Pith. sign in

REVIEW 2 major objections 6 minor 49 references

A Synthetic Business Cycle Approach to Counterfactual Analysis with Nonstationary Macroeconomic Data

T0 review · 2 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read This paper argues that conventional synthetic control on nonstationary macroeconomic outcomes can yield spurious counterfactuals, and proposes a trend-cycle decomposition estimator that is asymptotically unbiased.

desk verdict The SBC idea is solid and worth reviewing, but the main theorem's proof rests on a perfect-fit assumption that contradicts the paper's own Lemma A.1. read the letter →

arxiv 2505.22388 v1 pith:QPUDRPWV submitted 2025-05-28 econ.EM

Zhentao Shi , Jin Xi , Haitian Xie This is my paper
classification econ.EM MSC 62M1062P20
keywords syntheticcontrolspuriousregressionnonstationarypaneldatatrend-cycledecompositionbusinesscyclesynchronizationcounterfactualpredictionHamiltonfiltercausalinference
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

The paper's target is a failure mode: when synthetic control is applied to nonstationary macroeconomic outcomes, a good pre-treatment fit can be pure artifact of independently trending series, and post-treatment counterfactuals can be badly wrong. The authors call this the spurious synthetic control problem and propose a synthetic business cycle (SBC) estimator that first detrends every series, forecasts the treated unit's trend from its own past, and uses donor units only to synthesize the stationary cyclical component. They prove that the SBC estimator is asymptotically unbiased for the counterfactual outcome under a factor structure on cycles, and consistent when cycles have no idiosyncratic noise. In simulations it reduces counterfactual mean-squared error substantially relative to conventional synthetic control, especially under spurious trends, and in the German and Hong Kong applications it passes placebo tests that the conventional estimator fails. If the paper is right, practitioners get a simple recipe for causal counterfactual analysis of trending macro data without pre-testing for cointegration.

What carries the argument

The load-bearing object is the Hamilton-filter trend-cycle decomposition used in a divide-and-conquer estimator. Hamilton's filter defines the trend $\tau_{i,t}$ as the linear projection of $Y_{i,t}(0)$ on its own values $h$ periods earlier (plus a constant), and the cycle $c_{i,t}$ as the resulting forecast error; because the forecast error is stationary for unit-root and trend-stationary processes, the synthetic control step is run on stationary inputs. The SBC estimator then combines the treated unit's own extrapolated trend with a synthetic cycle $\sum_{i=2}^{N+1}\hat{w}_i \hat{c}_{i,t}$, with weights fitted by matching pre-treatment cycles. The factor-structure assumption on cycles and the high-level consistency of the filter are what make the weighted cycle combination approximate the treated cycle, while the trend is kept separate precisely to avoid spurious regression.

What would settle it

Run a Monte Carlo design like Model 1 of the paper (independent random walks with drifts) with $T_0=200$, force the cycle-matching objective in equation (3) to a strictly positive minimum rather than zero, and record the bias of the SBC estimator at each post-treatment date. If the bias does not shrink as $T_0$ grows, the theorem's assumption of a perfect pre-treatment fit is doing the work and the asymptotic claim fails in that setting.

Watch

Extended reading notes

Core claim

The central claim is that the conventional synthetic control estimator can attribute common movement to treatment effects when outcomes contain unit-specific nonstationary trends, and that the correct division of labor is to trust the treated unit's own history for trend and the donor pool only for cycles. Formalized, the paper posits $Y_{i,t}(0)=\tau_{i,t}+c_{i,t}$ with stationary cycles $c_{i,t}=\lambda_i' f_t+\varepsilon_{i,t}$, and estimates the counterfactual as the treated unit's extrapolated trend plus a weighted average of donor cycles. Theorem 1 states that as the pre-treatment sample grows, the SBC estimator satisfies $\hat{Y}_{1,t}(0)-Y_{1,t}(0)=\sum_{i=2}^{N+1}\hat{w}_i(\varepsilon_{i,t}-\varepsilon_{1,t})+o_p(1)$, so the idiosyncratic part of the treated unit's cycle is the only asymptotic source of error and the estimator is asymptotically unbiased; without idiosyncratic shocks it is consistent. In the empirical illustrations, the SBC estimator produces counterfactuals that track placebo-period data while conventional synthetic control drifts, and the donor weights for trend and cycle select different countries.

