World Population Futures (Sept 2001)

8/9/2019 World Population Futures (Sept 2001)

1/44

Aging, HIV/AIDS,and environmental

concerns drawincreased attention to

population projections.

Fertility trends are keyto projections of futureworld population.

New methods helpcommunicate theuncertainty of

projections.

World Population

Futuresby Brian ONeill and Deborah Balk

BULLETINA publication of the Population Reference Bureau

Population

Vol. 56, No. 3

September 2001


2/44

Population Reference Bureau (PRB)Founded in 1929, the Population Reference Bureau is the leader in providing timely, objectiveinformation on U.S. and international population trends and their implications. PRB informspolicymakers, educators, the media, and concerned citizens working in the public interestaround the world through a broad range of activities including publications, informationservices, seminars and workshops, and technical support. PRB is a nonprofit, nonadvocacyorganization. Our efforts are supported by government contracts, foundation grants, individ-ual and corporate contributions, and the sale of publications. PRB is governed by a Board of

Trustees representing diverse community and professional interests.

OfficersMichael P. Bentzen, Chairman of the Board

Partner, Hughes and Bentzen, PLLC, Washington, D.C.

Patricia Gober, Vice Chairwoman of the Board

Professor of Geography, Arizona State University, Tempe, Arizona

Peter J. Donaldson, President

Population Reference Bureau, Washington, D.C.

Montague Yudelman, Secretary of the Board

Senior Fellow, World Wildlife Fund, Washington, D.C.

Richard F. Hokenson, Treasurer of the BoardDirector of Demographic Research, Credit Suisse First Boston, New York

TrusteesFrancisco Alba, Professor, El Colegio de Mxico, D.F., Mxico

Jodie T. Allen, Senior Writer, U.S. News & World Report, Washington, D.C.Patty Perkins Andringa, Consultant and Facilitator, Bethesda, MarylandPape Syr Diagne,Director, Centre for African Family Studies, Nairobi, KenyaBert T. Edwards,Executive Director, Office of Historical Trust Accounting, Office of the Secretary,

U.S. Department of the Interior, Washington, D.C.

Klaus M. Leisinger,Executive Director, Novartis Foundation for Sustainable Development, Basel,Switzerland

Karen Oppenheim Mason,Director, Gender and Development, The World Bank, Washington, D.C.Francis L. Price, Chairman and CEO, Q3 Industries and Interact Performance Systems, Columbus, Ohio

Douglas Richardson,Director, Research and Strategic Initiatives, Association of American Geographers,and Founder and Director, GeoResearch Institute, Washington, D.C., and Bethesda, MarylandCharles S. Tidball, M.D., Professor Emeritus of Computer Medicine and Neurological Surgery,

School of Medicine and Health Sciences, George Washington University, Washington, D.C.

Barbara Boyle Torrey,Executive Director, Commission on Behavioral and Social Sciences,National Research Council, National Academy of Sciences, Washington, D.C.

Mildred Marcy, Chairwoman Emerita

Editor: Mary Mederios KentDesign/Production: Heather LilleyTwo photos in this publication were selected from M/MC Photoshare at: www.jhuccp.org/mmc.

The Population Bulletinis published four times a year and distributed to members of the Popu-

lation Reference Bureau. Population Bulletinsare also available for $7 (discounts for bulkorders). To become a PRB member or to order PRB materials, contact PRB, 1875 Connecticut

Ave., NW, Suite 520, Washington, DC 20009-5728; Phone: 800/877-9881; Fax: 202/328-3937;E-mail: [email protected]; Website: www.prb.org.

The suggested citation, if you quote from this publication, is: Brian ONeill and Deborah Balk,World Population Futures, Population Bulletin, vol. 56, no. 3 (Washington, DC: PopulationReference Bureau, September 2001). For permission to reproduce portions from the Popula-tion Bulletin, write to PRB, Attn: Permissions

2001 by the Population Reference BureauISSN 0032-468X

Printed on recycled paper


3/44

1

BULLETINA publication of the Population Reference Bureau

Population

Vol. 56, No. 3

September 2001

World Population FuturesProjecting Populations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

Figure 1.World Population Projections to 2050 and 2100: The United . . . . . .Nations, World Bank, U.S. Census Bureau, and IIASA . . . . . . . . . . . . . . . . 5

How Are Populations Projected? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6Figure 2. Projecting a Cohort of U.S. Women Ages 1519 in 2000 to 2005:

