You are here

Working Papers

health model

Salvatory R. Kessy, Michael Sherris, Andrés M. Villegas and Jonathan Ziveyi

Abstract: We present a stacked regression ensemble method that optimally combines dierent mortality models to reduce the mean squared errors of mortality rate forecasts and mitigate model selection risk. Stacked regression uses a supervised machine learning algorithm to approximate the horizon-specific weights by minimizing the cross-validation criterion for each forecasting horizon. The horizon-specific weights facilitate the development of a mortality model combination customized to each horizon. Unlike other model combination methods, stacked regression simultaneously solves model selection and estimates model combinations to improve model forecasts. Our numerical illustrations based on 44 populations from the Human Mortality Database demonstrate that stacking mortality models increases predictive accuracy. Using one-year-ahead to 15-year-ahead out-of-sample mean squared errors, we find that stacked regression improves mortality forecast accuracy by 13% - 49% and 19% - 90% over the individual mortality models for males and females, respectively. Therefore, combining the mortality rate forecasts provides lower out-of-sample point forecast errors than selecting the single best individual mortality method. Stacked regression ensemble also achieves better predictive accuracy than other model combination methods, namely Simple Model Averaging, Bayesian Model Averaging, and Model Confidence Set. Our results support the stacked regression ensemble approach over individual mortality models and other model combination methods in forecasting mortality rates. We also provide a user-friendly open-source R package, CoMoMo, that combines multiple mortality rate forecasts using dierent model combination techniques.

Keywords: Stacked regression, ensemble learning, cross-validation, model uncertainty, model combination, age-period-cohort model, mortality forecasting.


Rob Bauer, Inka Eberhardt, and Paul Smeets

Abstract: To understand what motivates individuals to look at their pension situation and take adequate savings decisions, we conduct two field experiments with 226,946 and 257,433 pension fund participants. We find peer-information statements do not increase the rate at which individuals check their pension information, but lottery-type financial incentives do. Offering a few large prizes rather than many small prizes is most effective. However, the uptake of pension information does not lead to improved pension knowledge nor to increased self-reported savings three weeks after our intervention.

Mike Sherris CEPAR

Dilan SriDaran, Michael Sherris, Andrés M. Villegas and Jonathan Ziveyi

Abstract: Given the rapid reductions in human mortality observed over recent decades and the uncertainty associated with their future evolution, there have been a large number of mortality projection models proposed by actuaries and demographers in recent years. However, many of these suer from being overly complex, thereby producing spurious forecasts, particularly over long horizons and for small, noisy datasets. In this paper, we exploit statistical learning tools, namely group regularisation and cross validation, to provide a robust framework to construct such discrete- time mortality models by automatically selecting the most appropriate functions to best describe and forecast particular datasets. Most importantly, this approach produces bespoke models using a trade-o between complexity (to draw as much insight as possible from limited datasets) and parsimony (to prevent overfitting to noise), with this trade-o designed to have specific regard to the forecasting horizon of interest. This is illustrated using both empirical data from the Human Mortality Database and simulated data, using code that has been made available within a user-friendly open-source R package StMoMo.

Keywords: Mortality projection, regularisation, cross validation, Age-period-cohort model 




Seda Peksevim

Foreword: Today, people’s greatest financial concern is no longer paying their short-term bills or credit-card debt. According to the new study by Zurich Insurance Group and the University of Oxford (2019), ‘retirement security is the top financial worry’ for workers in 14 out of 16 countries. Likewise, recent surveys on old-age income suggest that nearly half of the respondents from different parts of the world do not feel secure about having a comfortable retirement (AARP Foundation, 2018; Credit Suisse, 2020).

While a lack of retirement savings has turned out to be a global phenomenon, most studies cover the design of pension systems in developed countries, which face relatively few challenges compared to developing ones. Moreover, from a handful of papers on developing regions, there is a tendency to discuss pension- related issues in the context of specific countries or topics. To this end, this study aims to provide an overall and detailed picture of the public and private pension systems in the developing world, including the present challenges and future directions.

The first part of the paper presents an overview of public pensions in developing countries. It illustrates the impact of ageing on sustainability and the adequacy of pay-as-you-go plans, along with some suggestions for the future of state pensions. In the second part, the paper focuses on private pension systems in the developing world and discusses the reasons for low pension savings with respect to the issues of coverage, contribution, and investment performance. This section also concludes by proposing certain recommendations for private pensions in the light of financial as well as behavioural and technological developments.

This work was made possible by the invaluable research support from the Pensions Scholarship Trust and IPE Magazine. Special thanks are due to Prof. Metin Ercan and Prof. Vedat Akgiray for their encouragement and guidance in my doctoral studies on pensions. I am also grateful to many researchers and colleagues from different parts of the world, notably Prof. Olivia Mitchell, Prof. Christopher Sier, Prof. Umut Çetin, Dr. Oğuz Karahan, Yaşar Kemal Peştreli, Joseph Mariathasan, Ziga Vizintin, Wojciech Sieczkowski, and Manfred Jormakka, for their fruitful discussion and valuable suggestions.




