The impact of economic growth in mortality modelling for selected OECD countries

• Author: Malgorzata Seklecka,Athanasios A. Pantelous,Lydia Dutton
• Date: 01 April 2020
• Published date: 01 April 2020
• DOI: http://doi.org/10.1002/for.2640

Received: 12 February 2019 Revised: 30 July 2019 Accepted: 25 November 2019

DOI: 10.1002/for.2640

RESEARCH ARTICLE

The impact of economic growth in mortality modelling

for selected OECD countries

Lydia Dutton1Athanasios A. Pantelous2Malgorzata Seklecka1,3

1Department of Mathematical Sciences,

University of Liverpool, United Kingdom

2Department of Econometrics and

Business Statistics, Monash University,

Melbourne, Australia

3Group Accumulation Management,

Zurich Insurance Group Ltd, Fareham,

United Kingdom

Correspondence

Athanasios A. Pantelous, Department of

Econometrics and Business Statistics,

Monash Business School, Monash

University, WellingtonRoad, 20

Chancellors walk, Clayton VIC 3800,

Australia.

Email: athanasios.pantelous@monash.edu

Abstract

The health of a population is affected by social, environmental, and economic

factors. Pension providers and consultants, insurance companies, government

agencies and individuals in the developed world have a vested interest in under-

standing how the economic growth will impact on the life expectancy of their

population. Therefore, changes in death rates may occur due to climate and eco-

nomic changes. In this study, we extend a previous study into excess deaths as

a result of climate change to also provide a comprehensive investigation of the

impact of economic changes using annual female and male data for 5 devel-

oped OECD countries. Wefind that there is strong negative relationship between

mortality index, and climate and economic proxies. This model shows to pro-

vide better fitting and forecasting results both for females and males, and for all

countries studied.

KEYWORDS

economic change (GDP), longevity, climate change (temperature), mortality modelling, forecasting

MSC CLASSIFICATION

51; C52; C53; G22; G23; J11

1INTRODUCTION

Due to recent advances in sciences, humans are living,

on average, longer than ever before (e.g., Pitacco, Denuit,

Haberman, & Olivieri, 2009). As a direct consequence,

longevity is and will continue to be a key risk in the future

for governments and financial institutions (see, Barrieu

et al., 2012; French & O'Hare, 2014; Godinez-Olivares,

Boado-Penas, & Pantelous, 2016; O'Hare & Li, 2017a).

Therefore, appropriate mortality modelling and accurate

mortality forecasting are becoming increasingly important

(e.g., Lee & Carter, 1992; Lee, 1993; 2000; Lee & Miller,

2001; Renshaw & Haberman, 2003b; 2006; Currie, 2006;

Cutler, Deaton, & Lleras-Muney, 2006; Booth & Tickle,

2008; Plat, 2009; Li, Hardy, & Tan, 2009; French & O'Hare,

2013; Li, O'Hare, & Zhang, 2015; O'Hare & Li, 2017b;

Seklecka, Pantelous, & O'Hare, 2018, among manyothers).

Evidence of a significant relationship between economic

changes and trends in mortality for many developed

as well as developing countries has been presented by

many researchers (e.g., Ruhm, 2000; Preston, 2007; Tapia

Granados, 2008; 2011; 2012; French & O'Hare, 2014; Niu

& Melenberg, 2014; Rolden, van Bodegom, van den Hout,

& Westendorp, 2014; Seklecka, Md Lazam, Pantelous, &

O'Hare, 2019). In those studies, several macro-economic

determinants are used such as the unemployment rates,

Gross Domestic Product (GDP) or national income per

capita as exogenous determinants to capture the economic

changes in their models. Particularly, Brenner (2005) sug-

gests that economic growth not only reduces poverty

through an increase of real income, but also cause more

investments in new medicines, technologies and hospital

services, which may increase life expectancy. Results of

his time series analysis over medium- and long-term GDP

Journal of Forecasting. 2020;39:533–550. wileyonlinelibrary.com/journal/for © 2019 John Wiley & Sons, Ltd. 533

534 DUTTON ET AL.

display strong negative correlation between the time series

and mortality rates. In a similar direction, Swift (2011)

shows that GDP has a significant impact on life expectancy

for most of the 13 OECD countries considered in his study.