The Cohort-Component Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7Box 1.Accuracy of Population Projections . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8Box 2. Using Scenarios to Show Uncertainty . . . . . . . . . . . . . . . . . . . . . . . . . 10

Projecting Fertility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10Box 3. Using Probabilities to Account for Uncertainty . . . . . . . . . . . . . . . . . . 12Box 4. Explaining Fertility Decline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14Figure 3. Completed Fertility for European Women, Selected

Countries and Birth Cohorts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

Projecting Mortality. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20Figure 4. Projected Life Expectancy for 2000 in the 1980 and 2000 UN

Projection Series, Selected Regions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

Migration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

Projection Outcomes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29Figure 5.Alternate Projections for Brazil: UN and U.S. Census Bureau . . . . 29Figure 6.Alternate Projections for Nigeria: UN and U.S. Census Bureau . . . 30Figure 7. UN and IIASA World Population Projections, High and Low

Scenarios, 20002100 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31Figure 8.Annual World Population Growth and Population Growth Rate,

UN Projections, 19502050 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

Implications of Future Growth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32Figure 9.World Population Age 60 or Older in 2000 and 2100:

Six Scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34Figure 10.World Population by Region or Country: UN Projections

to 2050 and 2100. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

Appendix Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

Suggested Resources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40


4/44

2

About the Authors

Brian ONeillis an assistant professor (research) at the Watson Institute for International

Studies at Brown University. His research interests are in population-environment interactionsand the science and policy of climate change. He holds a doctoral degree in Earth systems sci-ence from New York University. He has written numerous articles on population and environ-mental issues and is the lead author ofPopulation and Climate Change (2001).

Deborah Balkis an associate research scientist at the Center for International Earth ScienceInformation Network (CIESIN) at Columbia University where she is project scientist for theSocioeconomic Data and Applications Center (SEDAC). She holds a doctoral degree in demogra-phy from University of California, Berkeley. Her research has focused on gender, fertility, andthe family, with more recent emphasis on interactions between population and the environment.She is currently working on studies of climate, population, and health in Africa and on geospa-tial demography of urban areas.

The authors would like to thank Melanie Brickman for her invaluable assistance and

Mary Kent for her thoughtful editing of thisBulletin

. For their commentary on earlier versions,we thank: John Bongaarts, Ed Bos, Thomas Buettner, Randy Bulatao, Bob Chen, DianaCornelius, Patricia Dickerson, Robert Engelman, Kees Klein Goldewijk, Anne Goujon, PeterJohnson, Nico Keilman, Ron Lee, Susan Motzer, Evert van Imhoff, and Hania Zlotnik. Wethank Wolfgang Lutz for providing recent data. The U.S. National Aeronautics and SpaceAdministration (NASA) contributed funding through SEDAC for this work and a more detailedreport, A Guide to Global Population Projections, by Brian C. ONeill, Deborah Balk,Melanie Brickman, and Markos Ezra, published in the online journalDemographicResearch, 2001.

2001 by the Population Reference Bureau


5/44

3

W

hat will the future inhabi-tants of the world be like?How many will there be, and

what kind of world will they live in?We can only speculate about theanswers to these questions, but we canbe reasonably sure that populationcharacteristics and social and environ-mental factors are likely to becomemore interconnected. Global environ-mental changes, for example, will bedriven in part by the evolving size,geographic distribution, and makeupof the worlds population. In turn,

changes in societies, economic sys-tems, and the environment will influ-ence population dynamics.

The nature of these linkages isunclear. Scientists do not agree onhow (and how much) demographics,in concert with social, economic, andcultural forces, affect the environ-ment; and they cannot know preciselyhow much socioeconomic and envi-ronmental factors will sway individu-als future decisions about when or

whether to have children, practicegood health, or move to a new coun-try. Yet the fact that forecasts of futurepopulation dynamics are inherentlyuncertain does not make them anyless important. Scientists and policy-makers are turning more attention topopulation projections. Their interestis driven by concern about the poten-tial effects of aging populations onsocial security systems and economicgrowth, the possibility of declining

populations in some industrialized

countries, the long-term conse-quences of HIV/AIDS, and the impli-cations of demographic trends for

long-term environmental changessuch as global warming and loss ofbiodiversity.

At the same time, researchers havetaken a renewed interest in themechanics of population projections,not only to improve accuracy, but alsoto make the results more useful andthe methodology easier to understandfor experts in a variety of academicfields and policy arenas. Demogra-phers are experimenting with creative

ways to express the uncertainty inher-

World Population Futuresby Brian ONeill and Deborah Balk

The size and characteristics of the worlds future population willdepend primarily on how many children women havebut thisindividual behavior will be influenced by future socioeconomic,political, health, and environmental trends.

