Modern government service delivery – 2018 Public Policy Essay Competition Winner

The Institute’s Public Policy Essay Competition attracted a host of young authors whose essays demonstrated that actuaries can apply their skills across a range of fascinating topics; from mental health to blockchain, road pricing, the gig economy, complexity in retirement, genetic testing and many more. The winning essay ‘Modern government service delivery’ by Hugh Miller was singled out for its clear communication, logical structure and innovative approach to a growing issue for governments – how to deliver important services within ever tightening budgets.

Background and recent trends

Government plays a significant role in securing the population’s welfare. Income support, the justice system, public housing, healthcare, education and child protection are all areas where we have asked government to assume responsibility for improving the lot of needy citizens.

Some government spending programs are new – for example, the roll out of the National Disability Insurance Scheme represents an important new support mechanism for disabled Australians. However, most of the systems listed above have endured for decades in a recognisable form. The Australian ‘welfare state’ including income support for the unemployed, disabled and single parents was rolled out in the early 1940s, and these supports remain today. Similarly, public housing expanded rapidly in the two decades after World War II and the policies behind the system (and many of the houses themselves) have changed little.

This stability in government service has done much to address the needs of the poor and vulnerable in society. But the world has not stood still; in recent times there have been some very significant changes that have profound implications for government services.

  1. Better data: Computer management systems for many government services has led to significant volumes of administrative data. Moreover, we usually have extended longitudinal data sets for services that extend back decades, often back to the 1990s. We can use this data to see long-term patterns of service use, including people who repeatedly require support. Moreover, we can see how such behaviours differ by segment; issues such as higher levels of long-term unemployment for Aboriginal and Torres Strait Islanders can only be properly understood when the data is collected.
  2. An increasing focus on outcomes: Many people have recognised that the amount of services delivered (e.g. how many cancer patients a hospital treats in a year) is not as important as understanding what outcomes are achieved (e.g. how many people go into remission). A focus on outcomes asks whether a service is genuinely effective in helping people. However, outcomes are usually more difficult to define, and measurement often requires more than standard administrative data. More challenging still is the recognition that equality in service provision does not lead to an equality of outcomes; for instance, children from low socio-economic backgrounds require significantly more support to achieve comparable educational outcomes to children from wealthy backgrounds (Lamb et al, 2015). This focus on outcomes has sharpened government thinking.
  3. Better prediction: We live in an age where prediction is easier and cheaper (Agrawal et al, 2016). With good data, computers can now make accurate predictions of future outcomes that account for many input characteristics. We can predict which job seekers will find employment, which prisoners will re-offend, or who will end up homeless. This means we can anticipate poor outcomes as well as observe them.
  4. The changing role of the not-for-profit sector: Before the government delivered services like welfare, health and education, it was charities and other not-for-profit organisations who supported the needy. Their role reduced as government spending increased, but we have seen a reversal of this trend in recent years. Some government services have been outsourced to not-for-profits (and sometimes, for-profits too). For example, in 1998 the Federal Government introduced the ‘Job Network’, a competitive environment where governments would pay third-party providers to assist job seekers into employment, a service previously provided by the public sector. The program aimed to improve competition and innovation, leading to better outcomes for job seekers. Nowadays third-party providers play important roles including prisons, homelessness services and public housing (termed ‘social housing’). This has also led to the increasing professionalisation of many of these organisations.
  5. Fiscal pressures: It is well-known that government has long-running fiscal challenges associated with the aging population; the old-age ratio (the number of people over 65 divided by the number aged 15-64) is currently about 23% whereas in 50 years it will be 38% (AIHW, 2015). However, other factors have contributed to increasing costs to government too.

Welfare: The rapid growth in the Disability Support Pension (DSP) population (up to 2011, when stricter tests were introduced) has increased welfare costs. Numbers on the DSP rose fourfold to 800,000 over the 30 years to 2011 (McVicar and Wilkins, 2013).

Housing: The opportunity cost of providing public housing (the difference between market rents and the subsidised rents paid by public housing tenants) has grown tremendously due to fast growth in property rents. Providing housing to those most in need adds to this, as they often earn less and require larger subsidies. For example, in NSW in 1960 only 15% of public housing tenants were on government benefits (as opposed to earned wages), compared to 94% today (FACS, 2014).

Crime and justice: While the incidence of crime has broadly decreased over time, prison populations have not. In Australia the prison population has grown from 26,000 to 40,500 over the decade to 2017 Q1 (ABS, 2017). Courts costs have also increased rapidly over time.

Healthcare: Cost inflation pressures in healthcare are well-known. Substantial increases in spending is due to a combination of growing costs (healthcare inflation has consistently run at 2.5% above CPI for decades, Miller 2016) and increasing community health expectations.

All these factors mean that governments are faced with the need to increase taxes or find ways to reduce the cost of service delivery.


What do these trends mean for government services?

First, there is more impetus to appropriately target services. Support should be given to those who most need it, and to those who can least afford to pay for it themselves. The welfare system in Australia has a high degree of income and means testing to restrict assistance to those most in need. Similarly, tax rules regarding the Medicare levy and private health insurance rebates have been designed to push wealthier people from the public to private system. Income targeting is useful for controlling fiscal costs, but ultimately has more to do with the allocation of resources rather than improvement in the way a service is delivered.