Furthermore, results for a wide range of developed and

developing countries suggest that middle-aged and older

people are much more vulnerable to economic fluctua-

tions in terms of their mortality (e.g., Ruhm, 2000; Tapia

Granados, 2011; Rolden et al., 2014).1

The main goal of this study is to investigate the relation-

ship between trends in mortality and trends in GDP per

capita change (as a proxy of the economic changes) using

the setup of the temperature-related mortality (TRM)

model proposed by Seklecka, Pantelous, and O'Hare

(2017); Seklecka et al. (2018) applying annual data over

the period of 1974–2013 for 5 developed OECD countries,

namely Australia, France, Germany, Japan and the U.K..

In addition, we explore the relationship between the time

dependent factor of the Lee and Carter (1992) model k1

t,2

and the logarithm of the average temperatureand GDP for

females and males at ages 0–89. Moreover we investigate

the long run relationship among time series data using the

Johansen (1988) cointegration test. Finally, we check cor-

relation between these factors according to various tests.

Our empirical analysis confirms that there is a rela-

tionship between mortality index, and climate and eco-

nomic proxies as strong negative correlation exists. The

GDP-related approach of the TRM model Seklecka et al.

(2017); Seklecka et al. (2018) outperforms all the mod-

els compared with, such as the celebrated Lee and Carter

(1992) (LC), the GDP-related model by Niu and Melenberg

(2014) (NM), the TRM by Seklecka et al. (2017); Seklecka

et al. (2018) both for male and female populations. For

completeness, in the comparison analysis we also con-

sider Plat (2009) (P) and O'Hare and Li (2012)(OL)

models.

The organization of this paper is as follows. In Section 2,

the relationship between economic growth and mortality

rates is explored. Section 3 focuses on the statistical anal-

ysis of the Human Mortality Database data. Results on

the long run relationship among time series data using

the Johansen (1988) cointegration test and also correlation

according to various tests are delineated. Based on those

1Certainly there are many factors which contribute to population mor-

tality levels (Cutler et al., 2006), such as political stability and cul-

ture/lifestyle choices. However,we will not consider those in the present

version of our paper.

2The Lee Carter model is given by

ln(mx,t)=b1

x+b2

xk1

t+𝜖x,t,

where b1

xand b2

xrepresent age effect, while k1

tis a random period effect.

The k1

tare estimated and refitted to ensure the model maps onto historic

data and the subsequent time series k1

tis used to forecast mortality rates.

results of Section 4, the GDP-related TRM model is intro-

duced, and the fitting estimation process is discussed. The

forecasting performance using ARIMA and Exponential

Smoothing processes are considered in Section 5. Finally,

Section 6 concludes the paper.

2ECONOMIC CHANGE AND

ENVIRONMENT

Since the mid-nineteenth century, a secular decline in

mortality rates is observed associated with long-term

improvements in economic conditions. The standard of

living has improved and economic instability and inse-

curity have been much reduced (Brenner, 1979). Clearly,

a strong and stable economy results in improvements

in healthcare, education, and social conditions which all

impact the mortality experience of a population. Some

studies suggest that in the long run, higher economic

output results in lower mortality (Preston, 2007; Barrieu

et al., 2012). The more immediate link between mortal-

ity and the economy was studied by Ruhm (2000) and

Tapia Granados (2008); Tapia Granados (2011) who find

that mortality rates increase during economic expansions.

Tapia Granados (2008) and Hanewald (2011) use both

real GDP and unemployment change as macroeconomic

indicators to study the links between these factors and

mortality changes for several developed countries. Particu-

larly, Hanewald (2011) concludes that the mortality index

of the Lee-Carter model, k1

t, and GDP levels are signif-

icantly correlated and that the mortality of various age

groups display signs of cointegration with GDP. Based on

these results, Niu and Melenberg (2014) propose a new

model that includes a real GDP factor (see also, Seklecka

et al., 2019).

Along with the literature above, studies in an envi-

ronmental economics research deliver various evidence

of the relation between economic growth and climate

changes, in particular,between GDP and temperature fluc-

tuations consider in (e.g., Dell, Jones, & Olken, 2009;

2014; Stern, 2007; Deschenes & Greenstone, 2011; Stern,

2013; Burke, Hsiang, & Miguel, 2015, among others).

IPCC (2014) report suggests that additional temperature

increase of around 2◦C may lead to losses of range between

0.2% and 2% of the Worlds real GDP. While, Burke et al.

(2015) find that climate fluctuation could reduce average

global GDP per capita by 23% by 2109 (mostly concen-

trated in poor countries). In other words, the effects of

climate change will not be uniformly distributed across

the globe and there are likely to be winners and losers

of it.

Figure 1 summarises the potential vulnerability to cli-

mate change on the World's map. In line with the eco-

nomic literature, many developing nations appear most