Honours project:: Modelling Dynamic Choice: Private Health Insurance
Supervisors: Professor Denzil Fiebig (University of New South Wales, and Centre for Health Economics Research
and Evaluation), Professor Jane Hall (Centre for Health Economics Research and Evaluation) and Dr Anne Young (Research Centre for Gender and Health, University of Newcastle)
University: University of New South Wales, and Centre for Health Economics Research and Evaluation

This paper uses ALSWH data on the young cohort over the first three waves to build and test a model for private hospital insurance choice in Australia. An understanding of why Australians choose private cover is pivotal to designing policies that will ensure a well functioning private health system. Further, the Australian market seems to exhibit a high level of persistence in this choice. Consumers that are insured tend to stay insured and vice versa. This is despite a wide range of legislation changes to try and induce movement into the private system. So far, researchers have been unable to quantify this persistence due to the limitations of crosssectional data and have therefore missed a crucial piece of the puzzle. Use of the ALSWH panel data set has allowed this study to track women as they move in and out of cover, measuring changes in decisions over time and over individuals.

Objective: To model consumer choice to purchase private hospital insurance in Australia using a dynamic framework to account for persistence, unobserved heterogeneity across individuals, and socio-demographic drivers.

Study design/setting: The final data set was created by using the first three waves of the Young Cohort of the ALSWH data and then defining variables consistently across all waves so that changes over time could be measured.

Results: Investigation of bi-variate relationships indicated self-assessed health, income, age, education, employment and child rearing variables were highly positively correlated with the choice to insure. Risk assessment variables such as smoking were negatively correlated. Other models that were tested but rejected included: models using a single cross-section; a static panel specification; and a fixed effects specification. These were rejected due to bias from non-dynamic specification. The final model specification uses a dynamic, random effects probit model, which allows for persistence in decisions, differences across individuals, and controls for sociodemographic groupings.

The dependant variable - the choice to insure - is measured through a binary choice, latent variable framework. The independent variables included were:
1. A lagged dependant variable (to account for persistence in decisions);
2. Exogenous socio-demographic controls such as income, employment and age;
3. A linear specification for the correlation between unobserved heterogeneity and the exogenous variables. Conclusions: Factors that lead to a greater propensity to insure included insurance status from last period, income (both personal and household), employment, self-assessed health, not being single, and being pregnant in the past 12 months. Factors that lead to a lower propensity to insure included: living in a regional or rural area and being born outside of Australia. These results shed further light on why Australians choose private health insurance. Specifically, if the Australian government wishes to improve access to the public system by ensuring greater movements into the private system, the results from this study will be helpful. The significance of dynamic choice specifications discovered by this study highlight the need for panel data sets, such as the ALSWH set, to provide a true understanding of health related behaviours.

To Contact Vineta::
Vineta Salale
University of New South Wales
Level 3, Quadrangle Building
School of Banking and Finance, UNSW
Sydney ,NSW 2052
Australia

Email: v.salale@gmail.com