The Economic Science Association organized a seminar series featuring junior job-market candidates in economics who use experimental methods. The virtual seminar series ran from October to December 2022 with sessions occuring at different times to accomodate audiences around the world. Each sessions consists of two to three speakers, each speaking for 20 minutes, followed by a discussant.

The scheduled speakers and discussants are listed below.

In a game with costly information acquisition, the ability of one player to acquire information directly affects her opponent's incentives of gathering information. Rational inattention theory then posits that the opponent's information acquisition strategy is a direct function of these incentives. This paper argues that people are cognitively limited in predicting their opponent's level of information, and hence lack the strategic sophistication that theory may predict. In an experiment involving a real effort attention task and a simple two player trading game, I study the ability of subjects to (1) anticipate the information acquisition of opponents in this strategic game, and (2) best respond to this when acquiring their own costly information. I do this by exogenously manipulating the difficulty of the attention task for both the player and their opponent. Predictions of behavior are generated by a novel theoretical model in which Level-K agents can acquire information a la rational inattention. I find an out-sized lack of strategic sophistication, driven largely by the cognitive difficulties of predicting opponent information. These results suggest a necessary integration of the theories of rational inattention and cognitive uncertainty in strategic settings.

This paper studies career concerns in groups where workers can allocate effort between their own tasks and other group members' tasks. We first theoretically investigate how workers allocate effort when they have career concerns. Then, we study their decisions when the labor market can only observe their aggregate performance rather than their individual performance. Consistent with theoretical predictions, our experimental results show that workers tend to allocate more effort to their own tasks when they have career concerns. Additionally, workers are more likely to support others when their aggregate performance is observed by the labor market. We also observe a number of inconsistencies between the data and theoretical predictions and apply the quantal response equilibrium (QRE) to explain the deviations.

We investigate how increasing the complexity of the message space can reduce misrepresentation in strategic communication. We develop a theoretical model that extends the standard cheap talk approach by i) allowing for communication about both a payoff relevant state and non-payoff relevant attributes which are correlated with the state, and ii) supposing that agents are boundedly rational in memorizing the additional attributes. We show that although babbling is the only equilibrium for standard cheap talk, there exists a new and informative equilibrium by introducing dimensions that payoff irrelevant. We adopt a novel experimental design to test our predictions and explore mechanisms that drive changes in the informativeness of communication. We find that increasing the size of the message space, while keeping the number of messages sent fixed, can significantly reduce misrepresentation, particularly when we decrease complexity for the receiver. We find that the informativeness of communication is dependent both on absolute complexity (the size of the message space) and relative complexity for the the sender versus the receiver. We also show that the relationship between complexity and information transmission is non-monotone and there is an interior optimum.

I study the role of impact beliefs in explaining why many people adopt low impact other-regarding behaviors, while only few people engage in high-impact behaviors. Using an online experiment, I test the explanatory power of two potential cognitive mechanisms. First, I propose that motivated impact beliefs might serve as a justification to only adopt low impact behaviors. In particular, I hypothesize that people overestimate impact of low cost behaviors, while underestimating the impact of high cost behaviors. Second, I hypothesize that people only vaguely think about impact of their behaviors, and rather rationalize their choices ex post. In line with a self-serving interpretation, I find that subjects on average slightly overestimate low impacts and underestimate high impacts. Beliefs are, however, robust to changes in incentives. Neither higher incentives for accuracy, nor changes in prices of donation affect beliefs. Reducing scope for ex post rationalization by eliciting beliefs before donation choices does not affect beliefs, but makes it more likely that subjects maximize impact by donating to both projects. Taken together this suggests that besides motivated beliefs, complexity of integrating impact and cost information may play a role in the low adoption levels of high impact behaviors.

This paper proposes a new method to measure beliefs and ambiguity attitudes towards events that are not necessarily equally likely and belong to a discrete set (i.e., discrete sources of uncertainty). Our method increases robustness to misspecification and allows flexibility in parametric choices compared to previous methods. We implement our method experimentally to both equal and different sources of uncertainty in two contexts: trust and coordination games. We find two main results. First, for equal sources of uncertainty, our method successfully reveals that subjects have context-independent beliefs on events, but context-dependent utility and weighting functions. This result indicates that comparing different sources of uncertainty requires a complete measurement of the utility and weighting functions. Second, different sources of uncertainty where the events are not equally likely lead to an increase in likelihood insensitivity, which indicates that the beliefs formation process of unknown events is cognitively demanding.

