Implications of Earnings Model Complexity to Costing Income Contingent Loans

Abstract

The implications of earnings model complexity to costing income contingent loans (ICL) are investigated. Models of hourly wage are developed using the first seven waves of the Household, Income and Labour Dynamics in Australia (HILDA) Survey, starting with a simple mean fit. The residuals are decomposed into permanent and transitory components, and it is shown that the variance components differ depending on employment state. The permanent component is modelled as a random walk, with both Gaussian and non-Gaussian permanent and transitory shocks. The hourly wage models are combined with models of hours per week. Annual earnings are simulated using a Monte Carlo process, and outstanding debt is estimated for a hypothetical ICL while keeping labour force state fixed. It is found that under both the Australian ICL scheme for higher education (HECS) and a modified scheme design, and under both full-time and part-time static employment states, dynamic stochastic earnings models lead to lower projected debt, and significantly lower subsidies than under static earnings models.