nep-exp 2026-08-31 papers

This nep-exp issue is ©2026 by Daniel Houser. It is provided as is without any express or implied warranty. It may be freely redistributed in whole or in part for any purpose. If distributed in part, please include this notice.

General information on the NEP project can be found at https://nep.repec.org. For comments please write to the director of NEP, Marco Novarese at <director@nep.repec.org>. Put “NEP” in the subject, otherwise your mail may be rejected.

NEP’s infrastructure is sponsored by the Griffith Business School of Griffith University in Australia.

By:	Francesco Capozza; Robert Dur; Pedro Rey-Biel
Abstract:	This paper examines public acceptability of randomized experimental research in economics using a preregistered conjoint experiment with 1, 471 U.S. respondents. Acceptability varies substantially across domains: health-promoting interventions rank highest, while covert political advertising falls below the scale midpoint. Participant vulnerability is the dominant design attribute, followed by country of implementation and funding source. We also ask who should set ethical standards for experimental research, finding that the public assigns greatest governance legitimacy to experiment participants, citizens, and scientists and least to elected representatives and lawyers. We complement these findings with a companion study of research-active economists (N = 461) across the same experimental scenarios. Economists rate experiments substantially higher overall, preserve the public’s acceptability hierarchy, and assign the highest governance legitimacy to scientists rather than experiment participants. Our findings have direct implications for IRB design, research communication, and the political economy of evidence-based policymaking.
Keywords:	experiments, preferences, IRB, economists
JEL:	A11 C81 C93 D83
Date:	2026
URL:	https://d.repec.org/n?u=RePEc:ces:ceswps:_12888

By:	Lee, Gyuchan
Abstract:	This paper examines how consumers respond to different forms of soda tax information under tax-inclusive pricing. Using a randomized pre–post discrete choice experiment (DCE) in China, we isolate informational effects while holding tax-inclusive prices constant. Difference-in-differences comparisons and mixed logit estimates show that informing consumers that the displayed soda price includes a soda tax reduces regular soda demand relative to a price-only control group, while explicit disclosure of the tax amount produces larger demand reductions and higher implied price elasticity. The estimates suggest that tax information primarily reduces the perceived attractiveness of regular soda, whereas explicit tax disclosure increases price sensitivity. These findings suggest that tax-inclusive pricing alone may not ensure full tax salience and that disclosure design may complement corrective taxation beyond the standard price effect.
Keywords:	Food Consumption/Nutrition/Food Safety
Date:	2026
URL:	https://d.repec.org/n?u=RePEc:ags:aaea26:404547

By:	Andries, Marianne; Bursztyn, Leonardo; Chaney, Thomas; Djourelova, Milena; Imas, Alex
Abstract:	We explore the mechanics of empathy. We show that information about an outgroup can activate and magnify empathy when presented in conjunction with an experience simulating their struggles. This response increases the willingness to help the struggling group. We provide evidence for this effect in an immersive virtual reality experiment where participants (``witnesses") experience a simulation of the struggle of unauthorized migrants (``protagonists"), then replicate these results in a series of controlled lab experiments. We show that information enhances the witnesses' empathetic response and drives them to engage in more prosocial behavior when it increases their perceived interpersonal similarity, or relatability to the protagonist---an effect we trace to attention: eye-tracking data reveals that information provision concentrates witnesses' gaze on the struggles of the protagonist instead of searching through peripheral elements of the scene. Conversely, only information packages that strengthen perceived relatability---an effect that can vary across subgroups with heterogeneous attributes---magnify empathy. Together, our evidence suggests that the ability to put oneself in the shoes of another person or group can be enhanced by activating empathy through simple, targeted, information provision.
Keywords:	Empathy; Similarity; Immigration
JEL:	C9 D8 D9 F22 Z1
Date:	2024–07
URL:	https://d.repec.org/n?u=RePEc:cpr:ceprdp:19197

By:	Brian Jabarian; Luca Henkel
Abstract:	This paper studies whether AI automation can improve organizational outcomes by reducing variance when collecting information. We conducted a large-scale natural field experiment in which 70, 000 job applicants were randomly assigned to be interviewed by human recruiters or AI voice agents. In both conditions, human recruiters evaluate the interviews and make hiring decisions. Applicants interviewed by AI agents are 12% more likely to receive job offers, and these gains translate into higher job starts and worker retention, with no decline in the productivity of hired workers. Analyzing interview transcripts reveals that AI voice agents achieve controlled variance: their interviews are more structured and consistent while remaining responsive to individual applicants, which is associated with more hiring-relevant information collected. These results demonstrate that automating information collection with AI can enhance decision quality through standardization.
Date:	2026–07
URL:	https://d.repec.org/n?u=RePEc:arx:papers:2607.28222

By:	Joshua Greubel; Henrik Guhling; Fabian Herweg
Abstract:	In an online experiment, we study how generative AI steers individuals toward dominated choices and, conversely, helps them avoid such mistakes. Participants choose between two virtual index funds, one of which strictly dominates the other. An AI chatbot increases optimal choices by over 20 percentage points when promoting the optimal fund but reduces them by almost 30 points when promoting the dominated fund. Its influence is undiminished when framed as bank-provided despite a disclosed interest. Incentivized human advice steers choices toward the promoted fund but is significantly less effective. Transcript analyses tentatively suggest that AI's advantage reflects more persuasive argumentation.
Keywords:	AI chatbots, conflicts of interest, dominated choices, financial advice, generative AI, persuasion, retail investors
JEL:	C91 D14 G11
Date:	2026
URL:	https://d.repec.org/n?u=RePEc:ces:ceswps:_12925

