Pith. sign in

REVIEW 5 minor 37 references

Moment Restrictions for Nonlinear Panel Data Models with Feedback

T0 review · 0 major / 5 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 All moment restrictions that tolerate panel feedback, characterized

desk verdict Complete characterization of feedback-robust moments in nonlinear panels, with a worked MPH application—the real deal, send it out. read the letter →

arxiv 2506.12569 v2 pith:R7KSQB57 submitted 2025-06-14 econ.EM stat.ME

Stéphane Bonhomme , Kevin Dano , Bryan S. Graham This is my paper · ORCID
classification econ.EMstat.ME MSC 62P2062F1262G05
keywords sequentialexogeneityfeedbackpaneldatadurationmodelsincidentalparameterssemiparametricefficiencyboundsmomentconditionsmixedproportionalhazards
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 asks which moment functions of an observed panel can estimate the parameter $\theta$ when the only structural assumption is a parametric outcome density $f_\theta(y_t|y_{t-1},x_t,a)$, while the feedback process, the distribution of unobserved heterogeneity, and the initial condition are allowed to be arbitrary. It proves a complete characterization: a moment function is valid for every such nuisance specification if and only if it satisfies one integration condition against the outcome density (heterogeneity robustness) and, for each time $s$, a condition that the integral over outcomes from $s$ onward does not depend on future covariates (feedback robustness). Because strict exogeneity is a special case, every feedback-robust moment also works under strict exogeneity, and the characterization shows exactly what is lost by allowing feedback. The same machinery provides moment conditions for average effects and the semiparametric efficiency bound, making the precision loss from accommodating feedback a computable quantity.

What carries the argument

The load-bearing object is the pair of conditions (9) and (10) in Theorem 2.1. Condition (9) is the heterogeneity-robustness condition: $\int \phi_\theta(y^T,x^T)\prod_{t=1}^T f_\theta(y_t|y_{t-1},x_t,a)\,dy_{1:T}=0$, inherited from functional differencing. Condition (10) is the feedback-robustness condition: for each $s=2,\ldots,T$, the integral $\int \phi_\theta(y^T,x^T)\prod_{t=s}^T f_\theta(y_t|y_{t-1},x_t,a)\,dy_{s:T}$ does not depend on $x_{s:T}$. Together they make the nuisance factors $g$, $\pi$, and $\nu$ cancel from the expectation by iterated expectations, and they yield the Fredholm-equation representation of Corollary 2.1.1, in which $E[\phi_\theta(Y^T,X^T) | Y^{T-1},X^T,A]$ decomposes into a sum of functions with conditional martingale-difference structure. In the mixed proportional hazards model, the same representation is re-expressed through the ratio transformation $(\tilde P_{\theta,t}, \bar P_\theta)$ of equation (20), which converts the exponential structure into Beta and Gamma components and makes the full set of FHR moments explicit.

What would settle it

In the two-period binary choice logit model with continuous heterogeneity and sequential exogeneity, write out condition (10) and solve for $\phi_\theta$: the theorem implies the only solution satisfying both (9) and (10) is the zero function, so exhibiting any nonzero $\phi_\theta$ that satisfies both conditions in this model would refute Theorem 2.1.

Watch

Extended reading notes

Core claim

The central discovery is Theorem 2.1: under mild regularity, a function $\phi_\theta$ of the full observed history satisfies $E_{\theta,\omega}[\phi_\theta(Y^T,X^T)] = 0$ for every allowable feedback process, heterogeneity distribution, and initial condition if and only if it satisfies conditions (9) and (10). Condition (9) says that integrating $\phi_\theta$ against the product of parametric outcome densities over all outcome periods gives zero, which removes the unobserved heterogeneity. Condition (10) says that, for each $s = 2,\ldots,T$, the integral of $\phi_\theta$ over $y_{s:T}$ against the densities $f_\theta(y_t|y_{t-1},x_t,a)$ is independent of $x_{s:T}$, which removes the feedback channel. Theorem 4.1 then shows that this set of feedback and heterogeneity robust (FHR) moment functions is exactly the orthocomplement of the nuisance tangent set, so efficient scores are projections of the score for $\theta$ onto the same set. In the multi-spell mixed proportional hazards model, the conditions collapse to an explicit characterization in terms of forward orthogonal deviations and the between variation, and the paper derives efficient score functions for parameters and for average structural hazards.

