Intentionally Overidentifying Local Projections

arXiv:2610.13269 · econ.EM · Submitted 2026-10-09 · Read on arXiv

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: I'm Ruth.

Warren: And I'm Warren, here with Iris.

Ruth: We're looking at the paper: "Intentionally Overidentifying Local Projections".

Warren: Here's the gist of the paper. Local projections (LPs) have become a standard tool for estimating dynamic effects in macroeconomics and finance because they estimate horizon-specific impulse responses directly and can readily accommodate a range of empirically relevant features The resulting (optimally weighted) IV estimator is never asymptotically less efficient than OLS under a finite-order model with martingale-difference shocks, and it can deliver substantial efficiency gains when data are persistent The paper proposes an alternative approach to Local Projections (LPs) by exploiting additional lags as excluded instrumental variables (IVs) to construct an optimally weighted IV estimator via the generalized method of moments (GMM) The method involves defining a target parameter regression, the local projection (LP), and then constructing a GMM(p,q) estimator by using lag vectors up to order q as instruments for lags up to order p, where q ≥ p ≥ d The optimal weighting is achieved through a HAC (heteroskedasticity and autocorrelation consistent) score variance estimator Efficiency Gains and Redundancy The analysis shows that the new method can yield substantial efficiency gains over LP OLS, arising from the specific structure of the model For horizon-one local projections with i.i.d.

Ruth: errors, additional lags are generally redundant excluded instrumental variables, but this redundancy does not extend to a general LP regression at horizons larger than one The optimally weighted GMM can produce a strict efficiency gain over OLS especially when the data are strongly persistent, arising primarily from the serial correlation of the GMM (residualized) score process Robustness and Inference Theory The paper develops IV-based LP inference that is robust to unknown persistence and moderate local model misspecification, whereas recursive VAR-based inference is generally not robust to the latter Uniform GMM inference is established under Assumptions 1–3, showing that the Wald statistic based on a general GMM(p,q) LP estimator is asymptotically pivotal uniformly over the persistence of the data Furthermore, Theorem 2 establishes persistence-uniform inference for GMM-LALP under local VARMA misspecification Empirical Validation Monte Carlo simulations and an empirical application demonstrate that the proposed method can yield meaningfully shorter confidence intervals for impulse responses than OLS in finite samples The GMM-LALP confidence intervals are systematically shorter than the LALP confidence intervals, reflecting the efficiency gain from using valid lagged instruments and optimally weighting the resulting moment conditions Empirical results reinforce these findings, showing that the optimal GMM weighting provides additional precision conditional on that same lag-augmented specification This paper proposes a new method for estimating impulse responses and conducting inference based on local projections by exploiting additional lags as excluded instruments and optimally combining the resulting moment conditions using GMM The authors show that these additional moments can deliver substantial efficiency gains relative to OLS particularly when the data are persistent They formally establish the robustness of the proposed inference to data persistence and model misspecification The optimal GMM weighting provides additional precision conditional on that same lag-augmented specification.

Warren: References BAKER, S.

Ruth: R., N.

Paper summary: Ruth: We just talked about the core idea of this paper, "Intentionally Overidentifying Local Projections," which is proposing a new way to estimate impulse responses using local projections by intentionally overidentifying them with extra lags as excluded instruments and then optimally combining those moment conditions through GMM.

Warren: So, to recap what we covered before, the thesis is that under a finite-order model with martingale-difference shocks, this resulting optimal weighted IV estimator is never asymptotically less efficient than OLS and it can actually give some substantial efficiency gains when the data are persistent.

Iris: And we talked about how they set up the GMM(p,q) estimator by using lag vectors up to order q as instruments for lags up to order p where q is greater than or equal to p and also greater than or equal to the degree of freedom d.

Ruth: That’s right, and then we discussed that even though extra lags are redundant for horizon-one projections with i.i.d errors, they become non-redundant when the horizon is larger than one, or when there's conditional heteroskedasticity present.

Warren: So the crucial finding is that this optimally weighted GMM can produce a strict efficiency gain over OLS especially when the data are strongly persistent because of that serial correlation in the GMM residualized score process.

Iris: They also worked hard on inference, developing IV-based LP inference that is robust to unknown persistence and moderate local model misspecification, which they contrast against recursive VAR-based inference.

Ruth: And they showed this new method is robust to data persistence and model misspecification, formally establishing that the Wald statistic based on a general GMM(p,q) LP estimator is asymptotically pivotal uniformly over the persistence of the data.

