|
on Experimental Economics |
| By: | Hans-Theo Normann; Nina Ruli\'e; Olaf Stypa; Tobias Werner |
| Abstract: | We analyze the delegation of pricing by participants, representing firms, to a collusive, self-learning algorithm in a repeated Bertrand experiment. In the baseline treatment, participants set prices themselves. In the other treatments, participants can either delegate pricing to the algorithm at the beginning of each supergame or receive algorithmic recommendations that they can override. Participants delegate more when they can override the algorithm's decisions. In both algorithmic treatments, prices are lower than in the baseline. Our results indicate that while self-learning pricing algorithms can be collusive, they can foster competition rather than collusion with humans-in-the-loop. |
| Date: | 2025–10 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2510.27636 |
| By: | Lee, I.; Lléo-Bono, A.; Rauh, C.; Tipoe, E. |
| Abstract: | This paper examines whether increasing individuals' awareness of their own confidence can influence financial behaviour. In a pre-registered online experiment with nearly 3, 000 U.S. adults, we test the effects of a novel metacognitive intervention: personalised feedback on implicit confidence about one's financial abilities, as measured by a custom Implicit Association Test (IAT), paired with an explanation of the importance of self-confidence in financial abilities. Treated participants show a significant reduction in "don't know" responses on financial literacy tests and their performance in an incentivised investment task significantly improves: treated participants are less likely to make clearly dominated choices, more likely to select efficient allocations, and choose portfolios closer to the efficient frontier. These effects persist two weeks later in a follow-up survey with obfuscated framing. Heterogeneity analyses show stronger effects for females and for participants who understate their confidence (i.e. whose reported confidence is lower than what their IAT suggests). |
| Keywords: | Confidence, Confidence Awareness, Personal Finance, Financial Literacy, Survey Experiment |
| JEL: | D14 D83 D91 G11 G53 |
| Date: | 2025–10–14 |
| URL: | https://d.repec.org/n?u=RePEc:cam:camjip:2527 |
| By: | Aaron S. Berman (California Institute of Technology); Saika Cer Askin (Chapman University, Economic Science Institute); Shapeng Jiang (Wuhan University); David Porter (Chapman University, Economic Science Institute); Jason Shachat (Chapman University, Economic Science Institute) |
| Abstract: | This study examines preference-based behavioral biases in social interactions between two distinct communities: students from Chapman University in the United States and Wuhan University in China. Using controlled experiments, participants interacted within or across communities in Dictator games. Two versions of the Dictator game were used: one where decisions were observable by both the experimenter and the recipient, and another where allocators could misreport outcomes with plausible deniability. Results revealed unexpected patterns, including similar allocation distributions across communities in the transparent task, and differing behaviors in the misreporting task, with Chapman allocators being more generous to out-group members and Wuhan allocators choosing more selfishly. The study challenges traditional theories of in-group favoritism and highlights the role of cultural differences and image concerns in decision-making. Findings contribute to understanding cross-cultural interactions, particularly in the context of increasing global connectivity. |
| Keywords: | In-group bias, Dictator game, Lying, Social image |
| JEL: | C92 D63 D91 |
| Date: | 2025 |
| URL: | https://d.repec.org/n?u=RePEc:chu:wpaper:25-12 |
| By: | Mo, Zhexun; Kaeppel, Katharina; Schröder, Carsten; Yang, Li |
| Abstract: | This paper investigates how the German public perceives Chinese Foreign Direct Investment (FDI) amid escalating geopolitical tensions and deepening China–EU economic interdependence, complemented by evidence on Chinese perceptions of German FDI. We combine data from a large-scale survey experiment embedded in the 2023 German Socio-Economic Panel Innovation Sample (N = 2, 365) with a descriptive survey in China (N = 2, 000). German respondents substantially overestimate the scale of Chinese investment—believing it accounts for about 30% of total inward FDI, compared to an actual share of roughly 1%—and evaluate it significantly less favorably than investment from other EU countries or the United States. In contrast, Chinese respondents express consistently positive views of German FDI, revealing a pronounced asymmetry in mutual perceptions. To probe