Research:
Research:
Working Papers:
“Semiparametric Estimation of Count Regression Models Revisited” (with Paul Rilstone)
This collaborative paper revisits and advances semiparametric approaches to estimating count regression models, which are essential for analyzing discrete, non-negative outcomes like event counts in economics (e.g., number of patents, hospital visits, or policy incidents). Building on existing literature, Gurmu, Rilstone, and Stern (1999),we refine estimation techniques to jointly estimate the regression parameters and the unknown distribution of unobserved heterogeneity. Key contributions include establishing robust identification conditions for the model parameters and deriving asymptotic properties, such as consistency, convergence rates, and asymptotic normality. These theoretical advancements ensure the estimator's efficiency under various data conditions. The paper demonstrates practical utility through simulations and empirical examples, highlighting its relevance for policy evaluation in fields like health economics.
Presented at:
Canadian Economics Association (CEA), Montréal, 2025 (Presenter)
30th International Panel Data Conference, France, 2025 (Co-author presentation)
Poster: CESG 2025, Ottawa
This solo-authored work explores sieve estimation methods for semiparametric count regression models, incorporating innovative basis functions for these models like soft-thresholded wavelets and shallow neural networks with adaptive activations to approximate the unknown nonparametric component. Unlike traditional polynomial bases, soft-thresholded wavelets offer superior handling of localized features, discontinuities, and sparsity in data through denoising techniques, while shallow neural networks with adaptive activations enhance adaptability to complex patterns via machine learning integration and improved gradient flow. The paper develops these techniques, proves their asymptotic properties (including consistency, rate of convergence, and asymptotic normality), and addresses challenges in count data, such as overdispersion and zero-inflation. The paper illustrates practical utility through Monte Carlo simulations and empirical applications, providing flexible and efficient tools for advanced modeling in econometrics and microeconometrics.
Presented at:
(Presented alongside a co-authored paper on count regression models)
Canadian Economics Association (CEA), Montréal, 2025 (Presenter)
Poster: CESG 2025, Ottawa
“Semiparametric Estimation of Panel Count Data: An Application to Terrorism”
(solo-authored, in progress).
This ongoing solo-authored paper applies semiparametric sieve estimation to panel count data models, focusing on longitudinal datasets where observations are repeated over time (e.g., across countries or regions). The framework uses wavelet and polynomial bases to model semiparametric structures, establishing identification and asymptotic efficiency while accounting for panel-specific issues like fixed effects and correlation. A key empirical application examines terrorism incidents, using global datasets to evaluate counter-terrorism policies' impacts on event frequencies. This work bridges theoretical econometrics with practical policy analysis in political economy, offering insights into how regulatory interventions can mitigate risks in high-stakes contexts.
Other Research
“Terrorism and Economic Growth”
First-year PhD paper, York University – November 2021
Empirical analysis of terrorism’s effect on growth, FDI, and military spending using Global Terrorism Database 1970–2019; finds trade openness significantly reduces terrorism and rejects the “Market Civilization and Its Clash with Terror” hypothesis.