Pith. sign in

REVIEW 3 major objections 5 minor 61 references

Recovering latent linkage structures and spillover effects with structural breaks in panel data models

T0 review · 3 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read A two-step Lasso and least-squares procedure recovers latent spillover networks with a breakpoint that is correct with probability approaching one.

desk verdict Solid new machinery for latent spillover networks with breaks, but the main inference theorem rests on an oracle-selection assumption that the proof overstates. read the letter →

arxiv 2501.09517 v1 pith:4UTQOAA7 submitted 2025-01-16 econ.EM

Ryo Okui , Yutao Sun , Wendun Wang This is my paper · ORCID
classification econ.EM MSC 62P2062F1262J07
keywords spillovereffectslatentnetworkstructuralbreakpaneldatahigh-dimensionalparameterdoublemachinelearningLassosuper-consistency
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

This paper claims that a latent, time-varying spillover structure in panel data can be recovered without pre-specifying who influences whom: the linkage matrix is treated as high-dimensional and sparse, and the moment it changes is an unknown breakpoint. The central result is that a two-step estimator—Lasso for a preliminary fit, then an unpenalized least-squares search over candidate break dates—finds the true breakpoint with probability approaching one as the number of units and time periods grow. That super-consistency means downstream estimates of spillover and private effects can be analyzed as if the breakpoint were known. The paper also claims that the private-effect parameter, estimated by double machine learning with a post-double-Lasso step, is $\sqrt{NT}$-consistent and asymptotically normal even though spillover estimates converge more slowly. Empirically, the procedure places a single break in cross-country R&D spillovers in 2009 and finds the spillover network became about 51 percent sparser afterward.

What carries the argument

The load-bearing mechanism is the two-step breakpoint refinement. Step 1 solves a Lasso objective with adaptive weights to get preliminary estimates of all coefficients and a candidate break, and the excess-risk lemma puts these in an $O_p(sD_{NT})$ neighborhood of the truth. Step 2 solves an unpenalized least-squares problem over all candidate break dates, holding the preliminary coefficients fixed, and cross-sectional variation in the pre/post coefficient differences forces the minimizer to be exactly $b_0$ with probability approaching one. For private-effect inference, the mechanism is the orthogonal score constructed from post-double-Lasso estimates of the sparse regressions of $y_{it}$ and $z_{it}$ on $X_{it}(b)$: the additional post-Lasso OLS step removes selection uncertainty, and the oracle selection assumption allows the slower spillover estimation error to be ignored.

What would settle it

Simulate the paper's data-generating process with a true break at $b_0$ and observe the empirical frequency of $\tilde b=b_0$ as $N,T$ grow; Theorem 4.1 predicts the frequency approaches one. If the true break magnitude $m$ in Assumption 2(ii) is made very small while $N,T$ are large, the exact-recovery probability should drop away from one, revealing the boundary of the super-consistency claim.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that a panel regression with an unknown sparse spillover matrix and a common unknown breakpoint can be estimated so that the breakpoint is pinned exactly in the limit. Theorem 4.1 states $P(\tilde{b}=b_0)\to 1$ as $N,T\to\infty$, making the refined breakpoint estimator super-consistent. Theorem 4.2 states that, under Assumptions 1–4, $\sqrt{NT}(\tilde{\delta}-\delta_0)$ converges in distribution to a mean-zero normal law with variance $(E(e^2))^{-1}E(u^2e^2)(E(e^2))^{-1}$, so the private effect can be inferred at the full $\sqrt{NT}$ rate despite the slower spillover estimates. The same asymptotic argument lets spillover-effect estimation proceed conditionally on the true breakpoint, which is what justifies the empirical comparison of the pre-2009 and post-2009 R&D networks.

Load-bearing premise

Assumption 4(iii) requires that the two Lasso selection steps eventually pick exactly the covariates that matter for the auxiliary regressions for every unit—an oracle property that is assumed rather than proved, and without it the $\sqrt{NT}$ private-effect result collapses.