By:	Thomas W. Graeber; Benjamin Enke
Abstract:	Economic decisions are often shaped by comparison points such as reference points, goals, expectations and peers. A long-standing puzzle is that decisions sometimes increase in these comparators (assimilation) and sometimes decrease in them (contrast). We develop a simple taxonomy that predicts the sign. The idea is that when absolute evaluations - translating a decision input into an output - are difficult, people use comparison points as information about how to navigate tradeoffs. This logic predicts that decisions decrease in input comparators - those tied to exogenous decision inputs (e.g., a typical hourly wage) - and increase in output comparators - those tied to endogenous decisions or outcomes (e.g., typical daily earnings). We test and confirm these predictions in experiments on labor supply, investment and belief updating, involving both social and expectations-based comparison points. We further show that comparison effects weaken both when decision uncertainty is experimentally reduced and when comparators are less informative, consistent with a mechanism of comparisons as information. Finally, a classification of the prior literature suggests that the input-vs.-output taxonomy organizes the sign of comparison effects across more than 100 experimental and field observational studies.
Keywords:	comparison effects, reference dependence, uncertainty, experiments
JEL:	D90
Date:	2026
URL:	https://d.repec.org/n?u=RePEc:ces:ceswps:_12924

By:	Elia Antoniou
Abstract:	This paper examines the source of algorithmic aversion, defined as the unwillingness to accept advice or decisions made by algorithms. Algorithmic aversion is conceptualized as a form of individual partiality, driven either by belief-based factors attributed to differences in perceived ability, or by preference-based factors which reflect a disamenity associated with selecting algorithms. To empirically test the predictions of the model, a preregistered online experiment was conducted, where participants evaluated answers to objective and subjective economic questions, with varying information on whether the source was human or algorithm. The results provide no evidence of algorithmic aversion in the evaluation task: participants did not systematically favor human-generated answers over algorithm-generated ones. These results suggest that algorithmic aversion may not be as robust or uniform as previously assumed.
Keywords:	Algorithmic Aversion; Human-AI interaction; Experimental Economics.
JEL:	D83 D90 C91
Date:	2025–11–25
URL:	https://d.repec.org/n?u=RePEc:ucy:cypeua:05-2025

By:	Roberto Galbiati (ECON - Département d'économie (Sciences Po) - Sciences Po - Sciences Po - CNRS - Centre National de la Recherche Scientifique); Emeric Henry (ECON - Département d'économie (Sciences Po) - Sciences Po - Sciences Po - CNRS - Centre National de la Recherche Scientifique); Nicolas Jacquemet (CES - Centre d'économie de la Sorbonne - UP1 - Université Paris 1 Panthéon-Sorbonne - CNRS - Centre National de la Recherche Scientifique, PSE - Paris School of Economics - UP1 - Université Paris 1 Panthéon-Sorbonne - ENS-PSL - École normale supérieure - Paris - PSL - Université Paris Sciences et Lettres - EHESS - École des hautes études en sciences sociales - CNRS - Centre National de la Recherche Scientifique - INRAE - Institut National de Recherche pour l’Agriculture, l’Alimentation et l’Environnement - ENPC - École nationale des ponts et chaussées - IP Paris - Institut Polytechnique de Paris); Itzhak Rasooly (University of London [London])
Abstract:	In this paper, we report the results of two experiments that attempt to elicit social image effects. The first experiment (N=1, 252) provides little evidence that individuals behave in more ‘prosocial' ways when their choices are disclosed to other participants. If anything, imposing observability appears to make the participants slightly less prosocial, although this effect is not statistically significant. The second experiment (N=750) generates similar results and further suggests that our results are not dependent on the omission (or inclusion) of ranking information. We discuss why our experiments fail to generate the results that we had expected and why our results differ from those in the published literature.
Keywords:	Social image, Laboratory experiment
Date:	2026
URL:	https://d.repec.org/n?u=RePEc:hal:cesptp:halshs-05719080

By:	Holden, Stein T. (Centre for Land Tenure Studies, Norwegian University of Life Sciences); Tione, Sarah (Centre for Land Tenure Studies, Norwegian University of Life Sciences)
Abstract:	We study how trust and trustworthiness respond to social distance and how beliefs, norms, and cognitive precision shape social exchange in a field experiment with irrigation farmers. Using a standard trust game with within-subject variation in social-distance framing, we distinguish interactions with an anonymous partner from the same irrigation block from interactions with a partner from another scheme within the same district. In addition to trust and reciprocal behavior, we elicit beliefs about expected returns, stated moral obligations to reciprocate, risk preferences, and decision precision measured through consistency in multiple price list choices. <p> The results reveal a clear asymmetry between trust and trustworthiness. Changes in trust across social-distance framings are primarily driven by changes in beliefs, whereas trustworthiness is more strongly associated with normative commitments and exhibits greater behavioral stability across contexts. Cognitive precision is positively related to baseline levels of trust but is associated with smaller framing-induced changes in trustworthiness, consistent with an interpretation of precision as stabilizing behavior rather than uniformly increasing prosociality. Risk preferences play a limited role once beliefs are taken into account.<p>By separating trust from trustworthiness and by distinguishing between behavioral levels and framing-induced changes, the paper provides new field evidence on the mechanisms underlying social exchange. The findings highlight the importance of beliefs for trusting behavior, norms for reciprocal behavior, and decision precision for the consistency of social decisions across contexts.
Keywords:	Social distance; Beliefs; Norms; Decision precision; Trust; Trustworthiness
JEL:	C92 C93 D01 D91
Date:	2026–08–05
URL:	https://d.repec.org/n?u=RePEc:hhs:nlsclt:2026_007

