|
on Experimental Economics |
| By: | Hayk Amirkhanyan (University of Warsaw, Faculty of Economic Sciences) |
| Abstract: | This paper reports the results of two experiments exploring the effects of monetary and non-monetary incentives on overconfidence and underconfidence. It additionally explores the moderating effects of gender. The participants of both experiments are university students taking their academic tests. Right before the tests, they predict weather their score will be higher or lower than the average in their class group. Overplacement is defined as expecting a better-than-average score but actually getting a below-average score. Similarly, underplacement occurs when a subject expects a below-average score but ends up above the average. Half of the subjects are incentivized: some receive money and others a handmade gift conditional on predicting correctly. Experiment 1 is conducted weekly for nine tests and involves students of one course. Experiment 2 involves students of four different courses, and the grade predictions are made on the final exam only. While Experiment 2 shows no significant results, the main finding from Experiment 1 is that the handmade gift reduces the likelihood of overplacement and underplacement for women, making them significantly better calibrated than men. Experiment 1 also shows that overplacement, but not underplacement, is more likely when the stakes are high (midterm and final tests compared to the weekly quizzes). Finally, there is limited support for the Dunning-Kruger effect. |
| Keywords: | overconfidence, overplacement, underconfidence, non-monetary incentives, gender differences, Dunning-Kruger effect |
| JEL: | D91 J16 C91 M52 |
| Date: | 2026 |
| URL: | https://d.repec.org/n?u=RePEc:war:wpaper:2026-37 |
| By: | Fabrizio Adriani (University of Leicester); Abigail Barr (University of Nottingham); Anna Hochleitner (Norwegian School of Economics); Silvia Sonderegger (University of Nottingham) |
| Abstract: | We investigate the impact of rising inequality on social instability through a novel multi-period laboratory experiment. In each period, members of two groups are randomly matched to play a battle of the sexes game, in which payoffs are either zero for both or positive but unequal. At the beginning of the game, the implied payoff inequality is low but, under some treatments, it increases over time. While under low initial inequality, unequal conventions – where one group is systematically better off – tend to emerge, increasing inequality causes destabilization. This is initiated by the group disadvantaged by the unequal convention and is particularly pronounced when not only the relative, but also the absolute position of the disadvantaged group is worsening. These findings are consistent with a simple model incorporating disadvantageous inequality aversion. Finally, we show that history matters; responses to current inequality depend on past experiences of inequality and stability. |
| Keywords: | Inequality; Conflict and Revolutions; Coordination; Group Behavior; Experiments |
| Date: | 2026–04 |
| URL: | https://d.repec.org/n?u=RePEc:not:notcdx:2026-04 |
| By: | Rod Falvey (Bond University); Tom Lane (Newcastle University); Shravan Luckraz (University of Nottingham Ningbo China); Shuo Yang (University of Nottingham Ningbo China); Wanjun Zheng (Zhejiang University of Finance and Economics) |
| Abstract: | This paper reexamines the Holmström’s moral hazard in teams for environments where team production problems are repeated a finite number of times. Motivated by the observation that many real-world team interactions are naturally repeated while much of the literature emphasizes static peer-evaluation rules or repeated-game constructions that rely on infinite horizons, we adapt a Galbraith Mechanism (GM)-style allocation mechanism to a finitely repeated principal–agent team production setting and evaluate its performance both analytically and in the laboratory. We construct a repeated GM allocation that mixes equal sharing with proportional rewards, characterize subgame-perfect equilibria that sustain higher aggregate effort than static equal-share benchmarks, and design credible finite-horizon history-dependent punishments to select efficient equilibria when multiplicity arises. A laboratory experiment on a twice-repeated game tests the mechanism’s static and dynamic predictions under alternative allocation treatments. Our results show that groups under the (repeated GM) proportional allocation exert significantly higher effort than under (one-shot GM) equal sharing, consistent with theoretical predictions. Observed behavior indicates increased cooperation with repeated interaction. |
| Keywords: | Team Production; Moral Hazard; Finitely Repeated Games; Galbraith Mechanism |
| Date: | 2026–05 |
| URL: | https://d.repec.org/n?u=RePEc:not:notcdx:2026-05 |
| By: | Erik O. Kimbrough (Chapman University); Brennan McDavid (Chapman University); Diba Vazirian (Chapman University) |
