Executive summary
The aim of this report is to provide guidance to regulators and supervisors of pension systems – as well as other relevant stakeholders – to develop and assess the mortality assumptions used in the context of the provision of retirement income. It looks at trends in mortality over the past several decades, including the impact that the COVID-19 pandemic has had, to provide context and inform future expectations of mortality trends. It also explains the different approaches taken to develop standard mortality tables for pensioners and annuitants, and summarises the tables used across OECD member countries. The report concludes with policy guidelines informed by international good practices.
It is important to understand the historical drivers of mortality to inform future expectations.
Changes in the drivers of historical improvements in mortality can provide insight as to what will happen in the future and help to inform modelling decisions to be in line with those expectations. For example, rapid economic development is often accompanied by accelerated improvements in life expectancy as developing countries catch up to their more economically advanced peers. Political and social drivers can also change the direction of trends, as the collapse of the Soviet Union did in Eastern Europe and the Baltic countries, or as the increase in drug overdoses have in the United States and others. Health policies and medical spending can drive improvements related to medical advances, such as the improvements in mortality from cardiovascular diseases that have driven strong reductions in mortality over several decades, though these gains are now slowing in many countries. Trends in social inequalities can also indicate whether life expectancy inequalities may grow or diminish over time.
The COVID-19 pandemic represented the largest shock to mortality and life expectancy in recent history. While mortality increased as a result of COVID-19 infections, particularly among those with comorbidities, the indirect impacts related to the response to the pandemic also had a significant effect on mortality. Mortality increased in many countries due to a lack of access to health care, either from a reluctance of patients to visit the doctor because of the risk of COVID-19 infection, or because of insufficient medical resources as medical facilities were overwhelmed by COVID-19 patients. Lockdown measures also impacted mortality, taking a toll on mental well-being and at times leading to detrimental behaviour such as substance abuse or physical violence. Some impacts were positive, however, with fewer fatalities from traffic accidents and other infectious diseases such as seasonal influenza. While the consequences of the pandemic are still being felt, its impact on mortality should nevertheless largely be temporary, and mortality levels are expected to normalise in the short term. Mortality assumptions used going forward should therefore reflect this expectation. However, there remains significant uncertainty around the long-term health impacts of COVID-19 and its effect on mortality. These effects will need to continue to be monitored to inform how they should affect mortality assumptions in the future.
Baseline mortality assumptions need to be representative of the population for which they will be used.
Baseline mortality assumptions, or the level of mortality for the base year excluding future mortality improvements, need to be based on data that is as similar as possible to the target population for which they will be used. Mortality levels vary significantly across population groups and the population(s) chosen to calibrate the model should reflect the characteristics of the target population. Populations of pensioners and annuitants tend to be skewed towards higher socio-economic groups, and therefore tend to have higher life expectancies than the population average. This selection effect represents around 2 to 2.5 additional years of life expectancy at age 65 on average in OECD member countries, and is therefore crucial to account for in the development of mortality assumptions.
Mortality improvements are essential to account for the expected increases in life expectancy over time.
Mortality improvements add an additional 1 to 1.5 years of life expectancy at age 65 on average, so are also crucial to take into account when assessing how long lifetime retirement incomes need to be paid. These assumptions should be based on a demographically stable population representative of the population to which they will apply. Mortality improvement assumptions are usually based on general population data, as there are often not sufficient data for pensioner and annuitant populations on which to accurately measure a long-term trend.
The model selected to project future mortality improvements should be compatible with future expectations, taking into account the trade-off between transparency and complexity. A variety of projection models are available, each of which demonstrates a range of advantages and drawbacks. Selection of the appropriate model will need to consider expectations regarding how mortality will evolve in the future and how to best match those expectations while aiming for parsimony.
The guidelines in this report can assist regulators and supervisors to develop and assess the appropriateness of mortality assumptions to ensure the sustainability of lifetime retirement income for pensioners.
The modelling of mortality and development of standard mortality tables involves a considerable amount of complexity and judgement at each step of the process. The guidelines included in this report aim to provide explanations regarding the implications of the different modelling choices available and their appropriateness for use in a particular context.