IIASA Population Projection Results

1996 Population Projections (Table)
Notes - Bibliography - Contents
Definition of 13 World Regions
Six Aggregated World Regions
Two Economic Regions

We reproduce here the complete set of results of the population projections realized at IIASA in 1996 and presented in The Future Population of the World: What Can We Assume Today? Revised and Updated Edition edited by Wolfgang Lutz and published by Earthscan.

Besides the results themselves, you will find below some excerpts from the book explaining the methodology used, the assumptions made and analyzing the results. Most of the text is directly extracted from the book itself.

Why another set of Global Population Projections?

The chapters of The Future Population of the World: What Can We Assume Today? Revised and Updated Edition (1996) show that the generation of new world population projections is useful provided that the approach followed (a) uses the most recent demographic data; (b) reflects the current scientific knowledge of the relevant disciplines; (c) considers several possible future trends in all three demographic components (fertility, mortality, and migration); (d) leaves room for substantive considerations of justifiable alternative future trends in the components; and (e) possibly improves the way projections deal with uncertainty or considers the interactions between the demographic variables and their socioeconomic and natural environments. The projections presented in this volume try to meet these criteria.

Users of population projections generally demand two types of information from population projections: (a) one most likely projection that can be used as a guideline, and (b) information about the range of uncertainty for sensitivity analyses and for testing the robustness of certain strategies (or policies). Several users -- especially environmental modelers -- demand projections until the end of the 21st century.

To meet these demands the alternative assumptions defined by the group of experts are extended to the end of the 21st century. They provide users with one most likely path, a broad range of alternative scenarios, and a fully probabilistic projection.

Methodology

Our projections are carried out using a multiregional cohort-component model, with five-year age groups. In contrast to global projections produced by the UN and the World Bank, which are based on individual country estimates, we have disaggregated the globe into 13 major world regions. By doing so, much of the world heterogeneity is taken into account, and we need not bother with national particularities, especially with respect to migration.

The crucial difference with respect to existing global population projections lies in the specification, justification, and combination of alternative scenario assumptions and in the definition of the first probabilistic world population projections.

The approach chosen involves a number of steps. First, selected experts with different backgrounds in the field of fertility, mortality, and migration analysis are asked to think about the future of the three demographic components. If necessary, different experts are invited to different regions. The experts are asked to suggest possible high and low assumptions for future fertility, mortality, and migration levels up to the year 2030 and discuss them in a group setting. The experts must argue their points substantively and justify their views. These elements are documented in chapters 3 to 14 of the book .

Next those alternative views expressed on future levels of fertility, mortality and migration which are sometimes only of a qualitative nature, are transformed into alternative quantitative assumptions on different future paths in each region considered. The high and low assumptions are assumed to correspond roughly to a 90 percent confidence interval of a subjective probability distribution. In other words, on each side only about 5 percent of all possible cases should lie outside the values considered. Assuming a normal distribution as representing the intuitive subjective probability distribution, the assumed central value is both the most likely case and the mean of the high and low values.

Synthesis of Expert Views on Future Fertility, Mortality, and Migration

The specific process by which these subjective probability distributions were defined and combined in practical terms is described in detail in Lutz et al. (1995) and summarized in Chapters 15 (reproduced partly below) and 16 of the book . A basic outline of the process is given in Chapter 2 .

Future fertility

Assumptions about the future course of fertility may be derived from several sources of information. First, past trends may be analyzed to assess the status of individual countries and regions in the process of demographic transition. Chapter 3 presents an assessment of this type on today's developing countries. This assessment was a major guideline in making specific assumptions for individual regions (see Table 1) under the basic premise that the fertility transition that has started almost everywhere will essentially continue. While low values assume a rapid pace in the continuation of the fertility transition, the high values assume a much slower pace, in some cases even a stagnation, of the declining trend. In parts of Asia this even implies increasing fertility.

For the near future the intentions expressed by women in representative surveys can be taken as an additional important piece of evidence. Chapter 4 analyzes this information for a number of countries. Sub-Saharan Africa has, by far, the highest fertility levels and the lowest levels of contraceptive prevalence, but in all countries in this region the desired fertility rates, which have been derived from surveys, are clearly lower than actual levels -- for example, in Kenya by almost two children. This fact, together with indications of increasing contraceptive prevalence, leads to the expectation of significant fertility declines in the next decade, even in sub-Saharan Africa which until recently has shown constant high fertility. In North Africa the fertility transition is already much further advanced, and more than one-third of the female population uses contraceptives. Desired fertility rates in this region indicate that further significant fertility declines are to be expected unless dramatic events, such as a rise in pronatalist fundamentalism, change the pattern.

The three Asian countries considered in Chapter 4 , Indonesia, Sri Lanka, and Thailand, already have total fertility levels below 3.0 with further declines expected. Latin America has an intermediate position comparable with that in North Africa. A noteworthy feature of this region is that over recent decades fertility was rather stagnant at this intermediate level, but the most recent evidence indicates further decline.

In the long run, without doubt, the general socioeconomic and cultural changes will determine fertility levels. But these trends are by no means easier to forecast than the path of fertility change itself. One more specific aspect in this context is the role of government policies and family-planning programs. As pointed out in Chapter 5 , experience with such programs has shown that, over several decades, they can have an important effect if they are well integrated into other government policies and particularly if the socioeconomic development of the population has reached a point at which limitation of family size is considered advantageous and a real option by sizable segments of the population. Since many countries have been adopting such programs, they may well accelerate the fertility decline.

