The mathematical side of modelling people’s saving, superannuation, housing and portfolio decisions through their lifetime using Python.
The project solves and then estimates structural parameters for a high dimensional dynamic model of people’s saving, superannuation, housing and portfolio decisions through their lifetime. The model has applications to understand people’s retirement adequacy, the distribution of wealth and income, people’s saving and consumption decisions in response to crises such as the COVID pandemic and can help policy design regarding pensions and housing markets.
This presentation will focus on the mathematical contribution of the paper and its computational implementation in Python (using MPI and Numba). In particular, to address the high dimensionality of the model, the paper develops an extension of the “endogenous grid method” method to quickly solve high-dimensional, discontinuous and non-convex dynamic problems using generalised super-derivatives in function space. The paper then makes use of parallel HPC to estimate the model parameters using the cross-entropy method. The presentation will first discuss the contributions of the solution method for the dynamic programming problem and then discuss implementation of the structural estimation on HPC, with a view to share insights and receive feedback on the computational implementation.
This talk will be followed by Hacky Hour at 3pm on Teams.
For more information about ResTech please visit our website restech.unsw.edu.au