By:	Holden, Stein T. (Centre for Land Tenure Studies, Norwegian University of Life Sciences)
Abstract:	This paper examines how decision imprecision affects the measurement of risk and time preferences in Multiple Price List (MPL) tasks. Using experimental data with alternative elicitation procedures and repeated binary choices, we document systematic relationships between inconsistency in choices and the distribution of elicited switching behavior.<p> We show that lower decision precision compresses switching behavior toward intermediate rows, reducing the prevalence of corner outcomes. In the risk domain, this compression shifts choices relative to the risk-neutral threshold, increasing the likelihood that observations are classified as risk-loving in low-probability environments. Consistent with this, decision precision affects both the level and probability sensitivity of inferred risk premia. <p>In the time domain, lower precision is associated with lower estimated discount rates, implying that less precise individuals appear more patient in elicitation data. <p>Across domains, these results indicate that commonly used measures such as risk premia and discount rates reflect not only underlying preferences but also variation in decision precision and the structure of elicitation tasks. We therefore interpret these measures as reduced-form summaries of switching behavior rather than structural preference parameters. The evidence is based on a large population-based field experiment in a developing country context, providing external validity to recent findings on cognitive imprecision that have primarily been established in smaller laboratory samples.
Keywords:	Multiple price lists; Decision precision; Stochastic choice; Risk preferences; Time preferences
JEL:	C91 C93 D81 D91
Date:	2026–08–05
URL:	https://d.repec.org/n?u=RePEc:hhs:nlsclt:2026_006

By:	Louis Fréget; Ariell Reshef; Claudia Senik
Abstract:	Generative AI is increasingly promoted as a remedy for loneliness. We test this claim in a large-scale pre-registered experiment on a large sample of French adults. Participants assigned to treatment were asked to hold at least one personal conversation a day with a generative AI tool of their choice for 28 days; the control group continued as usual. The intervention raised socio-emotional AI use substantially. This led to an increase in subjective loneliness, a decline in life satisfaction, and a rise in depressive affect and in objective isolation, with more dinners taken alone. The effects are consistently adverse across outcome categories.
Keywords:	personal AI conversations, generative AI, subjective well-being, loneliness, social connectedness
JEL:	C93 I31 I12 O33
Date:	2026
URL:	https://d.repec.org/n?u=RePEc:ces:ceswps:_12882

By:	Rigissa Megalokonomou; Juliana Silva-Goncalves; Roel van Veldhuizen
Abstract:	We study the role of self-promotion and career advice in sustaining gender differences in labor market outcomes. We conduct a pre-registered experiment in which "advisers" advise "workers" to attempt either a more or a less ambitious task. We find that women promote themselves less than men and, as a result, are 12 percentage points less likely to be advised to choose the more ambitious task. This gender gap in advice persists across both quantitative and qualitative self-assessments and is robust to variation in advisers' information sets-including when advisers observe workers' actual performance-but is eliminated and even reversed when advisers are informed of the gender gap in self-promotion.
Keywords:	Advice, Gender, Self-Promotion, Randomized Experiment
JEL:	C91 D91 J16
Date:	2026–07
URL:	https://d.repec.org/n?u=RePEc:crm:wpaper:26189

By:	Cristina Cattaneo; Daniela Grieco; Nicola Lacetera; Mario Macis
Abstract:	We conducted a survey experiment with over 8, 000 participants in Germany, Hungary and Italy to examine how information about the economic impact of immigration affects the support for more open immigration policies. Respondents who received projections showing higher future incomes under more open immigration policies reported greater support for these policies. The treatment also partly changed how prior attitudes affected policy preferences. Structural estimates suggest that majority support for immigration above current levels would require annual per-capita income gains of about 1.2% in Italy and 1.7% in Germany, and still larger or unattainable gains in Hungary. These increases are large relative to recent historical growth rates, suggesting that economic benefits alone are unlikely to generate vast support for more open immigration policies.
Keywords:	immigration policy, survey experiments, public opinion.
JEL:	C99 D02 D64 D91 F22 J15 P16
Date:	2026
URL:	https://d.repec.org/n?u=RePEc:ces:ceswps:_12879

By:	Philip Oreopoulos; Michael Liut; Alp Sungu; Nina Low
Abstract:	Large language models (LLMs) can make tutoring more scalable, but only if students use them to reason through mistakes rather than avoid effort. We study this question in a randomized field experiment with more than 6, 000 middle-school students in Hamilton County Schools using NUMI, a research-based computer-assisted learning (CAL) platform. Students were randomized to AI versus CAL-only support, mastery versus non-mastery progression, and one of two math topics. One week later, they completed a delayed assessment covering practiced and unpracticed material. AI students progressed more slowly and attempted fewer questions, but answered more accurately conditional on reaching an attempt. The clearest mechanism appeared after mistakes: AI improved next-attempt correctness and reduced attempts needed to return to a correct answer, while increasing time spent on each question receiving structured support. Mastery increased three-correct-in-a-row attainment but did not by itself improve delayed learning. The most encouraging delayed-test evidence appears when AI is embedded in the mastery workflow, with marginally significant gains concentrated on practiced Exercise 1 material. The results suggest that LLM tutoring can add value over standard CAL, but its value depends on structure that turns mistakes into productive learning moments.
JEL:	C93 I21 I24 I28 O33
Date:	2026–08
URL:	https://d.repec.org/n?u=RePEc:nbr:nberwo:35621