Load-bearing premise

The proof of Theorem 1 assumes a perfect pre-treatment fit in the cycle-matching step: the treated unit's estimated cycle must be reproduced exactly by the donor cycle combination in every pre-treatment period, which is impossible when idiosyncratic shocks are present, and the asymptotic unbiasedness expansion depends on this cancellation.

Editorial extensions

If this is right

  • A good pre-treatment fit by conventional synthetic control on nonstationary outcomes is not evidence that the counterfactual is valid; a placebo test with a shifted event date can expose the spurious drift.
  • The SBC estimator is asymptotically unbiased for each post-treatment counterfactual period as the pre-treatment sample grows, regardless of whether the treated unit's trend is cointegrated with donor trends.
  • When cycles follow an exact low-dimensional factor structure with no idiosyncratic shocks, the SBC estimator is consistent for the counterfactual outcome.
  • In the paper's simulations, SBC cuts post-treatment mean-squared error relative to conventional synthetic control in spurious-trend designs, often by an order of magnitude under non-negative weights, and remains competitive when half the donors are cointegrated with the treated unit.
  • In the German reunification and Hong Kong handover case studies, SBC assigns trend and cycle to different donor groups and yields a larger estimated treatment effect with placebo trajectories that stay close to actual data.

Reading between the lines

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

  • An extension the authors leave implicit is inference: with multiple treated units or longer post-treatment windows, Theorem 1's characterization could be turned into confidence intervals for average treatment effects.
  • The same trend-cycle split could be paired with other one-sided filters or model-based trend forecasts, which would make the approach robust to Hamilton filter misspecification.
  • A testable prediction of the paper is that the gap between SBC and conventional synthetic control widens as the treated unit's nonstationary trend becomes more idiosyncratic relative to the donor pool.
  • The placebo evidence suggests that outcome-only synthetic control applications on trending data should routinely report both trend and cycle weights, since the divergence itself diagnoses spurious matching.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 6 minor

Summary. The paper proposes a synthetic control method for nonstationary panel outcomes, termed the synthetic business cycle (SBC) estimator. The procedure decomposes each unit's outcome into a trend and a cyclical component using the Hamilton filter, forecasts the treated unit's trend from its own history, and uses donor cycles to construct a synthetic cycle. The central theoretical claim (Theorem 1) is that the estimator is asymptotically unbiased for the counterfactual outcome under a factor structure for the cyclical components. The paper also provides Monte Carlo simulations and empirical applications to German reunification and the return of Hong Kong, where SBC is compared with a conventional synthetic control estimator.

Significance. The paper addresses a real and important gap: conventional synthetic control applied to nonstationary macroeconomic outcomes can produce spurious counterfactuals. The divide-and-conquer idea of separating trend and cycle is intuitive and economically motivated, and the simulation evidence is extensive and largely favorable to the proposed method. If the theoretical result were rigorously established, the paper would be a valuable contribution to the causal panel literature. The simulations and the two empirical illustrations are concrete strengths, as is the clear framing of the spurious synthetic control problem. However, the proof of the main theorem contains a substantial gap: it relies on an unstated perfect pre-treatment fit assumption that is not in the theorem's hypotheses and is inconsistent with a lemma proved in the same appendix. As a result, the central theoretical claim is not established in its current form.

