pith. sign in

arxiv: 2605.15119 · v1 · pith:WTHC6P6Mnew · submitted 2026-05-14 · 💰 econ.EM

Identification and Estimation of Staggered Difference-in-Differences with Network Spillovers

Hayato Tagawa This is my paper

Pith reviewed 2026-05-15 02:48 UTC · model grok-4.3

classification 💰 econ.EM
keywords staggered difference-in-differencesnetwork spilloverscausal inferencepolicy evaluationparallel trendsspatial dependence
0
0 comments X

The pith

A staggered difference-in-differences method separates a unit's own adoption effect from spillovers generated by others using a prespecified exposure summary.

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

This paper builds a difference-in-differences framework for staggered policy rollouts in which units can influence one another. For each treated group and time since adoption, it isolates the direct effect of a unit's own treatment, the spillover effect from other units' treatments, and the combined total effect that occurs under the actual pattern of adoption. Identification rests on parallel-trends comparisons among units that share the same summary measure of spillover exposure at the baseline and target dates. Spillover parameters are first recovered from never-treated units and then evaluated for treated units under the exposure levels those units actually experience. The resulting estimators allow for spatial dependence in errors, and the approach is illustrated both in simulations and in an application to the Community Health Centers program where spillovers explain a sizable share of the estimated impact on older-adult mortality.

Core claim

For each treated cohort and event time the framework identifies three distinct quantities: the effect of a unit's own adoption, the spillover effect produced by other adopters, and the total effect realized under the observed rollout. These quantities are recovered by comparing outcome changes among units that have identical values of a prespecified spillover-exposure summary at the relevant baseline and target dates, with the spillover component learned exclusively from never-treated units and then applied to the exposure distribution faced by treated cohorts.

What carries the argument

A prespecified summary measure of spillover exposure that groups units for parallel-trends comparisons at baseline and target dates.

If this is right

  • Standard staggered DID estimators that ignore spillovers can miss or misattribute the total policy effect.
  • The proposed estimators recover the three components with small bias and produce confidence intervals whose coverage is close to the nominal level under spatial dependence.
  • In the Community Health Centers rollout, the spillover component accounts for a substantial share of the effect on older-adult mortality.
  • Monte Carlo results confirm that ignoring spillovers leads to incorrect inference about the total effect while the new estimators maintain good finite-sample performance.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The method could be applied to other settings where a low-dimensional exposure summary can be pre-specified without full knowledge of the underlying network.
  • Policymakers could use the separated estimates to choose rollout sequences that either amplify or dampen desirable spillovers.
  • Robustness checks that vary the functional form of the exposure summary would be a natural next step when the summary itself is not uniquely determined by theory.

Load-bearing premise

Spillover effects recovered from never-treated units remain valid when applied to treated cohorts under the exposure distribution those cohorts actually face, together with conditional parallel trends holding for all units that share the same exposure summary.

What would settle it

If the spillover effect estimated from never-treated units differs systematically from the spillover effect implied for treated units once both groups are restricted to identical exposure-summary values at the same dates, the separation of own, spillover, and total effects fails.

Figures

Figures reproduced from arXiv: 2605.15119 by Hayato Tagawa.

Figure 1
Figure 1. Figure 1: Spatial diffusion of spillover exposure in Community Health Centers data [PITH_FULL_IMAGE:figures/full_fig_p034_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Event-time decomposition of Community Health Center mortality effects. [PITH_FULL_IMAGE:figures/full_fig_p037_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Control-group spillover effect. Notes: Gray points are cohort-event fitted contrasts in the never-treated source sample when avail￾able. Intervals are 95 percent pointwise intervals based on spatial HAC covariance estimates. 37 [PITH_FULL_IMAGE:figures/full_fig_p037_3.png] view at source ↗
read the original abstract

This paper develops a difference-in-differences framework for staggered policy adoption when units can be affected by other units' adoption. For each treated cohort and event time, the framework separates the effect of own adoption, the spillover effect generated by other adopters, and the total effect under the realized rollout. Identification uses a prespecified summary of spillover exposure and parallel trends comparisons among units with the same exposure at the baseline and target dates. Spillover effects are learned from never-treated units and evaluated for treated cohorts under the exposure distribution they face. We construct estimators for these effects and an inference procedure that allows for spatial dependence. Monte Carlo simulations illustrate that standard DID estimators that ignore spillovers can miss the total effect, whereas the proposed estimators have small bias for these effects and the associated confidence intervals have coverage close to the nominal level. In an empirical study of the Community Health Centers rollout, estimated spillovers account for a substantial share of the effect on older-adult mortality.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 2 minor

