present/future bias (insignificant) but with higher asset level are likely to be more future bias (p<0.10). Results suggest that these effects can only be attributed to an increase in income due to the credit transfer program. Credit programs may promote patience and risk-taking abilities that are critical for successful entrepreneurial activities. Learning About Type: The Value of Being Willing to Fail Ingrid Burfurd RMIT University Learning and decision-making are key concepts in information and behavioural economics. Using a simple model that incorporates risk aversion and cognitive sophistication, we focus on the signal extraction problem that arises when investments and type are substitutes and when subjects encounter a series of two pass/fail hurdle tasks. The first hurdle task is relatively low-stakes, while the second hurdle offers larger rewards for success but requires more costly investments in order to succeed. Stage 1 of the experiment provides subjects with an opportunity to risk failure in order learn their type. This information allows subjects to efficiently tailor their investments in Stage 2, thus maximising expected payoffs across both stages of the game. We find that optimal learning is rare. Between 14 and 28 percent of subjects gather information, depending on their level of experience. Roughly a quarter of subjects’ investments completely eliminate information. Most of these subjects make decisions that are strongly correlated with risk aversion, while other subjects are risk-tolerant but myopic. Our data suggest that individuals undervalue the information that can be gained by exposing themselves to failure, and we present a rational path by which risk-averse or myopic individuals might systematically make costly over-investments in goal-oriented tasks. Cognitive Heterogeneity and Complex Belief Elicitation Tom Wilkening University of Melbourne (Ingrid Burfurd)

1 Interest rate is low (20%) and repayment starts after two months of taking the loan. Other MFIs in Bangladesh charge 25 percent interest rates. 

Abstract: The Stochastic Becker-DeGroot-Marshaak (SBDM) mechanism is a theoretically elegant way of eliciting incentive compatible beliefs under a variety of risk preferences. However, the mechanism is complex and there is concern that some participants may misunderstand the incentive properties of the mechanism. We use a two-part design where we first identify participants with high and low cognitive ability and elicit beliefs from both groups using either the SBDM or Introspection. We find that reports are more accurate in the SBDM for participants with high cognitive ability in problems where calculating correct beliefs is cognitively costly. However, participants with low cognitive ability have low accuracy in the SBDM and perform worse than those in the introspection mechanism in problems where calculating correct beliefs is easy. Our results show that complexity is an important consideration in developing elicitation mechanisms and identifies cognition as an important consideration in interpreting elicited beliefs. Financial Delegation in Late Adulthood Sylvain Hohn Queensland University of Technology (Anup Basu, Uwe Dulleck, Nicholas Cherbuin and J. Henry) There is a growing body of literature suggesting that people are willing to pay a control premium to retain decision-making rights. In our study, we investigate how the control premium is affected when participants are given the option to revoke the delegation of financial decision-making tasks to their spouses. In a framed field experiment, couples aged 60 – 89 in South-East Queensland were incentivised to solve 10 financial decision-making tasks, which were presented in an increasing order of difficulty. They were allowed to transfer tasks to their spouses. Our results indicate, that the option to revoke the delegation of tasks increases the willingness to transfer decision-making responsibility. Participants were more likely to transfer financial decision-making tasks when they were allowed to revoke the delegation decision. This resulted in a significantly higher number of tasks solved. By this view, the control premium participants were willing to pay, was lower when they were allowed to revoke the delegation decision. We also find a strong gender effect. Whilst we do not find gender differences in financial competence, men are significantly less likely to transfer financial decision-making tasks. Our results have implications for policy makers who seek to increase the transfer of financial decision-making rights in Australia. Day 2: Tuesday 26 November 2019 Session 3A – Cooperation 2 The Ghost of Institutions Past: History as an Obstacle to Fighting Tax Evasion Aaron Kamm University of Melbourne (Christian Koch and Nikos Nikiforakis) Can a history of evasion affect tax compliance after a major institutional reform? We address this question in a novel experiment varying the quality of past and present institutions. We observe that tax compliance is substantially higher in good-quality than bad-quality institutions when there is no history of tax evasion. When a bad-quality institution is replaced with a good-