Table 1. Alternative fertility assumptions used in projections.

In terms of specific numerical assumptions, the high value in 2030-2035 was assumed to be lower than the current fertility level for all developing regions except East and Central Asia. This is due to the overwhelming evidence of a continuation of the demographic transition process in parts of the world that still have high fertility. For this reason no constant fertility scenario was specified (as done by the UN, 1994). However, the high values do not assume a smooth fertility transition but rather an interruption or at least retardation of the process. This could be due to setbacks in socioeconomic development, cultural or religious movements such as a rise of fundamentalism, or strong heterogeneity with pre- and post-transitional populations living together in one region. The low values indicate a very rapid fertility decline that does not stop at 2.1 but continues to subreplacement fertility as it did in most industrialized countries. The resulting central values (the mean of the high and low values) lie somewhat above replacement level in most developing regions and significantly above in Africa, Central Asia, and the Middle East. The fertility assumptions for China are based on the discussion in Chapter 6 , and are set at a low value of 1.5 and a high value of 3.0 in 2030-2035.[1]

The process of making assumptions for today's industrialized countries is in one way more difficult and in another way easier than for today's high-fertility countries. It is more difficult because the direction of change is unclear: Will fertility increase or decrease? It is easier because the margin of uncertainty may be assumed to be smaller. For a modern urban society with increasing female economic activity, it is unlikely that fertility will be significantly above replacement level in the future. But it seems equally unlikely that fertility will remain permanently below a very low level such as about half of replacement because, at an individual level, the desire for children seems to be deeply rooted in women and men and, at a societal level, very rapid population aging results in serious problems (see deliberations in Chapter 11 ).

Until a few years ago, the United Nations and other institutions preparing population forecasts assumed that fertility would increase to replacement level and that subreplacement fertility was only a transitory phenomenon. This assumption is supported by the argument of homeostasis as discussed in Chapter 11 . In this view, fertility levels are not seen as the sum of individual behavior, but as one aspect of the evolution of a system in which individual behavior is a function of the status of the system (see Vishnevsky, 1991). Under such a systems approach the assumption of replacement fertility in the long run seems a defendable possibility. Therefore, we assumed a TFR between 2.1 and 2.3 in 2030-2035 as the high-fertility assumption in the five industrialized regions.

It is difficult, however, to find many researchers who support this view. Too much evidence points toward low fertility. The return to replacement fertility has been criticized as an assumed magnetic force without empirical support (Westoff, 1991). Many significant arguments support an assumption of further declining fertility levels. They range from the weakening of the family in terms of both declining marriage rates and high divorce rates, to the increasing independence and career orientation of women, and to a value change toward materialism and consumerism.

These factors, together with increasing demands and personal expectations for attention, time, and also money to be given to children, are likely to result in fewer couples having more than one or two children and an increasing number of childless women. Also, the proportion of unplanned pregnancies is still high, and future improvements in contraceptive methods are possible.

In conclusion one can say that the bulk of evidence suggests that fertility will remain low or further decline in today's industrialized societies. How far it will decline is not clear at this time. The expert group decided to settle for a TFR of 1.3 as the low value to be reached by 2030-2035. This value is below the low-fertility assumptions in some other projections, but it is far from being impossible. Actually several populations have already experienced such levels for extended periods. Taking into account the different ethnic composition in North America, the fertility range in this region was chosen to be slightly higher (1.4 to 2.3) than in Europe and Japan.

Future mortality

Mortality conditions at one point in time can be conveniently summarized by period life expectancy at birth, an indicator that results from a life table based on all age-specific mortality rates observed at that time. In this section, this indicator is used to define mortality assumptions (see Table 2).[2]

The uncertainties about future improvements or possible declines in life expectancy in today's high-mortality countries differ from those in low-mortality countries. With the exception of Eastern Europe, the latter countries have seen impressive increases in life expectancy and segments of their populations are approaching ages that were once considered a biological upper limit to the human life span. Hence assumptions about future improvements crucially depend on whether such a limit exists and will soon be reached (see Chapter 12 ; Manton, 1991). In regions that still have very low life expectancy, this question is irrelevant, and future mortality conditions will be determined by the efficiency of local health services, the spread of traditional (e.g., malaria) and new (AIDS) diseases, and the general level of subsistence.

Table 2. Alternative mortality assumptions to 2030-2035.

Currently, life expectancy at a national level is highest in Japan: 82.5 years for women and 76.3 years for men. Only 30 years ago in 1960-1964, female life expectancy was 71.7 and male life expectancy was 66.6, implying an average increase of more than 3.5 years per decade for women. Eastern Europe, on the other hand, had the same life expectancy as Japan during the early 1960s but has had almost no improvement since then, bringing it to only 67.3 for men and 75 for women at present. During the 1980s the Soviet Union even experienced a decline in male life expectancy, and current estimates for Russia are as low as 59 years. Western Europe and North America took an intermediate position; life expectancy increased steadily by about two to three years per decade. The recent trend clearly points toward further improvements, and the analysis of age- and cause-specific mortality trends does not give any indications of improvements leveling off (Valkonen, 1991). Also studies of occupational mortality differentials in the Nordic countries -- which are more homogeneous than countries in many other regions -- still show significant inequalities by social class (Andersen, 1991), which may be taken as an indication of the possibility of further improvements if the higher-mortality groups change their lifestyles. Vaupel and Lundström (in Chapter 12 ) show in the case of Sweden that at very high ages mortality also continues to decline.