| Abstract: | This paper studies delegation to artificial intelligence in a setting where human principals retain the consequences of delegated choices. Participants wrote prompts instructing ChatGPT-4o mini how to choose on their behalf in three canonical economic domains: risky choice, intertemporal choice, and social allocation. We then elicited the compensation participants required to let the AI’s choices count for payment and compared participants’ own choices to choices generated from their prompts. The design produces two central empirical objects: a revealed measure of reluctance to delegate, captured by willingness to accept compensation for AI delegation, and a behavioral measure of alignment, captured by the share of decisions on which the participant and AI made the same choice. We supplement the original experiment with two benchmarks: a human-agent follow-up in which other participants attempted to implement the same prompts, and an ex post robustness exercise using GPT-5.5. The results show substantial reluctance to delegate despite moderate-to-high alignment, and they suggest that misalignment reflects not only model limitations but also the difficulty of communicating delegable preferences through short natural-language prompts. |
| Keywords: | AI, Trust, Experiments, Preferences |
| JEL: | C91 D91 D83 D81 O33 |
| Date: | 2026 |
| URL: | https://d.repec.org/n?u=RePEc:chu:wpaper:26-11 |
| By: | Adrian Hillenbrand; Hans-Theo Normann; Matthias Potarca; Tobias Werner |
| Abstract: | Rule-based pricing tools are widespread in digital commerce, yet we know little about how their design shapes market outcomes. In a controlled market experiment, participants use dashboards to build pricing algorithms competing in a sequential Bertrand game over multiple periods. We vary design features commonly found in commercial repricing tools: warnings about price wars, pre-configured strategies, and advice from a large language model. Most treatment variations raise market prices with effects driven by an increase in starting prices and more cooperative algorithm designs. The results matter for competition policy, platform regulation and current discussions on regulating algorithm design tools. |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2609.26861 |
| By: | Bah, Tijan (World Bank); Batista, Catia (Nova School of Business and Economics) |
| Abstract: | Irregular migration to Europe by sea, though risky, remains one of the most popular migration options for many Sub-Saharan Africans. This study examines the drivers of irregular migration decisions using an incentivized lab-in-the-field experiment in rural Gambia, the African country with the highest per-capita rate of irregular migration to Europe. Potential migrants substantially overestimate the probability of dying en route, reporting an average belief of 51 percent compared with a best available estimate of about 20 percent. In the experiment, reducing the probability of dying en route from 50 to 20 percent raises willingness to migrate irregularly by 3.0 percentage points, while reducing the probability of obtaining legal residence from 50 to 30 percent lowers it by 2.1 points. Our results highlight the importance of potential migrants' prior beliefs in shaping responses to information and suggest that information policies that fail to account for these beliefs may backfire. |
| Keywords: | international migration, irregular migration, migration intentions, information campaigns, prior updating, migration policy, lab-in-the-field experiment, West Africa |
| JEL: | C93 D83 F22 J17 O15 |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:iza:izadps:dp18966 |
| By: | Tanaka, Mari (University of Tokyo); Okudaira, Hiroko (Doshisha University); Sakka, Mariko (University of Tsukuba); Yamaguchi, Shintaro (University of Tokyo) |
| Abstract: | We test whether employers can shift household allocation by targeting men as fathers. Among 1, 225 male employees in four Japanese organizations, we randomized a two-hour work-life balance training and an information treatment correcting misperceptions of colleagues' support. Training increased fathers' weekend childcare by about one hour per day in young-child households and raised spouses' reported paid work; spouses' housework also fell. Training strengthened beliefs that paternity leave benefits workplace functioning but not leave intentions. Information modestly shifted perceived support but affected neither intentions nor information-seeking. Employer training can affect household allocation without changing formal rights or work arrangements. |
| Keywords: | workplace intervention, paternity leave, household time allocation, second-order beliefs, field experiment |
| JEL: | J13 J16 J22 M54 |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:iza:izadps:dp18913 |
| By: | Bloomfield, Juanita; Balsa, Ana; Cid, Alejandro |