quality one, however, tax compliance remains low, as if the institutional change had not occurred. The reason is that the institutional change appears to leave expectations about future compliance largely unaffected. This suggests that past incentives should not be ignored as they are in traditional models of compliance. Our analysis indicates that the higher evasion has been historically, the stronger incentives may need to be to overcome the 'ghost of institutions past'. Cooperation and Distributional Conflict with a Costly Contract Option Sookie Xue Zhang Zhongnan University of Economics and Law (Ralph-Christopher Bayer) An early study by Bayer (2016) shows that positive contributions in a joint venture investment game can be welfare damaging if a distributive contest follows. We ask whether exante contracting can prevent ex-post rent seeking in such partnership venture situations. As expected, the majority of participants implemented the contract with full cooperation when the division rule aligns individual incentives. We further study the egalitarian social norm as an alternative division rule that transforms the two-stage cooperation and contest game into a standard public-goods game if followed. Interestingly, we find that implementing the contract seems to improve and stabilize average contributions among those participants who opted in. More over, implementing the contract has a significant positive effect on individual contributions if the implementation cost is high. When the implementation cost is low, the decline trend of contracting frequency is postponed and participants fought harder in contest stage if they contributed more. Pledge-and-Review Bargaining: A Preliminary Test in the Laboratory James Tremewan University of Auckland (Steffen Lippert) We implement a simple game in the laboratory to test the main insight of the model of Pledge-and-Review bargaining in Harstad (2019): when uncertainty over disagreement payoffs is only resolved after pledges of contributions to a public good have been made (but before unanimity voting on whether those pledges will be implemented) positive contributions can be sustained. We find that this uncertainty does lead to greater contributions, but only for subjects who perform well in the cognitive reflection test. It appears that the reasoning required to reach the (efficient) subgame perfect Nash equilibrium is too challenging for most subjects. Interestingly, despite contributing one’s full endowment being the payoff maximizing strategy given subjects’ actual behaviour in the treatment with uncertainty, the bulk of subjects do not learn, and contributions in fact fall over time. A treatment where all players’ disagreement payoffs are equal to their endowment for sure leads to a sustained high level of cooperation, suggesting that the voting mechanism itself is sufficient to allow coordination on high levels of contributions without a role for uncertainty. Value Co-creation and Value Appropriation in Strategic Alliance: An Experimental Study of Coopetition Yimei Hu Aalborg University (Huanren (Warren) Zhang)

Strategic alliances are competitive in nature as they incorporate both cooperation and competition: partners cooperate to use their complementary resources for joint value creation while simultaneously compete to appropriate the created value. Previous empirical coopetition studies usually adopt case studies or quantitative analysis based on survey data. These research designs can hardly capture each agent’s subtle and paradoxical coopetition mindsets and behaviors, and usually overlooks how the partner’s decision on cooperation and competition dynamically influences the focal party’s strategies. Moreover, because coopetition behaviors can be unverifiable or even unobservable, interview and survey data usually fail to provide complete and objective information on competitive relationships. We study the tension between value creation and value appropriation both theoretically and experimentally by modeling coopetition as dyadic repeated interactions. Using an experimental method, we aim to explore how economic agents balance value creation and value appropriation in strategic alliance. In specific, we investigate two research questions: 1). How does competition for existing value influence resource allocation between value creation and value appropriation,? 2). How does strategic relationship in one market condition influence future relationship when the market condition changes? The experimental results demonstrate that, contrary to the game theoretical analysis and to the conventional wisdom, existing competition does not undermine value creation within the alliance. Instead, past experience has shown to have a significant influence on alliance performance and sustainability: relationships that successfully establish trust at the start are likely to sustain high cooperation level and achieve the efficient outcome, even after the market condition has changed. The persistent effect of past experience can be explained by conditional strategies: players adjust their allocation of resources in the direction matching their opponents’ investment level in the previous round. The results imply that a positive attitude towards competitive alliance, especially in the presence of intense competition, is the prerequisite for a conducive alliance. The study makes one of the first attempts to employ the experimental method to investigate the trade-off between value creation and value appropriation, which contributes coopetition research methodologically. Facing contradictory suggestions from existing literature, our findings suggest that strategic alliance managers need to focus more on establishing benign initial interactions, rather than arbitrarily avoid value co-creation opportunities with competitors. Integrating transaction cost economics, resource-based view, and game theory, our study contributes to the growing strategic alliance and coopetition literature. Though the research setting lines in dyadic inter-firm strategic alliance, our findings may also shed light on understanding coopetition and interactions between agents at intraorganizational and ecosystem levels. Session 3B – Labour Economics Negotiating Down the Gender Wage Gap Chau Tran Monash University (Pushkar Maitra and Ananta Neelim) This paper examines the gender wage gap in Vietnam. We decompose this gender wage gap into explained and unexplained components. Using nationally representative data, we find that the unexplained component is the entire source of the gender wage gap. How much of this unexplained component reflects pure discrimination? Using surveys and lab in-the-field experiments we obtain data on observables (endowments) and unobservables (the propensity to negotiate and risk preferences) that are likely to have significant effects on wages and labour