With the advent of technological advancements, human-bot interactions are becoming increasingly more common. Attitudes that promote or prevent dishonesty in human-human interactions may not carry over to human-bot interactions. Previous studies show that people lie more in a human-bot interaction compared to a human-human interaction because of a lack of social bonds with bots. However, those studies mostly focus on the simplest of situations, reporting to a web link or a chatbot that can only answer limited questions. Technological progress has radically transformed the way of human-bot interaction. Nowadays many of the chatbots are equipped with voice capabilities and can answer and deal with complicated questions. The presence of voice can create stronger social bonds (Kumar and Epley 2021), which in turn might reduce deception. Interacting with voice can also cue people into behaving more morally to avoid negative self-image. To test this, we conducted a 3x2 online experiment with 325 participants using a modified version of coin-tossing experiment (Fischbacher and Föllmi-Heusi 2013). We compare the reported heads between a voice bot and a human voice for both genders. In addition, we also vary the level of sophistication of the voice bot (voice bot vs. AIbot). We find that on average there is no difference in the propensity of lying when reporting to a human voice, a voice bot, or an AI bot. However, we find that there is more extreme liars (reporting all heads) when reporting to a voice bot compared to reporting to a human voice and the level of sophistication of a voice bot decreases the proportion of extreme liars. Consistent with the past studies in gender difference in deception, we find that there is no gender effect in reporting to a human voice. Interestingly, female and male participants behave differently in human-bot interactions. Female participants are less likely to lie to a voice bot compared to male participants. We also find that participants in our experiments lie more to a female voice bot than to a male voice bot. Although on average the level of voice bot sophistication does not affect lying, we find that male participants lie less with higher voice bot sophistication, but only when reporting to a female voice bot. Our male participants lie more to a female voice bot as opposed to a female human voice. However, we observe no effect on dishonesty when participants report to a male voice suggesting that social or self image concerns are triggered by listening to a male voice, irrespective of it being sourced from a human or a bot.

Individuals are frequently tempted to delay tasks to the detriment of their overall well-being – even when they know they shouldn’t. I propose that costly procrastination increases when excuses, or justifications, for myopic behavior are present in a decision-making environment. I use a survey to document an emotional cost of procrastination that is attenuated by excuses and a within-subject longitudinal laboratory experiment that manipulates the availability of an excuse to demonstrate the behavioral impact of said cost. I show that participants exploit uncertainty (that can be costlessly resolved) over future work amounts as an excuse to select options that require less immediate work but risk increasing total work. Forty percent of participants minimize overall work when given full information but avoid the information necessary to do so when given the option. This pattern of behavior increases procrastination four-fold when an optional excuse is present. Excuse-based procrastination reduces immediate work at the expense of increasing future work by 30 percent. The stark difference in how patient an individual appears across situations that are identical except for the presence of an excuse shows how important understanding the supply of excuses is for measuring time preferences. My results also help explain long-standing puzzles, such as persistent naivety.

In this paper I introduce the MRC framework, which presents two theories about how moral rules drive cooperative behaviour: blame avoidance, or an imperative to avoid doing what one considers as blameworthy, and praise seeking, or an imperative to do what one considers as praiseworthy. Using this new framework, I test the extent to which these two moral rules and a set of preference-based theories (selfishness, inequality aversion, reciprocity, spite, social efficiency and maximin) can explain people’s attitudes to cooperation in two co-operation problems: social dilemmas, where the individual and social optima are misaligned, and common interest games, where the individual and social optima are aligned. My results suggest that (i) blame avoidance, inequality aversion and maximin preferences are the best candidate explanations of people’s attitudes to cooperation in both problems; (ii) praise seeking, reciprocity, social efficiency, and selfishness are also explanations of attitudes to cooperation in common interest games; and (iii) spite is the least promising explanation of attitudes to cooperation in either co-operation problem.

Image concerns are a common motive for dishonest behaviour. However, lying to appear more socially desirable can come at a cost. There is plenty of evidence that some patients misreport sensitive information to their doctors to avoid feeling judged or because they are ashamed of their habits. Patients' misreports are a cause for concern because they can significantly affect diagnoses, treatments and ultimately, their health. In this study, we ask whether it is possible to enhance patients' disclosure of truthful information to their doctors. We conducted a field experiment in Mexico City, covering six clinics and 1,174 patients. We experimentally varied the script doctors used to ask about two health-related habits: alcohol and vegetable consumption. In addition, we randomly assigned patients in three clinics to either a female or a male doctor. This setup enables us to establish the causal effect of gender matching and question formulation on patients' reports. We observed evidence of an effect of the gender of the doctor on patients' shared information and, therefore, on dishonest behaviour. Male patients report less alcohol to female than to male doctors. We do not observe such a difference for female patients. Also, we found that the way doctors ask these questions does not affect patients' answers. Our results imply that simple strategies to avoid social-desirable answers in clinical settings might not be enough to improve the quality of patients' reports. However, concordance of some characteristics between patients and doctors may enhance honesty.

Financial advisors frequently propose narratives to explain the recent performance of the stock market and to forecast what it will do next. When advisors’ incentives are not fully aligned with those of the individuals they are advising, there is scope for self-interested persuasion using narratives. In this paper, we use an experiment to learn about the underlying mechanisms that govern this form of persuasion. Our results reveal several insights. First, we characterize the strategies advisers use to construct persuasive narratives. Second, we show that advisors are able to shift investors’ beliefs about the future performance of a company. This harms investors and benefits advisors. Third, we identify the types of narratives that investors adopt. Finally, we evaluate the efficacy of several potential policy interventions aimed at protecting investors. We find that narrative persuasion is difficult to protect against.