Editorial extensions

If this is right

  • Breakpoint estimation uncertainty can be ignored in practice: confidence regions for spillover and private effects constructed as if $b_0$ were known are asymptotically valid.
  • The private effect is estimable at the full $\sqrt{NT}$ rate with a normal limiting law, so standard tests and confidence intervals apply to it.
  • Researchers can compare the estimated network before and after the break edge by edge, which is how the paper establishes that the R&D network became sparser.
  • The method gives a concrete empirical date—2009—for the shift in cross-country R&D spillovers, consistent with the financial crisis narrative.
  • The same machinery extends to time-varying private effects, multiple breaks, and latent group heterogeneity in private effects.

Reading between the lines

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

  • If the super-consistency behavior carries over to finite samples beyond the reported simulations, the same pipeline could be transplanted to other latent-network panel settings—crime networks, educational spillovers, or trade linkages—whenever a policy shock or crisis may have rewired the network.
  • The $\sqrt{NT}$ inference for the private effect rests on the oracle selection assumption, which is imposed rather than derived; a practical diagnostic that checks whether the selected covariate sets are stable across sample splits would tell an applied user how much to trust the normal approximation.
  • The empirical mechanism could be tested directly: if reduced R&D spending in key European source countries caused the sparser post-2009 network, then re-estimating without those countries should shrink or eliminate the estimated density drop.
  • The paper does not provide an asymptotically nondegenerate confidence set for the breakpoint, only the singleton $\{\tilde b\}$; developing interval inference for $b_0$ would require a separate argument.
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

3 major / 5 minor

Summary. This paper develops a multi-step estimator for panel data models with latent spillover networks and a possible structural break. The method first obtains a preliminary breakpoint and coefficients via adaptive Lasso, refines the breakpoint by least squares, and then estimates the private effect using a double machine learning procedure with double Lasso and post-Lasso, while re-estimating spillover effects by post-Lasso. The main theoretical results are super-consistency of the refined breakpoint (Theorem 4.1) and √(NT)-consistency and asymptotic normality of the DML private-effect estimator (Theorem 4.2). An empirical application to 24 OECD countries finds a single break in R&D spillover effects in 2009 and a sparser spillover network after the break.

Significance. The paper addresses an important and timely problem: detecting structural breaks in high-dimensional spillover parameters with a latent network, and it combines tools from change-point analysis, Lasso, and DML in a novel way. The super-consistency proof in Theorem 4.1 is detailed and appears to follow standard change-point arguments, and the simulation study is reasonably comprehensive. However, the central theoretical contribution for the private-effect estimator is compromised by an internal gap between Assumption 4(iii) and the proof of Theorem 4.2, which requires exact variable selection rather than the superset property actually assumed. The empirical headline about sparser R&D spillovers is plausible but is presented without uncertainty quantification. If the theoretical gap can be closed (e.g., by strengthening the selection assumption and acknowledging the loss of generality) and the empirical claims are appropriately qualified, the paper would be a useful contribution to the high-dimensional panel break literature.

