Pith. sign in

REVIEW 3 major objections 3 minor 28 references

Good Controls Gone Bad: Difference-in-Differences with Covariates

T0 review · 3 major / 3 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The paper claims that standard DiD estimators require covariate effects to be constant across groups and time, and proposes DID-INT to identify the ATT when that fails.

desk verdict The paper's central claim about heterogeneous treatment effects isn't proven—its own proof uses homogeneous effects—but the CCC taxonomy and simulation evidence make it worth a careful look. read the letter →

arxiv 2412.14447 v2 pith:ZOJITA4B submitted 2024-12-19 econ.EM

Sunny Karim , Matthew D. Webb This is my paper · ORCID
classification econ.EM MSC 62P20
keywords difference-in-differencescommoncausalcovariatestime-varyingaveragetreatmenteffectonthetreatedstaggeredadoptionparalleltrendstwo-wayfixedeffectscovariateinteractions
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper argues that standard difference-in-differences estimators silently assume a 'two-way common causal covariates' (CCC) condition: each covariate's effect on the outcome must be the same across treated and control groups and across time periods. When covariate effects differ by group or period, TWFE, Callaway-Sant'Anna, imputation, and FLEX estimators carry a bias term that does not cancel, which the paper demonstrates by proofs and Monte Carlo simulations. It introduces the Intersection Difference-in-differences (DID-INT) estimator, which regresses the outcome on group-by-time intersection dummies and on covariates interacted with group, time, or group-by-time dummies, and then forms difference-in-differences of the estimated intersection coefficients. The paper claims DID-INT identifies the average treatment effect on the treated under the standard DiD assumptions without the CCC assumption, and that it extends to staggered treatment rollout.

What carries the argument

The central object is the group-by-time intersection dummy $I(s,t)$ and the associated regression $Y_{i,s,t} = \sum_s \sum_t \lambda_{s,t} I(s,t) + f(X_{i,s,t}) + \epsilon$, run without a constant. The four versions of DID-INT differ in $f(X)$: homogeneous, state-varying (covariates interacted with $I(s)$ only), time-varying (interacted with $I(t)$ only), and two-way (interacted with both). The ATT for a treated group and period is the difference between the treated and control cells of the long difference $\lambda_{s,t} - \lambda_{s,t^{-}_{s}}$, and the overall ATT is a weighted average that omits forbidden comparisons. This construction is what makes the covariate effects cancel even when $\gamma_{s,t}$ varies across cells.

What would settle it

Run the paper's Monte Carlo design with two-way CCC violations and treatment effects that depend on a covariate, such as $\tau_i = X_{1i}$, while keeping Assumptions 1-4; if the two-way DID-INT estimates do not center on the true ATT across many replications, the identification claim as stated fails.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that a covariate that is a 'good control' in one pair of groups and periods can become a source of bias in another: the effect of the covariate on the untreated outcome is written $\gamma^k_{s,t}$, and unless $\gamma$ is constant across states and years the conventional estimators misestimate the ATT. DID-INT removes this problem by allowing the covariate effect to be fully cell-specific: it runs $Y_{i,s,t} = \sum_s \sum_t \lambda_{s,t} I(s,t) + f(X_{i,s,t}) + \epsilon$, with $f(X)$ chosen as covariate-by-group, covariate-by-time, or covariate-by-group-by-time interactions, and estimates the ATT for a treated cell as the long difference of the $\lambda$'s in the treated cell minus the same long difference for a not-yet-treated control cell. Because the covariate terms are absorbed at the group-by-time level, the bias terms from CCC violations cancel before the DiD comparison is made. The paper states this identification result as Theorem 2, requiring only binary treatment, overlap, conditional parallel trends, and no anticipation, and the abstract extends the claim to heterogeneous treatment effects and staggered rollout.

Load-bearing premise

The proof of Theorem 2 assumes homogeneous treatment effects when simplifying the bias terms, and the practical guarantee assumes the model-selection algorithm finds the correct covariate functional form; if either fails, the bias terms need not cancel.