These considerations in the low-mortality countries prompted us to assume future improvements that are more significant than those assumed by most other national and international population projections. As a high variant life expectancy is assumed to increase by three years per decade to 2030-2035. It is clearly not impossible since it is still below recently observed improvements in Japan and parts of Western Europe. The resulting central variant of a two-year improvement per decade also seems plausible, as supported by Mesle (1993). A broader margin was chosen for men in the European part of the former USSR because of greater uncertainties: assuming a continuation of bad conditions, the low value was set at zero years; the upper value was fixed at 4.0 because of the potential that life expectancy will greatly improve if the lifestyles in the region become similar to the Western patterns.

Health conditions in the developing regions have generally experienced very impressive improvements since World War II. Life expectancy in all developing countries has increased by more than 20 years since 1950-1955, when it was estimated to be about 40 years for both men and women. Some regions, such as East and West Asia, have experienced especially rapid increases. Latin America, which already had higher life expectancies during the 1950s, has achieved lower mortality levels than Africa and Asia. In the Caribbean, the mortality decline is particularly impressive.

These rapid improvements in mortality have led institutions that produce population projections to regularly change their assumptions. The UN raised the life expectancy limit for men from 72.6 years, assumed in 1973, in steps to 82.5 years, assumed in 1988. Similar adjustments were made for women (see Chapter 7 ).

Africa has the greatest mortality uncertainties of all regions. In Chapter 8 , Garenne concludes his survey of African mortality by pointing out that past trends in Africa have been induced by technology transfers from the West, which affected virtually all countries in a short period of time. Public health, nutrition, economic development, and modern education were the key determinants of mortality decline. There are reasons for assuming that this declining trend of the past 30 years will not continue in Africa, and that differences between countries will prevail or even increase: one reason is that infectious diseases, especially the HIV virus, may spread further; another is that problems of basic subsistence and food supply may continue (see Chapter 10 ). In Chapter 9 , Bongaarts presents a new calculation of the possible impact of AIDS on mortality. Translated into life expectancies and projected further into the future, his calculations result in a considerable range of uncertainties. For sub-Saharan Africa, the values for the scenario assumptions include uncertainties about AIDS as well as other uncertainties, such as possible wars and food shortages. The most optimistic case assumes further increases of four years per decade during a catching-up process with the rest of the world. The low scenario assumes declines in life expectancy of two years per decade. Consequently, the central value shows only very modest improvements of one year per decade.

In the other developing regions, a general increase in life expectancy of between one and three years per decade has been assumed. In South Asia and Pacific Asia, which are also seriously affected by AIDS, the range was chosen to be greater assuming no improvements in the low case and four years per decade in the high case. In South Asia and China, where the sex differential in mortality is still unusually low because of differential treatment of girls and boys, life expectancy is assumed to increase somewhat more rapidly for women over the coming decades. For the Islamic regions of North Africa and the Middle East, a minimum increase of 0.5 years and a maximum increase of 4 years was chosen reflecting the somewhat greater potential for improvements in these regions.[3]

Future interregional migration

Of the three components of population change, international migration is the most difficult to examine for two reasons: less reliable and representative statistical information is available for assessing past and present migration levels; and migration patterns tend to show much less continuity. Recent immigration trends in Western Europe clearly demonstrate the volatility of migration trends. During the early 1970s West Germany had an annual net migration gain of more than 300,000; five years later this number had declined to only 6,000 and 3,000 during the early 1980s. During 1985-1989, however, the annual net gain increased sharply to 378,000, 100 times that of the previous period. Few other countries have these extreme fluctuations, but the traditional immigration countries -- USA, Canada, and Australia -- show remarkable ups and downs. Annual net migration to Australia declined from 112,000 in the early 1970s to 54,000 in the late 1970s. During the 1980s it increased again to over 100,000. For the United States and Canada, only immigration figures are available, which for the USA shows a steady increase from around 280,000 per year during the early 1960s to around 800,000 during the early 1980s. In 1990-1991 the figures have declined to under 700,000.

Table 3. Matrix of assumed high values of annual net migration flows (in thousands).

Aside from the migration streams into Western Europe, North America, and Australia/New Zealand, labor migration within Asia has been remarkable over recent decades (see Chapter 13 by Zlotnik). For instance, during the late 1980s more than 430,000 workers left the Philippines annually, with some 280,000 of them going to the Middle East. South Korea lost an average of 170,000 workers per year during the early 1980s, India lost 240,000, and Pakistan lost 130,000 during the same period. The largest proportion of these workers went to the oil-rich countries in the Middle East. During the late 1980s these migratory streams within Asia have shown significant declines.

