Cette page existe uniquement en anglais
Home > Activities > Committees >Age Structure and Public Policy > Conference Report
Nihon university Population Research Institute (NuPRI), Tokyo, 19-21 March 2001
Organized by the IUSSP Committee on Population Age Structure and Public Policy and NuPRI
This was the second in a series of three meetings organized by the IUSSP Committee on Population Age Structure and Public Policy. The first meeting, which took place in Pukhet, Thailand, in November 2000, covered a wide range of issues related to age-structural transitions and focused on countries in a late transitional stage, with a certain emphasis on South-Eastern Asia. The third meeting, to take place in Taipei, Taiwan, in December 2001, aims at examining policy issues, namely the allocation of Public and private resources across generations.
The purpose of this second Conference, co-organized by the IUSSP Committee on Population Age Structure and Public Policy and the Nihon university Population Research Institute (NuPRI), was to examine the consequences of strong fertility declines and population ageing in post-transitional countries. The conference included seven sessions, that covered the following topics: age-structural transitions in industrialized countries (an overview of the process); the consequences of age-structural transitions on the family, the government and the market; the status of the elderly in industrialized countries; newly emerging research opportunities and policy implications of age-structural transitions. One or various papers were presented in each session, and a discussion followed. The key outcome of the conference will be a book based on a selection of the papers presented.
1. Age-structural Transitions: Overview>
The two theoretical papers presented in this session put the ageing process in a historical perspective. In the first paper, "Age-Structural Transitions in Industrialized Countries", Ian Pool referred to ageing as the end-phase of the process of age-structural transitions. As it follows a phase of population waves, population ageing is not a monotonic process; it is confounded by other processes of age structural mutation. In other words, the size of cohorts of ages under 60 can co-vary with structural ageing. Although the demographic community has placed less attention to changes in the sizes of cohorts than to the rates that determine these changes, the relative size of cohorts has important political implications. Disordered cohort flows can lead to intergenerational competition for resources and challenge the capacity to supply them. using modal age groups as the main indicator of the trajectories and speed of ageing, he observed that the early-ageing countries (Germany, Greece, Spain, for instance) will see the 60-74 age group become the mode soon, and the speed of transition from the mode at 0-14 to the mode at 60-74 will be fast. But in the late-ageing countries (Australia, New Zealand, the uS), 60-74 will not be the modal age group even by 2050. Countries in which the age transition is slower (France for instance) will have a more plateaued age distribution, while others (Italy) will move from one that is left-skewed to one that is heavily skewed to the right. Due to earlier fertility swings there can be reversals in the process, with countries going from higher to lower modal age groups. Finally, two or more waves can hit simultaneously, raising the mentioned possibility of inter-generational competition for Public and family resources. Successive waves and troughs require significant adaptation of policies and investment strategies. This is particularly relevant for current developing countries, many of which face more marked waves and troughs than developed countries.
In his paper "Age structural transitions in historical perspective: What can we learn from the past", Nico Keilman explained the assumptions that are made in various theories and methods to infer information from observed age structures. Stable population theory builds on strong assumptions, but has been used successfully for near-stable historical populations. The method is quite robust against changes in mortality. However, for industrialized countries at present, the stability assumption is unrealistic. The variable growth rate method rests on much weaker assumptions, but permits less powerful conclusions. The most detailed information on fertility and mortality may be derived from having series of births and deaths counts, in combination with one or more age structures. unlike these two methods, the inverse projection and back projection methods are not based on modeling assumptions; they go from counts to rates under assumptions on schedules only. Keilman observed that existing projections have tended to underestimate mortality improvements and therefore the number of elderly people. He suggested three ways of improving projections of the elderly in the future: 1. Including cause of death in mortality extrapolations. He questioned this option: including cause of death had not made a significant difference in existing projections; also, the independence assumption on which traditional models are based did not seem to be realistic. 2. Including risk factors, lifestyles, and health status as independent factors in mortality extrapolations. unfortunately, few examples show that this increases accuracy. 3. Quantifying uncertainty. Accepting that a certain level of error is to be expected. Stochastic forecast models are based on this principle, i.e. thinking of future population as a whole predictive distribution instead of a number.