| Abstract: | Despite growing investment in education technology, evidence on how to integrate digital tools eectively into public education systems remains limited. Substituting classroom instruction with computer-aided learning has often failed to improve outcomes, even though schools remain the most direct route to scale educational innovations. This paper studies an alternative approach: delivering personalized educational software to households, where smartphone access is widespread, while experimentally varying school involvement. We evaluate Leo Leo, an adaptive early literacy app for young children, using a randomized controlled trial in 105 public schools in Nuevo León, Mexico. Schools were assigned to either (i) a family-only intervention providing access to the app, (ii) a combined family-and-school intervention that additionally provided teachers and principals with information on childrens progress, or (iii) a control group that received neither the app nor school support. The home-based technology generates robust gains in early literacy, with suggestive positive spillovers to early numeracy, when complemented by school support; the family-only intervention produces no signicant learning gains. Descriptive mechanism analyses are most consistent with greater child engagement with the software as the leading channel. We nd no indication that changes in parental inputs play a role, based on a household survey with a lower response than the student assessment. Overall, the results show that while home-based delivery of educational technology can enhance scalability, institutional involvement remains critical for translating scalable technology into learning gains. |
| JEL: | I21 C93 O15 |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:idb:brikps:14736 |
| By: | Simon Gaechter (University of Nottingham); Jonathan F. Schulz (George Mason University); Christian Thoeni (University of Lausanne) |
| Abstract: | Can common institutions produce different gains from cooperation across culturally diverse environments? We address this question using standardized repeated public-goods experiments, with and without punishment, in 43 diverse countries (n = 3, 876). The design permits direct comparison of voluntary cooperation, conditional cooperation, peer-enforced cooperation, and punishment behavior under identical incentives. We document large cross-societal differences: realized cooperation gains vary by 81 percentage points between the highest- and lowest-gain subject pools. We interpret these patterns through a belief-based model: everyday experiences of cooperation shape expectations about others’ cooperation and the use of punishment. Consistent with model predictions, cooperation and punishment correlate with a cross-country Cooperative Environment Index. Mechanism experiments (n = 1, 092) exogenously vary exposure to strong enforcement. This exposure raises later beliefs and cooperation after enforcement is removed. Together, the findings provide evidence that common institutions can generate unequal gains because their effectiveness depends on expectations about cooperation and its enforcement. |
| Keywords: | Cross-cultural experiments; voluntary cooperation; peer punishment; institutional quality; cultural values; cooperative environment index; ABC model of cooperation; mechanism experiments |
| Date: | 2026–06 |
| URL: | https://d.repec.org/n?u=RePEc:not:notcdx:2026-06 |
| By: | Yuchen Hu |
| Abstract: | A switchback experiment alternates an entire system between treatment and control over time. It is especially useful when interactions between units can undermine standard unit-level experiments. Switchback experiments are commonly implemented on fixed temporal grids, which impose highly structured restrictions on when treatment can switch. We study a broader class of random-duration switchbacks, in which treatment and control alternate across runs whose durations are drawn from a common distribution. Under finite carryover, we show that the mean squared error of the Horvitz-Thompson estimator has a simple frequency-domain representation governed by how temporal outcome patterns align with design-induced imbalance and contamination patterns. This representation yields a tractable worst-case design criterion that can be computed directly from the duration distribution. Optimizing over even a simple two-rate family produces a run-age-dependent switching rule that reduces the asymptotic worst-case mean squared error by at least 32\% relative to the standard independently randomized fixed-block switchback. |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2609.13698 |
| By: | Zihao Chen |