major comments (2)
  1. [Appendix A, proof of Lemma A.1] The proof asserts 'we follow Abadie et al. (2010) in assuming a perfect pre-treatment fit, that is, the minimization problem in (3) attains zero.' This perfect-fit assumption is not part of Assumptions 1-2. With continuous idiosyncratic errors and T0 > N, exact equality of the estimated cycles is an event of probability zero. More importantly, Lemma A.1, proved in the same appendix, shows that the OLS weights converge to a limit that is generally not equal to the infeasible factor-matching weights w0. Consequently, the limiting weights do not satisfy Λ'_{-1}w∞ = λ1, and the factor mismatch term (λ1 - Σŵ_iλ_i)'f_t does not vanish in probability. The displayed substitution, which replaces F0λ1' - F0Σw_iλ_i' by idiosyncratic and estimation-error terms, is therefore invalid, and the asymptotic characterization Ŷ1,t(0)-Y1,t(0) = Σŵ_i(ε_i,t - ε_1,t) + op(1) does not follow. The theorem's claim of asymptotic unbiasedness may still be salvageable in a weaker, unconditional form when E[f_t]=0, but the proof as written does not establish the stated result under Assumptions 1-2.
  2. In deriving the probability limit of (1/˜T0)(Ĉ'_{-1})(ε_1 + û_1 - ε_{-1}w0 - û_{-1}w0), the term involving ε'_{-1}ε_{-1}w0 enters with a minus sign; the displayed expression in the manuscript retains a plus sign. Algebraically, ε'_{-1}(-ε_{-1}w0) = -ε'_{-1}ε_{-1}w0, so the limit is -Σε_{-1}w0, not +Σε_{-1}w0. This sign error affects the limiting weights derived in the lemma: the limit is shifted by a term involving Σε and is generally different from w0. The sign error reinforces the inconsistency with the perfect-fit substitution in Theorem 1 and means the lemma's statement, as written, is not correct.
minor comments (6)
  1. [Section 3.1, equation (2)] The coefficient αi2 should be αi,2, and the indexing of the lags is inconsistent: the regressors are Y_{t-h}, ..., Y_{t-h-p+1}, so the coefficient on the j-th lag should be αi,j.
  2. [Section 5.1] The conventional synthetic control comparison is based on a version without the covariates used by Abadie et al. (2015). The text acknowledges this, but the abstract and introduction may be read as directly challenging those estimates; please clarify the scope of the comparison.
  3. [Section 3.1] The method produces forecasts only for horizons up to h; the empirical figures should make clear that only these horizons are shown, or the paper should explain how longer-horizon forecasts would be obtained.
  4. [Section 3.3, Lemma 2] The statement of Lemma 2 is vague about the class of processes for which the projection coefficients are consistent; citing Hamilton (2018) without further justification may be insufficient for the nonstationary cases discussed.
  5. [Section 4, Table 1] The MSE ratios are reported without Monte Carlo standard errors, and some ratios are close to 1; reporting uncertainty in the simulation results would help assess whether small differences are meaningful.
  6. [Section 3.2, just before Theorem 1] The term 'unbiased' is used loosely: the estimator is defined for a fixed post-treatment t and unbiasedness is asymptotic; consider rephrasing to avoid confusion.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the synthetic business cycle counterfactual is an out-of-sample construction from the treated unit's own trend history and donor cycles, not a refit of the target outcome.

full rationale

The claimed derivation is not circular in the sense defined by the review protocol. The synthetic business cycle estimator separates the counterfactual into a trend component, obtained by extrapolating the treated unit's own pre-treatment history through the Hamilton filter (Eq. 2), and a cyclical component, imputed from donor cycles with weights fitted only to pre-treatment cyclical discrepancies (Eq. 3); the post-treatment object Yhat_1,t(0) = tauhat_1,t + chat_1,t is therefore an out-of-sample construction rather than a refit of the target. Theorem 1's asymptotic characterization is derived from explicit assumptions, namely the factor representation of cycles (Assumption 1), high-level filter-accuracy conditions (Assumption 2), and the standard perfect pre-treatment fit assumption attributed to Abadie et al. (2010); none of these inputs is the post-treatment outcome or an equivalent of it. The paper's self-citations (Phillips and Shi 2021; Mei, Phillips, and Shi 2024; Shi and Huang 2023; Hsiao, Shi, and Zhou 2022) are used as background or as discussion pointers, not as the load-bearing justification for the main identification. A technical concern exists in the proof: the perfect pre-treatment fit assumption appears inconsistent with the limiting-weight expression in Lemma A.1 when idiosyncratic errors are continuously distributed; however, that is a correctness and rigor issue, not a circular reduction of the prediction to its inputs. The empirical evaluations, with simulations based on known DGPs and placebo tests relative to external benchmarks, further indicate that the claimed advantages are not manufactured by the construction.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

