Inference on Common Trends in a Cointegrated Nonlinear SVAR

James A. Duffy; Xiyu Jiao

arxiv: 2507.22869 · v2 · submitted 2025-07-30 · 💰 econ.EM · math.ST· stat.TH

Inference on Common Trends in a Cointegrated Nonlinear SVAR

James A. Duffy , Xiyu Jiao This is my paper

Pith reviewed 2026-05-19 02:50 UTC · model grok-4.3

classification 💰 econ.EM math.STstat.TH

keywords nonlinear cointegrationcommon trendsSVARvariance ratio testasymptoticsnonstationary time seriesstructural models

0 comments

The pith

A modified variance ratio test accurately infers the number of common trends in nonlinearly cointegrated systems.

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

This paper addresses inference on the number of common stochastic trends in time series generated by cointegrated nonlinear structural vector autoregressions. It introduces a modification to a standard test that remains valid when cointegrating relations take a nonlinear form of known type. The authors establish the necessary asymptotic results by developing a dual linear process approximation for certain nonstationary autoregressive processes. A sympathetic reader would care because correctly identifying the number of trends is crucial for understanding long-run dynamics in economic data that exhibit regime switches or other nonlinearities. Without the modification, tests tend to suggest more independent trends than are truly present.

Core claim

The paper establishes that for data from a two-regime piecewise affine SVAR with nonlinear cointegration, the modified test correctly identifies the number of common trends present, whereas the standard test overstates this number.

What carries the argument

A modified multivariate variance ratio test that accounts for nonlinear cointegration through a dual linear process approximation for deriving its limiting distribution.

If this is right

The number of common trends can be inferred reliably even with nonlinear relationships.
The unmodified test will incorrectly suggest extra trends when nonlinearity is present.
The approach works for stable nonstationary processes that admit the required approximation.
Correct trend inference supports better understanding of long-run economic equilibria.

Where Pith is reading between the lines

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

Similar modifications might apply to other tests for cointegration in nonlinear settings.
This could influence how regime-dependent models are used in forecasting long-term economic variables.
Extensions to cases where the nonlinear form is not fully known would broaden applicability.

Load-bearing premise

The form of the nonlinear cointegration relationship is known beforehand.

What would settle it

Generate artificial data from a nonlinear cointegrated model with a known number of trends and verify whether the modified test recovers that number accurately while the original test does not.

read the original abstract

We consider the problem of performing inference on the number of common stochastic trends when data is generated by a cointegrated CKSVAR (a two-regime, piecewise affine SVAR; Mavroeidis, 2021), using a modified version of the Breitung (2002) multivariate variance ratio test that is robust to the presence of nonlinear cointegration (of a known form). To derive the asymptotics of our test statistic, we prove a fundamental LLN-type result for a class of stable but nonstationary autoregressive processes, using a novel dual linear process approximation. We show that our modified test yields correct inferences regarding the number of common trends in such a system, whereas the unmodified test tends to infer a higher number of common trends than are actually present, when cointegrating relations are nonlinear.

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.

Desk Editor's Note private letter to a colleague

The paper gives a modified Breitung test that corrects for known nonlinear cointegration in two-regime CKSVARs, supported by a new LLN result via dual linear process approximation.

read the letter

The main point is that this paper modifies the Breitung variance ratio test so it has the right asymptotics when cointegrating relations follow the known piecewise affine form in a CKSVAR. The unmodified version tends to overstate the number of common trends, and the fix plus the supporting limit theory is what they put forward as new. They derive the asymptotics from a fresh LLN-type result for stable nonstationary AR processes that uses a dual linear process approximation to handle the nonlinearity after the transformation is applied. That combination is the concrete advance over the Mavroeidis model and the original Breitung test. The work is useful for anyone who already works with regime-switching cointegrated systems and needs a practical adjustment to an existing test rather than an entirely new procedure. The central claim holds up on its own terms if the approximation delivers uniform control on the error, which the abstract states it does. The soft spot is exactly where the stress-test note points: in the two-regime case the transformed series carry regime-dependent coefficients, and it is not obvious from the given material whether the approximation error vanishes at the required rate near the threshold or when roots approach unity. The abstract gives no simulation results or explicit bounds, so the practical size control remains unverified in the information available. This is narrow enough that it will mainly interest specialists in nonlinear macroeconometric time series. A reader who needs to test common trends in piecewise affine setups will get a usable adjustment if the derivations check out. I would send it to peer review so the approximation and any Monte Carlo evidence can be examined directly.