Because of the volatility of these trends and the great role that short-term political changes play in both the receiving and sending countries, it is more difficult to speculate about the future of migratory streams than about future fertility or mortality trends. Furthermore, net migration always results from the combination of two partly independent migration streams, namely, that entering and that leaving a certain region for whatever reasons. These reasons are sometimes grouped into political (asylum seekers), economic (expected differentials in standard of living), and recently environmental factors. This last category remains rather vague because worsening environmental conditions often result in worsening economic conditions. As pointed out in Chapter 14 , for all categories the potential for further increases in interregional migration is great because of better communications between the regions, cheap mass transport facilities, and the persistent gap between the North, which is not only richer but also rapidly aging and most likely in need of young labor, and the South, which has many young people with only limited opportunities in their home countries.

The actual extent of future South-North migration streams depends not only on the pull and push factors, but also on the immigration policies in the receiving countries. If the European Union, for instance, decides to enforce a policy of virtually closed borders to the outside because of popular demand within the EU, this may well result in a situation of almost no net migration. Hence in the projections, zero net migration was chosen for the low variant. This does not mean that borders are entirely closed to migrants; it only assumes that the number of in-migrants approximately equals the number of out-migrants.

For the high-migration scenarios, annual net migration gains of 2 million in North America, 1 million in Western Europe, 350,000 in the Pacific OECD region, and 50,000 in the Middle East have been assumed. The distribution of migrants from the assumed sending regions to the receiving regions and migration patterns among the less developed regions are based mostly on current migratory streams as described in Chapter 13 . Table 3 gives the migration matrix assumed in the high-migration case. In the low-migration case, all cells are zero; in the central scenario, values are half of those given in Table 3. The high scenario assumes an annual net migration loss of 375,000 in North Africa, 310,000 in sub-Saharan Africa, 370,000 in China and other Centrally Planned Asia, 560,000 in Pacific Asia, 495,000 in South Asia, 150,000 in Eastern Europe, and 225,000 in the European part of the former Soviet Union. Model migration schedules by Rogers and Castro (1981) were used to determine the age patterns of migrants.

It is almost certain that actual migration streams will not be constant over time. But assuming that the short-term fluctuations average around certain levels of constant annual net migration is the most practicable way to deal with this volatility.

Scenarios with Independent Fertility, Mortality, and Migration Trends

Specific model assumptions

In addition to the general assumptions described above, every projection must make several specific choices concerning the age patterns of fertility, mortality, and migration as well as the specific patterns of change over time and many other more technical details. The most important choices are summarized in this section.

The projections were generated using DIALOG, a software program developed by Sergei Scherbov; DIALOG is specifically designed for multistate population analysis and projection. The projections were done in five-year intervals for five-year age groups. These projections differ from most other projection models because we define separate age groups up to age 120; extending the age categories beyond the usual 80+ category helps to avoid serious biases that occur when significant proportions of the population in low-mortality regions come into the highest age group.

All countries were grouped into thirteen world regions that can be regrouped into six continents as well as into two groups of less developed and more developed regions. The population data for the starting year, 1995, were taken from the Population Reference Bureau (PRB) World Population Data Sheet 1995, which gives the most recent estimates for total population size as well as fertility and mortality levels in all countries. Since this source does not provide the full age distribution of the population, age distributions were taken from the United Nations 1994 assessment by proportionally adjusting the age groups to match the PRB total population figures in the case of discrepancies between the two sources. Age patterns of fertility and mortality in the starting year were derived using the same procedure.

As to the assumptions of fertility, mortality, and migration, the 1995 values and the alternative values chosen for 2030-2035 have been described above. A more detailed table giving the specific time path of the values and their extensions to 2100 is given in Tables 4 and 5. The values for 1995-2000 and 2000-2005 reflect the assumption of a quick opening up of the uncertainty range over the first five years followed by linear interpolation ("sausage model''). This particular decision was based on the observation that, under strictly linear interpolation, the range of uncertainty given for the first interval is far too small; therefore, between 1990 and 1995, many countries showed empirical trends that were outside the range assumed by the projections based on 1990 data.

Fertility assumptions for the second half of the 21st century are usually considered to be more art than science. Typically long-range projections assume some kind of a convergence toward the so-called replacement level. This is done for two reasons: first, it is convenient; second, no plausible alternatives seem to exist. But there is no good theory to explain how some magic hand should raise fertility to a level slightly above two children per woman after completion of the demographic transition. Only a vague and unproven homeostasis hypothesis would suggest such a pattern.

Table 4. Alternative Assumptions for Total Fertility Rates in 13 World Regions: 1995, 2000, 2030-2035, 2080-2085

In this analysis we suggest another rationale. Ultimate fertility is assumed to be a function of region-specific population densities within a predefined range of ultimate TFRs. After the completion of the demographic transition, low population densities are assumed to be associated with higher fertility, and vice versa.[4] Fertility is assumed to reach a level between 1.7 and 2.1 children per woman depending on density (for the central scenario) by the year 2080. The projected density for 2030 (central scenario) is taken as a criterion. The least densely populated region -- South America with 28 persons per km² projected for 2030 -- is given the highest TFR (2.1), while beyond 300 persons per km² (only South Asia, with a density of 478 persons per km² by 2030) a TFR of 1.7 is assumed. For densities between 28 and 300, intermediate fertility levels are obtained through linear interpolation. The low-fertility (and high-fertility) assumptions are set at 0.5 children below (and above) the value obtained for the central scenario. Table 4 and 5 give assumed fertility and mortality levels for the period 2080-2085 for each of the three scenarios. Beyond 2080-2085, levels are assumed to remain constant. Between the periods 2030-2035 and 2080-2085, levels are derived from linear interpolation.

