Data and Method. Method. The living standard at a specified point in time, (C/N) t, can be expressed using the following framework: C C Y L N Y L N

Transcription

1 The Cos of Living Longer: Proecions of he Effecs of Prospecive Moraliy Improvemen on Economic Suppor Raios for Foureen More Advanced Economies Nick Parr, Macquarie Universiy ( Jackie Li, Nanyang Technological Universiy ( Leonie Tickle (Macquarie Universiy ( Paper prepared for he 2014 European Associaion of Populaion Sudies (EAPS) Conference in Budapes, Hungary June 2014 Absrac The aims of his paper are hreefold; (1) o forecas moraliy for a wide range of more developed counries from ; (2) o proec he effecs of he forecas moraliy paerns on economic suppor raios assuming coninuaion of curren feriliy, migraion and labour force paricipaion; and 3) o calculae changes o labour force paricipaion which would offse hese effecs. The moraliy forecass are prepared for foureen counries using he Poisson Common Facor Model proposed by Li (2013). The moraliy forecass show ha he proeced gains in life expecancy are greaes in Japan, Ausralia and Eas-Cenral Europe, and are leas in Neherlands, Norh America and Sweden. Preliminary resuls for he proecions show ha he suppor raios are proeced o fall mos over he period o 2050 in Japan, Eas-Cenral and Souhern Europe, and leas in Sweden and Ausralia. However, excep for Poland, some recovery i suppor raios is proeced for he Eas-Cenral and Souhern European counries pos The dependency of he esimaed effecs of moraliy improvemen on suppor raios and he iniial age srucure and he assumed levels of feriliy, migraion and labour force paricipaion is discussed.

2 Inroducion The economic implicaions of prospecive populaion rends, paricularly he aging of populaions, are imporan concerns for conemporary more developed socieies (Kupiszewski e al. 2006, Ausralian Governmen 2002, 2007, 2010, UNPD 2010, Bloom, Canning and Fink, 2010, Saczuk, 2013). These economic implicaions include hose for public expendiure on pensions, social securiy, healh and aged care and on he proporionae size of he economically acive populaion. In his paper we resric our aenion o he laer: we consider prospecive changes in economic suppor raios for an exensive range of more developed counries. The prospecive age changes of populaions will be affeced by feriliy and migraion rends, boh pas and fuure, as well as by boh pas and fuure moraliy rends. The effecs of hese changes on economic suppor raios will also depend on prospecive age-specific paerns of labour force paricipaion. However i would appear ha he poenial effecs of prospecive changes in moraliy have no received he same aenion as eiher poenial variaions in migraion or poenial changes in feriliy. Therefore we focus on he effecs of prospecive moraliy change. We highligh he inernaional variabiliy of hese effecs. Such variabiliy is o be expeced boh because of he beween-counry variabiliy in proeced moraliy change and because of he inernaional variabiliy in base populaions, fuure feriliy, fuure migraion and fuure labour force paricipaion raes, all of which inerac wih he effecs of moraliy change. Jus as pas moraliy changes will affec fuure changes in populaion age srucure, and hence economic suppor raios, so fuure moraliy changes over a specified period of ime will affec suppor raios beyond he end of ha period. Hence we also consider he very long run implicaions of moraliy changes hrough consideraion of he sable populaions o which here would be convergence under he (purely hypoheical) mainenance of consan end-of-period values for all daa inpus, and model-based esimaes of he pahs of convergence (Gues and Parr 2014). Recen Moraliy Trends Our moraliy forecass are based on exrapolaive mehods, and will herefore reflec recen pas moraliy rends. In 2010 Japan has he highes overall life expecancy a birh, afer impressive increases for males (averaging 0.24 years per year) and especially for females (0.28

3 years per year) over he preceding 40 years ( Wang e al., 2012). Ausralia also saw moderaely fas increases over his period. In Europe, average improvemen since 1970 has been very fas in Porugal, moderae in Ialy, Swizerland, Spain, and he UK, and slow in he Neherlands and Sweden. Afer sagnaion or even for males declines over 1970 o 1990, life expecancy a birh increased dramaically from 1990 in he Eas-Cenral European counries Czech Republic, Hungary and Poland ( for males and for females since 1990); Biak (2013) oulines some of he healh policy and oher facors conribuing o hese changes. Improvemens since 1970 have been slow in he USA and Canada. I can be observed ha for mos counries, male improvemens have ousripped female improvemens. Daa and Mehod Mehod The living sandard a a specified poin in ime, (C/N), can be expressed using he following framework: C C Y L N Y L N (1) where indicaes he ime period and he oher symbols in his ideniy refer o he following naional aggregaes: C is consumpion of goods and services, N is he effecive number of consumers (eiher oal populaion or a needs-weighed populaion, for example as defined by Culer e al. 1990), Y is oupu of goods and services (gross domesic produc, GDP), and L is employmen (for example, oal hours worked). The raio C/N is consumpion per capia, C/Y is he consumpion share of GDP, Y/L is average labour produciviy, and L/N is he employmen o populaion raio. In his paper we resric our aenion o proecions of he suppor raio, L/N. The value of LN can be decomposed as he sum of age-specific componens:

