|
on Experimental Economics |
| By: | Chen Wang; Shichao Han; Shan Huang |
| Abstract: | Participants in online experiments often enroll over time, which can compromise sample representativeness due to temporal shifts in covariates. This issue is particularly critical in A/B tests, online controlled experiments extensively used to evaluate product updates, since these tests are cost-sensitive and typically short in duration. We propose a novel framework that dynamically assesses sample representativeness by dividing the ongoing sampling process into three stages. We then develop stage-specific estimators for Population Average Treatment Effects (PATE), ensuring that experimental results remain generalizable across varying experiment durations. Leveraging survival analysis, we develop a heuristic function that identifies these stages without requiring prior knowledge of population or sample characteristics, thereby keeping implementation costs low. Our approach bridges the gap between experimental findings and real-world applicability, enabling product decisions to be based on evidence that accurately represents the broader target population. We validate the effectiveness of our framework on three levels: (1) through a real-world online experiment conducted on WeChat; (2) via a synthetic experiment; and (3) by applying it to 600 A/B tests on WeChat in a platform-wide application. Additionally, we provide practical guidelines for practitioners to implement our method in real-world settings. |
| Date: | 2025–02 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2502.18253 |
| By: | Cruces, Guillermo (University of Nottingham); Tortarolo, Dario (World Bank); Vazquez-Bare, Gonzalo (UC Santa Barbara) |
| Abstract: | This paper develops a framework to analyze partial population experiments, a generalization of the cluster experimental design where clusters are assigned to different treatment intensities. The framework allows for heterogeneity in cluster sizes and outcome distributions. The paper studies the large-sample behavior of OLS estimators and cluster-robust variance estimators and shows that (i) ignoring cluster heterogeneity may result in severely underpowered experiments and (ii) the clusterrobust variance estimator may be upward-biased when clusters are heterogeneous. The paper derives formulas for power, minimum detectable effects, and optimal cluster assignment probabilities. All the results apply to cluster experiments, a particular case of the framework. The paper sets up a potential outcomes framework to interpret the OLS estimands as causal effects. It implements the methods in a large-scale experiment to estimate the direct and spillover effects of a communication campaign on property tax compliance. The analysis reveals an increase in tax compliance among individuals directly targeted with the mailing, as well as compliance spillovers on untreated individuals in clusters with a high proportion of treated taxpayers. |
| Keywords: | cluster experiments, randomized controlled trials, spillovers, partial population experiments, two-stage designs, property tax, tax compliance |
| JEL: | C01 C93 H71 H71 H26 H26 H21 H21 O23 |
| Date: | 2025–02 |
| URL: | https://d.repec.org/n?u=RePEc:iza:izadps:dp17692 |
| By: | Kevin He; Ran Shorrer; Mengjia Xia |
| Abstract: | We conduct an incentivized laboratory experiment to study people's perception of generative artificial intelligence (GenAI) alignment in the context of economic decision-making. Using a panel of economic problems spanning the domains of risk, time preference, social preference, and strategic interactions, we ask human subjects to make choices for themselves and to predict the choices made by GenAI on behalf of a human user. We find that people overestimate the degree of alignment between GenAI's choices and human choices. In every problem, human subjects' average prediction about GenAI's choice is substantially closer to the average human-subject choice than it is to the GenAI choice. At the individual level, different subjects' predictions about GenAI's choice in a given problem are highly correlated with their own choices in the same problem. We explore the implications of people overestimating GenAI alignment in a simple theoretical model. |
| Date: | 2025–02 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2502.14708 |
| By: | Carcillo, Stéphane (Sciences Po); Valfort, Marie-Anne (Paris School of Economics); Vergara Merino, Pedro (CREST-ENSAE) |
| Abstract: | This paper presents the first rigorous evaluation of school-based interventions aimed at reducing LGBTphobia. We focus on a classroom intervention that addresses the issue of LGBT harassment through perspective-taking and narrative exchange. Using a field experiment in France with more than 10, 000 middle and high school students, we find robust evidence of strong positive effects, with variations across gender, age, and socio-economic status. We argue that changing perceptions of group norms is a key channel driving these heterogeneous effects. |