The method depends on two user-chosen filter settings (h,p), on the factor structure of cycles, on high-level filter accuracy, and on an exact pre-treatment fit in the proof. No new physical entities are introduced.

free parameters (2)
  • Hamilton filter horizon h = 2 in simulations; 4 in empirical applications
    Sets how much of the series is classified as trend versus cycle; no data-driven selection or sensitivity analysis reported.
  • Hamilton filter lag order p = 2 in all exercises
    Number of own lags used for trend extrapolation; chosen following Hamilton (2018), not varied.
assumptions (3)
  • domain assumption Cyclical components follow a common factor structure ci,t = lambda_i' ft + epsilon_i,t with bounded factors and positive definite factor second moments (Assumption 1).
    This is the identifying assumption that makes donor cycles informative about the treated unit's cycle.
  • domain assumption Filter estimation error satisfies Assumption 2: pointwise op(1) and sum of squared errors op(T0).
    High-level assumption on the detrending filter; verified for a random walk example but not for all DGPs.
  • ad hoc to paper Perfect pre-treatment fit in the synthetic control minimization on cycles (Appendix A proof of Theorem 1).
    The proof assumes the minimization in (3) attains zero, which requires the treated cycle to lie exactly in the span of donor cycles during the pre-treatment sample.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Synthetic Business Cycle Approach to Counterfactual Analysis with Nonstationary Macroeconomic Data." pith.science (2026). https://pith.science/paper/QPUDRPWV

@misc{pith2026250522388,
  author       = {Pith},
  title        = {Pith review of: A Synthetic Business Cycle Approach to Counterfactual Analysis with Nonstationary Macroeconomic Data},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QPUDRPWV}},
  note         = {Machine review of arXiv:2505.22388}
}
read the original abstract

This paper investigates the use of synthetic control methods for causal inference in macroeconomic settings when dealing with possibly nonstationary data. While the synthetic control approach has gained popularity for estimating counterfactual outcomes, we caution researchers against assuming a common nonstationary trend factor across units for macroeconomic outcomes, as doing so may result in misleading causal estimation-a pitfall we refer to as the spurious synthetic control problem. To address this issue, we propose a synthetic business cycle framework that explicitly separates trend and cyclical components. By leveraging the treated unit's historical data to forecast its trend and using control units only for cyclical fluctuations, our divide-and-conquer strategy eliminates spurious correlations and improves the robustness of counterfactual prediction in macroeconomic applications. As empirical illustrations, we examine the cases of German reunification and the handover of Hong Kong, demonstrating the advantages of the proposed approach.

Figures

Figures reproduced from arXiv: 2505.22388 by the authors.

