The time to decline: tracing a cohort’s descendants in below replacement populations
© The Author(s) 2018
Received: 26 September 2017
Accepted: 5 January 2018
Published: 23 January 2018
A number of contemporary populations are exhibiting sustained fertility at levels substantially below long-term replacement. Nonetheless, relatively few populations are actually diminishing in size. Here, we approach that apparent paradox by analyzing the time before the number in a birth cohort, and its descendants, falls below the initial number in the cohort. First, models are examined with constant below replacement fertility, cohort extinction at age 75 or 85, and no mortality below the highest age attained. For a net reproduction rate (NRR) of 0.75, it takes 150 years for the cohort’s descendants to be fewer than the cohort’s original size if persons live to age 85, and over 130 years if persons live to age 75. If the NRR is at least 0.60, it takes a century before the descendants are fewer in number than the original cohort. Second, projections are done for the USA 2012, Italy 2012, and Hong Kong 2011 assuming that fertility and mortality remain constant. The results resemble the projections. For example, in Italy, with actual mortality and an NRR of 0.70, it takes over 125 years before the descendants of a cohort are fewer in number than the initial cohort. A relatively simple equation for the long term “time to decline” is presented, showing that it depends primarily on the level of fertility, secondarily on longevity, and only modestly on the mean age of fertility.
Replacement level is when, on average, every woman has one daughter. The net reproduction rate (NRR), the average number of daughters a cohort of women bears subject to a given set of fertility and mortality rates, is then 1. Over the last three decades, below replacement fertility has spread to characterize most of Europe, overseas Europe, and East Asia. For the 2010–2015 period, the United Nations Population Division estimated that Europe had an NRR of 0.763, below replacement since 1975–1980; Eastern Asia an NRR of 0.699, below replacement since 1990–1995; and Northern America an NRR of 0.896, below replacement since 1970–1975. The Chinese Autonomous Region of Macau has the lowest fertility of any listed entity, with an NRR of 0.577 (UN Population Division 2015, File FERT/5). In contrast, only Germany, Japan, and some countries in Southern and Eastern Europe have a negative rate of natural increase (i.e., more deaths than births). The largest rate of natural decrease, in Bulgaria 2010–2015, was only 0.57% per year, a rather slow rate of decline (UN Population Division 2015, File POP/3).
The explanation for the broad prevalence of below replacement fertility and the modest extent of negative natural increase is population momentum, an important concept introduced in Keyfitz (1971). Population momentum is the factor an initial population will grow (or decline) after it has an immediate shift to replacement level fertility. In general, growing populations tend to continue growing because they have large cohorts at the reproductive ages and small cohorts at older ages. Preston (1986) examined growing populations that experienced a fall in fertility to replacement level and found that, over the transition to zero growth, the population under age 30 remained virtually constant, the population between the ages of 30 and 60 increased by a factor approximately equal to the pre-shift NRR, and the population 60 to 90 increased by a factor approximately equal to the square of that NRR. An in-depth discussion of population momentum can be found in Schoen (2006, Chap. 3).
Momentum, however, is a population level concept that does not provide a cohort level perspective or explicitly consider the descendants of a cohort. Here, we do so, starting with a birth cohort and tracing its descendants over time. The goal is to determine, in terms of a fixed level and pattern of below replacement fertility, how many years pass before the sum of the surviving members of the cohort, and the number of its living descendants, falls below the initial number in the cohort.
Projecting a cohort and its descendants in a simplified model
The first approach used here projects a birth cohort and its descendants in the usual interval-by-interval manner. We then derive an analytical procedure for projecting the model population to any future point. The principal goal is to find the time at which total population size falls below the number in the initial cohort.
The simplified projection model
Let the initial population consist only of females in the first age group, and consider only female births. Assume that age-specific fertility remains constant over time and that the population is closed to migration. For simplicity, to focus on fertility, and to recognize the low mortality prevailing in most contemporary below replacement populations, assume that there is no mortality below the highest age attained, with the cohort becoming extinct on attaining that age. In separate calculations, we take that highest age to be 75 or 85 years.
The standard cohort-component approach to population projection advances an initial population using a Leslie matrix, i.e., a projection matrix that has fertility values in its first row and survival probabilities on its subdiagonal (Preston et al. 2001, Chap. 6). While projections have become quite sophisticated (cf. Sevcikova et al. 2016), here, we use a basic approach that proceeds using 5-year age and time intervals and continues past the point where the total number in the population is less than the initial number in the cohort.
The model population projection procedure
Projections were carried out for NRR levels from 0.50 to 1.0 and for fertility patterns with a mean age of 25, 29, and 33 years. Those bounds span the range of fertility levels and patterns in most below replacement fertility populations.
A base fertility pattern, roughly following patterns in the USA during the late 20th century, was created for each mean age of fertility, and those patterns are shown in Appendix Table 4. In each base pattern, the sum of the fertility values is one.
An analytical alternative to interval-by-interval projection
In the model described above, long-term projections can be made analytically, as well as interval by interval. To do so, we now derive an expression for total population size at any time point based on two parameters, μ and r.