| Keywords: | LGBT, discrimination, social norms |
| JEL: | C93 J15 J16 J71 |
| Date: | 2025–02 |
| URL: | https://d.repec.org/n?u=RePEc:iza:izadps:dp17683 |
| By: | Mahyar Habibi; Zahra Khanalizadeh; Negar Ziaeian |
| Abstract: | Interdependencies between units in online two-sided marketplaces complicate estimating causal effects in experimental settings. We propose a novel experimental design to mitigate the interference bias in estimating the total average treatment effect (TATE) of item-side interventions in online two-sided marketplaces. Our Two-Sided Prioritized Ranking (TSPR) design uses the recommender system as an instrument for experimentation. TSPR strategically prioritizes items based on their treatment status in the listings displayed to users. We designed TSPR to provide users with a coherent platform experience by ensuring access to all items and a consistent realization of their treatment by all users. We evaluate our experimental design through simulations using a search impression dataset from an online travel agency. Our methodology closely estimates the true simulated TATE, while a baseline item-side estimator significantly overestimates TATE. |
| Date: | 2025–02 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2502.09806 |
| By: | O'Szabo, Rebeka; Deritei, David; Battiston, Federico |
| Abstract: | Collective intelligence–-the ability of groups to solve diverse problems-–has been explored using laboratory experiments, computer simulations, and questionnaires. These instruments, however, suffer from limitations, such as external validity in the case of laboratory experiments and self-reporting bias in the case of questionnaires. Here we investigate the exploration-exploitation dynamics of small teams using high-frequency, observational data from escape rooms: a non-interventional yet controlled environment where naturally occurring teams solve connected sequences of exploration and exploitation tasks. We find that more effective teams tend to coordinate throughout problem-solving, exhibit balanced communication patterns, and maintain a dynamic alternation between exploration and exploitation tasks. In contrast, members of less effective teams often work in isolation, participate in problem-solving unequally, and lacking the capacity to exploit information efficiently. Moreover, we show that the effect of collaborative behavior depends on the task: exploitation benefits from team-wide communication and dominance of key members, while exploration requires balanced participation. Additionally, positive exchanges accelerate the compilation of exploitation tasks while negative communication decelerates them. These findings expand the external validity of experimental work on collective intelligence to an organic non-interventional setting and highlight the importance of understanding team performance through behavior instead of team demographics. |
| Date: | 2024–12–12 |
| URL: | https://d.repec.org/n?u=RePEc:osf:osfxxx:4rwpd_v1 |
| By: | Buser, Thomas (University of Amsterdam); Sangi, Sahar (University of Amsterdam) |
| Abstract: | Competitive environments often leave room for "dirty" practices such as sabotage, retaliation, or dishonesty. We use an online experiment to document aggregate levels and individual differences in the willingness to engage in dirty competition and in the willingness to enter competitions where the opponent may play dirty. We then use the experimental data to validate a set of survey questions that capture willingness to engage in dirty competition above general willingness to compete. We elicit these questions in a representative survey panel and show that willingness to engage in dirty competition is a strong predictor of holding a management or supervisory position and of working in the private – versus the public – sector, but also of worse self-esteem, worse social relationships, and increased feelings of guilt and shame. Men, younger people, and lower-educated people are on average more willing to engage in dirty competition. |
| Keywords: | preferences, personaility, career choice |
| JEL: | D91 J24 |
| Date: | 2025–02 |
| URL: | https://d.repec.org/n?u=RePEc:iza:izadps:dp17676 |
| By: | Bandiera, Antonella (ITAM); , Rojas Daniel |