major comments (3)
  1. [Section 4.2 / Appendix C] Assumption 4(iii) only requires the adaptive-Lasso selected sets to equal fixed supersets s*_i of the true active sets with probability approaching one, and explicitly disclaims variable-selection consistency; however, the proof of Theorem 4.2 in Appendix C asserts 'P(ˆsi = s0_i, ∀i) → 1 ... by Assumptions 4(ii) and 4(iii)' and then replaces the selected sets by the true sets s0_i. This equality is not a consequence of Assumption 4(iii). When redundant variables are present, the post-Lasso OLS estimators in (3.4) contain extra components, and the proof's variance bounds (e.g., the O_p(s/√{NT}) terms after (C.2)) rely on |s0_i| being bounded rather than on |s*_i|; without a bound on sup_i|s*_i| and an explicit control of redundant-variable contributions, the √{NT}-consistency and normal limit in Theorem 4.2 are not established. This also contradicts the claim in Remark 3 that the result depends on an oracle property, since Assumption 4(iii) is weaker than exact selection. The gap is fixable in principle, but it is load-bearing for the paper's central DML result.
  2. [Section 4.2, Assumption 4(i)] The errors are assumed i.i.d. across time and units and independent of X_it(b), whereas Assumption 1 allows strong mixing and contemporaneous exogeneity. The proof of Theorem 4.2 relies on serial independence in several places (e.g., the cross-fitting independence for zero-mean/variance computations and the martingale-type CLT for the sum of u_it e_it). Since the paper motivates the DML for panel time series with possibly dependent data, this strong condition substantially narrows the scope of the main inference result. The text says that some conditions are stronger than necessary and can be relaxed, but it does not indicate how Assumption 4(i) could be relaxed; please either provide such a relaxation or state clearly that the √(NT) result currently applies only under full independence.
  3. [Section 7 / Abstract] The central empirical claim that the R&D spillover network 'becomes sparser' after the 2009 break, with a roughly 51% reduction in network density, is based solely on point estimates of the thresholded/post-Lasso adjacency matrix. No standard errors, confidence sets, or robustness to tuning parameters are reported for the density or for the individual link indicators. Given the admitted non-uniformity of the oracle property in Remark 3, the observed sparsening could in principle be a selection artifact. To support the headline, please provide uncertainty quantification (e.g., subsampling or bootstrap over the selection step, or sensitivity analysis across tuning parameters) and/or downgrade the claim to a descriptive finding.
minor comments (5)
  1. [Section 7] The text lists 'six countries, namely France, Italy, Canada, the Netherlands, and Spain' but names only five countries; also 'Japanese, Korean' should be 'Japan, Korea'.
  2. [Assumption 1(i)] The notation E[(... /N)^2] < M/N is ambiguous; it should be stated as the mean square of the cross-sectional average being O(1/N), or equivalently as a bound on the variance of that average.
  3. [Table 6] The table reports p-values and mentions 'weakly significant estimates (at 15%)' without justifying the choice of a 15% significance level; conventional thresholds such as 5% or 10% would be more standard.
  4. [General] The paper does not include a data availability statement or replication code, which would be important for an empirical economics readership and would also help verify the empirical findings.
  5. [Section 3 / Remark 2] The distinction between the proposed 'double Lasso' and the 'double selection' procedure of Belloni et al. (2014) could be made clearer; the post-Lasso step is described, but its role in eliminating selection uncertainty is only fully explained in the proof of Theorem 4.2.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; asymptotic results are proved from explicit assumptions and the self-citations are independent technical lemmas.

full rationale

The paper's central derivations do not reduce to their own inputs. Theorem 4.1 (super-consistency of the refined breakpoint estimator) is proved in Appendix B from Assumptions 1-2 via excess-risk bounds and maximal inequalities; the conclusion P(tilde-b = b0) -> 1 is not assumed or renamed from an input. Theorem 4.2 is a conditional result: under Assumption 4(iii), which explicitly allows superset selection ('the selected sets need not exactly match the true set of nonzero coefficients'), plus the i.i.d. and sparsity conditions, the proof derives the sqrt(NT)-normal limit. The proof later replaces hat-s_i with s0_i; the paper itself acknowledges reliance on an 'oracle property' in Remark 3 and states the result is not uniform. This is a gap between an imposed high-level condition and the proof's event, not a circular step in the sense of using the target result as a premise. The appendix's DML proof treats the nuisance estimation error explicitly, bounding the non-leading terms and identifying the u_it*e_it term as the source of the asymptotic variance; the variance formula is not inserted as an input. The empirical breakpoint of 2009 and the roughly 51% drop in network density are estimation outputs from the data, not fitted quantities relabeled as predictions. Self-citations occur (Dzemski and Okui 2024 for concentration lemmas; Okui and Wang 2021 for grouped panel break methods; Lumsdaine, Okui, and Wang 2023 for panel break techniques), but they support technical steps or extensions and do not carry the load of the main theorems. The cited lemmas are standard maximal inequalities stated in Appendix D with explicit mixing and tail assumptions that do not include the theorems' conclusions, and the same estimates would be needed under any authorship. No pattern of self-definitional reasoning, fitted inputs called predictions, or uniqueness imported from prior work by the same authors is present. The identified oracle-selection issue is better classified as a correctness or robustness concern than as circularity.