Load-bearing premise

The paper's construction assumes the parametric outcome density $f_\theta(y_t|y_{t-1},x_t,a)$ is correctly specified; if the true outcome process lies outside that family, equations (9) and (10) no longer describe valid estimating equations and the moments target a pseudo-true value.

Editorial extensions

If this is right

  • Every FHR moment is also valid when strict exogeneity holds, so the characterization nests and extends the functional-differencing moments of the strict-exogeneity case.
  • Because the FHR set is the orthocomplement of the nuisance tangent set, semiparametric efficiency bounds under feedback follow directly, turning the information loss from allowing feedback into a computable number.
  • In the multi-spell mixed proportional hazards model, all FHR moment functions are enumerated; the efficient score for $\beta$ and the efficient moment for the average structural hazard are explicit, and a plug-in estimator using an efficient $\theta$ estimate attains the bound for the average structural hazard.
  • Locally efficient estimators based on arbitrary working models for feedback and heterogeneity remain consistent even when the working models are misspecified, and they attain the efficiency bound when a working model is correct.
  • Some models, such as binary logit and Gaussian random coefficients with continuous heterogeneity and sequential exogeneity, admit no nontrivial FHR moment functions, so $\theta$ cannot be regularly estimated at the $\sqrt{N}$ rate in those settings.

Reading between the lines

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

  • The same pair of conditions could be used as a numerical oracle test in any new model: verify whether nonzero solutions of (9) and (10) exist before attempting GMM estimation, which would flag identification failures of the binary-logit type before any estimation is run.
  • For models where the integral equation is not invertible in closed form, the paper's regularized-inverse strategy suggests a general recipe: choose mean-zero functions $\psi_{\theta,t}$, apply spectral or Tikhonov regularization to solve (14), and study the bias-variance tradeoff as the regularization parameter shrinks; this is an extension the paper only sketches.
  • The numeric MPH comparison suggests an operational rule for practice: report both the strict-exogeneity and the feedback efficiency bounds, since the losses appear concentrated on regressor coefficients rather than on duration-dependence parameters, and use a locally efficient score only for the coefficients whose bounds differ materially.
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

0 major / 5 minor

Summary. This paper studies nonlinear panel data models in which the conditional outcome density fθ(yt|yt−1,xt,a) is parametric but the feedback process, heterogeneity distribution, and initial condition are unrestricted. It defines feedback and heterogeneity robust (FHR) moment functions as functions of the observables with zero mean under every admissible nuisance specification. Theorem 2.1 characterizes the complete set of FHR moments by conditions (9) and (10): a heterogeneity condition (9) together with a sequence of feedback-robustness conditions (10). Theorem 3.1 gives the analogous characterization for moment functions identifying average effects, Theorem 4.1 shows that the orthocomplement of the nuisance tangent set coincides with the FHR moments, and Theorem 4.2 develops locally efficient estimation based on working models. The results are specialized to the multi-spell mixed proportional hazards (MPH) model, for which Lemma 2.2 provides a complete characterization of FHR moments in terms of forward orthogonal deviations and the between variation, and Section 5 derives efficiency bounds and reports numerical comparisons of asymptotic standard errors.

