Hedging out the HEDGE Bias: Identifying Dynamic Partial-Equilibrium Effects with Panel LP-IV
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Economics. The day in economics: theory, mechanisms and the evidence beneath them.
Ruth: We're looking at the paper: "Hedging out the HEDGE Bias".
Warren: Here's the gist of the paper. Cross-sectional designs are frequently employed to estimate dynamic partial-equilibrium effects arising from aggregate treatments, but these estimates are known to be biased (the HEDGE bias) when exposures to the treatment and aggregate confounders are correlated or when confounders exhibit feedback into the treatment.
Ruth: This paper introduces the heterogeneous-exposure local-projection instrumental variables (HELP-IV) framework specifically designed to address this bias.
Warren: HELP-IV compares units while holding the path of aggregate confounders fixed by instrumenting both the treatment and this confounder path with exposure-weighted shock proxies.
Ruth: The authors analytically prove that HELP-IV successfully identifies the true partial-equilibrium effect, extending its application to scenarios involving multiple exposure margins, multiple confounders, forward-looking outcomes, and cumulative multipliers.
Paper summary: Ruth: So we’ve talked about how cross-sectional designs can be biased when treatment and confounders are correlated, leading to what they call the HEDGE bias. The central thesis of this paper is that this bias ruins our ability to estimate dynamic partial-equilibrium effects.
Warren: They introduce the heterogeneous-exposure local-projection instrumental variables, or HELP-IV framework, which is designed specifically to solve this problem by holding the confounder path fixed using exposure-weighted shock proxies for both variables.
Iris: The authors analytically prove that this HELP-IV framework successfully identifies the true partial-equilibrium effect, extending its application to scenarios with multiple exposures and multiple confounders.
Ruth: They characterize the HEDGE bias under general aggregate dynamics and develop a regression framework to identify these dynamic partial-equilibrium effects across a broad range of environments.
Warren: The key mechanism they use is instrumenting the treatment and that confounder path with exposure-weighted shock proxies, which lets them remove both components of the bias—omitted history and the HEDGE bias itself.
Iris: They derive an explicit formula for this dynamic partial-equilibrium effect in backward-looking environments, and they characterize how traditional two-way fixed effects instrumental variables estimates contain both an exposure-correlation component and an aggregate feedback component.
Ruth: The paper shows that HELP-IV removes both components of the HEDGE bias simultaneously because its treatment instrument is unpredictable from pre-window information while its path instruments hold the confounder path fixed.
Warren: This means for someone running a simulation, if you use this method, you should be able to identify the true effect even when things are messy in terms of correlation and feedback.
Iris: The framework is formally established through propositions demonstrating that HELP-IV identifies the partial-equilibrium effect and that the resulting estimator converges to it under various regularity conditions.
Ruth: They also compared it against other methods, showing that while decomposition removes one component of the HEDGE bias, it fails when offsets change the treatment response used for normalization.
Conclusion: Ruth: So, looking at this work by Endong Wang and Ren Zhang, they’ve put forward a way to systematically deal with the HEDGE bias in dynamic partial-equilibrium effect estimation using the HELP-IV framework.
Warren: What does that mean for us who are using these statistics every day? It suggests that when we use cross-sectional data to look at how things change over time, we need a more careful way to account for how exposures and confounders interact.
Iris: The authors show that by instrumenting both the treatment and the path of the confounders with specific proxies, you can isolate the true effect from these confounding interactions.
Ruth: In simple terms, it’s about building a statistical tool that helps us find what happens to an outcome when we shock one thing while keeping another related thing stable.
Warren: It means that for policy and modeling, instead of relying on estimates from traditional methods that might be biased in complex settings, we can use this framework to get more accurate measurements.
Iris: The results show that applying this method yields partial-equilibrium effects that are smaller than conventional estimates, which gives us a different picture when we translate these findings into broader macroeconomic models.
Ruth: It’s about getting a cleaner signal from the data, even when the underlying economic system has feedback loops and complex correlations built in.
Endong Wang, Ren Zhang
University of Mannheim · Texas State University
econ.EM
Submitted: 2026-11-10
Updated: 2026-11-12
Project page: https://edmanuel7.github.io/MIBS_Mar26.pdf
License: http://creativecommons.org/licenses/by-nc-nd/4.0/
Importance score: 90/100
The gist: Cross-sectional designs are frequently employed to estimate dynamic partial-equilibrium effects arising from aggregate treatments…
Key concepts
- HEDGE Bias
- This bias occurs in cross-sectional designs when the treatment exposure and aggregate confounders are correlated, or when confounders influence the treatment itself. It causes estimates to be systematically wrong, leading to biased results for dynamic partial-equilibrium effects.