At the end of the session, the discussant, Kevin Kinsella, welcomed Pools cohort approach but thought that modes had to be backed with figures showing the relative importance of a cohort over the others. Some participants argued that, in some cases, simultaneous waves do not lead to competition for resources. The discussant also commented on Keilmanīs final 3 points. He wondered whether there were ways to use causes of death without introducing the independence assumption. He mentioned the appeal of including risk factors to understand cohort differences and wondered whether cohort projections should be more widely used. Finally, he questioned the use of quantifying a wide range of uncertainty for policy purposes. This issue would appear in other papers and be further discussed during the meeting. Questions were also raised on whether declines in life expectancy at old ages (85, 90) were a rare phenomenon or a general trend. Some disagreed with the selectivity argument (less selection as more people reach old ages) and asked whether observed declines could be due to cohort factors.
2. Age-structural Transitions and the Family
The four papers presented in this session focused on two main topics: impacts of ageing on the family and challenges related to provision of care by the family. Regarding the first topic, in the paper "Ageing, Intergenerational Transfers and Saving in Japan", Andrew Mason, Naohiro Ogawa and Takehiro Fukui find, using household data, that the future of savings in Japan will depend greatly on the evolution of extended families and on the role of the family in old age support. Most studies on the subject are based on estimates of consumption and savings rates that do not take changes in family structure into consideration. However, evidence suggests that using current age-earning consumption profiles is misleading. First, consumption behaviour is influenced by inter-generational flows that are not captured in conventional analyses. For instance, in nuclear households, consumption increases if parents are richer, indicating that inter-generational flows may be in a downward direction, at least in this type of household. Second, the shift toward nuclear families and inter-generational independence appears to be strengthening the lifecycle motive for saving. If this is the case, and the trend towards more nuclear families continues, population ageing will have a greater effect on saving than would otherwise be the case. In the paper "Family and Kinship Networks in the Context of Ageing Societies", Michael Murphy wonders what changes have and will occur in family constellations up to 2050. He uses demographic microsimulation to elucidate historical, present and future kinship patterns. The results of this innovative analysis show that, up to a certain point, mortality declines tend to offset the effects of fertility declines and marital instability on family and kinship networks. By 2050, for instance, 1/3 of those aged 55 will not be in a partnership; however, at older ages, mortality improvements will offset lower lifetime marriage rates, so more aged people will be living with a partner than ever before. The average number of parents alive for those aged 60 in 2050 will be double the number today; in contrast, those born in 2000 will have only half as many living children at 60 as those born in 1950. More grandchildren will have grandparents and more of them, but the opposite does not hold: the average number and proportion of grandchildren will progressively decline. The number of ever-born sibs will drop sharply among children (but no so much for adults due to mortality improvements). The proportion of step-families will double from 2000 to 2050 and step-sibs will be a third of all sibs at the end of the period. These changes have important social implications. Grandparental rights over their grandchildren will need to be acknowledged, for instance. The possible consequences of ageing for inter-generational relationships or the increase in partial relationships involving step and supplanted parents provide material for further research.
In discussing the paper by Mason et al., Timothy Smeeding praised the micro approach and the methods used, but suggested that more be done with the data. He wondered about the future importance of savings and the non-demographic influences on savings (taxes and benefits, housing values, bequest motivation, among others). He recommended to focus on income since this, and not wealth, drives savings; however, given their implicit correlation, the effect of each is hard to disentangle. He also recommended that better measures of extended/nuclear households be used. Regarding Murphys paper, there was speculation about the possible implications of the forecasted changes for women, for care giving and for child investment, among others. Participants raised questions on whether the uK was a typical case and whether future cohorts may behave differently than present ones.