the origins and malleability of these views, we implement three randomized interventions in Germany: a factual correction of actual FDI shares and two narrative framings emphasizing either positive or negative aspects of Chinese investment. While factual information modestly improves perceptions of FDI’s economic benefits, none of the treatments meaningfully shift deeper, ideologically anchored attitudes toward Chinese investment. Quantile treatment effect analyses indicate, however, that these interventions reduce anti-China biases among respondents who are initially more receptive to Chinese FDI. Taken together, the results—consistent across direct evaluations, conjoint choice experiments, and Willingness-to-Accept (WTA) measures—underscore the limits of informational interventions in reshaping entrenched geopolitical or identity-based biases toward foreign investment. (Stone Center on Socio-Economic Inequality Working Paper) |
| Date: | 2025–10–29 |
| URL: | https://d.repec.org/n?u=RePEc:osf:socarx:74k3v_v1 |
| By: | Moustapha Sarr; Noémi Berlin; Tarek Jaber-Lopez |
| Abstract: | In a lab-in-the-field experiment, we investigate the influence of social norms on 300 parents’ beliefs regarding the nutritional quality of food items and their subsequent food choices. We use a 3 × 2 between-subject experimental design where we vary two factors: 1-the social norm provided to parents: a descriptive norm (what other parents choose) vs. an injunctive norm (what other parents approve of), and 2-the recipient of the food decisions made by parents: their own child vs. an unknown child. Parents participate in a two-stage process. In the first stage, we elicit their beliefs regarding the nutritional quality of various food items and ask them to make a food basket without specific information. In the second stage, based on their assigned treatment, they receive specific information and repeat the belief elicitation and the food basket selection tasks. We find that only the descriptive norm significantly reduces parents’ overestimation rate of items’ nutritional quality. Injunctive norm significantly improves the nutritional quality of both, the parent’s and child’s baskets. Descriptive norm significantly improves the nutritional quality of child’s baskets only when parents are choosing for unknown child. |
| Keywords: | social norms, information provision, food choices, food beliefs, parent, child |
| JEL: | C93 D12 D91 |
| Date: | 2025 |
| URL: | https://d.repec.org/n?u=RePEc:drm:wpaper:2025-42 |
| By: | Julia Baarck; Mathias Dolls; Lisa Windsteiger |
| Abstract: | This study investigates how information about intergenerational and intragenerational inequality shapes fairness views and policy preferences. Using a large-scale randomized survey experiment with 4, 900 respondents in Germany, we test how exposure to information on wealth and age disparities affects (i) perceptions of distributive and intergenerational fairness, and (ii) support for redistributive and future-oriented policies. We find that respondents generally underestimate existing inequalities. Moreover, providing accurate information about the extent of age and wealth disparities has little impact on left-leaning and centrist individuals but elicits a backlash among right-leaning respondents: the information increases their perceived fairness of the status quo and lowers their support for redistributive and future-oriented measures. We attribute these counterintuitive responses to skepticism about the credibility and neutrality of the provided information. Overall, the findings highlight the limits of informational interventions and the potential for factual communication to backfire in politically polarized contexts. |
| Keywords: | fairness views, inequality, policy preferences, survey experiment |
| JEL: | D63 D72 J48 D83 D91 H23 |
| Date: | 2025 |
| URL: | https://d.repec.org/n?u=RePEc:ces:ceswps:_12234 |
| By: | Leyens, Alexa Feisthauer; Philipp Börner, Jan; Hartmann, Monika; Storm, Hugo |
| Abstract: | Round effects can arise in any kind of economic experiment, where participants have to make decisions over a course of various rounds. They are associated with changes of preferences and patterns of response variance, they may counteract or distort experimental treatment effects and ultimately result in the biased estimation of the “true” underlying preferences. To investigate how exactly round effects occur and can be captured statistically, we have designed and conducted a multi-round study with German agricultural students. We develop a novel Bayesian Probabilistic Programming approach to assess to which extent preference learning, institutional learning and fatigue effects influence the response behaviour of survey participants. We find strong evidence for fatigue effects, since the response behaviour of our participants became increasingly variant over the course of the experiment. We could also falsify the assumptions of institutional and preference learning for our participant group. Our results emphasize the importance of modelling round effects in economic experiments. The results and the developed modelling framework will be of interest to both, experimental agricultural economists and policy developers, who interpret and apply findings from business simulation games and similar multi-period experimental studies. |