Photo removed forcopyright reasons.


6/44

4

ent in all projections and with newapproaches for projecting populationsize and other characteristics. With agrowing wealth of census and surveydata and medical studies from aroundthe world, researchers are also refin-ing theories about how reproductivebehavior and childbearing prefer-

ences may change, and on likelyimprovements in life expectancy.

This Population Bulletinexplainsprojection methodology and discussesvarious approaches for expressinguncertainty. It analyzes the key assump-tions on which most global projectionsare based: baseline demographic dataand trends in future fertility, mortality,and migration. The Bulletinalsoreviews the conceptual basis forprojecting demographic variables, dis-cusses the extent to which environ-mental factors are or should be takeninto account, and compares assump-tions made by different institutions. Itconcludes with a discussion of whatglobal population projections implyabout the kind of world our descen-dents will inhabit.

ProjectingPopulationsThe population of the world (or of anygeographic area) can be projectedinto the future based on currentknowledge about population size andage structure, rates of birth, death, andmigration, and assumptions about howquickly these rates will change. Theprojection results, or output, mayinvolve very different geographic

areas, time horizons, or populationcharacteristics, and they may be tar-geted for a number of different uses.For global or national populations, atime horizon of less than 15 yearsmight be considered short-term; 15 to50 years, medium-term; and morethan 50 years could be considered along-range projection.1 The accuracy,geographic coverage, and populationcharacteristics typically vary dependingon whether the projections are short-,

medium-, or long-term.

Spatial dimensions can range fromlocal areas like counties or cities tothe entire world. Local-area projec-tions tend to use shorter time hori-zons, often less than 10 years, whereasnational and global projections canextend decades into the future, andin some cases, for more than a cen-

tury. Short- and medium-term projec-tions are more likely than long-termprojections to include more than thenumber and age and sex profile ofthe future population. They may pro-ject such socioeconomic characteris-tics as educational and labor forcecomposition, ethnicity, urban resi-dence, or household type.2

The intended user of the projec-tion results, or output, usually deter-mines the level of detail and timehorizon. Businesses often use projec-tions for marketing research; theygenerally want a single most likelyforecast of population classified bysuch socioeconomic categories asincome and consumption habits (inaddition to age and sex) and by placeof residence. Government plannersmay be concerned with populationaging and its potential social and eco-nomic impact. They might want, for

example, longer-term projections ofthe likely health status and livingarrangements of the elderly.

Governments and the public policycommunity, including advocacygroups, often are more interested in arange of likely scenarios that reflectthe potential influence of a policyrather than a single best guess offuture population size. Those con-cerned with the environmental effectsof population growth, for example,

may be interested in how policies toreduce fertility might affect futurepopulation size. In addition, they maywant to study how environmentalchange might affect demographicchange, and vice versa. Rapid popula-tion growth might promote overuseof agricultural land, for example,which would deplete resources, andin turn, encourage migration out ofthe area, which would slow popula-tion growth. Researchers studying

global environmental changes often


7/44

5

use population projections as a vari-able in models that predict energyconsumption, food supply, and globalwarming.3 These studies usuallyrequire projections with long timehorizons (a century or longer) andseveral scenarios rather than a singlemost likely projection.

Demographers often are uncom-fortable making projections morethan a few decades into the future,when most of the population will bemade up of people not yet born.Nonetheless, long-term global projec-tions are increasingly in demand byglobal change researchers.

While individual researchers andinstitutions have made significantcontributions to the methods used toproject population, especially at thenational level (or below), global pro-jections have been the province ofrelatively few institutions: the UnitedNations (UN), the U.S. CensusBureau, the World Bank, and theInternational Institute for Applied Sys-tems Analysis (IIASA), based in Aus-tria. They use different methodologies,make varying assumptions aboutfuture fertility, mortality, and migra-tion trends, and begin with slightly dif-

ferent estimates of current populationsize. Their results tend to fall within arelatively small band for the next 50years, then diverge as the time horizonlengthens (see Figure 1).