4 (2) where H x, denoes hours worked per employed personin age and sex group x a ime and hence, E x, is he number of employed persons in age and sex group x a, S x is he number of people in of he labour force in age and sex group x a, and N x, is he populaion in age and sex group x a, and N is he oal populaion a. In his paper for each counry we consider proecions for he following measures of he suppor raio: (1) The raio of hours worked o oal populaion (2) The raio of hours worked o a (consumpion) weighed populaion, using he weighs proposed by Culer e al (1990) 1. Proecions of populaion N x, are prepared using he sandard cohor componen mehod (e.g. Pollard e al. 1990), and proecions of hours worked by applying assumed fuure age and sex specific labour force paricipaion raes S x, /N x,, employmen raes (E x, /S x, ), and mean hours worked per employed person H x, o he proeced numbers in he populaion by age. The effec of proeced moraliy change is calculaed by comparing he proeced values of L/N for a proecion using our bes esimaes of forecased fuure moraliy o hose produced by a proecion wih he same assumpions for feriliy, migraion and labour force paricipaion bu which also assumed moraliy raes remain consan a he levels for he base year of he proecion (i.e. 2010). The forecass of moraliy are developed using he Poisson Common Facor mehod of Li (2013). This is a coheren forecasing mehod, and was applied here o ensure coherence beween sexes, wih each individual counry forecased independenly (see Appendix A for deails of he mehodology). For all counries we firsly consider proecions over he period and comparison of proeced rends under he forecas moraliy change wih hose proeced wih moraliy consan a he 2010 level (and idenical assumpions for feriliy, migraion and labour force paricipaion). Secondly we consider he sable populaions o which he populaions would converge if he values of he feriliy, moraliy, migraion and labour force paricipaion were o 1 Culer e al. (1990) apply a weigh of 0.72 o he 0-19 populaion, 1.00 o he populaion and 1.27 o he 65 and over populaion.

5 remain consan a he 2050 levels used, and heir conrass wih hose generaed by moraliy a 2010 levels (Pollard 1973, Cerone 1987, Espenshade e al. 1982). This is because he proeced moraliy change beween 2010 and 2050, in addiion o conribuing o populaion change beween 2010 and 2050, will have implicaions for populaion change pos 2050.Finally we consider valuaions of he effec of he forecas change in moraliy over he enire long run pah owards he sable populaion, using a mehod developed by Gues and Parr (2014). The Gues- Parr mehod addiionally requires he formulaion of assumpions for a social discoun rae, ρ, which allow an equivalisaion of he social value of consumpion a differen ime poins and he very long run rae of produciviy growh, g (Samuelson 1958) (see Appendix B for deails of he Gues-Parr mehod). Values of 2.0 per cen per annum for ρ and 1.5 per cen per annum for g were adoped for all counries. We consider paerns for he following counries; Ausralia, Canada, Czech Republic, Hungary, Ialy, Japan, Neherlands, Poland, Porugal, Spain, Sweden, Swizerland, Unied Kingdom, and Unied Saes of America. The choice of counries was resriced o more developed counries for which all he requisie daa inpus could be idenified. We endeavoured o selec a range of counries which were diverse in erms of heir geographical regions and in heir recen moraliy rends, giving preference for inclusion o counries wih larger populaions over hose from he same region wih smaller populaions. Daa The moraliy forecass were based on daa from he Human Moraliy Daabase (HMD 2013). The populaion, feriliy, migraion, and labour force daa were sourced from he websies of a range of inernaional agencies (Unied Naions, OECD, Eurosa) and official naional daa sources. Assumpions Base Populaion The proporionae age disribuions of he base (2010) populaions differ significanly (Table 1). The Unied Saes, Ausralia, he UK, he Neherlands, Sweden and Canada have he highes

6 percenages aged under 20 years, whils Japan, he Souhern European counries (Ialy, Spain and Porugal), and he Eas-Cenral (Czech Republic, Hungary and Poland) have he low percenages in his age range. In conras, for he percenage of is populaion aged 65 or over Japan has he highes value, followed a a disance by Ialy. The Unied Saes, Ausralia and Poland have he lowes iniial percenages aged over 65. The range of variaion for he percenage of populaion in he main working age range (20-64) is fairly small. The percenages in his age range are highes for he Eas-Cenral counries and for Spain, and lowes for Sweden, and Japan. Feriliy The feriliy assumpion for all counries is ha all age-specific feriliy raes remain consan a he 2010 level. The assumed Toal Feriliy Raes are highes for Sweden, and for hree of he mosly English-speaking counries (Ausralia, UK and USA) (Table 1). Japan, he Eas-Cenral and Souhern European counries have he lowes feriliy levels. Ne Migraion For all counries annual ne migraion in oal and by age and sex was assumed consan a he average level for he period. Where ne migraion was posiive he absolue numbers of migrans by age and sex were used, whils for he wo counries wih negaive ne migraion, Japan and Poland, consan age-specific ne migraion raes were applied. The choice of he average migraion over his period as opposed o he level of he mos recen year was made in view of quie significan year-o-year volailiy in ne migraion. In erms of he absolue numbers he assumed ne migraion o he Unied Saes far exceeds ha for all oher counries considered. Spain and Ialy, Ausralia, Canada and he UK all had ne immigraion over 200,000. When viewed as a rae per 1000 populaion ne migraion is highes for Spain, Swizerland and Ausralia, followed by Canada, Ialy, and Sweden. Moraliy Proecion The Poisson Common Facor mehod of Li (2013) (Appendix A) was applied o each individual counry independenly, o produce forecass ha are coheren beween he sexes. The mehod 2 Following US Census Bureau (2013) in he apparen absence of suiable daa for he Unied Saes an assumed level of 725,000 per annum was applied o he proporionae age and sex disribuion of ne inernaional migraion for Canada