Assumptions & free parameters 3 free parameters · 8 assumptions · 0 invented entities

The central claims rest on standard high-dimensional Lasso assumptions plus an imposed oracle selection property. No new physical or latent entity is introduced beyond the unknown spillover coefficients, which are the model's parameters of interest.

free parameters (3)
  • Tuning parameters lambda_NT,B and lambda_NT,A = not reported
    Penalty strengths in the Lasso steps are chosen by the researcher; asymptotic theory gives only rate conditions, but finite-sample breakpoint and network estimates depend on the penalty magnitude.
  • Adaptive weights phi_ij,B and phi_ij,A = not reported
    Weights are data-dependent and several constructions are described (heteroscedastic, HAC), but no single implementation is fixed, which affects variable selection in practice.
  • Number of groups G in heterogeneous private-effect extension = G=1 by BIC in application
    G is selected by information criterion in Section 7; it is not a parameter of the benchmark model but is tuned in the empirical analysis.
assumptions (8)
  • domain assumption Exogeneity E[W_it(b) u_it] = 0
    Assumption 1(i) rules out endogeneity and limits cross-sectional correlation of the error term.
  • domain assumption Thin tails and strong mixing for covariates and errors
    Assumptions 1(ii)-(iii) supply concentration inequalities used throughout the proofs.
  • domain assumption Uniform eigenvalue bound on W_it W_it'
    Assumption 1(v) prevents any covariate element from diverging as dimension grows.
  • domain assumption Break signal m > 0 and break not near sample boundaries
    Assumption 2 establishes that the break is identifiable and that both regimes have enough observations.
  • domain assumption Sparsity and compatibility conditions
    Assumption 3 requires s >= sqrt(N), s D_NT N^c -> 0, and a restricted eigenvalue or compatibility condition for the Lasso.
  • ad hoc to paper Perfect variable selection by the double Lasso
    Assumption 4(iii) imposes selection consistency with probability approaching one. It is not derived for the adaptive Lasso in this panel setting, and Theorem 4.2's root-NT rate depends on it.
  • domain assumption i.i.d. errors independent of covariates
    Assumption 4(i) is used for the central limit theorem in Theorem 4.2 and is stronger than the mixing assumptions used elsewhere.
  • standard math Background lemmas from Dzemski and Okui (2024) and Bai and Perron (1998)
    Published concentration inequalities and change-point lemmas are cited and used in the proofs of Lemmas 4.1, A.1, A.2, and Theorem 4.1.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Recovering latent linkage structures and spillover effects with structural breaks in panel data models." pith.science (2026). https://pith.science/paper/4UTQOAA7

@misc{pith2026250109517,
  author       = {Pith},
  title        = {Pith review of: Recovering latent linkage structures and spillover effects with structural breaks in panel data models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4UTQOAA7}},
  note         = {Machine review of arXiv:2501.09517}
}
read the original abstract

This paper introduces a framework to analyze time-varying spillover effects in panel data. We consider panel models where a unit's outcome depends not only on its own characteristics (private effects) but also on the characteristics of other units (spillover effects). The linkage of units is allowed to be latent and may shift at an unknown breakpoint. We propose a novel procedure to estimate the breakpoint, linkage structure, spillover and private effects. We address the high-dimensionality of spillover effect parameters using penalized estimation, and estimate the breakpoint with refinement. We establish the super-consistency of the breakpoint estimator, ensuring that inferences about other parameters can proceed as if the breakpoint were known. The private effect parameters are estimated using a double machine learning method. The proposed method is applied to estimate the cross-country R&D spillovers, and we find that the R&D spillovers become sparser after the financial crisis.

Figures

Figures reproduced from arXiv: 2501.09517 by the authors.

