Principal investigator: Omar Barroso Khodr
Author (s): Thomas Buchmueller; Terence C. Cheng; Ngoc TA Pham; and Kevin E. Staub
Original title: The effect of income-based mandates on the demand for private hospital insurance and its dynamics
Location of the Intervention: Australia (a country in Oceania)
Sample Size: 101,670 observations and 18,407 individuals.
Sector: Health Economics
Primary Variable of Interest: Participation in Private Hospital Insurance (in English) Purchase of Hospital Insurance (PHI)
Type of Intervention: Impact of the Medicare (public health service) Tax Surcharge on the demand for private hospital insurance.
Methodology: Dynamic Panels; Generalized Method of Moments; Instrumental Variables; Hansen's J-Test.
Summary
This study analyzes the impact of the Medicare Levy Surcharge (MLS) on the demand for private hospital insurance in Australia. The MLS affects high-income individuals who do not purchase this type of insurance. Identification is based on variations in liability for payment of the penalty, caused by fluctuations in income and a reform that raised exemption thresholds. Using data from the HILDA longitudinal survey, we estimate dynamic models that capture persistence in purchase decisions, influenced by unobserved factors and temporal dependence. Being subject to the MLS increases the probability of purchase by 2% to 3% in the current year, reaching 13% after a decade of continuous exposure. The penalty has an asymmetrical effect: its impact is twice as great among those who become obligated to pay. Finally, the obligation affects younger individuals more intensely.
- Policy Problem
The authors highlight several important political issues surrounding Australia's private health insurance system and its interaction with the public health system. They point to the declining proportion of the population purchasing private health insurance plans after the creation of the Medicare program in 1984 as a central issue. Since Medicare offers universal coverage, private insurance ceased to be the primary form of healthcare financing and became a supplementary option, leading to a reduction in demand for private plans. This presented a challenge for policymakers in that country, as lower private coverage meant greater dependence on the public system, especially public hospitals, which already face pressures regarding capacity and waiting times.
In this way, the government introduced a series of policies aimed at encouraging greater adherence to private health insurance. One policy concern is adverse selection, in which older and sicker individuals primarily join private plans, increasing costs. To combat this, the Lifetime Health Coverage (VHC) policy was introduced, which penalizes people who delay purchasing insurance beyond age 30 by charging higher premiums. This aimed to encourage younger and healthier individuals to enter the market earlier and balance risk groups. Another policy issue involves affordability, which was addressed through subsidies. Initially subject to income testing, subsidies were later expanded to a flat 30% discount for all families, regardless of income, in an attempt to make coverage more attractive and reduce pressure on Medicare.
Subsequently, another policy challenge identified by the authors is how to design tax-based incentives and penalties to influence consumer behavior. The Medicare Surcharge (MLS) requires higher-income families to purchase private health insurance or pay an additional tax. The problem here is the balance between fairness and effectiveness: while the surcharge increases incentives for wealthier individuals to purchase insurance, the initially set nominal limits failed to adjust for inflation, pushing more families into the penalty bracket over time. Later reforms adjusted the limits and introduced income-based surcharge brackets, increasing surcharge rates for very high-income individuals while simultaneously reducing the premium discount for these same groups. The authors point out that these effects created a complex policy environment in which incentives to purchase private coverage varied according to income bracket, and the effects of reforms could be offset depending on family circumstances.
In summary, the article points to three interconnected policy issues: (1) how to sustain participation in private health insurance in the context of a strong universal system, (2) how to avoid adverse selection while keeping premiums affordable, and (3) how to balance subsidies and tax penalties across income groups to ensure fairness and efficiency in reducing pressure on the public hospital system.
- Policy Implementation Context
The study is based on ten years of data (2004-2013) from the survey. Household, Income and Labor Dynamics in Australia (HILDA), a nationally representative longitudinal survey initiated in 2001. The survey collects comprehensive information on family structure, labor force participation, income, health insurance, health status, and expenditures by interviewing all family members aged 15 and older each year. Between 2004 and 2010, each annual wave included over 17.000 personal observations, and from 2011 onward, the sample increased to over 23.000 observations per year due to a supplementary sample. For analysis purposes, researchers focused on respondents from primary income units, excluding secondary units, individuals under 18 years of age, and responses with missing or ambiguous information. This resulted in an unbalanced panel of 101.670 observations across 18.407 individuals.
Thus, information on health insurance coverage in HILDA comes from two main sources. Starting in 2005, respondents were asked annually about their private health insurance expenses in a self-administered questionnaire, although this did not specify whether coverage included hospital care. However, since premiums for general care plans are typically low, the data allow researchers to differentiate between them. The authors explain that three research periods (2004, 2009, and 2013) included more detailed questions about health insurance, identifying the type of coverage (hospital, general, or combined) and whether the plan was individual or family. By combining expense data with these detailed survey responses, researchers were able to construct a consistent measure to determine whether an individual had private hospital insurance each year.