7 imposes coherence in proeced male and female moraliy for he same counry. Following Booh e al. (2002) he lengh of he fiing period used varied somewha beween counries. For some counries he overall moraliy rend (measured by K ) has changed quie considerably in he recen pas, paricularly for he Czech Republic, Hungary, and Poland. As a resul, for hese counries fiing periods around 20 years or even less have been used. Daa availabiliy can also impose resricions on fiing period. In all oher cases, a fiing period saring in eiher 1970 or 1980 was used depending on paerns in he moraliy rend. The number of addiional facors, n, was deermined separaely for each counry based on he crieria oulined in Appendix A, and proecion commenced from he acual raes in he final year of he fiing period. Labour Force Paricipaion Our labour force proecions used he average values over he period for age and sex specific labour force paricipaion, employmen, and hours worked per employed, shown in Table 2. Following he Global Financial Crisis (GFC) in 2008 here were significan increases in unemploymen raes in some counries, mos noably hose in Souhern Europe. Thus he values of labour force paricipaion measures we use represen averages of pre and pos-gfc values. Age-sandardised measures of labour force paricipaion are used in order o purge he poenially significan effecs of he beween-counry variaion in populaion age disribuions from he comparisons. The key values o noe in Table 2 are he sandardised mean hours per person aged 15 and over in he wo righ-hand columns. These values are weighed sums of he producs of agespecific labour force paricipaion raes, employmen raes and he mean hours worked per employed person. Unsurprisingly for every counry we sudied he average hours worked per male exceeded he average per female. Surprisingly he range of variaion beween counries for males is greaer han for females. For males Japan has by far he highes average number of hours per person. This reflecs is high paricipaion rae, low unemploymen and an unusually high average hours worked per employed male (across some age groups he averages exceed 48 per week). The English-speaking counries, Swizerland and he Czech Republic also have high hours worked per person. Excep for he Czech Republic hese counries have above average male labour force paricipaion raes, all have above average hours worked per male employed, and in he case of Swizerland and Ausralia low unemploymen raes are also a facor. Hungary,

8 Ialy, Poland, and Spain have he lowes hours worked per male. This is he produc of heir lower male paricipaion raes, higher unemploymen raes and lower mean hours per employed. Swedish females have he highes mean hours worked per person, a paern driven by heir much higher paricipaion rae (Table 2). The USA, Canada and Porugal also have high average hours per female. For USA and Porugal females i is due o a relaively high paricipaion rae and a relaively high hours average worked per employed female. For females in Canada i is driven by relaively high paricipaion rae and a slighly higher han average employmen rae. Ialy has he lowes hours worked per female, primarily due o a much lower labour force paricipaion rae. Despie a high female paricipaion rae and a low unemploymen rae, he Neherlands has one of he lowes hours worked per female. This is driven by an unusually low average hours per employed, a paern linked o widespread par-ime work. Spain also has a low average hours per female, due o is low paricipaion rae and high unemploymen rae. There are some considerable beween-counry differences in he paern of variabiliy of hours worked per person by age (no shown). For boh males and females in he age group Swizerland has he highes average hours per person, wih Ausralia, Canada and he UK also having relaively high values. Swizerland, Japan, Porugal, Ausralia, he UK and he USA have he highes values for males and females aged For males aged beween aged 25 and 59 he hours worked by Japanese men far exceed hose for he oher counries. For females Porugal has he highes hours worked over he main reproducive ages (i.e ), whils for hose beween he ages of 40 and 59 he Czech Republic, and Sweden have he highes raes. For males in he age range Japan, Swizerland and Sweden have he highes values, whils for females in his age range Sweden, and he USA have he highes values, followed by Japan. For boh sexes Japan and he USA have he highes hours worked for males aged over 65. Resuls Moraliy Proecions A he sar of he proecion period in 2010 for boh sexes he greaes differences in life expecancy a birh are beween he Easern European counries and he oher counries, wih he former having significanly lower values (Table 3). Japan has he highes life expecancy a birh

9 for boh sexes. Life expecancies for he Unied Saes, whils higher han hose for Easern Europe, are noiceably lower han hose for he Wesern European counries and he oher mosly English-speaking counries. The resuls of he moraliy proecions show ha in all he counries we consider beween 2010 and 2050 he increase in male life expecancy a birh is expeced o be greaer han ha for females (Table 3): convergence beween male and female moraliy is expeced due o he use of he Poisson Common Facor Model. Excep for males in he USA, he proeced increases in life expecancy are greaer beween 2010 and 2030 han beween 2030 and For boh sexes he greaes increases in life expecancy a birh are proeced o occur in Japan, followed by hree Eas-Cenral European counries (Table 3, Figures 1 and 2)). The proeced gains in life expecancy in Ausralia and he UK are also relaively large. For boh sexes he Neherlands has he smalles proeced increase in life expecancy a birh. The proeced gains in life expecancy a birh are also relaively small for he Norh American counries and Sweden for boh sexes, and for Spanish females. Figures 1 and 2 show he widening gap beween he proeced life expecancies a birh for Japan and hose for he oher counries. For females he proeced life expecancy a birh in 2050 for Japan is proeced o be a huge 5.3 years greaer han he nex highes counry (Ausralia). By 2050 for boh sexes Ausralia is proeced o have significanly higher life expecancy a birh han all he European counries we consider. The Eas-Cenral European counries are proeced o sill have lower life expecancies in 2050 han mos of he Wesern European counries for boh sexes, alhough he gaps will be significanly smaller. However for boh sexes in 2050 life expecancy a birh for he Czech Republic is proeced o exceed hose for he USA, he Neherlands, and Porugal, and for females life expecancy a birh is also proeced o be greaer in Poland han in he Unied Saes. For Japan, relaively seep decreases in cenral deah raes are proeced for all age groups and boh genders. For mos ages increases in male o female age-specific moraliy raios are proeced, wih he excepions being a he wo exremes of he age range (under 10 and over 90). For he Eas-Cenral European counries moraliy declines are proeced for all age and sex groups. However he reducions in he log(m x ) are proeced o be considerably more rapid in he younger ages han in he older ages. Male o female raios for moraliy raes are proeced o remain more or less consan for all age groups. The proeced paerns of moraliy improvemen

