Experimental Design

Angrist, Imbens, and Rubin formalize the LATE framework — originally introduced in Imbens and Angrist (1994) — within the Rubin Causal Model, providing a detailed treatment of the assumptions required for causal interpretation of IV estimates. This paper introduces the complier taxonomy (always-takers, never-takers, compliers, defiers) that is now standard in the IV literature. The practical implication is that IV estimates should be interpreted as local to the complier subpopulation, not as average effects for the entire population.

Athey and Imbens provide a modern, rigorous treatment of the econometrics behind randomized experiments. They cover design, analysis, and inference issues such as stratification, clustering, and multiple hypothesis testing. It is an excellent reference for researchers running field experiments.

Bruhn and McKenzie compare different randomization methods—simple, stratified, and pairwise—in practice and show that stratified randomization substantially improves balance on baseline covariates and increases statistical power. They provide practical recommendations for choosing among randomization procedures in field experiments.

Dunning provides a systematic framework for identifying and analyzing natural experiments across the social sciences. The book covers as-if random assignment, instrumental variables, regression discontinuity, and difference-in-differences through a unified design-based lens, making it essential reading for researchers exploiting natural variation for causal inference.

Fisher's classic book lays the foundations of experimental design, introducing concepts like randomization, blocking, and factorial designs. The 'lady tasting tea' example from this book remains one of the most famous illustrations of hypothesis testing and the logic of controlled experiments.

Harrison and List provide an influential taxonomy of field experiments, distinguishing artefactual, framed, and natural field experiments from conventional lab experiments. The paper helps establish field experiments as a mainstream methodology in economics.

Rubin formalizes the 'potential outcomes' framework that is now central to causal inference. The idea is simple but powerful: each unit has a potential outcome under treatment and under control, and the causal effect is the difference. This paper is the origin of what is now called the Rubin Causal Model.

Acquisti and Fong conduct a correspondence experiment using social media profiles to study hiring discrimination based on religion and sexual orientation. They find no significant national-level discrimination against Muslim or gay candidates, but significant anti-Muslim discrimination emerges in Republican-leaning areas. The paper illustrates how online information creates new channels for employment discrimination that vary with local attitudes.

Bandiera, Barankay, and Rasul use a field experiment in a fruit-picking firm to study how switching from relative to piece-rate pay affects productivity. They demonstrate that social preferences among workers matter for incentive design, bridging experimental economics and management.

Banerjee, Duflo, and colleagues conduct a large-scale RCT across six countries, demonstrating that a multifaceted anti-poverty program produces sustained economic gains for the ultra-poor. The study is notable for its multi-site design, which provides rare multi-country evidence on how the same intervention performs across diverse contexts. It demonstrates both the power of randomized evaluation at scale and the importance of bundled interventions when individual components may be insufficient.

Bertrand and Mullainathan send fictitious resumes with randomly assigned names to employers and find that 'white-sounding' names receive 50% more callbacks in this famous audit study. It is one of the most widely cited field experiments in social science and a powerful example of how randomization can identify discrimination.

Bloom and colleagues conduct a large-scale randomized experiment at a Chinese travel agency, finding that working from home leads to a 13% performance increase. The study becomes a landmark reference in management and labor economics for its clean experimental design applied to a practical workplace question.

Camuffo and colleagues conduct a randomized controlled trial with 116 Italian startups, randomly assigning half to receive training in a 'scientific' approach to entrepreneurial decision-making (formulating and testing hypotheses before committing resources). Treated startups perform better, are more likely to pivot, and are not more likely to drop out, providing experimental evidence that structured decision-making improves entrepreneurial outcomes.

Camuffo and colleagues conduct four randomized controlled trials with 759 firms across Italy, the UK, and India, replicating and extending their earlier finding that training entrepreneurs to adopt a 'scientific' approach to decision-making improves venture performance. The multi-site, multi-country design provides strong evidence on the external validity of the original RCT findings.

Chatterji, Findley, Jensen, Meier, and Nielson make the case for using field experiments in strategy research and provide practical guidance for doing so. They discuss internal validity, external validity, and ethical considerations specific to strategy scholars.

Crepon and colleagues evaluate a job placement assistance program in France using a two-step clustered randomization design that varies treatment intensity across 235 labor markets. The paper's key contribution is identifying displacement effects: treated job seekers gain at the expense of untreated competitors, particularly in weak labor markets and among workers with similar skills. This innovative experimental design allows estimation of both direct and indirect (general equilibrium) effects of active labor market policies.