By:	Etienne Dagorn; Isaac Parkes; Alberto Prati
Abstract:	Two teachers grading the same student can reach very different conclusions— yet little is known about which teacher characteristics drive such disagreement. We investigate this question through a large-scale online experiment in which over 1, 600 French secondary-school teachers independently evaluated the same set of fictitious student transcripts, generating hundreds of ratings per transcript. This design allows us to decompose grade variation into a component attributable to the student (transcript fixed effects) and one attributable to the teacher (rater fixed effects). After controlling for transcript fixed-effects, we find that teacher leniency accounts for about 40% of the remaining grade variance. These differences are largely unrelated to teachers' gender, subject, teaching style, or job satisfaction, but are positively associated with altruism (measured via an incentivized dictator game) and exhibit a U-shaped relationship with age, with middle-aged teachers being the most strict. An incentivized task reveals that most teachers perceive themselves as more lenient than their colleagues - in particular women - but middle-aged teachers less so. Finally, teachers' stated tolerance for inter-rater disagreement comfortably exceeds the disagreement we actually observe - in contrast with findings from other professional domains. These results speak to the design of fair assessment panels and suggest that differential leniency, while substantial, may be partly predictable and therefore correctable.
Keywords:	Teacher grading, grading bias, teacher leniency, noise audit, inter-rater disagreement, experimen
Date:	2026–08–12
URL:	https://d.repec.org/n?u=RePEc:cep:cepdps:dp2204

By:	Melis Kartal; Edoardo Cefala; Sylvia Kritzinger; Jean-Robert Tyran
Abstract:	We combine a simple theoretical framework with a survey experiment to investigate how individuals sort fact from fiction and update their assessments in societally important but politicized topics, such as climate change and public health. We show that cognitive ability improves news discernment and in particular, the ability to make correct assessments that contradict one’s prior issue positions. However, when we disaggregate results by topic, higher cognitive ability sometimes amplifies motivated decision-making rather than attenuating it. Our framework and findings suggest a hard–easy pattern: more complex issues elicit greater motivated reasoning; simpler ones elicit less. In addition, overconfidence reduces responsiveness to new information, contributing to the persistence of misperceptions. Higher institutional trust appears to weaken motivated reasoning, suggesting that institutional quality may play an important role in constraining opinion polarization.
Keywords:	misinformation, cognitive ability, motivated reasoning, overconfidence, online survey experiment, institutions
JEL:	D72 D83 D91 I21 L82
Date:	2026
URL:	https://d.repec.org/n?u=RePEc:ces:ceswps:_12932

By:	Joaquín Gómez-Miñambres (Lafayette College and Economic Science Institute, Chapman University); Brice Corgnet (emlyon business school); Roberto Hernán González (Burgundy School of Business)
Abstract:	We study how workersâ€™ non-binding self-set goals affect their production and whether goal nudges (ability-calibrated goal hints) make goal setting more effective. In a real-effort task, participants first complete an ability pretest and are then assigned to one of three treatments: a no-goal baseline, a personal goal treatment with no external hint, or a personalized goal nudge that participants may accept or modify. We document a positive link between goals and production. However, merely asking workers to set goals is ineffective: when left unguided, workers anchor on conservative goals well below those induced by goal nudges, leading to underperformance. In contrast, goal nudges increase production and substantially reduce on-the-job leisure, hence inducing higher effort and engagement. These gains are concentrated in early periods, when workers are most uncertain about their own ability, and attenuate as they accumulate task experience. We develop a theory in which this effect operates through self-efficacy. Goal nudges act as informative signals that increase workersâ€™ beliefs about their likelihood of success, thereby raising self-set goals and, in turn, increasing effort. Our findings suggest that personalized goal nudges offer organizations a low-cost complement to monetary incentives, particularly valuable for workers who are new to a task or role.
Keywords:	non-monetary incentives, goal setting, self-efficacy, goal-nudges, laboratory experiments
JEL:	C91 D23 D83 M54
Date:	2026
URL:	https://d.repec.org/n?u=RePEc:chu:wpaper:26-10

By:	Erik Berwart; Sean K. Higgins; Sheisha Kulkarni; Santiago Truffa
Abstract:	How do biased beliefs about the interest rate distribution affect search, negotiation, and loan terms in consumer credit markets? In collaboration with Chile's financial regulator, we conducted a randomized controlled trial with 112, 063 loan seekers. Randomly eliciting beliefs about interest rates led participants to search more and obtain lower rates. Most participants underestimated both interest rate levels and dispersion. Showing them a price comparison tool we built using administrative data on the universe of consumer loans caused them to update their beliefs, negotiate more, obtain lower rates without searching more, and be more likely to take out a loan.
JEL:	D14 D18 D83 G51
Date:	2026–08
URL:	https://d.repec.org/n?u=RePEc:nbr:nberwo:35605

By:	Marty Haoyuan Chen; Ginger Zhe Jin
Abstract:	We study voluntary disclosure when partially naive receivers observe a public signal correlated with the sender’s private information. We develop a framework in which the public signal changes how receivers interpret silence, and derive a closed-form disclosure threshold nesting classical unraveling. A laboratory experiment varying the correlation confirms our predictions: disclosure decreases as the signal becomes more favorable, and receiver guesses upon nondisclosure rise toward it. The public signal can crowd out disclosure, leaving receivers less informed than without it. This informational loss peaks at intermediate correlation, where the signal shifts beliefs yet remains too noisy to substitute for disclosure.
JEL:	C91 D82 D83 D91 L15
Date:	2026–08
URL:	https://d.repec.org/n?u=RePEc:nbr:nberwo:35625