10 for Ausralia and he UK exhibi similariies wih he fases decreases in log(m x ) occurring below age 20 and in he age range (more so for year olds in Ausralia), wih lile change for year olds. For he USA and Canada deerioraing moraliy is proeced for he over 90 age range. For he Neherlands for boh sexes proeced reducions in log(m x ) are greaer in he younger ages han in he older ages, and male o female raios in age-specific moraliy are proeced o increase slighly. Table 4 shows ha for boh sexes life expecancy a age 65 is proeced o increase mos in Japan. Of all he counries we consider he USA has he smalles proeced increase in life expecancy a age 65 for males, and for females only he Neherlands has a smaller proeced gain han he USA. Proeced Changes in Suppor Raios o 2050 In 2010 Japan, Swizerland, and he Czech Republic have he highes suppor raios (Table 5). These paerns reflec he relaively age-specific high hours worked per person in hese counries (Table 2). For he Czech Republic a favourable age srucure is also a facor, whils in Japan he high iniial hours worked o populaion raio is despie is proporionaely very large over 65 age group (Table 1). For boh forms of he suppor raio, Ialy has he lowes value by a significan margin, a paern which reflecs is relaively low hours worked per person for males and paricularly for females, as well as is large relaive populaion aged over 65. The Neherlands, and Hungary also have relaively low suppor raios. These counries all have low sandardised hours worked (Table 2). Beween 2010 and 2050 he suppor raios are proeced o decrease in all he counries we consider. This resul is driven by he proeced ageing of hese populaions over his period. For every counry he decrease according o he consumpion needs-weighed measure is greaer han ha for he unweighed measure. Poland and Japan are proeced o experience he greaes decreases for boh forms of he suppor raio over he period 2010 o 2050 in absolue erms (Table 6). This reflecs he rapid ageing produced by very low feriliy, rapidly improving moraliy, paricularly in he older ages, and ne oumigraion, paricularly by young aduls. Two Eas-Cenral European counries, he Czech Republic and Hungary, also have relaively large decreases in suppor raios, as do he hree Souhern European counries. These are all counries wih very low feriliy. In 2010 his

11 very low feriliy would have been beneficial o he suppor raios, by reducing he proporionae size of he child and young adul ages. By 2050 he effec of very low feriliy exends more widely across he life span, and any beneficial effecs on he suppor raio accordingly are much reduced. Sweden has he smalles proeced decrease in he suppor raio. The USA, Ausralia, he UK and he Neherlands are also proeced o have relaively small decreases (Table 6). These counries share a paern of relaively high feriliy and significan ne immigraion. A slower rae of populaion ageing combined wih relaively high raes of labour force paricipaion in he older ages are oher conribuory facor behind Sweden s mainenance of is suppor raio. For mos counries he proeced reducion in he suppor raio is greaer beween 2010 and 2030 han beween 2030 and The excepions are he hree Eas-Cenral European counries, Porugal, Spain and Japan. By 2050 Swizerland and Japan are proeced o have he highes labour force o populaion raio, followed by he Czech Republic, USA, Ausralia, and Sweden (Table 5). These counries have relaively high age-sandardised labour force paricipaion raes (Table 2). The proeced suppor raio defined relaive o he age-weighed populaion for Japan is significanly worse han for he oher afore-menioned counries, due o is much older proeced populaion. Ialy is proeced o sill have he lowes value for boh forms of he suppor raio. This is linked o is very low raes of labour force paricipaion, especially for females. However, by 2050, as a resul of is large proeced reducions beween 2010 and 2050, Poland is proeced o have he second lowes value for boh forms of he suppor raio. Hungary, Spain and he Neherlands are also proeced o have relaively low proeced suppor raios. Suppor Raios for Terminal Sable Populaions Since he forecas moraliy declines no only affec suppor raios during he 2010 o 2050 period bu also will affec hose beyond 2050, Table 5 also presens he suppor raios for he sable populaions produced by he assumed levels of feriliy, migraion and labour force paricipaion and wih moraliy a he forecas level for 2050 and Table 7 he differences beween hese suppor raios for he erminal sable populaions (TSP) and he 2010 and 2050 values. For nine of he foureen populaions we consider he sable populaion suppor raios are