Regarding the second topic, logit regression models done by Naohiro Ogawa, Robert Retherford and Yasuhiko Saito in their paper "Care of the Elderly and Womenīs Labour Force Participation in Japan" show that severity of parents disability has no effect on whether a co-residing daughter or daugther-in-law is actively working. If type of work is broken down into three categories (full-time, part-time, traditional sector), having a parent that is unable to do at least one activity of daily living (ADL) actually increases the likelihood that the daughter works full-time and decreases the likelihood of her working part-time. The authors conjectured that institutionalization or long bouts of hospitalization frequently occur when a parent becomes partly or fully "unable", thereby creating the need for more resources in the household and freeing the daughter to work full-time. However, the positive effect of disabilitys severity on womens work is not as strong as the negative effect of parents age. The authors hypothesis is that this negative effect is due to a cumulation of illness episodes, not so much to severity of disability at a particular time. The authors indicated that the validity of these results may change with increases in female participation in the labour force, mandatory long-term insurance for the elderly and changes in values. In her presentation on Issues in the Provision of Care to Elderly People in Industrialized Countries, Emily Grundy cited emerging issues in elder care. Among them, the needs for long-term and acute care for an increasing number of oldest old, the quality and the access to care, as well as issues related to the provision of care (is it a private or a Public responsibility? What is the role of the family, the State and the market?). In Britain, the percentage of persons, particularly women, aged 80 and over that need daily help is increasing together with the number of elderly people. While the number of elderly people with 2 or more ADL limitations is projected to multiply by two by 2031, the amount of GDP spent in care for the old in various industrialized countries is still low. Family support is at present strong but the provision of intensive care and co-residence may decline with the increased prevalence of divorce and female labour force participation, for instance. Grundy wondered whether market and family support were interchangeable.
In the discussion, referring to the first paper, Jenny Gierveld wondered whether health status was a sufficient indicator of need for care to the elderly. She emphasized the need to convince policy makers that care is an issue for those 85 and over, a majority of which are women. Participants also pointed to the fact that the negative correlation between parentīs age and womenīs labour force participation does actually indicate that care for the elderly has an impact on womenīs work, regardless of the cross-section effect of disabilitys severity.
3. Age-structural Transitions and the Government
The future of Public health care costs and Public pensions were the subject of two very different papers presented in this session. Timothy Smeeding presented an empirical study co-authored by Deborah Freund on "The Future Costs of Health Care in Ageing Societies: Is the Glass Half Full or Half Empty?". According to Smeeding, concern about future health care costs for the elderly ignores the increased willingness and ability of the current and future elderly to pay, and the likely future benefits of new and better medical treatments. Most forecasts are based on two simple parameters, age-specific spending (empirical results showed that the elderly spend between 4.5 and 7.5 times more than the young) and age structure; assumptions are made about the growth of real health care spending, both Public and private. Since technological change is usually assumed constant, demographic change drives all health care costs. However it is mostly due to technological change that spending has increased and that, according to the authors, will continue to rise. But the benefits of these technological changes in terms of morbidity and mortality are frequently disregarded in Public cost-benefit analyses. On the other hand, time series analyses show that demand for health care is very income elastic and that income is increasing among the elderly. Out-of-pocket expenses are increasing parallel to income, but mostly among low income groups. If this is the case, the issue of cost distribution (Public/private mix) should be carefully considered: Should the "Public spending only" mold be broken? The authors suggested that future income security policy and health care policy be combined. Public funds should not be universal but targeted to the less well off.
Taking a much more theoretical approach, Shripad Tuljapurkarīs paper "Risk Metrics for the Management of Social Security Programs" argued for the use of probabilistic forecasts in population analysis and presented a model for the inclusion of uncertainty in the dynamics of pension systems. Tuljapurkar argued that, since trajectories wander (random walks are not straight lines) and variance cumulates, it should be accepted that forecasts are intrinsically uncertain. Decisions are gambles and risks must be estimated. Presenting ranges (i.e. accepting uncertainty) gives more options. Stochastic forecasts show that convergence towards the mean is not certain. Therefore, regarding pensions, policies trying to affect an indebted system, i.e. the average value of Public funds (C) so that it remains constant, may still lead to lower than average balances. To guard against this downside risk, the policy could rather be such that the average values did not remain constant, but increased more than the standard deviation. This would be challenging since the standard deviation scales up with time. Another option would be to find policies that reduce the variance. One possibility would be an adaptive benefit and tax scheme (increasing retirement age is an option, although its effect, according to Tuljapurkar, is modest). Being able to increase the rate of return would reduce both the average and the variance. But it would only make sense if a growth in surplus dominated any demographic shifts that could draw down the balance.
