|
I'm a postdoctoral research/teaching fellow at the University of British Columbia.
Primary Fields: Econometrics, Macroeconomics Curriculum Vitae Github: @wupeifan peifan [dot] wu [at] ubc [dot] ca |
Working Paper
Differentiable State Space Models and Hamiltonian Monte Carlo Estimation [draft] [online appendix] [slides]
Related Julia package: DifferentiableStateSpaceModels.jl
Joint with David Childers, Jesús Fernández-Villaverde, Jesse Perla, Cameron Pfiffer, Christopher Rackauckas
Abstract: We propose a methodology to take dynamic stochastic general equilibrium (DSGE) models to the data based on the combination of differentiable state space models and the Hamiltonian Monte Carlo (HMC) sampler. First, we introduce a method for differentiating perturbation solutions of DSGE models with respect to the model's parameters. The resulting output can be used for various computational tasks requiring gradients, such as building an HMC sampler to estimate first- and second-order approximations of DSGE models. The availability of derivatives also enables a general filter-free method to estimate nonlinear, non-Gaussian DSGE models by sampling the joint likelihood of parameters and latent states. We show that the gradient-based joint likelihood sampling approach is superior in efficiency and robustness to standard Metropolis-Hastings samplers by estimating a canonical real business cycle model and a medium-scale New Keynesian DSGE model.
Related Julia package: DifferentiableStateSpaceModels.jl
Joint with David Childers, Jesús Fernández-Villaverde, Jesse Perla, Cameron Pfiffer, Christopher Rackauckas
Abstract: We propose a methodology to take dynamic stochastic general equilibrium (DSGE) models to the data based on the combination of differentiable state space models and the Hamiltonian Monte Carlo (HMC) sampler. First, we introduce a method for differentiating perturbation solutions of DSGE models with respect to the model's parameters. The resulting output can be used for various computational tasks requiring gradients, such as building an HMC sampler to estimate first- and second-order approximations of DSGE models. The availability of derivatives also enables a general filter-free method to estimate nonlinear, non-Gaussian DSGE models by sampling the joint likelihood of parameters and latent states. We show that the gradient-based joint likelihood sampling approach is superior in efficiency and robustness to standard Metropolis-Hastings samplers by estimating a canonical real business cycle model and a medium-scale New Keynesian DSGE model.
Market Structure, Labor Share, and Firm Investment [draft] [slides]
Joint with Alberto Polo
Abstract: This paper provides a theory that links industrial market structure with firm investment and employment decisions consistent with the empirical literature. We demonstrate that the impact of market structure on a single firm as the impact originated from the firm's direct competitors. We develop a business cycle model with a continuum of industries. Within each industry, firms play a dynamic duopoly game by choosing production quantities and investing in capital. As firms are not atomic within one industry, The strategic substitutabilities between the firms generate time-varying industry-level market structure and price markups. The implications of this model are consistent with the empirical evidence found in literature: i) industry-level market concentration and labor shares are negatively correlated; ii) firm investment and industry-level market concentration are negatively correlated. We also find heterogeneity in investment sensitivities to the business cycle for firms facing different opponents, supported with new empirical evidence we found.
Joint with Alberto Polo
Abstract: This paper provides a theory that links industrial market structure with firm investment and employment decisions consistent with the empirical literature. We demonstrate that the impact of market structure on a single firm as the impact originated from the firm's direct competitors. We develop a business cycle model with a continuum of industries. Within each industry, firms play a dynamic duopoly game by choosing production quantities and investing in capital. As firms are not atomic within one industry, The strategic substitutabilities between the firms generate time-varying industry-level market structure and price markups. The implications of this model are consistent with the empirical evidence found in literature: i) industry-level market concentration and labor shares are negatively correlated; ii) firm investment and industry-level market concentration are negatively correlated. We also find heterogeneity in investment sensitivities to the business cycle for firms facing different opponents, supported with new empirical evidence we found.