Figure 1
Figure 1. Information criterion to determine the number of breaks and groups [PITH_FULL_IMAGE:figures/full_fig_p028_1.png] view at source ↗
Figure 2
Figure 2. Heatmap of R&D spillover in the two regimes [PITH_FULL_IMAGE:figures/full_fig_p029_2.png] view at source ↗
Figure 3
Figure 3. Heatmap of R&D spillover in the two regimes [PITH_FULL_IMAGE:figures/full_fig_p066_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

61 extracted references · 60 canonical work pages

  1. [1]

    Babina, A

    T. Babina, A. Bernstein, and F. Mezzanotti. Financial disruptions and the organization of innovation: Evidence from the great depression. Review of Financial Studies, 36: 0 4271--4317, 2023

    work page 2023

  2. [2]

    J. Bai. Common breaks in means and variances for panel data. Journal of Econometrics, 157: 0 78--92, 2010

    work page 2010

  3. [3]

    Bai and P

    J. Bai and P. Perron. Estimating and testing linear models with multiple structural changes. Econometrica, 66: 0 47--78, 1998

    work page 1998

  4. [4]

    B. H. Baltagi, Q. Feng, and C. Kao. Estimation of heterogeneous panels with structural breaks. Journal of Econometrics, 191: 0 176--195, 2016

    work page 2016

  5. [5]

    B. H. Baltagi, C. Kao, and L. Liu. Estimation and identification of change points in panel models with nonstationary or stationary regressors and error term. Econometric Reviews, 36: 0 85--102, 2017

    work page 2017

  6. [6]

    Belloni, V

    A. Belloni, V. Chernozhukov, and C. Hansen. Inference on treatment effects after selection among high-dimensional controls. The Review of Economic Studies, 81: 0 608--650, 2014

    work page 2014

  7. [7]

    Bonhomme and E

    S. Bonhomme and E. Manresa. Grouped patterns of heterogeneity in panel data. Econometrica, 83: 0 1147--1184, 2015

    work page 2015

  8. [8]

    Bonhomme, T

    S. Bonhomme, T. Lamadon, and E. Manresa. Discretizing unobserved heterogeneity. Econometrica, 90: 0 625--643, 2022

    work page 2022

Show all 61 references
  1. [9]

    Campello, J

    M. Campello, J. R. Graham, and C. R. Harvey. The real effects of financial constraints: Evidence from a financial crisis. Journal of Financial Economics, 97: 0 470--487, 2010

    2010

  2. [10]

    J. Chen, D. Li, Y. Li, and O. Linton. Estimating time-varying networks for high-dimensional time series. Journal of Econometrics, forthcoming, 2024

    2024

  3. [11]

    Chernozhukov, D

    V. Chernozhukov, D. Chetverikov, M. Demirer, E. Duflo, C. Hansen, W. Newey, and J. Robins. Double/debiased machine learning for treatment and structural parameters. The Econometrics Journal, 21: 0 C1--C68, 2018

    2018

  4. [12]

    Chernozhukov, W

    V. Chernozhukov, W. K. H\" a rdle, C. Huang, and W. Wang. LASSO -driven inference in time and space. Annals of Statistics, 49: 0 1702--1735, 2021

    2021

  5. [13]

    D. T. Coe and E. Helpman. International R & D spillovers. European Economic Review, 39: 0 859--887, 1995

    1995

  6. [14]

    D. T. Coe, E. Helpman, and A. W. Hoffmaister. International R&D spillovers and institutions. European Economic Review, 53: 0 723--741, 2009

    2009

  7. [15]

    Comola and S

    M. Comola and S. Prina. Treatment Effect Accounting for Network Changes . The Review of Economics and Statistics, 103: 0 597--604, 2021

    2021

  8. [16]

    de Paula , I

    A. de Paula , I. Rasul, and P. Souza. Identifying network ties from panel data: Theory and an application to tax competition. Review of Economic Studies, forthcoming, 2024

    2024

  9. [17]

    Dzemski and R

    A. Dzemski and R. Okui. Confidence set for group membership. Quantitative Economics, 15: 0 245--277, 2024

    2024

  10. [18]

    Ertur and W

    C. Ertur and W. Koch. Growth, technological interdependence and spatial externalities: Theory and evidence. Journal of Applied Econometrics, 22: 0 1033--1062, 2007

    2007

  11. [19]

    Ertur and A

    C. Ertur and A. Musolesi. Weak and strong cross-sectional dependence: A panel data analysis of international technology diffusion. Journal of Applied Econometrics, 32: 0 477--503, 2017

    2017

  12. [20]

    Fan and R

    J. Fan and R. Li. Variable selection via nonconcave penalized likelihood and its oracle properties. Journal of the American Statistical Association, 96 0 (456): 0 1348--1360, 2001