| Keywords: | Agricultural Finance, Research and Development/Tech Change/Emerging Technologies, Research Methods/Statistical Methods |
| URL: | https://d.repec.org/n?u=RePEc:ags:aes024:355341 |
| By: | Cortes, Kalena; Holzman, Brian; Gentry, Melissa; lambert, miranda |
| Abstract: | This study examines how digital incentives influence survey participation and engagement in a large randomized controlled trial of parents across seven Texas school districts. We test how incentive amount and information about vendor options affect response behavior and explore differences by language background. Incentivized parents were more likely to start and complete surveys and claim gift cards, though Spanish-speaking parents exhibited distinct patterns—greater completion rates but lower redemption rates, often selecting essential-goods vendors. Increasing incentive value and providing advance information both improved engagement. Findings inform the design of equitable, effective digital incentive strategies for diverse populations. |
| Date: | 2025–11–03 |
| URL: | https://d.repec.org/n?u=RePEc:osf:socarx:rtj4e_v1 |
| By: | Joan Costa-Font; Nicolo Gatti (Università Cattolica del Sacro Cuore; Dipartimento di Economia e Finanza, Università Cattolica del Sacro Cuore); Gilberto Turati (Università Cattolica del Sacro Cuore; Dipartimento di Economia e Finanza, Università Cattolica del Sacro Cuore); Daniel Wiesen |
| Abstract: | We study the extent to which exposure to COVID-19 has affected the prosociality of healthcare professionals. Drawing on evidence from an incentivized experiment and a companion survey of healthcare professionals (N = 194) at a large Italian hospital, we find that exposure to COVID-19 predicts their altruistic motivation. Healthcare professionals who either worked in COVID-19 wards or had a close relative or friend severely affected by the virus exhibited a higher propensity to prioritize patient welfare over personal gain. Controlling for demographic characteristics and personality traits, our estimates indicate a 5 percentage point increase in prosociality among exposed individuals. Conversely, personally contracting COVID-19 is related to a 6 percentage point decline in prosociality. Our results highlight that experiences of need shape prosocial preferences in hospital settings. |
| Keywords: | COVID-19, provider altruism, healthcare professionals, medical decision-making, prosocial behavior, experiment. |
| JEL: | C91 D64 D81 I12 |
| Date: | 2025–10 |
| URL: | https://d.repec.org/n?u=RePEc:ctc:serie1:def145 |
| By: | Garlick, Robert; Field, Erica; Vyborny, Kate |
| Abstract: | This paper studies whether commuting barriers constrain women’s labor supply in urban Pakistan. This study randomized offers of gender-segregated or mixed-gender commuting services at varying prices. Women-only transport more than doubled job application rates, while mixed-gender transport had minimal effects on men’s and women’s application rates. Women valued the women-only service more than large price discounts for the mixed-gender service. The results are similar for baseline labor force participants and non-participants, suggesting there are many “latent jobseekers’’ close to the margin of participation. These findings highlight the importance of safety and propriety concerns in women’s labor decisions. |
| Date: | 2025–11–10 |
| URL: | https://d.repec.org/n?u=RePEc:wbk:wbrwps:11256 |
| By: | Aguiar, Felipe; Lapple, Doris; Buckley, Cathal |