In the discussion of the paper by Smeeding and Freund, Andrew Mason noted that a better explanation of the link age-health-expenditure was required. Smeedings paper assumed that poor health started at age 65; an alternative assumption could be that poor health is limited to the 2 years prior to death. Alternative assumptions had very different implications for health expenditure. Participants noted also that the age pattern of use is very much supply-sided. Mason remarked that the uS spends more than other countries in health care and recommended cross-section analysis to evaluate the role of institutions in cost-benefit analyses. There was also discussion about the consequences of changing the Public/private mix and about the universal versus targeted approaches. Regarding Tuljapurkarīs presentation, Mason questioned the methods ease of use and thus its political "utility". He recommended that the method be made more accessible. Other participants asked whether the model was useful for economic forecasts, which involve a high degree of uncertainty.
4. Age-structural Transitions and the Market
The five papers presented in this session covered a variety of issues related to the economics of population ageing: the global economic impacts of ageing, selection effects in mortality and in annuity markets, the use of time and of human resources in ageing societies. In the first paper, "How important are the Global Economic Impacts of Ageing? Early Results from a Multiregional Overlapping Generations Model Designed to Address such Questions", Landis MacKellar and Tatiana Ermolieva reported on a simulation done with the most recent version of a linked overlapping generations model developed at IIASA. The emphasis was on establishing model properties rather than on substance. However, the main features of the preliminary baseline scenario were presented for all regions (uS, Japan, other industrialized countries, less developed countries). Model simulation properties were described in detail for Japan using a medium and alternatively a low fertility scenario. The underlying theme of the paper was the global impact of demographic changes, linked through capital flows. Some of the baseline global results were: Gradual reduction in rates of GDP per capita growth from 1995 to 2050 in all regions but the uS; Japan moves from surplus to deficit in the account balance, the uS moves from deficit to surplus, other industrialized countries remain in surplus and less developed countries remain in deficit. A comparison of these and the global results obtained using an alternative (low) fertility hypothesis showed that lower fertility translates in higher GDP per capita in all regions and particularly in the LDCs. In Japan, the uS and other industrialized countries, the smaller, older population gives rise to reduced net foreign assets and net factor payments from abroad (total GDP is lower and so are savings). The balancing entry in the global scenario were the LDCs, whose net foreign asset position moves in a positive direction.
In the discussion of the paper, Ian Pool praised the consideration of age-composition effects, often neglected. Since the burdens of ageing are not equally shared around the globe, he asked whether LDCs were paying for the ageing of industrialized countries. He however questioned the division of the world in four regions.
Lawrence Carterīs paper "Long-run Relationships in Differential uS Mortality Forecasts by Race and Sex: Test for Co-integration" examined differentials in observed and forecasted age-sex-specific mortality in the uS using the Lee-Carter method. Forecast trajectories do not present clear indications of either convergence or divergence of mortality rates by sex and, particularly, by race. The overall curvilinear patterns of life expectancy differentials are of very slow movements towards race divergence and either convergence or stability by sex; life-table entropy measures lead to similar results. In no instance is closure of differentials expected by 2065. Results suggest that degenerative diseases will continue to increase the disparity between white and non-white mortality. Carter noted that this type of extrapolative forecasts are meant to portray trajectories, not to offer an understanding of patterns. The results nevertheless suggested that there is probably no uniform protocol to eliminate mortality differentials by race and by sex. David McCarthy and Olivia Mitchellīs paper "Estimating International Adverse Selection in Annuities" offered new evidence of adverse selection in mortality in international annuities markets, for both compulsory and voluntary annuities. using various methods to compare mortality tables, international data show that adverse selection associated with the purchase of individual annuities reduces mortality by at least 25 per cent. This finding applies to all countries (except Chile and Germany), even to those where retirement benefits are subject to compulsory annuitization (for instance the uK). However, differences among countries in mortality patterns of both annuitants and total population are significant. This implies that the choice of mortality table to be used in countries with incomplete data (uS and uK tables are used in most developing countries) has an important effect in valuing annuities. In Japan, the system of mortality tables used to value annuities does not fit well into international norms: annuitant mortality is higher than predicted. This, according to the authors, required further investigation, especially since none of the tables used in Japan is derived from the actual experience of annuitants.
Regarding Carterīs paper, Shripad Tuljapurkar wondered whether it was realistic to assume linear mortality declines for the future; it is possible that remaining mortality gains are lesser than those observed during the twentieth century. He wondered whether the lack of co-integration was a reasonable finding, and, referring to both papers, whether researchers knew how to disaggregate mortality data (differentials always exist, regardless of whether disaggregating was conceptually appropriate). On McCarthy and Mitchells paper, if countries disaggregated to have mortality for annuitants and then a national-level annuity was created, how would countries aggregate? Other participants noted that it was important to know whether adverse selection existed because annuitants were a selected a sample of the population or because annuities created additional incentives to stay healthy.