Evidence from the experimental laboratory and the field gives contradictory conclusions on whether people play mixed strategy equilibrium. Laboratory data reject the hypothesis of equilibrium mixing, while field data from professional sports supports it. We explain this contradiction by showing that the strategic value of being unpredictable is high in professional sports. Therefore, professional sports players have considerable incentives to be unpredictable and, more generally, to play a mixed-strategy equilibrium. This paper exploits a unique natural experiment that occurred in major league baseball to estimate the strategic value of being unpredictable. Specifically, we study the advantage gained by the Houston Astros when they stole signs in the 2017 season. They used electronic devices to record communication between the opposing team players, taking away their ability to be unpredictable. We report statistically and economically significant effects of being strategically (un)predictable in professional sports.

There is evidence that the average person accepts income inequality that is based on differences in individual achievement. However, a particular achievement is not equally difficult for everyone. I study how information about differences in difficulties affects redistributive decisions. I consider two sources of difficulties: external circumstances and individual ability. Participants have to redistribute the income earned by achievement within pairs with information about the relative difficulty of the task within the pair. I find that participants strongly compensate the member of the pair who had a harder job in producing if the difficulties come from an external source, but fully ignore the relative difficulties when they arise from individual ability. This is true for both when participants redistribute between themselves and another subject and when they redistribute between two other subjects. Nevertheless, when involved, participants choose allocations that benefit them the most: those with externally harder tasks make more selfish, more egalitarian, and more compensating allocation choices, while those with easier tasks are more likely to simply reward performance, even though they learned that performance differences were caused partly by external differences in the difficulty of the task.

This paper investigates the role news media plays in promoting hatred through the news coverage of mass shootings. I first show through observational data, that media treats mass shootings differently depending on the motive behind the shooting. Any time a shooting targets a specific ethnicity/race/religion/gender, i.e., the shooting is hate-motivated, its media coverage differs in two aspects: 1) higher coverage; 2) more focus on the shooter. I then show that there is more public interest in hate-motivated mass shootings, based on online searching behavior. Finally, I provide suggestive evidence that immediately following a hate-motivated mass shooting, there is an increase in the number of hate crimes against the same victimized group. Based on these findings and the existing literature, I hypothesize that the way hate-motivated mass shootings are covered in the news spreads more hatred. I test my hypotheses by conducting an online information provision experiment where I manipulate how a mass shooting targeting immigrants is reported in the news. In particular, I examine how Democrats and Republicans in the U.S., who have different ex-ante views about immigrants, react to the news emphasis on the ideology/background of the shooter. Results from the experiment show that details on the shooter's ideology increases Republican subjects' support for the shooter. Emphasis on the shooter's background increases Democrat subjects' support for the shooter and the shooter's anti-immigration ideology.

I investigate both theoretically and experimentally the economics of empathy and its implications for redistribution. I first define empathy as an accurate simulation of how one would feel if they were in another’s position and separate it from altruism. I then formalize empathy and investigate its implication for redistribution using a motivated reasoning model. In the model, a wealthy individual with a certain prior consumption experience decides both what to think about a poor individual's situation and how much to redistribute to them. This wealthy individual has selfish motivation not to be empathetic towards the poor person –- they can conveniently understate the poor person’s marginal utility of consumption to justify less redistribution. A wealthy individual with a more varied personal experience of consumption has higher cognitive dissonance due to a larger divergence between such understatement and their experience-based knowledge. In this way, wealthy individuals with a more varied experience have higher empathy, and therefore, redistribute more. I design a laboratory experiment with real effort tasks to randomly assign personal experiences and test the model propositions on empathetic beliefs and redistribution choices. Experiment results corroborate the main propositions of the theory -- compared to subjects with more varied experience, subjects with uniform experience are more subject to motivated reasoning, and as a result, are less empathetic and less generous towards their partner in the context of task redistribution.

We study the impact of motivated beliefs, the phenomenon that people believe what they want to believe, on market performance in a laboratory market for a state-contingent common value asset. Motivated beliefs are induced so that traders have polarized preferences over the states. The main findings are that (1) these induced motivated beliefs do not have a significant impact on overall market efficiency, but (2) they do impact traders’ final asset holdings and price spread, and (3) the induced polarization persists after receiving private signals and trading in the market. Other findings suggest that trading reduces motivated traders’ belief distortions when financial stakes are more intense and signal precision is higher. The induced ego-relevant motivation is significantly stronger than the homegrown motivation to believe in higher payoffs.

Mental load is a widespread but invisible psychological burden. It mainly affects women, by constantly loading them with pressing concerns related to household management and children’s well-being. In this study, we investigate whether mental load reduces labor productivity and leads to a self-selection into less cognitive and remunerative jobs, and whether its impact is gender-differentiated. We conduct a lab-in-the-field experiment with more than 700 participants in Nairobi, in which we randomly trigger thoughts related to mental load and then ask to perform an automatic or a more cognitively-demanding task. Results show that mental load reduces productivity in the automatic task, but not in the cognitively demanding one. At the same time, when given the opportunity, treated participants are more likely to self-select precisely into the less remunerative automatic task. A heterogeneity analysis shows that mental load reduces only women's labor productivity in the automatic task, while the self-selection effect is driven by men. This study provides evidence of an understudied psychological channel that, by creating a negative performance loop, widens the gender productivity gap and contributes to the reinforcement of the psychological poverty traps identified in the literature.