By:	Abdurrahman Bekir Aydemir; Giacomo Battiston; Riccardo Franceschin; Asli Ebru Sanlıtürk
Abstract:	Do migrants want to conform to the host country social norms? We examine this question in the fertility domain with an RCT evaluating the impact of receiving information about native fertility norms on the fertility preferences, intentions, and beliefs of forced migrants. The treatment reduces migrants' expectations about natives' actual and desired fertility. Information lowers the probability of intending to have a child within two years and within five years, and the ideal number of children, both of which are correlated with future pregnancies among sample households, and affects concerns about stigmatization based on family size. Effects are stronger among men and more religious individuals. In these groups, they persist even 7-8 months after the intervention. They are not driven by priming effects and point to a specific role of natives' norms. Indeed, a second treatment providing information about the fertility of migrants from the same group has no statistically significant impact on intentions and weaker effects on the ideal number of children. Our findings are consistent with a framework in which migrants are misinformed about natives' social norms and update their beliefs when exposed to new information.
Keywords:	Forced migration; Fertility preferences; Social norms; Information frictions; Cultural integration; Belief updating; Randomized controlled trial.
JEL:	J31 F22 D83 C93
Date:	2026–07
URL:	https://d.repec.org/n?u=RePEc:crm:wpaper:26205

By:	Gorodnichenko, Yuriy; Yin, Xiao
Abstract:	We combine a customized survey and randomized controlled trial (RCT) to study the effect of higher-order beliefs on U.S. retail investorsâ€™ portfolio allocations. We find that investorsâ€™ higher-order beliefs about stock market returns are correlated with but distinct from their first-order beliefs. Furthermore, the differences between the two vary systematically according to investor characteristics. We use information treatments in the RCT to create exogenous differential variations in first- and higher-order beliefs. We find that an exogenous increase in first-order beliefs increases the portfolio share allocated to the stock market (risky assets), while an exogenous increase in higher-order beliefs reduces it.
Keywords:	Expectations; Surveys; Higher-order beliefs; Asset pricing
JEL:	G11 G12 G51 D84 C83
Date:	2024–07
URL:	https://d.repec.org/n?u=RePEc:cpr:ceprdp:19205

By:	Justin E. Holz; Ricardo Perez-Truglia; Andrew Simon; Alejandro Zentner
Abstract:	Navigating the tax system often requires information and judgment that taxpayers may lack. Although taxpayers sometimes rely on human experts, AI could provide comparable assistance at substantially lower cost. Yet there is no causal evidence on how taxpayers use AI tax agents or how these tools affect their behavior. We fill this gap with a field experiment in the context of property tax appeals. We recruited a sample of 645 households in Dallas County, Texas. All households received access to a website providing personalized information, filing instructions, and supporting evidence for an appeal. Half were randomly assigned to a version of the website that also included an AI chatbot capable of answering questions and providing guidance tailored to their circumstances. Chatbot take-up was high: 78% initiated a conversation. Access to the chatbot increased the probability of filing an appeal without a human agent by 9.1 pp, from 41.4% to 50.5%. Evidence from clickstream data and conversation transcripts suggests that the chatbot helped households exercise judgment—that is, evaluate and act on the available information. The increase in appeal filing was smaller among less-advantaged households, providing suggestive evidence that the chatbot widened existing disparities. Our results demonstrate the potential of AI tax assistance while highlighting that who benefits depends not only on access, but also on how taxpayers use the technology.
JEL:	C93 H71
Date:	2026–08
URL:	https://d.repec.org/n?u=RePEc:nbr:nberwo:35632

By:	Brañas, Pablo; Cabrales, Antonio; Espinosa, María Paz; García-Muñoz, Teresa
Abstract:	In this paper we explore how individual social preferences correlate with political support for redistribution. We ran an incentivized experiment with a large representative sample of the Spanish population. Our participants took six decisions that elicited their social preferences. Their choices could result in a different total surplus and different distributions of the surplus between the subject and an anonymous counterpart. In our sample, social preferences are unrelated to political support for distributive policies. The main correlates for support of redistribution are the beliefs concerning the importance of effort versus luck for success (fairness), the trust in government institutions (effectiveness) and the perceived importance of the poverty problem (need).
JEL:	C93 D31 D72
Date:	2024–07
URL:	https://d.repec.org/n?u=RePEc:cpr:ceprdp:19274

By:	Philip Oreopoulos; Nina Low
Abstract:	Generative AI has been promoted as the technology that could transform education by providing every student a personal tutor. We provide some of the first large-scale experimental evidence, from a two-year cluster randomized trial in 18 Tennessee middle schools in which randomly assigned students used Khan Academy with its AI tutor, Khanmigo, configured to coach rather than give answers, during existing daily remedial mathematics sessions. Assignment raises math achievement by 1.3 national percentile ranks per term, or about 0.06 to 0.08 standard deviations over a school year; the implied effect of a full year of active participation reaches 0.14 standard deviations. These gains resemble those from Khan Academy practice without AI assistance. One explanation is that students used the tutor infrequently and, when they did, rarely engaged it in substantive mathematical dialogue: 96 percent of students tried Khanmigo at least once, but the median student messaged it on only a third of the days they practiced, and in only 17 percent of the exercise sessions in which they made a mistake. Messages that students did send were mostly bare answers or clicks on suggested prompts. The binding constraint appears to be engagement: realizing the promise of AI tutoring will require getting students to use it, not just giving them access.
JEL:	I21 I24 J24 O33
Date:	2026–08
URL:	https://d.repec.org/n?u=RePEc:nbr:nberwo:35620