    2001

  13. [21]

    Goldsmith-Pinkham and G

    P. Goldsmith-Pinkham and G. W. Imbens. Social networks and the identification of peer effects. Journal of Business & Economic Statistics, 31: 0 253--264, 2013

    2013

  14. [22]

    Hahn and H

    J. Hahn and H. R. Moon. Panel data models with finite number of multiple equilibria. Econometric Theory, 26: 0 863--881, 2010

    2010

  15. [23]

    H \'a jek and A

    J. H \'a jek and A. R \'e nyi. Generalization of an inequality of K olmogorov. Acta Mathematica Hungarica, 6 0 (3-4): 0 281--283, 1955

    1955

  16. [24]

    Han, C.-S

    X. Han, C.-S. Hsieh, and S. I. M. Ko. Spatial modeling approach for dynamic network formation and interactions. Journal of Business & Economic Statistics, 39: 0 120--135, 2021

    2021

  17. [25]

    E. A. Hanushek and D. D. Kimko. Schooling, labor-force quality, and the growth of nations. American Economic Review, 90: 0 1184--1208, 2000

    2000

  18. [26]

    Hardy and C

    B. Hardy and C. Sever. Financial crises and innovation. European Economic Review, 138: 0 103856, 2021

    2021

  19. [27]

    Hardy, R

    M. Hardy, R. M. Heath, W. Lee, and T. H. McCormick. Estimating spillovers using imprecisely measured networks. arXiv:1904.00136v3, 2020

    1904 arXiv

  20. [28]

    C. P. Himmelberg and B. C. Petersen. R & D and internal finance: A panel study of small firms in high-tech industries. The Review of Economics and Statistics, 76: 0 38--51, 1994

    1994

  21. [29]

    Hsieh and L

    C.-S. Hsieh and L. F. Lee. A social interactions model with endogenous friendship formation and selectivity. Journal of Applied Econometrics, 31: 0 301--319, 2016

    2016

  22. [30]

    Hsieh, L.-F

    C.-S. Hsieh, L.-F. Lee, and V. Boucher. Specification and estimation of network formation and network interaction models with the exponential probability distribution. Quantitative Economics, 11: 0 1349--1390, 2020

    2020

  23. [31]

    N. Islam. Growth empirics: A panel data approach. The Quarterly Journal of Economics, 110: 0 1127--1170, 1995

    1995

  24. [32]

    W. Keller. International technology diffusion. Journal of Economic Literature, 42: 0 752--782, September 2004

    2004

  25. [33]

    Keller and S

    W. Keller and S. R. Yeaple. The gravity of knowledge. American Economic Review, 103: 0 1414--44, 2013

    2013

  26. [34]

    Kolar, L

    M. Kolar, L. Song, A. Ahmed, and E. P. Xing. Estimating time-varying networks. The Annals of Applied Statistics, 4: 0 94 --123, 2010

    2010

  27. [35]

    S. Lee, M. H. Seo, and Y. Shin. The lasso for high dimensional regression with a possible change point. Journal of the Royal Statistical Society. Series B (Statistical Methodology), 78: 0 193--210, 2016

    2016

  28. [36]

    S. Lee, Y. Liao, M. H. Seo, and Y. Shin. Oracle estimation of a change point in high-dimensional quantile regression. Journal of the American Statistical Association, 113: 0 1184--1194, 2018

    2018

  29. [37]

    Lewbel, X

    A. Lewbel, X. Qu, and X. Tang. Social networks with unobserved links. Journal of Political Economy, 131: 0 898--946, 2023

    2023

  30. [38]

    D. Li, J. Qian, and L. Su. Panel data models with interactive fixed effects and multiple structural breaks. Journal of the American Statistical Association, 111: 0 1804--1819, 2016

    2016

  31. [39]

    Y.-N. Li, D. Li, and P. Fryzlewicz. Detection of multiple structural breaks in large covariance matrices. Journal of Business & Economic Statistics, 41: 0 846--861, 2023

    2023

  32. [40]

    R. L. Lumsdaine, R. Okui, and W. Wang. Estimation of panel group structure models with structural breaks in group memberships and coefficients. Journal of Econometrics, 223: 0 45--65, 2023