Many decisions people make have the potential to increase overall welfare by creating positive externalities. This is especially the case for decisions that are made under uncertainty: Starting a business, developing new technologies, or investing into new ventures, are all decisions that involve uncertainty about personal gains and losses, but have potential positive externalities for wider society. The likelihood of a person taking such prosocial risk is however affected by whether and how societies reward this decision. One channel that may affect such rewards are individuals’ distributive preferences. How potential externalities affect distributive preferences is however an open question. In this paper, I provide a theoretical framework and a first experimental test of how potential externalities affect distributive preferences when income is earned under uncertainty.

Governments, central banks, and private organizations frequently face the challenge of convincing their audience to take a particular action. This paper studies whether these senders' persuasion may benefit from adapting the communication strategy to their receivers' strategic environment. In a laboratory experiment, I test a prediction arising in models of information design (Bergemann & Morris, 2019): Optimal communication involves public signals when the receivers face strategic complements. In contrast, private signals are optimal with strategic substitutes. As a test of the receivers' responses to different communication strategies, this paper studies obedience constraints, which capture the receivers' incentives to trust signals. In the experiment, the senders often succeed in strategically using information to persuade their audience. As predicted, public signals enhance persuasion with strategic complements. The senders themselves respond to this force by engaging more frequently in public communication when the receivers' game features strategic complements. However, public communication strategies perform equally well as private ones, even with strategic substitutes. Therefore, while the advantage of public signals depends on the audience's strategic environment, they are more persuasive than expected if the behavioral benefits of public communication are not considered.

I report results from a controlled laboratory experiment designed to investigate the effect of different feedback structures on belief-updating in an ego-relevant task. Treatments vary whether subjects receive feedback through a signal with a noise or a comparison component. I find that the signal structure is an important determinant of how subjects update their beliefs. This is driven by men and women exhibiting different biases depending on whether the signal is noisy or comparative. Men underweight bad news when the signal has a noise component and women underweight good news when the signal has a comparison component. These findings have implications for policies aiming to reduce the well-established gender gap in self-confidence through feedback provision.

Social and economic networks are pervasive and influence many decisions people make. In dynamic environments, networks may change over time as individuals form and sever connections, and therefore, it is also important to understand how choices and outcomes affect the structure of networks. This study experimentally analyzes a repeated local public goods game with periodic opportunities for individuals to revise their direct connections, thereby altering the prevailing network structure. Treatments vary the frequency of network revision and the cost of supporting links to examine the relationship between choices, outcomes, and network dynamics.

This paper provides the first experimental evidence that information receivers consider the size of the signal space, which is the number of possible signals. When purchasing signals, the willingness to pay for signals increases as the size of the signal space increases, even if the signal accuracies are fixed. The results also reveal the preference reversal phenomenon: when the signal is free, individuals no longer prefer larger signal space. Preference reversal suggests that the motivation for a larger signal space is intrinsic, for instance, curiosity. These experimental findings of this paper have two main practical implications. First, when selling their services, information providers can make their services look more attractive by simply increasing the number of possible signals. Second, the preference for larger signal space suggests the key assumption in the information design literature, which is the signal space is equal to the action space, might not always hold. I found leading theoretical frameworks cannot explain these experimental findings.

This paper investigates the role played by self-confidence in college applications. After documenting how much gender and social differences in self-confidence explain inequalities in access to college, we show that correcting underconfidence helps to close the gap in college applications. We collect unique data on student self-confidence using experiments on more than 2,000 French college applicants. We match this data with administrative data on real college applications. After measuring student under- (or over-) confidence in their academic abilities, we randomly corrected miscalibrated beliefs about relative academic ability by providing some students information on their real rank in the grade distribution. Our results show that the best female students and the best students from low socio-economic status systematically underestimate their rank in the grade distribution. As a result, they apply to less selective programs. Providing information on the correct position in the grade distribution made the best students apply to more ambitious programs with stronger effects for female and low-SES students.

I investigate how workers respond to learning that they are more/less trusted by a manager relative to their peers. In my experimental treatment, a manager decides how to allocate a low and high impact task between two workers after viewing their responses to some personality questions. In the baseline, a random device makes this decision on the manager's behalf. Being assigned a low or high impact task is inconsequential for the workers' earnings and only determines how much the workers' effort is worth to the manager. I find that while low impact workers provide less effort for the manager in the treatment compared to the baseline, there is no symmetric positive effect on the high impact workers. The negative effect on low impact workers is partly driven by their negative emotional response to learning the manager intentionally chose to trust the other worker with the high impact task. My results underscore the negative consequences of managers disclosing their subjective preference over workers, even when this preference is materially inconsequential for the workers' earnings.