- HELP-IV Framework
- This is a new regression design that compares units by holding the path of aggregate confounders fixed. It uses exposure-weighted shock proxies to instrument both the treatment and the confounder paths, allowing it to remove both components of the HEDGE bias simultaneously.
- Partial-Equilibrium Effect ($φ^P h$)
- This is the true, underlying dynamic partial-equilibrium effect that researchers aim to estimate. The HELP-IV framework is designed to accurately identify this specific effect, even when dealing with multiple exposures, confounders, or forward-looking outcomes.
- Two-Way Fixed Effects Instrumental Variables (TWFE-IV)
- This is a traditional estimation method used for dynamic effects. The paper analyzes its limitations by showing it suffers from HEDGE bias because it fails to account for both exposure correlation and aggregate feedback in its estimates.
Terminology
Summary
Cross-sectional designs are frequently employed to estimate dynamic partial-equilibrium effects arising from aggregate treatments, but these estimates are known to be biased (the HEDGE bias) when exposures to the treatment and aggregate confounders are correlated or when confounders exhibit feedback into the treatment. p. 0 This version: November 9, 2026
Cross-sectional designs are widely used to estimate dynamic partial-equilibrium effects of an aggregate
treatment. We show analytically that these estimates are biased when exposures to the treatment and
to aggregate confounders are correlated or the confounders feed back into the treatment. A recent paper by Donaldson (2026) terms this bias the heterogeneous-exposure-to-dynamic-general-equilibrium (HEDGE) bias. The
central contribution of this paper is to characterize the HEDGE bias under general aggregate dynamics and
to develop a regression-based framework, heterogeneous-exposure local-projection instrumental variables
(HELP-IV), which identifies the dynamic partial-equilibrium effects across a broad range of environments. We show analytically that these estimates are biased when exposures to the treatment and
to aggregate confounders are correlated or the confounders feed back into the treatment. This paper
develops the heterogeneous-exposure local-projection instrumental variables (HELP-IV) framework,
which compares units while holding the confounder path fixed by instrumenting the treatment and this
path with exposure-weighted shock proxies. We prove that HELP-IV identifies the partial-equilibrium
effect. We further extend the framework, with the corresponding inference, to multiple exposure margins
and confounders, forward-looking outcomes, and cumulative multipliers.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
The central contribution of the paper is characterizing the HEDGE bias under general aggregate dynamics and developing a regression-based framework (HELP-IV) that identifies dynamic partial-equilibrium effects across a broad range of environments. p. 1 A recent paper by Donaldson (2026) terms this bias the heterogeneous-exposure-to-dynamic-general-equilibrium (HEDGE) bias. The
central contribution of this paper is to characterize the HEDGE bias under general aggregate dynamics and
to develop a regression-based framework, heterogeneous-exposure local-projection instrumental variables
(HELP-IV), which identifies the dynamic partial-equilibrium effects across a broad range of environments. We start by deriving an explicit formula for the dynamic partial-equilibrium effect in a backward-looking
environment and by characterizing in closed form the HEDGE bias of the TWFE-IV estimate, which
contains an exposure-correlation component and an aggregate-feedback component. Intuitively, the effect
is the outcome response divided by the impact response of the treatment, in the experiment that holds the
confounder path fixed. To hedge out the HEDGE bias, we develop for this environment the benchmark HELP-IV design, which
augments the TWFE-IV local projection with the path controls and instruments boxed below. At each
horizon ℎ, HELP-IV regresses the outcome 𝑌𝑖,𝑡+ℎ of unit 𝑖 on the treatment 𝐺𝑡 and on the confounder path
𝑅𝑡
, .Where the paper says it
Where the paper says it
Where the paper says it
The framework is formally established through several propositions demonstrating that HELP-IV identifies the partial-equilibrium effect (phi P h) and that the resulting estimator converges to it under various regularity conditions. p. 2 We prove that, with valid instruments, 𝑏
HELP
ℎ
identifies the dynamic partial-equilibrium effect. Intuitively, the regression identifies its ℎ + 2 coefficients
through ℎ + 2 moment conditions, one for each instrument, and identifying 𝑏
HELP
ℎ
is equivalent to taking a
linear combination of these conditions that eliminates the coefficients on the path controls.Where the paper says it
Empirical applications include re-estimating state fiscal multipliers and regional stock-wealth effects. p. 0 We further extend the framework, with the corresponding inference, to multiple exposure margins
and confounders, forward-looking outcomes, and cumulative multipliers. Finally, we apply HELP-IV to
state fiscal multipliers and regional stock-wealth effects, where it yields partial-equilibrium effects that