    2023

  33. [41]

    E. Manresa. Estimating the structure of social interactions using panel data. Working paper, 2016

    2016

  34. [42]

    S. M. Miller and M. P. Upadhyay. The effects of openness, trade orientation, and human capital on total factor productivity. Journal of Development Economics, 63: 0 399--423, 2000

    2000

  35. [43]

    Musolesi

    A. Musolesi. Basic stocks of knowledge and productivity: Further evidence from the hierarchical B ayes estimator. Economics Letters, 95: 0 54--59, 2007

    2007

  36. [44]

    OECD Science, Technology and Industry Scoreboard 2009

    OECD. OECD Science, Technology and Industry Scoreboard 2009. OECD Publishing, Paris, 2009

    2009

  37. [45]

    Trade and Economic Effects of Responses to the Economic Crisis, OECD Trade Policy Studies

    OECD . Trade and Economic Effects of Responses to the Economic Crisis, OECD Trade Policy Studies. OECD Publishing, Paris, 2010 a

    2010

  38. [46]

    OECD Employment Outlook 2010

    OECD . OECD Employment Outlook 2010. OECD Publishing, Paris, 2010 b

    2010

  39. [47]

    OECD Science, Technology and Industry Outlook 2012

    OECD . OECD Science, Technology and Industry Outlook 2012. OECD Publishing, Paris, 2012

    2012

  40. [48]

    Okui and W

    R. Okui and W. Wang. Heterogeneous structural breaks in panel data models. Journal of Econometrics, 220: 0 447--473, 2021

    2021

  41. [49]

    J. H. Park and Y. Sohn. Detecting structural changes in longitudinal network data. Bayesian Analysis, 15: 0 133--157, 2020

    2020

  42. [50]

    Peia and D

    O. Peia and D. Romelli. Did financial frictions stifle R&D investment in E urope during the great recession? Journal of International Money and Finance, 120: 0 102263, 2022

    2022

  43. [51]

    B. v. P. d. l. Potterie and F. Lichtenberg. Does foreign direct investment transfer technology across borders? The Review of Economics and Statistics, 83: 0 490--497, 2001

    2001

  44. [52]

    Pritchett

    L. Pritchett. Where has all the education gone? The World Bank Economic Review, 15: 0 367--391, 2001

    2001

  45. [53]

    Psacharopoulos and H

    G. Psacharopoulos and H. A. Patrinos. Returns to investment in education: A decennial review of the global literature. Education Economics, 26: 0 445--458, 2018

    2018

  46. [54]

    Qian and L

    J. Qian and L. Su. Shrinkage estimation of common breaks in panel data models via adaptive group fused lasso. Journal of Econometrics, 191: 0 86--109, 2016

    2016

  47. [55]

    P. M. Romer. Endogenous technological change. Journal of Political Economy, 98: 0 S71--S102, 1990

    1990

  48. [56]

    Safikhani and A

    A. Safikhani and A. Shojaie. Joint structural break detection and parameter estimation in high-dimensional nonstationary VAR models. Journal of the American Statistical Association, 117: 0 251--264, 2022

    2022

  49. [57]

    Vandenbussche, P

    J. Vandenbussche, P. Aghion, and C. Meghir. Growth, distance to frontier and composition of human capital. Journal of Economic Growth, 11: 0 97--127, 2006

    2006

  50. [58]

    D. Wang, Y. Yu, and A. Rinaldo. Optimal change point detection and localization in sparse dynamic networks. The Annals of Statistics, 49: 0 203--232, 2021

    2021

  51. [59]

    Wang and L

    W. Wang and L. Su. Identifying latent group structures in nonlinear panels. Journal of Econometrics, 220: 0 272--295, 2021

    2021

  52. [60]

    X. Zhu, R. Pan, G. Li, Y. Liu, and H. Wang. Network vector autoregression. The Annals of Statistics, 45: 0 1096--1123, 2017

    2017

  53. [61]

    H. Zou. Adaptive lasso and its oracle properties. Journal of the American Statistical Association, 101 0 (476): 0 1418--1429, 2006

    2006

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.