For a sufficiently large t, an interval whose length is examined below, matrix F t becomes a rank one matrix, i.e., it can be represented as the product of a column vector, u, and a row vector, v' (Schoen 2006, p. 28). Here, e(0) = 85, there is no mortality during the first 17 age intervals, and F is consistent with zero growth. Hence 17 × 1 column vector u is a vector of ones. The 1 × 17 row vector v' has first element 1/μ and jth element Σi = j f i /μ (Schoen 2006, p. 157, #2b).
Thus, in the long term, Eq. (12) shows that total population size at time t is a constant factor, c/μ, times a decreasing exponential term in negative r. The constant factor represents the size of the implicit initial stable population divided by the mean age of fertility (approximately the length of a generation).
Thus, in the long term, time to decline t* can be expressed in terms of two parameters, the mean age of fertility, μ, and the intrinsic growth rate, r, with Eq. (5) relating r to μ and the NRR.
The results for the simplified projection model
The interval-by-interval projection results
Time to decline (t*) values, in single years, by level and mean age of fertility
μ = 25 years
μ = 29 years
μ = 33 years
e(0) = 75 years
e(0) = 85 years
e(0) = 75 years
e(0) = 85 years
e(0) = 75 years
e(0) = 85 years
There are huge differences in the time to decline by NRR level. The higher the NRR, the larger t*. For NRRs of 0.65, t* is a century or more; for NRRs of 0.75, t* is 133 or more years; and for NRRs of 0.95, t* is well over five centuries. Differences by mean age of fertility are present, but relatively small. There are no “tempo” effects, as fertility is held constant.
As to be expected, the time to decline is less when life expectancy is 75 than when it is 85 years. Proportionally, however, the effect of a 10-year smaller e(0) is not all that large relative to t*. When the NRR is 0.95, the difference in t* is some 65 to 85 years (10–13%), while when the NRR is 0.75, differences are about 15 to 20 years (again 10–12%).
Equation (14) suggests that a higher mean age of fertility is associated with a longer time to decline. However, parameter c is a function of r, which in turn is a function of μ; hence, a change in μ can introduce offsetting changes in r and c. Table 1 shows that while there is a tendency for t* to increase when the mean age of fertility rises, there are many exceptions to that pattern.
μ (in years)
where TFR* is the corresponding total fertility rate (TFR). The TFR is set equal to 2.05 times the NRR, using the customary sex ratio at birth of 105 males per 100 females. A TFR of 1.3, that is women having an average of 1.3 children, is considered very low fertility, but even sustained fertility at that level implies a time to decline of about 110 years.
The analytical results
Analytical projections using Eq. (14) are also presented in Table 1 for e(0) = 85. The equation-based values implicitly assume that the population is approximately stable, which initially is far from the case. Stability takes several generations to arise, even more when one begins with a single cohort, and is slower for higher mean ages of fertility.
For NRRs of 0.70 or above (0.75 when μ = 33 years), the equation-based t* values are quite close to the projections. For NRRs less than 0.60, the Eq. (14) values are poor.
The trajectory of a cohort
For observed populations, a rough rule of thumb is that the population becomes approximately stable up to age x after about (60 + x) years (Schoen 2006). Here, we begin with a single cohort age 0, and Table 1 suggests that total population size is not roughly stable for up to 150 years. At the 150-year time point, growth is approximately stable (i.e., equal to exp(5r)), though projected and stable values still differ by 1½–3% when μ = 33 years.
The total size of a cohort and its descendants, as calculated by projection and by Eq. (10), for NRR = 0.75 and μ = 29 years
Time (t) [years]
Projected total size [PT(t)]
Growth rate [PT(t)/PT(t − 5)]
Total size from Eq. (10)
− 0.000 54
− 0.000 06
− 0.000 23
− 0.000 28
Column (2) of Table 2 shows that for the first three time periods, before the cohort reaches age 15, the projected total population remains at 1. Between times 15 and 80 years, population size steadily increases, as the cohort and its offspring reproduce. At time 85 years, when the initial cohort has just died, total population size drops by nearly 1. After that point, Fig. 1 shows how total population size fluctuates in waves of diminishing amplitude around the stable population size given by the Eq. (10)-based decreasing exponential [the plotted values are shown in column (4) of Table 2].
Column (3) presents the interval by interval population growth rates. After some 130 years, the growth rate of the projected population becomes quite close to the ultimate stable population growth rate of 0.9518. Column (5) compares the Leslie and analytical projections, showing the proportional error in the Eq. (10) projection. The direction of the error varies because of the fluctuations in the growth rate of the projected population. After 120 years, the two trajectories are within 1% of each other, with the difference inconsequential after 145 years.