| Abstract: | Online experiments must often be evaluated before long-term outcomes mature. Under rolling enrollment, these outcomes are observed only for early enrollees, while short-term surrogates are available for everyone. We compare seven estimators across eleven data-generating processes, spanning partial mediation, drift, outcome sparsity, and enrollment-time labeling, with up to $R = 2, 000$ replications over more than 500 method-by-scenario cells. We find a sharp robustness-efficiency tradeoff: the surrogate index delivers large efficiency gains when surrogacy holds but its coverage collapses under violations, while PPI-family methods stay asymptotically valid under random labeling at smaller gains. We give the finite-sample variance of PPI++ in the all-units parameterization for a fixed predictor, a joint asymptotic distribution for the two estimators under cross-fitting, and a Hausman-type estimator-disagreement diagnostic, then quantify the detection-damage gap: in the partial-mediation design, where we locate both edges, a band of violations destroys surrogate-index coverage yet is too small to detect on most datasets. On the 64, 000-customer Hillstrom experiment the diagnostic rarely flags a violation that biases the surrogate index, and a Cauchy-kernel hybrid of the two estimators inherits 10.7% relative bias; on the 14-million-user Criteo experiment the violation is detected, and subsampling traces detection turning on with scale as damage persists. PPI++ has limits: at a rare-conversion $n = 30, 000$ Criteo subsample its empirical coverage is 85.0%. We recommend prespecifying PPI++ with the exact variance as the primary analysis under random labeling and adequate labeled outcome counts, reading the diagnostic as a warning, not a certificate, and treating the surrogate index as a sensitivity analysis. |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2609.30528 |
| By: | Shonchoy, Abu (Florida International University and J-PAL); Fujii, Tomoki (Singapore Management University); Raihan, Selim (University of Dhaka) |
| Abstract: | Migration into stable urban manufacturing jobs offers large returns, yet rural-to-urban migration remains limited in developing countries. Using a field experiment in Bangladesh, we sequentially relax constraints on information, skills, liquidity, and migration risk for poor rural youth seeking jobs in the urban apparel sector. Information and training alone have limited effects. By contrast, combining training with a stipend, paid internship, and transition support increases migration, employment, and earnings, reduces poverty, and produces intra-household spillovers. We find that reducing job-search uncertainty and migration risk is central to labor-market entry, and skills-training programs may require complementary financing, migration assistance, and job-search support for scalable impacts. |
| Keywords: | field experiment, internal migration, vocational training, internship, Bangladesh, apparel sector |
| JEL: | J61 O15 R23 |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:iza:izadps:dp18912 |
| By: | Lata Gangadharan (Monash University); Jona Krutaj (University of Nottingham); Maria Claire Villeval (University of Lyon) |
| Abstract: | Heterogeneous returns from contributions to a public good create a normative conflict between equality and efficiency. In a laboratory experiment, we proposed an indicative menu of contribution principles including one featuring a decentralized redistribution mechanism ensuring earnings equality in exchange for fully efficient contributions. Although a majority of individuals, when in the position of an impartial observer, considered this principle to be the most appropriate and expected others to agree, they failed to act on it. Designating a leader who endorsed this principle and made non-binding recommendations enabled a majority of groups to adopt it successfully. This resulted in full contributions and earnings equalization through redistribution from advantaged to disadvantaged members, effectively resolving the conflict. |
| Keywords: | Normative conflict; Redistribution; Efficiency; Leadership; Reciprocity; Experiment |
| Date: | 2026–03 |
| URL: | https://d.repec.org/n?u=RePEc:not:notcdx:2026-03 |
| By: | Sascha O. Becker; Ana Fernandes; Nurlan Lalayev; Doris Weichselbaumer |
| Abstract: | We study gender differences in employers’ responses to commuting distance using data from a large-scale correspondence test in Germany, Switzerland, and Austria. A 10 km increase in driv- ing distance reduces women’s probability of receiving an interview invitation by 1.8 percentage points, around 9 percent of the mean female invitation rate, while commuting distance has no corresponding effect for men. The gender difference is statistically significant and robust to alter- native distance measures and thresholds, firms’ rural or urban location, and local labor-market tightness. The female distance penalty does not vary detectably with applicants’ reported marital status or the presence of children, providing no support for screening based on these observed indicators of household responsibilities. |
| Keywords: | Gender Discrimination, Commuting, Labor Market, Field Experiment, Germany, Switzerland, Austria |
| JEL: | C93 J16 J61 J71 R41 |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:jku:econwp:2026-08 |
| By: | Beknazar-Yuzbashev, George (University of Chicago, The Normal Lab); Capozza, Francesco (UB, IEB, and CESifo); Rutherford, Kylan (New York University); Stalinski, Mateusz (University of Warwick and CAGE,) |