Summary. The paper develops a staggered DiD framework for settings with network spillovers. For each treated cohort and event time, it separates the own-adoption effect, the spillover effect from other adopters, and the total effect under the realized rollout. Identification relies on a prespecified summary of spillover exposure together with parallel-trends comparisons among units that share the same exposure at baseline and target dates; spillovers are learned from never-treated units and then evaluated for treated cohorts under their realized exposure distribution. Estimators and a spatial-dependence-robust inference procedure are proposed, with Monte Carlo evidence of small bias and near-nominal coverage, and an empirical illustration using the Community Health Centers rollout.

Significance. If the conditional parallel-trends assumption given the exposure summary holds, the framework supplies a practical way to recover own, spillover, and total effects in applications where units are interdependent. The simulation results and the empirical finding that spillovers account for a substantial share of the mortality effect demonstrate the practical relevance of separating these components rather than relying on standard DiD estimators that ignore spillovers.

major comments (1)
  1. [§3] §3 (Identification): The strategy learns spillover effects from never-treated units and applies them to treated cohorts under the realized exposure distribution. This requires that the prespecified exposure summary is sufficient to restore conditional parallel trends and that no selection into treatment timing occurs on potential responses to spillovers. If higher-order network connections or time-varying exposure intensity are omitted from the summary, the decomposition into own-adoption, spillover, and total effects will generally be biased.
minor comments (2)
  1. [§5] §5 (Simulations): The Monte Carlo design reports small bias and coverage close to nominal levels, but the paper should report the precise network-generating processes and the range of exposure-summary specifications examined so that readers can assess sensitivity to misspecification of the summary.
  2. [§6] §6 (Empirical application): Clarify how the exposure summary was constructed for the Community Health Centers data and whether results are robust to alternative summaries (e.g., including higher-order neighbors).

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the careful reading and positive recommendation. The comment correctly identifies the central identifying assumption of the framework. We address it directly below and have revised the manuscript to expand the discussion of this assumption and its practical implications.

read point-by-point responses

  1. Referee: [§3] §3 (Identification): The strategy learns spillover effects from never-treated units and applies them to treated cohorts under the realized exposure distribution. This requires that the prespecified exposure summary is sufficient to restore conditional parallel trends and that no selection into treatment timing occurs on potential responses to spillovers. If higher-order network connections or time-varying exposure intensity are omitted from the summary, the decomposition into own-adoption, spillover, and total effects will generally be biased.

    Authors: We agree that identification requires the prespecified exposure summary to be sufficient for conditional parallel trends (Assumption 3.1) and that treatment timing is independent of potential outcomes conditional on baseline exposure and covariates. The manuscript states this requirement explicitly in Section 3 and emphasizes that the researcher must choose the summary on the basis of economic theory and the network structure. If higher-order connections or time-varying intensity are omitted, the decomposition can indeed be biased; this is an inherent feature of any reduced-form exposure-summary approach rather than a flaw unique to our framework. We view the transparency of the assumption as a strength, because it makes the required conditions clear and allows researchers to incorporate richer summaries when data permit. In the revised manuscript we have added a new paragraph in Section 3.2 that (i) reiterates the sufficiency requirement, (ii) provides guidance on constructing summaries that capture higher-order or intensity effects, and (iii) suggests simple robustness checks that vary the exposure measure in both the Monte Carlo design and the empirical application. revision: yes

Circularity Check

0 steps flagged

No significant circularity; identification relies on external assumptions

full rationale

The paper develops a DiD framework separating own-adoption, spillover, and total effects using a prespecified spillover exposure summary and conditional parallel trends comparisons among units with matching exposure at baseline and target dates. Spillovers are learned from never-treated units and applied to treated cohorts. No equations or steps reduce by construction to fitted parameters, self-definitions, or self-citation chains. The central claims rest on stated identifying assumptions (conditional parallel trends given the summary) that are external and falsifiable, not derived from the target quantities themselves. This is a standard non-circular identification argument.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on conditional parallel trends assumptions and the validity of learning spillover effects from never-treated units for application to treated cohorts.

axioms (2)
  • domain assumption Parallel trends hold among units with the same prespecified spillover exposure summary at baseline and target dates.
    Invoked for identification of own and spillover effects via comparisons within exposure groups.
  • domain assumption Spillover effects estimated from never-treated units apply to treated cohorts under their realized exposure distribution.
    Allows separation and evaluation of spillover component for treated units.