Projecting three observed populations
The time to decline based on observed rates in three contemporary populations
Hong Kong 2011
Life expectancy at birth
Total fertility rate
Net reproduction rate
Mean age of fertility
Variance of age of fertility
Time to decline (t*)
Ratio of total population size to initial cohort size at year:
The USA 2012, with an NRR of 0.92, requires over 340 years before the descendants of a cohort are fewer than the original number in the cohort. Even at time 200, the descendants of the initial cohort are 62% more numerous. Hong Kong 2011, with an NRR of 0.76, has a time to decline over 155 years, but at time 200, the descendants of the initial cohort number only about two thirds the size of the initial cohort. Italy 2012, with an NRR of 0.70, has a time to decline of just over 125 years, and at time 200, the descendants of the original cohort number only 43% of their initial number.
The introduction of actual mortality rates seems to have only a small effect, as the observed population rates yield times to decline quite similar to those in Table 1. For example, Italy has a time to decline of 126.7 years with an NRR of 0.70, a life expectancy of 84.6 years, and a mean age of fertility of 31.4 years. For an NRR of 0.70 and a life expectancy of 85 years, Table 1 has a time to decline of 132.1 years for μ = 29 years and 127.1 years for μ = 33 years.
Summary and conclusions
Population momentum means that a growing population continues to increase in size for some years after its fertility falls to replacement. Here, we take a birth cohort, individual level, perspective, and examine how long it takes, under below replacement fertility, before the number of survivors of a cohort, plus its living descendants, falls below the number in the initial cohort. With no mortality below the highest age attained (75 or 85), and constant below replacement fertility, we find that t*, the time to decline, varies dramatically with the level of fertility, moderately with longevity, and modestly with the mean age of fertility.
As the cohort reproduces, the succession of generations prolongs the onset of population decrease. When e(0) is 85 and the NRR is 0.95, it takes nearly 650 years for total population size to be less than 1. For an NRR of 0.50 or below, t* is 85 years, the point at which the initial cohort dies. Projections with the observed vital rates of three contemporary below replacement populations suggest that the model simplifications have only a minor effect on the time to decline.
Analytically, Eq. (10) allows a simplified projection in terms of life expectancy, the NRR, and the mean age of fertility. That procedure tracks the projected population size quite closely after about 150 years.
Even with an NRR of 0.65, an e(0) of 75, and no migration, the time to decline is almost a century. A cohort produces its own momentum, propelled by the arrival of succeeding generations.
Helpful comments and suggestions from Lowell Hargens and Samuel H. Preston are acknowledged with thanks, as is assistance from Carl Boe.
No funding was received for the work reported.
Availability of data and materials
No data were used in the paper, apart from the publicly available sources cited and the fertility schedules, produced by me, that are shown in Appendix Table 4.
I am the Inaugural Hoffman Professor of Family Sociology and Demography, Distinguished Senior Scholar of the Department of Sociology, and Affiliate of the Population Research Institute at Pennsylvania State University, University Park. I received a PhD in Demography from the University of California, Berkeley in 1972.
The author declares that he has no competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
- Arias, E., Heron, M., & Xu, J. G. (2016). Life table for females, table 3, National Vital Statistics Reports, Vol 65, No 8, November 28. Hyattsville: National Center for Health Statistics.Google Scholar
- Coale, A. J., & Trussell, T. J. (1974). Model fertility schedules: variations in the age-structure of childbearing in human populations. Population Index, 40, 185–258.View ArticleGoogle Scholar
- Hong Kong Special Administrative Region, Census and Statistics Dept. Hong Kong Life Tables 2011–2066. Downloaded November 26, 2017 from www.statistics.gov.hk/pub/B1120016072017XXXXB0100.pdf.
- ISTAT. 2016. Life tables of the resident population, Italy – Females – Year: 2015, downloaded November 26, 2017 from demo.istat.it/tvm2016/index.php?lingua=eng.
- Keyfitz, N. (1971). On the momentum of population growth. Demography, 8, 71–80.View ArticleGoogle Scholar
- Keyfitz, N. (1977). Introduction to the mathematics of population (2d ed.). Reading: Addison-Wesley.Google Scholar
- Preston, S. H. (1986). The relation between actual and intrinsic growth rates. Population Studies, 40, 343–351.View ArticleGoogle Scholar
- Preston, S. H., Heuveline, P., & Guillot, M. (2001). Demography: measuring and modeling population processes. Oxford: Blackwell.Google Scholar
- Schoen, R. (2006). Dynamic population models. Dordrecht: Springer.Google Scholar
- Sevcikova, H., Li, N., Kantorova, V., Gerland, P., & Raftery, A. E. (2016). Age-specific mortality and fertility rates for probabilistic population projections. In R. Schoen (Ed.), Dynamic demographic analysis (pp. 285–310). Dordrecht: Springer.View ArticleGoogle Scholar
- Sykes, Z. (1973). Intrinsic age-specific birth rates: a new method of fertility analysis, Unpublished manuscript. Baltimore: Department of Population Dynamics, Johns Hopkins University.Google Scholar
- United Nations, Population Division. (2015). World population prospects: the 2015 revision. New York: United Nations.Google Scholar
- United Nations Statistical Division. 2014. United Nations Demographic Yearbook 2013. Downloaded November 24, 2017 from unstats.un.org/unsd/demographic/products/dyb/dyb2013.htm.