Warren: They also established persistence-uniform inference for GMM-LALP even when there's local VARMA misspecification.

Conclusion: Ruth: So we’re wrapping up with a look at the title and authors of this work, "Intentionally Overidentifying Local Projections." The paper introduces this new method for estimating impulse responses and conducting inference based on local projections by exploiting additional lags as excluded instruments and optimally combining the resulting moment conditions using GMM.

Warren: The main implication is that these additional moments can deliver substantial efficiency gains relative to OLS particularly when the data are persistent.

Iris: This means we get more precise estimates for impulse responses, especially when the traditional methods would normally give us wide confidence intervals because of persistence.

Ruth: It’s about using those extra lags intelligently to improve the precision of our forecasts and understanding how shocks propagate over time in a way that accounts for data persistence.

Warren: The work formalizes this by establishing robust inference to unknown persistence and moderate local model misspecification, showing that the optimal GMM weighting provides additional precision conditional on that same lag-augmented specification.

CUFE University of Mannheim · Indiana University

econ.EM

Submitted: 2026-10-09

Updated: 2026-11-13

Project page: https://nbloom.people.stanford.edu/research

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 90/100

The gist: Local projections (LPs) have become a standard tool for estimating dynamic effects in macroeconomics and finance because they estimate horizon-specific impulse responses directly and can readily…

Key concepts

Local Projections (LPs)
LPs are a standard tool used in macroeconomics to estimate dynamic effects by directly calculating horizon-specific impulse responses. They are widely used but can be inefficient under certain conditions, prompting the need for alternative methods.
Generalized Method of Moments (GMM)
GMM is a statistical technique that estimates parameters by using moment conditions. In this paper, it is used to construct an optimally weighted estimator by employing additional lags as instruments for the main regression lags.
Efficiency Gains
The new GMM-LALP method achieves significant efficiency gains over standard LP OLS, particularly when data exhibit strong persistence. This improvement stems from using valid lagged instruments and optimally weighting the resulting moment conditions to reduce estimation error.

Terminology

Summary

Local projections (LPs) have become a standard tool for estimating dynamic effects in macroeconomics and finance because they estimate horizon-specific impulse responses directly and can readily accommodate a range of empirically relevant features The resulting (optimally weighted) IV estimator is never asymptotically less efficient than OLS under a finite-order model with martingale-difference shocks, and it can deliver substantial efficiency gains when data are persistent

Methodology and Model Setup

The paper proposes an alternative approach to Local Projections (LPs) by exploiting additional lags as excluded instrumental variables (IVs) to construct an optimally weighted IV estimator via the generalized method of moments (GMM) The method involves defining a target parameter regression, the local projection (LP), and then constructing a GMM(p,q) estimator by using lag vectors up to order q as instruments for lags up to order p, where q ≥ p ≥ d The optimal weighting is achieved through a HAC (heteroskedasticity and autocorrelation consistent) score variance estimator

Efficiency Gains and Redundancy

The analysis shows that the new method can yield substantial efficiency gains over LP OLS, arising from the specific structure of the model For horizon-one local projections with i.i.d. errors, additional lags are generally redundant excluded instrumental variables, but this redundancy does not extend to a general LP regression at horizons larger than one The optimally weighted GMM can produce a strict efficiency gain over OLS especially when the data are strongly persistent, arising primarily from the serial correlation of the GMM (residualized) score process

Robustness and Inference Theory

The paper develops IV-based LP inference that is robust to unknown persistence and moderate local model misspecification, whereas recursive VAR-based inference is generally not robust to the latter Uniform GMM inference is established under Assumptions 1–3, showing that the Wald statistic based on a general GMM(p,q) LP estimator is asymptotically pivotal uniformly over the persistence of the data Furthermore, Theorem 2 establishes persistence-uniform inference for GMM-LALP under local VARMA misspecification

Empirical Validation

Monte Carlo simulations and an empirical application demonstrate that the proposed method can yield meaningfully shorter confidence intervals for impulse responses than OLS in finite samples The GMM-LALP confidence intervals are systematically shorter than the LALP confidence intervals, reflecting the efficiency gain from using valid lagged instruments and optimally weighting the resulting moment conditions Empirical results reinforce these findings, showing that the optimal GMM weighting provides additional precision conditional on that same lag-augmented specification

Conclusion

This paper proposes a new method for estimating impulse responses and conducting inference based on local projections by exploiting additional lags as excluded instruments and optimally combining the resulting moment conditions using GMM The authors show that these additional moments can deliver substantial efficiency gains relative to OLS particularly when the data are persistent They formally establish the robustness of the proposed inference to data persistence and model misspecification The optimal GMM weighting provides additional precision conditional on that same lag-augmented specification.