Providing feedback on performance relative to a comparison group is a common approach to foster motivation and enhance later performance. In this study, we investigate how the comparison group for such feedback should be chosen. We show in a field experiment taking place in high schools in China that relative performance feedback has a greater positive effect on performance when the comparison group is small than when it is large. We argue that a likely explanation is that the first and last places in a ranking can act as reference points. In small groups, participants’ ranks tend to be closer to these extreme ranks leading to a greater motivation to increase performance in order to reach the first place or avoid the last place.

We organized an online mentoring program, in which college students from an elite university provided voluntary assistance to middle-school students from a less developed area in China. By designing a random phase-in of college students and random matching between mentors and mentees, this study provides a comprehensive evaluation of the effect on mentors. We find that one-semester mentoring service can significantly improve mentors’ skill of caretaking, raise their appreciation of government anti-poverty effort, generate more egalitarianism in the redistribution, increase compromise between equality and efficiency, and create familiarity favoritism. Personal specific experience matters; inexperienced mentors bear larger impact, mentors who encountered difficult communication are more affected, and those who observed different links across mentee’s performance, effort, and family background lean towards different directions in social preference.

In credence goods markets such as healthcare service markets, after a buyer encounters a problem that requires treatment, only sellers have the expertise to determine which type of treatment is sufficient to address the buyer’s problem. Although sufficient treatment maximizes buyers’ expected utility, sometimes it cannot guarantee a 100% success rate. When a treatment failure happens, buyers cannot determine whether it is caused by undertreatment or bad luck after sufficient treatment. Buyers may express their dissatisfaction with sellers in costly ways by engaging in “crying behavior” that result in compensation. To avoid the costly aftermath from such “crying behavior”, sellers will “defend” themselves by overtreating to minimize the probability of treatment failure. However, the high cost of overtreatment incurred by both sellers and buyers will result in a Pareto-inefficient outcome, as compared to the situation where buyers do not cry and sellers choose sufficient treatment. To encourage sellers’ sufficient treatment and discourage buyers’ crying behavior, I extend the one-shot game to a repeated game and introduce a reputation system in which the seller’s treatment history and the buyer’s reactions are publicly visible, and I show that there exists a perfect public equilibrium in which the seller and buyer will frequently play the Pareto-efficient strategy profile. I also hypothesize that education, which primes sellers’ and buyers’ knowledge that they can be better-off when playing the Pareto-efficient strategy profile, can also encourage them to play the efficient strategy profile. My laboratory experimental results show that most sellers overtreat while most buyers cry when neither reputation nor education is present. Buyers are significantly less likely to cry when the reputation system is introduced. When both reputation and education are present, sellers are significantly more likely to choose sufficient treatment and significantly less likely to overtreat.

This paper explores beliefs about how economic inequality changes society, and establishes a causal link between such inequality externality beliefs and redistributive preferences. Using two representative surveys of 6,244 U.S. citizens, we show that nearly every individual believes that inequality affects society in one way or another. A large and consistent majority believes that more economic inequality leads to negative societal outcomes through channels such as increased crime or worsening economic factors. These beliefs are widespread across incomes and party lines, and less divided on these dimensions than comparable economic fairness beliefs. We establish a causal link from inequality externality beliefs to individuals' redistributive preferences by using exogenously provided video information treatments, and estimate the importance of externality beliefs for redistributive preferences to be roughly two-thirds that of classical fairness views. Our results also show that inequality externality-based arguments are generally less polarized and polarizing than classical fairness-based arguments, indicating that these two motives behind individuals' preferences for redistribution are structurally distinct.

We study the effects of group identity on cooperation in experimental two-person strategic games. Our results show that in games of strategic complements, group identity only matters initially when a participant interacts with another participant of the same group by delivering more cooperative choices, but this effect disappears in later rounds. The Nash equilibrium with standard preferences describes behavior well regardless of the identity condition. However, in treatments with strategic substitutes, we find that participants are persistently more competitive when playing with another participant of a different group than when there are no identity groups. This is because, in games of strategic substitutes, participants have spiteful preferences toward participants that are members of the other group which leads to choices that are more competitive than the Nash equilibrium.

The behaviour in the strategic situations is affected by people's own cognitive ability to make rational decisions and their beliefs on opponents' moves. The literature has focused on explaining irrational play assuming non-equilibrium belief and unbounded ability. This paper identifies subjects' iterative reasoning ability in dominance solvable games with beliefs manipulated using known rational computer opponents. We further study whether subjects can learn from a novel tutorial, which teaches iterative dominance, and whether subjects perceive the tutorial useful by asking them to predict their performance. We find that 90% of the subjects cannot play equilibrium, and the tutorial has striking effect on improving reasoning ability - 44% of our subjects learned to play equilibrium after the tutorial. In addition, more cognitively able subjects have higher original reasoning ability and more likely to learn. Finally, we find that tutorial is believed effective in improving performance, even for subjects who do not have ability improvement via tutorial.