The use of human resources in Japan was the theme of Atsushi Seikeīs paper "An Economic Analysis of Age Discrimination - The Impact of Mandatory Retirement and Age Limitations in Hiring on the utilization of Human Resources in an Ageing Society". Seike finds that mandatory retirement has a negative effect on the utilization of older employees. Furthermore, those employees with relatively higher human capital are those that completely retire from the labour market. He also finds that age limitations in hiring prevent middle-aged and older workers from finding jobs. To eliminate these barriers, the author supported the need to establish measures against employment discrimination based on age. However, there are reasons for employers to maintain these practices; namely, the current seniority-based wage and promotion systems justify mandatory retirement and age limitations in hiring. In Seikes view, substantial reform of seniority-based wage and promotion systems is necessary. Anne Gauthier and Timothy Smeedings paper "Historical Trends in the Patterns of Time use of Older Adults" examined time use of older adults from the 1960s to the 1990s in the united States, the united Kingdom and the Netherlands. While the increase in older peoples education, health and wealth would allow for an increase in time spent actively in old age, their results did not provide strong evidence of such trend. They suggest that time spent on paid work has slightly decreased for men aged 45 to 64, while slightly increasing for women 45 to 54. There has been a net increase in time devoted to leisure activities for people aged 45 to 64 in all three countries, but not for those aged 65 and over. The share of active leisure (e.g. social and cultural activities, sport, unpaid work) and passive leisure (e.g. watching television) has followed different trends in each country. Time spent in active leisure has remained stable in the Netherlands, increased in the uK and declined in the uS.
In the discussion, Robert Retherford noted that the empirical evidence supplied by these papers was important but that some analysis was required. This was challenging in Gauthier and Smeedings paper, because institutional issues need to be taken into account in cross-country analyses, and the survey they used includes little background information. On Seikes paper, he suggested to analyze the probabilities of working with and without mandatory retirement. He also asked what the role of unions was in maintaining seniority-based wages (Seike confirmed that the role of unions was significant). Participants also asked whether Seike was suggesting to abolish or to postpone mandatory retirement. He suggested abolishing it (i.e. making it voluntary).
5. Changing Status of the Elderly in Industrialized Countries
Jenny Giervelds paper "Gender and Well-Being; the Elderly in the Industrialized World" aimed at describing aspects of late life heterogeneity with emphasis on its gender aspects and to present country comparative data on the position of older women relative to mens in the industrialized world. Her main message was that data to study ageing by sex are scarce; she recommended that efforts be made to get sex-specific data for the population 60 and over. Based on the information available, her findings showed that ageing is at present a female experience. First, the proportion of non-married elderly is much higher among women in most industrialized countries, so more of them live alone. Second, Public or private pensions based on personal earnings in the labour market are still rare among women. In the future, men may experience similar situations if their life expectancy increases, particularly since instability in relationships is rising.
The discussant, Michael Murphy, raised a series of questions regarding social heterogeneity and inequalities: Are gender differences stronger among old people than among the young? Some participants noted that they are. Is this due to cohort or period factors? Is heterogeneity an indication of inequality? Should variety be accepted? What is the relationship between inter-generational and intra-generational inequalities? He also noted that societies which had not discriminated against women had not survived, and wondered whether residuals of that "advantageous" pattern remained. Some participants asked about the role of friendship networks at old age. Jenny Gierveld answered that although women have greater support networks, these are not substitutes for the family.