market outcomes. Controlling for propensity to negotiate in the wage regression can reduce the unexplained component of the wage gap by up to 30 percent. Including risk in the wage regression results in up to a 15 percent decline of the unexplained component. Controlling for both of the variables can reduce the unexplained component by up to 37 percent. Our results suggest that disregarding gender differences in preferences can lead to an overestimation of the magnitude of gender discrimination in labour markets. Scorecards, Rankings and Gateways: Remuneration and Conduct in Financial Services Le (Lyla) Zhang Macquarie University (Elizabeth Sheedy and Dominik Steffan) The financial services industry typically provides performance pay to staff – a practice that has been linked to misconduct. This lab-in-the-field, real-effort experiment with 318 finance professionals investigates the impact of performance pay on behaviour, in an environment where compliance with policy is imperfectly monitored. For the first time we examine the remuneration structures which have been proposed by the industry and its regulators to mitigate misconduct: the balanced scorecard and the compliance gateway. Relative to fixed remuneration, both these remuneration structures produce significantly worse compliance outcomes. The study also examines the role of relative performance information which is found to boost productivity. Peer feedback has no significant impact on the proportion of participants choosing to comply with policy, but for those who sometimes violate policy, peer feedback reduces the compliance rate. Membership of a professional association has no significant impact on the proportion choosing to comply throughout, but for those who sometimes violate policy, members have higher compliance rates than non-members. Perceptions of peer compliance predict actual compliance behaviour in all treatments, highlighting the importance of social norms or culture for behavioural outcomes. Promotion and Demotion in Multi-Stage Contests Jingjing Zhang University of Technology Sydney (Jonathan Levy) We develop a multi-stage contest design where heterogeneous agents face the prospect of promotion and the threat of demotion from one stage to the next. The theoretical model illustrates how the efficacy of promoting and demoting agents in incentivizing effort depends on ability differences between agents. Specifically, if abilities are homogeneous, the principal is better off pooling agents in one division. However, if abilities are heterogeneous, the principal is better off assigning agents to separate divisions based on ability level, while allowing for agents to be promoted and demoted after each stage of play. The experimental results support the use of promotion and demotion in multi-stage contests when abilities are heterogeneous. In contrast with the theoretical predictions, we did not find significant differences in total effort between the pooled contest and the contest with promotion and demotion when abilities were homogeneous. We believe this discrepancy between the theory and the experimental findings might be due to our participants’ desire to achieve a higher status.

Signals from On High and the Power of “Growth Mindset”: A Field Experiment in Workplace Diversity Andreas Leibbrandt Monash University (Jeffrey A. Flory, Christina Rott and Olga Stoddard) White males occupy most high-profile positions in the largest U.S. corporations. Many firms have set ambitious goals to increase demographic diversity among employees, but there is a dearth of empirical evidence on effective ways to do so. We run a large field experiment with a Fortune 500 company of over 15,000 employees to test several approaches suggested by the literature. By randomly varying a small portion of the content in recruiting materials seen by over 6,000 prospective applicants in two different populations at different stages in their career progress, we test different types of signals aimed to increase interest from racial minorities and women. We find that self-selection into high-profile positions at two different early-career stages exhibits a substantial gap by race even among those who have expressed interest and already begun the process. We also find that the race gap, and self-selection by ethnic minorities, can be strongly influenced by several treatments, even at a relatively late stage in the process of applying. Some treatments increase application rates by minorities by as much as 30 percentage points (63%), others are particularly powerful at closing the gap and raising applications by women of color. The heterogeneities we find by gender, race, age, and stage of career progress help shed important light on the underlying drivers of self-selection barriers among underrepresented groups. Session 3C – Market 3 The Impact of Organisational Form on Contract Design and Buyer-Seller Interaction Andreas Ortmann UNSW (NE Cole, PC Yu and JJ Zhang) We investigate the impact of organisational form on the contracting behaviour of firms, the responses of agents of different cognitive and informational types to those contracts, and firms’ anticipations of agents’ responses. Specifically, we address theoretically and empirically the difference in behaviour of for- and not for-profit firms. We construct a model to predict differences in how for- and not-for-profit firms contract with agents of different cognitive (sophisticated or naıve about the likelihood of outcomes) and informational (informed or uninformed about organizational form) characteristics. A pilot experiment provided some (noisy) confirmatory support for our model's predictions. Cost Containment in Pollution Auctions Lana Friesen University of Queensland (Lata Gangadharan, Peyman Khezr and Ian A. MacKenzie) This article investigates supply reserves in pollution permit auctions. A supply reserve is a fixed quantity of permits that is automatically released if the initial clearing price is sufficiently high. The main rationale for using such a reserve is for cost containment: to lower the final clearing price. We show the inclusion of a reserve does exactly the opposite and provide corroborating experimental evidence. Relative to a benchmark without a supply reserve, we find that the introduction supply reserve will actually increase the clearing price, increase the