Sequential Cyclicality of Firm Births and Deaths [slides]
Joint with Mi Luo, Can Tian, Gang Zhang
Abstract: This paper documents the sequential pattern of firm entry and exit and the economic condition in the US at the state level. First, at both annual and quarterly frequencies, firm entry leads the economic conditions and is positively associated with subsequent firm exit. Moreover, such a leading pattern is also robust at the major MSA level. It suggests that firm formation provides a reasonable signal for local economic conditions. Second, no sequential pattern is observed between firm exit and economic conditions. Third, the sequential pattern is consistent across states, suggesting a national pattern rather than the reallocation of labor between states. We highlight the importance of the sequential pattern between entry and business cycle when targeting startup subsidies.
Joint with Mi Luo, Can Tian, Gang Zhang
Abstract: This paper documents the sequential pattern of firm entry and exit and the economic condition in the US at the state level. First, at both annual and quarterly frequencies, firm entry leads the economic conditions and is positively associated with subsequent firm exit. Moreover, such a leading pattern is also robust at the major MSA level. It suggests that firm formation provides a reasonable signal for local economic conditions. Second, no sequential pattern is observed between firm exit and economic conditions. Third, the sequential pattern is consistent across states, suggesting a national pattern rather than the reallocation of labor between states. We highlight the importance of the sequential pattern between entry and business cycle when targeting startup subsidies.
Firm Demographics and the Great Recession [slides]
Joint with Gian Luca Clementi and Berardino Palazzo
Abstract: The last U.S. recession stands out not only for its depth, but also for the rather slow recovery the followed it. What is less well known is that the number of productive units also dropped substantially, while it barely budged in occasion of the 1981 recession, and kept increasing during the 1991 and 2001 recessions. To the extent that the stock of establishments is a very slow–moving variable, a recession characterized by an unusually large drop in establishments will necessarily be followed by a slow recovery. In order to evaluate the quantitative significance of this simple mechanism, we build a general equilibrium business cycle model with heterogeneous firms and endogenous entry and exit, and calibrate it so that the implied firm–level and aggregate dynamics are consistent with the empirical evidence. Preliminary results show that, when hit with a negative total factor productivity shock, the model produces a much more persistent decline in employment and output than an off-the-shelf model with a fixed number of establishments.
Joint with Gian Luca Clementi and Berardino Palazzo
Abstract: The last U.S. recession stands out not only for its depth, but also for the rather slow recovery the followed it. What is less well known is that the number of productive units also dropped substantially, while it barely budged in occasion of the 1981 recession, and kept increasing during the 1991 and 2001 recessions. To the extent that the stock of establishments is a very slow–moving variable, a recession characterized by an unusually large drop in establishments will necessarily be followed by a slow recovery. In order to evaluate the quantitative significance of this simple mechanism, we build a general equilibrium business cycle model with heterogeneous firms and endogenous entry and exit, and calibrate it so that the implied firm–level and aggregate dynamics are consistent with the empirical evidence. Preliminary results show that, when hit with a negative total factor productivity shock, the model produces a much more persistent decline in employment and output than an off-the-shelf model with a fixed number of establishments.
Notes
- Setup instruction and examples for ComputeCanada
QuantEcon
I'm a member of the QuantEcon team. In particular, I involve in QuantEcon DataScience and QuantEcon Julia lectures.
- Presentation for JupyterCon
- Presentation for UBC Jupyter Day
- Presentation for JupyterCon
- Presentation for UBC Jupyter Day
Miscellany
- I have been a supporter of Liverpool Football Club since 2003.
- I am a certified bartender of New York State.
- I have been playing piano for more than 20 years. Former member of Clavier team at Tsinghua University.
- Meet Mocha (American bobtail) and Yasmine (English Longhair).
- I am a certified bartender of New York State.
- I have been playing piano for more than 20 years. Former member of Clavier team at Tsinghua University.
- Meet Mocha (American bobtail) and Yasmine (English Longhair).