By:	Timothy Sudijono; Edgar Dobriban; Eric Tchetgen Tchetgen
Abstract:	We study minimax-optimal designs and estimators for estimating the sample average treatment effect in finite population randomized experiments, where both design and estimator are unrestricted. For binary potential outcomes, we show this minimax risk is equivalent to the minimax risk $\rho_n^$ of an estimation problem with $2$ unknown parameters. We leverage this reduction to establish a second-order risk expansion $\rho_n^ = n^{-1} - Cn^{-4/3} + o_n(n^{-4/3})$ for an explicit constant $C$ related to the Airy function. The minimax risk is attained by Bernoulli randomization with a nonlinear shrinkage estimator. Our results show that standard procedures such as complete randomization with difference in means are only minimax optimal up to first order in $n.$ We derive further results on admissibility of these procedures and discuss the practical implications of our results.
Date:	2026–08
URL:	https://d.repec.org/n?u=RePEc:arx:papers:2608.13822

By:	Guiso, Luigi; Jappelli, Tullio
Abstract:	We implement a survey experiment to study whether awareness of the consequences of hydrogeological risk affects peopleâ€™s willingness to fight it. To do so, we leverage a representative panel of 5, 000 Italian individuals interviewed at quarterly frequency, starting in October 2023. We elicit survey participantsâ€™ willingness to contribute to a public fund to finance investment to secure areas exposed to hydrogeological risk under different information treatments. We find that disclosing information about the consequences of hydrogeological risk causes individuals to increase both support for public funding and individual willingness to pay for the policy. Compared to the control group, individuals exposed to the treatment were 9 percentage points more likely to contribute to the fund and more willing to contribute an additional â‚¬29. Applying the information treatment to the whole working age population could raise as much as â‚¬0.26 billion per year. We provide evidence that individual willingness to pay depends on individual knowledge that the success of the policy depends critically on the willingness to pay of other citizens.
JEL:	H31 H2 H23
Date:	2024–07
URL:	https://d.repec.org/n?u=RePEc:cpr:ceprdp:19280

By:	Tyack, Nicholas; Arouna, Aminou; Dembele, Urbain; Goeschl, Timo
Abstract:	We study how confidence bias affects investment in learning via experimentation, a mechanism critical for technology adoption under uncertainty. We hypothesize that bias direction and strength predict how willingness to experiment diverges from unbiased agents. We measure revealed and stated demand for experimenting with drought-resistant crop varieties of 1, 957 farmers in West Africa, a climate change hotspot. Consistent with our hypothesis, confidence bias strongly predicts willingness to experiment. The effect, however, is driven exclusively by underconfident agents, among whom females are overrepresented. In deteriorating environments, this behavioral friction undercuts effective technology diffusion and risks trapping individuals in maladapted production environments.
Keywords:	Research Methods/Statistical Methods
Date:	2026
URL:	https://d.repec.org/n?u=RePEc:ags:aaea26:404524

By:	Albarracín, Dolores; Hyde, Timothy (Department of Economics, Oberlin College)
Abstract:	Flood risk disclosure is expanding rapidly in real estate markets, but if disclosures move demand by commanding attention rather than by conveying information, theirwelfare effects are ambiguous. We examine this tension in a discrete choice experiment where 1, 498 respondents make hypothetical choices among real Gulf Coast properties under a control condition and four disclosure formats varying in informational content and presentation. Absent disclosure, insurance costs barely influence choices; by contrast, when respondents are shown precise flood insurance estimates, they are willing to pay $3.34 in annualized housing costs to avoid each annual premium dollar. Using a separate beliefs elicitation, we estimate each format’s technical information content and construct a Bayesian benchmark for how much choices should respond. Observed responses exceed this benchmark by a factor of four to six; informational content accounts for at most 30 percent of the response to any format.
Keywords:	Flood risk, Risk disclosure, Salience, Attention, Willingness to pay, Stated preferences
JEL:	D83 D91 Q54 R31 G22
Date:	2026–07–01
URL:	https://d.repec.org/n?u=RePEc:cxv:wpaper:2603

By:	Guy Aridor; Winston Chou; Nathan Kallus; Antoine Scheid; Allen Tren; Kevin Zielincki
Abstract:	We study an experiment with 8.5 million users on Netflix's recommender system to measure how improvements in recommendation technology affect the set of products that get consumed. Improvements increase total consumption and users' reliance on recommendations while diffusing recommendations and consumption away from the most popular titles (``superstars") toward a larger number of moderately popular titles (``middle-tail"), with minimal effects on the most niche titles (``long-tail"). Our results challenge the notion that recommender systems polarize consumption -- raising the consumption shares of the head and tail at the expense of the middle -- and suggest that the returns to investing in middle-tail products grow as algorithms improve and platforms scale.
Date:	2026–08
URL:	https://d.repec.org/n?u=RePEc:arx:papers:2608.21274

By:	Evgeny Kagan
Abstract:	Problem Definition: How a queue of jobs is arranged and presented to workers is an important design problem in service operations. This includes choosing the order in which jobs are performed, how much say workers have in setting that order, and how much queue and arrival information workers receive. Methodology/Results: To better understand how queue design (job ordering, autonomy, visibility) affects worker performance (speed, quality), we run a series of pre-registered online experiments. We use a new, real-effort task in which workers fulfill order-picking jobs of varying complexity that arrive dynamically over time. Our results are as follows: (1) When workers choose their own picking order, we reproduce the field finding that Easy First (EF) ordering is associated with worse performance than First-in-first-out (FIFO), and show that this is mainly due to worker self-selection rather than due to the ordering itself; (2) Exogenously imposed EF ordering improves work quality (picking accuracy) relative to both FIFO and discretionary ordering; (3) Imposing an ordering may reduce speed for the most capable workers; (4) Seeing a new job arrival leads to a short-term productivity burst; however, removing job arrival and queue information altogether does not affect performance in the long term. Managerial implications: Our results provide guidance on which queue design works best for a given performance goal (speed or quality) and worker ability level. We also identify personality measures that can help managers screen for error-prone workers.
Date:	2026–07
URL:	https://d.repec.org/n?u=RePEc:arx:papers:2607.28820