| Abstract: | A key measure to reduce chemical fertilizer application, and thereby mitigate greenhouse gas (GHG) emissions from agriculture, is the introduction of clover into grazing grass. However, adoption of this measure remains low. In this context, information and expectations can influence adoption decisions around clover. So far, there is little evidence of how farmers update their expectations in response to information. In this study, we used an information experiment to assess how providing information affects farmers’ expectations about clover, and how this in turn influences subsequent attitudes. We collected data from over 300 dairy farmers in Ireland, which were randomly assigned into two information treatment groups and one active control group. While both treatments provided information about the reduction in chemical fertilizer associated with adoption, each treatment framed the information differently. To elicit farmers’ expectations, we combined qualitative open-ended questions and quantitative point estimates. As for the subsequent attitudes, we elicited farmers’ intentions and willingness to accept (WTA) clover adoption. We estimated treatment effects by employing a two-stage least squares regression. To examine responses from the open-ended questions, we used three text analysis methods: wordclouds, keyness, and topic analyses. We document that farmers have biased expectations about clover adoption. They underestimate the reduction of chemical fertilizer that is possible with adoption, and we provide causal evidence that information reduces misperceptions by up to 19%. Yet, through the text analysis, we discover that information increases the likelihood of having not only a positive change, but also a negative change in opinions around clover adoption. Lastly, there was no meaningful impact of the updated expectations on intentions and WTA, which underlines the complexity of adoption decisions. Nonetheless, our findings are relevant to help construct accurate expectations that can facilitate more widespread adoption of clover. |
| Keywords: | Environmental Economics and Policy |
| URL: | https://d.repec.org/n?u=RePEc:ags:aes024:355328 |
| By: | Li, Xiaofei |
| Abstract: | While on-farm experiments offer invaluable insights for precision management decisions, their scope is usually confined to the specific conditions of individual farms and years, which limits the derivation of more broad and reliable decisions. To address this limitation, aggregating data from multiple farms into a comprehensive dataset appears promising. However, the quantifiable value of this experiment network remains elusive, despite the common agreement of the existence of this value. This study conducted a simulation-based assessment of the economic value of large-scale on-farm experiments, using crop variety selection as a case study. A hypothetical region was simulated comprising one thousand corn production fields of diverse soil types and weather conditions. Each field was implemented with an on-farm variety trial. Yields for each variety were simulated based on presumed true yield responses to soil types and weather conditions that are derived from historical Mississippi variety trial data. By constructing aggregated on-farm experiment data set of farms, the individualized optimal variety for each field was recommended, and the associated yields were predicted. The production profitability for all fields was calculated based on current market prices. Results revealed a substantial improvement in farming profitability when employing the individualized optimal variety selection derived from the large-scale experiment network, compared to the scenario of only using farm’s own data. Furthermore, the simulation study also reveal that the profitability improvement diminishes when the scale of the experiment network decreases or when the number of trial varieties per field reduces. The simulation results underscored the economic benefits of broader farmer participation in on-farm experiment network and more intensive trials by each participant. |
| Keywords: | Crop Production/Industries |
| Date: | 2024–09–29 |
| URL: | https://d.repec.org/n?u=RePEc:ags:haaewp:373440 |
| By: | Assenza, Tiziana; Huber, Stefanie; Mogilevskaja, Anna; Schmidt, Tobias |
| Abstract: | We use a randomized experiment in the Bundesbank Online Panel-Households (n ≈ 3, 900) to show that the estimated link between inflation expectations and household consumption flips sign depending on survey wording. This finding reconciles prior contradictory results and has direct implications for central bank survey design. Our experiment systematically varies elicitation framing of consumption question along three dimensions: the reference unit (individual vs. household), the time horizon (past one, 3, or 12 months), and the question type (attitudinal, planned, qualitative and quantitative recall-based). We find that the time horizon and question type significantly influence the estimated relationship between inflation expectations and durable consumption. While the average effect is weak, its sign and magnitude vary strongly with question design. Planned spending and attitudinal questions, such as whether it is a good time to buy, produce very similar negative associations, suggesting that respondents interpret the former as a proxy for future intentions. In contrast, quantitative recall-based questions on past spending yield a modestly positive link, especially for shorter horizons. These results highlight the critical role of survey design in shaping behavioral measurements, offering a novel explanation for mixed findings in the literature and guidance for both research and policy. |
