Estimating Linear Dynamic Panels with Recentered Moments

Round 1

Reviewer 1 Report

This paper conducts a comprehensive analysis of dynamic panel data model estimation using bias-corrected moments (profile scores). Despite the existence of similar method of moment estimators in the literature, the paper introduces extensions to encompass more general forms of heteroskedasticity, higher-order autoregressive models, and unit roots. However, these extensions contribute to the paper's complexity.

The sole comment for the author is as follows:

1. Assumption 1 labels individual constants as "fixed effects," but subsequently assumes them to be i.i.d. random variables with existing fourth moments. The author should clarify the need for imposing a more restrictive i.i.d. random specification, considering that the matrix A eliminates individual-specific effects from the moments. It appears that Theorem 4 holds even if the individual effects are considered fixed constants.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

Please check the attached file

Comments for author File: Comments.pdf

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 3 Report

The author uses the word "endogenous" in the sense that the regressors are correlated with the composite error term, including individual effects and the idiosyncratic error term. This is not incorrect, of course. However, I find this a bit confusing because many papers in this literature distinguish between endogeneity and exogeneity depending on whether the regressors are correlated with the idiosyncratic error or not by assuming that the regressor is correlated with individual effects. In fact, the author assumes "strict exogeneity" in Assumption 3.  Therefore, I suggest dropping the word "endogeneity" when it is used in the sense that the regressor is correlated with the individual effects.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Dear author, thank you for your response to my review report. I appreciate how you have addressed my suggestions and I believe that it has improved your paper.

It is not required that you mention the name "implicit indirect inference estimator" in footnote 3 (lines 1118–1119). The name you use (RMM) is of course much better and your motivation is now clearly explained in the introduction. Typically, one tries to reduce the number of (foot)notes as much as possible.

I both like the idea presented in this paper and the estimation method in your previous work with Xuewen Yu. It is really new and provides useful insights when estimating dynamic panel data models.

Warm regards