Second, there’s been a push towards prevention and return on investment. Evidence has been growing that some types of spending can be tied to improved outcomes years later. For example, there is significant evidence that early childhood education improves school performance, which in turn has been strongly linked to higher employment rates and income in adulthood. For governments, this means more tax and lower social security spending. The idea of early investments that produce positive fiscal impacts later has a natural appeal.

Third, there are opportunities to innovate in how services are delivered. Much recent excitement revolves around ‘impact investment’ where payments are tied to the outcomes achieved. The most sophisticated form of this is the social impact bond, where private investors contribute capital and assume the risk of a program’s effectiveness; if it succeeds, the government pays investors a higher return, justifiable as the outcomes have led to fiscal savings elsewhere. While there can be complexity in designing these bonds, they can offer private investment in social sector services, risk-sharing of outcomes, and disciplined measurement of the effectiveness of a program.


In this section we give recent examples of how these ideas are playing out in specific areas.

Income support system

Australia has one of the most sophisticated outcomes-based payment frameworks in the world for employment outcomes. Job seekers with work requirements are referred to third-party employment service providers. There are separate services for able-bodied job seekers and those with disabilities (who may have part-time work requirements depending on their assessed capacity). A large proportion of revenue for these providers is outcome-based; they receive government payments when they place job seekers in employment. These outcome payments are structured to improve longer-term outcomes:

  • The bulk of outcome payments are paid at 12 weeks and 26 weeks after job placement, and only if the employment is sustained. This encourages providers to appropriately train and place job seekers.
  • Outcome payments are larger for job seekers with greater barriers to employment. Job seekers are streamed based on their characteristics and allocated to different funding bands. This ensures providers still have an incentive to help those with the largest barriers to employment.
  • The streaming described in the second bullet relies on good information on barriers to employment. The program is rich in data; job seekers are interviewed to ascertain their barriers (such as housing instability, education and duration unemployed). Providers must maintain and submit good records to receive payments. As a result, there is high quality data for the program on which to base streaming rules and other policy decisions.
Public housing

Public housing is particularly amenable to long-range analysis, since durations in housing tend to be long and ongoing costs high. The New Zealand investment approach to social housing demonstrates some of the types of analysis that are possible (Taylor Fry, 2015):

  • There is notional potential saving for households that have more bedrooms than they need. This often occurs when adult children leave the household, or because the property was larger than needed initially. In New Zealand’s case, the theoretical saving from ‘right-sizing’ public housing is larger than the long-term cost of providing additional housing places for everyone on the waiting list.
  • There are often regional imbalances. In some places, less needy people can be housed in vacant properties while needier people have longer waits in other regions. Better targeting of new properties, informed by an understanding of public housing demand, can ameliorate this issue.
  • With average public housing durations measured in decades, there is also a recognition that there may better options for people with a real but temporary housing need. New Zealand has invested heavily in emergency and transitional support for such households, which can also reduce long-term fiscal costs.

Policing, courts and corrections can all benefit from analytics and long-term predictions. For courts, the progression of a case through the court system can be predicted in ways that courtrooms are used more efficiently. For corrections, prediction models of recidivism are already used in many jurisdictions internationally to help make parole decisions.

More broadly, criminal offending is highly concentrated amongst a small fraction of the population. One Swedish study found 1% of the population accounted for 63% of violent crimes (Falk et al, 2014).

Identification of these repeat offenders, particularly those predicted to be at high risk of future offending, enables better and earlier support, such as mental health or substance abuse assistance. Again, such programs are designed with the goal of improving outcomes and reducing future justice sector costs.


Government service delivery covers a wide range of programs with different challenges and different levels of maturity in their use of data and prediction. However, there is a clear trend towards sophisticated analysis to enable these services to be delivered better. Data and models will play an increasingly important part in how programs are targeted, measured and assessed for effectiveness.


  • Agrawal, A., Gans, J., & Goldfarb, A. (2016). The simple economics of machine intelligence. Harvard Business Review, 17.
  • AIHW. (2015). Australia’s welfare 20151.
  • Australian Bureau of Statistics, 2017, Corrective Services, Australia, cat. no. 4512.0,
  • FACS. (2014). Social Housing in NSW: A discussion paper. Department of Families and Community Services, NSW Government.
  • Falk, Ö., Wallinius, M., Lundström, S., Frisell, T., Anckarsäter, H., & Kerekes, N. (2014). The 1% of the population accountable for 63% of all violent crime convictions. Social psychiatry and psychiatric epidemiology, 49(4), 559-571.
  • Lamb, S., Jackson, J., Walstab, A., & Huo, S. (2015). Educational opportunity in Australia 2015: Who succeeds and who misses out. Mitchell Institute.
  • McVicar, D., & Wilkins, R. (2013). Explaining the growth in the number of recipients of the Disability Support Pension in Australia. Australian Economic Review, 46(3), 345-356.
  • Miller, H. (2016). The rise and fall of inflation indices. General Insurance Seminar, November 2016.
  • Taylor Fry (2015). Baseline valuation of the social housing system. Prepared for the Ministry of Social Development, New Zealand.

CPD: Actuaries Institute Members can claim two CPD points for every hour of reading articles on Actuaries Digital.