Figure 1
Figure 1. Diagram of the procedure The inputs of the learning process is Yi,t for i = 1, 2, . . . , N + 1 and t = 1, . . . , T0. The output is an estimator of the counterfactual (Y1,t(0))T0+h t=T0+1. Objects in squares are nonstationary, whereas those in circles are stationary. The pre-treatment / post-treatment and treated / control combinations partition the plane into four quadrants, with the top-right one for counterfactu… view at source ↗
Figure 2
Figure 2. GDP path decomposition for Germany and donor countries [PITH_FULL_IMAGE:figures/full_fig_p025_2.png] view at source ↗
Figure 3
Figure 3. Treatment effect estimates of German reunification [PITH_FULL_IMAGE:figures/full_fig_p026_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Comparison of synthetic weights Weights assigned to donor countries using three different approaches. SC (grey) represents the synthetic control method applied to the raw GDP series. SC trend (white) and SBC (black) represent synthetic control applied to the decomposed…
Figure 5
Figure 5. Figure 5: Placebo reunification in 1975 This figure applies the synthetic business cycle (SBC) and the conventional synthetic control (SC) method to the placebo reunification dated 1975. Panel (a) shows Germany’s synthetic GDP estimated using both the synthetic business cycle es…
Figure 6
Figure 6. Figure 6: GDP path decomposition for Hong Kong and donor countries [PITH_FULL_IMAGE:figures/full_fig_p030_6.png]
Figure 7
Figure 7. Figure 7: Treatment effect estimates of return of Hong Kong [PITH_FULL_IMAGE:figures/full_fig_p031_7.png]
Figure 8
Figure 8. Figure 8: Comparison of synthetic weights Weights assigned to donor economies using three different approaches. SC (grey) represents the synthetic control method applied to the raw GDP series. SC trend (white) and SBC (black) represent synthetic control applied to the decomposed…
Figure 9
Figure 9. Figure 9: Placebo return of Hong Kong in 1987 This figure applies the synthetic business cycle and the conventional synthetic control method to the placebo return of Hong Kong dated 1987. Panels (a) and (b) impose and relax the non-negativity constraint on the weights, respectiv…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

49 extracted references · 45 canonical work pages

  1. [1]

    Abadie, A. (2021). Using synthetic controls: Feasibility, data requirements, and methodological aspects. Journal of Economic Literature\/ 59\/ (2), 391--425

    work page 2021

  2. [2]

    Diamond, and J

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

    work page 2010

  3. [3]

    Diamond, and J

    Abadie, A., A. Diamond, and J. Hainmueller (2015). Comparative politics and the synthetic control method. American Journal of Political Science\/ 59\/ (2), 495--510

    work page 2015

  4. [4]

    Abadie, A. and J. Gardeazabal (2003). The economic costs of conflict: A case study of the basque country. American Economic Review\/ 93\/ (1), 113--132

    work page 2003

  5. [5]

    Johnson, A

    Acemoglu, D., S. Johnson, A. Kermani, J. Kwak, and T. Mitton (2016). The value of connections in turbulent times: Evidence from the united states. Journal of Financial Economics\/ 121\/ (2), 368--391

    work page 2016

  6. [6]

    Andersson, J. J. (2019). Carbon taxes and CO _2 emissions: Sweden as a case study. American Economic Journal: Economic Policy\/ 11\/ (4), 1--30

    work page 2019

  7. [7]

    Athey, D

    Arkhangelsky, D., S. Athey, D. A. Hirshberg, G. W. Imbens, and S. Wager (2021). Synthetic difference-in-differences. American Economic Review\/ 111\/ (12), 4088--4118

    work page 2021

  8. [8]

    Castell \'o n, and T

    Asatryan, Z., C. Castell \'o n, and T. Stratmann (2018). Balanced budget rules and fiscal outcomes: Evidence from historical constitutions. Journal of Public Economics\/ 167 , 105--119

    work page 2018

Show all 49 references
  1. [9]

    Bayati, N

    Athey, S., M. Bayati, N. Doudchenko, G. Imbens, and K. Khosravi (2021). Matrix completion methods for causal panel data models. Journal of the American Statistical Association\/ 116\/ (536), 1716--1730

    2021

  2. [10]

    Avila-Montealegre, O. and C. Mix (2024). Common trade exposure and business cycle comovement. Journal of International Economics\/ 152 , 103998

    2024

  3. [11]

    Li, and M

    Bai, C., Q. Li, and M. Ouyang (2014). Property taxes and home prices: A tale of two cities. Journal of Econometrics\/ 180\/ (1), 1--15

    2014

  4. [12]

    Bai, J. (2009). Panel data models with interactive fixed effects. Econometrica\/ 77\/ (4), 1229--1279

    2009

  5. [13]