| Abstract: | This paper examines the effectiveness of media literacy interventions in combating misinformation among in-transit migrants in Mexico and Colombia. We conducted experiments to study whether an established strategy for fighting misinformation works for this understudied yet particularly vulnerable population. We evaluate the effect of digital media literacy tips on migrants' ability to identify false information and their intentions to share migration-related content. We find that these interventions can effectively decrease migrants' intentions to share misleading migration-related information, with a significantly larger reduction observed for false content than accurate information. We also find that prompting participants to think about accuracy can unintentionally obscure sharing intent by acting as a nudge. Additionally, the interventions decreased trust in social media as an information source while maintaining trust in official sources. The findings suggest that incorporating digital literacy tips into official websites could be a cost-effective strategy to reduce misinformation circulation among migrant populations. |
| Date: | 2024–12–13 |
| URL: | https://d.repec.org/n?u=RePEc:osf:osfxxx:md42a_v1 |
| By: | Chakraborty, Anujit (University of California, Davis); Henkel, Luca (Erasmus University Rotterdam) |
| Abstract: | In prosocial decisions, decision-makers face interpersonal uncertainty–uncertainty about how their choices impact others' utility. We use three approaches to show how it shapes classic patterns of prosocial behavior like ingroup favoritism, merit-based fairness, and self-favoring behavior. First, we compare standard allocation decisions with decisions where we remove social consequences but retain uncertainty, revealing strikingly similar patterns across both. Second, we exogenously vary interpersonal uncertainty to estimate the aversion to interpersonal uncertainty and quantify how it combines with preferences to determine prosocial decisions. Finally, we show that self-reported interpersonal uncertainty systematic ally predicts behavior across individuals, choice patterns, and behavioral interventions. |
| Keywords: | prosocial behavior, decision-making under uncertainty, interpersonal uncertainty, ingroup favoritism, merit-based fairness, self-favoring behavior |
| JEL: | C91 D01 D91 |
| Date: | 2025–02 |
| URL: | https://d.repec.org/n?u=RePEc:iza:izadps:dp17708 |
| By: | Liu, Chuanjun (Sichuan University); Zou, Lemei; Wu, Junhong; Wang, Taolin; Abbas, Syed Zain |
| Abstract: | Previous studies have highlighted the paradoxical nature of unethical pro-organizational behavior (UPB), yet no measurement tools specifically targeting this paradox have been developed. This study introduces a new thought experiment measure based on the five-phase development framework by Hinkin (1995, 1998). Phase one, concept clarification, adopted the widely accepted UPB definition from Umphress et al. (2010). Phase two developed and initially validated seven UPB scenarios with high content and ecological validity, capturing the paradox that employees experience regarding UPB (Study 1). Phase three demonstrated that the employees’ UPB choices in these scenarios showed high internal consistency and structural validity (Study 2). Phase four demonstrated high criterion and discriminant validity compared to Umphress et al.’s UPB scale, along with high test-retest reliability (Study 3). Phase five examined the cognitive process underlying UPB through the lens of dual-process theory (Study 4). Cognitive load manipulations (with or without load) revealed that employees under cognitive load made faster decisions but exhibited lower UPB choice frequency, thereby supporting the paradoxical nature of UPB and its cognitive underpinnings. Theoretical and practical implications were discussed. |
| Date: | 2024–04–13 |
| URL: | https://d.repec.org/n?u=RePEc:osf:osfxxx:uamxy_v1 |
| By: | Diem, Andrea (Swiss Coordination Centre for Research in Education); Gschwendt, Christian (University of Bern); Wolter, Stefan C. (University of Bern) |
| Abstract: | A university degree is a risky investment because of the non-negligible risk of having to drop out of university without graduating. However, the costs of this risk are controversial, as it is often argued that even an uncertified year of study has a value in the labor market. To determine this value causally, however, alternatives to studying must also be considered, which is done here with the help of a discrete choice experiment with a representative sample of over 2, 500 HR recruiters. The result is that dropping out of university with a major closely related to an advertised job leads to similar labor market outcomes as if someone had not studied at all. Without a direct link to a job, however, dropping out of university significantly reduces lifetime earnings. Furthermore, HR recruiters clearly prefer applicants who have used the years without studying for human capital accumulation in an alternative way, for example in the form of a traineeship. |
| Keywords: | dropouts, hiring decisions, discrete choice experiment, sheepskin effect, willingness to pay, tertiary education |
| JEL: | I26 J23 J24 J31 M51 |
| Date: | 2025–02 |
| URL: | https://d.repec.org/n?u=RePEc:iza:izadps:dp17693 |