Future age-specific fertility rates have been obtained by proportionally changing the age-specific rates of the starting year in accordance with the TFR level chosen. Extensive analysis of changing age profiles of fertility under different fertility levels has indicated that this is a reasonable approximation.

With respect to mortality, the United Nations (in the bi-annual assessments as well as in long-range projections) presents only one fixed path for each region without any uncertainty variants. Maximum life expectancy at birth is assumed to be 87.5 years for women and 82.5 years for men, and improvements in life expectancy end once the maximum figure is reached (between 2075 and 2150, depending on the region). For countries that already have very low mortality this implies very little prospects for further improvements in the future. In sharp contrast, our mortality assumptions span a very broad range in the long term, reflecting the great uncertainty about possible maximum mortality levels. In the low-mortality case, mortality rates continue to decline with a speed that corresponds to the past experience of the more successful countries. The high-mortality case even assumes some worsening in certain regions (see Table 5). Life-expectancy increases after 2030-2035 are generally smaller than before 2030, with constant mortality under the high-mortality assumption, a one-year increase per decade under the central-mortality assumption, and a two-year increase per decade under the low-mortality assumption. In regions with small mortality differentials by sex, future increase of the differentials has been assumed, as described in Table 5.

Table 5. Alternative Assumptions for Life Expectancy at Birth in 13 World Regions: 1995, 2000, 2030-2035, 2080-2085

A Brass-type relational logits model was used to produce a new set of age-specific mortality rates at a given level of mortality. For sub-Saharan Africa and South Asia special life tables have been applied; these tables include the specific age pattern of AIDS mortality which has been used by Bongaarts (Chapter 9 ). For age-specific migration, the model migration schedules by Rogers and Castro (1981) have been applied.

Most important results

Population size by region

World Population

Table 6 lists the total population sizes for the six continental divisions, the two economic divisions, and the world total in 1995 and population projections for 2020, 2050, and 2100. The table comprises nine cells that present the results for all possible combinations of assumptions of low, central, and high fertility and mortality under central-migration assumptions (see also Tables)

Table 6. Projections of total population (in millions) according to alternative fertility and mortality assumptions (and central-migration assumptions): 2020, 2050, and 2100.

Table 6 is different from the usual presentation of population projection results in two important ways: first, the user can freely choose any combination of the three fertility and mortality assumptions; second, the user may combine different scenarios in different continents. Hence, the user can derive the total world population resulting from high fertility and low mortality in Africa but low fertility and low mortality in Europe. Traditionally, uncertainty variants, such as the low variant in UN projections or the low-fertility/low-mortality scenario given in the 1994 edition of this book, present figures assuming lower than expected fertility in all parts of the world. This strong assumption of a perfect correlation between the regions results in the most extreme low figures for total world population. In the more likely case of no strict conformity of trends in different parts of the world, results tend be close to the central scenario because, for example, higher than expected fertility in Africa may be compensated by lower than expected fertility in East Asia. The figures given for the world in Table 6 thus refer only to the very unlikely case that all regions follow the same trends. The figures given for the two economic divisions, also assume that the same trends occur in each division.

Given the possible combinations of scenarios for different parts of the world, it is difficult to summarize the information contained in the table in a few sentences. If every region followed the central-fertility, central-mortality, and central-migration assumptions -- which are assumed to be the most likely ones as seen today -- world population would increase from 5.7 billion in 1995 to 7.9 billion in 2020, 9.9 billion in 2050, and 10.4 billion in 2100. This time path is plotted in Figure 1 which shows that this set of assumptions implies a clear leveling off of world population growth toward the end of the 21st century. Before 2050 world population still increases by another 4 billion people. It surpasses the 10 billion mark in 2050-2055, then reaches an all-time high of 10.6 billion around 2080, and starts to decline somewhat to 10.35 billion by the end of the 21st century. In other words, the world population under the central scenario falls short of doubling from its present size of 5.75 billion (end 1995).

Figure 1. Alternative projected paths of world population size (nine scenarios, central migration).

The most extreme case of high population growth results from the combination of high-fertility assumptions in each region with low-mortality assumptions in each region. In this highly unlikely case, the world population would double before the middle of the next century and reach an incredible 22.7 billion by the year 2100. High fertility combined with high mortality -- which is a more probable combination -- would only result in 15.1 billion people. Hence, given a high-fertility case, alternative mortality assumptions can make a difference in total population size that is larger than today's world population.

On the low-fertility side, the combination of low fertility with high mortality yields the lowest path of world population growth. If these rather unlikely assumptions were strictly applied to all world regions, world population would still increase by another 30 percent until 2030 and thereafter would enter a population decline that gains momentum during the second half of the century and might bring world population down to a low 3.9 billion by 2100. This would be significantly below today's total population size. It would equal the population in 1973-1974. Because of the significant population aging implied by this scenario, an inverse momentum of population growth (shrinking momentum) suggests further population decline during the 22nd century because smaller and smaller cohorts would enter the main reproductive ages. But aside from the incredible population aging implied by this, this scenario cannot be called desirable because it is associated with the highest mortality and thus a great deal of human suffering. In many parts of the world this scenario assumes very little or no future progress in mortality reduction; in sub-Saharan Africa it even assumes declining trends in life expectancy. The most desirable, under the criteria of a high quality of life and low population growth, is the low-fertility/low-mortality scenario, which results in a further increase of population size to 8.5 billion by 2050 followed by a more moderate decline to 6.5 billion by 2100.