Global Projection SeriesThe UN assumed the leadership rolein the production of projections andthe dissemination of their resultsbeginning in the 1950s, long before

the U.S. Census Bureau, the WorldBank, and IIASA began to produceglobal projections. Between 1951 and2001, the UN produced 17 sets of esti-mates and projections covering allcountries and areas of the world. Until1978, the UN published new revisionsapproximately every five years; sincethen, it has published revisions everytwo years. These medium-term projec-tions, published in the UNs WorldPopulation Prospectsseries, include vari-

ous scenarios with different assump-

tions about future birth rates and,more recently, include alternative sce-

narios for average life expectancy andmigration.The UN projections are available in

print, online through the UN website,and on CD-ROM, and they are themost widely cited throughout theworld. UN projections are used forplanning by individual governmentsand by the UN and other internationalagencies, as well as by the media, aca-demics, and research institutions.

The World Bank was the second

major institution to produce countryand global population projections.The World Bank first publishedcountry-level population projectionsin the annual World Development Reportin 1978, although they prepared ear-lier projections for internal use. TheWorld Bank projections did notextend as far into the future as didthe UN series, but they did identifythe year in which each countryspopulation was projected to stop

growing. Later editions of the World

0

2

4

6

8

10

Population in billions

U.S. Census Bureau

World Bank

IIASA* median

UN medium

2000 2025 2050 2075 2100

Figure 1World Population Projections to 2050and 2100: The United Nations, WorldBank, U.S. Census Bureau, and IIASA

*International Institute for Applied Systems Analysis.

Sources: United Nations, Long-Range World Population Projections Basedon the 1998 Revision(1999); U.S. Census Bureau, International DataBase, accessed online at: www.census.gov/ipc/www, July 10, 2001;The World Bank, World Development Indicators 2001 CD-ROM; W.Lutz, W. Sanderson, and S. Scherbov, Nature(Aug. 2, 2001): 543-46;and unpublished data from IIASA.


8/44

6

Development Reportcontained popula-tion projections to 2000 and 2025.About every two years between 1984and 1995, the World Bank producedlong-term projections of world popu-lation out to 2150.4While the WorldBank no longer publishes long-termprojections, it continues to create

projections for use in projects andplanning within the World Bank, forexample, to anticipate the demandfor pensions, education resources,and health care. Since 1997, theWorld Bank has included medium-term projections of country popula-tions, which are updated annuallyand available on their World Develop-ment IndicatorsCD-ROM.

The U.S. Census Bureau has beencompiling and evaluating interna-tional population statistics since the1950s, primarily by assisting the statis-tical offices in less developed coun-tries and by preparing estimates ofpopulation and vital rates. The Cen-sus Bureau has published projectionsfor all countries and for world regionsin the World Population Profilesince1985.5 The Census Bureau publishesprojections prepared under one set ofassumptions, and prints the results for

15 to 25 years into the future. WorldPopulation Profile: 1998includes pro-jection results for countries andregions through 2025. Projectionsthrough 2050 are offered in an onlineservice that is updated more often.The Census Bureau projections areused by other U.S. government agen-cies to help manage and design for-eign assistance programs, and forlong-range planning and other uses,as well as by national governments

and nongovernmental organizationsaround the world.The Population Project at IIASA

first produced a set of long-rangeglobal population projections in 1994and updated them in 1996 and 2001.6

IIASA projections are made for 13regions of the world through 2100.The earlier projections used three sce-narios of fertility, mortality, and migra-tion, which yield a possible 27 outputscenarios. Additional projection series

can be created by combining different

migration scenarios with different sce-narios for fertility and mortality ineach region.

How Are

PopulationsProjected?The population of a geographic areagrows or declines through the interac-tion of just three variables: fertility,mortality, and migration. To projectthe size of a population at a futuredate, demographers generally makean assumption about levels of fertilityand mortality and about how manypeople will move in or out of the areaduring the projection period. The netpopulation increase or decrease overthe period (derived from the numberof births and in-migrants minus thenumber of deaths and out-migrants)is added to the baseline population toproject the future population size.

Nearly all national and global popu-lation projections are produced fromassumptions about these three demo-graphic variables using some variant

of the cohort-component method.7

Under the cohort-component method,an initial population for a country orregion is grouped into cohorts definedby age and sex. Women ages 15 to 19in 2000 would make up one cohort ofthe population, for example. Eachcohort is projected forward accordingto assumed migration and mortalityrates for that age and sex group. TheU.S. Census Bureau estimates that inthe year 2000, for example, there were

9,672,000 females ages 15 to 19 resid-ing in the United States. The CensusBureau projects that by the year 2005,when members of this cohort will beages 20 to 24, the cohort will havegrown by 230,000 to number 9,902,000(see Figure 2). This cohort will loseabout 115,000 women from deathsover the period, while it gains about345,000 women from internationalmigration (the Census Bureau projectsthat 345,000 more women in this age

group will move into the United States

Assumptionsused to project

population arebased on expert

opinion.


