Many countries with dual public and private healthcare systems use financial incentives to encourage the take-up of private health insurance (PHI) either through government subsidies to those who buy, or a tax penalty to those who do not, or a mix of both (Colombo and Tapay, 2004). In Australia, since the late-1990s, the government has used both premium subsidies (known as rebates) and tax penalties (known as “Medicare Levy Surcharge”) to encourage people to buy PHI.

Currently in Australia both rebates and the Medicare Levy Surcharge (MLS) are means tested and share the same income thresholds for each tier. The means testing begins at incomes above $AU90,000 for singles and $AU180,000 for families. This means the policy mainly affects high income earners (in our administrative tax data, only around 16% of people are affected). There are three discrete thresholds at which the rebate (a fixed percentage of the premium) decreases, while simultaneously the MLS (a fixed percentage of income) increases (see Table 1 for precise details). The intent is for the increase in the MLS penalty (which increases the incentive to insure) to offset the reduced incentive to insure due to the lower rebate. We estimate the net effect of the MLS/rebate policy on demand for PHI, considering both the extensive (take-up) and intensive (premium spending) margins.

The Australian MLS/rebate policy is an interesting case study because it attempts to support demand for PHI in a fiscally (and equitably) mindful way. This will be increasingly important for nations dealing with ageing populations, increasing health care costs, and substantial fiscal constraints (see Lorenzoni et al., 2019, WHO, 2020). Government subsidies are very commonly used to encourage people to take-up PHI, but it is very expensive. In comparison, tax penalties are not as commonly used and under studied, and if they can be used effectively together with subsidies, more countries could consider implementing them.

Like many insurance markets, Australian PHI is subject to community rating, whereby everyone can purchase insurance for the same price, with insurers sharing the risk through a risk equalization scheme. While equitable for consumers, this setting can lead to a vicious cycle of adverse selection where healthier people, who are forced to pay substantially above actuarially fair rates for insurance, drop out, putting upward pressure on premiums, leading to further drop outs. Such a pattern was observed in Australia in the late 1990s and the MLS and rebate were introduced to address it Smits et al. (2022). However, rebates are expensive and in 2012 the new system of means-tested incentives was introduced to better target this scheme. This was controversial at the time – the opposition government opposed the measure and cited analysis by Deloitte that predicted that over five years “1.6 million Australians would drop cover and 4.3 million would downgrade their cover”.1 In reality, these claims turned out to be overstated, with the trend in coverage remaining fairly stable through this period, and the introduction of means testing associated with an overall small increase in take-up of PHI (Bilgrami et al., 2021). However, while the extent of downgrading was likely exaggerated, the policy does in theory create an incentive to downgrade. Our study is the first to investigate this empirically.

Another important feature of the Australian MLS/rebate policy is that it is ideally structured for empirical evaluation. The policy tier thresholds provide numerous sources of exogenous variation where the financial incentives to purchase insurance change, providing opportunities to understand how price responsive consumers are in a dual public–private system at different levels of income. The availability of large administrative data also allows us to obtain precise estimates for the effect of this policy.

To formally estimate the effect of the MLS and rebate changes, we estimate regression discontinuity design (RDD) specifications separately for each income threshold. We use a 10% random sample of all registered tax-filers in Australia from the 2017–18 financial year, which is the latest year available at the time of writing.

Intuitively, in an RDD design, we compare people with incomes just above and below the relevant threshold. We can obtain causal estimates if there are no other discontinuous changes in outcomes at the income threshold which could affect PHI coverage. In our setting, this assumption may not hold due to sorting on the running variable in an effort to avoid the penalty. To deal with this, we identify a window around the threshold where sorting occurs and remove those observations from the analysis Barreca et al. (2016), essentially extrapolating through that range of the running variable.

Our baseline estimates indicate that the net effect of the MLS and rebate reduction at the tier 1 threshold is an increase in PHI take-up of 1.1–1.4 percentage points (ppts) for families and 3–3.5 ppts for singles, but no robust effects at other thresholds. We also estimate a jump in premium spending of $44-$55 at the families tier 1 threshold and $124-$125 at the singles threshold, but again, no effect at other thresholds. Considering that at the tier 1 thresholds where the MLS kicks in, around 90% of families and around 70% of single people are already insured, the policy only explains a small fraction of demand. Moreover, the fact that people do not downgrade their cover, even at higher thresholds, where the withdrawal of the penalty provides a clear incentive to do so, suggests the high income earners affected by these these policies are highly price inelastic.