revenue from the auction, and increase auction efficiency. The clearing price also increases in the level of the trigger price and relative size of the reserve. This has important implications for supply reserves currently in use, such as the Cost Containment Reserve (CCR) within the US Regional Greenhouse Gas Initiative (RGGI). What Makes Pyramid Schemes Work? Kenan Kalaycı University of Queensland Ponzi and pyramid schemes promise a large return for investors with no risk. In reality, most investors lose money when the scheme collapses due to lack of new entrants. In an online experiment without deception, we invite participants to invest their endowment in a pyramid-like investment scheme with a negative expected return. We find that about half of the subjects are willing to invest in the scheme regardless of age, gender, years of schooling, trust levels, and income. Our results suggest that pyramid schemes have inherently attractive qualities beyond the deceptive claims of the scheme founders. Peer Effects in Public Support for Pigouvian Taxation Erte Xiao Monash University (Lingbo Huang)   We examine how peer effects can impact public support for Pigouvian taxation in a market experiment with negative externalities. Our experimental data indicate that the support rate increases significantly when tax supporters are selected to communicate with voters. Nonetheless, left to their own devices, people are generally reluctant to communicate with and influence others. Among those who are willing, both tax supporters and objectors are equally likely to volunteer and are equally persuasive. As a result, the overall positive peer effect disappears. These findings offer an explanation for the continuing low public support for social-welfare enhancing tax policies.   Session 3D – Risk 3 Classification of Distributional Preference under Risk Edwin Ip University of Exeter (Joseph Vecci) We propose a new method that classifies distributional preference under risk using three binary choice lottery questions. Leaders are asked to make decisions that affect how payoffs are distributed among their followers. Based on their decisions, they are classified according to 2x3 preference types representing different social welfare functions, including egalitarian maximising equality of outcome, egalitarian maximising equality of opportunity and different types of utilitarian. In our baseline, we find that 69% of the leaders express egalitarian preferences maximising the equality of outcomes and only 12% of the leaders express any type of utilitarian preferences. However, preference may not translate to behaviour if followers can

choose to reward their leaders for their decisions. The leaders’ behaviour may change depending on the institution. If followers can only observe their own outcomes but not the choices that their leaders make when they choose whether to reward their leaders, then leaders’ behaviour become significantly less egalitarian, with only 41% making decisions consistent with equality of outcome egalitarian preferences and 33% now making decisions consistent with utilitarian preferences. On the other hand, if followers can observe the leaders’ choice sets and decisions before choosing whether to reward their leaders, then the leaders’ decisions are not statistically different from those in the baseline. Our results show that what social welfare function leaders use to select policies may depend on the transparency of their decision making and whether they are rewarded based on how they make decisions or only on the outcomes of their decisions. Risky Mistakes and Revisions Zachary Breig University of Queensland (Paul Feldman) We adapt Andreoni and Harbaugh's (2013) risk elicitation method to study of "mistakes" in choices over lotteries. We utilize repetitions, revisions, and revealed preference in order to disentangle multiple potential explanations of randomness in choice behavior. Subjects revise their decisions over 85 percent of the time, although this drops by 15 percentage points when they are reminded of what they chose earlier. Revised decisions appear to be better: using a parameterized model of risky decision making, revised decisions give higher levels of expected utility regardless of how the parameter is estimated. The richness of the data allows for the separation of revisions, violations of rationality, deviations from different predicted behaviors, and purposeful randomization. Motivated False Memory Xiaojian Zhao Monash University (Soo Hong Chew and Wei Huang) People often forget and sometimes fantasize. This paper reports a large-scale experiment on memory errors and their relation to preferential traits including time preference, attitudes toward risk and ambiguity, and psychological traits such as anticipatory feelings. We observe systematic incidences of false memory in favor of positive events and positive amnesia in forgetting past negative events. Both positive delusion and positive confabulation significantly relate to present bias, but this is not the case for positive amnesia. In an intra-person, multiple-self model, we demonstrate that positive false memory, rather than selective amnesia, serves to enhance confidence in one's future self in equilibrium, thereby accounting for our empirical findings. Voluntary Contributions to Public Goods with Heterogeneous, Uncertain or Ambiguous Individual Productivity Zack Dorner University of Waikato (Steven Tucker and Gazi Hassan)