In this way, descriptive evidence shows that more than half of the sample had private health insurance, with higher rates among families compared to single individuals. Coverage generally increased between 2004 and 2008, but fell around 2008, coinciding with revisions to the Medicare Contribution Surcharge (MLS) income limits. This temporary decline was followed by renewed growth in coverage until another drop occurred after 2011. These trends provide the context for the study's analysis of how policy changes and economic conditions have influenced private health insurance participation in Australia.
- Evaluation Details
The study constructs an indicator of MLS Surcharge liability using income data from the HILDA survey. To align with the Australian Tax Office's definition of income for MLS purposes, gross income declared in the HILDA was adjusted accordingly. Based on this, a binary indicator was created to capture whether an individual's income exceeded the MLS income thresholds in a given year. Trends in MLS liability show that between 2004 and 2008, the proportion of individuals subject to the surcharge increased from 28% to 38%. However, when the MLS income thresholds were raised in 2009, liability dropped drastically to 24%, before stabilizing below the indexed thresholds in subsequent years.
The authors point out that despite these smooth aggregate patterns, household-level data reveal significant changes. According to them, before the policy change, about 14% of households that were not responsible for insurance in one year became responsible the following year due to increased income, but this number fell to 4% in 2008. Similarly, before 2008, about 16% to 18% of households exited responsibility each year due to reduced income. In 2008, however, almost half of the previously responsible households ceased to be subject to the surcharge, reflecting the impact of the increased limit. To assess how this policy change influenced private insurance uptake, researchers compared insurance coverage by income level in 2004 and 2009. The analysis showed that individuals with incomes between $50.000 and $70.000—subject to the surcharge in 2004 but exempt in 2009—experienced a noticeable decline in coverage. Meanwhile, coverage patterns for the highest and lowest income groups remained stable, suggesting that the decline in insurance coverage in 2009 was directly linked to the policy change.
The authors explain that, in addition to MLS liability, econometric models incorporate a wide range of covariates to capture the determinants of health insurance demand. Since income influences both surcharge liability and insurance adherence, the models include income, income squared, and interactions with household type. Demographic and socioeconomic controls encompass age, marital status, education, occupation, and health status, while behavioral variables include smoking and alcohol consumption as proxies for risk preferences. Geographic factors such as state of residence and distance are also considered. In this way, summary statistics indicate that the sample is predominantly female (54%), with a mean age of 48 years. Individuals with private insurance tend to be older, wealthier, healthier, and less likely to smoke compared to those without insurance, aligning with findings from previous Australian studies.
- Method
The econometric method used by the authors was developed to capture the persistence in private health insurance purchase decisions, while also addressing potential biases from unobserved preferences and policy-induced incentives. Purchase decisions tend to be highly persistent for two reasons: I) unobserved heterogeneity—such as risk tolerance or attitudes toward public versus private hospitals; and II) state dependence, where past choices directly affect current ones due to switching costs or inertia. To account for these factors, the authors employ a model... Dynamic data dashboard which includes lagged insurance decisions, fixed effects and relevant policy variables.
Formally, the model specifies Participation in Private Hospital Insurance as the dependent variable (in English). Purchase of Hospital Insurance – PHI) for the individual i in time t Based on their previous social security status, liability under the Medicare Surcharge (MLS), individual income and other characteristics, individual fixed effects and annual fixed effects. The main policy variable, responsibility for MLSThis captures whether an individual is above the income threshold that triggers the tax penalty for not having private insurance. The coefficient of this variable (γ – gamma) indicates whether MLS effectively increases insurance adherence. The lagged dependent variable (ρ – rho) captures the state dependency, where prior insurance status increases the likelihood of continued coverage in future periods. The authors emphasize that persistence implies that the short-term effects of MLS can accumulate into larger long-term effects.
To estimate this model, the authors use the Generalized Method of Moments (MGM), following Arellano and Bond (1991) and Blundell and Bond (1998). This approach addresses the endogeneity of the lagged dependent variable, using deeper lags in insurance participation as instruments. In this way, it also considers the possible correlation between current policy variables and past error terms. To ensure robustness, the authors test the serial correlation in the errors and the validity of the instrument using the test of Arellano-Bond and the test J de Hansen, respectively. To avoid the proliferation of instruments, they use the collapsed instrument matrix suggested by Roodman (2009).
Ultimately, the chosen specification includes three lags of the dependent variable, balancing parsimony with statistical validity. The two central parameters of interest are γ, which captures the contemporary effect of MLS policy on health plan enrollment, and ρ, which measures persistence through state dependence. Together, they allow the authors to distinguish between the immediate and long-term effects of MLS policy on the demand for private health plans.
