About Me
I am a PhD candidate at the Research School of Economics, Australian National University.
My research interests are:
- Computational methods in dynamic optimization and structural estimation
- Economic growth
- Empirical industrial organization
My chair supervisor is Professor Fedor Iskhakov.
Research
-
Growth and (Re)distribution in China and Korea: Bayesian Estimation of a Non-Stationary Heterogeneous Agent Model
I analyze growth and (re)distribution jointly in China (1982–2019) and Korea (1962–2019) inside a non-stationary heterogeneous-agent model. Two innovations make Bayesian estimation feasible: a certainty-equivalent-anchored quasi-rational equilibrium concept which replaces full rational expectations with a finite-horizon deterministic forecast; and a deep-learning pipeline that amortizes the agent's decision problem, the equilibrium computation, and the posterior itself. Counterfactual experiments find that TFP catch-up is the dominant driver of output growth yet leaves the wealth distribution untouched, that the education expansion and the fertility decline likewise buy per-capita growth with little distributional footprint, that state investment promotes growth while potentially concentrating wealth, that a thin social safety net buys growth at a large cost to consumption at the bottom, and that among the model's estimated primitives idiosyncratic labor-productivity risk has the largest marginal impact on wealth inequality.
-
Policy-Tree Descendant Algorithm for Solving Consumption–Saving Problems with a Safety Net
This paper develops a new algorithm for solving life-cycle consumption–saving problems that feature social insurance in the form of an exogenous consumption floor. Despite the proliferation of this setup in the literature, the fact that such safety-net programs render the dynamic optimization problem non-convex and the optimal consumption policy discontinuous is generally overlooked. Applicable for the general class of piecewise-affine consumption–saving models, the policy-tree descendant (PTD) algorithm developed in this paper accurately tracks the discontinuity points of the policy function and computes it using a graph-theoretic representation without the need to discretize the state space. In a series of simulation exercises, we compare the performance of the PTD algorithm to the existing endogenous grid methods for non-convex dynamic optimization problems. Being grid-free, PTD pays its computational cost per simulated path rather than per parameter configuration, and no part of its cost grows with the dimension of the exogenous state space. In our experiments the accuracy of the PTD solver matches the endogenous grid benchmarks, while in the empirically relevant regime of heterogeneous agents—few simulated paths per parameter configuration—they are two to three orders of magnitude faster.
-
Land Offers and Fiscal Competition Between City Governments in China
I analyze the fiscal competition between city governments in China by structurally estimating a Bertrand pricing game model. The model characterizes the land pricing strategy of city governments as they use land sales discounts to attract industrial firms. The estimation results imply that city governments can generate a huge amount of fiscal revenue from landing industrial firms, which is around 45% of the firm's yearly output. By counterfactual experiments, I show that the impact of this kind of fiscal competition on resource allocation is small. Simulation results also show that fiscal centralization and increasing urban wages would result in a modest average land price increase.