Computing HA models
Useful papers & technics:
Endogenous Grid Method (EGM)
- [EXPECTATION 1] Useful tool to approximate expectations over normally distributed shocks: Gauss-Hermite + Monomial Rule.
- [EXPECTATION 2] Computing expectations of value functions using polynomials: Judd et al. (2017)
- [SIMULATION] Non-stochastic simulation routine: Young (2010)
- [COLOCATION] Collocation Method: Collocation
- [ECM] Envelop Condition Method (ECM): Maliar (2013)
- [SMM] Estimating HA models :: Sobol sequence
- [VAR, AR(1)] Discretizing VAR: Farmer et Toda (2013)
int C++ // (for discrete time)
- [HOUSING] Solve standard housing macroeconomic models with DC-EGM, Sommer & Sullivan AER (2018). Useful note here, code available upon-request.
- [AIYAGARI] Solve the Aiyagari model in 0.04 – 0.14 seconds with EGM + Young method (2010). Useful note (by Josep Pijoan-Mas) is available here. Download my code (iterate on marginal utilities or value functions with code available upon-request).
- [SGM] Solve the stochastic growth model using EGM, Barillas & Villaverde (2007), code here.
- [ENTREPRENEURSHIP] Solve Cagetti & DeNardi (2006) in 3s with DC-EGM, code available upon-request.
MatLab % (for continuous time)
- [AIYAGARI] Solve Aiyagari in 0.13 seconds with Envelope Condition Method (ECM), many codes available here: HATC project.
- [AIYAGARI] Aiyagari in Continous Time with Jump-Drift Process. Code is available upon-request, note: here.
- [HANK] Heterogenous Agent New Keynesian (HANK) (Kaplan et al. (2018)) model and the code available here: (not yet available), note: here.
Computational speed - comparison between EGM, DC-EGM and VFI
Comparison of performance and accuracy of EGM, DC-EGM and VFI methods on occupational choice and entrepreneurship models à la Cagetti & De Nardi (2006). The presence of discrete choice (occupational choice) makes EGM inaccurate. DC-EGM encompasses generated kinks very well, while being substantially faster than standard VFI.
||Speed (in s)