The voluntary contribution mechanism (VCM) to the linear public goods game is associated with a large and long run literature. More recent papers have investigated impacts on contributions of heterogeneity, risk and ambiguity in marginal per capita return (MPCR) from the public good. As is generally common throughout the public goods literature, it is assumed a dollar contributed by one individual to the public good is the same as a dollar contributed by another individual. We investigate a neglected, but highly relevant set up. In our experiment, the contribution of one individual to the public good may be less productive than another. This set up has many examples in the field. For example, farmer investment in mitigation of negative impacts on water quality is dependent on a range of geophysical attributes, which are not fully understood. Hence, for any individual farmer, the actual impact of their mitigation action on water quality is subject to ambiguity. Other examples of heterogeneous and uncertain or ambiguous risk on productivity of contributions to public goods include workers in a shared project, public participation in research, and healthy behaviours with a public good aspect such as vaccination and hand washing. Our design consists of four between subject treatments in a standard VCM public goods game. Subjects are in fixed groups of four, for ten periods. First, the control with no heterogeneity or risk and an MPCR of 0.6. Second, heterogeneity of productivity with certainty, with two group member’s contributions subject to an MPCR of 0.3, and two of 0.9. These are randomly reallocated in each period, and subjects know their MPCR before making a contribution. Third, each individual group member’s contribution is subject to an independent 50-50 lottery of being 0.3 or 0.9, and is unknown at the time of contribution. Fourth, each individual group member’s contribution is either 0.3 or 0.9, but is subject to ambiguous risk and unknown at the time of contribution. Individual risk and ambiguity preferences are also collected. Our findings can be compared with previous research into public goods contributions under other sources of risk and heterogeneity. We can inform whether heterogeneity, uncertainty or ambiguity in productivity of contributions to public goods should be included in future experimental research to add realism. Session 4A – Leadership Gender, Leadership and the Nature of Employment Contracts Sherry Li University of Auckland (Ananish Chaudhuri and Erwann Sbai) This paper is concerned with gender differences in labor market interactions between employers and workers. These may go some way to explaining observed differences in remuneration. We focus on two situations—one where the employers are interacting with the same employee over time, and another where the interaction is random. Employers can either rely on mutual trust and reciprocity, or they can impose a penalty to ensure compliance. We find that female employers make more generous offers and consequently earn less compared to male employers in a random matching treatment but not in a fixed matching situation. Female employers tend to choose penalty contracts over trust contracts more often than male employers, but this tendency declines over time. There are no pronounced gender differences in the level of shirking under a trust contract but female workers are more likely to shirk in the presence of a penalty contract. Overconfidence as a Strategy in Leadership Striving Changxia Ke Queensland University of Technology

(Alice Solda, Lionel Page and William von Hippel) This paper investigates whether individuals become strategically overconfident when being a leader give access to valuable resources and to what extent overconfident leaders are detrimental to the group. To address this question, we designed a laboratory experiment in which we exogenously manipulate the incentives to be a group leader. After completing an effort task, participants are matched in groups of four. Each group has to select a leader who will make a series of risky binary choices on behalf of the group. For each decision, the probability to obtain the desirable outcome depends on the likelihood that the leader was the best performer of the group. Thus, selecting the best performer of the group as the leader maximizes the social outcome. Before casting votes to select a leader, group members are allowed to communicate among themselves via a group chat box. In one treatment, the leader receives equal payment as other group members. In another treatment, the leader receives a bonus on top of the part of the payoff that is identical to other group members. Results show that leaders in the treatment with the bonus (i) are less likely to be the best performer in their group and (ii) make overconfident choices that earn less money for their group compared to leaders who did not receive the bonus. However, while we observe that participants in both treatments are overconfident regarding their absolute and relative performance, participants in the bonus treatment are not significantly more overconfident. Insurance Advice as a Signalling Device in Markets: Evidence from a Lab Experiment Ben Grodeck Monash University (Franziska Tausch and Erte Xiao) We design and test a novel advice mechanism aimed at promoting trust and cooperation in markets with asymmetric information. Sellers are given the option to advise buyers whether to purchase third-party insurance against the potential losses from sellers’ noncooperative behavior. A Bayesian model suggests that both cooperative and noncooperative sellers will advise buyers not to purchase the insurance. Once this advice has been given, noncooperative sellers are less likely to pursue self-interest due to guilt aversion. Experimental data from a buyer-seller game shows that, compared to a market where there is no opportunity to give advice, in the market that contains the advice mechanism, sellers are more likely to cooperate and buyers are more likely to purchase the product from the sellers who advise not to purchase the insurance. The number of sellers who advise not to purchase the insurance increases over time. As a result, more efficient transactions take place when the advice mechanism is implemented. The Influence of Leader Reputation on Followers’ Response to Cheap-Talk Messages Philip J. Grossman Monash University (Mana Komai Molle) We consider how followers respond to three attributes – Generosity, Trust, and Trustworthiness – that define a leader’s reputation. Subjects play a series of games, dictator and trust, followed by a sequential leader/follower investment game. The former games provide measures of our reputational attributes. In the investment game, the leader, informed of the investment’s payoff, sends a cheap-talk message recommending invest/not invest to the uninformed followers. Followers also receive reputational information or irrelevant personal information, or no