| Keywords: | Expectations; household decision making; survey methodology; framing effects; measurement; inflation (economic) |
| JEL: | C83 D12 D84 E31 |
| Date: | 2025–11–04 |
| URL: | https://d.repec.org/n?u=RePEc:tse:wpaper:131083 |
| By: | Andrew Leal |
| Abstract: | Employers are concerned not only with a prospective worker's ability, but also their propensity to avoid shirking. This paper proposes a new experimental framework to study how Principals trade-off measures of ability and prosocial behavior when ranking Agents for independent jobs. Subjects participate in a simulated, incentivized job market. In an initial session, subjects are Workers and generate a database of signals and job results. Managers in subsequent sessions observe the signals of Worker behavior and ability and job details before a rank-and-value task, ranking and reporting a value for each Worker for two distinct jobs. Results highlight Managers' preference for ability over prosocial behavior on average, especially for Managers in STEM fields. There is evidence of homophily: the relative value of prosocial behavior is higher for highly prosocial Managers, compensating for ability or even surpassing it in value. |
| Date: | 2025–10 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2510.27625 |
| By: | Yue Fang; Geert Ridder |
| Abstract: | In cluster randomized controlled trials (CRCT) with a finite populations, the exact design-based variance of the Horvitz-Thompson (HT) estimator for the average treatment effect (ATE) depends on the joint distribution of unobserved cluster-aggregated potential outcomes and is therefore not point-identifiable. We study a common two-stage sampling design-random sampling of clusters followed by sampling units within sampled clusters-with treatment assigned at the cluster level. First, we derive the exact (infeasible) design-based variance of the HT ATE estimator that accounts jointly for cluster- and unit-level sampling as well as random assignment. Second, extending Aronow et al (2014), we provide a sharp, attanable upper bound on that variance and propose a consistent estimator of the bound using only observed outcomes and known sampling/assignment probabilities. In simulations and an empirical application, confidence intervals based on our bound are valid and typically narrower than those based on cluster standard errors. |
| Date: | 2025–11 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2511.05801 |
| By: | Albrecht, Sabina; , Danyelle; Dolnicar, Sara (The University of Queensland) |
| Abstract: | The warm glow of charitable and environmentally sustainable actions has received much attention in the academic literature. Some studies suggest the power of the feel-good effect of doing good is severely underrated as a driver of human behaviour. This study curbs the enthusiasm by evaluating whether different forms of warm glow change behaviour across online and field experimental studies. Second-hand clothes shopping is the setting of our investigation because multiple forms of warm glow apply. Behaviour change toward reuse is highly desirable to lower the negative environmental impact of the fashion industry. In line with prior research, we find promising evidence in survey studies that priming the warm glow effect of second-hand shopping could increase second-hand shopping by up to 16%. However, when tested in a field study involving second-hand charity shops, messages communicating the warm glows of second-hand shopping performed no better than a general attention-generating message. Our study shows that charitable and environmentally sustainable behaviour does generate warm glow, but that the perception of warm glow might not be enough to motivate purchasing behaviours. |
| Date: | 2025–11–05 |
| URL: | https://d.repec.org/n?u=RePEc:osf:socarx:jgec8_v1 |
| By: | Agnieszka Zalejska-Jonsson; Sara Wilkinson |
