According to the 1996 Revision of United Nations World Population Prospects, the total fertility rate (TFR) is estimated to be, in 1990-1995, at or below the level of 2.1 children in 51 countries or areas of the world (Table 1), whose combined population (2.6 billion in 1997) amounts to 44 per cent of the global population. According to the medium variant of the 1996 Revision, there are 9 additional countries in which fertility is expected to decrease to or below replacement level between 1990-1995 and 1995-2000: Armenia, Guadeloupe, Netherlands Antilles, New Zealand (where TFR already has been below 2.1 in 1980-1990), Puerto Rico, Republic of Moldova, Reunion, Sri Lanka and Trinidad and Tobago.
In practically all countries of the more developed regions fertility is currently significantly below the level necessary for the replacement of generations; in many countries in these regions the period TFR has stayed at below-replacement level for several decades (Figure 1). In the last decade fertility has decreased to levels below replacement in several countries from the less developed regions, including all countries in the populous region of Eastern Asia (except Mongolia) and a rapid fertility transition in South-Eastern Asia and Latin America brings an increasing number of developing countries close to that threshold. As a result, the methodological issues of projecting fertility at levels below replacement is acquiring particular importance.
The three fertility variants of the United Nations projections are referred to as high, medium and low, depending upon the assumed levels of future fertility. The high-, medium- and low-fertility variants for each country are all thought to provide reasonable and plausible future trends in fertility. The low and high-fertility variants are usually thought to bracket the probable range of future population change for each country; nevertheless, the fertility change for a given country could progress, and occasionally has, at a pace outside the high-low bracket. The constant-fertility scenario is an illustrative projection prepared for analytical purposes. This scenario is based on the assumption that future fertility for each country will remain unchanged at the level calculated for l990-1995.
Comparison of observed world population change with that projected in the whole series of Revisions shows that for large aggregations of countries (such as development regions or major areas or the world) the medium-variant projection has well described future world population growth, at least for year horizons of 20-25 years. In that sense, the medium variant projection for such large aggregations has been described as "most likely".
The fertility assumptions for each country in the United Nations projections are operationalized, first, by assuming future changes in the total fertility rate; and, secondly, assuming or calculating age-specific fertility patterns that are jointly consistent with the assumed TFR and the current national age patterns of fertility.
For high- or medium level fertility countries (where fertility rates are above replacement level), the pace of the assumed fertility decrease (operationalized in the medium variant as the 'target period' in which TFR will reach and pause at replacement level) is determined by taking into account a range of socio-economic factors, such as population policies and programmes, adult literacy, school enrolment levels, economic conditions, infant mortality and nuptiality, as well as historical, cultural and political factors.
The analysis of past fertility trends in countries that have already completed the demographic transition does not suggest any obvious future trend. For instance, although the downward trends typically continue after having crossed the replacement level, no cross-sectional relationship emerges between how long a country's TFR has been below replacement and how low the TFR is (Figure 1). There exists no compelling and quantifiable theory of reproductive behaviour in low-fertility societies. Therefore, the methods of formulating fertility assumptions for countries whose fertility level is currently below replacement represent an evolving attempt at better approximating future reality.
||Bosnia and Herzegovina
||Republic of Korea
||Dem. Peoples's Rep. of Korea
||10,350 ||United Kingdom
||United States of America
||China, Hong Kong SAR
In the 1996 and in the previous Revisions the pace (and direction) of future fertility change for each low fertility country depends upon the recent fertility experience of the country -- whether fertility has recently been declining, rising or remaining relatively constant. For all countries, however, it is assumed that fertility in the medium variant will ultimately rise toward replacement level, either by continuing to rise to that level if fertility is already rising or by beginning to rise sometime in the future (usually after 2000 or 2005) if fertility has been constant or is declining.
In particular, if recorded fertility was declining in 1990-1995, TFR was assumed to decline at its past pace through 1995-2000 and then remain at its 1995-2000 level through 2000-2005. Subsequently, TFR was assumed to rise towards the replacement level at a pace of 0.07 child per 5-year period. If recent recorded fertility was relatively constant, TFR was assumed to remain constant through 1995-2005 and then to increase towards the replacement level at a pace of 0.07 child per projection period. If fertility was rising recently, TFR was assumed to continue rising to 2.1 at a pace of 0.07 child per projection period.
Similar approaches are taken for the high- and low-fertility variants. For the low variant, if recorded fertility was declining in 1990-1995, it was assumed that TFR will decline further at a pace of 0.1 child per 5 years until it reaches 1.5, or will remain constant at its lowest level if the actual TFR was below 1.5. If fertility was relatively constant or rising, TFR was expected to start falling at a pace of 0.1 child per projection period until it reaches 1.5, or remain constant at its lowest recorded below-1.5 level.