Editorial extensions

If this is right

  • Researchers can include time-varying covariates without assuming their effects are constant across groups and periods, provided they use the fully interacted two-way DID-INT.
  • TWFE, CS-DID, imputation, and FLEX are all biased when the two-way CCC is violated, and the bias grows with the degree of violation in the paper's simulations.
  • The two-way DID-INT is unbiased across all simulated DGPs but is the least efficient estimator, so there is a bias-variance tradeoff.
  • A modified TWFE with fully interacted covariates also removes the CCC bias under common treatment timing, but it remains vulnerable to staggered-adoption negative weights, while DID-INT is not.
  • The model-selection algorithm can turn non-parallel pre-trend figures into parallel ones by residualizing outcomes with group-by-time interacted covariates.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The appendix proof of Theorem 2 appears to import Assumption 13 (homogeneous treatment effects) when it replaces $E[\tau_{s,t} | s \in S_T, t, f(X)]$ with $\tau$, so the advertised robustness to heterogeneous treatment effects is not demonstrated by that proof.
  • The model-selection algorithm chooses the covariate functional form by visual inspection of pre-trend plots; identification therefore carries an informal, data-dependent step that would need a formal test to be fully automatic.
  • An immediate testable extension is to run the paper's Monte Carlo with treatment effects that vary with a covariate, which directly checks whether the Theorem 2 claim holds beyond homogeneous effects.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 3 minor

Summary. This paper introduces the 'common causal covariates' (CCC) assumption for difference-in-differences with covariates, distinguishing state-invariant, time-invariant, and two-way CCC violations. It argues that existing estimators (TWFE, CS-DID, imputation, FLEX) are biased when the two-way CCC is violated, and proposes a new estimator, DID-INT, which residualizes outcomes after interacting covariates with group and time dummies. The paper claims that DID-INT identifies the ATT under Assumptions (1)–(4) alone, without any CCC assumptions, and also under heterogeneous treatment effects and staggered adoption. The theoretical results are supplemented by Monte Carlo simulations and an empirical illustration using Canadian Labour Force Survey data.

Significance. The paper identifies a real and important gap in the DiD-with-covariates literature: the implicit assumption that covariate effects are constant across groups and time. The CCC taxonomy (state-invariant, time-invariant, two-way) is a useful conceptual contribution, and the Monte Carlo comparisons across estimators are informative. The paper also provides open-source software (Julia and Stata) for the proposed estimator, which is a practical strength. However, the central identification claim for DID-INT is not supported by the proof as written: the proof of Theorem 2 relies on a homogeneous treatment effect assumption that is neither stated in the theorem nor reflected in the abstract's heterogeneous-treatment-effects claim. Consequently, the paper's headline contribution—that DID-INT identifies the ATT under the standard DiD assumptions without extra covariate restrictions—is not established.

major comments (3)
  1. [Section 10 / Theorem 2 / Appendix B] The proof of Theorem 2 relies on Assumption 13 (homogeneous treatment effects), but the theorem and the abstract claim identification under heterogeneous treatment effects. Appendix B's Eq. (9) assumes E[Y_{i,s,t}|s in ST, t, f(X)] = lambda_{s,t} + sum gamma^k_{s,t} E[X^k|...] for the treated post-period cell. This holds only if the OLS coefficient on X in that cell equals the untreated-potential-outcome coefficient gamma_{s,t}. When treatment effects vary with X, the slope instead converges to gamma_{s,t} + Cov(X, tau(X))/Var(X), so the cancellation in Eq. (17) fails and theta_{s,t} is generally not tau. Therefore the central claim that DID-INT identifies the ATT without any CCC assumptions is unproven.
  2. [Appendix E, Eq. (43)] Appendix E derives the DR-DID estimand under time-varying covariates and two-way CCC violations, obtaining an explicit nonzero bias term in Eq. (42), but then concludes with Eq. (43): theta_DRDID = E[hat theta_DRDID] = tau. This contradicts Theorem 9, which states theta_DRDID != tau in that setting. The displayed equation should be theta_DRDID != tau. As printed, the appendix's conclusion is internally inconsistent with the stated theorem and with the preceding derivation.
  3. [Section 9 / Assumption 3] The identification result in Theorem 2 depends on Assumption 3, which conditions on the 'correct functional form of covariates identified from the model selection algorithm defined in Section 9.' The model selection algorithm is based on visual inspection of parallel-trends plots and is not shown to recover the true f(X) under any formal criterion. If the algorithm selects an incorrect functional form, conditional parallel trends need not hold, and the bias terms in Eq. (17) that the proof cancels would not cancel. The paper provides no formal guarantee for this step, so the identification argument is incomplete.
minor comments (3)
  1. [General] There are numerous typographical errors, including 'SUTV A' for SUTVA, 'accross' for across, 'the the' instances, and inconsistent equation numbering between the main text and appendices. A careful proofreading pass is needed.
  2. [Section 13.1] The text says Panel (d) uses a DGP 'Y^4' where 'Assumption (2) is violated'; this appears to be a typo for 'the two-way CCC is violated' (or similar). Also, the statement that Assumption (2) is violated for Y^2 in the main Monte Carlo seems inconsistent with the earlier description that Assumption (2) holds for Y^1 and is violated for Y^2; the text should clarify which assumptions are actually violated in each DGP.
  3. [Section 7 / Appendix A] Theorem 1 states identification under Assumptions (1)-(4), (8), (9), and (10), but the proof in Appendix A uses Assumption 13 (homogeneous treatment effects) to write E[tau_{s,t}|...] = tau. The theorem statement should either include Assumption 13 or the proof should handle heterogeneous treatment effects explicitly.