    Bai, J. and S. Ng (2021). Matrix completion, counterfactuals, and factor analysis of missing data. Journal of the American Statistical Association\/ 116\/ (536), 1746--1763

    2021

  6. [14]

    Pfeifer, and K

    Becker, M., G. Pfeifer, and K. Schweikert (2021). Price effects of the austrian fuel price fixing act: a synthetic control study. Energy Economics\/ 97 , 105207

    2021

  7. [15]

    Feller, and J

    Ben-Michael, E., A. Feller, and J. Rothstein (2021). The augmented synthetic control method. Journal of the American Statistical Association\/ 116\/ (536), 1789--1803

    2021

  8. [16]

    Beveridge, S. and C. R. Nelson (1981). A new approach to decomposition of economic time series into permanent and transitory components with particular attention to measurement of the ‘business cycle’. Journal of Monetary Economics\/ 7\/ (2), 151--174

    1981

  9. [17]

    Billmeier, A. and T. Nannicini (2013). Assessing economic liberalization episodes: A synthetic control approach. Review of Economics and Statistics\/ 95\/ (3), 983--1001

    2013

  10. [18]

    Botosaru, I. and B. Ferman (2019). On the role of covariates in the synthetic control method. The Econometrics Journal\/ 22\/ (2), 117--130

    2019

  11. [19]

    Masini, and M

    Carvalho, C., R. Masini, and M. C. Medeiros (2018). Ar C o: An artificial counterfactual approach for high-dimensional panel time-series data. Journal of Econometrics\/ 207\/ (2), 352--380

    2018

  12. [20]

    Garcia, and L

    Chamon, M., M. Garcia, and L. Souza (2017). FX interventions in B razil: A synthetic control approach. Journal of International Economics\/ 108 , 157--168

    2017

  13. [21]

    Di Giovanni, J., A. A. Levchenko, and I. Mejean (2018). The micro origins of international business-cycle comovement. American Economic Review\/ 108\/ (1), 82--108

    2018

  14. [22]

    Doudchenko, N. and G. W. Imbens (2016). Balancing, regression, difference-in-differences and synthetic control methods: A synthesis. Technical report, National Bureau of Economic Research

    2016

  15. [23]

    crown jewel

    Eliason, P. and B. Lutz (2018). Can fiscal rules constrain the size of government? A n analysis of the “crown jewel” of tax and expenditure limitations. Journal of Public Economics\/ 166 , 115--144

    2018

  16. [24]

    Gobillon, L. and T. Magnac (2016). Regional policy evaluation: Interactive fixed effects and synthetic controls. Review of Economics and Statistics\/ 98\/ (3), 535--551

    2016

  17. [25]

    Granger, C. W. and P. Newbold (1974). Spurious regressions in econometrics. Journal of Econometrics\/ 2\/ (2), 111--120

    1974

  18. [26]

    Hamilton, J. D. (2018). Why you should never use the H odrick- P rescott filter. Review of Economics and Statistics\/ 100\/ (5), 831--843

    2018

  19. [27]

    Hodrick, R. J. and E. C. Prescott (1997). Postwar us business cycles: an empirical investigation. Journal of Money, Credit, and Banking\/ 29\/ (1), 1--16

    1997

  20. [28]

    Hou, L., K. Li, Q. Li, and M. Ouyang (2021). Revisiting the location of fdi in china: A panel data approach with heterogeneous shocks. Journal of Econometrics\/ 221\/ (2), 483--509

    2021

  21. [29]

    Hsiao, C., S. H. Ching, and S. K. Wan (2012). A panel data approach for program evaluation: Measuring the benefits of political and economic integration of H ong K ong with mainland C hina. Journal of Applied Econometrics\/ 27\/ (5), 705--740

    2012

  22. [30]

    Shi, and Q

    Hsiao, C., Z. Shi, and Q. Zhou (2022). Transformed estimation for panel interactive effects models. Journal of Business & Economic Statistics\/ 40\/ (4), 1831--1848