For the high variant, if fertility was actually declining in 1990-1995, it was assumed to remain constant in 1995-2000, and then begin to rise to 2.5 at a pace of 0.1 child per projection period. If fertility was constant or rising, the TFR was assumed to begin (continue) rising immediately at a pace of 0.1 child per projection period.
These assumptions are translated into varying lengths of periods when fertility is expected to remain below replacement level in particular countries or areas. Table 2 shows the projection periods when the currently below-replacement countries are expected to return to replacement in the medium and high variants. (Fertility will always stay below replacement in the low variant). According to the medium variant, this will happen before 2010 in only 3 countries; fertility is expected to reach replacement from 2010 to 2030 in 16 countries and from 2030 to 2050 in 15 countries; in 17 countries replacement level is not reached until after the projection horizon (2050).
Corresponding TFR trajectories are shown in Figure 2 for the seven regions where fertility was below replacement in 1990-1995. While these trajectories follow similar patterns, they differ in several respects. For instance, in the medium variant fertility is projected to return at replacement level in 2010-2015 in Northern America, in the following quinquennium in Australia/New Zealand and in 2020-2025 in Eastern Asia. In the regions of Europe the restoration of replacement fertility is projected to occur much later - and beyond the projection's horizon in all regions except Northern Europe.
Therefore, the central assumption of this projection approach is that fertility will remain below replacement level for many decades to come although it will eventually move toward and pause at replacement level many years in the future. Intrinsically, the 'magnetic force' toward replacement (Westoff, 1991) is based on the homeostatic argument of the demographic transition theory: an initial equilibrium between high birth rates and high death rates is disturbed by declining mortality which in turn triggers a fertility decline that brings birth and death rates back to an equilibrium at low levels. Most explicitly developed by Vishnevsky (1976,1991), the homeostasis argument suggests that fertility levels ultimately result from the development of the 'demographic system' that aims at its own inherent goals of self-maintenance and survival, rather than are the sum of individual behaviour. Fertility levels that dropped substantially below replacement levels or stayed below replacement for relatively long time are construed as aberrations or overshootings that will be inevitably reversed in the future. Although that hypothesis is not specific enough to be empirically tested (Lutz, 1994), it remains very influential (see, e.g. Day, 1995), partly because reproductive intentions stated in opinion surveys consistently reveal a strong trend toward the two-child family as a normative goal.
|Bosnia and Herzegovina||after 2050||2025-2030||Netherlands||2040-2045||2025-2030|
|China, Hong Kong SAR||after 2050||2030-2035||Republic of Korea||2035-2040||2015-2020|
|Cuba||2040-2045||2015-2020||Russian Federation||after 2050||2025-2030|
|Czech Republic||after 2050||2025-2030||Singapore||2025-2030||2010-2015|
|Dem. People's Rep. of Korea||1995-2000||1995-2000||Slovakia||2045-2050||2020-2025|
|Estonia||after 2050||2030-2035||Spain||after 2050||2040-2045|
|Greece||after 2050||2030-2035||Ukraine||after 2050||2030-2035|
|Hungary||after 2050||2025-2030||United Kingdom||2025-2030||2015-2020|
|Ireland||2025-2030||2005-2010||United States of America||2010-2015||2000-2005|
For countries which are in the midst of the transition from high to low fertility, age-specific fertility patterns matching those projected total fertility rates are calculated by interpolating (with respect to TFR) between the most recent available national age-specific fertility pattern and model replacement-level fertility patterns. These models were constructed considering the age-specific fertility patterns of countries that have reached replacement level. Three types of the age pattern of fertility at replacement level have been identified by looking at existing age patterns in countries where fertility has already reached replacement level (see Table 3). These models are:
The currently used assumption for low-fertility countries is that country-specific age patterns of fertility will remain stable during the entire projection period. In fact, continuing shifts in child-bearing patterns of low-fertility populations are the norm rather than an exception. However, those shifts do not necessary occur in the same direction. As a result, fertility schedules often tend to further diverge rather than converge as fertility decreases below replacement. For instance, around 1970 Japan and the Russian Federation had similar levels of fertility and similar fertility pattern. However, the Japanese fertility pattern became older as TFR decreased from 2.1 in 1970 to 1.5 in 1990-1995, with the mean age at child-bearing increasing from 27.4 years in 1975 to 29.3 years in 1994, while the Russian pattern became younger while experiencing a similar drop in TFR from 2.1 in 1965 to 1.4 in 1990-1995, the mean age at child-bearing decreasing from 26.4 years in 1975 to 24.6 years in 1994 (Figure 3). In the space of 20 years a relatively small difference of 1 year in the mean age of child-bearing had widened to 5 years.