Significance. If the results hold, they constitute a substantial contribution to nonlinear panel data econometrics: a complete, constructive characterization of moment conditions that are robust to heterogeneous feedback, efficiency bounds that quantify the cost of allowing feedback, and a novel MPH analysis. The central theorem is supported by a clean iterated-expectations sufficiency argument and a necessity argument based on quadratic-mean differentiability and score orthogonality, with explicit regularity conditions. The MPH derivations are lengthy but internally consistent, and the numerical experiments illustrate the practical efficiency trade-offs. The paper is careful to state the semiparametric scope: all results are within the model where fθ is correctly specified, and the working-model construction in Section 4.2 is clearly distinguished from parametric random-effects maximum likelihood.

minor comments (5)
  1. [Appendix A.2] In the proof of Corollary 2.1.1, the density in the display following equation (13) is written as fθ(yT | yT−1, xT−1, a); it should be fθ(yT | yT−1, xT, a).
  2. [Supplemental Appendix B.1] In the proof of Lemma A.1, after integrating over y_{T−1:T} and x_T, the conditioning set in the displayed conditional expectation is written as Y^{T−1}=y^{T−1}; based on the preceding line it should be Y^{T−2}=y^{T−2}.
  3. [Lemma 2.2] The statement says 'for t = 1, ..., T' but the summation runs to T−1; it should be 'for t = 1, ..., T−1'. It would also help to note explicitly that the conditioning set excludes eP_t, so the zero-mean condition is over the independent increment eP_t.
  4. [Section 5.2] The working heterogeneity density eπ(v)=1/v is improper and therefore is not an element of Ω as defined in Section 1.2; the authors should clarify that it is used as a limiting working model and that the resulting locally efficient moment functions are verified directly to satisfy the FHR conditions.
  5. [Section 1.2] The paper would benefit from an explicit remark that all results are conditional on correct specification of the parametric density fθ; under misspecification, conditions (9)-(10) define moments for a pseudo-true value rather than for the true θ.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the FHR characterization is derived from the defining zero-mean property, and the MPH/efficiency results rest on external structural facts and explicit working models, not on the conclusions.

full rationale

The paper's central derivation is self-contained rather than circular. Theorem 2.1 defines FHR moments by E_{θ,ω}[φ]=0 for all nuisances ω and then proves, without assuming the conclusion, that this property is equivalent to conditions (9) and (10). Sufficiency follows by iterated expectations: condition (10) is used to remove dependence on future regressors, and condition (9) integrates out the latent heterogeneity; no equation is assumed equal to the target result. Necessity uses quadratic-mean differentiability and score functions, with Lemma A.1 recovering (9)–(10) by choosing mean-zero nuisance scores that span the relevant conditional L2 spaces; this is a standard spanning argument, not a re-importation of the conclusion. Theorem 4.1 identifies the orthocomplement of the nuisance tangent set with the FHR moments via the same Lemma A.1, so the efficiency analysis builds on, rather than presupposes, the characterization. The MPH results (Lemma 2.2 and the efficiency scores) use the exponential/Helmert structure of the MPH model, which is imported from external work (Lancaster, Hahn, Ridder–Woutersen) and then derived in the supplement; the derivation does not rely on the existence of FHR moments to establish those distributional facts. The locally efficient estimators in Theorem 4.2 are valid because the working-model scores are elements of the orthocomplement by construction, and their consistency under misspecification follows from Theorem 2.1; no fitted parameter is relabeled as a prediction. The numerical experiments fix θ0 and use Monte Carlo integration to approximate asymptotic standard errors, so no estimation from data is disguised as a forecast. The paper's self-citations (Bonhomme 2012; Bonhomme et al. 2023; Dano 2023) are contextual or extensions, not load-bearing premises: the central characterization does not reduce to any cited result. The explicit reliance on correct specification of fθ is a scope condition of the semiparametric model, not an internal circular step.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

No free parameters are fitted in the theoretical derivation; the simulation parameters in Section 5.2 are fixed DGP inputs, not estimated. The axioms are the standard semiparametric modeling assumptions plus quadratic mean differentiability for necessity. The paper introduces no new particles or forces; it introduces new mathematical objects (forward orthogonal deviations for MPH, new moment functions) but these are constructions, not unexplained entities.

