Policy
Simulation
Group

Pension Projection Models

Key Feature: Stochastic Microsimulation

The PSG models are stochastic in two different senses of the word.

First, like all microsimulation models, the PSG models use Monte Carlo methods to generate sampling variability in the simulated population.

Second, the PSG models use Monte Carlo methods to realize stochastic macro time-series scenarios from any VAR(2) model that is estimated with historical data. The annual realized values of these demographic and economic variables represent the macro environment in which the lives of sample individuals play out. The macro-micro relationship is top-down without any micro-to-macro feedback as in dynamic stochastic general equilibrium models.

See section 1.2.5 of the SSASIM Guide for an estimated VAR(2) model of inflation and nominal asset returns that is often used in stochastic PSG model runs.

See Methods for Stochastic Trust Fund Projection (January 2003) for an empirical comparison of VAR models with time-series models that can represent series with a time-varying mean.

Other key features.


This page was last revised on February 17, 2016.