| Abstract: | Urbanization has significantly altered land use, reducing natural habitats and exacerbating environmental issues such as increased temperatures, altered water cycles, and poor air quality. The study investigates whether and how buyers perceive greenery in housing developments, the impact of green infrastructure on consumer interest, and the potential financial benefits of investing in sustainable residential designs. The study builds upon research in consumer behavior, urban planning, and real estate economics to explore property buyers’ preferences. The study employed an experimental design embedded within a survey to test the effect of green infrastructure on buyers’ interest. Five courtyard designs with varying levels of greenery were developed in collaboration with landscape architects. These were visualized using 3D modeling techniques and virtual reality tools, allowing participants to explore different courtyard layouts. Respondents were randomly assigned one of the five courtyard scenarios and asked to evaluate its attractiveness and their interest in purchasing an apartment in the development. Data was collected from 922 participants in Sweden and 614 in Australia using structured online surveys. The analysis utilized structural equation modeling (SEM) to assess relationships between courtyard greenery, attractiveness, and purchasing interest. The study concludes that green infrastructure plays a significant role in shaping buyers' perceptions of residential developments. Findings indicate that greener courtyards enhance property attractiveness, social engagement, and environmental benefits, leading to increased willingness to pay. Additionally, study reflects how cultural and environmental contexts shape buyers’ perceptions of green spaces, compering results obtained in Sweden and Australia. The study’s experimental approach provides a novel methodology for evaluating the economic and psychological effects of greenery in residential developments |
| Keywords: | buyers perception; Green Infrastructure; housing; sustainable cities |
| JEL: | R3 |
| Date: | 2025–01–01 |
| URL: | https://d.repec.org/n?u=RePEc:arz:wpaper:eres2025_230 |
| By: | Manuel C. Kathan (University of Augsburg); Raphaela Roeder (University of Augsburg); Sebastian Utz (University of Augsburg); Martin Nerlinger (University of St. Gallen - School of Finance; Swiss Finance Institute) |
| Abstract: | We examine how changes in competition affect firms’ carbon performance. Exploiting reductions in import tariffs as a quasi-natural experiment that increases competitive pressure, we find that stronger competition improves firms’ carbon efficiency through lower Scope 1 and 2 emission intensities. These results remain robust to alternative specifications, heterogeneous treatment effects, and placebo tests. Mechanism analyses indicate systematic differences in firms’ strategic responses. High-emission firms tend to adopt visible environmental actions and reallocate resources toward intangible assets, whereas low-emission firms increase investment and financing activities. Overall, our results highlight competition as a determinant of corporate decarbonization, suggesting that market forces can complement regulatory approaches to improving firms’ environmental performance. |
| Keywords: | Carbon Emissions, Environmental Performance, Sustainability, Competition, Import Tariffs, Quasi-Natural Experiment |
| Date: | 2025–10 |
| URL: | https://d.repec.org/n?u=RePEc:chf:rpseri:rp2588 |
| By: | Stefano Carattini; Fabian Dvorak; Ivana Logar; Begum Ozdemir-Oluk |
| Abstract: | Corporate social responsibility and the private provision of (global) public goods are of key interest to economists and policymakers. Over the last few years, many more private companies made their operations carbon neutral. It is an empirical question how consumers value carbon-neutral and low-carbon products, which we address as follows. First, we provide a meta-analysis of the literature. We analyze consumers’ demand for carbon-neutral and low-carbon products, based on an overall sample of 29, 666 participants. The focus is on average willingness to pay for carbon reductions as well as on the characteristics of the underlying literature, which is mainly based on stated preferences and controlled environments. Second, we leverage information on prices and product characteristics from one of the largest online marketplaces, Amazon’s. Using a hedonic approach, we infer from revealed preferences on consumers’ valuation of carbon- neutral products. The staggered process of carbon-neutral certification leads to a series of quasi-natural experiments, which we use for identification purposes. We find that the literature suggests a positive willingness to pay for carbon reductions that exceeds most estimates of the social cost of carbon. However, this finding is not supported by the hedonic analyses, where we do not find evidence that consumers value carbon neutrality. |
| Keywords: | corporate social responsibility, pro-social behavior, stated and revealed preferences, meta-analysis, hedonic analysis, carbon-neutral labels |
| JEL: | D12 D22 H41 Q51 Q54 |
| Date: | 2025 |
| URL: | https://d.repec.org/n?u=RePEc:ces:ceswps:_12232 |
| By: | Boyer, Christofer; Field, Erica; Lehrer, Rachel; Morrison, Andrew; Piras, Claudia |