Based on a within-subjects experimental design, we examine strategic ambiguity within strategic and non-strategic environments. Specifically, we elicit beliefs and preferences that reveal a subject's perception towards ambiguity and their aversion to ambiguity. The three strategic settings include a modified Tullock contest, a minimum-effort coordination game, and a classic zero-sum Rock-Paper-Scissors game. Additionally, we examine a non-strategic standard three-color Ellsberg task. We employed matching probabilities to elicit these views towards ambiguity for both natural and artificial settings. The elicited matching probabilities were then used to estimate each subject's ambiguity aversion and their perception towards ambiguity in each environment using alpha-maxmin expected utility preferences with a neo-additive core. We find remarkable stability of ambiguity-seeking behavior across the four environments. Our ambiguity perception analysis provides evidence that subjects perceive a lot of ambiguity in games like the coordination game. In contrast, they perceived smaller amounts of ambiguity in the modified Tullock contest with a dominant strategy. Using the elicited beliefs to estimate individual-level ambiguity aversion and ambiguity perception parameters, we determined each subject's optimal choices and compared them to actual decisions made within our experiments. We found that the alpha-maxmin model failed to explain most of the data from the Coordination and Rock-Paper-Scissors tasks but could predict 136 out of 142 subjects' behavior for the modified Tullock contest.

Upstream reciprocity, known colloquially as “pay-it-forward", is the act of reciprocating an act of kindness to an unsuspecting third party. In the present study we contrast it to its evil counterpart: reciprocating an unkind act to an unsuspecting third party which is conveniently termed "negative upstream reciprocity". The comparison is a novelty in the upstream reciprocity literature in psychology but also enriches the economics literature on indirect reciprocation and pro-social behavior. Two experimental interventions are proposed where subjects tackle ancillary tasks and questionnaires. In the control group subjects can only help the next participant while in the treatment they can only harm them; direct reciprocation is ruled out in both instances. Kind or unkind acts are available at a monetary cost but they are received in a non-monetary fashion to avoid income effects. Pay-forward behavior is measured through a monetary dictator game. Our results show that both kindness and unkindness are paid forward at approximately equal rate (extensive margin) but the amount of retaliation is larger when paying forward an unkind act (intensive margin). Our results also show that motivations for the observed behavior are not symmetric, highlighting the role of gratitude when kindness is being paid forward.

How do students form beliefs about how their future career will depend on their choice of college major? Using both nationally representative survey data and surveys that we administered among undergraduates at the Ohio State University, we document that U.S. freshmen hold systematically incorrect beliefs about the relationship between majors and occupations. Students appear to stereotype majors, greatly exaggerating the likelihood that they lead to their most distinctive jobs (e.g., counselor for psychology, journalist for journalism, teacher for education). A stylized model of major choice suggests that stereotyping boosts demand for "risky" majors: ones with rare stereotypical careers and low-paying alternative jobs. In a field experiment among the same Ohio State sample, providing statistical information on career frequencies to first-year college students has significant effects on their intended majors (and, less precisely, on their choices of which classes to enroll in), with larger effects on students considering risky majors. Finally, we present a model of belief formation in which stereotyping arises as a product of associative memory. The same model predicts—and the survey data confirm—that students also overestimate rare non-stereotypical careers and careers that are concentrated within particular majors. The model also generates predictions regarding role model effects, with students exaggerating the frequency of career-major combinations held by people they are personally close to.

Natural language is essential for human decision-making and interaction. It is used to exchange information and convey the consequences of choices and the rules of interaction. In languages with grammatical gender, entity nouns (e.g., ``the student'' or ``the worker'') can be written in a male, a female or, more recently, to promote gender equality and to include people identifying as non-binary, in gender-inclusive forms. There is an ongoing debate if the usage of male, female, or gender-inclusive forms triggers gender stereotypes and leads to potential biases in decision-making. We conducted a controlled experiment and systematically varied the framing of the instructions, either using the male, the female, or the gender-inclusive form. Participants made economic decisions measuring their degree of fair sharing, cooperation, and honesty in simple economic games. We study if participants behaved differently if their self-reported gender matched the grammatical gender used in the instructions. This line of research is important for experimental methodology on writing instructions in languages with grammatical gender and for practitioners in law and organizations.

Is majority voting an effective way of aggregating information in small to medium sized groups when voters have to acquire information from sources of varying quality? We investigate this question in a theoretically informed lab experiment, where participants can obtain relevant information by choosing between two sources of instrumental but imperfect information. Both sources are informative on a true state of nature, but one is more informative than the other. The less informative source grants an additional ‘consumption benefit’. There are thus strong incentives to free-ride on the information acquisition of others. To understand whether such free riding harms the quality of group decisions, we compare four different information environments between subjects. We vary the observability of group-members’ information acquisition choices and the signal realisations. In the baseline, signal source choices and signal realisations are unobservable. In the “Source” treatment, the signal source choices are observable, but the realisations are not. In the “Source and Signal” treatment, the signal source choice and the signal realisation are both observable. In the final treatment, “Endogenous”, participants decide whether they wish to reveal their source-choice and/or realisation of their signal. They can misrepresent their information. We find that in the “source and signal” treatment, participants acquire low quality information significantly more often. However, this does not harm their eventual outcomes since this treatment yields a high average payoff per person. This suggests that free-riding on other group-members’ information acquisition does not harm overall decision quality in an otherwise information rich environment. In sum, we present the first evidence regarding the effect of acting in various information environments, on outcomes such as endogenous information acquisition, voting patterns and payoffs.