Circularity Check

2 steps flagged · score 6.0 of 10

Theorem 2's proof cancels the covariate bias only after substituting Appendix A Eq. (6), which was derived under homogeneous-treatment-effects Assumption 13, so the abstract's heterogeneous-effects claim reduces by construction to a constant-τ input.

  1. other [Theorem 2 (Section 10); Appendix B, Equations (9)-(18); Appendix A, Equation (6); Assumption 13]
    "Imposing Assumptions (1) to (4), we can replace ... in Equation (15) with two-way version of the term shown in Equation (6) to get: τ + ... Canceling out the relevant terms in Equation (17), we can show that: θs,t = E[θ̂s,t] = τ. Equation (6) uses Assumption (13), which implies E[τ_{s,t}|s∈ST,t∈T^s,f(X̃)] = τ."

    The proof of Theorem 2 does not derive τ from Assumptions (1)-(4); it substitutes the 'two-way version' of Appendix A Eq. (6), and that expression is obtained only after imposing Assumption 13 (homogeneous treatment effects). Appendix B Eq. (9)-(10) also writes the treated-post cell conditional mean as λs,t + Σ γ_{s,t}E[X|...], omitting any E[τ|...] term, which is valid only if τ is constant within the cell or mean-independent of X. Substituting Eq. (6) makes the bias terms cancel by construction, so θ=τ is an algebraic restatement of the homogeneous-τ input, not identification of the ATT under heterogeneous treatment effects.

  2. fitted input called prediction [Section 9 (Model Selection Algorithm); Assumption 3, Section 5; Section 4 step 1]
    "The algorithm relies on plotting a sequence of parallel trends figures, conditional on the appropriate functional form of covariates, and stopping where pre-trends seem plausible through visual inspection. ... Assumption (3) in this setup holds conditional on the correct functional form of covariates identified from the model selection algorithm defined in Section 9."

    The 'correct' functional form f(X) is selected by inspecting residualized outcome trends in the same dataset later used for estimation, and the stopping rule is 'pre-trends seem plausible'. Hence the selected specification is guaranteed by the selection criterion to make the pre-trend figure look parallel, so the paper's claim that covariate residualization 'can recover parallel trends' is in part a property manufactured by the search rather than independent evidence. Subsequent inference also ignores this outcome-based specification search, so unbiasedness of the chosen DID-INT variant is conditional on a data-dependent event that the formal theorem does not account for.

full rationale

The paper's negative results about TWFE, CS-DID, imputation, and FLEX under CCC violations are derived under the explicitly stated homogeneous-treatment-effects Assumption 13 and are not the circular part. The central circular step is Theorem 2: the proof substitutes the 'two-way version' of Appendix A Eq. (6) into the DID-INT estimand, and Eq. (6) is derived only after invoking Assumption 13 (E[τ|s,t,f(X)] = τ). Thus the cancellation that yields θ=τ is produced by assuming the treatment effect is constant in covariates, not by Assumptions (1)-(4). For heterogeneous τ(X), the treated-post OLS slope on X equals γ_{s,t} + Cov(X,τ)/Var(X), so the residual λ_{s,t} carries an extra mean-zero term and the Eq. (17) cancellation fails; the abstract's heterogeneous-effects claim is therefore an assumption restated as a conclusion. A secondary practical circularity is the model selection algorithm: f(X) is chosen by inspecting residualized outcome trends from the estimation sample and stopping when pre-trends 'seem plausible', so the 'recovered parallel trends' are in part manufactured by the selection criterion, and inference ignores this selection. The self-citations (Karim et al. 2024, 2025) concern software, inference routines, and a related estimator, and are not load-bearing for the identification argument, so they do not raise the score further. The Monte Carlo exercises are internal consistency checks with DGPs matching the estimator's assumed linear structure; they do not constitute external validation, but that is a simulation-design limitation rather than a separate circularity. Overall, one central prediction reduces by construction to a hidden homogeneity input; the rest of the identification algebra is coherent once that input is admitted.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The central claims rest on linearity of the outcome model in covariates, the homogeneity of treatment effects used in the proof, and the informal correctness of the model-selection algorithm. The Monte Carlo design uses data-fitted coefficients and hand-chosen violation degrees.

