Testing for a general class of functional inequalities

Kyungchui Song, Yoon‐Jae Whang, Sokbae LeeEconometric 2017

TL;DR

This paper introduces a novel, general, and easy-to-implement bootstrap-based method for testing inequality restrictions on nonparametric functions. This approach unifies many existing nonparametric testing problems and offers improved power properties, even in complex scenarios with degenerate limiting distributions, making it applicable to a wide range of economic models.

What Problem Does This Paper Address?

Many economic theories impose inequality restrictions on nonparametric functions, which are crucial for understanding phenomena like auction behavior, game theory outcomes, wage inequality, and consumer preferences. However, testing these restrictions, especially when they involve complex functional forms or apply uniformly over a range of parameters, has been challenging. Existing methods often struggle with issues like pointwise asymptotics, conservativeness, or lack of applicability to continuous covariates. This paper aims to provide a unified and robust framework for testing these general functional inequalities.

Methods

The core of the paper's contribution is a new bootstrap-based test for functional inequalities. The test statistic is based on a one-sided $L_p$ functional of kernel-type nonparametric estimators. A key innovation is the use of a bootstrap procedure, specifically a nonparametric bootstrap applied to kernel-based test statistics, which allows for computing critical values. The authors introduce a "contact set" estimation approach to improve power properties, particularly in cases where the null hypothesis holds with slackness. They establish uniform asymptotic validity of the bootstrap test over a broad class of distributions, including those with degenerate limiting distributions, and derive the local power function.

Datasets

The paper uses two real-world datasets for empirical illustrations:

  • Timber auction data: Used to test implications from auction theory, specifically examining conditional quantile functions related to bidding behavior.
  • CPS data extract (Acemoglu and Autor, 2011): Used to test functional inequalities in the context of wage inequality, specifically differences-in-differences in conditional quantiles of log hourly wages.

Key Insights and Ideas

  • Unification of Testing Problems: The proposed framework provides a single, flexible method to address a wide array of nonparametric testing problems involving inequality restrictions, unifying previously disparate approaches.
  • Bootstrap for Robustness and Power: The bootstrap procedure is crucial for handling complex asymptotic distributions and achieving good finite sample power. The contact set estimation further enhances power by adapting critical values to the data.
  • Handling Degeneracy: A significant achievement is the development of a test that remains asymptotically valid even when the limiting distribution of the test statistic is degenerate, a common issue in testing complex functional inequalities.
  • Dual Convergence Rates: The test exhibits interesting local power properties, achieving parametric convergence rates ($\sqrt{n}$) in some directions while maintaining slower but still nontrivial rates in others, demonstrating its adaptability.
  • Empirical Relevance: The method is demonstrated to be useful in analyzing real-world economic data from auctions and wage inequality studies, areas where existing methods might be insufficient.

Key Results

  • The proposed bootstrap test is shown to be asymptotically valid uniformly over a large class of distributions.
  • The test exhibits good power properties in Monte Carlo experiments, often outperforming existing methods.
  • The bootstrap critical values, particularly those derived using contact set estimation, lead to substantial power improvements.
  • The local power analysis reveals dual convergence rates, with a faster rate achieving the parametric $\sqrt{n}$ rate in certain scenarios.
  • Empirical applications to auction data and wage inequality data demonstrate the test's practical usefulness and ability to detect violations of economic restrictions.

Limitations and Future Work

  • Tuning Parameters: While the bootstrap procedure helps, the test still relies on tuning parameters (e.g., bandwidth for kernel estimation, constants for contact set estimation), which require careful selection.
  • Multiscale Approaches: The authors suggest that exploring multiscale versions of the test could lead to adaptive rate-optimality against smooth alternatives, an interesting avenue for future research.
  • Complementarity of Tests: The paper notes that their test, along with sup-norm based tests (like CLR) and other methods, can be complementary, each performing better in different scenarios. This suggests that choosing the most appropriate test might depend on the specific characteristics of the data and the hypothesis being tested.

Summary updated

Abstract

In this paper, we propose a general method for testing inequality restrictions on nonparametric functions. Our framework includes many nonparametric testing problems in a uni ed framework, with a number of possible applications in auction models, game theoretic models, wage inequality, and revealed preferences. Our test involves a one-sided version of Lp functionals of kernel-type estimators (1 < p < É) and is easy to implement in general, mainly due to its recourse to the bootstrap method. The bootstrap procedure is based on nonparametric bootstrap applied to kernel-based test statistics, with estimated 'contact sets'. We provide regularity conditions under which the bootstrap test is asymptotically valid uniformly over a large class of distributions, including the cases that the limiting distribution of the test statistic is degenerate. Our bootstrap test is shown to exhibit good power properties in Monte Carlo experiments, and we provide a general form of the local power function. As an illustration, we consider testing implications from auction theory, provide primitive conditions for our test, and demonstrate its usefulness by applying our test to real data. We supplement this example with the second empirical illustration in the context of wage inequality
BibTeX
@inproceedings{song2017testing,
  author = {Kyungchui Song and Yoon‐Jae Whang and Sokbae Lee},
  title = {{Testing for a general class of functional inequalities}},
  booktitle = {Econometric},
  year = {2017},
}
Figure