6. Newly Emerging Research Opportunities
In this session, Kevin Kinsella presented the first conclusions of the NAS Panel Report on Data Needs for an Ageing World, which was prepared by a group of 17 experts from 6 countries to: - Identify new scientific opportunities for research. - Review existing data and the quality of current international data collection efforts. -Outline data collection efforts. -Identify obstacles. The main challenge for the panel was to identify the audience: was the report aimed to researchers or to policy-makers? Their intention was to have an impact on policy-makers so more data are made available. Since ageing covers a wide range of topics, the panel recommended areas in which international comparison would be most useful: 1. Labour force participation and retirement. 2. Patterns of income and savings behaviour. 3. Family structures and inter-generational transfers. 4. Health and disability. 5. Subjective assessments of wellbeing (socio-behavioural approach, based on indicators other than health and wealth). They identified a series of crosscutting issues that required political attention: -International applicability. -The interrelated nature of the subject matter. -The value of data linkage: the expansion of available data, on one side, and the growing concern about individual protection and respondent confidentiality on the other. The challenge was to find ways to exploit data while protecting the individual. -The diversity among the older population. -The importance of socio-economic differences. -The need to deal with uncertainty: the ageing experience may change with time. The panel also recommended the development and use of multi-disciplinary, longitudinal (micro), cross-national research designs and remarked the need for dialogue between policy-makers and researchers. Some of the model examples to build upon would be: the Berlin Ageing Study; the uS Health and Retirement Study; the Taiwan Study on the Elderly; the German Socioeconomic Panel.
In the discussion, Nico Keilman stressed the need for more attitudinal data to do forecasts. He indicated that recommendations should include suggestions on how to make data available (emphasis on access). On the need for micro studies, he indicated that the challenge is to combine these with the "macro" political interests. He also wondered whether politicians were interested in having more data and understanding their meaning, or only in finding those data that support their policies.
7. Conclusions and Policy Implications of Age-structural Transitions
Gustavo de Santis presented the main conclusions of the meeting. He highlighted its main findings and mentioned the issues that had not been discussed, while also providing an overview of policy-relevant matters and discussing possible policy interventions. He noted that ageing was both a process (forecasted, studied in academia) and a result (which raised policy concerns due to its economic and social consequences). All methods used to forecast ageing had their pros and cons. Traditional (cohort-component) methods were easy to understand and to implement, but they were often used uncritically and gave an unjustified notion of certainty; parametric forecasts summarized the main information in a few parameters and could be used to incorporate model-dependent uncertainty, but they were scarcely known and did not work well for all demographic variables; stochastic projections included uncertainty, so they gave a range of options for the future, but confidence intervals were model-dependent and they were difficult to apply and to understand. Micro-simulations, which had also been shown in the conference to project family networks, furnished less mechanic, more dynamic forecasts, but they were difficult to control, did not provide confidence intervals, and its results were not necessarily reliable (if peoples behaviours were context-dependent, for instance).
Regarding the consequences of ageing, the seminar had touched upon economic, health-related, social and gender issues. If anything, ageing appeared to have a modest impact on economic growth; often, this effect was brought about by institutions (e.g. early retirement) more than by ageing per se. Regarding health, despite improved health conditions at all ages, inequalities did not seem to be diminishing. Parallel to ageing (but not necessarily due to it), health expenses were on the rise; the affordability of necessary expansions of basic services in the future was at stake. Another issue of concern was the increasing loneliness among the elderly. De Santis wondered whether interventions such as giving incentives to cohabitation and to de-institutionalization could solve this problem. Then there was also the issue of time use: the elderly had more free time and also more potential than in the past. This opened significant possibilities for mutual help (i.e. care of the elderly would not only be confined to the family network); he noted that this issue had not been discussed in the Conference. Finally, regarding gender issues, de Santis said that womens disadvantage at old age was the result of better survival chances coupled with poorer health and of free conjugal choices. He questioned whether pension systems should be revised to further reward women, and noted that changes were required in other areas (labour market, Public assistance to manageing a household), before women reached old age. He wondered to what extent the effects of ageing on "private" economics should become a policy concern. In his view, there was a trade-off between the direct participation of the state and the state control on private activities: the smaller the Public coverage, the more "market failures" became a policy concern.
Some topics were not discussed in the Conference. Among them: thresholds for old age (constant or evolving?); optimum equilibrium between Public and private forces in the provision of pensions and health assistance (does Public pay-as-you-go, as opposed to private funding, discourage savings and economic growth? Is Public health assistance less expensive and more effective?); policies to attenuate ageing (should governments encourage fertility? Migration?).
The conference provided an excellent opportunity to present state-of-the-art studies on age-structural transitions and to discuss challenges ahead, both at the private and Public levels. In general, though, the conference placed a stronger emphasis on challenges than on responses to ageing. The forthcoming meeting of the Committee, to take place in Taipei in December 2001, will hopefully review policy issues more in depth.