Referee Report

2 major / 2 minor

Summary. The paper proposes a modified Breitung (2002) multivariate variance-ratio test for inferring the number of common stochastic trends in a cointegrated CKSVAR (two-regime piecewise-affine SVAR) with known nonlinear cointegration. To obtain the asymptotics, the authors prove a new LLN-type result for stable nonstationary autoregressive processes via a novel dual linear-process approximation. They claim that the modified test delivers correct inferences on the number of trends, while the unmodified version tends to overstate the number when cointegration is nonlinear.

Significance. If the dual linear-process approximation and associated error bounds hold with the required uniformity, the result would extend reliable trend inference to a practically relevant class of nonlinear cointegrated systems. The work supplies a concrete, implementable modification of an existing test together with a supporting limit theory, which could be useful for macroeconometric applications involving regime-dependent cointegration.

major comments (2)

[Section 3 (LLN derivation)] The central claim that the modified test has the same limiting null distribution as in the linear case rests on the new LLN result obtained via the dual linear-process representation. The manuscript does not supply explicit uniform error bounds or convergence rates for the approximation error, particularly near regime boundaries or when autoregressive roots approach unity; without these rates it is not possible to confirm that the approximation error is negligible at the order needed for the variance-ratio statistic.
[Section 5 (Monte Carlo)] The simulation design used to illustrate size control and power of the modified versus unmodified test is not described in sufficient detail (e.g., specific parameter values for the piecewise-affine cointegrating relations, persistence levels, or sample sizes). This makes it difficult to assess whether the reported finite-sample behavior actually corroborates the asymptotic claim under the conditions where the dual approximation is most fragile.

minor comments (2)

[Section 2] Notation for the regime indicator and the known nonlinear transformation should be introduced once and used consistently; occasional switches between different symbols for the same object reduce readability.
[Abstract and Section 4] The statement that the unmodified test 'tends to infer a higher number of common trends' would benefit from a precise reference to the relevant theorem or corollary that quantifies the direction of the size distortion.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments, which help clarify the presentation of our asymptotic results and improve the transparency of the Monte Carlo evidence. We address each major comment below and outline the revisions we will make to strengthen the manuscript.

read point-by-point responses

Referee: [Section 3 (LLN derivation)] The central claim that the modified test has the same limiting null distribution as in the linear case rests on the new LLN result obtained via the dual linear-process representation. The manuscript does not supply explicit uniform error bounds or convergence rates for the approximation error, particularly near regime boundaries or when autoregressive roots approach unity; without these rates it is not possible to confirm that the approximation error is negligible at the order needed for the variance-ratio statistic.

Authors: We agree that explicit uniform error bounds and convergence rates would make the argument more transparent. The current draft establishes the LLN via the dual linear-process approximation and shows that the remainder term vanishes in probability under the maintained stability conditions, but does not derive explicit rates. In the revision we will add an appendix that supplies uniform bounds on the approximation error, with explicit rates that hold uniformly over a compact set of autoregressive roots strictly inside the unit circle and over the regime boundaries. These bounds will be shown to be of sufficiently small order for the variance-ratio statistic to inherit the same limiting null distribution as in the linear case. revision: yes
Referee: [Section 5 (Monte Carlo)] The simulation design used to illustrate size control and power of the modified versus unmodified test is not described in sufficient detail (e.g., specific parameter values for the piecewise-affine cointegrating relations, persistence levels, or sample sizes). This makes it difficult to assess whether the reported finite-sample behavior actually corroborates the asymptotic claim under the conditions where the dual approximation is most fragile.

Authors: We accept that the simulation section currently lacks sufficient detail for full reproducibility and for evaluating performance near the most delicate regions of the parameter space. In the revised manuscript we will expand Section 5 with a complete description of the data-generating process, including the exact functional forms and parameter values of the piecewise-affine cointegrating relations, the specific autoregressive coefficients (with persistence levels ranging from 0.7 to 0.95), the sample sizes (T = 100, 200, 500), and the number of replications. We will also add a new set of experiments that deliberately place the design near regime boundaries and close to unit roots to directly assess whether the finite-sample behavior aligns with the strengthened asymptotic results. revision: yes