| By: | Hein, Ilka; Cecil, Julia (Ludwig-Maximilians-Universität München); Lermer, Eva (LMU Munich) |
| Abstract: | Artificial intelligence (AI) is increasingly taking over leadership tasks in companies, including the provision of feedback. However, the effect of AI-driven feedback on employees and its theoretical foundations are poorly understood. We aimed to close this research gap by comparing perceptions of AI and human feedback based on construal level theory and the feedback process model. Using these theories, our objective was also to investigate the moderating role of feedback valence and the mediating effect of social distance. A 2 × 2 between-subjects design was applied to manipulate feedback source (human vs. AI) and valence (negative vs. positive) via vignettes. In a preregistered experimental study (S1) and subsequent direct replication (S2), responses from NS1 = 263 and NS2 = 449 participants were studied who completed a German online questionnaire asking for feedback acceptance, performance motivation, social distance, acceptance of the feedback source itself, and intention to seek further feedback. Regression analyses showed that AI feedback was rated as less accurate and led to lower performance motivation, acceptance of the feedback provider, and intention to seek further feedback. These effects were mediated by perceived social distance. Moreover, for feedback acceptance and performance motivation, the differences were only found for positive but not for negative feedback in the first study. This implies that AI feedback may not inherently be perceived as more negatively than human feedback as it depends on the feedback's valence. Furthermore, the mediation effects indicate that the shown negative evaluations of the AI can be explained by higher social distance and that increased social closeness to feedback providers may improve appraisals of them and of their feedback. Theoretical contributions of the studies and implications for the use of AI for providing feedback in the workplace are discussed, emphasizing the influence of effects related to construal level theory. |
| Date: | 2024–12–22 |
| URL: | https://d.repec.org/n?u=RePEc:osf:osfxxx:uczaw_v1 |
| By: | Vazirani, Ashish; Bhattacharjee, Titas |
| Abstract: | New venture investment decisions reflect misvaluations as many deserving proposals are rejected, while most invested proposals fail to provide the required financial returns. This suggests that investors overvalue and undervalue investment proposals. We have used the arguments of dual process theory and salience theory to explain the cause of misvaluations. An experimental setup is used where the treated group is given time constraints to make the investment. Results show that investors under time constraints overvalue investment proposals, while reference-based comparison of salient information signals results in undervaluation. Results further show a significant impact of time constraints on the reference for comparison of proposals. Investors under time constraints consider only the last proposal as a reference to compare salient information signals, in contrast, they consider information from multiple previous proposals in the absence of time constraints. This study provides the specific source of misvaluations as per the context of overvaluation and undervaluation of new venture investment proposals. |
| Date: | 2024–04–04 |
| URL: | https://d.repec.org/n?u=RePEc:osf:osfxxx:az4yr_v1 |
| By: | Jahani, Eaman; Manning, Benjamin; Zhang, Joe; TuYe, Hong-Yi; Alsobay, Mohammed Abdulrahman M; Nicolaides, Christos (University of Cyprus); Suri, Siddharth; Holtz, David (University of California, Berkeley) |
| Abstract: | The recent surge in generative AI has led to new models being introduced almost every month. In light of this rapid progression, we pose and address a central question: to what extent must prompts evolve as the capabilities of generative AI models advance? To answer this question, we conducted an online experiment with N = 1, 893 participants where each participant was incentivized to write prompts to reproduce a target image as closely as possible in 10 consecutive tries. Each participant was randomly and blindly assigned to use one of three text-to-image diffusion models: DALL-E 2, its more advanced successor, DALL-E 3, or a version of DALL-E 3 with automatic prompt revision. In total, we collected and analyzed over 18, 000 prompts and over 300, 000 images. We find that task performance was higher for participants using DALL-E 3 than for those using DALL-E 2. This performance gap corresponds to a noticeable difference in the similarity of participants’ images to their target images, and was caused in equal measure by: (1) the increased technical capabilities of DALL-E 3, and (2) endogenous changes in participants’ prompting in response to these increased capabilities. Furthermore, while participants assigned to DALL-E 3 with prompt revision still outperformed those assigned to DALL-E 2, automatic prompt revision reduced the benefits of using DALL-E 3 by 58%. Our results suggest that for generative AI to realize its full impact on the global economy, people, firms, and institutions will need to update their prompts in response to new models. |