information at all. We investigate how the reputational information affects group cooperation. While, on average, reputation information has no significant impact on group coordination, leaders with a positive (negative) reputation significantly increases (decreases) group coordination. All three reputational attributes are positively and significantly correlated with followers’ willingness to adhere to the suggested action. Finally, we find that the reputational attributes are reliable indicators of leader behavior. Session 4B – Social Preference 2 Accounting for Knaves by Punishing the Virtuous: Fairness in Experimental Market Equilibrium Aaron Nicholas Deakin University Much evidence exists to suggest that humans often behave inconsistent with Homo Economicus – selfishness is a convenient but inaccurate assumption. Despite this, evidence also continues to suggest that humans behave consistent with selfishness when interacting in competitive markets: “this general result remains one of the most robust findings in experimental economics” (John List, JPE 2017). However, behaviour consistent with selfishness need not imply actual selfish preferences. This difference is not trivial, since the (selfish) competitive equilibrium may no longer be surplus maximising in the presence of non-selfish preferences. In light of the inability to distinguish between other-regarding and self-regarding preferences in competitive markets, I develop an experiment to scrutinize agents’ preferences in a competitive setting through information revelation. I find that while prices in the competitive market converge to standard (selfish) equilibrium predictions, agents’ choices reveal non-selfish preferences. It is as if some agents have a preference for the type of agent they exchange with. The key implication of this is that with sufficient information about seller profits, competitive markets effectively punish buyers with nonselfish preferences relative to those without. Markets have traditionally been viewed as moral-free institutions: agents are constrained to behave similarly regardless of whether they are selfish or other-regarding. Instead, our evidence suggests that competition may discriminate against those with other-regarding preferences by rendering their preferences materially costly and ineffectual. In turn, the presence of such preferences may help identify markets that are unlikely to be surplus-maximising despite being in (selfish) competitive equilibrium. Do We Prefer Praise from Friends or Strangers? An Experiment on Esteem-Seeking in Random and Repeated Interactions Amy Beth Corman University of Melbourne (Nilss Olekalns and Hugh Sibly) People enjoy receiving esteem, and dislike dis-esteem. Esteem seeking will thereby influence people’s actions and thus economic outcomes. We present findings from an economic experiment designed to identify some key determinants and impacts of esteem seeking in a competitive environment. The experiment's two between-subject treatments are designed to determine whether repeated interaction with another individual increases or decreases esteem seeking activities.

Participants are paired in each of five rounds with either the same other participant, or a different other participant in each round depending on the treatment. Each round begins with a real effort task which gives a performance tally. A participant then decides if they want to purchase any credits and these are added to their performance tally to give their score. The participant’s payment for the round is equal to the tally minus the credits bought. In this way a participant can buy a better score.

In each round the participant's matched partner observes the participant's score, but not their performance. In order to provide additional saliency to the observed score, partners provide feedback on the participant’s score using a Likert scale (from very poor score to very good score). The feedback reflects the level of esteem a participant receives from their partner. In this environment a participant’s purchase of credits, or an increase in their performance, are indicative of esteem seeking.

It is difficult to predict, a priori, whether repeated interactions with a given person increase or decrease esteem seeking relative to random interactions with different people. Esteem seeking is costly, so when possible, it is in the interests of both parties in a repeated interaction to implicitly agree not to esteem seek. Thus, we might expect lower esteem seeking on repeated interactions. On the other hand, it may be that participants get into a ‘rat race’, with esteem seeking by one participant leading to further esteem seeking by their partner. In this case it might be expected that there is greater esteem seeking with repeated interactions relative to random interactions.

We find that, across all participants, the number purchasing credits is more volatile across rounds in random matching than repeated matching. However, when averaged across rounds approximately the same proportion of participants (18-19%) purchase credits in each treatment. However, the average number of credits purchased (by those who purchase credits) is approximately 40% higher in the repeated matching treatment.

Approximately 23% of females purchase credits, whereas 13% of males purchase credits. Both males and females in the random matching treatment, and males in the repeated matching treatment purchase approximately the same number of credits. However, males in the repeated matching treatment purchase nearly 3 times the amount of credits than in the random matching treatment.

Overall there appears to be greater esteem seeking when people are repeatedly matched than when they are randomly matched. This is driven by slightly higher effort by both men and women in repeated interactions and a greater purchase of credits by males when there is a repeated interaction. Honesty, Ability, Norm, and Socioeconomic Status: Experimental Evidence from Bangladesh Dina Tasneem American University of Sharjah (Minhaj Mahmud) In a real effort task experiment, we study the (dis)honesty of undergraduate students in Bangladesh. Consistent with earlier studies, when they self-report their performance, a significant fraction of students cheats to varying degrees. We find that an individual’s own ability, as well as social norms in terms of beliefs about peers’ behavior, are the two most important factors influencing (dis)honesty in our experiment. In particular, a higher actual performance in the real effort task reduces both the likelihood and extent of cheating, while the