free parameters (2)
  • Monte Carlo degree of CCC violation increments = 10, 50, 100, 250, 500
    In Appendix G, the severity of CCC violations is set by hand to these five values; the bias magnitudes in Table 2 depend directly on this choice.
  • Weights w_{s,t} in ATT aggregation = not specified
    Equation (5) aggregates θ_{s,t} with weights w_{s,t}, but the paper never defines how these weights are chosen across multiple treated and control groups; this leaves a free implementation choice.
assumptions (4)
  • standard math SUTVA (observed outcome equals treated or untreated potential outcome)
    Assumption 1 in Appendix A, used to substitute potential outcomes in the proof of the ATT estimand.
  • domain assumption Untreated outcome is linear in f(X) with cell-specific coefficients (Eq. 6)
    The bias cancellations in Appendices B and D require this linear model; the paper claims no assumptions on covariates but this linearity is needed.
  • domain assumption Homogeneous treatment effects (Assumption 13)
    The proof of Theorem 2 replaces E[τ|s,t,f(X)] with τ, but Theorem 2 and the abstract claim identification under heterogeneous effects.
  • ad hoc to paper Model selection algorithm recovers the true f(X)
    Section 9's visual pre-trends algorithm must select a functional form for which conditional parallel trends holds; no formal guarantee is given.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Good Controls Gone Bad: Difference-in-Differences with Covariates." pith.science (2026). https://pith.science/paper/ZOJITA4B

@misc{pith2026241214447,
  author       = {Pith},
  title        = {Pith review of: Good Controls Gone Bad: Difference-in-Differences with Covariates},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZOJITA4B}},
  note         = {Machine review of arXiv:2412.14447}
}
read the original abstract

This paper introduces the two-way common causal covariates (CCC) assumption, which is necessary to get an unbiased estimate of the ATT when using time-varying covariates in existing Difference-in-Differences methods. The two-way CCC assumption implies that the effect of the covariates remain the same between groups and across time periods. This assumption has been implied in previous literature, but has not been explicitly addressed. Through theoretical proofs and a Monte Carlo simulation study, we show that the standard TWFE and the CS-DID estimators are biased when the two-way CCC assumption is violated. We propose a new estimator called the Intersection Difference-in-differences (DID-INT) which can provide an unbiased estimate of the ATT under two-way CCC violations. DID-INT can also identify the ATT under heterogeneous treatment effects and with staggered treatment rollout. The estimator relies on parallel trends of the residuals of the outcome variable, after appropriately adjusting for covariates. This covariate residualization can recover parallel trends that are hidden with conventional estimators.

Figures

Figures reproduced from arXiv: 2412.14447 by the authors.