pith-pipeline@v0.9.0 · 5459 in / 1451 out tokens · 30930 ms · 2026-05-15T02:48:00.285312+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

58 extracted references · 58 canonical work pages

  1. [1]

    arXiv preprint arXiv:2105.03737 , year =

    Difference-in-differences with spatial spillovers , author =. arXiv preprint arXiv:2105.03737 , year =

    work page arXiv

  2. [2]

    Railroads and American Economic Growth: A “Market Access” Approach *

    Railroads and American economic growth: A “market access” approach , author =. The Quarterly Journal of Economics , volume =. 2016 , publisher =

    work page 2016

  3. [3]

    The review of economic studies , volume =

    Identification of endogenous social effects: The reflection problem , author =. The review of economic studies , volume =. 1993 , publisher =

    work page 1993

  4. [4]

    Social Networks and the Identification of Peer Effects

    Social networks and the identification of peer effects , author =. Journal of Business & Economic Statistics , volume =. 2013 , publisher =

    work page 2013

  5. [5]

    Econometrica , volume =

    Worms: identifying impacts on education and health in the presence of treatment externalities , author =. Econometrica , volume =. 2004 , publisher =

    work page 2004

  6. [6]

    Identification and estimation of spillover effects in randomized experiments

    Gonzalo Vazquez-Bare , keywords =. Identification and estimation of spillover effects in randomized experiments , journal =. 2023 , issn =. doi:https://doi.org/10.1016/j.jeconom.2021.10.014 , url =

    work page doi:10.1016/j.jeconom.2021.10.014 2023

  7. [7]

    Programme evaluation and spillover effects

    Programme evaluation and spillover effects , author =. Journal of Development Effectiveness , volume =. 2016 , publisher =

    work page 2016

  8. [8]

    Annals of statistics , volume =

    Average treatment effects in the presence of unknown interference , author =. Annals of statistics , volume =

    work page

  9. [9]

    arXiv preprint arXiv:2104.03802 , year =

    Average treatment effects in the presence of interference , author =. arXiv preprint arXiv:2104.03802 , year =

    work page arXiv

  10. [10]

    Estimating the total treatment effect in randomized experiments with unknown network structure

    Estimating the total treatment effect in randomized experiments with unknown network structure , author =. Proceedings of the National Academy of Sciences , volume =. 2022 , publisher =

    work page 2022

  11. [11]

    Design-based inference for spatial experiments under unknown interference

    Design-based inference for spatial experiments under unknown interference , author =. The Annals of Applied Statistics , volume =. 2025 , publisher =

    work page 2025

  12. [12]

    The Review of Economic Studies , author =

    Revisiting Event-Study Designs: Robust and Efficient Estimation , author =. The Review of Economic Studies , volume =. 2024 , doi =

    work page 2024

  13. [13]

    Spillover effects in empirical corporate finance

    Spillover Effects in Empirical Corporate Finance , author =. Journal of Financial Economics , volume =. 2021 , doi =

    work page 2021

  14. [14]

    arXiv preprint arXiv:2306.12003 , year =

    Difference-in-Differences with Interference , author =. arXiv preprint arXiv:2306.12003 , year =

    work page arXiv

  15. [15]

    JUE Insight: Difference-in-differences with geocoded microdata

    JUE insight: Difference-in-differences with geocoded microdata , author =. Journal of Urban Economics , volume =. 2023 , publisher =

    work page 2023

  16. [16]

    Handbook of regional and urban economics , volume =

    Causal inference in urban and regional economics , author =. Handbook of regional and urban economics , volume =. 2015 , publisher =

    work page 2015

  17. [17]

    JUE Insight: The (non-)effect of opportunity zones on housing prices

    JUE Insight: The (non-) effect of opportunity zones on housing prices , author =. Journal of Urban Economics , volume =. 2023 , publisher =

    work page 2023

  18. [18]

    Local Economic Development, Agglomeration Economies, and the Big Push: 100 Years of Evidence from the Tennessee Valley Authority

    Local economic development, agglomeration economies, and the big push: 100 years of evidence from the Tennessee Valley Authority , author =. The Quarterly journal of economics , volume =. 2014 , publisher =

    work page 2014

  19. [19]

    Journal of econometrics , volume =

    Difference-in-differences with variation in treatment timing , author =. Journal of econometrics , volume =. 2021 , publisher =

    work page 2021

  20. [20]

    Place-Based Policies, Creation, and Agglomeration Economies: Evidence from China’s Economic Zone Program