Exogenously given reputational information is known to improve cooperation. This paper experimentally studies how people create such information through reporting of partner’s action choices, and whether the endogenous monitoring helps sustain cooperation, in an indefinitely repeated prisoner’s dilemma game. The experiment results show that most subjects report their opponents’ action choices, thereby successfully cooperating with each other, when reporting does not involve a cost. However, subjects are strongly discouraged from reporting when doing so is costly. As a result, they fail to achieve strong cooperation norms when the reported information is privately conveyed only to their next-round interaction partner. Costly reporting occurs only occasionally, even when there is a public record whereby all future partners can check the reported information. However, groups can then foster cooperation norms aided by the public record, because reported information gets gradually accumulated and becomes more informative over time. These findings suggest that the efficacy of endogenous monitoring depends on the quality of platforms that store reported information.

We experimentally investigate the theory of global games in a simultaneous three-agent market entry game with strategic substitutes. The payoff from staying out is constant, whereas the payoff from entering depends on a random state, a heterogeneous cost of entry, and decreases in the number of entrants. The game predicts multiple Nash equilibria for intermediate state values. The (global) game, however, where agents observe a noisy but precise private signal about the state has a unique equilibrium where agents adopt threshold strategies that are ordered by the entry cost. This equilibrium persists in the limit and characterizes the unique equilibrium that is selected in the game without noise. The experiment provides support for the theory. Aggregate and individual behavior follow comparative static predictions. A majority of subjects adopt threshold strategies with few mistakes. Finally, a majority of outcomes in the game without noise correspond to the equilibrium selected by the theory.

We study the effects of overconfidence in a sequential social learning setting. In a lab experiment, we let subjects form beliefs about their own and others’ information by tying the accuracy of their signal to their score on a trivia quiz. Their beliefs on the expected scores allow us to measure and study the effects of confidence on social learning. Our results show that there are two distinct effects of confidence that manifest in their behavior of breaking herds. First, environments that promote relative overconfidence increase the occurrence of herd breaks. Second, environments that promote absolute underconfidence increase herd breaks after subjects realize their true performance. Both cases lead to a welfare improvement compared to their opposite environment. These findings can be explained by a model of social learning where rational agents have information structures that induce overconfidence about their relative signal accuracy.

Since Allais (1953) documented violations of the Independence axiom in decision makers, many experimental studies have produced the same type of behavior. As they are not consistent with standard Expected Utility theory, many models that account for such violations have been developed. The most well-known are the Rank-Dependent Utility of Quiggin (1982) and the Cumulative Prospect Theory of Tversky and Kahneman (1992). In this paper, we propose a theoretical model and a laboratory experiment that generate an alternative type of violations that cannot be consistent with these models and are distinct from the Allais Paradox. We provide and axiomatize a representation in which a decision maker can exhibit different risk attitudes, which we refer to as a Dual Expected Utility representation. In this representation, there will be two different utility functions over the prizes, and the decision maker will use one of them when disappointment probabilities are low and shift to the other once the disappointment probability exceeds a threshold. It contains the standard EU representations as special cases. Unlike the representations that distort weights using a fixed utility function, the DEU allows the concavity/convexity of the utility functions to change. This observation leads to our experiment design, which consists of a series of binary comparison tasks of two lotteries. Specifically, we change the amount of the most disappointing prize while fixing the other prizes in both lotteries. The main implication shares the core idea of Allais as we expect a riskier lottery may be more preferred to a less risky lottery when the probability of a disappointing prize is high and vice versa. However, the design is distinct because we keep the ranks of the prizes constant while varying only the relative size of the disappointment. A potential choice reversal in this series of comparison tasks will be a simple yet direct test against the Expected Utility, the Rank-Dependent Utility, and the Cumulative Prospect Theory. We expect to extend the results further to narrow down possible thresholds for our DEU representation and use the thresholds to predict the presence of Allais behavior in different scenarios of Common Consequence and Common Ratio examples.

In Ecuador, otherwise, equal low-income students exhibit 74% lower educational ambitions than their high-income peers even after controlling for a rich set of covariates such as cognitive ability, psychological traits, family education, and school fixed effects. The lower ambitions impact students’ school achievements. For instance, educational aspirations account for 18% of the standard deviation in students’ test scores. I examine if a “light-touch” role model intervention can reshape this dynamic. To do so, I study the results from a clustered randomized control trial targeting more than 16,500 high school students from 108 schools in Ecuador. I found suggestive evidence that the treatment mitigates –but does not close- the aspirations gap between low and high-income students by 15 percentage points (p-value = 0.06). However, the effect only last when assessing a raw estimation in the short run. Despite that, I found evidence that the intervention exhibited subgroup effects. Among low-income students, those who benefit the most are boys with highly educated fathers at home. I also find that students with high test scores on a standardized test also exhibit a significantly higher benefit from the intervention. My findings contribute to the growing literature on aspirations and inequality by offering suggestive evidence that a light-touch video intervention with the potential to scale up could mitigate the aspirational failures among low-income students. The paper also urges policymakers to give more attention to the psychosocial dimensions of poverty since they are interrelated with future economic outcomes.