Figure 1
Figure 1. Unconditional and Corrected Parallel Trends in DGP with CCC Violations [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. State X Year coefficients of earnings on a dummy for university (BA or better) [PITH_FULL_IMAGE:figures/full_fig_p014_2.png] view at source ↗
Figure 3
Figure 3. Illustration of different types of covariates [PITH_FULL_IMAGE:figures/full_fig_p016_3.png] view at source ↗
Figures from the paper (12 more)
Figure 4
Figure 4. Figure 4: Modeling Covariates in DID-INT 17 [PITH_FULL_IMAGE:figures/full_fig_p017_4.png]
Figure 5
Figure 5. Figure 5: Model Selection Algorithm 19 [PITH_FULL_IMAGE:figures/full_fig_p019_5.png]
Figure 6
Figure 6. Figure 6: Example 1: DGP with time-invariant CCC violations [PITH_FULL_IMAGE:figures/full_fig_p020_6.png]
Figure 7
Figure 7. Figure 7: Example 2: DGP with two-way CCC violations [PITH_FULL_IMAGE:figures/full_fig_p021_7.png]
Figure 8
Figure 8. Figure 8: Parallel trends for weekly earnings regression: earningsi,t = α0 + β 0 t year + ϵi,t. (41) When the two-way CCC holds, the known-DGP outcome is generated as follows: Y 1 i,s,t = y0 + X k bγ kX k i,s,t. (42) where, bγ k ’s are the estimated coefficients from the regress…
Figure 9
Figure 9. Figure 9: Kernel Densities of the standard TWFE, modified TWFE and two-way DID-INT [PITH_FULL_IMAGE:figures/full_fig_p031_9.png]
Figure 10
Figure 10. Figure 10: Kernel Densities of state-varying, time varying and two-way DID-INT [PITH_FULL_IMAGE:figures/full_fig_p033_10.png]
Figure 11
Figure 11. Figure 11: Kernel Densities of the CS-DID and two-way DID-INT [PITH_FULL_IMAGE:figures/full_fig_p034_11.png]
Figure 12
Figure 12. Figure 12: Kernel Densities of the Imputation estimator and two-way DID-INT [PITH_FULL_IMAGE:figures/full_fig_p035_12.png]
Figure 13
Figure 13. Figure 13: Kernel Densities of FLEX and the two-way DID-INT [PITH_FULL_IMAGE:figures/full_fig_p035_13.png]
Figure 14
Figure 14. Figure 14: Example of Two One-Way CCC Violations For the DGP where there are two-way CCC violations (Y 2 i,s,t) and two one-way violations (Y 5 i,s,t), we estimate both the two-way DID-INT and the two one-way DID-INT a 1000 times, and compare their kernel densities. The results …
Figure 15
Figure 15. Figure 15: Kernel Densities of two one-way and two-way DID-INT [PITH_FULL_IMAGE:figures/full_fig_p050_15.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

28 extracted references · 26 canonical work pages

  1. [1]

    (2005) Semiparametric difference-in-differences estimators, The review of economic studies 72(1), 1--19

    Abadie, A. (2005) Semiparametric difference-in-differences estimators, The review of economic studies 72(1), 1--19

    work page 2005

  2. [2]

    Diamond, and J

    Abadie, A., A. Diamond, and J. Hainmueller (2010) Synthetic control methods for comparative case studies: Estimating the effect of california’s tobacco control program, Journal of the American statistical Association 105(490), 493--505

    work page 2010

  3. [3]

    Duflo, and S

    Bertrand, M., E. Duflo, and S. Mullainathan (2004) How much should we trust differences-in-differences estimates?, The Quarterly journal of economics 119(1), 249--275

    work page 2004

  4. [4]

    Jaravel (2017) Revisiting event study designs, Available at SSRN 2826228

    Borusyak, K., and X. Jaravel (2017) Revisiting event study designs, Available at SSRN 2826228

    work page 2017

  5. [5]

    Jaravel, and J

    Borusyak, K., X. Jaravel, and J. Spiess (2024) Revisiting event-study designs: robust and efficient estimation, Review of Economic Studies 91(6), 3253--3285

    work page 2024

  6. [6]

    Difference-in-Differences when Parallel Trends Holds Conditional on Covariates

    Caetano, C., and B. Callaway (2024) Difference-in-differences when parallel trends holds conditional on covariates, arXiv preprint arXiv:2406.15288

    work page Pith review arXiv 2024

  7. [7]

    Callaway, S

    Caetano, C., B. Callaway, S. Payne, and H. S. Rodrigues (2022) Difference in differences with time-varying covariates, arXiv preprint arXiv:2202.02903

    arXiv 2022

  8. [8]

    Callaway, B., and P. H. Sant’Anna (2021) Difference-in-differences with multiple time periods, Journal of Econometrics 225(2), 200--230

    work page 2021

Show all 28 references
  1. [9]

    Rothstein, and M

    Card, D., J. Rothstein, and M. Yi (2024) Industry wage differentials: A firm-based approach, Journal of Labor Economics 42(S1), S11--S59

    2024

  2. [10]

    d’Haultfoeuille (2020a) Two-way fixed effects estimators with heterogeneous treatment effects, American Economic Review 110(9), 2964--2996

    De Chaisemartin, C., and X. d’Haultfoeuille (2020a) Two-way fixed effects estimators with heterogeneous treatment effects, American Economic Review 110(9), 2964--2996