12 below hose for The counries in which greaes reducions in he suppor raios beween 2050 and he erminal populaion, Ausralia and Sweden have relaively high feriliy and ne immigraion, and hence relaively young proeced populaions. The counries for which he raios for he erminal sable populaion are higher han for 2050, Czech Republic, Hungary and Ialy and Porugal, all are counries wih very low feriliy, low bu posiive ne immigraion and relaively rapid moraliy improvemen which is concenraed in he younger ages. The reducions in suppor raios over he enirey of he ransiion beween 2010 and he erminal sable populaion are larges, boh in absolue and in percenage erms, for he wo counries wih very low feriliy, ne oumigraion and rapid moraliy improvemen, Poland and Japan. In Poland he percenage decreases (of 24.6 per cen for he unweighed L/N and 30.9 per cen for he weighed measure) are paricularly large. However, due o he proeced pos 2050 recoveries in suppor raios, he differences beween he suppor raios of he iniial and erminal are lowes for wo oher counries wih very low feriliy, Porugal and Hungary. As a resul of he recovery in is value he Czech Republic has he highes suppor raios for is erminal sable populaion, followed by Porugal, USA and Swizerland. Japan ranks only fifh. Ialy sill has he lowes suppor raios. Due o he large proeced reducion, Poland has he second lowes suppor raio for is erminal sable populaion. The Neherlands and Spain also have relaively low values. The Effecs of Moraliy Improvemen on Suppor Raios Table 7 shows ha in all he counries we consider proeced moraliy change, according o our bes esimaes, conribues o he proeced reducion in he suppor raio over he period from 2010 o The proeced effecs of moraliy improvemen are greaes for he Czech Republic and Japan. These are also he wo counries in which he forecas increases in life expecancy are greaes (Table 3). In Japan he proeced improvemens in moraliy a older ages are especially large relaive o hose for oher counries, and his conribued o he derimenal effecs of moraliy improvemen on he suppor raios. Since labour force paricipaion by Japanese females is much lower han for males, he paricularly large forecas improvemens in female moraliy are also a facor in he reducion of he suppor raio. Moreover he effec of his moraliy improvemen is magnified a populaion age srucure which even in 2010 is he oldes

13 in he World and which ages very rapidly largely. The effecs of proeced moraliy improvemen are also relaively large for Poland and Hungary, boh also counries in which he forecas improvemens in life expecancy are large and which are proeced o age rapidly. In absolue erms Sweden, USA, Canada, he Neherlands and Ausralia have he smalles proeced moraliy effecs on heir suppor raios. Excep for Ausralia, he forecas improvemens in life expecancy in hese counries are relaively small. All hese counries have somewha younger populaions and slower raes of ageing which are linked o relaively high feriliy and ne immigraion by young aduls. Their relaive youhfulness elevaes he effecs of moraliy improvemen on he working ages relaive o hose on he pos-reiremen ages. The proporionae conribuion of he forecas moraliy improvemens o he reducions in he suppor raios is greaes for he Czech Republic (44.8% for he reducion in L/N and 41.6% of he reducion for L/N*) and for he UK (42.1% and 43.5% respecively) and Hungary (40.4% and 37.7%). For he Czech Republic and for Hungary hese are proporionae o large oal reducions in he suppor raios, whils for he UK hey are relaive o a small reducion. The conribuions of moraliy are smalles for Spain, Sweden, Canada, Porugal and Poland. The firs hree of hese counries have relaively small forecas improvemens in life expecancy. The laer wo have very low feriliy and an absence of immigraion. For all counries and boh forms of he suppor raios he proeced effec of moraliy change are negaive boh for he 2010 and for he 2030 o 2050 period, excep for Sweden for he unweighed measure for 2030 o For all counries excep for Sweden he suppor raio lowering effecs of moraliy change are greaer beween 2030 and 2050 han beween 2010 and This would be linked o he effecs of moraliy change becoming increasingly concenraed on he pos reiremen ages and he increasing weigh placed on he pos-reiremen age groups due o populaion ageing. Gues-Parr Valuaions of he Effecs of Moraliy Using he mehod of Gues and Parr (2014) we evaluae he effec of he forecas moraliy changes beween 2010 and 2050 condiional on feriliy, migraion and labour force paricipaion remaining a he assumed levels hroughou fuure ime, on here being no furher moraliy improvemen pos 2050 and assuming a ne social discoun rae of 1.5% per annum.

14 The values in he Toal Social Value column of Table 9 represen he effec of he proeced moraliy change over he enire ransiion owards he erminal sable populaion as a muliple of he consumpion value of one hour worked per week per capia (i.e. ignoring differences beween counries in labour produciviy). By his measure he effec of he proeced size of he effec of moraliy change is greaes for Japan, followed by he hree Eas-Cenral European counries. The proeced effec of moraliy change is smalles for he Unied Saes, he Neherlands, Canada and Sweden. The values in he fourh column of Table 9, which show he moraliy effecs expressed as percenages of he value of he consan moraliy series, indicae he (immediae and susained) percenage increases (relaive o he average level) in labour force paricipaion which would compensae for he effec of he forecas change in moraliy, if he percenage increases were uniformly spread across all age and sex groups. The greaes value is 7.8 per cen for Poland. This indicaes ha in Poland an increase of labour force paricipaion on 7.8 per cen would be needed o compensae for he forecas moraliy improvemen. Japan has he second larges (relaive o is higher paricipaion rae) value in percenage erms, followed by he wo oher Eas-Cenral European counries. The Norh American counries have he smalles percenage changes in labour force paricipaion required o offse forecas moraliy improvemen, followed by he Neherlands and Sweden. The final column of Table 9 effecively shows he moraliy effecs as a percenage of he 2010 GDP, and serves o indicae he very large magniudes of hese effecs if susained over he very long run. Conclusions Whils for all he counries we consider life expecancies, boh for males and for females, are forecas o increase, he forecas paerns of improvemen are diverse. The gap beween he highes (Japan) and lowes (Hungary) life expecancies a birh is forecas o increase over ime. Life expecancy for Ausralia is forecas o exceed hose for European counries, wih he differences increasing over ime. Wihin Europe our forecass show a narrowing of he life expecancy gaps beween Wesern European counries and Eas-Cenral European counries, wih life expecancies for he Czech Republic exceeding hose of wo Wes European counries by