Differences in the age pattern of fertility have consequences on the size and the age-structure of the population. The impact of the assumed age pattern of fertility on the projection output is illustrated by the application of the Japanese schedule of 1990-1995 to Russian data. Assuming all other projection inputs (including TFR trajectory for the projection period), the Japanese age pattern of fertility yields a 20 per cent decrease in population size from 1995 to 2050 compared with a 23 per cent decrease when the Russian 1990-1995 schedule is applied. This is a three percent difference in the size of the 2050 population. The differences in the 2050 age structure are however more striking: for instance, under the Japanese fertility schedule the age groups 0-4 and 5-9 become 15 per cent larger, while the group 45-49 is 5 per cent smaller.
It is thus clear that the age-pattern of fertility is a significant factor in projecting populations, however the current lack of an operational theory of reproductive behaviour in low-fertility societies precludes the design of an appropriate methodology for assuming future changes in fertility pattern, and in practice the age pattern of fertility is usually kept constant in the future.
In all that precedes, the expression "replacement level " has been used interchangeably with a TFR level of 2.1 when discussing low fertility, and the value 2.1 has in effect been used as the replacement level. This is correct in first approximation for all the countries where fertility has already reached replacement level, because these countries also have low mortality. In practice, of course, "replacement level" means fertility level such that the net reproduction rate is equal exactly to 1, and depends on the sex ratio at birth and on the proportion of women surviving to the mean age of child-bearing. In real life the sex ratio at birth for most countries varies in a narrow range, from 1.05 to 1.07, but there are exceptions, notably the Republic of Korea, where it is 1.15 and China (1.10). Survivorship to the mean age of child-bearing, in the countries with low fertility, varies from .99 in Japan to .97 in the Russian Federation. As a result, the replacement-level TFR, strictly speaking, is 2.08 in Japan, 2.12 in the Russian Federation, 2.17 in China and 2.22 in the Republic of Korea.
We will continue to use the TFR value of 2.1 as a proxy for "replacement level" with the understanding that it is only a loose definition.
The medium variant of the 1996 Revision yields a stable population size for Northern Europe during the first half of the 21st century; in Western Europe, during the first quarter of the century there will be zero population growth, followed by a 10 per cent population decrease in the second quarter. Populations of Eastern and Southern Europe will be decreasing during the entire projection period (Table 4) and the pace of negative population growth will accelerate; by the mid-century they are expected to lose 18 and 16 per cent of their 1995 size, respectively. On the contrary, the populations of Northern America, Eastern Asia and Australia/New Zealand are projected to keep growing; their sizes in 2050 will be from 21 to 43 per cent higher than in 1995.
|Low variant||Medium variant||High variant|
When reviewing table 4, it must be noted that these 7 regions include 5 countries whose total fertility rates are above replacement level in 1990-1995 (Republic of Moldova, Iceland, Albania, New Zealand and Mongolia). In addition, it must be noted that large immigration flows in many below-replacement countries that are projected to last well into the 21st century counteract the fading or negative natural increase of their populations.
Under the fertility assumptions of the high variant, population growth in Eastern Asia, Northern America and Australia/New Zealand is further amplified; the virtual stabilization in Northern Europe becomes a slow positive growth, while the population of Eastern, Southern and Western Europe remains nearly constant. Conversely, under the low variant all currently low-fertility regions are projected to experience in the second quarter of the 21st century high rates of negative population growth (Figure 4).
Naturally, fertility assumptions have the heaviest impact on projected annual numbers of births and, therefore, on the size of youngest population groups. Figure 5 shows, for all developed regions combined, that the medium variant yields relatively minor changes in the size of the age group 0-4 between 1995 and 2050, while the high variant implies a 36 per cent growth and the low variant projects a 53 per cent decrease; as a result in 2050 the size of the age-group 0-4 in the medium variant will be twice the size in the low variant and in the high variant it will be three times the size in the low variant.