    2022

  23. [31]

    Ibragimov, R. and S. Sharakhmetov (2002). The exact constant in the rosenthal inequality for random variables with mean zero. Theory of Probability & Its Applications\/ 46\/ (1), 127--132

    2002

  24. [32]

    Shen, and Q

    Li, X., Y. Shen, and Q. Zhou (2024). Confidence intervals of treatment effects in panel data models with interactive fixed effects. Journal of Econometrics\/ 240\/ (1), 105684

    2024

  25. [33]

    Masini, R. and M. C. Medeiros (2021). Counterfactual analysis with artificial controls: Inference, high dimensions, and nonstationarity. Journal of the American Statistical Association\/ 116\/ (536), 1773--1788

    2021

  26. [34]

    Masini, R. and M. C. Medeiros (2022). Counterfactual analysis and inference with nonstationary data. Journal of Business & Economic Statistics\/ 40\/ (1), 227--239

    2022

  27. [35]

    Mayoral, L. (2013). Heterogeneous dynamics, aggregation, and the persistence of economic shocks. International Economic Review\/ 54\/ (4), 1295--1307

    2013

  28. [36]

    McCracken, M. W. and S. Ng (2016). FRED-MD : A monthly database for macroeconomic research. Journal of Business & Economic Statistics\/ 34\/ (4), 574--589

    2016

  29. [37]

    Mei, Z., P. C. Phillips, and Z. Shi (2024). The boosted H odrick- P rescott filter is more general than you might think. Journal of Applied Econometrics\/ 39\/ (7), 1260--1281

    2024

  30. [38]

    Seidel, F

    Meinhardt, V., B. Seidel, F. Stille, and D. Teichmann (1995). Transferleistungen in die neuen bundesländer und deren wirtschaftliche konsequenzen. Sonderheft 154, Deutsches Institut für Wirtschaftsforschung (DIW Berlin), Berlin

    1995

  31. [39]

    Miao, K., P. C. Phillips, and L. Su (2023). High-dimensional vars with common factors. Journal of Econometrics\/ 233\/ (1), 155--183

    2023

  32. [40]

    Moon, H. R. and M. Weidner (2017). Dynamic linear panel regression models with interactive fixed effects. Econometric Theory\/ 33\/ (1), 158--195

    2017

  33. [41]

    Onatski, A. and C. Wang (2021). Spurious factor analysis. Econometrica\/ 89\/ (2), 591--614

    2021

  34. [42]

    Peri, G. and V. Yasenov (2019). The labor market effects of a refugee wave: Synthetic control method meets the mariel boatlift. Journal of Human Resources\/ 54\/ (2), 267--309

    2019

  35. [43]

    Phillips, P. C. (1986). Understanding spurious regressions in econometrics. Journal of Econometrics\/ 33\/ (3), 311--340

    1986

  36. [44]

    Phillips, P. C. and Z. Shi (2021). Boosting: W hy you can use the HP filter. International Economic Review\/ 62\/ (2), 521--570

    2021

  37. [45]

    Shen, D., P. Ding, J. Sekhon, and B. Yu (2023). Same root different leaves: Time series and cross-sectional methods in panel data. Econometrica\/ 91\/ (6), 2125--2154

    2023

  38. [46]

    Shi, Z. and J. Huang (2023). Forward-selected panel data approach for program evaluation. Journal of Econometrics\/ 234\/ (2), 512--535

    2023

  39. [47]

    Stock, J. H. and M. W. Watson (1988). Variable trends in economic time series. Journal of Economic Perspectives\/ 2\/ (3), 147--174

    1988

  40. [48]

    Stock, J. H. and M. W. Watson (1999). Business cycle fluctuations in US macroeconomic time series. Handbook of Macroeconomics\/ 1 , 3--64

    1999

  41. [49]

    Xiong, R. and M. Pelger (2023). Large dimensional latent factor modeling with missing observations and applications to causal inference. Journal of Econometrics\/ 233\/ (1), 271--301

    2023

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.