are smaller than conventional estimates and deliver different implications when ported to macroeconomic
models. We further extend the framework, with the corresponding inference, to multiple exposure margins
and confounders, forward-looking outcomes, and cumulative multipliers. Finally, we apply HELP-IV to
state fiscal multipliers and regional stock-wealth effects, where it yields partial-equilibrium effects that
are smaller than conventional estimates and deliver different implications when ported to macroeconomic
models. We further extend the framework, with the corresponding inference, to multiple exposure margins
and confounders, forward-looking outcomes, and cumulative multipliers. Finally, we apply HELP-IV to
state fiscal multipliers and regional stock-wealth effects, where it yields partial-equilibrium effects that
are smaller than conventional estimates and deliver different implications when ported to macroeconomic
models. To assess the performance of HELP-IV in finite samples and its practical importance, we conduct a set
of Monte Carlo simulations and two empirical applications. The simulations, which cover all our designs
and environments, show that HELP-IV recovers the target effect, with confidence intervals whose coverage
is generally close to the nominal level. We also adapt
this inference where our designs go beyond this setting. First, the instruments may be estimated in a first
step, as in composite HELP-IV or when shocks from a structural vector autoregression (SVAR) replace
unavailable proxies. 2 The extension
of HELP-IV to multiple exposure margins, multiple confounders, and transformed treatments and outcomes
broadens this scope further. On the other hand, most HELP-IV designs are one-step LP-IV regressions,
so they can be estimated with standard statistical packages, and inference builds on standard IV methods.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Improve identification of dynamic partial-equilibrium effects by developing a regression based framework called heterogeneous-exposure local-projection instrumental variables (HELP-IV). This framework compares units while holding the confounder path fixed by instrumenting both the treatment and this path with exposure weighted shock proxies, which proves that HELP-IV identifies the partial equilibrium effect.
-
Extend identification to multiple exposure margins by identifying a separate partial equilibrium effect for each margin through interaction between treatment, path controls, and instruments with each margin, as shown in Corollary 1. This allows the framework to study distinct sources of fiscal transmission like prime contract spending versus subcontract spending.
-
Enhance robustness in multiple confounder settings by developing composite HELP-IV and targeting-rule HELP-IV designs to handle different classes of applications while maintaining identification under general equilibrium dynamics. Composite HELP-IV is preferred when proxies are noisy and the horizon is long, while targeting-rule HELP-IV is the most precise design when a targeting rule is available.
-
Develop forward looking partial equilibrium effects by extending composite HELP-IV into EHELP-IV, which uses an instrument that combines the treatment proxy with dated confounder proxies to fix both ex post and in expectations formed at dates t through t + h. This allows for the identification of the information constrained effect when news shocks are few.
-
Improve estimation accuracy by employing Monte Carlo simulations to assess identification and inference performance across all designs and environments, showing that HELP-IV recovers the target effect with confidence intervals whose coverage is generally close to the nominal level.
Abstract
Cross-sectional designs are widely used to estimate dynamic partial-equilibrium effects of an aggregate treatment. We show analytically that these estimates are biased when exposures to the treatment and to aggregate confounders are correlated or the confounders feed back into the treatment. This paper develops the heterogeneous-exposure local-projection instrumental variables (HELP-IV) framework, which compares units while holding the confounder path fixed by instrumenting the treatment and this path with exposure-weighted shock proxies. We prove that HELP-IV identifies the partial-equilibrium effect. We further extend the framework, with the corresponding inference, to multiple exposure margins and confounders, forward-looking outcomes, and cumulative multipliers. Finally, we apply HELP-IV to state fiscal multipliers and regional stock-wealth effects, where it yields partial-equilibrium effects that are smaller than conventional estimates and deliver different implications when ported to macroeconomic models.
Sources
Related papers
- Stationarity with Occasionally Binding Constraints
- Backtesting High-Dimensional Expected Shortfall
- Higher-Order Neyman Orthogonality in Moment-Condition Models
- Online Bayesian Inference for a Dirichlet Process Mixture of Binary Logit Models
- Debiased Conditional Independence Tests for Causal Effects
- Sharp Testable Implications of Encouragement Designs