| Abstract: | We study whether content-level moderation that reduces exposure to social media toxicity, holding platform access constant, affects mental health. We deploy, for randomly assigned users, a cross-platform browser extension that hides posts and comments exceeding a prespecified toxicity threshold, and follow N = 664 desktop users over a six-week intervention across two waves-one during the 2024 U.S. election campaign and one outside of it. Contrary to basic intuition, filtering toxicity worsens mental health: treatment raises a standardized index of GAD-7 and PHQ-8 symptoms by 0.15σ, and the share of respondents above a moderate-symptom screening cutoff rises by about 12 pp. A QALY-based calibration implies a welfare cost of about $158 per treated participant. Heightened loneliness and a diminished relative moral self-view emerge as potential channels. We conclude that curbing toxic content can impose unintended psychological costs, even where it yields other social benefits. |
| Keywords: | mental health, toxic content, hate speech, moderation, social media, field experiment JEL Classification: C93, D83, I10, I31, L82, L86, Z13 |
| Date: | 2026 |
| URL: | https://d.repec.org/n?u=RePEc:cge:wacage:817 |
| By: | Sarah Cattan; Pulkit Bajpai; Julian Edbrooke-Childs; Emily Stapley; Monica Costa-Dias; Jenna Jacob; Emily Goodacre; Angelika Labno; Vaishnavi Dhas; Emily Orchard; Erin Mckeaveney; Grizelda Khaling; Anoushka Kapoor; Ashley New; Imran Rasul |
| Abstract: | Concerns around the exposure of young people to violence lie at the top of the UK policy agenda. We evaluate a flagship violence-reduction intervention in London, known as Your Choice, that targets adolescents at risk of exposure to violent crime. The intervention trains youth practitioners in CBT, who deliver these techniques to adolescents over 12-18 weeks. Delivery is embedded within pre-existing services that London Local Authorities are statutorily required to provide, so accounting for constraints to scale-up from the outset. Many challenges to retaining voltage at-scale relate to system-wide factors that are hard to measure quantitatively. We therefore combine qualitative and quantitative evidence to understand intervention impacts moving from an internal pilot to a full scale efficacy trial. While the pilot showed the promise of Your Choice, the intervention faced challenges to scale. We document four reasons for this: (i) initial pilot participants differ from those recruited at scale-up; (ii) uncertainty over future funding and staff turnover; (iii) binding workload capacity constraints for practitioners; (iv) a rise in per participant intervention costs. We derive implications for future evaluations of similar interventions in public services that operate near full capacity, to help build the evidence base on how to overcome these fundamental sources of voltage drops. |
| Date: | 2026–01–30 |
| URL: | https://d.repec.org/n?u=RePEc:bri:uobdis:26/852 |
| By: | Patrick Baylis; Judson Boomhower; Bob Horton; Chris Mulverhill |
| Abstract: | Losses from natural disasters can be mitigated through investments in property protection, yet adoption of low-cost risk-reducing measures remains low. In wildfire-exposed areas, managing vegetation and other fuels next to homes decreases structure vulnerability, but few properties comply with best practices despite widespread outreach efforts. We report on two randomized field experiments that incentivized thousands of Oregon homeowners to participate in pre-existing fire department programs to reduce wildfire risk. Across both experiments, financial incentives of $100 to $500 resulted in up to 12% of homeowners scheduling free defensible space assessments, while information-only approaches achieved assessment takeup rates of 3% at most. Responses to subsidies are largest among owners of more valuable homes, while baseline wildfire risk does not predict responsiveness. On-the-ground and remotely-sensed measures of vegetation provide noisy estimates of treatment effects on endline vegetation. Finally, we consider the extent to which the estimated behavioral elasticities from these experiments might shed light on potential homeowner responses to pricing of wildfire risk reductions in home insurance premiums, a practice that is becoming more common due to technological and regulatory changes. |
| JEL: | C93 Q54 Q58 |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:nbr:nberwo:35764 |
| By: | Becker, Sascha (University of Warwick); Fernandes, Ana (Bern University of Applied Sciences); Lalayev, Nurlan (University of Warwick); Weichselbaumer, Doris (University of Linz) |