    Lu, Yi and Wang, Jin and Zhu, Lianming , title =. American Economic Journal: Economic Policy , volume =. 2019 , month =. doi:10.1257/pol.20160272 , url =

    work page doi:10.1257/pol.20160272 2019

  21. [21]

    Difference-in-differences techniques for spatial data: Local autocorrelation and spatial interaction

    Difference-in-differences techniques for spatial data: Local autocorrelation and spatial interaction , author =. Economics Letters , volume =. 2015 , doi =

    work page 2015

  22. [22]

    A Simple Approach to Staggered Difference-in-Differences in the Presence of Spillovers

    Estimating Difference-in-Differences in the Presence of Spillovers , author =. 2017 , type =

    work page 2017

  23. [23]

    A Framework for Separating Individual-Level Treatment Effects From Spillover Effects

    A framework for separating individual-level treatment effects from spillover effects , author =. Journal of Business & Economic Statistics , volume =. 2021 , publisher =

    work page 2021

  24. [24]

    Journal of econometrics , volume =

    Doubly robust difference-in-differences estimators , author =. Journal of econometrics , volume =. 2020 , publisher =

    work page 2020

  25. [25]

    Aronow and Cyrus Samii , title =

    Peter M. Aronow and Cyrus Samii , title =. The Annals of Applied Statistics , number =. 2017 , doi =

    work page 2017

  26. [26]

    Biometrika , volume =

    Causal inference with misspecified exposure mappings: separating definitions and assumptions , author =. Biometrika , volume =. 2024 , publisher =

    work page 2024

  27. [27]

    1998 , publisher =

    General equilibrium treatment effects: A study of tuition policy , author =. 1998 , publisher =

    work page 1998

  28. [28]

    Journal of the American statistical Association , volume =

    A generalization of sampling without replacement from a finite universe , author =. Journal of the American statistical Association , volume =. 1952 , publisher =

    work page 1952

  29. [29]

    Identification of treatment response with social interactions

    Identification of treatment response with social interactions , author =. The Econometrics Journal , volume =. 2013 , publisher =

    work page 2013

  30. [30]

    Econometrica , volume =

    Causal inference under approximate neighborhood interference , author =. Econometrica , volume =. 2022 , publisher =

    work page 2022

  31. [31]

    Spillover Effects in the Presence of Unobserved Networks

    Spillover effects in the presence of unobserved networks , author =. Political Analysis , volume =. 2021 , publisher =

    work page 2021

  32. [32]

    Management Science , volume =

    Experimenting in equilibrium , author =. Management Science , volume =. 2021 , publisher =

    work page 2021

  33. [33]

    Social dynamics , volume =

    Policy interventions, low-level equilibria, and social interactions , author =. Social dynamics , volume =

    work page

  34. [34]

    The quarterly journal of economics , volume =

    Do labor market policies have displacement effects? Evidence from a clustered randomized experiment , author =. The quarterly journal of economics , volume =. 2013 , publisher =

    work page 2013

  35. [35]

    Mediation and Spillover Effects in Group-Randomized Trials: A Case Study of the 4Rs Educational Intervention

    Mediation and spillover effects in group-randomized trials: a case study of the 4Rs educational intervention , author =. Journal of the American Statistical Association , volume =. 2013 , publisher =

    work page 2013

  36. [36]

    Market Externalities of Large Unemployment Insurance Extension Programs

    Market externalities of large unemployment insurance extension programs , author =. The American Economic Review , pages =. 2015 , publisher =

    work page 2015

  37. [37]

    2024 , institution =

    A Simple Approach to Staggered Difference-in-Differences in the Presence of Spillovers , author =. 2024 , institution =

    work page 2024

  38. [38]

    What’s trending in difference-in-differences? A synthesis of the recent econometrics literature

    Jonathan Roth and Pedro H.C. Sant'Anna and Alyssa Bilinski and John Poe , keywords =. What's trending in difference-in-differences? A synthesis of the recent econometrics literature , journal =. 2023 , issn =. doi:https://doi.org/10.1016/j.jeconom.2023.03.008 , url =

    work page doi:10.1016/j.jeconom.2023.03.008 2023

  39. [39]

    Journal of econometrics , volume =

    Difference-in-differences with multiple time periods , author =. Journal of econometrics , volume =. 2021 , publisher =

    work page 2021

  40. [40]

    Journal of econometrics , volume =

    Estimating dynamic treatment effects in event studies with heterogeneous treatment effects , author =. Journal of econometrics , volume =. 2021 , publisher =

    work page 2021

  41. [41]

