Mortality Leads and Lags

• Author: Maria Efthymiou,Andreas Milidonis
• DOI: http://doi.org/10.1111/jori.12187
• Date: 01 April 2017
• Published date: 01 April 2017

MORTALITY LEADS AND LAGS

Andreas Milidonis

Maria Efthymiou

ABSTRACT

Mortality risk varies geographically, especially in the Asia-Pacific (APAC)

region, where economic development is quite diverse. We present a newly

collected data set on aggregate population mortality from 11 countries in

APAC, which we rank based on their economic development. Using lead–lag

analysis, we identify short-term predictability in mortality risk across

countries. Mortality improvements seem to appear faster in more developed

than less developed countries. Such predictability is useful for longevity risk

financing and in producing cross-country mortality indices. We propose

ways in which our results can benefit institutions manage their exposure in

APAC mortality and longevity risk.

INTRODUCTION

Motivated by the high growth in the insurance and reinsurance industry in the Asia-

Pacific (APAC) region, and also the scarcity of available mortality data, this article

contributes to theliterature in the following three ways. First,we use a newly collected

data set of mortalityrates in the APAC region to analyze multipopulationmortality risk

at an aggregate population level, in a group of, frontier, emerging and developed

countries, representing about 42 percent of the world population. Second, given the

heterogeneity in economic development in APAC, we examine short-term, lead–lag

mortality dynamics (in the sense of Granger, 1969) across several populations. This

analysis allows us to identify intertemporal (across time) predictability among

Andreas Milidonis (contact author) is at the Department of Accounting & Finance, School of

Economics & Management, University of Cyprus, P.O. Box 20537, CY-1678 Nicosia, Cyprus.

Milidonis can be contacted via e-mail: andreas.milidonis@ucy.ac.cy. This work has benefited

from Milidonis’ former affiliation with the Insurance Risk & Finance Research Centre, Nanyang

Business School, Nanyang Technological University, Singapore. Maria Efthymiou is at the

Department of Accounting & Finance, School of Economics & Management, University of

Cyprus. Financial support by the Insurance Risk and Finance Research Centre (IRFRC), at

Nanyang Business School, NTU, Singapore, is acknowledged. We would like to thank two

anonymous reviewers for their comments and participants at the following seminars and

conferences for their comments: National Chengchi University, Taiwan; 18th International

Congress in Insurance, Mathematics and Economics (Shanghai, China); Longevity 10

Conference (Santiago, Chile).

© 2016 The Journal of Risk and Insurance. Vol. 84, No. S1, 495–514 (2017).

DOI: 10.1111/jori.12187

495

mortality time series, in countrieswith heterogeneous and rapidly changing economic

development. Third, the direction of predictability suggests that mortality improve-

ments appear with a lag in developing regions, thus opening up opportunities in

mortality and longevity risk management, not only within APAC but also for non-

APAC institutions considering expanding in this region (Blake, Cairns, and Dowd,

2003; Biffis and Blake, 2014; Blake et al., 2013; Blake, Boardman, and Cairns, 2014).

Therefore, we propose ways in which longevity risk transfer makes sense in a setting

where short-term predictability is present or is expected to appear.

The academic literature on multipopulation mortality risk is relatively new and

focuses almost entirely on non-APAC countries. Most research has focused on

developing models and improving existing models to capture the joint dynamics of at

least two populations (e.g., Li and Lee, 2005; Cairns et al., 2011; Dowd et al., 2011; Li

and Hardy, 2011; O’Hare and Li, 2012; Zhou et al., 2013; Chuang and Brockett, 2014;

Chen, MacMinn, and Sun, 2015). In addition, some articles have focused on

comparing existing models in the literature (e.g., Stallard, 2006). Another strand of the

literature is using well-known techniques from econometrics or other disciplines to

model multipopulation mortality (e.g., Hatzopoulos and Haberman, 2013; Lin, Liu,

and Yu, 2013; Wang, Huang, and Liu, 2013; Yang and Wang, 2013; Zhou et al., 2014).

In this article, we take a broader view and present a systematic comparison of

intertemporal flows in mortality changes from one country to another. We expect

mortality changes to be associated with each country’s degree of economic

development. For example, Pritchett and Summers (1996) and more recently Baird,

Friedman, and Schady (2011) show both long-term and short-term causal evidence

from income to mortality. Hence, as the APAC region is characterized by a wide

variation in economic development, we expect mortality improvements to appear

with a lag in less developed regions.

To test this hypothesis, we rank countries based on their economic development

following the rankings of three established ranking bodies: the World Bank (World

Bank, 2014), the Morgan Stanley Capital International (MSCI) indices (MSCI, 2014),

and the Financial Times Stock Exchange (FTSE) indices (FTSE, 2012). To refine the

economic development classification, we rank countries based on their gross national

income over time (a common criterion for the three ranking bodies). Then, we conduct

bivariate Granger-causality (henceforth BGC) regressions among the mortality time

series for the 11 countries in our sample, from 1950 onward. Our results are suggestive

of support for the hypothesis that mortality improvements would appear faster in

more than less economically developed countries.

We focus on short-term predictability instead of long-term mortality convergence

(Yang and Huang, 2011; Yang and Wang, 2013) for two reasons. First, we use a data set

with a relatively short horizon compared to recent improvements in life expectancy

that have been taking place. For example, life expectancy over the past century has

increased by about 30 years, while the gap in life expectancy between the more and

less economically developed countries has also been maintained at about 30 years

(Cutler, Deaton, and Lleras-Muney, 2006). Hence, searching for convergence in life

expectancy, or a long-term relation among countries with such diverse economic

development, is not justified by the literature.

496 THE JOURNAL OF RISK AND INSURANCE