Economics Research You Can Actually Understand

Veetil (2026). Applies network analysis to India's import dependency structure to map the potential transmission of economic sanctions through trade networks. Identifies critical chokepoints in energy, semiconductors and pharmaceutical precursors — sectors where India's current diversification strategy leaves significant exposure. An empirically grounded contribution to a pressing policy debate.

Ciganovic, D'Amario & Tancioni (2026). Double machine learning (DML) has transformed causal inference in cross-sectional econometrics, but applying it to time series data with autocorrelation and non-stationarity requires new theoretical guarantees. This paper extends DML to time series settings with rigorous inference theory, enabling causal estimation in macroeconomic and financial panel data.

Rawat (2026). A comprehensive review mapping the intersection of reinforcement learning methods and economic modelling — covering dynamic programming for optimal control, multi-agent RL for game theory, and deep RL for heterogeneous agent macroeconomics. Essential reading for economists entering the RL literature and for RL researchers looking for economically meaningful problem domains.

Li & Zhang (2026). Proposes a formal linear model of geopolitical competition that can be estimated from observable trade and alliance data, making predictions about equilibrium coalition structures and conflict probability. Unlike most formal geopolitical models, it remains tractable enough to test empirically against historical data — a significant methodological contribution.

Tchuente (2026). Decentralised Autonomous Organisations have become significant economic actors but their governance dynamics are poorly understood. This paper models DAO governance as a monitoring game and identifies capacity breakpoints above which rational voters rationally abstain, leading to endogenous voting power concentration — explaining the whale-dominated dynamics observed in real DAOs.

Epping, Caplin, Duhaime, Holmes, Martin & Trueblood (2026). A field experiment run with a real labelling workforce studying how cognitive bias affects the quality of human annotations for rare-event detection tasks. Demonstrates that standard quality control methods fail to detect or correct the specific bias patterns that affect rare-event labelling — with direct implications for AI safety and medical AI data pipelines.

Luo, MacKay & Chater (2026). Challenges the standard approach of deriving macroeconomic models from individual optimisation, instead proposing entropy as a measure of the aggregate properties of exchange economies. Agent-based simulations show that entropy captures macroeconomic dynamics that microfounded models miss — a heterodox contribution with significant implications for how we model recessions and crises.

Chen, Tamer & Yao (2026). Online (sequential) learning methods from machine learning can update estimates as new data arrives without re-fitting from scratch — a critical capability for real-time economic forecasting. This paper establishes theoretical guarantees for online learning in semiparametric models, bridging a gap between the econometric and machine learning literatures.

Kock, Frazier, Smith & Nott (2026). Copula models are essential tools for modelling multivariate financial dependence structures, but their inference is sensitive to marginal distribution misspecification. Bayesian modular inference provides a framework for fitting the dependence structure robustly even when the marginal models are approximately wrong — directly relevant for risk management under model uncertainty.

Veetil (2026). Applies network analysis to India's import dependency structure to map the potential transmission of economic sanctions through trade networks. Identifies critical chokepoints in energy, semiconductors and pharmaceutical precursors — sectors where India's current diversification strategy leaves significant exposure. An empirically grounded contribution to a pressing policy debate.

Ciganovic, D'Amario & Tancioni (2026). Double machine learning (DML) has transformed causal inference in cross-sectional econometrics, but applying it to time series data with autocorrelation and non-stationarity requires new theoretical guarantees. This paper extends DML to time series settings with rigorous inference theory, enabling causal estimation in macroeconomic and financial panel data.

Rawat (2026). A comprehensive review mapping the intersection of reinforcement learning methods and economic modelling — covering dynamic programming for optimal control, multi-agent RL for game theory, and deep RL for heterogeneous agent macroeconomics. Essential reading for economists entering the RL literature and for RL researchers looking for economically meaningful problem domains.

Li & Zhang (2026). Proposes a formal linear model of geopolitical competition that can be estimated from observable trade and alliance data, making predictions about equilibrium coalition structures and conflict probability. Unlike most formal geopolitical models, it remains tractable enough to test empirically against historical data — a significant methodological contribution.

Tchuente (2026). Decentralised Autonomous Organisations have become significant economic actors but their governance dynamics are poorly understood. This paper models DAO governance as a monitoring game and identifies capacity breakpoints above which rational voters rationally abstain, leading to endogenous voting power concentration — explaining the whale-dominated dynamics observed in real DAOs.

Epping, Caplin, Duhaime, Holmes, Martin & Trueblood (2026). A field experiment run with a real labelling workforce studying how cognitive bias affects the quality of human annotations for rare-event detection tasks. Demonstrates that standard quality control methods fail to detect or correct the specific bias patterns that affect rare-event labelling — with direct implications for AI safety and medical AI data pipelines.

Luo, MacKay & Chater (2026). Challenges the standard approach of deriving macroeconomic models from individual optimisation, instead proposing entropy as a measure of the aggregate properties of exchange economies. Agent-based simulations show that entropy captures macroeconomic dynamics that microfounded models miss — a heterodox contribution with significant implications for how we model recessions and crises.

Chen, Tamer & Yao (2026). Online (sequential) learning methods from machine learning can update estimates as new data arrives without re-fitting from scratch — a critical capability for real-time economic forecasting. This paper establishes theoretical guarantees for online learning in semiparametric models, bridging a gap between the econometric and machine learning literatures.

Kock, Frazier, Smith & Nott (2026). Copula models are essential tools for modelling multivariate financial dependence structures, but their inference is sensitive to marginal distribution misspecification. Bayesian modular inference provides a framework for fitting the dependence structure robustly even when the marginal models are approximately wrong — directly relevant for risk management under model uncertainty.