| Abstract: | We study gender differences in employers’ responses to commuting distance using data from a large-scale correspondence test in Germany, Switzerland, and Austria. A 10 km increase in driving distance reduces women’s probability of receiving an interview invitation by 1.8 percentage points, around 9 percent of the mean female invitation rate, while commuting distance has no corresponding effect for men. The gender difference is statistically significant and robust to alternative distance measures and thresholds, firms’ rural or urban location, and local labor-market tightness. The female distance penalty does not vary detectably with applicants’ reported marital status or the presence of children, providing no support for screening based on these observed indicators of household responsibilities. |
| Keywords: | gender discrimination, commuting, labor market, field experiment, Germany, Switzerland, Austria |
| JEL: | C93 J16 J61 J71 R41 |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:iza:izadps:dp18950 |
| By: | Busso, Matias (Inter American Development Bank); Gonzalez-Velosa, Carolina (Inter American Development Bank); Muñoz-Morales, Juan (IÉSEG School of Management); Ruiz, Juanita (Georgetown University); Valencia, Juan (Inter American Development Bank); van der Werf, Cynthia (Inter American Development Bank) |
| Abstract: | This paper studies whether government-issued skill certification improves labor market integration when workers' skills are hard to verify. We evaluate a randomized offer of access to an occupation-specific certification program in Colombia among locals and migrants, two groups with different ex ante signaling capacity. The program certifies existing skills rather than providing training. Among locals, the offer increases employment by reducing inactivity and raises hours worked. The effects persist unevenly over time and are robust to attrition bounds and alternative outcome measures. Among migrants, estimates are less precise and provide no evidence of effects in employment, hours, or wages. The results show that a common skill certification generates different patterns of labor-market adjustment among locals and migrants. |
| Keywords: | signaling, certifications, skills, migration, Colombia |
| JEL: | J24 J61 J64 I26 O15 D82 |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:iza:izadps:dp18963 |
| By: | Isaac Cheah; Ethan Pancer; Jianping Huang; Aimee Pink; Béatrice Parguel (DRM - Dauphine Recherches en Management - Université Paris Dauphine-PSL - PSL - Université Paris Sciences et Lettres - CNRS - Centre National de la Recherche Scientifique); Fong Keng-Highberger |
| Abstract: | Research examining artificial intelligence's (AI) role in food choice largely asks whether AI helps people make better decisions, treating it as an information source that improves the accuracy or personalization of recommendations. We suggest that this framing captures one point on a broader continuum of how AI impacts food decisions. AI can act, often simultaneously, as an advisor consulted for information, a decision partner that actively shares in constructing the choice, or an ambient influence that shapes what appears desirable and available before deliberate choice begins. As AI becomes more conversational and more embedded across the food environment, this continuum is increasingly populated in less visible regions, yet the field remains concentrated on the advisor role. We focus on the decision-partner role, where AI participates in value construction, goal clarification, attribute weighting, and choice justification, raising questions about agency, responsibility, and delegation. We articulate four mechanisms that describe this participation, namely preference offloading, responsibility reallocation, confidence calibration, and food-decision skill erosion. Because food choice is frequent, often habitual, and commonly involves trade-offs among hedonic and utilitarian goals, it is an ideal domain in which these process-level effects can compound. We draw on evidence from consumer psychology, automation research, and cognitive science, noting where boundary conditions remain untested in food contexts, and close with a research agenda organized around agency, reliance, responsibility, confidence, and cognitive effort. When AI shares in constructing a food choice, the values being optimized may no longer be entirely the consumer's own. |
| Keywords: | Artificial intelligence, Food choice, Decision-making, Consumer psychology, Algorithm reliance |
| Date: | 2027–01 |
| URL: | https://d.repec.org/n?u=RePEc:hal:journl:hal-05752256 |
| By: | Joshua T. Dean; Christine L. Exley; Muriel Niederle; Heather Sarsons |
| Abstract: | Gender-neutral attitudes judge a behavior the same way regardless of whether a man or a woman does it. Yet surveys of gender attitudes typically ask respondents to evaluate statements about men's or women's behaviors in isolation, never eliciting how the same behavior is evaluated when assigned to the other gender. Such responses therefore cannot reveal whether respondents perceive society as holding gender-neutral views. We introduce paired (role-reversed) elicitation, which varies the gender assigned to otherwise identical behaviors. In a representative sample of Americans, responses to standard statements suggest that few respondents perceive society as holding traditional views toward women. In isolation, this may be seen as evidence that society has moved past gender roles that confine choices. The paired elicitation, however, reveals a gender-neutrality gap: only about half of respondents report the same beliefs across the standard and role-reversed statements, even though these statements only vary in terms of which gender is assigned to which behavior. Thus, a substantial portion of respondents believe that the social acceptability of different actions depends on gender. Paired responses also uncover substantial heterogeneity and reveal broader underlying beliefs driving responses, such as beliefs about work and family rather than gender-specific views. |
