Our Methodological Philosophy
The Economic Science Society maintains that the gold standard of economic research is rigorous causal identification. Whether analyzing historical datasets, leveraging algorithms, or deploying field interventions, the choice of methodology must isolate true causal mechanisms from mere correlation.
Novel Observational Estimators
The ESS champions the advancement of econometric techniques designed to extract robust causal insights from real-world, non-experimental data. Our members lead the field in applying and refining:
- Advanced Design-Based Inference: Developing modern approaches to Difference-in-Differences, Regression Discontinuity, and Instrumental Variables to overcome endogeneity in complex settings.
- Synthetic Controls & Matching: Utilizing sophisticated techniques to construct robust counterfactuals when randomized control groups are unavailable.
- High-Dimensional Panel Data: Creating cutting-edge methods for handling dynamic, longitudinal data structures to track causal effects over time.
Foundational Research:
- Manski, C. F. (1990). "Nonparametric Bounds on Treatment Effects." The American Economic Review.
- Angrist, J. D., & Krueger, A. B. (1991). "Does Compulsory School Attendance Affect Schooling and Earnings?" The Quarterly Journal of Economics.
- Imbens, G. W., & Angrist, J. D. (1994). "Identification and Estimation of Local Average Treatment Effects." Econometrica.
Artificial Intelligence & Machine Learning
Recognizing the transformative potential of computational power, the ESS promotes the principled integration of AI and machine learning into the causal framework. We focus on ensuring algorithmic tools enhance, rather than obscure, identification:
- Causal Machine Learning: Applying frameworks like Double/Debiased Machine Learning (DML) and Causal Forests to accurately estimate heterogeneous treatment effects.
- Generative Agent-Based Modeling: Utilizing AI-informed agents within agent-based models (ABM) to simulate complex, emergent market dynamics.
- High-Dimensional Confounding: Utilizing regularization algorithms for principled variable selection, allowing researchers to control for massive sets of covariates.
- Unstructured Data Analysis: Leveraging NLP and Large Language Models to reliably extract causal variables from text and visual archives.
Frontier Research:
- Athey, S., & Imbens, G. W. (2016). "Recursive partitioning for heterogeneous causal effects." PNAS.
- Chernozhukov, V., et al. (2018). "Double/debiased machine learning for treatment and structural parameters." The Econometrics Journal.
- Park, J., et al. (2023). "Generative Agents: Interactive Simulacra of Human Behavior." ACM UIST.
Innovative Research Design
While observational data and AI represent vital new frontiers, the ESS continues to value the power of tightly controlled research design as a core component of the causal toolkit:
- Field & Natural Experiments: Designing robust interventions that bridge the gap between controlled environments and real-world behavior.
- Adaptive Designs: Utilizing algorithms to dynamically adjust experimental parameters in real-time, maximizing information gain.
Foundational Research:
- Smith, V. L. (1962). "An Experimental Study of Competitive Market Behavior." Journal of Political Economy.
- Miguel, E., & Kremer, M. (2004). "Worms: Identifying Impacts on Education and Health." Econometrica.
- Banerjee, A., Duflo, E., et al. (2015). "A multifaceted program causes lasting progress for the very poor." Science.
Open Science & Replicability
Methodological advancement requires a foundation of absolute transparency. The ESS advocates for standardizing open science practices across the discipline:
- Pre-registration: We strongly encourage the public logging of pre-analysis plans to separate exploratory analysis from confirmatory testing.
- Computational Reproducibility: We mandate comprehensive guidelines for sharing raw data, clean pipelines, and well-annotated code in public repositories.