15 2050. The forecas increases in USA life expecancy are relaively slow and our bes esimaes indicae ha by 2050 hey may even fall below hose for he Czech Republic for boh sexes and for females also for Poland. The forecas paerns of improvemen by age are similarly diverse. These forecass are essenially based on exrapolaions of recenly observed rends. The proecions show ha, in he absence of changes in paricipaion raes, feriliy or migraion, economic suppor raios will decrease in all he counries we consider. However he magniude of he decrease varies considerably beween counries, wih he larges proeced reducions (for Poland and Japan), depending on he choice of measuremen of he suppor raio, being wo or hree imes more han double imes he smalles reducions) over he very long run. The larges reducions are proecions for counries which assume he coninuaion of very low feriliy and also ne oumigraion and herefore in which annual birhs decrease indefiniely. Whils over he period o 2050 he reducions in suppor raios are generally greaer in very low feriliy counries, for hose which also have ne immigraion recoveries in he suppor raios are proeced pos 2050, as an immigraion-creaed floor for birhs is creaed and differences in populaion numbers beween age ranges increasingly become smaller and he produc of immigraion and moraliy. Thus he proecions indicae he poenial long-run benefi of posiive immigraion for he age srucures of counries wih persisenly very low feriliy. They also show he poenial for analyses which only consider he more immediae fuure (eg 40 years) o creae a misleading impression of prospecive suppor raio rends for very low feriliy counries. Our paper is disincive in is aemp o isolae he effecs of forecas improvemen in moraliy on suppor raios. The proecions show, unsurprisingly,hahe effecs of he forecas improvemens in moraliy are generally o decrease suppor raios. For all he counries we consider he effec of forecas moraliy improvemen conribues less han half he proeced reducion in he suppor raio beween 2010 and However, he effecs of moraliy will increase in imporance over ime as he populaion ages. Anoher disincive feaure of our paper is is applicaion of a mehod, recenly developed by Gues and Parr (2014), o provide a single-figure index of he cumulaive effec over ime of moraliy change on he suppor raio. The resuls illusrae he large size of he moraliy effecs cumulaed over he very long run and some of he combinaions of changes in labour force paricipaion which could compensae for hese effecs. The changes can be calculaed condiional on he disribuion of paricipaion rae changes beween age groups and he gradien

16 of he changes over ime. A number of more developed counries, including Ausralia and he UK, have recenly sough o increase reiremen ages and labour force paricipaion more broadly in response o prospecive increases in longeviy and populaion ageing more generally (Ausralia 2010). The mehod used can evaluae he magniude of increases which are needed. I is imporan o noe ha he proeced effecs of moraliy are condiional on he assumed fuure levels of feriliy, migraion, labour force paricipaion, he iniial age srucure of he populaion and he assumed raes of discoun: for any one counry he effec of moraliy improvemen would be smaller if labour force paricipaion raes in he laer working ages increased or if ne immigraion increased. Similarly he changes o paricipaion which could counerbalance he moraliy changes also depend on he values of he oher demographic and economic variables in he model.

17 Acknowledgemens This research was suppored by an Ausralian Acuarial Research Gran provided by he Ausralian Acuaries Insiue. We graefully acknowledge research assisance provided by (Ree) Yongqing Chen, Amy Lo, and Daniel Ciarliero.

18 Technical Appendix A: The Poisson Common Facor Mehod for Forecasing Moraliy Coherenly The moraliy proecions were prepared using he Poisson Common Facor mehod of Li (2013). This appendix summarises he specificaion of his model. Considering a paricular counry and assuming ha he force of moraliy is consan wihin each age-ime cell and so is equal o he cenral deah rae, Li (2013) uses: D x,, i Poisson e x,, i m x,, i ~, (1A) n ln m a B K b k, (2A) x,, i x, i x 1 x, i,, i, in which D x,, i, e x, i,, and m x,, i are he number of deahs, cenral exposed-o-risk, and cenral deah rae a age x in year for sex i (i = 1 for females and i = 2 for males). The parameer depics he overall moraliy schedule across age for sex i, sexes, and b x i, k, i, B K x a x, i is he common facor for boh, is he h addiional sex-specific facor. The common facor represens he main long-erm rend in moraliy change for boh sexes, and he addiional facors represen he shor-erm discrepancy from he main rend for each sex. More specifically, K is he imespecific moraliy index of he common facor wih and k i, B x as he age-specific sensiiviy measure,, is he ime componen of he h addiional facor wih b x, i, as he age-specific sensiiviy measure. There are a oal of n addiional facors. The Lee-Carer model as applied wihin a Poisson framework (Brouhns e al, 2002) can be regarded as a special case of he PCFM, in which he common facor is se o zero and here is only one sex-specific facor (i.e. n = 1). The parameers are esimaed by (condiional) maximum likelihood via an ieraive x updaing scheme, under he consrains B 1, K 0, b 1, and k 0. x x x, i,, i, The number of addiional facors, n, is deermined based on he Bayesian Informaion Crierion (BIC) values, he residual plos, he rends of he addiional parameers, and he amoun of daa being sudied. The moraliy index K is modelled by a random walk wih drif as usual: K 1, (3A) K

