AEA RCT Registry currently lists 11660 studies with locations in 170 countries.
While market-based sanitation (MBS) has successfully expanded access for many, purely market-driven approaches often fail to reach the poorest and most vulnerable households due to significant financial barriers. To address this equity gap, we are conducting a two-arm cluster randomized controlled trial of a poverty-targeted sanitation subsidy program integrated with iDE's MBS programming in the Wolayita zone of rural Ethiopia. 104 villages (gotts) were randomly assigned to either a control group receiving status-quo MBS programming or a treatment group receiving MBS integrated with targeted financial discount vouchers for eligible poor households. Following a baseline household listing in September–October 2024, subsidies were distributed between January and February 2025. We are retur...
Digitalization in banking is leaving elderly clients at risk of losing access to financial services, yet little is known about how to effectively encourage technology adoption at an advanced age. In a companion field experiment (AEARCTR-0013985), we find that while internet banking training is highly effective conditional on participation, take-up of training among the group of elderly non-adopters is very low (2% initial response rate). This survey experiment investigates how the design of training invitation materials can increase older adults’ willingness to participate in internet banking training. We conduct an online survey experiment with around 1,000 German respondents aged 50-75 who do not regularly use internet banking. Respondents are presented with a hypothetical invitation ...
This study aims to improve agricultural productivity in Northern Ghana by identifying the most effective method for selecting "seed farmers" to disseminate information about Integrated Soil Fertility Management (ISFM) technologies (e.g., improved maize varieties, fertilizer application, and intercropping). While selecting central individuals in a social network is known to speed up technology diffusion, traditional methods of mapping these networks are often too expensive and time-consuming for practical use in developing countries. This Randomized Controlled Trial (RCT) tests a novel, low-cost targeting method designed to identify influential farmers without a full census. The study compares four specific strategies for selecting seed farmers: a) Network Theory-Based Targeting: Using ...
The goal of this research is to investigate biased beliefs about personal saving behavior. Biased beliefs in this study are operationalized as overplacement (one of three forms of overconfidence), comparing an individual’s beliefs about personal saving behavior to an individual’s beliefs about saving behavior of others. We will investigate whether specificity of referent characteristics affects the level of overplacement. Moreover, we will investigate whether the specificity of the referent characteristics interacts with the perceived importance of saving. We designed an online experiment where the specificity of the referent characteristics are systematically varied. The results of the experiment will have relevance for understanding the role motivation plays in overplacement about per...
Young graduates in Côte d’Ivoire exhibit strong preferences for public-sector employment and appear to hold biased beliefs regarding sectoral wages and employment probabilities. This study evaluates whether providing accurate information about (i) average wages by sector, (ii) employment shares across sectors, or (iii) both, affects job search effort allocation, sectoral preferences, reservation wages, and employment outcomes. Individuals are randomly assigned to a control group or one of three treatment arms receiving sector-specific information tailored by gender and education level. Outcomes include sector-specific search effort, job offers received, employment status, wages, and sectoral sorting. The intervention is informational only and delivered via phone and optional video.
This randomized controlled trial studies how households interpret rising public debt and how these interpretations affect inflation expectations and consumption behavior. The study measures two primary outcomes: (1) inflation expectations and related macroeconomic beliefs, and (2) intertemporal consumption responses elicited through a structured hypothetical windfall allocation task that allows estimation of marginal propensities to consume, save, and repay debt. To distinguish behavioral responses from liquidity constraints, the survey includes a short balance-sheet module classifying households into liquidity types consistent with Heterogeneous Agent New Keynesian (HANK) models (poor hand-to-mouth, wealthy hand-to-mouth, unconstrained).
The incentive compatibility of the Becker-DeGroot-Marschak (BDM) mechanism implies valuations should not depend on the market offer range or where the induced value falls within that range. In our first experiment (AEARCTR-0016635), varying the support set's upper bound while holding the induced value fixed produces monotonic increases in mean willingness to accept. Risk preferences are not correlated with the size of the shift as hypothesized. This second experiment varies the induced value within a fixed range to test whether responses shift with value location. A similar shift in this environment implies a mechanical adjustment to the location of the induced value with respect to the range. Taken together, the experiments evaluate whether valuations are sensitive to response scale wi...
Does the degree of AI involvement in writing affect the success of written output? Specifically, we ask two questions. First, for social media posts (e.g., LinkedIn, X/Twitter), does AI-assisted writing improve engagement metrics such as impressions, likes, comments, shares, and follower growth? Second, for academic papers, does AI-assisted writing improve citation counts, journal acceptance rates, peer review quality scores, and altmetric attention? The deeper question is whether AI assistance generates a voltage gain or voltage drop when scaled from the author’s initial draft to published output. AI-polished writing may gain surface-level engagement but lose the idiosyncratic voice and rough edges that drive deeper resonance. Or it may do the opposite. We don’t know yet—which is prec...
Workplace sexual harassment is highly prevalent and harms women and the economy. We survey the UK population to provide the first estimates under a single definition of the prevalence of sexual harassment, its harms, people’s awareness of sexual harassment law, and indicators of policy effectiveness including reporting and redressal. In a separate survey, we elicit participants’ beliefs over these quantities and document the distribution of beliefs. We then experimentally vary information on prevalence, harms and policy (in)effectiveness and estimate impacts on indicators of stated and revealed preferences for policy and civil society action. Finally, we compare policymakers’ beliefs with citizens’.
This study investigates how the development of AI agent design and governance choices shape quality and productivity outcomes in an enterprise setting. We have advised a large enterprise running a `lab-in-the-field' style experiment to measure how different agentic AI approaches affect work. The experiment compares the effectiveness of developing bespoke agents using local knowledge, using a centrally developed agent, and using standard, out-of-the-box generative AI tools (Microsoft Copilot M365). This research will provide insights into how organizations can effectively leverage agentic AI in complex knowledge workflows.