    American economic review , volume =

    Two-way fixed effects estimators with heterogeneous treatment effects , author =. American economic review , volume =. 2020 , publisher =

    work page 2020

  42. [42]

    Journal of the Association of Environmental and Resource Economists , volume =

    The role of parallel trends in event study settings: An application to environmental economics , author =. Journal of the Association of Environmental and Resource Economists , volume =. 2021 , publisher =

    work page 2021

  43. [43]

    The Effect of Minimum Wages on Low-Wage Jobs*

    The effect of minimum wages on low-wage jobs , author =. The Quarterly Journal of Economics , volume =. 2019 , publisher =

    work page 2019

  44. [44]

    Mimeo , year =

    Two-Stage Differences in Differences , author =. Mimeo , year =

    work page

  45. [45]

    SSRN Electronic Journal , author =

    Two-Way Fixed Effects, the Two-Way Mundlak Regression, and Difference-in-Differences Estimators , author =. SSRN Electronic Journal , year =. doi:10.2139/ssrn.3906345 , url =

    work page doi:10.2139/ssrn.3906345

  46. [46]

    Political Analysis , volume =

    On the use of two-way fixed effects regression models for causal inference with panel data , author =. Political Analysis , volume =. 2021 , publisher =

    work page 2021

  47. [47]

    Review of Economics and Statistics , pages =

    Difference-in-differences estimators of intertemporal treatment effects , author =. Review of Economics and Statistics , pages =. 2024 , publisher =

    work page 2024

  48. [48]

    Fuzzy Differences-in-Differences

    Fuzzy differences-in-differences , author =. The Review of Economic Studies , volume =. 2018 , publisher =

    work page 2018

  49. [49]

    Two-way fixed effects and differences-in-differences estimators with several treatments , journal =

    Clément. Two-way fixed effects and differences-in-differences estimators with several treatments , journal =. 2023 , issn =. doi:https://doi.org/10.1016/j.jeconom.2023.105480 , url =

    work page doi:10.1016/j.jeconom.2023.105480 2023

  50. [50]

    2024 , institution =

    Difference-in-differences with a continuous treatment , author =. 2024 , institution =

    work page 2024

  51. [51]

    Potential Outcome Modeling and Estimation in Did Designs with Staggered Treatments

    Staggered Adoption DiD Designs with Misclassification and Anticipation , author =. 2025 , eprint =

    work page 2025

  52. [52]

    Journal of Development Economics , volume =

    Using technology to prevent fraud in high stakes national school examinations: Evidence from Indonesia , author =. Journal of Development Economics , volume =

    work page

  53. [53]

    The War on Poverty's Experiment in Public Medicine: Community Health Centers and the Mortality of Older Americans

    The War on Poverty's experiment in public medicine: Community health centers and the mortality of older Americans , author =. American Economic Review , volume =. 2015 , publisher =

    work page 2015

  54. [54]

    APPLICATIONS OF FUNCTIONAL DEPENDENCE TO SPATIAL ECONOMETRICS

    Applications of Functional Dependence to Spatial Econometrics , author =. Econometric Theory , pages =. 2024 , publisher =

    work page 2024

  55. [55]

    Difference-in-Differences Estimates under Selective Migration

    Difference-in-Differences Under Network Interference , author =. 2025 , eprint =

    work page 2025

  56. [56]

    Limit theorems for network dependent random variables

    Denis Kojevnikov and Vadim Marmer and Kyungchul Song , keywords =. Limit theorems for network dependent random variables , journal =. 2021 , issn =. doi:https://doi.org/10.1016/j.jeconom.2020.05.019 , url =

    work page doi:10.1016/j.jeconom.2020.05.019 2021

  57. [57]

    Central limit theorems and uniform laws of large numbers for arrays of random fields

    Nazgul Jenish and Ingmar R. Prucha , keywords =. Central limit theorems and uniform laws of large numbers for arrays of random fields , journal =. 2009 , issn =. doi:https://doi.org/10.1016/j.jeconom.2009.02.009 , url =

    work page doi:10.1016/j.jeconom.2009.02.009 2009

  58. [58]

    On spatial processes and asymptotic inference under near-epoch dependence

    Nazgul Jenish and Ingmar R. Prucha , keywords =. On spatial processes and asymptotic inference under near-epoch dependence , journal =. 2012 , issn =. doi:https://doi.org/10.1016/j.jeconom.2012.05.022 , url =

    work page doi:10.1016/j.jeconom.2012.05.022 2012