Table 6, together with Figure 1, can be used to evaluate the relative impacts of the difference in the assumed mortality and fertility alternatives. The nine cells in the table and the nine lines in the figure show a clear ranking of the demographic variables, with fertility being the dominating variable but mortality still being highly significant. The impact of alternative mortality assumptions on total population size is more significant in the low-fertility case than in the high-fertility case. In the high-fertility case by 2100 the low-mortality assumptions yield a population size that is 50 percent higher than that of the high-mortality assumption; in the low-fertility case the differential is 65 percent. This differential is clearly big enough to justify the analysis of alternative mortality assumptions in addition to alternative fertility assumptions.

As stated above, Figure 1 only gives the scenarios in which fertility and mortality trends in all regions follow the same pattern. Combining different assumptions in different regions results in several additional lines that all lie between the extremes given. Most are close to the central scenario; the positions of the lines depend on the degree to which trends in one region compensate for those in another. A more systematic analysis of these aspects is provided in the context of the probabilistic population projections in Chapter 16 .

Regional Population Distribution

The global population growth described above results from a great diversity of different regional patterns. Under the assumptions of central fertility, mortality, and migration the population on the African continent grows the most rapidly. Its population almost triples in size over the next five decades from 720 million in 1995 to more than 2 billion in 2050. Accordingly, the share of the world population that lives in Africa increases from 13 percent to 21 percent (see Table 15.5). In the high-fertility/low-mortality case Africa's population more than quadruples by 2050, whereas under the low-fertility/high-mortality scenario, it does not even double. By 2100 the pessimistic mortality assumptions of this scenario lead to a population size that is smaller than today's.

Table 7. Distribution of world population by continent, according to three scenarios and under central migration: 1995 and 2050

The population of East Asia, which is currently home to one-third of the world's population, increases by 41 percent under the central assumptions until the year 2050. Because this growth lies well below the world average of 73 percent, the proportion of the world population living in this region declines by six percentage points to 28 percent. West Asia, on the other hand, gains five percentage points and holds 30 percent of the world population; by 2050 the population in this region more than doubles under the central assumptions. The reason for this difference lies predominantly in the fact that the fertility transition is already much further advanced in East Asia. Latin America currently has 8 percent of the world's population; it is likely to maintain this proportion because its growth is expected to be close to that of the world average.

Europe and North America both grow below the world average under all three scenarios. The fraction of the world population living in Europe falls from the current 14 percent to 7-8 percent by 2050. Looking at economic divisions, today's more developed regions decline from 22 percent to only 12-15 percent depending on the scenario, while the less developed regions grow from 78 percent to 85-88 percent (Table 7).

Figure 2. Alternative projected paths of population size in North America.

Alternative migration assumptions show sizable impacts on total population size only in the low-fertility industrialized regions of Western Europe and North America. Figure 2 illustrates the impacts of high- and low-migration scenarios in North America over the 21st century. As indicated above the high-immigration scenario for North America assumes average annual net migration gains of 2 million, whereas the low-migration assumption sets net migration at zero. The central scenario assumes a net migration of 1 million per year. Fertility assumptions in North America (for 2030-2035) vary from a TFR of 1.4 (low) to one of 2.3 (high). All the scenarios plotted in Figure 2 are based on central-mortality assumptions and alternative fertility and migration assumptions.

The figure shows that under all scenarios the total population size of North America increases from the present 300 million, at least for the next 30 years. After that time the combination of low fertility and low migration results in a significant decline of total population size, reaching about 160 million by the end of the 21st century. Combining the low-fertility assumption with high migration, however, results in a stable total population size of some 400 million. This difference of 240 million is due to alternative migration assumptions. Combined with high-fertility assumptions, the impact of high- and low-migration assumptions is greater in absolute terms (because the immigrants also have higher fertility) but smaller in relative terms. The combination of high fertility with low mortality is very close to the central scenario until the middle of the 21st century but then results in higher population growth later in the century. The unlikely combination of high fertility and high immigration results in significant exponential population growth in North America, coming close to 1 billion by the end of the 21st century. Government policies have more influence on immigration than on either fertility or mortality; therefore, immigration is clearly the key policy variable for influencing the extent of population growth in North America over the next century.

Population aging

Universal population aging (as measured by the mean age of the population or the proportion of the population above age 60) is the most consistent and dominant population trend of the next century. In all regions, the scenario projections show an aging population. Population aging has long been recognized as a cause of concern in the industrialized countries, but projections clearly show that over the coming decades it is likely to hit less developed countries even harder than it has industrialized countries because of more rapid fertility declines.

Table 8. Projections of percentages of the population above age 60 according to alternative fertility and mortality assumptions (and central-migration assumptions): 2020, 2050, and 2100.