| JEL: | D1 J16 |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:nbr:nberwo:35761 |
| By: | Baader, Malte (Dept. of Economics, Norwegian School of Economics and Business Administration); Barron, Kai (ESCP Business School); Hochleitner, Anna (Centre for Applied Research, Norwegian School of Economics and Business Administration); Kaiser, Jonas P. (The Technical University of Berlin) |
| Abstract: | In two studies using AI-assisted qualitative interviews, we ask a representative sample of the German electorate (N = 1, 013) to reflect on the most important societal events of the last 15 years. We document that far-right supporters construct systematically different narratives about the recent past. First, they recall different events—with a strong focus on migration-related issues. Second, even when they do consider the same events, they are more likely to blame the establishment and describe events more negatively and with greater emotional intensity. We further document that these narratives are associated with policy preferences, and that far-right supporters are more inclined than supporters of other parties to prioritize cultural over economic issues. Together, our findings reveal that partisan divides extend to how citizens remember and interpret shared national experiences. |
| Keywords: | Polarization; Narratives; AI-Assisted Qualitative Interviews; Explanations; Mental Models |
| JEL: | C90 D01 D84 |
| Date: | 2026–09–18 |
| URL: | https://d.repec.org/n?u=RePEc:hhs:nhheco:2026_014 |
| By: | Ritsu Yano; Yoshiyuki Nakazono; Jun Takahashi |
| Abstract: | Many central banks, including the Bank of Japan, define price stability as 2% inflation. Do households agree? We answer this question with a conjoint experiment in which Japanese respondents chose between hypothetical economies that differed in their inflation and unemployment rates. We find that households prefer zero inflation. An economy with 2% inflation is chosen significantly less often than one with 0% inflation. On average, respondents are indifferent between 0% and −2% inflation, although men and younger respondents prefer 0% to deflation. We also find that households weigh unemployment more heavily than inflation: for a one-percentage-point fall in inflation, they accept a rise in unemployment of only about 0.65 percentage points. The results point to a gap between the inflation rate households prefer and the 2% that the Bank of Japan targets. This gap raises the possibility that the weak anchoring of Japanese households’inflation expectations at 2% partly reflects their preference for inflation closer to zero. |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:toh:tupdaa:90 |
| By: | Yi-Lin Tsai (Arvin); Yung-Hsiu (Arvin); Lai |
| Abstract: | Marketers now deploy generative AI agents as synthetic consumers to pretest visual assets such as logos, packaging, and advertising at a fraction of human-panel cost. However, this procedure assumes that a model seeing a visual cue can also perceive its consumer meaning, which is largely untested. We stress-test the assumption using six canonical visual marketing experiments, varying the two levers managers control: model generation (GPT-4o-mini vs. GPT-5.4-mini) and input format (plain text vs. JSON). Every resulting configuration passed the manipulation checks; however, none of the configurations reproduced more than two of the six human effects, and the remainder were nonsignificant. The one exception was a significant reversal of the human pattern. Providing conceptual or empirical evidence through in-context learning steers average responses toward the human effect. Yet steering has a limit: even when it succeeds, a configuration reproduces less than half of the natural spread of human responses and so understates consumer heterogeneity. We integrate these results into an AI governance protocol (Calibrate, Intervene, Deploy) that delineates when synthetic consumers can responsibly screen creatives and when human panels remain necessary. |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2609.25677 |
| By: | Malte Baader (Norwegian School of Economics, FAIR); Kai Barron (ESCP Business School, WZB Berlin); Anna Hochleitner (Norwegian School of Economics, FAIR, SNF); Jonas P. Kaiser (TU Berlin, Berlin School of Economics) |