    2020

  3. [11]

    --- --- --- (2023) Two-way fixed effects and differences-in-differences with heterogeneous treatment effects: A survey, The Econometrics Journal 26(3), C1--C30

    2023

  4. [12]

    Deb, P., E. C. Norton, J. M. Wooldridge, and J. E. Zabel (2024) A flexible, heterogeneous treatment effects difference-in-differences estimator for repeated cross-sections, Technical report, National Bureau of Economic Research

    2024

  5. [13]

    Denning, J. T., E. R. Eide, K. J. Mumford, R. W. Patterson, and M. Warnick (2022) Why have college completion rates increased?, American Economic Journal: Applied Economics 14(3), 1--29

    2022

  6. [14]

    Fortin, N. M. (2006) Higher-education policies and the college wage premium: Cross-state evidence from the 1990s, American Economic Review 96(4), 959--987

    2006

  7. [15]

    (2021) Difference-in-differences with variation in treatment timing, Journal of Econometrics 225(2), 254--277

    Goodman-Bacon, A. (2021) Difference-in-differences with variation in treatment timing, Journal of Econometrics 225(2), 254--277

    2021

  8. [16]

    Heckman, J. J., H. Ichimura, J. A. Smith, and P. E. Todd (1998) Characterizing selection bias using experimental data,

    1998

  9. [17]

    Heckman, J. J., H. Ichimura, and P. E. Todd (1997) Matching as an econometric evaluation estimator: Evidence from evaluating a job training programme, The review of economic studies 64(4), 605--654

    1997

  10. [18]

    (2018) Relative education and the advantage of a college degree, American Sociological Review 83(4), 771--801

    Horowitz, J. (2018) Relative education and the advantage of a college degree, American Sociological Review 83(4), 771--801

    2018

  11. [19]

    Strumpf, N

    Karim, S., E. Strumpf, N. Austin, and M. D. Webb (2025) Which policy works and where? estimation and inference of state level treatment effects using difference-in-differences, Slides presented at the Canadian Economics Association Annual Conference, accessed July 11, 2025

    2025

  12. [20]

    Karim, S., M. D. Webb, N. Austin, and E. Strumpf (2024) Difference-in-differences with unpoolable data, arXiv preprint arXiv:2403.15910

    2024 arXiv

  13. [21]

    (1958) Investment in human capital and personal income distribution, Journal of political economy 66(4), 281--302

    Mincer, J. (1958) Investment in human capital and personal income distribution, Journal of political economy 66(4), 281--302

    1958

  14. [22]

    Kreif, R

    O’Neill, S., N. Kreif, R. Grieve, M. Sutton, and J. S. Sekhon (2016) Estimating causal effects: considering three alternatives to difference-in-differences estimation, Health Services and Outcomes Research Methodology 16, 1--21

    2016

  15. [23]

    Roth (2023) A more credible approach to parallel trends, Review of Economic Studies 90(5), 2555--2591

    Rambachan, A., and J. Roth (2023) A more credible approach to parallel trends, Review of Economic Studies 90(5), 2555--2591

    2023

  16. [24]

    Rios-Avila, F., P. H. Sant'Anna, and B. Callaway (2021) CSDID: Stata module for the estimation of Difference-in-Difference models with multiple time periods , Statistical Software Components, Boston College Department of Economics

    2021

  17. [25]

    (2025) Julia as a universal platform for statistical software development, The Stata Journal 25(2), 255--284

    Roodman, D. (2025) Julia as a universal platform for statistical software development, The Stata Journal 25(2), 255--284

    2025

  18. [26]

    Roth, J., P. H. Sant'Anna, A. Bilinski, and J. Poe (2022) What's trending in difference-in-differences? a synthesis of the recent econometrics literature, arXiv preprint arXiv:2201.01194

    2022 arXiv

  19. [27]

    H., and J

    Sant’Anna, P. H., and J. Zhao (2020) Doubly robust difference-in-differences estimators, Journal of Econometrics 219(1), 101--122

    2020

  20. [28]

    Abraham (2021) Estimating dynamic treatment effects in event studies with heterogeneous treatment effects, Journal of Econometrics 225(2), 175--199

    Sun, L., and S. Abraham (2021) Estimating dynamic treatment effects in event studies with heterogeneous treatment effects, Journal of Econometrics 225(2), 175--199

    2021

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.