By:	Vigild, Daniel Juhász (ROCKWOOL Foundation); Bjerre-Nielsen, Andreas (University of Copenhagen); Mortensen, Laust Hvas; Sekara, Vedran
Abstract:	Artificial intelligence is increasingly being used in government, changing the ways in which citizens relate to public institutions. One important aspect, the impact of AI on trust in public institutions, has been the subject of substantial efforts that document an "AI penalty" on trust in public institutions when they use AI. However, little is known about why or how such a penalty arises. In two preregistered survey experiments (N=6 084) we measure the effect of AI on trust across four real-life use cases of governmental AI. First, we find large differences across the four domains—disclosing AI usage in social services and health care settings decreases trust, while trust is unaffected for policing and unemployment use cases. Second, we experimentally test a specific mechanism (algorithmic nimbyism) that considers personal exposure to the decision-making outcomes as a driver of the negative impact of AI on trust. Despite not finding any causal evidence for the notion of algorithmic nimbyism, we used rich administrative data on survey participants’ medical history and labour force participation to explore heterogeneous treatment effects that revealed descriptive and suggestive evidence of algorithmic nimbyism. Future work could consider further exploring the combination of experimental and observational data to empirically examine explanations of the impact AI has on relations between citizens and public institutions.
Date:	2026–08–08
URL:	https://d.repec.org/n?u=RePEc:osf:socarx:xrejw_v1

By:	Andrés Barrios Fernández; Josefina Eluchans Errázuriz; Fernanda Ramírez-Espinoza
Abstract:	Large socioeconomic gaps in higher-education access and completion persist even among similarly talented students with access to generous funding, pointing to an important role for information frictions and unequal support. We evaluate whether a brief, scalable program combining a customized information package with four one-hour group counseling sessions delivered during the school day by external counselors can improve postsecondary trajectories when students face complex financial-aid and higher-education decisions. In a randomized controlled trial with 26, 853 students in 229 Chilean high schools, we assign schools to three groups: control, information only, and information plus group counseling. Linking administrative records and survey data, we find that information alone improves some dimensions of knowledge but does not significantly change funding applications or enrolment. By contrast, information plus counseling increases applications, enrolment, persistence, and short-degree completion. The counseling arm also generates larger gains in financial-aid knowledge and shifts students toward more reliable information sources, consistent with uncertainty reduction and procedural guidance as key mechanisms. Treatment effects are largest for students with less family support and weaker initial informational endowments, suggesting that counseling partly substitutes for missing support at home. The results indicate that in high-stakes, high-complexity settings, policy-relevant barriers extend beyond lack of information to difficulties in processing and acting on information. Structured counseling interventions of this kind can mitigate counselor scarcity and improve higher-education trajectories
Keywords:	school counselling, school counselling, higher education access, financial aid, access barriers
Date:	2026–08–12
URL:	https://d.repec.org/n?u=RePEc:cep:cepdps:dp2205

By:	Gary Charness (University of California); Francesco Feri (Royal Holloway University of London); Matthew O. Jackson (Stanford University & Santa Fe Institute); Miguel A. Meléndez-Jiménez (Universidad de Málaga); Matthias Sutter (Max Planck Institute for Behavioral Economics, University of Cologne & University of Innsbruck)
Abstract:	We provide a first causal analysis of the behavioral consequences of the friendship paradox—the fact that people’s friends in a network have more connections than average. We find that people’s behavior is biased by their network position: they do not best respond to what they should infer the average behavior of the population to be, but instead simply to the average behavior of their friends. Moreover, we find that they fail to learn to overcome such a bias when relocated within the network, varying their observational environment. In these games of complements, the friendship paradox generates a systematic upward distortion in actions, increases behavioral dispersion, and persists despite learning opportunities.
Keywords:	Friendship paradox, networks, learning, experiment
JEL:	C91 D01 D85 D90
Date:	2026–08
URL:	https://d.repec.org/n?u=RePEc:ajk:ajkdps:426

By:	Nunnari, Salvatore; Proto, Eugenio; Rustichini, Aldo
Abstract:	Rational choice theories assume that voters accurately assess the outcomes of policies. However, many important policiesâ€”such as regulating prices and introducing Pigouvian taxationâ€”yield outcomes through indirect or equilibrium effects that may differ from their direct effects. Citizens may underestimate these effects, leading to a demand for bad policy, that is, opposition to reforms that would increase welfare or support for reforms that would decrease it. This appreciation might be linked to cognitive functions, raising important research questions: Do cognitive abilities influence how individuals form preferences regarding policies, especially untried reforms? If so, what is the underlying mechanism? We use a simple theoretical framework and an experiment to show that enhanced cognitive abilities may lead to better policy choices. Moreover, we emphasize the crucial role of beliefs about other citizens' cognitive abilities. These findings have important policy implications as they suggest that educational programs developing cognitive skills or interventions increasing trust in others' understanding could improve the quality of democratic decision-making in our societies.
Keywords:	Voting; Cognition; Experiment
JEL:	C90 D72 D91
Date:	2024–07
URL:	https://d.repec.org/n?u=RePEc:cpr:ceprdp:19217

