Hello! I am a PhD candidate in Economics at Emory University, currently on the job market. My research focuses on empirically and theoretically examining the impact of economic frictions on macroeconomic dynamics.
The wealth effect is a critical channel through which economic shocks propagate. This paper examines the persistent and heterogeneous effects of net worth shocks across U.S. counties, focusing on the local collateral constraints and downward nominal wage rigidity and their interactions. We develop a two-agent general equilibrium model that demonstrates how the interaction between collateral constraints and DNWR shapes local wealth effects on consumption, unemployment and house prices. Using a new county-level dataset spanning 2003-2019 and a semi-varying coefficient approach, we identify large heterogeneities and non-linearities in shock transmission: the marginal propensity to consume out of wealth varies from 11 to 3 cents per dollar between low- and high-friction counties, with shock responses amplified 2-5 times when both frictions bind. These findings highlight how local financial and labor market conditions determine heterogeneous responses to aggregate shocks, informing the design of regionally targeted economic policies.
CountyPlus is an open-source panel dataset that covers 3000+ U.S. counties from 2003 to 2019. It consists of 100+ variables, including demographic, geographic, household balance sheet, local economy indicators. Specially, this dataset estimates household consumption, measure of local financial friction, and measure of local nominal friction (Downward Nominal Wage Rigidity).
(Version 0.0.2 is available now! Newly added 04-19 identified net worth shock and spatial weight matrices)
This package provides object-oriented API to define, train, evaluate and adapt multi-linear Adaptive Sparse Grid (ASG) interpolations. It can be used to solve high-dimensional discrete/continuous time models; perform high-dimensional non-parametric estimations; and help with other high-dimensional numerical exercises.
with Cheng Ding, Vivian Yue, and Aliaksandr Zaretski
with Kaiji Chen, and Kai Zhao
with Yunyun Jiang and Haitao Zheng
with YunYun Jiang and Haitao Zheng
with Shanshan Wang, Jie Hu and Haitao Zheng
Presenter
Presenter & Session chair
High-dimensional models, either numerical or empirical, are known to suffer from the curse of dimensionality. Starting from the basic idea of interpolation, this blog post provides an intuitive explanation of how ASG works and how it can further reduce the curse of dimensionality.
The important paper by Barles and Souganidis (1991) provides a powerful framework to analyze a class of 2nd-order non-linear PDEs which are commonly used in economics. Among all the required conditions, the monotonicity of the numerical scheme is critical and often violated in practice. This blog post explains how to analyze the monotonicity of a wide class of numerical schemes with multiple examples that share similar structures with some commonly used economics models. One can follow the examples to design their own numerical schemes.
Discretizing PDE needs careful treatment on boundary conditions, esp. at points where multiple boundary conditions must hold simultaneously: are my boundary conditions conflict with each other? How to accommodate multiple boundary conditions? etc. However, there lacks formal or informal discussion about this topic. This blog intuitively introduces the ideas of properly discretizing the boundary problem at such special points while emphasizing the importance of respecting the boundary conditions.
The collocation method is widely used in macroeconomic modeling to approximate value functions (or policy functions) through interpolation techniques. Despite its advantages, many textbooks and online materials overlook an important scenario: cases where the transition matrix (often assumed to describe a Markov process governing uncertainty in dynamic programming) is exogenous but depends on the current state of the system. This scenario, involving endogenous risk, is highly relevant in economic applications. For instance, an individual's unemployment risk in general equilibrium may depend on the aggregate unemployment rate, which itself evolves as an aggregate state. Addressing such cases requires specific formulations and methodological adjustments to extend the collocation method. In this post, I first establish the notation for collocation and outline the construction of the discretized system. I emphasize that endogenous risk, compared to standard exogenous risk, significantly increases the number of equations required to interpolate the expected value function. Specifically, the size of the system expands proportionally to the number of supporting nodes for endogenous states. I then discuss how to solve the resulting pseudo-linear system and compute its Jacobian matrix analytically. Finally, I explore several topics for efficiently implementing the collocation method.
Optim.jl is a powerful Julia package that offers a wide range of optimization algorithms. Among these, interior-point Newton (IPNewton) is the only gradient-based method capable of solving non-linearly constrained problems. Despite its flexibility, IPNewton can magically accelerate the algorithm by requiring users to provide the Jacobians and Hessians for both objective function and constraint functions. Even though automatic differentiation is available today, user-provided information still has the best performance when it is available. However, specifying these data structures and functions is challenging because of the math complexity and the non-transparent internal routines of the algorithm. In this post, I clearly outline the ideas behind and math formulas for these data structures (primarily matrices) to help readers better understand the official documentation and tailor the method to their specific problems.