belief that peers are cheating increases both the likelihood and extent of cheating. Additionally, a lower perceived fear of detection increases the extent of cheating, but does not increase the likelihood of cheating. Among the two most important indicators of socioeconomic status that we considered, such as parents’ education and income, only mother’s level of education shows a significant negative effect on the likelihood of cheating. Migration and Long-Term Demographic Change: Can We Control the Numbers? Klaus Abbink Monash University (Behnud Mir Djawadi) In many Western societies, mass immigration has been one of the most divisive policy issue in recent years. Seemingly moderate inflows of migrants can have substantial demographic consequences in the long run, due to (1) higher fertility of the migrant population, (2) its younger age distribution, and (3) the possibility of family unification (chain migration). Yet demography hardly makes an appearance in the policy debate, even in media outlets that are critical of mass immigration. This suggests that population dynamics are not widely understood and systematically underestimated. We design an incentivised lab experiment in which we confront our student subjects with different migration scenarios. All subjects are in the role of the government and have to decide how many migrants to allow into the domestic country so that after 60 years from now, the share of migrants reaches a target threshold of 20% of the total population. Subjects are presented 30 different scenarios varying in the fertility rate, the age structure, the opportunity for family unification, and the annual inflow of migrants Subjects press a start button that initiates a constant flow of population “B” to enter the country of “A” per year. As soon as the subject presses the stop button the flow of migration stops. The further population development is projected using the cohort-component method. We are interested in which of the elements affect subjects’ ability to control the population dynamics. Our results show that all our preregistered hypotheses are supported: In the scenario with all three elements subjects overshoot the target. Adding any element leads to later stopping points. Overshooting is even more pronounced in additional scenarios that closely resemble the German situation since the opening of the borders in 2015. Our findings call for better public information about demographic mechanisms and a more long-term strategy in immigration policies. Session 4C – Theory and Experiments Planar Beauty Contests Mikhail Anufriev University of Technology Sydney (John Duffy and Valentyn Panchenko) We introduce a planar beauty contest game where agents must simultaneously guess two, endogenously determined variables, a and b. The system of equations determining the actual values of a and b is a coupled system; while the realization of a depends on the average forecast for a, 𝑎, as in a standard beauty contest game, the realization of b depends on both 𝑎 and on the average forecast for b, 𝑏. Our aim is to better understand the conditions under which agents learn the steady state of such systems and whether the eigenvalues of the system matter for the convergence or divergence of this learning process. We find that agents are able to learn the steady state of the system when the eigenvalues are both less than 1 in absolute value (the sink property) or when the steady state is saddle-path stable with the one root outside the unit circle

being negative. By contrast, when the steady state exhibits the source property (two unstable roots) or is saddle-path stable with the one root outside the unit circle being positive, subjects are unable to learn the steady state of the system. We show that these results can be explained by either an adaptive learning model or a mixed cognitive levels model, while other approaches, e.g., naive or homogeneous level-k learning, do not consistently predict whether subjects converge to or diverge away from the steady state. A Systematic Test of the Independence Axiom: Around Certainty Ritesh Jain Academia Sinica (Kirby Nielsen) We experimentally investigate the Independence Axiom - a central tenet of the expected utility theory. In our experiment, we test this axiom on the entire probability simplex. This method allows us to simultaneously study both the certainty effect and the reverse certainty effect. A preliminary analysis suggests that the Independence Axiom is violated systematically across the entire simplex. Moreover, the nature of violations is consistent with the reverse certainty effect as opposed to the accepted experimental knowledge which suggests the prevalence of the certainty effect. Our experiment contributes to the existing literature by studying the Independence Axiom on the entire simplex and is the first one to document the prevalence of the reverse certainty effect. Exploration vs Exploitation, Impulse Balance Equilibrium and a Specification Test for the El Farol Bar Game Paul Pezanis-Christou University of Adelaide The paper reports on market-entry experiments that manipulate both payoff structures and payoff levels to assess two stationary models of behaviour: Exploration vs Exploitation (EvE, which is equivalent to Quantal Response Equilibrium) and Impulse Balance Equilibrium (IBE). These models explain the data equally well in terms of goodness-of-fit whenever the observed probability of entry is less than the symmetric Nash equilibrium prediction; otherwise IBE marginally outperforms EvE. When assuming agents playing symmetric strategies, and estimating the models with session data, IBE yields more theory-consistent estimates than EvE, no matter the payoff structure or level. However, the opposite occurs when the symmetry assumption is relaxed. The conduct of a specification test rejects the validity of the restrictions on entry probabilities imposed by EvE for agents with symmetric strategies in 50 to 75% of sessions and it always rejects it in the case of IBE or when the data is pooled over sessions, which indicates that the symmetric variant of these models has little empirical support. Evolution of Cooperation in Noisy Repeated Games with Social Structures Huanren (Warren) Zhang University of South Denmark (Tim Cason and Vai-Lam Mui)

Social structure has been shown to be important for the emergence and persistence of cooperation. Most previous studies, however, have only focused on static networks and investigate how cooperation emerges in exogenously fixed social structures. In this paper, we