assumptions (4)
  • domain assumption The observed outcome density is correctly specified as fθ(yt|yt-1,xt,a) and is the only prior restriction on the data generating process.
    Section 1.1, equation (1). All FHR moment conditions are defined relative to this parametric family; misspecification invalidates the moment restrictions.
  • domain assumption Unobserved heterogeneity A, feedback process g, and initial condition ν are unrestricted but have a common, known support.
    Section 1.2, the paragraph defining Ω. The common support makes the zero-mean property well-defined across all ω.
  • standard math For necessity results, the root density ω ↦ ℓ^{1/2}(θ,ω|y^T,x^T) is differentiable in quadratic mean at some ω* and moment functions are square-integrable in a neighborhood.
    Theorem 2.1 Part (B) and Theorem 3.1 Part (B). Standard regularity in semiparametric estimation, citing van der Vaart (2000) and Newey-McFadden (1994).
  • domain assumption Latent heterogeneity A is time-invariant and enters the outcome density additively in the MPH and Poisson examples; the general characterization allows multidimensional A but the negative examples show this matters.
    Equation (1) and Section 1.1; the random-coefficients example in Section 6.1 shows that with multidimensional A and Gaussian errors only trivial FHR moments exist.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Moment Restrictions for Nonlinear Panel Data Models with Feedback." pith.science (2026). https://pith.science/paper/R7KSQB57

@misc{pith2026250612569,
  author       = {Pith},
  title        = {Pith review of: Moment Restrictions for Nonlinear Panel Data Models with Feedback},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/R7KSQB57}},
  note         = {Machine review of arXiv:2506.12569}
}
read the original abstract

Many panel data methods, while allowing for general dependence between covariates and time-invariant agent-specific heterogeneity, place strong a priori restrictions on feedback: how past outcomes, covariates, and heterogeneity map into future covariate levels. Ruling out feedback entirely, as often occurs in practice, is unattractive in many dynamic economic settings. We provide a general characterization of all feedback and heterogeneity robust (FHR) moment conditions for nonlinear panel data models and present constructive methods to derive feasible moment-based estimators for specific models. We also use our moment characterization to compute semiparametric efficiency bounds, allowing for a quantification of the information loss associated with accommodating feedback, as well as providing insight into how to construct estimators with good efficiency properties in practice. Our results apply both to the finite dimensional parameter indexing the parametric part of the model as well as to estimands that involve averages over the distribution of unobserved heterogeneity. We illustrate our methods by providing a complete characterization of all FHR moment functions in the multi-spell mixed proportional hazards model. We compute efficient moment functions for both model parameters and average effects in this setting.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

37 extracted references · 36 canonical work pages

  1. [1]

    Identification of Average Marginal Effects in Fixed Effects Dynamic Discrete Choice Models

    Aguirregabiria, V. and Carro, J. M. (2021). Identification of average marginal effects in fixed effects dynamic discrete choice models. arXiv preprint arXiv:2107.06141 . Aguirregabiria, V. and Mira, P. (2010). Dynamic discrete choice structural models: A survey. Journal of Econometrics , 156(1):38 –

    work page Pith review arXiv 2021

  2. [3]

    Windmeijer, F

    Cambridge university press. Windmeijer, F. (2008). The Econometrics of Panel Data , volume 46 of Advanced Studies in Theo- retical and Applied Econometrics , chapter GMM for panel data count models, pages 603 –

    work page 2008

  3. [20]

    Chamberlain, G. (2023). Identification in dynamic binary choice models. SERIEs, 14(3):247–251. Chernozhukov, V., Fern´ andez-Val, I., Hahn, J., and Newey, W. (2013). Average and quantile effects in nonseparable panel models. Econometrica, 81(2):535–580. Chesher, A., Rosen, A. M., and Zhang, Y. (2024). Robust analysis of short panels. arXiv preprint arXiv:...