Finkelstein and colleagues analyze the Oregon Health Insurance Experiment, in which uninsured low-income adults are selected by lottery for the chance to apply for Medicaid. Using this randomized controlled design with IV to handle noncompliance, they estimate the local average treatment effect of Medicaid coverage on health care utilization, financial strain, and self-reported health. The study demonstrates the practical difference between intent-to-treat and LATE estimates in a real-world experiment where not all lottery winners enrolled.

Friebel, Heinz, and Zubanov conduct a long-term randomized field experiment in a large Eastern European retail chain, in which the CEO asked treated store managers to reduce employee quit rates. The intervention decreased the quit rate by a fifth to a quarter, lasting nine months before petering out, but reappearing after a reminder. However, there is no treatment effect on sales, illustrating that reducing turnover does not automatically translate into improved store performance.

Gornall and Strebulaev conduct a large-scale correspondence experiment, sending approximately 80,000 pitch emails from fictitious startups to 28,000 venture capitalists and angel investors. By randomly varying the entrepreneur's name to signal gender and race, they find that female entrepreneurs received 9% more interested replies and Asian-surname entrepreneurs received 6% more responses than White-surname entrepreneurs, indicating favorable rather than adverse bias. The paper provides large-scale experimental evidence on investor response patterns by entrepreneur demographics in entrepreneurial finance.

Grant conducts field experiments showing that briefly exposing workers to the beneficiaries of their work significantly increased their motivation and performance. This paper is a well-known example of experimental design applied within organizational behavior research.

Hoogendoorn, Parker, and van Praag conduct a field experiment with 573 students randomly assigned to 49 startup teams that varied in cognitive ability dispersion. They find an inverted U-shaped relationship between ability dispersion and team performance, with moderately diverse teams in ability outperforming both homogeneous and highly dispersed teams. The random assignment to teams ensures that ability composition is exogenous, providing clean experimental identification of the effect of team cognitive diversity on venture performance.

Hurst, Lee, and Frake conduct a reverse audit study in partnership with a U.S. healthcare startup, sending recruitment emails to approximately 8,400 job seekers with randomly varied descriptions of the firm's organizational hierarchy. Featuring a flatter hierarchy did not significantly affect applicant pool size but significantly decreased women's representation, because women perceived flatter structures as offering fewer career advancement opportunities and greater workload burdens.

Jia, Luo, Fang, and Liao conduct a field experiment examining how AI assistance affects creative work through a sequential division of labor. They find that AI augmentation improves average output quality but reduces the novelty of top-performing work, with effects moderated by employee skill level. The paper provides causal evidence on the productivity implications of human-AI collaboration in knowledge work.

Kang and colleagues conduct a résumé audit study sending fictitious applications to real employers, finding that minority applicants who 'whitened' their résumés received significantly more callbacks. The study combines a correspondence experiment with qualitative interviews, providing a powerful example of how audit studies can identify discrimination in hiring.

Pongeluppe conducts a randomized controlled trial of a business training program offered to residents of Brazilian favelas, complementing the experiment with quantile regressions, field visits, and interviews. The results show that training improves economic outcomes such as income and entrepreneurship participation, but also intensifies participants' experiences of favela-related stigma, revealing that socioeconomic mobility can simultaneously generate material benefits and psychosocial costs.

Angrist and Pischke write one of the most influential modern textbooks on applied econometrics, organizing the field around a design-based approach to causal inference. The book provides essential treatments of instrumental variables, difference-in-differences, and regression discontinuity, each grounded in the potential outcomes framework. It remains the standard reference for graduate students learning to evaluate and implement identification strategies.

Duflo, Glennerster, and Kremer write a comprehensive practical guide to running randomized experiments in development economics. The chapter covers all stages from design to analysis, including power calculations, stratification, dealing with attrition, and estimating treatment effects with imperfect compliance. It has become required reading for anyone designing a field experiment.

Gerber and Green write a comprehensive textbook on field experiments covering randomization, blocking, clustering, noncompliance, and attrition. The book provides rigorous treatment of experimental design principles with practical guidance drawn from political science and public policy applications. It is particularly valuable for its coverage of complications that arise in real-world experiments, including how to handle noncompliance through intent-to-treat analysis and instrumental variables.

List, Sadoff, and Wagner provide rules of thumb for sample size, treatment assignment, and other design decisions in field experiments in this practical guide. It is a useful starting point for researchers planning their first experiment.

Wooldridge's graduate textbook is the standard reference for cross-section and panel data econometrics. Chapters 10-11 provide a thorough treatment of fixed effects, random effects, and related panel data methods, while later chapters cover general estimation methodology (MLE, GMM, M-estimation) with panel data applications throughout. The book covers both linear and nonlinear models with careful attention to assumptions.