use a computational experiment to study the simultaneous evolution of strategic behaviors and social structures in a dynamic setting. In the computational model, players play infinitely repeated prisoner’s dilemmas with each other. Players adjust their repeated game strategies using reinforcement learning. We investigate the evolution of strategies with and without partner selection: without partner selection, each player randomly proposes to another player in period. The simulation results show that the evolution of social structure differ greatly depending on whether players have repeated interactions: in the one-shot setting we observe sparse networks with heterogeneous behaviors and preferences for partners, while in the repeated setting where players can use repeated game strategies, we observe dense networks with homogeneous behaviors and preferences. Both dynamics network formation (partner selection) and repeated interactions can promote cooperation. Repeated interactions are shown to be more effective in promoting cooperation compared to partner selection. Moreover, partner selection does not influence the strategic behaviors and cooperation levels when players can use conditional strategies in the repeated setting. The evolved strategies and network structures differ greatly in the one-shot setting and in the repeated setting. In the one-shot setting, the population is quickly divided into cooperators and defectors, where cooperators only propose to interact with other cooperators. While some defectors also propose to the cooperators, almost none of these proposals are accepted. The social network also becomes sparser over time and with a gradual decrease in cooperation level. In the repeated setting, selection of strategies and partners becomes more homogeneous over time. In the noise-free environment, only strategies that can lead to full cooperation remain in the population, and the network has the tendency to become a complete network. In the noisy environment, the non-forgiving and non-lenient strategies either exhibit low frequencies or are totally eliminated from the population, while popular players gradually emerge in the social network: many players assign relatively high probabilities to the same players as partners, causing some popular players to receive a high payoff. We also find that although imperfect information reduces the average payoff, it increases the average cooperation level in the repeated setting, which is consistent with previous studies. However, such conducive effect of errors is absent in the one-shot setting. Session 4D – Matching and Search The Role of Advice in a University Admission Matching Clearinghouse Pablo Guillen University of Sydney (Alexander Kiefer and Mark Melatos) The University Admission’s Centre (UAC) manages domestic student applications to universities in NSW and the ACT. There are more than 50,000 such application each year. Applicants are asked to rank up to five courses and are provided with advice from UAC. Namely, they are advised to not to rank a less preferred course above a more preferred one regardless of the likelihood of acceptance. Some universities with guarantee entry programs advise student to put them at the top of their ranking. The latter is a misleading piece of advice as, if followed, students could miss out on a more preferred course. We test the effect of UAC’s

and university advice in an online experiment with a 2x2 factorial design. Participants will be applicants from the 2018 cohort who are thus familiar with the system. Information Acquisition and Preference Formation in College Admissions Siqi Pan University of Melbourne (Rustamdjan Hakimov and Dorothea Kübler) Every year during college admissions, students and their parents devote considerable time and effort to information acquisition in order to form their preferences over a large number of universities. In this paper, we explore ways to reduce wasteful information acquisition using a market design approach. We theoretically and experimentally compare the widely recommended direct student-proposing deferred acceptance mechanism (DSP-DA) with a sequential university-proposing deferred acceptance mechanism (SUP-DA). Consistent with the theory, SUP-DA leads to less wasteful information acquisition and higher student welfare than DSP-DA. In addition, we observe significantly smaller deviations from optimal search strategies under SUP-DA. We also investigate the effects of providing historical cutoff scores, which is a common practice in many real-life college admissions markets. The Complementary Duet of Vehicular Diverging: An Experimental Approach Mingyue (Selena) Sheng University of Auckland (Addison Pan) This study considers a complementary duet method for achieving the “ideal” economically efficient vehicular diverging, i.e., the combination of charging toll fee on the highway and applying an average pricing structure for public transportation. We investigate commuter’s route-choice behaviour by exploring equilibria in various coordination games. Specifically, we conduct a laboratory experiment to emulate transportation route-choice games. We vary the highway entry fee structure (among a fixed level of fee, a two-price charge, or without toll), the pricing scheme for public transport (between a constant and an average cost structure), and the road capacity (before and after an expansion). Our empirical evidence not only validates the Downs-Thomson paradox but also provides strong experimental support for adopting the duet method, as it outperforms other approaches by lowing congestion to a socially optimal level. Moreover, we conduct the quantal response justification for commuter’s choice behaviour. As a result, we observe much less variability in entry rates from round to round, which significantly reduces welfare loss. Our analysis of the data would lead to new insights on how subjects indeed respond to the introduction of a congestion toll and shift to public transport, thereby deliver credible policy implications for traffic management on congested roads and highways. The Effect of Time Cost on Search Behavior Maroš Servátka Macquarie University We experimentally investigate the effect of time search cost on behavior in a decision task framed as selling houses. Sequential search in such a scenario is often costly and time-consuming. Previous literature has shown that if the search cost is monetary, people search

less, but despite the unavoidable nature of time search cost, its effect on behavior has previously been unexplored. In our experiment, we implement four treatments: no search cost, unannounced time delay before a new price offer is available, announced time delay, and an appropriately calibrated monetary cost. Being able to discriminate between the effect of monetary cost, time cost per se, and people’s (in)ability to perceive time delay has implications for theoretical modeling of search behavior and is critical to obtaining the empirical information that can guide the process of formulating models with increased empirical validity. Understanding the origins of people continuing search when the opportunity cost of time is higher than the potential increase in payoffs will allow to design mechanisms capable of improving decision quality, especially in unfamiliar or infrequently encountered situations.