Table 8 has the same structure as Table 6 but gives the proportions above age 60 instead of total population. Currently only 9.5 percent of the world population is above age 60. Since the early 1950s this percentage has slowly increased from 8.1 percent in 1950 to 8.4 percent in 1970 and 8.8 percent in 1985. Over the coming decades this percentage is projected to increase much more rapidly under all possible scenarios. Under the three central assumptions it increases to 13.2 percent in 2020, 19.6 percent in 2050, and 26.8 percent in 2100. These projected increases are much higher than any increases that have been experienced today in industrialized countries. Currently, Western Europe, the region with the oldest population, has 18.6 percent of the population above age 60.

Even under the extreme high-fertility and low-mortality scenario, which results in more than 22 billion people in 2100, the percentage of elderly increases to above 17 percent. Under the low-fertility/low-mortality assumptions the aging figures become truly daunting. By the end of the 21st century, this scenario shows that 42.5 percent of the population is above age 60. In industrialized countries half of the population is projected to be above age 60 under these assumptions, which are often called most desirable by ecologists and the media because they result in the lowest total population size. If mortality is high, fewer people survive to an old age, lowering the old-age dependency ratio. But even in this case of undesirable high mortality one-third of the population will be above age 60 by the end of the 21st century.

Given the dynamics of the population age structure one can safely say that in the foreseeable future the proportion of the population above age 60 will not return to today's low level. Only major disasters affecting the elderly could bring about such a situation, and then only for a limited period. Continued exponential growth at very high rates (which would be necessary to produce a younger age structure through fertility) can be ruled out for several obvious reasons.

Figure 3. World population (in millions) under high-fertility and high-mortality scenario in all regions: 1995 and 2050.

Figure 4. World population (in millions) under low-fertility and low-mortality scenario in all regions: 1995 and 2050.

Figures 3 and 4 depict age pyramids for 2050 of the two extreme scenarios in terms of the resulting age structure -- namely, that of high fertility combined with high mortality (Figure 3) and that of low fertility combined with low mortality in all regions (Figure 4). The shaded area gives the 1995 age pyramid for comparison. The high-fertility/low-mortality scenario (Figure 3) results in an age distribution that closely resembles the shape of the 1995 pyramid but at 2.3 times its population size (13.3 billion). Although it may not be directly visible from the graph, the population in the 2050 pyramid is clearly older than that in the 1995 pyramid both in terms of mean age and the proportion above age 60.

Under the low-fertility and low-mortality assumptions (Figure 4), the 2050 age distribution no longer has a pyramid shape but resembles that of a bottle; this shape is already visible today in the age pyramid of some very low-fertility countries. Figure 4 shows that, although the population in the 2050 age pyramid is 50 percent larger than the population in the 1995 pyramid, it is made up of fewer children. The proportion above age 60 consequently increases from 9.5 percent to 26.6 percent under this scenario. The pyramids graphically depict this dramatic aging within only 55 years.

Figure 5. Alternative projected trends in the proportion of population above age 60 in China and CPA.

Among the developing regions, China (together with Centrally Planned Asia) is certain to experience the most rapid aging because fertility has already declined dramatically. Figure 5 shows the increase in the proportion of the population above age 60 under five different scenarios for this region. Currently, only 9.2 percent of the population is above age 60. This percentage is much lower in China than in all industrialized regions. But a steep increase in this proportion is already preprogrammed in the age structure of China. Over the next 15 years a moderate increase in the age structure occurs followed by a steep increase; these increases show up under all alternative fertility and mortality assumptions. Only after 2035, when the large cohorts born before 1975 are over age 60, does this proportion stabilize under certain scenarios. Within less than four decades China's old-age dependency burden will be higher than North America's and about the same as that in Western Europe today. Western Europe's old-age dependency burden has built up over more than a century, so the region has been able to develop social-security schemes for the elderly. Nevertheless, Western Europe is confronted with serious problems in the pension system today and will have to deal with more problems in the future.

To avoid some of these problems China must take immediate and intensive actions. China's authorities are well aware of this issue and are considering not pushing programs to lower fertility further in order to avoid the large imbalances in the future age structure.

Population growth versus population aging

When comparing the information on projected aging to that on projected population growth, one very important feature becomes apparent. The scenarios that result in the highest population growth show only moderate population aging, while those showing a leveling off in population size imply very rapid population aging. Generally, very rapid population growth and very rapid aging are considered undesirable, but the scenarios described above clearly show that one cannot avoid both. An intermediate path, such as that given by the central scenario, avoids extremes but still combines substantial population growth with substantial aging at the global level. Since the problems caused by rapid growth and rapid aging are felt more at the regional level rather than at the global level, solutions to the problems must work at the former level. This issue must be discussed within a more regional context. The epilogue to this volume takes up some of the policy issues related to the aging versus population growth dilemma.

Comparison with Other World Population Projections

In Chapter 2 we argue that there is a need for additional world population projections that supplement those prepared by the United Nations, World Bank, and others. Because the differences in results often lie in the approach taken by the organization -- such as how to deal with uncertainty, whether to vary mortality assumptions in addition to fertility assumptions, whether to consider dependencies between the regions and between the components -- it is difficult to directly compare our results with those of the other agencies. But all sets of projections have one common feature: they include one projection that is presented as the most likely one, called the medium variant by the UN and the central scenario by us. Table 9 summarizes the most likely projections of the UN, the World Bank, the US Bureau of the Census, and IIASA.