    work page Pith review arXiv 2023

  4. [26]

    Chamberlain, G. (2010). Binary response models for panel data: Identification and information. Econometrica, 78(1):159–168. Chamberlain, G. (2022). Feedback in panel data models. Journal of Econometrics , 226(1):4 –

    work page 2010

  5. [33]

    Hence, since Eθ,ωη [ϕθ(Y T , XT )] = 0 for all η ∈ Bκ(η∗), we have Eθ,ω∗[ϕθ(Y T , XT )Sη(Y T , XT )′] =

    Next, by (B)(i) and (B)(ii), we can apply Lemma 5.4 in Newey and McFadden (1994), and conclude that η 7→ Eθ,ωη ϕθ(Y T , XT ) is differentiable at η∗ with derivative Eθ,ω∗[ϕθ(Y T , XT )Sη(Y T , XT )′] where Sη(Y T , XT ) denotes the score at η∗. Hence, since Eθ,ωη [ϕθ(Y T , XT )] = 0 for all η ∈ Bκ(η∗), we have Eθ,ω∗[ϕθ(Y T , XT )Sη(Y T , XT )′] =

    work page 1994

  6. [34]

    matrix. Its orthocomplement is T ⊥ θ0,ω0,K = ϕ ∈ RK | E[ϕ] = 0, E[ϕ′ϕ] < ∞ with E [ϕ′s] = 0, for all s ∈ Tθ0,ω0,K , or, equivalently, T ⊥ θ0,ω0,K = ϕ ∈ RK | E[ϕ] = 0, E[ϕ′ϕ] < ∞ with E [ϕSη′] = 0, for all scores Sη of smooth parametric submodels } , by Lemma A.1 in Newey (1990). We have ϕθ0 ∈ T ⊥ θ0,ω0,K if and only if E ϕθ0(yT , xT )Sη(yT , xT )′ =

    work page 1990

  7. [35]

    forward orthogonal transforms

    Γ(T − t) (1 − ept)T −t−1deps:T −1 = E h ψθ(Y0, eP T −1, P , XT −1) Y0 = y0, eP s−1 = eps−1, P = p, Xs = xs i , which does not depend on xs:T −1. The second equality follows from the fact that the “forward orthogonal transforms” eP s:T −1 are independent of eP s−1, P , Xs by part (i) of Lemma D.1 and Lemma D.2. This shows that, for s ∈ {3, . . . , T− 1}, ψ...

    work page 1994

  8. [36]

    TX t=1 ∂ ln fθ(Yt | Yt−1, Xt, A) ∂β | Y T , XT # (71) = E

    By the Projection Theorem (e.g Theorem 11.1 in Van der Vaart (2000)) we conclude that, Π(φθ(Y0, YT , XT )|T ⊥ θ,ω,L) = T −1X t=1 φ⊥ θ,t(Y0, eP t, P , Xt), as claimed. D.2.1 MPH efficient score under feedback We can use Lemma D.3 to derive an explicit expression of the efficient score forθ = (α, β′, γ)′. Given the historical and continued importance of the...

    work page 2000

Show all 37 references
  1. [37]

    leads and lags

    lnYt. This yields ∂ ln Λα(Yt) ∂α = ln Yt ∂ ln λα(Yt) ∂α = 1 α + ln Yt. In the parameterization fP1, P , with Y1 = P 1 α 1 e−X ′ 1 β α − γ α Y0 = eP 1 α 1 P 1 α e−X ′ 1 β α − γ α Y0 and Y2 = (1 − eP1) 1 α P 1 α e−X ′ 1 β α − γ α Y0 we have ∂ ln Λα(Y1) ∂α = ln Y1 = − 1 α (X ′ 1β...