| Date: | 2024–07–22 |
| URL: | https://d.repec.org/n?u=RePEc:osf:osfxxx:9rhku_v1 |
| By: | Campos, Francisco (World Bank); Frese, Michael (Leuphana University Lüneburg); Iacovone, Leonardo (World Bank); Johnson, Hillary C. (World Bank); McKenzie, David (World Bank); Mensmann, Mona (University of Cologne) |
| Abstract: | A randomized experiment in Togo found that personal initiative training for small businesses resulted in large and significant impacts for both men and women after two years. We revisit these entrepreneurs after seven years, and find long-lasting average impacts of personal initiative training of $91 higher profits per month, which is larger than the 2-year impacts. However, these long-term impacts are very different for men and women: the impact for men grows over time as they accumulate more capital and increase self-efficacy, whereas the impact for women is flat or declines, and capital build-up is much more limited. |
| Keywords: | microentrepreneurship, business training, personal initiative, firm growth |
| JEL: | O12 O17 L26 J24 J16 D22 |
| Date: | 2025–02 |
| URL: | https://d.repec.org/n?u=RePEc:iza:izadps:dp17672 |
| By: | Cavanagh, Jack; Fliegner, Jasmin Claire; Kopper, Sarah; Sautmann, Anja |
| Abstract: | The use of randomized controlled trials (RCTs) in the social sciences has greatly expanded, resulting in newly abundant, high-quality data that can be reused to perform methods research in program evaluation, to systematize evidence for policymakers, and for replication and training purposes. However, potential users of RCT data often face significant barriers to discovery and reuse. This paper proposes a metadata schema that standardizes RCT data documentation and can serve as the basis for one—or many, interoperable —data catalogs that make such data easily findable, searchable, and comparable, and thus more readily reusable for secondary research. The schema is designed to document the unique properties of RCT data. Its set of fields and associated encoding schemes (acceptable formats and values) can be used to describe any dataset associated with a social science RCT. The paper also makes recommendations for implementing a catalog or database based on this metadata schema. |
| Date: | 2023–02–06 |
| URL: | https://d.repec.org/n?u=RePEc:wbk:wbrwps:10296 |
| By: | Premand, Patrick; Rohner, Dominic Patrick |
| Abstract: | Conflict undermines development, while poverty, in turn, breeds conflict. Policy interventions such as cash transfers could lower engagement in conflict by raising poor households' welfare and productivity. However, cash transfers may also trigger appropriation or looting of cash or assets. The expansion of government programs may further attract attacks to undermine state legitimacy. To investigate the net effect across these forces, this paper studies the impact of cash transfers on conflict in Niger. The analysis relies on the large-scale randomization of a government-led cash transfer program among nearly 4, 000 villages over seven years, combined with geo-referenced conflict events that draw on media and nongovernmental organization reports from a wide variety of international and domestic sources. The findings show that cash transfers did not result in greater pacification but—if anything—triggered a short-term increase in conflict events, which were to a large extent driven by terrorist attacks by foreign rebel groups (such as Boko Haram) that could have incentives to “sabotage” successful government programs. |
| Date: | 2023–02–06 |
| URL: | https://d.repec.org/n?u=RePEc:wbk:wbrwps:10293 |
| By: | Thomas Henning; Siddhartha M. Ojha; Ross Spoon; Jiatong Han; Colin F. Camerer |
| Abstract: | This paper explores how Large Language Models (LLMs) behave in a classic experimental finance paradigm widely known for eliciting bubbles and crashes in human participants. We adapt an established trading design, where traders buy and sell a risky asset with a known fundamental value, and introduce several LLM-based agents, both in single-model markets (all traders are instances of the same LLM) and in mixed-model "battle royale" settings (multiple LLMs competing in the same market). Our findings reveal that LLMs generally exhibit a "textbook-rational" approach, pricing the asset near its fundamental value, and show only a muted tendency toward bubble formation. Further analyses indicate that LLM-based agents display less trading strategy variance in contrast to humans. Taken together, these results highlight the risk of relying on LLM-only data to replicate human-driven market phenomena, as key behavioral features, such as large emergent bubbles, were not robustly reproduced. While LLMs clearly possess the capacity for strategic decision-making, their relative consistency and rationality suggest that they do not accurately mimic human market dynamics. |