| Abstract: | In two studies using AI-assisted qualitative interviews, we ask a representative sample of the German electorate (N = 1, 013) to reflect on the most important societal events of the last 15 years. We document that far-right supporters construct systematically different narratives about the recent past. First, they recall different events—with a strong focus on migration-related issues. Second, even when they do consider the same events, they are more likely to blame the establishment and describe events more negatively and with greater emotional intensity. We further document that these narratives are associated with policy preferences, and that far right supporters are more inclined than supporters of other parties to prioritize cultural over economic issues. Together, our findings reveal that partisan divides extend to how citizens remember and interpret shared national experiences. |
| Keywords: | polarization; narratives; ai-assisted qualitative interviews; explanations; mental models; |
| JEL: | D01 C90 D84 |
| Date: | 2026–09–18 |
| URL: | https://d.repec.org/n?u=RePEc:rco:dpaper:586 |
| By: | Bugarčić, Milica; Hanić, Hasan |
| Abstract: | The paper presents the results of a meta-analysis of empirical literature focused on the study of behavioral biases in financial decision-making. The findings indicate that biases such as loss aversion, present bias, status quo bias, overconfidence, and herding systematically shape both individual and collective financial behavior. Their influence is evident across diverse institutional and cultural contexts, confirming that behavioral deviations are structural rather than incidental features of economic behavior. The study highlights the need to integrate longitudinal and neuroeconomic approaches to achieve a deeper understanding of the interaction between cognitive, emotional, and contextual factors in decision-making processes. By incorporating psychological dimensions into economic theory, behavioral economics provides a more comprehensive and realistic framework for interpreting market dynamics and enhancing the quality of financial decision-making. |
| Keywords: | behavioral economics, behavioral finance, financial decion-making, cognitive biases |
| JEL: | D14 G53 I22 G41 |
| Date: | 2025 |
| URL: | https://d.repec.org/n?u=RePEc:zbw:esconf:343874 |
| By: | Matussek, Martin; Wendland, René; Hendriks, Patrick |
| Abstract: | This study examines how artificial intelligence (AI), specifically large language models (LLMs), influences social loafing in group work. Social loafing describes reduced individual effort in teams. To explore this in AI-supported settings, 28 participants were placed into seven groups of four and completed a validated brainstorming and evaluation task on a text-based platform (Discord) with assistance from an LLM chatbot. The results show that AI does not inherently increase social loafing. While some participants offloaded effort onto the AI, most used it strategically to extend and refine their own ideas. Qualitative findings indicate that AI changes key mediators of social loafing, such as responsibility diffusion and group cohesion, depending on how it is perceived and engaged with. These results challenge assumptions that AI necessarily causes disengagement. Instead, AI reshapes, rather than replaces, human effort, creating both risks and opportunities for future collaborative work. |
| Date: | 2026–07 |
| URL: | https://d.repec.org/n?u=RePEc:dar:wpaper:162316 |
| By: | Zhenhao Fu; Ruipeng Xu; Qibing Ren |
| Abstract: | Individually protective decisions can produce avoidable collective failures. As large language model (LLM) agents take on greater roles in financial decision-making, financial AI safety must therefore be considered not only at the level of individual agents, but also at the level of the systems they jointly create. We study this problem with FRAIL, a controlled experimental framework that places LLM agents in three dynamic financial environments---bank runs, debt rollover, and reward crowdfunding---where agents' decisions reshape the financial conditions faced by others. Across seven leading LLMs, we find widespread collective fragility even when no agent is instructed to destabilize the system: 77\% of baseline bank-run episodes and 83\% of debt-rollover episodes end in failure. We then compare three interaction mechanisms based on compensated commitments, centralized commitment agreements, and participant-led coalitions. All three improve aggregate outcomes, but no single mechanism performs best across all financial structures. Across mechanisms, successful stabilization shares a common temporal pattern: broad commitment forms early, before defensive behavior becomes self-reinforcing. Our findings show that individually capable agents do not automatically form safe financial systems, highlighting system-level evaluation and interaction design as central problems for financial AI safety. Code is available at https://anonymous.4open.science/r/FinFra il-CF26. |
| Date: | 2026–09 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2609.30940 |