Current fertility levels and fertility assumptions for the future also have important implications for the projected pace and extent of population ageing (Figure 6). Ageing will proceed in all variants: even the fertility increase incorporated into the high variant will not prevent the proportion of persons aged 65 or over from increasing considerably between 1995 and 2050, with that increase ranging from 6 percentage points in Northern America to 13 percentage points in Southern Europe. While in 1995 the highest proportion of persons in the age group 65 or over was recorded in Northern Europe (15.4 per cent), Southern Europe in all variants is expected to become the leading region as early as the year 2000 and will retain that position until the end of the projection period when the proportion of the age group 65 or over is expected to be from 27.4 per cent in the high fertility variant to 36.0 per cent in the low fertility variant. The high proportion of the age group 65 or over in Western Europe (15.0 per cent in 1995) will further increase by 2050 when it is expected to range from 24.3 per cent (high variant) to 32.3 per cent (low variant). Northern Europe will see a slower pace of ageing and the region's age structure will converge with that of Eastern Europe and will be significantly younger than those in Southern and Western Europe. In Eastern Asia the relatively recent fertility transition will cause especially rapid ageing, increasing the proportion of the population aged 65 and over from 6.8 per cent in 1995 to between 17.3 and 25.2 per cent in 2050, according to the variant.
The differences in the proportion of the population group aged 65 or over among projection variants will be increasing with time at accelerating pace and are expected to reach, in all regions, around 8 percentage points between the extreme variants (Figure 6).
The historical experience up to date does not suggest the stabilization of fertility at replacement level. Instead, total fertility rates often cross the 2.1-line and go further down. Their further trajectories are varied. Very often fertility decreases further substantially and stays for long periods below replacement level (Figure 1); that fall may be either straight and fast or fluctuating and gradual. What is particularly important is that new empirical evidence tends to continually revise downward the level of TFR. For instance, a decade or two ago low fertility variants of national demographic projections in Europe assumed TFR's higher than 1.5 (TFR level of 1.5 is consistent with two children per mother on average and the childlessness level of 25 per cent). Now, a TFR in the range of 1.2 to 1.5 is assumed as the low variant in demographic projections of most low-fertility countries and the medium variant is almost universally set below replacement (Eurostat, 1991; Keilman and Cruijsen, 1992).
Moreover, these shifts in assumptions happen against the background of even lower fertility levels achieved in large parts of several European countries (e.g. in Eastern Germany, Northern Italy, and certain most urbanized regions of the Russian Federation the TFR was at or below 1.0 in the 1990s). Longer full-time education of women, more cohabitation or living alone and increasing female labor-force participation are often identified as relevant social trends leading to later motherhood, more children born out of wedlock, a rising voluntary childlessness and further declining family sizes.
On the other hand, in the Nordic countries, in particular in Sweden and Norway, fertility substantially increased in the late 1980s and approached or even surpassed the replacement level. This reversal of fertility decline may have been associated with large-scale social policies aimed at creating conditions (through significantly extending child care facilities, and increasing family allowances) which allow women to combine professional careers with motherhood. (However, since 1990 Sweden's fertility has been declining fast again).
As a result, the 'post-transition' fertility levels became less homogenous in the first half of the 1990s than they were a decade ago, which highlights the methodological challenges of projecting their future trends.
In an attempt to better reflect these features of below-replacement fertility, the United Nations is suggesting a new approach to formulating fertility assumptions in the 1998 Revision. First, two types of target TFR levels are introduced in the medium variant instead of the previously used replacement level (i.e., TFR at 2.1). For countries with current (i.e. most recent available) TFR below 1.5, the medium-variant target level is 1.7; for countries with current TFR between 2.1 and 1.5, the target level is 1.9. Second, completed cohort fertility of the youngest generation for which it can be safely estimated (i.e. cohorts born in 1962) is used as a yardstick for long-term trends: the new medium-variant target fertility is computed as the average of the estimated completed fertility of the 1962 cohort and either 1.7 or 1.9 as defined above. In the absence of recent information on cohort fertility, the target becomes simply either 1.7 or 1.9.
In making assumption about the future path of fertility, the level of the TFR in the year 2000 is determined by continuing the recent trends, upward, downward or about constant; the TFR is assumed to pause at the level reached in 2000 until 2005, before starting a linear move toward the selected target level, increasing by 0.07 children during each 5-year period of the projection.
This new procedure takes into account the recorded experience of long periods below replacement level and represents a move towards the cohort method of fertility projection which is currently implemented in most countries of the European Economic Area (Cruijsen, 1994).
The major rationale behind the cohort method is that it explicitly takes into account past fertility history of the individual country and therefore better captures the lifetime number of births. While the application of the period method under conditions of profound changes in timing of fertility (currently, towards postponement of births in most low-fertility populations) significantly overstates the degree of fertility decline, the trends in cohort total fertility are smoother and of smaller magnitude. If there is no more change in the future cohort fertility, the period TFR in the long range will become close to the current cohort TFR. An important additional consideration is that an increasing number of countries produce necessary data on cohort fertility.