| Date: | 2025–02 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2502.15800 |
| By: | Marina Kontalexi; Alexandros Gelastopoulos; Pantelis P. Analytis |
| Abstract: | Theoretical work on sequential choice and large-scale experiments in online ranking and voting systems has demonstrated that social influence can have a drastic impact on social and technological systems. Yet, the effect of social influence on online rating systems remains understudied and the few existing contributions suggest that online ratings would self-correct given enough users. Here, we propose a new framework for studying the effect of social influence on online ratings. We start from the assumption that people are influenced linearly by the observed average rating, but postulate that their propensity to be influenced varies. When the weight people assign to the observed average depends only on their own latent rating, the resulting system is linear, but the long-term rating may substantially deviate from the true mean rating. When the weight people put on the observed average depends on both their own latent rating and the observed average rating, the resulting system is non-linear, and may support multiple equilibria, suggesting that ratings might be path-dependent and deviations dramatic. Our results highlight potential limitations in crowdsourced information aggregation and can inform the design of more robust online rating systems. |
| Date: | 2025–02 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2502.19861 |
| By: | Clarke, Natasha; Pechey, Emily; Shemilt, Ian (University College London); Pilling, Mark Andrew Dr (University of Cambridge); Roberts, Nia; Marteau, Theresa Mary; Jebb, Susan; Hollands, Gareth J (University College London) |
| Abstract: | Background: Overconsumption of food and consumption of any amount of alcohol increases the risk of non-communicable diseases. Calorie (energy) labelling is advocated as a means to reduce energy intake from food and alcoholic drinks. However, there is continued uncertainty about these potential impacts, with a 2018 Cochrane review identifying only a small body of low-certainty evidence. This review updates and extends the 2018 Cochrane review to provide a timely reassessment of evidence for the effects of calorie labelling on people's selection and consumption of food or alcoholic drinks. Objectives: – To estimate the effect of calorie labelling for food (including non-alcoholic drinks) and alcoholic drinks on selection (with or without purchasing) and consumption. – To assess possible modifiers – label type, setting, and socioeconomic status – of the effect of calorie labelling on selection (with or without purchasing) and consumption of food and alcohol. Search methods: We searched CENTRAL, MEDLINE, Embase, PsycINFO, five other published or grey literature databases, trial registries, and key websites, followed by backwards and forwards citation searches. Using a semi-automated workflow, we searched for and selected records and corresponding reports of eligible studies, with these searches current to 2 August 2021. Updated searches were conducted in September 2023 but their results are not fully integrated into this version of the review. Selection criteria: Eligible studies were randomised controlled trials (RCTs) or quasi-RCTs with between-subjects (parallel group) or within-subjects (cross-over) designs, interrupted time series studies, or controlled before-after studies comparing calorie labelling with no calorie labelling, applied to food (including non-alcoholic drinks) or alcoholic drinks. Eligible studies also needed to objectively measure participants' selection (with or without purchasing) or consumption, in real-world, naturalistic laboratory, or laboratory settings. Data collection and analysis: Two review authors independently selected studies for inclusion and extracted study data. We applied the Cochrane RoB 2 tool and ROBINS-I to assess risk of bias in included studies. Where possible, we used (random-effects) meta-analyses to estimate summary effect sizes as standardised mean differences (SMDs) with 95% confidence intervals (CIs), and subgroup analyses to investigate potential effect modifiers, including study, intervention, and participant characteristics. We synthesised data from other studies in a narrative summary. We rated the certainty of evidence using GRADE. Main results: We included 25 studies (23 food, 2 alcohol and food), comprising 18 RCTs, one quasi-RCT, two interrupted time series studies, and four controlled before-after studies. Most