Circularity Check

0 steps flagged

No circularity: asymptotics rest on independent new LLN result

full rationale

The paper's derivation chain establishes the limiting distribution of the modified Breitung variance-ratio statistic under known-form nonlinear cointegration by proving a new LLN-type result for stable nonstationary autoregressive processes, obtained via a dual linear-process approximation. This LLN is introduced as a fundamental auxiliary result rather than being fitted to the target data or defined in terms of the test statistic itself. No load-bearing step reduces by construction to a self-citation, a fitted parameter renamed as a prediction, or an ansatz smuggled from prior work by the same authors; the unmodified test's tendency to overstate the number of trends is shown via the same external asymptotic benchmark. The derivation is therefore self-contained against the stated assumptions.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim depends on the assumption that nonlinear cointegration is known in form and that the processes satisfy conditions allowing the dual linear process approximation; no free parameters or invented entities are introduced in the abstract.

axioms (2)

domain assumption Nonlinear cointegration is of a known form.
Explicitly stated in the abstract as the setting for which the modified test is robust.
domain assumption The data-generating process consists of stable but nonstationary autoregressive processes admitting a dual linear process approximation.
Invoked to establish the LLN-type result and test asymptotics.

pith-pipeline@v0.9.0 · 5665 in / 1344 out tokens · 76451 ms · 2026-05-19T02:50:53.497727+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Foundation/AbsoluteFloorClosure.lean reality_from_one_distinction unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

prove a fundamental LLN-type result for a class of stable but nonstationary autoregressive processes, using a novel dual linear process approximation

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

18 extracted references · 18 canonical work pages

[1]

Aruoba, S. B., M. Mlikota, F. Schorfheide, and S. Villalvazo (2022): SVARs with occasionally-binding constraints, Journal of Econometrics, 231, 477--499

work page 2022
[2]

Berkes, I. and L. Horv \'a th (2006): Convergence of integral functionals of stochastic processes, Econometric Theory, 22, 304--22

work page 2006
[3]

(2002): Nonparametric tests for unit roots and cointegration, Journal of Econometrics, 108, 343--363

Breitung, J. (2002): Nonparametric tests for unit roots and cointegration, Journal of Econometrics, 108, 343--363

work page 2002
[4]

Duffy, J. A. and S. Mavroeidis (2024): Common trends and long-run identification in nonlinear structural VARs , arXiv:2404.05349

work page arXiv 2024
[5]

Duffy, J. A., S. Mavroeidis, and S. Wycherley (2023): Stationarity with Occasionally Binding Constraints, arXiv:2307.06190

work page arXiv 2023
[6]

--- -.1pt --- -.1pt --- (2025): Cointegration with Occasionally Binding Constraints, arXiv:2211.09604v3

work page arXiv 2025
[7]

Engle, R. F. and C. W. J. Granger (1987): Co-integration and error correction: r epresentation, estimation, and testing, Econometrica, 251--276

work page 1987
[8]

Granger, C. W. J. (1986): Developments in the study of cointegrated economic variables. Oxford Bulletin of Economics & Statistics, 48

work page 1986
[9]

Hall, P. and C. C. Heyde (1980): Martingale Limit Theory and Its Application, Academic Press

work page 1980
[10]

Ikeda, D., S. Li, S. Mavroeidis, and F. Zanetti (2024): Testing the effectiveness of unconventional monetary policy in Japan and the United States, American Economic Journal: Macroeconomics, 16, 250--286

work page 2024
[11]

(1991): Estimation and hypothesis testing of cointegration vectors in Gaussian vector autoregressive models, Econometrica, 1551--1580

Johansen, S. (1991): Estimation and hypothesis testing of cointegration vectors in Gaussian vector autoregressive models, Econometrica, 1551--1580

work page 1991
[12]

--- -.1pt --- -.1pt --- (1995): Likelihood-based Inference in Cointegrated Vector Autoregressive Models, O.U.P

work page 1995
[13]

Kristensen, D. and A. Rahbek (2010): Likelihood-based inference for cointegration with nonlinear error-correction, Journal of Econometrics, 158, 78--94

work page 2010
[14]

(2021): Identification at the zero lower bound, Econometrica, 69, 2855--2885

Mavroeidis, S. (2021): Identification at the zero lower bound, Econometrica, 69, 2855--2885

work page 2021
[15]

Revuz, D. and M. Yor (1999): Continuous Martingales and Brownian Motion, Berlin: Springer, 3 ed

work page 1999
[16]