By:	Ba, Bocar A.; Ndiaye, Abdoulaye; Rivera, Roman G.; Whitefield, Alexander
Abstract:	We study how negative sentiment around an industry impacts beliefs and behaviors, focusing on demands for racial justice after the murder of George Floyd and the salience of the ``defund the police" movement. We assess stakeholder beliefs on the impact of protests on the stock prices of police-affiliated firms. In our survey experiment, laypeople and finance professionals predicted more negative stock price outcomes when they lacked details on the products supplied by such firms. Exposure to narratives about the context of the protests further reduced the prediction accuracy of these groups. In contrast, product information improved the prediction accuracy of respondents. Turning to real-life behavior, we find that mutual funds exposed to protests were 20% less likely to hold police stocks, after the protests, than funds in areas without protests. Political support for maintaining police funding, though in the majority, declined by 4.3 percentage points in protest areas. The salience of the ``defund the police" narrative led to significant overreactions in both financial predictions and real-life behavior.
Keywords:	Narratives; Reasoning; Surveys
JEL:	D72 D74 D83 G41
Date:	2024–07
URL:	https://d.repec.org/n?u=RePEc:cpr:ceprdp:19308

By:	Peter Hull
Abstract:	I study the optimal design and analysis of randomized experiments for estimating finite-population average treatment effects when potential outcomes are known to be bounded, as with binary outcomes. Among all assignment mechanisms and a broad class of affine estimators, worst-case mean-squared error (MSE) is minimized by independent random assignment and an unconventional regression of the support-midpoint-centered outcome on the recentered treatment, with no intercept. This contrasts with the usual prescription of balanced complete randomization and difference-in-means estimation: when outcomes are bounded, randomness in the realized treatment share is informative. The worst-case gain over full-sample complete randomization is asymptotically small, but gains can be first-order relative to other designs: complete within-pair randomization and pair-fixed-effect regression have twice the worst-case MSE. I extend the result to allow for arbitrary estimators. Independent random assignment remains optimal, and the generally-nonlinear optimal estimator can meaningfully reduce worst-case MSE.
Date:	2026–08
URL:	https://d.repec.org/n?u=RePEc:arx:papers:2608.09812

By:	Albers, Thilo N.H.; Bischof, Daniel (Aarhus University); Kersting, Felix; Kosse, Fabian (University of Würzburg)
Abstract:	Why do materially secure citizens embrace populist attitudes? Research on the populist right has focused primarily on economically vulnerable voters and left-behind places, leaving unclear why support extends to citizens who are objectively well off. We argue that a belief-based source of grievance helps explain this puzzle: citizens may underestimate their relative economic standing and feel deprived despite material security. Using a survey module embedded in the Innovation Sample of the German Socio-Economic Panel, we show that pessimistic income-rank misperceptions predict stronger populist attitudes conditional on actual rank. Instrumental-variable and within-household analyses strengthen a causal interpretation. A preregistered survey experiment randomly provides personalized information about respondents' actual income percentile. The control arm replicates the observational pattern; the treatment corrects beliefs and shifts income satisfaction, but does not reduce populist attitudes. Economic grievance can therefore travel upward through perceived status, while becoming politically resistant once organized by populist narratives.
Date:	2026–08–03
URL:	https://d.repec.org/n?u=RePEc:osf:socarx:2s7mc_v1

By:	Jean-Pierre H. Dubé; Ilya Morozov; Franklin She; Anna Tuchman
Abstract:	We study how in-store display ads communicate information about a product’s price and quality, both directly through ad content and indirectly through inferences consumers draw from the presence and content of the ad. We run an incentivized experiment in a simulated online store, randomizing ad exposure, ad content, and the advertised product’s price, while also eliciting participants’ beliefs and choices. Display ads shift beliefs about price and quality, even when the ad copy does not directly reveal either attribute. We provide evidence that these belief shifts have a causal effect on searches and purchases of the advertised product. Using a stylized search model, we show that while visibility explains most of the ad effects, belief updating explains a statistically and economically significant share. Casting ad content choice as an information design problem, we show that advertisers may find it optimal to conceal product information even when it exceeds consumers’ expectations.
JEL:	M3 M30 M31 M37
Date:	2026–08
URL:	https://d.repec.org/n?u=RePEc:nbr:nberwo:35596

By:	Bau, Natalie; Henning, David J.; Low, Corinne; Millett Steinberg, Bryce
Abstract:	Early fertility is thought to be one of the key barriers to female human capital attainment in sub-Saharan Africa, yet contraceptive take-up remains puzzlingly low, even among highly-educated populations with healthcare access. We study a barrier to hormonal contraceptive uptake that, while recognized in the qualitative literature, has not been causally tested: the persistent (incorrect) belief that these contraceptives cause later infertility. This belief creates a perceived tradeoff between current and future reproductive control. We use a randomized controlled trial with female undergraduates at the flagship university in Zambia to test two interventions to increase contraceptive use. Despite high rates of sexual activity, low rates of condom-use, and near zero desire for current pregnancy, only 5% of this population uses hormonal contraceptives at baseline. Providing a non-coercive conditional cash transfer to visit a local clinic â€“ greatly reducing access costs â€“ only temporarily increases contraceptive use. However, pairing this transfer with information addressing fears that contraceptives cause infertility has a larger initial effect and persistently increases contraceptive take-up over 6 months. This treatment reduces beliefs that contraceptives cause infertility and leads to the take-up of longer-lasting contraceptives like injections. Compliers are more likely to cite fear of infertility as the reason for not using contraceptives at baseline. IV estimates indicate that eliminating the belief that contraceptives cause infertility would more than triple contraceptive use.
Date:	2024–07
URL:	https://d.repec.org/n?u=RePEc:cpr:ceprdp:19264