    1994

  2. [38]

    Chamberlain, G

    Cambridge University Press, Cambridge. Chamberlain, G. (1986). Asymptotic efficiency in semi-parametric models with censoring. Journal of Econometrics, 32(2):189 –

    1986

  3. [46]

    Chamberlain, G. (1984). Handbook of Econometrics , volume 2, chapter Panel Data, pages 1247 –

    1984

  4. [57]

    and Card, D

    Ashenfelter, O. and Card, D. (1985). Using the longitudinal structure of earnings to estimate the effect of training programs. Review of Economics and Statistics , 67(4):648 –

    1985

  5. [67]

    and Chen, X

    Ai, C. and Chen, X. (2012). The semiparametric efficiency bound for models of sequential moment restrictions containing unknown functions. Journal of Econometrics , 170(2):442–457. Al-Sadoon, M. M., Li, T., and Pesaran, H. (2017). Exponential class of dynamic binary choice pan...

    2012

  6. [79]

    Heckman, J. J. and Borjas, G. J. (1980). Does unemployment cause future unemployment? defini- tions, questions and answers from a continuous time model of heterogeneity and state dependence. Economica, 47(187):247–283. Heckman, J. J. and Singer, B. (1984). The identifiability ...

    1980

  7. [112]

    Carrasco, M., Florens, J.-P., and Renault, E. (2007). Linear inverse problems in structural economet- rics estimation based on spectral decomposition and regularization. Handbook of econometrics, 6:5633–5751. Chamberlain, G. (1980). Analysis of covariance with qualitative data...

    2007

  8. [131]

    and Powell, J

    Blundell, R. and Powell, J. L. (2003). Advances in Economics and Econometrics: Theory and Ap- plications, Eighth World Congress , volume 2, chapter Endogeneity in nonparametric and semi- parametric regression models, pages 312 –

    2003

  9. [135]

    Newey, W. K. and McFadden, D. (1994). Large sample estimation and hypothesis testing.Handbook of econometrics, 4:2111–2245. Nickell, S. (1980). Estimating the probability of leaving unemployment. Econometrica, 47(5):1249 –

    1994

  10. [218]

    Chamberlain, G. (1992). Comment: sequential moment restrictions in panel data. Journal of Business and Economic Statistics , 10(2):20 –

    1992

  11. [241]

    Honor´ e, B. E. and Hu, L. (2004). Estimation of cross sectional and panel data censored regression models with endogeneity. Journal of Econometrics , 122(2):293–316. Honor´ e, B. E. and Kyriazidou, E. (2000). Panel data discrete choice models with lagged dependent variables. ...

    2004

  12. [261]

    and Bond, S

    Arellano, M. and Bond, S. (1991). Some tests of specification for panel data: Monte carlo evidence and an application to employment equations. The review of economic studies , 58(2):277–297. Arellano, M. and Bonhomme, S. (2011). Nonlinear panel data analysis. Annu. Rev. Econ.,...

    1991

  13. [340]

    Blundell, R., Griffith, R., and Windmeijer, F. (2002). Individual effects and dynamics in count data models. Journal of Econometrics , 108(1):113 –

    2002

  14. [351]

    Botosaru, I., Loh, I., and Muris, C. (2024). An adversarial approach to identification. Brown, B. W. and Newey, W. K. (1998). Efficient semiparametric estimation of expectations. Econometrica, 66(2):453–464. Buchinsky, M., Hahn, J., and Kim, K. i. (2010). Semiparametric inform...

    2024

  15. [357]

    Blundell, R

    Cambridge University Press. Blundell, R. W. and Powell, J. L. (2004). Endogeneity in semiparametric binary response models. The Review of Economic Studies , 71(3):655–679. Bonev, P. (2020). Nonparametric identification in nonseparable duration models with unobserved heterogene...

