Appendix — Revision and Evidence Map

Reading and Software Map

Selective foundational and current literature with dated official implementation guidance

Reading and Software Map

How this map was assembled

This is a selective teaching map, not a systematic review. Links and software documentation were checked on 1 August 2026. Foundational readings establish estimands; recent readings address known failure modes. Live documentation can change, so record the version used in an empirical project.

Identification and inference

PurposeReadingTeaching use
potential outcomesRubin (1974)assignment mechanisms and causal effects
IV and LATEAngrist, Imbens and Rubin (1996)complier interpretation
weak instrumentsAndrews, Stock and Sun (2019)robust diagnostics and inference
clusteringAbadie et al. (2023)when and why to cluster
OVB sensitivityCinelli and Hazlett (2020)observed-variable benchmarking
IV sensitivityCinelli and Hazlett (2025)exclusion and instrument-confounding violations
average-case sensitivityZhang and Zhao (2026)average-strength rather than only worst-case confounding bounds

Policy evaluation designs

DesignReadingTeaching use
applied DIDCard and Krueger (1994)design reconstruction and historical debate
group-time DIDCallaway and Sant’Anna (2021)heterogeneous staggered effects
imputation event studyBorusyak, Jaravel and Spiess (2024)supported untreated-outcome imputation
DID sensitivityRambachan and Roth (2023)violations of exact parallel trends
practical RDDCattaneo, Idrobo and Titiunik (2020)local estimation and diagnostics
matchingStuart (2010)design before outcome analysis
overlap weightsLi, Morgan and Zaslavsky (2018)target population under overlap
synthetic controlAbadie, Diamond and Hainmueller (2010)transparent comparative case study
synthetic DIDArkhangelsky et al. (2021)unit and time weighting

Flexible causal analysis and transport

PurposeReadingTeaching use
double MLChernozhukov et al. (2018)orthogonal scores and cross-fitting
generalized random forestsAthey, Tibshirani and Wager (2019)local moments and heterogeneity
R-learnerNie and Wager (2021)residual-on-residual CATE learning
efficient policy learningAthey and Wager (2021)constrained policy value
broad 2024 reviewFeuerriegel et al. (2024)causal ML workflow and limitations
external validity frameworkEgami and Hartman (2023)population, treatment, outcome and context
transport reviewDegtiar and Rose (2023)assumptions and estimators

Live software guidance

TaskOfficial documentationAudit note
group-time DIDdiddeclare control group, anticipation, aggregation and simultaneous bands
high-dimensional fixed effects/IVfixestinspect formula, fixed effects, clustering and current version
RDDrdrobustretain bandwidth, kernel and robust bias-correction output
synthetic DIDsynthdiddocumentation described beta status at the review date; verify design support
causal/generalized forestsgrfassess overlap, calibration and honest policy evaluation
double MLDoubleMLrecord learners, tuning, folds, score and repeated-split stability

Software names do not define estimands. Reconstruct the identification memo before translating any example code.

Reproducibility standards

Reading protocol

For any paper, extract only six lines before reading its regression table:

  1. target estimand;
  2. assignment/counterfactual source;
  3. identifying assumptions;
  4. support and inference level;
  5. strongest design-specific diagnostic;
  6. population, treatment and time boundary.

Return to the Course Home

Copyright © 2026