Determine the finite-sample value of an additional unsmoothed GMM optimization step

Determine whether re-optimizing the corresponding unsmoothed GMM criterion after obtaining the smoothed GMM estimator provides a finite-sample benefit, particularly in panel instrumental-variables quantile regression with individual fixed effects and serial dependence.

Background

The paper’s proposed estimator is based on smoothed moment conditions because smoothing makes the criterion differentiable and facilitates both numerical optimization and asymptotic analysis. After estimating the smoothed GMM model, one possible alternative is to re-optimize the analogous unsmoothed GMM criterion.

The authors note that prior results for linear iid IVQR models suggest smoothing can improve pointwise mean squared error, while potentially reducing estimated variation across quantiles. Because the finite-sample consequences of an additional unsmoothed optimization step are unresolved, determining whether that step improves estimation would clarify practical implementation choices for the proposed panel estimator.

References

After obtaining the smoothed GMM estimator, one could re-optimize the corresponding unsmoothed GMM criterion. We do not take this as the baseline estimator. In linear iid IVQR models, \citet{KaplanSun17} find that smoothing can improve pointwise mean squared error, although it may reduce the estimated variation across quantiles. Thus, the finite-sample benefit of an additional unsmoothed optimization step is not clear.

— GMM for Panel Data Quantile Regression  (2610.12824 - Galvao et al., 9 Oct 2026) in Section 2, subsection “Two-step GMM estimator under over-identification”