19 2 where is he drif erm and ~ 0, is he error erm. The compued drif erm is ofen negaive, which indicaes a general moraliy improvemen over ime. The ime componen of an addiional facor is modelled by an auoregressive (AR) model of order p: where k, i, 0, i, 1, i, k 1, i, 2, i, k2, i,... p, i, k p, i,, i, 0,i,, 1,i,,, p, i,, (4A), i, ~ i, 2 are model parameers and 0, Fuure deah raes (in year > T) are hen proeced as: n Kˆ K b kˆ k is he error erm. mˆ x,, i mx, T, i exp Bx T x, i,, i, T, i,, (5A) 1 in which he proecion sars wih he laes se of daa in year T. The resuling male-o-female raio of deah raes (in year > T) can be expressed as: mˆ mˆ m m exp kˆ k b k k x,,2 x, T,2 n b ˆ x,2,,2, T,2, x,1,,1, x,,1 x, T,1 1 T,1,. (6A) This raio converges o a consan if he proeced ime componens ˆ and k ˆ,2, converge as k,1, well. Accordingly, he choice of he AR(p) model s order is based on (a) he parial auocorrelaion funcions (PACF) of he ime componen; (b) he auocorrelaion funcions (ACF) of he residuals; and (c) wheher he ime componen approaches a consan in he proecion. The las crierion ensures ha he deah raes of females and males a each age do no diverge in he long run.

20 Technical Appendix B: The Gues-Parr Mehod for Socially Evaluaing Very Long Run Demographic Pahs Gues and Parr (2014) propose a mehod which may be used o evaluae very long run effecs of conrass in demographic processes and heir effecs on populaion age srucure and of labour force paricipaion on living sandards, defined as naional consumpion per capia. The framework i uses is as follows: C C Y E N N Y L N N x, x, H x, (1B) x x, where C is consumpion of goods and services a ime, N is he effecive number of consumers (eiher oal populaion or a needs-weighed populaion (eg as defined by Culer e al or according o age profiles of consumpion for he relevan populaion see Naional Transfer Accouns Proec (2014)), Y is oupu of goods and services (gross domesic produc, GDP), and L is employmen (for example, oal hours worked). The raio C /N is consumpion per effecive consumer, C /Y is he consumpion share of GDP, Y /L is average labour produciviy, and L /N is he employmen o populaion raio or suppor raio, which is equal o one minus he oal dependency raio. H x, denoes hours worked per employed person in age and sex group x a ime and hence L, H, E, where E x, is he number of employed persons in age and sex group x x x x x, and N x, is populaion in age and sex group x. In (2) he populaion age-srucure affecs living sandards hrough he age-specific variaion in H x, and E x, /N x,. In order o isolae he role of L /N,, C /Y is assumed o be invarian o populaion age srucure and Y /L grows a a consan rae, g, which is deermined independenly from he age composiion. Hence he living sandard a is:

21 C C L N L N 0 1 g (2B) 0 The value of a proeced populaion over he enire pah of fuure values of is: V (3B) C 0 1 g L L0 0 1 N ρ is a social discoun rae which equivalises living sandards for fuure ime poins o values a he sar poin for he proecion, 0, and depends on udgemens abou he social value of he consumpion of presen relaive o fuure generaions (Samuelson 1958). Each value of C /N can be decomposed as follows: C C s, C C s, N N s, N N s, (4B) where (C s, /N s, ) is a consan which equals he living sandard for he erminal sable age disribuion of he proecion series. Accordingly he value of he series is expressed as wo componens: a sable populaion componen and a ransiion pah componen. Subsiuing from (4B) ino (3B): V L N s, 0 1 N N s, C 0 1g Ls, 1g L L s, (5B) And he difference in social values of any wo demographic proecions, A and B, can be expressed as he sum of a difference in sable populaion componens plus a difference in he ransiion pah componens. V B V A 0 1 Ls,, A L C g s,, B 1g L, A L s,, A L, B L s,, B L0 0 1 N s,, A N s,, B 01 N, A N s,, A N, B N s,, B (6B)

22 Here A refers o he proecion in which moraliy changes according o he forecas and B he proecion wih moraliy consan a he iniial (2010) values. Thus V(A) V(B) evaluaes he very long run effec of he forecas moraliy change. ( ) ( ) is finie when in which case he value of he sable populaion componen in (6B) is simply calculaed as he sum of a geomeric series. The elemens of he ransiion pah componen can be calculaed for values of up o and including he las calculaed value T. The residual elemen of he ransiion pah componen of he value can be esimaed hrough model-based impuaion. Polynomials involving only negaive powers of were fied o he elemens of he valuaion beween 0 and T and he residual elemen esimaed from he inegral of his funcion beween T and. Soluion analyses calculaing changes o fuure labour force paricipaion (or migraion or feriliy or labour produciviy) which would produce changes in value equal o he esimaed effec of moraliy on social value may hen be performed.