Tj stheim, and C

Ter \"a svirta, T., D. Tj stheim, and C. W. J. Granger (2010): Modelling Nonlinear Economic Time Series, O.U.P

work page 2010
[17]

(2020): Some notes on nonlinear cointegration: a partial review with some novel perspectives, Econometric Reviews, 39, 655--673

Tj stheim, D. (2020): Some notes on nonlinear cointegration: a partial review with some novel perspectives, Econometric Reviews, 39, 655--673

work page 2020
[18]

(1990): Non-linear Time Series: a dynamical system approach , O.U.P

Tong, H. (1990): Non-linear Time Series: a dynamical system approach , O.U.P

work page 1990

[1] [1]

Aruoba, S. B., M. Mlikota, F. Schorfheide, and S. Villalvazo (2022): SVARs with occasionally-binding constraints, Journal of Econometrics, 231, 477--499

work page 2022

[2] [2]

Berkes, I. and L. Horv \'a th (2006): Convergence of integral functionals of stochastic processes, Econometric Theory, 22, 304--22

work page 2006

[3] [3]

(2002): Nonparametric tests for unit roots and cointegration, Journal of Econometrics, 108, 343--363

Breitung, J. (2002): Nonparametric tests for unit roots and cointegration, Journal of Econometrics, 108, 343--363

work page 2002

[4] [4]

Duffy, J. A. and S. Mavroeidis (2024): Common trends and long-run identification in nonlinear structural VARs , arXiv:2404.05349

work page arXiv 2024

[5] [5]

Duffy, J. A., S. Mavroeidis, and S. Wycherley (2023): Stationarity with Occasionally Binding Constraints, arXiv:2307.06190

work page arXiv 2023

[6] [6]

--- -.1pt --- -.1pt --- (2025): Cointegration with Occasionally Binding Constraints, arXiv:2211.09604v3

work page arXiv 2025

[7] [7]

Engle, R. F. and C. W. J. Granger (1987): Co-integration and error correction: r epresentation, estimation, and testing, Econometrica, 251--276

work page 1987

[8] [8]

Granger, C. W. J. (1986): Developments in the study of cointegrated economic variables. Oxford Bulletin of Economics & Statistics, 48

work page 1986

[9] [9]

Hall, P. and C. C. Heyde (1980): Martingale Limit Theory and Its Application, Academic Press

work page 1980

[10] [10]

Ikeda, D., S. Li, S. Mavroeidis, and F. Zanetti (2024): Testing the effectiveness of unconventional monetary policy in Japan and the United States, American Economic Journal: Macroeconomics, 16, 250--286

work page 2024

[11] [11]

(1991): Estimation and hypothesis testing of cointegration vectors in Gaussian vector autoregressive models, Econometrica, 1551--1580

Johansen, S. (1991): Estimation and hypothesis testing of cointegration vectors in Gaussian vector autoregressive models, Econometrica, 1551--1580

work page 1991

[12] [12]

--- -.1pt --- -.1pt --- (1995): Likelihood-based Inference in Cointegrated Vector Autoregressive Models, O.U.P

work page 1995

[13] [13]

Kristensen, D. and A. Rahbek (2010): Likelihood-based inference for cointegration with nonlinear error-correction, Journal of Econometrics, 158, 78--94

work page 2010

[14] [14]

(2021): Identification at the zero lower bound, Econometrica, 69, 2855--2885

Mavroeidis, S. (2021): Identification at the zero lower bound, Econometrica, 69, 2855--2885

work page 2021

[15] [15]

Revuz, D. and M. Yor (1999): Continuous Martingales and Brownian Motion, Berlin: Springer, 3 ed

work page 1999

[16] [16]

Tj stheim, and C

Ter \"a svirta, T., D. Tj stheim, and C. W. J. Granger (2010): Modelling Nonlinear Economic Time Series, O.U.P

work page 2010

[17] [17]

(2020): Some notes on nonlinear cointegration: a partial review with some novel perspectives, Econometric Reviews, 39, 655--673

Tj stheim, D. (2020): Some notes on nonlinear cointegration: a partial review with some novel perspectives, Econometric Reviews, 39, 655--673

work page 2020

[18] [18]

(1990): Non-linear Time Series: a dynamical system approach , O.U.P

Tong, H. (1990): Non-linear Time Series: a dynamical system approach , O.U.P

work page 1990