However, even when a comparison is restricted to the most likely case, it is impeded by the fact that different projections have different time horizons and different forms of spatial aggregation. For this reason the information presented in Table 9 refers only to two points in time (2025 and 2100) and to the total world population. While the degree of increase between the two points in time gives a fair representation of the total path resulting from the assumptions made by each agency, the global aggregation clearly hides some important regional differences. For example, the UN 1994 medium variant and the IIASA 1996 central scenario yield almost exactly the same total world population size for 2025 but have different population figures for specific world regions and different proportions above age 60. One reason for this is that the UN assumes lower mortality improvements in industrialized countries and higher improvement in Africa than we do.

Table 9. Most likely projections produced by the United Nations, the US Bureau of the Census, the World Bank, and IIASA.

Naturally the differences between the projections are larger in 2100 than in 2025. The UN long-range projections produce the highest population size and the lowest proportion above age 60. This is mostly due to higher fertility assumptions and, to a certain extent, to lower life-expectancy assumptions. But these projections already date back to 1992 when most agencies were still assuming higher fertility. In fact, the recent projections of population size of all the agencies are lower than their previous forecasts. This reflects the fact that over the past five years fertility has declined in all world regions and AIDS and other infectious diseases have cast serious doubts on further mortality improvements in several developing regions.

As to the treatment of uncertainty, the World Bank 1994 published only one projection to the year 2150 without any uncertainty variants. The UN 1994 assessment published three variants (high, medium, and low) up to the year 2050, which are only distinguished by different fertility assumptions -- mortality and migration assumptions remain constant. The UN 1992 long-range projections provide five fertility variants up to the year 2150. UN 1994 and UN 1992 both present an additional "Constant-Fertility Scenario,'' which can only be intended for illustrative purposes because the resulting population size of more than 100 billion by the end of next century belongs in the category of "impossible scenarios.'' Aside from this outlier, the UN 1992 uncertainty variants for 2100 range between 6 and 19 billion. The IIASA 1996 scenarios presented here show population sizes ranging between 4 and 23 billion. This broader range is entirely due to the fact that our scenarios consider mortality uncertainty in addition to the fertility uncertainty. When limited to the scenarios assuming central mortality, the IIASA range is very close to the UN range.

Notes

[1] In all regions, 1990 age patterns of fertility were kept unchanged and scaled down (or up) according to the specification of the total fertility rate.

[2] UN model life tables (UN, 1982) corresponding to the respective region were used as input for 1995 to obtain age-specific mortality rates for today's developing regions. For the remaining regions, observed mortality rates were used. For regions with assumed high AIDS mortality special sets of age-specific mortality were used.

[3] Future age-specific mortality patterns were derived from model life tables (UN, 1982; Coale and Guo, 1989). Coale-Guo "west tables'' were extended up to life expectancies of 100 to provide mortality schedules for very high life expectancies.

[4] Some very preliminary cross-sectional analyses suggest that such a relationship may indeed exist among societies that have fully passed through the demographic transition. It would also correspond to an ecological approach which typically thinks in terms of "carrying capacity'' and "crowdedness.'' But clearly much more research is needed on this issue. Here the rationale is assumed simply to present an alternative to the ubiquitous replacement-level assumption which also lacks sound scientific foundations.

Bibliography

Andersen, O., 1991, Occupational impacts on mortality declines in the Nordic countries, in W. Lutz, ed., Future Demographic Trends in Europe and North America: What Can We Assume Today? Academic Press, London, UK.

Coale, A.J., and Guo, G., 1989, Revised regional model life tables at very low levels of mortality, Population Index 55(4):613-643.

Lutz, W., Sanderson, W., and Scherbov, S., 1995, Probabilistic Population Projections Based on Expert Opinion, WP-95-123, International Institute for Applied Systems Analysis, Laxenburg, Austria.

Manton, K.G., 1991, New biotechnologies and the limits to life expectancy, in W. Lutz, ed., Future Demographic Trends in Europe and North America: What Can We Assume Today? Academic Press, London, UK.

Mesle, F., 1993, The future of mortality, in R. Cliquet, ed., The Future of Europe's Population: A Scenario Approach, The Council of Europe, European Population Committee, Strasbourg, France.

Rogers, A., and Castro, L., 1981, Model Migration Schedules, RR-81-30, International Institute for Applied Systems Analysis, Laxenburg, Austria.

UN, 1982a, Model Life Tables for Developing Countries, Population Studies, No. 77, United Nations, New York, NY, USA.

UN, 1994, World Population Monitoring, 1993, ESA/P/WP.121, United Nations, New York, NY, USA

Valkonen, T., 1991, Assumptions about mortality trends in industrialized countries: A survey, in W. Lutz, ed., Future Demographic Trends in Europe and North America: What Can We Assume Today? Academic Press, London, UK.

Vishnevsky, A., 1991, Demographic revolution and the future of fertility: A systems approach, in W. Lutz, ed., Future Demographic Trends in Europe and North America: What Can We Assume Today? Academic Press, London, UK.

Westoff, Ch.F., 1991, The return to replacement fertility: A magnetic force? in W. Lutz, ed., Future Demographic Trends in Europe and North America: What Can We Assume Today? Academic Press, London, UK.

The Future Population of the World: What Can We Assume Today? Revised and Updated Edition

Contents

Definition of 13 World Regions

Six Aggregated World Regions

Two Economic Regions