Algorithms are often more efficient in strategic decision-making than humans, but humans show aversion against the use of algorithmic support. In non-algorithmic contexts, overconfident individuals are less likely to delegate decisions. The aim of this study is to analyze whether individuals are averse against algorithms or rather overconfident concerning their capabilities. To disentangle the effect of overconfidence and control over algorithms on the use of algorithmic support, I conduct an online experiment. I examine whether the use and the willingness-to-pay for algorithmic support significantly differs, if individuals can either delegate their decisions to an algorithm or instead can receive algorithmic advice. I show that the use of algorithmic support is the lowest for low performing individuals, who would profit the most from algorithmic help. These individuals tend to overestimate their abilities. Through nudging I can significantly increase the usage rate of algorithmic support for the low-performing participants. The nudge comprises the information that these individuals belong to the bottom 20% of participants. This results bear important insights into how algorithmic support systems can be designed in order to increase their acceptance.

Inequality gives rise to the emergence of social status. Some are driven to showcase a desirable status, while others experience stigma from low status environments. We study how eligibility for a social benefit influences preferences for competitiveness. We exogenously manipulate the benefit eligibility in which we expanded the eligibility criteria. Expanding the eligibility criteria did not change the competitiveness of the low performing group, but the middle-income group was affected by the eligibility of the benefits with a 51% reduction in competitiveness. We conclude that the social benefit eligibility reduces competitiveness across treatments for the middle performing group.

Individual forecasts of economic variables show widespread overreaction to recent news, but laboratory experiments on belief updating typically find underinference from new signals. We provide new experimental evidence to connect these two seemingly inconsistent phenomena. Building on a classic experimental paradigm, we study how people make inferences and revise forecasts in the same information environment. Participants underreact to signals when inferring about underlying states, but overreact to signals when revising forecasts about future outcomes. This gap in belief updating is largely driven by the use of different simplifying heuristics for the two tasks. Additional treatments suggest that the choice of heuristics is affected by the similarity between cues in the information environment and the belief-updating question: when forming a posterior belief, participants are more likely to rely on cues that appear similar to the variable elicited by the question.

In absence of policy interventions like affirmative action, employers’ biased beliefs about underrepresented groups may not correct on their own due to less hiring of and subsequent learning about that group. This paper explores whether temporary affirmative action can improve representation even beyond the policy instance in settings where employers hold biased beliefs about performance. I experimentally elicit employer hiring decisions and beliefs about employee performance in two experimental treatments: a control treatment without affirmative action and a temporary affirmative action treatment. While beliefs and hiring are biased against women in the control condition, I find in the temporary affirmative action treatment that representation improves even after affirmative action is lifted. Exposure substantially increases the likelihood that women are hired after affirmative action is lifted. This increase is partially driven by employers gradually learning that their beliefs about women’s performance are biased downward.

This experiment seeks to study the impact of uncertainty and attitudes towards uncertainty on charity donations. We use a modified dictator game, where the donations received by the beneficiaries (environmental NGOs) are exposed to different levels of uncertainty. We elicit attitudes towards uncertainty, risk aversion, ambiguity aversion, likelihood insensitivity, and pessimism. We aim to test if different levels of uncertainty (risk and ambiguity) impact altruistic behavior. We find that different levels of ambiguity have a significant effect on altruistic behavior, individuals give more when in presence of a low level of ambiguity, and they donate less when in presence of a high level of ambiguity. We also find that the effect of attitudes toward ambiguity also depends on the level of ambiguity. We find that ambiguity aversion has a positive impact on charitable giving when there is a low level of ambiguity, meanwhile, it has no effect when there is a high level of ambiguity. On the contrary, we find that pessimism decreases charitable giving only with a high level of ambiguity.

In this paper, I use consumer transaction data from two supermarkets to study if carbon labels cause consumers to reduce their grocery carbon emissions. I exploit a difference-in-difference setting where one supermarket unexpectedly, from one day to the next, introduced labels on around 3000 products with their carbon footprint, while another similar supermarket did not. My sample includes more than 30,000 consumers making 400,000 orders, and I follow them up to one-year post-treatment. I find an average reduction in total emissions of around 2.5 percent from informing consumers about the products' carbon emissions, corresponding to about 1.4 kg CO2 less per order. This effect size is smaller than found in most other studies based on lab or field experiments but still economically meaningful. It corresponds to estimates in the literature of a carbon tax on meat and dairy products of about 20$ per tonne of CO2. The main driver was that customers in the treated store reduced their beef consumption by 8-16 percent and increased their consumption of products with a lower carbon footprint, such as pork, poultry, and vegetarian products.