References

BAKER, S. R., N. BLOOM, AND S. J. DAVIS (2016): “Measuring economic policy uncertainty,” Quarterly Journal of Economics, 131, 1593–1636

BARNICHON, R. AND C. BROWNLEES (2019): “Impulse Response Estimation by Smooth Local Projections,” The Review of Economics and Statistics, 101, 522–530

DE VOS, I. AND G. EVERAERT (2026): “GLS estimation of local projections: Trading robustness for efficiency,” Journal of Econometrics, 253, 106182

DINH, V. H., D. NIBBERING, AND B. p. 1 p. 1 WONG (2024): “Random Subspace Local Projections,” Review of Economics and Statistics, Forthcoming

DUFOUR, J.-M. AND E. RENAULT (1998): “Short run and long run causality in time series: theory,” Econometrica, 66, 1099–1125

FERNÁNDEZ-VILLAVERDE, J., J. F. RUBIO-RAMÍREZ, T. J. SARGENT, AND M. W. WATSON (2007): “ABCs (and Ds) of Understanding VARs,” American Economic Review, 97, 1021–1026

FERREIRA, L. N., S. MIRANDA-AGRIPPINO, AND G. p. 1 p. 1 RICCO (2025): “Bayesian Local Projections,” The Review of Economics and Statistics, 107, 1424–1438

GONÇALVES, S., A. M. HERRERA, L. p. 1 p. 1 KILIAN, AND E. p. 1 p. 1 PESAVENTO (2024): “State-dependent local projections,” Journal of Econometrics, 105702

HALL, A. R., A. INOUE, K. JANA, AND C. p. 8 p. 8 SHIN (2007): “Information in generalized method of moments estimation and entropy-based moment selection,” Journal of Econometrics, 138, 488–512

HANSEN, L. P. (1982): “Large Sample Properties of Generalized Method of Moments Estimators,” Econometrica, 50, 1029–1054

HUBER, F., C. MATTHES, AND M. p. 3 PFARRHOFER (2025): “General Seemingly Unrelated Local Projections,” ArXiv:2410. p. 417105

INOUE, A., Ò. JORDÀ, AND G. p. 1 M. KUERSTEINER (2026): “Inference for local projections,” The Econometrics Journal, 29, 2–26

INOUE, A. AND L. KILIAN (2020): “The uniform validity of impulse response inference in autoregressions,” Journal of Econometrics, 215, 450–472

LUSOMPA, A. (2023): “Local projections, autocorrelation, and efficiency,” Quantitative Economics, 14, 1199–1220

LÜTKEPOHL, H. (2005): New Introduction to Multiple Time Series Analysis, Berlin: Springer

MONTIEL OLEA, J. AND M. PLAGBORG-MØLLER (2021): “Local projection inference is simpler and more robust than you think,” Econometrica, 89, 1789–1823

MONTIEL OLEA, J., M. PLAGBORG-MØLLER, E. p. 0 p. 0 QIAN, AND C. p. 1 p. 1 WOLF (2026): “Double robustness of local projections and some unpleasant varithmetic,” Econometrica, 94, 1313–1343

MONTIEL OLEA, J., J. STOCK, AND M. p. 2 p. 2 p. 2 p. 2 WATSON (2021): “Inference in structural vector autoregressions identified with an external instrument,” Journal of Econometrics, 225, 74–87

RAMEY, V. A. (2016): “Macroeconomic shocks and their propagation,” Handbook of Macroeconomics, 2, 71–162

SIMS, C. A. (1980): “Macroeconomics and reality,” Econometrica, 48, 1–48

SIMS, C. A., J. H. STOCK, AND M. W. WATSON (1990): “Inference in linear time series models with some unit roots,” Econometrica, 58, 113–144

STOCK, J. H. AND M. W. WATSON (2018): “Identification and estimation of dynamic causal effects in macroeconomics using external instruments,” Economic Journal, 128, 917–948

XU, K.-L. (2025): “Local Projection Based Inference under General Conditions,” Forthcoming, The Review of Economic Studies, rdag034

——— (2026): “Impulse Responses, Efficiency Bounds, and Newey-West Inference,” Working paper, Indiana University Bloomington