    2004

  16. [375]

    Ghanem, D., Sant’Anna, P

    Springer Science & Business Media. Ghanem, D., Sant’Anna, P. H., and W¨ uthrich, K. (2022). Selection and parallel trends. arXiv preprint arXiv:2203.09001. 39 Graham, B. S., Pinto, C., and Egel, D. (2012). Inverse probability tilting for moment condition models with missing da...

    2022 arXiv

  17. [424]

    and Bover, O

    Arellano, M. and Bover, O. (1995). Another look at the instrumental variable estimation of error- components models. Journal of econometrics , 68(1):29–51. Arellano, M. and Honor´ e, B. (2001). Panel data models: some recent developments. In Handbook of econometrics, volume 5,...

    1995

  18. [438]

    Hahn, J. (1994). The efficiency bound of the mixed proportional hazard model. The Review of Economic Studies, 61(4):607–629. Hahn, J. (1997). A note on the efficient semiparametric estimation of some exponential panel models. Econometric Theory, 13(4):583 –

    1994

  19. [588]

    Heckman, J. J. (1991). Identifying the hand of past: distinguishing state fependence from hetero- geneity. American Economic Review, 81(2):75 –

    1991

  20. [624]

    41 Wooldridge, J

    Springer, 3rd edition. 41 Wooldridge, J. M. (1997). Multiplicative panel data models without the strict exogeneity assump- tion. Econometric Theory, 13(5):667–678. Woutersen, T. M. (2000). Essays on the integrated hazard and orthogonality concepts . Brown University. A Proofs ...

    1997

  21. [660]

    and Bond, S

    Blundell, R. and Bond, S. (2000). Gmm estimation with persistent panel data: an application to production functions. Econometric Reviews, 19(3):321 –

    2000

  22. [866]

    Stefanski, L. A. and Carroll, R. J. (1990). Deconvolving kernel density estimators. Statistics, 21(2):169–184. van der Laan, M. J. and Robins, J. M. (2003). United Methods for Censorted Longitudinal Data . Springer-Verlag, New York. Van der Vaart, A. W. (2000). Asymptotic stat...

    1990

  23. [956]

    Lancaster, T. (1990). The Econometric Analysis of Transition Data . Number 17 in Econometric Society Monographs. Cambridge University Press, Cambridge. Lee, W. (2020). Identification and estimation of dynamic random coefficient models . PhD thesis, The University of Chicago. 4...

    1990

  24. [1079]

    Graham, B. S. and Powell, J. L. (2012). Identification and estimation of average partial effects in ‘irregular’ correlated random coefficient panel data models. Econometrica, 80(5):2105 –

    2012

  25. [1266]

    and Pakes, A

    Olley, S. and Pakes, A. (1996). The dynamics of productivity in the telecommunications equipment industry. Econometrica, 64(6):1263–1297. Pakel, C. and Weidner, M. (2023). Bounds on average effects in discrete choice panel data models. arXiv preprint arXiv:2309.09299 . Ridder,...

    1996

  26. [1318]

    Chamberlain, G

    North-Holland, Amsterdam. Chamberlain, G. (1985). Longitudinal Analysis of Labor Market Data , chapter Heterogeneity, omit- ted variable bias, and duration dependence, pages 3 –

    1985

  27. [1385]

    Bonhomme, S., Dano, K., and Graham, B. S. (2023). Identification in a binary choice panel data model with a predetermined covariate. SERIEs: Journal of the Spanish Economic Association , 14(3-4):315 –

    2023

  28. [1589]

    Robins, J. (1986). A new approach to causal inference in mortality studies with a sustained exposure period—application to control of the healthy worker survivor effect. Mathematical modelling, 7(9- 12):1393–1512. Robins, J. M. (2000). Marginal structural models versus structu...

    1986

  29. [2152]

    Granger, C. W. J. (1969). Investigating causal relations by econometric models and cross-spectral methods. Econometrica, 37(3):424 –

    1969

Pith tools

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