- Main results
The study results show that the estimated effect of the Medicare Surcharge (MLS) on the demand for private health insurance (PSI) depends heavily on how persistence and unobserved heterogeneity are modeled. For example, in the naive MLS model, which ignores individual fixed effects and prior insurance status, the estimated effect of MLS is large and significant: being subject to the surcharge increases the likelihood of acquiring PHI by 11 to 14 percentage points. However, these estimates are upward biased, as higher-income households face MLS and already have a stronger inherent preference for private insurance.
In this way, when individual fixed effects are introduced, the estimated impact of MLS decreases dramatically, with γ (coefficient) falling to approximately 0,012–0,016, implying that MLS liability increases PHI participation by only 2–4% relative to reference rates. This correction reflects the importance of accounting for persistent differences in insurance demand among households. Subsequently, the dynamic fixed effects model further refines the analysis by controlling for both fixed effects and lagged participation in private health plans (PHI).
Thus, the results show a strong state dependency, where having private insurance in one year increases the probability of maintaining coverage the following year by 0,49, and insurance for two consecutive years increases the probability of coverage in the third year by 0,85. This persistence means that even modest short-term policy effects can accumulate into larger changes in the long term. Dynamic estimates of the MLS effect are approximately 1,3–1,5 percentage points in the first year, corresponding to a 2–3% increase relative to the average. For single individuals, the effect is similar but estimated with less precision.
In this context, dynamic simulations illustrate how the effect accumulates over time. A temporary MLS transit for one year increases PHI demand by 1,3 percentage points initially, with the effect slowly declining but remaining positive even 10 years later (+0,54 points). In contrast, a permanent MLS transit produces a sustained and increasing effect: after 10 years, PHI share is 7,1 percentage points higher than the baseline, equivalent to a 13% increase above the average. Thus, although the short-term effects are small, the long-term impacts of MLS on insurance demand are considerably greater due to persistence.
The authors emphasize that the results are robust when focusing on the 2007-2011 reform window, when MLS income thresholds were increased. In this subsample, the estimated effect of MLS liability is slightly stronger, with demand for PHI increasing by 1,8 percentage points (3,2% relative to the mean), consistent with the main findings, but suggesting somewhat larger policy impacts in the context of the reform.
Finally, the control variables behave as expected: insurance coverage increases with age (especially at age 31 due to the Lifetime Health Insurance policy), increases with income (albeit at a decreasing rate), and shows strong persistence over time. This reinforces previous evidence that both demographic incentives and policy penalties shape long-term PHI participation in Australia.
- Lessons in Public Policy
The study provides new evidence on the effectiveness of the Medicare Surcharge (MLS) as a policy tool to encourage enrollment in private health insurance plans in Australia. The authors find that MLS has a moderate, but statistically significant, effect on the likelihood of purchasing private health insurance plans. This aligns with the results of previous work on MLS and resembles results from studies on the similar individual mandatory coverage program in the US (Affordable care – ACA), suggesting that tax penalties linked to health plan coverage may modestly influence individual behavior across different health systems.
Thus, from a political standpoint, the results indicate that financial incentives can drive demand toward private health plans, but the broader question of whether such policies reduce pressure on public health systems remains unresolved. Previous studies in Australia, Spain, and the United Kingdom suggest that the fiscal costs of subsidies for private health plans often exceed the savings generated by reduced public spending. Therefore, while MLS modestly increases enrollment in private health plans, its overall effectiveness as a policy tool—particularly in terms of cost reduction and easing pressure on the public sector—remains uncertain and likely limited.
From an academic standpoint, the article contributes to the broader literature on incentives and demand for health plans, highlighting the role of persistence in purchasing decisions. Using a dynamic panel model, the study demonstrates that persistence arises from both unobserved individual heterogeneity and genuine dependence on the state, where prior insurance status strongly influences future decisions. The estimated one-year autoregressive effect (0,49) is substantially larger than in comparable studies, suggesting that once individuals opt for a private health plan, they are very likely to remain insured in subsequent years. This persistence implies that even modest short-term policy effects, such as those of MLS, can accumulate into significant long-term impacts.
Finally, the main implication is that static models underestimate the effects of policies because they ignore the cumulative influence of state dependency over time. Understanding why individuals remain persistently insured—whether due to switching costs, risk preferences, or other behavioral factors—remains an important area for future research. Ultimately, the study suggests that while MLS is moderately effective in changing behavior, the policy balance between fiscal costs and public system relief needs further analysis, and the dynamics of persistence should be incorporated into health insurance policy assessments.
References
Arellano, M. & Bond, S., 1991. Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. Review of Economic Studies, 58, pp.277–297.
Blundell, R. & Bond, S., 1998. Initial conditions and moment restrictions in dynamic panel data models. Journal of Econometrics, 87(1), pp.115–143.
Roodman, D., 2009. A note on the theme of too many instruments. Oxford Bulletin of Economics and Statistics, 71(1), pp.135–138.