Michael Keane

Abstract: The Retirement income system in South Korea is a patchwork of different programs, none of which is particularly effective at reducing poverty among senior citizens. Program benefits are very poorly targeted, meaning a very large fraction of the elderly receive modest benefits that are generally inadequate to lift a houshold out of poverty. The consequence is that South Korea has the highest elderly poverty rate in the OECD. I will argue that a better targeted system - with more generous benefits aimed at the fewer recipients - is likely to be welfare enhancing.

However, it seems clear that elderly poverty in Korea cannot be addressed merely by reforming the retirement income system. To address the root causes of the problem, structural reforms are needed to break down Korea's dual labor market system. Under that system, a large share of workers are in informal jobs where they do not make or receive mandated retirement contributions.


Katja Hanewald

Martin Eling, Omid Ghavibazoo and Katja Hanewald

Abstract: We investigate the relationship between self-reported willingness to take financial risks and ownership of life insurance and long-term care insurance. For a representative sample of individuals aged 50+ from 14 countries and controlling for demographic and socioeconomic determinants of insurance demand, we find a positive link between willingness to take financial risks and ownership of both long-term care insurance and life insurance. The link is stronger for whole life insurance compared to term life insurance and long-term care insurance. Two robustness tests that (i) use risky asset ownership instead of willingness to take financial risks and (ii) focus on specific demographic and socioeconomic groups confirm the results for life insurance, while the results for long-term care insurance are less clear. Our empirical results cannot be explained by the classical expected utility framework and thus support recent research indicating that alternative models (e.g., prospect theory) are needed to explain insurance demand.

Keywords: Risk attitudes; Long-term care insurance; Life insurance; SHARE data

Mengyi Xu

Yu Fu, Michael Sherris and Mengyi Xu

Abstract: China and the U.S. are two contrasting countries in terms of functional disability and long-term care. China is experiencing declining family support for long-term care and developing private long-term care insurance. The U.S. has more developed public aged care and private long-term care insurance than China. Changes in the demand for long-term care are closely related to levels of and trends in mortality and functional disability. To understand future potential demand for long-term care, we compare mortality and functional disability experiences in both China and the U.S using multi-state latent factor intensity model to estimate time trends and systematic uncertainty in transition rates. The estimation results show that if trends continue, both countries will experience longevity improvement with morbidity compression and a declining proportion of the older population with a functional disability. Although the elderly Chinese have an estimated shorter life expectancy, they are expected to spend a smaller proportion of that future lifetime functionally disabled in contrast to the U.S. Systematic uncertainty is shown to be significant in future trends in disability rates and our model estimated higher uncertainty in trends for the Chinese elderly, especially for urban residents.

Keywords: Functional disability; life expectancy; systematic trend and uncertainty; multi-state latent factor intensity model


David Rodgers

Abstract: Older male labour force participation in Australia plummeted in the last three decades of the twentieth century. This paper investigates the extent to which this fall in participation was due to a large share of these older men having fought in World War II. Australian World War II veterans could access retirement and disability benefits that the rest of the population could not; they also had worse health than non-veterans. Sharp quasi-random variation in service across birth cohorts in Australia enables the effects of WWII service to be measured. WWII veterans had slightly lower labour force participation over prime working years (around 1 percentage point lower in their late 40s), and then much lower participation from the age of 60 onwards (around 17 percentage points lower in their early 60s). Sixty was the access age for the retirement pension given to veterans. Survey data indicate that the retirement expectations of veterans, when young, were clustered at this age. Various measurements of the health effects of WWII service indicate ill health is unlikely to explain the sharp fall in labour participation at sixty. Despite these findings, in aggregate, WWII service can explain only a modest share of the fall in older male labour participation in Australia in the late twentieth century, suggesting other factors must be responsible for most of the fall. These results contribute to the litera- ture on the labour supply effects of retirement and disability programs. In particular, they are consistent with earnings tests having large effects on participation and program parameters framing retirement decisions.

Warwick McKibbin

Roshen Fernando and Warwick J. McKibbin

Abstract: This paper updates the analysis of the global macroeconomic consequences of the COVID-19 pandemic in McKibbin and Fernando (2020c) with data as of late October 2020. It also extends the focus to Asian economies and explores four alternative policy interventions coordinated across all economies. The first three policies relate to fiscal policy: an increase in transfers to households of an additional 2% of GDP in 2020; an increase in government spending on goods and services in all economies of 2% of GDP in 2020; an increase in government infrastructure spending in all economies in 2020. The fourth policy is a public health intervention similar to the approach of Australia that successfully manages the virus (flattens the curve) through testing, contact tracing and isolating infected people, coupled with the rapid deployment of an effective vaccine by mid-2021.

The policy that is most supportive of a global economic recovery is the successfully implemented public health policy. Each of the fiscal policies assists in the economic recovery with public sector infrastructure having the most short-term stimulus and longer- term growth benefits.

Keywords: COVID-19, pandemics, infectious diseases, risk, macroeconomics, DSGE, CGE, G- Cubed