studies were conducted in real-world field settings (16/25, with 13 of these in restaurants or cafeterias and three in supermarkets); six studies were conducted in naturalistic laboratories that attempted to mimic a real-world setting; and three studies were conducted in laboratory settings. Most studies assessed the impact of calorie labelling on menus or menu boards (18/25); six studies assessed the impact of calorie labelling directly on, or placed adjacent to, products or their packaging; and one study assessed labels on both menus and on product packaging. The most frequently assessed labelling type was simple calorie labelling (20/25), with other studies assessing calorie labelling with information about at least one other nutrient, or calories with physical activity calorie equivalent (PACE) labelling (or both). Twenty-four studies were conducted in high-income countries, with 15 in the USA, six in the UK, one in Ireland, one in France, and one in Canada. Most studies (18/25) were conducted in high socioeconomic status populations, while six studies included both low and high socioeconomic groups, and one study included only participants from low socioeconomic groups. Twenty-four studies included a measure of selection of food (with or without purchasing), most of which measured selection with purchasing (17/24), and eight studies included a measure of consumption of food. Calorie labelling of food led to a small reduction in energy selected (SMD −0.06, 95% CI −0.08 to −0.03; 16 randomised studies, 19 comparisons, 9850 participants; high-certainty evidence), with near-identical effects when including only studies at low risk of bias, and when including only studies of selection with purchasing. There may be a larger reduction in consumption (SMD −0.19, 95% CI −0.33 to −0.05; 8 randomised studies, 10 comparisons, 2134 participants; low-certainty evidence). These effect sizes suggest that, for an average meal of 600 kcal, adults exposed to calorie labelling would select 11 kcal less (equivalent to a 1.8% reduction), and consume 35 kcal less (equivalent to a 5.9% reduction). The direction of effect observed in the six non-randomised studies was broadly consistent with that observed in the 16 randomised studies. Only two studies focused on alcoholic drinks, and these studies also included a measure of selection of food (including non-alcoholic drinks). Their results were inconclusive, with inconsistent effects and wide 95% CIs encompassing both harm and benefit, and the evidence was of very low certainty. Authors' conclusions: Current evidence suggests that calorie labelling of food (including non-alcoholic drinks) on menus, products, and packaging leads to small reductions in energy selected and purchased, with potentially meaningful impacts on population health when applied at scale. The evidence assessing the impact of calorie labelling of food on consumption suggests a similar effect to that observed for selection and purchasing, although there is less evidence and it is of lower certainty. There is insufficient evidence to estimate the effect of calorie labelling of alcoholic drinks, and more high-quality studies are needed. Further research is needed to assess potential moderators of the intervention effect observed for food, particularly socioeconomic status. Wider potential effects of implementation that are not assessed by this review also merit further examination, including systemic impacts of calorie labelling on industry actions, and potential individual harms and benefits. |
| Date: | 2025–01–17 |
| URL: | https://d.repec.org/n?u=RePEc:osf:osfxxx:pwhs5_v1 |
| By: | Jibang Wu; Chenghao Yang; Simon Mahns; Chaoqi Wang; Hao Zhu; Fei Fang; Haifeng Xu |
| Abstract: | This paper develops an agentic framework that employs large language models (LLMs) to automate the generation of persuasive and grounded marketing content, using real estate listing descriptions as our focal application domain. Our method is designed to align the generated content with user preferences while highlighting useful factual attributes. This agent consists of three key modules: (1) Grounding Module, mimicking expert human behavior to predict marketable features; (2) Personalization Module, aligning content with user preferences; (3) Marketing Module, ensuring factual accuracy and the inclusion of localized features. We conduct systematic human-subject experiments in the domain of real estate marketing, with a focus group of potential house buyers. The results demonstrate that marketing descriptions generated by our approach are preferred over those written by human experts by a clear margin. Our findings suggest a promising LLM-based agentic framework to automate large-scale targeted marketing while ensuring responsible generation using only facts. |
| Date: | 2025–02 |
| URL: | https://d.repec.org/n?u=RePEc:arx:papers:2502.16810 |