The target medium-variant TFR levels of the 1998 Revision are the function of the most recently recorded period and cohort fertility of the individual country, and are not tied anymore uniformly to replacement fertility. They yield much more diverse national fertility trajectories than was the case in the previous Revisions.
The target TFR levels for the high variant are set at 0.4 child above the target of the medium variant; the target for the low variant is set at 0.4 child below the target of the medium variant or is assumed to remain constant at its latest recorded level, whichever is lower. In each case the level in the year 2000 is determined in relation to the medium variant level, and after the year 2000 the TFR moves toward its target level at a pace of 0.10 child per 5-year period of the projection. Therefore, the high and low variants also become country-specific and diversified.
The new, lower, fertility assumptions would yield slower positive population growth (or earlier and faster negative population growth) and accelerated ageing than in the previous Revisions. A simulation with the Austrian data illustrates these differences. (Austria was chosen because it has published estimated 1962 cohort fertility and is among the countries with the lowest cohort fertility). This compares the projection output of the medium variant of the 1996 Revision with the projection output generated on the basis of new fertility assumptions, while the baseline population, mortality and migration assumptions are kept the same. In the 1996 Revision fertility was assumed to move towards the replacement level, which, though, was expected to be reached beyond the projection horizon. In the new simulation scenario the TFR target is assumed to be 1.68, which is the average of the estimated fertility of the 1962 cohort (1.65) and the general target of 1.70 for countries with current period fertility below 1.50. The difference in the target level is not the only cause of the difference in the projected populations. While the pace of the transition toward the target level is the same in both cases, with the new approach fertility would go to a lower level in 2000-2005 than in the 1996 Revision (1.30 instead of 1.42), before starting its climb back up. This results from new data showing an acceleration of the downward trend in 1994 and 1995. Figure 7 shows the assumed future trajectories of fertility in the two scenarios .
The difference in the total population size in 2050 is relatively modest (7 per cent). However, under the new fertility assumptions the shape of the age pyramid in 2050 shifts (Figure 8). Since the cohorts that in 2050 will reach age 55 and above were born before 1995, their sizes in 2050 are identical in both scenarios. The cohort born in 1995-2000, aged 50-55, will be 5 per cent smaller under the fertility assumptions of the 1998 Revision; the following 4 five-year age groups, aged 30 to 49, will be 8 per cent smaller; in the age group 20-24 the difference is 10 per cent, in the age group 15-19 it is 16 per cent; in the age group 10-14 -- 20 per cent, and in the youngest age groups (5-9 and 0-4) it reaches 23 and 27 per cent, respectively.
Day, L. H. (1995). Recent fertility trends in industrialized countries: toward a fluctuating or a stable pattern? European Journal of Population/Revue Européenne de Démographie (Dordrecht, Netherlands), Vol. 11, No. 3, pp. 275-88.
Cruijsen, H. (1994). National population projections in the European Economic Area: a field in motion. Paper prepared for the ECE/Eurostat Joint Working Session on Demographic Projections, Mondorf-les-Bains, Luxembourg, 1-4 June.
Eurostat (1991). Two long term population scenarios for the European Community. Paper prepared for the International Conference on Human Resources in Europe at the Dawn of the 21st century, Luxembourg, 27-29 November.
Keilman, N. and H. Cruijsen (1992). National population forecasting in industrialized countries. Amsterdam: NIDI/CBGS (Publication No. 24).
Lutz, W. (1994). Future reproductive behavior in industrialized countries. In Wolfgang Lutz, ed., The Future Population of the World: What Can We Assume Today? London: Earthscan, pp. 267-294.
Van Hoorn, W. and N. Keilman (1997). Birth expectations and their use in fertility forecasting. Eurostat Working Paper. Series "Population and social conditions". E4/1997-4.
Vishnevsky, A. (1976). Demograficheskaya revolutsiya [Demographic Revolution, in Russian]. Moscow: Statistika.
(1991). Demographic revolution and the future of fertility: a systems approach. In Wolfgang Lutz, ed., Future Demographic Trends in Europe and North America: What Can We Assume Today? London: Academic Press, pp. 257-270.
Westoff, Ch. F. (1991). The return to replacement fertility: a magnetic force? In Wolfgang Lutz, ed., Future Demographic Trends in Europe and North America: What Can We Assume Today? London: Academic Press, pp. 227-234.
Return to Population Division Publications Archive Home Page