9/44

7

than will move out between 2001 and2005). Similar calculations are madefor each age group and for both sexes.

New cohorts are added at the bot-tom of the age structure by birthsover the projection period. The num-ber of births is projected by applyingassumed birth rates to the base popu-

lation. The Census Bureau furtherdivides the U.S. population by racialand ethnic groupso that eachcohort is defined by age, sex, andrace or ethnicity. The Census Bureauassumes slightly different fertility,mortality, and migration rates foreach racial and ethnic group.

The cohort-component methodwas the major innovation in the evo-lution of projection methodology. Itwas first proposed by the Englisheconomist Edwin Cannan in 1895,and was then reintroduced bydemographer Pascal Whelpton in the1930s, formalized in mathematicalterms by P.H. Leslie in the 1940s, andfirst used to produce a global popula-tion projection by demographerFrank Notestein in 1945.8 SinceNotesteins 1945 projection, thecohort-component method hasbecome the dominant means of pro-

jecting population. It has remainedessentially unchanged, but it has beenextended by incorporating popula-tion characteristics such as region ofresidence or educational status (mul-tistate projections) and by innova-tions in ways to demonstrate theuncertainty in projection results.9

The cohort-component model isnothing more than a particularly use-ful accounting scheme: It works outthe inevitable consequences of the size

and age structure of the population atthe beginning of the period and thefertility, mortality, and migration ratesassumed to prevail over the projectionperiod. The real work in producingprojections lies not in refining themechanics of the model itself, but inestimating the population size and agestructure in the base period and inforecasting future trends in fertility,mortality, and migration.

Although approaches may differ,

the assumptions used to produce

global population projections arebased on expert opinion informed bycurrent conditions, past trends, andtheories about why and how muchfertility, mortality, and migration arelikely to change. Demographers drawon specialized knowledge about thecomponents of population changeto develop the assumptions used

in projections.

Baseline DataPopulation projections must beginwith an estimate of the baseline data:the number of people in each age andsex cohort of the population at thebeginning of the projection period.The primary sources of baseline dataare national population censuses,which are carried out about once a

decade in most countries of the world.

Ages 75+

2000

Ages 75+

2005

70-74 70-74

65-69 65-69

25-29

9,902,000women

ages 20-24

9,672,000women

ages 15-19

10-14 10-14

5-9 5-9

20-24

25-29

+345,000migrants

-115,000deaths

15-19

0-4 0-4

Figure 2Projecting a Cohort of U.S. Women Ages 1519 in2000 to 2005: The Cohort-Component Method

Source: Data from U.S. Census Bureau. Adapted from J. Cohen, How Many People

Can the Earth Support?(1995): figure 7.2.


10/44

8

Projections from the United Nations (UN)and the World Bank have become more accu-rate over time, as measured by their ability toforecast the population for 2000. UN projec-tions of the world population size in 2000

made in the early 1970s were off by 6 percentto 7 percent, while projections made in the1990s were off by less than 1 percent. Butmost of this improvement in projection accu-racy reflects the fact that more recent projec-tions had less time to go wrong before 2000.

When comparing projections with equal timehorizons10 years into future, for exam-plethere is little evidence of improvement.

Projections of population size tend to bemore uncertain, or less accurate, under par-ticular circumstances.1 They are less accurate

(1) for less developed countries than formore developed countries, partly because lessdeveloped countries tend to have limited andless reliable data; (2) for smaller countriesthan larger ones, perhaps stemming from thegreater attention devoted to larger countries;(3) in younger and older age groups than inmiddle age groups because incorrect assump-tions about fertility and mortality have agreater effect at older and younger ages; and(4) at the country level than at regional orglobal levels because errors at the countrylevel partly cancel each other when aggre-

gated to regions or to the world. Countriesare more susceptible to errors from migra-tion assumptions, and regions are moreinfluenced by larger countries, for whichprojections tend to be more accurate.

