Towards a HANK Model for Canada: Estimating a Canadian Income Process

Heterogenous agent models rely on good estimates of the distribution of individual income growth to model the consumption behaviour of households and its macroeconomic implications. I study the distribution of income growth among Canadian workers and find that it is characterized by large, infrequent shocks rather than small but frequent ones. This distribution is very similar to US findings.

I provide estimates for the parameters of households’ stochastic income process based on T4 annual tax return data. On the methodological side, I show how the estimation process can be accelerated, allowing me to run millions of simulations in milliseconds. To achieve this, I use CUDA-based methods to distribute estimation across multiple NVIDIA Tesla graphical processing units.

My estimates represent a key first step in developing quantitatively realistic Heterogeneous Agent New Keynesian (HANK) models for the Canadian economy. HANK models are important tools for understanding consumption behaviour and analyzing the transmission mechanism of monetary policy. The estimated process in this paper may also prove useful in other contexts where an empirically realistic representation of household earnings dynamics is vital.