| Abstract: | We experimentally evaluate a novel approach to IPV prevention that harnesses social media to recruit and engage men in a virtual support group delivered by trained male facilitators via WhatsApp. The program succeeded in recruiting men at high risk of committing IPV through self-targeting alone: 52% of partners of men who enroll in the program in response to social media ads report experiencing IPV at baseline, more than four times the national average and nearly twice the rates observed in men recruited through targeted and untargeted invitations. Moreover, on average, participation in the program reduced the probability that female partners report sexual violence at endline by 20%. Treatment effects are concentrated among younger men (-36%), men who exhibit violence at baseline (-27%), and among those whose wives report that they do not drink alcohol (-40%). Program effects are also highly sensitive to group composition, which was randomly assigned. Segregating individuals based on baseline risk appears to magnify program impacts on high-risk individuals, and hence the program impact overall. |
| Keywords: | intimate partner violence;gender based violence;virtual support groups;engaging men;behavioral science;Peru;IPV;GBV |
| JEL: | J12 J16 O15 O54 C93 |
| Date: | 2025–11 |
| URL: | https://d.repec.org/n?u=RePEc:idb:brikps:14362 |
| By: | Felipe Valencia-Clavijo |
| Abstract: | Large language models (LLMs) are increasingly examined as both behavioral subjects and decision systems, yet it remains unclear whether observed cognitive biases reflect surface imitation or deeper probability shifts. Anchoring bias, a classic human judgment bias, offers a critical test case. While prior work shows LLMs exhibit anchoring, most evidence relies on surface-level outputs, leaving internal mechanisms and attributional contributions unexplored. This paper advances the study of anchoring in LLMs through three contributions: (1) a log-probability-based behavioral analysis showing that anchors shift entire output distributions, with controls for training-data contamination; (2) exact Shapley-value attribution over structured prompt fields to quantify anchor influence on model log-probabilities; and (3) a unified Anchoring Bias Sensitivity Score integrating behavioral and attributional evidence across six open-source models. Results reveal robust anchoring effects in Gemma-2B, Phi-2, and Llama-2-7B, with attribution signaling that the anchors influence reweighting. Smaller models such as GPT-2, Falcon-RW-1B, and GPT-Neo-125M show variability, suggesting scale may modulate sensitivity. Attributional effects, however, vary across prompt designs, underscoring fragility in treating LLMs as human substitutes. The findings demonstrate that anchoring bias in LLMs is robust, measurable, and interpretable, while highlighting risks in applied domains. More broadly, the framework bridges behavioral science, LLM safety, and interpretability, offering a reproducible path for evaluating other cognitive biases in LLMs. |
| Date: | 2025–11 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2511.05766 |
| By: | Adamus, Magdalena (Masaryk University); Guzi, Martin (Masaryk University); Ballová Mikušková, Eva (Slovak Academy of Sciences) |
| Abstract: | The paper investigates gender biases and differential treatment of women and men in the business start-up phase. A sample of 498 entrepreneurs from Slovakia participated in an online experiment and evaluated three fictitious business plans in terms of the applicants’ competence, likeability, and business ability. The start-ups were positioned in three different sectors—cosmetics production, services provision, and software development—where men’s and women’s chances of success may be viewed differently. Following Goldberg’s paradigm, half of the evaluators received business plans presented as written by female and half by male applicants; otherwise the plans were identical. Results imply that female applicants are assessed similarly to male applicants, but more masculine evaluators assess women’s business plans and their potential in entrepreneurship more critically. The study advises caution in recommending more female evaluators in the business plan assessment. If women who become involved in entrepreneurship are excessively masculine and masculinity is associated with a less favourable evaluation of potential female entrepreneurs, such policies could backfire against women, putting them in a more disadvantaged position. |
| Keywords: | Goldberg paradigm, start-up, entrepreneurship, masculinity, gender identity, gender-role theory |
| JEL: | J16 M13 L26 |
| Date: | 2025–11 |
| URL: | https://d.repec.org/n?u=RePEc:iza:izadps:dp18245 |