BASU, D. (2024): “Frisch–Waugh–Lovell theorem-type results for the k-Class and 2SGMM estimators,” Statistics & Probability Letters, 213, 110188

BREUSCH, T., H. QIAN, P. SCHMIDT, AND D. p. 8 WYHOWSKI (1999): “Redundancy of moment conditions,” Journal of Econometrics, 91, 89–111

DE VOS, I. AND G. EVERAERT (2026): “GLS estimation of local projections: Trading robustness for efficiency,” Journal of Econometrics, 253, 106182

DINH, V. H., D. NIBBERING, AND B. WONG (2024): “Random Subspace Local Projections,” Review of Economics and Statistics, Forthcoming

DUFOUR, J.-M. AND E. RENAULT (1998): “Short run and long run causality in time series: theory,” Econometrica, 66, 1099–1125

FERNÁNDEZ-VILLAVERDE, J., J. F. RUBIO-RAMÍREZ, T. J. SARGENT, AND M. W. WATSON (2007): “ABCs (and Ds) of Understanding VARs,” American Economic Review, 97, 1021–1026

FERREIRA, L. N., S. MIRANDA-AGRIPPINO, AND G. RICCO (2025): “Bayesian Local Projections,” The Review of Economics and Statistics, 107, 1424–1438

GONÇALVES, S., A. M. HERRERA, L. KILIAN, AND E. PESAVENTO (2024): “State-dependent local projections,” Journal of Econometrics, 105702

HALL, A. R., A. INOUE, K. JANA, AND C. SHIN (2007): “Information in generalized method of moments estimation and entropy-based moment selection,” Journal of Econometrics, 138, 488–512

HALL, P. AND C. C. HEYDE (1980): Martingale Limit Theory and Its Application, New York: Academic Press

HORN, R. A., AND C. R. JOHNSON (2013): Matrix Analysis, Cambridge: Cambridge University Press, 2 ed

MONTIEL OLEA, J., M. PLAGBORG-MØLLER, E. QIAN, AND C. WOLF (2026): “Double robustness of local projections and some unpleasant varithmetic,” Econometrica, 94, 1313–1343

MONTIEL OLEA, J., J. STOCK, AND M. WATSON (2021): “Inference in structural vector autoregressions identified with an external instrument,” Journal of Econometrics, 225, 74–87

RAMEY, V. A.: “Macroeconomic shocks and their propagation,” Handbook of Macroeconomics, 2, 71–162

SIMS, C. A.: “Macroeconomics and reality,” Econometrica, 48, 1–48

SIMS, C. A., J. H. STOCK, AND M. W. WATSON (1990): “Inference in linear time series models with some unit roots,” Econometrica, 58, 113–144

STOCK, J. H. AND M. W. WATSON (2018): “Identification and estimation of dynamic causal effects in macroeconomics using external instruments,” Economic Journal, 128, 917–948

XU, K.-L.

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

  1. The AI can estimate impulse responses using an optimally weighted Generalized Method of Moments (GMM) estimator that exploits additional lags as excluded instrumental variables, which is shown to be never asymptotically less efficient than OLS under finite-order models with martingale-difference shocks.

  2. The system can provide robust inference regarding impulse responses even when the data exhibits persistence or moderate model misspecification, as the proposed method is robust to both data persistence and moderate model misspecification.

  3. The AI can generate meaningfully shorter confidence intervals for impulse responses than OLS in finite samples by using a GMM-LALP procedure, which is demonstrated in simulations to yield better coverage than the standard Local Projection (LP) estimator.

  4. The system can determine the optimal number of lag controls to use by characterizing the trade-off between the degree of misspecification and the number of additional lag controls, recommending a specific setting like p = d + 1 and q = d + 2 for practical application.

  5. The AI can perform uniform inference over data persistence under specified assumptions, as evidenced by Theorem 1, which states that the Wald statistic based on a GMM-LP estimator is asymptotically pivotal uniformly over the persistence of the data.

  6. The system can handle local misspecification in VARMA models by using lag augmentation (p = d + s) where s is chosen based on misspecification order, allowing inference to remain robust even when the model error exceeds order T - 1/2.

  7. The AI can perform a bootstrap implementation for GMM-LP procedures, utilizing a recursive wild bootstrap adapted to GMM–LP to generate confidence intervals that are shown to be systematically shorter than LALP confidence intervals.

Sources

Related papers