An Agent-based Model of Household Spending Using a Random Assignment Scheme

This paper describes a way to model household expenditure using an agent-based microsimulation approach. Work in this area is dominated by econometric models, which attempt to estimate the parameters of an equation linking household characteristics with expenditure patterns. However, as Klevmarken (1997) notes, the representation of behaviour in these models is limited by the current state of the art in economics. A way to overcome these limitations is proposed based on a random assignment scheme where data is obtained from a donor case that is in some way similar to the receiving unit. The paper begins with a brief account of methods for modelling behaviour currently used in microsimulation. It then introduces the random assignment scheme in an example describing how household incomes can be projected over time. This is illustrated by describing two models that have been implemented using the approach. One of them is to forecast changes in household expenditure in response to variations in income. The other models the effect of the level of unemployment on household spending patterns. These examples provide the basis for a discussion of the strengths and weaknesses of the random assignment met.