Monday, 7 September 2026
17 new items across 4 threads · 32 screened · 48 of 50 sources reached
1 item
archival
Uses newly consulted Central Bank of Chile board minutes, 1925-1933, to show profit-seeking motives, not just ideology or credibility, shaped adherence to the interwar gold exchange standard.
15 items
Derives minimax-optimal designs and estimators for bounded (e.g. binary) potential outcomes, replacing balanced randomization and difference-in-means with an unconventional recentered regression.
Shows fast p-hacking always worsens publication-bias distortion in reported estimates while slow p-hacking can mitigate it under strong selection, and proposes a correction for the true effect distribution.
Develops a missing-data identification framework with Neyman-orthogonal moments to purge life-cycle bias from intergenerational income mobility estimates built on single-age income snapshots.
Adapts bias-aware minimax methods from factor-model panels to panel causal inference, correcting imputed counterfactuals for treatment effect estimates with valid uncertainty intervals.
Proposes a stochastic-gradient likelihood estimator (SEGA) for limited dependent variable models that avoids the high-dimensional latent-variable integration intractable likelihoods usually require.
Shows stationarity in short panels forces a common quantile-regression slope rather than genuinely quantile-varying effects, and develops a fixed-T estimator consistent with that restriction.
Builds a GMM framework for peer effects under conditional random assignment to urns that separately identifies endogenous and contextual effects and corrects finite-urn variance bias.
Derives a sharp second-order minimax risk expansion for treatment-effect estimation in randomized experiments, showing Bernoulli randomization with a nonlinear estimator attains the bound.
Characterizes what panel treatment effects can still be learned when comparison units may themselves be exposed to spillovers of unknown form, without a prespecified exposure mapping.
Introduces a winsorized-mean drop-in robustification of 2SLS (W-2SLS) that attains minimax-sharp rates under adversarial contamination of instruments or outcomes.
Develops asymptotically optimal treatment-assignment rules for settings where sampling uncertainty is compounded by partial identification of the target parameter.
Proposes conformal prediction intervals for individual (not just conditional-mean) causal effects, localizing calibration via causal-forest-weighted similarity to a query unit.
Extends synthetic control to compositional (vector-of-shares) outcomes by mapping to log-odds and recovering the Fréchet barycenter under the Aitchison simplex geometry.
Shows standard long-difference and panel fixed-effects estimators used to measure climate adaptation are biased and can disagree in sign on the direction of adaptation, undermining a widely used empirical strategy.
Formalizes RAG-based policy learning as nearest-neighbor matching under the potential-outcomes framework and decomposes regret into retrieval and estimation components.
1 item
Surveys 352 experts and finds no consensus, more than a decade on, on whether Albouy's (2012) critique overturns Acemoglu-Johnson-Robinson's (2001) settler-mortality instrument for institutions.
No new items today.
Fetched and recorded, but not written up. Filter by source or search titles and authors.