23 References Ausralian Governmen (2002). Inergeneraional repor Canberra: Commonwealh of Ausralia. hp:// index.hml. Ausralian Governmen (2007). Inergeneraional repor Canberra: Commonwealh of Ausralia. hp:// Ausralian Governmen (2010). Ausralia o 2050: Fuure challenges. Canberra: Commonwealh of Ausralia. hp:// IGR_2010.pdf. Biak, J. (2013). Moraliy scenarios for 27 European Counries, , In M. Kupiszewski (Ed.), Inernaional Migraion and he Fuure of Populaions and Labour in Europe, Springer, Bloom, D., Canning, D. & Fink, G. (2010). Populaion Aging and Economic Growh, in M. Spence & D. Leipziger (Eds.), Globalizaion and Growh: Implicaions for a Pos-Crisis World, The World Bank: Washingon, D.C., Booh, H., Maindonald, J., & Smih, L. (2002). Applying Lee-Carer under condiions of variable moraliy decline. Populaion Sudies, 56, Booh, H. & Tickle, L. (2008). Moraliy modelling and forecasing: A review of mehods. Annals of Acuarial Science, 3 (I/II), Cerone, P. (1987). On sable populaion heory wih immigraion. Demography, 24 (3), Culer, D., Poerba, J., Sheiner, L., and Summers, L. (1990). An aging sociey: opporuniy or challenge? Brookings Papers on Economic Aciviy, Espenshade, T. J., Bouvier, L. F., and Arhur, W. B. (1982). Immigraion and he Sable Populaion Model. Demography 19 (1), Human Moraliy Daabase (HMD). (2013). Universiy of California, Berkeley (USA) and Max Planck Insiue for Demographic Research (Germany). Kupiszewski, Marek, Biak, Jakub, and Nowok, Beaa (2006) Impac of fuure demographic changes in Europe. Cenral European Forum for Migraion Research Working Paper 6/2016. Li, J. (2013). A Poisson common facor model for proecing moraliy and life expecancy oinly for females and males. Populaion Sudies 67(1): Naional Transfer Accouns proec (2014) (hp:// Gues, Ross and Parr, Nick (2014) A Mehod for Socially Evaluaing he Effecs of Long Run Demographic Pahs on Living Sandards. Paper o be presened o he European Associaion of Populaion Sudies Conference in Budapes, Hungary June 2014 Pollard, A.H., Farha Yusuf, and Pollard, G.N. (1990) Demographic Techniques. Pergamon: Oxford. Pollard, J. H. (1973). Mahemaical models for he growh of human populaions. London: Cambridge Universiy Press. Saczuk, K. (2013). Labour Force Paricipaion Scenarios for 27 European Counries , In M. Kupiszewski (Ed.), Inernaional Migraion and he Fuure of Populaions and Labour in Europe, Springer, Samuelson, P. A. (1958) An exac consumpion- loan model of ineres wih or wihou he social conrivance of money. Journal of Poliical Economy 66 (6), Unied Naions Populaion Division (UNPD) (2010). World Populaion Policies New York: Unied Naions.

24 Unied Saes Census Bureau (2013) Mehodology and Assumpions for he 2012 Naional Proecions. Washingon DC: US Census Bureau. Dae accessed 19 Sepember 2013 hp://

25 Tables and Figures Figure 1 Proeced Change in Male Life Expecancy a Birh Life Expecancy a Birh Year Ausralia Canada Czech Hungary Ialy Japan Neherlands Poland Porugal Spain Sweden Swizerland UK USA

26 Figure 2: Proeced Change in Female Life Expecancy a Birh Life Expecancy a Birh Year Ausralia Canada Czech Hungary Ialy Japan Neherlands Poland Porugal Spain Sweden Swizerland UK USA

27 Table 1: Summary Measures of Inpu Demographic Daa Toal Populaion in Millions (2010) Annual Ne Migraion in Thousands Ne Migraion Rae (per 1000) % Aged 0- % Aged Counry 19 (2010) 65+ (2010) TFR Ausralia Canada Czech Rep l Hungary Ialy Japan Neherlands Poland Porugal Spain Sweden Swizerland UK USA ,

28 Table 2: Age-Sandardised* Labour Force Paricipaion Raes, Employmen Raes, Hours Worked Per Employed Person and Hours Worked Per Person Aged 15+ Counry LFPR (%) Employmen Rae (%) Mean Hours Worked Per Employed Mean Hours Worked Per Person Aged 15+ Male Female Male Female Male Female Male Female Ausralia Canada Czech Rep Hungary Ialy Japan Neherlands Poland Porugal Spain Sweden Swizerland UK USA *: Direc sandardisaion using persons aged 15+ in Ausralia by age as he sandard.

29 Table 3: Proeced Life Expecancy a Birh for Males and Females for Seleced Counries 2010, 2030 and 2050 Life Expecancy a Birh Absolue Change Percenage Change Counry Males Ausralia Canada Czech Republic Hungary Ialy Japan Neherlands Poland Porugal Spain Sweden Swizerland UK USA Females Ausralia Canada Czech Republic Hungary Ialy Japan Neherlands Poland Porugal Spain Sweden Swizerland UK USA

30 Table 4: Proeced Life Expecancy a Age 65 for Males and Females for Seleced Counries 2010, 2030 and 2050 Life Expecancy a Age 65 Absolue Change Percenage Change Counry Males Ausralia Canada Czech Republic Hungary Ialy Japan Neherlands Poland Porugal Spain Sweden Swizerland UK USA Females Ausralia Canada Czech Republic Hungary Ialy Japan Neherlands Poland Porugal Spain Sweden Swizerland UK USA

31 Table 5: Proeced Values of Suppor Raios Incorporaing Forecas Changes in Moraliy Terminal Sable Populaion Counry L/N L/N* L/N L/N* L/N L/N* L/N L/N* Ausralia Canada Czech Rep Hungary Ialy Japan Neherlands Poland Porugal Spain Sweden Swizerland UK USA Noes: L/N is he raio of oal hours worked per week o oal populaion. L/N* is he raio of oal hours worked per week o consumpion needs-weighed populaion

32 Table 6: Proeced Changes in Suppor Raios for , , and Change Change Change Percenage Change Counry L/N L/N* L/N L/N* L/N L/N* L/N L/N* Ausralia Canada Czech Rep Hungary Ialy Japan Neherlands Poland Porugal Spain Sweden Swizerland UK USA