Projecting vital rates has also proved to bedifficult. UN projections of fertility rates haveconsistently been too high for most regions ofthe world. In Latin America, for example, theestimates of fertility rates at the start of theprojection period often were too high, whichcontributed to excessively high projections offuture fertility rates. In addition, most projec-tions by the UN and other organizationsanticipated a halt to declines in fertility, whilein many countries fertility continued to fall

well below replacement level.The UN has generally been too pessimistic

about increases in life expectancy. Projectionsfor North America in the 1970s failed to fore-see the persistent rise in life expectancy

above 70 years. A lack of accurate base periodor baseline data also contributes to inaccu-racy in the projected life expectancy in manycountries. Projections for India in 1975 and1980, for example, underestimated life

expectancies by several years because baselineestimates were too low. The forecasts of lifeexpectancy in Africa are an exceptiontheyconsistently have been too optimistic, missingespecially the flattening in life expectancyafter 1985, in part because of HIV/AIDS.

UN projections of urban populationgrowth in less developed countries have alsogenerally been too high.2 The most recentprojections, made in 1999, foresaw an urbanpopulation in 2000 that is 9 percent smallerthan the UN had projected in 1980. This dif-

ference is not caused primarily by slower thanexpected growth of total populationprojec-tions of total population have been revised byonly 2 percent over the same periodbutrather to overestimating the rate of urbaniza-tion itself. The reasons for a slower thanexpected growth of urban population in lessdeveloped countries are not clear, but evi-dence suggests that weak expansion of urbanindustries, population aging, and policiesaffecting population distribution may haveplayed a role.

Although analysis of past errors can pro-

vide insight into the projection process, suc-cess or failure in projecting population underone set of conditions does not necessarilyimply continued success or failure under adifferent set of conditions in the future. Inaddition, as would be expected, errors grow

with the duration of the projection. Thus theperformance of past projections a fewdecades into the future becomes less relevantas the projection horizon stretches to 100

years or more.

References1. John Bongaarts and Rodolfo A. Bulatao, eds.,

Beyond Six Billion: Forecasting the Worlds Population

(Washington, DC: National Academy Press,2000).

2. Martin P. Brockerhoff, Urban Growth in Devel-oping Countries: A Review of Projections andPredictions, Population and Development Review25, no. 4 (2000): 757-78.

Box 1

Accuracy of Population Projections


11/44

9

In more developed countries, cen-sus results tend to be complete andprovide a solid base for making pro-jections (see Box 1). Fertility andmortality rates are calculated fromrecorded birth and death statisticsand population estimates based oncensuses. This information allows

demographers to produce a relativelyconsistent picture of historical popu-lation change. Even in these coun-tries, however, internationalmigration statistics are incomplete,and net migration is often estimatedfrom the differences between birthand death rates and assumed popula-tion change.10

Estimating the base population andvital rates for less developed countriesis more difficult because demographicdata are incomplete and often inaccu-rate. Over the past 20 years, however,data collection efforts have increasedsubstantially around the world. Whenthe UN produced its 1998 revision ofWorld Population Prospects, 83 percentof all countries or areas had post-1985census data available on populationsize and age structure.

Vital rates for many less developedcountries are derived from surveys

and are less accurate than rates basedon the complete birth and deathrecords that are available in moredeveloped countries. Information onadult mortality is usually the leastcompletebirths and child deathsare more likely to be recorded. Coun-tries accounting for 40 percent ofglobal population in 1998 lacked anyrecent data on adult mortality, whichmakes it difficult to estimate baselinepopulation size and the age and sex

structure of the population as well asto estimate mortality trends.The UN Population Division pro-

duces the most widely used estimatesof population size, age structure, andvital statistics (birth and death rates).Obtaining and evaluating data makeup the bulk of the Population Divi-sions demographic work. UN demog-raphers use statistical techniques to,for example, make sure that estimatesof vital rates are consistent with esti-

mates of population size and age

structure. A history of high fertilityrates would be consistent with a youngage structure, while a history of lowfertility would be associated with anolder age profile.

The Census Bureau and WorldBank make their own estimates ofbaseline data. While the UN and both

of these other organizations rely onthe same data sources and use similartechniques for estimating demo-graphic variables, they may employdifferent assumptions about censusundercounts and vital rates, and theymay obtain and incorporate new datasources at different times. The CensusBureau might use a lower fertility ratefor Brazil than the UN, for example,because it adopted the results of anew demographic survey before UNdemographers had a chance to evalu-ate and incorporate the results.

In practice, these differences havebeen very small at the global level.Estimates of the 1990 world popula-tion from the Census Bureau and theUN 1998 series differed by less than0.1 percent. For individual countries,differences can be larger: In 11 coun-tries the differences in population sizeestimates were 10 percent or more.

In its most recent projections,IIASA used baseline data on popula-tion size, total fertility rates, and lifeexpectancies from the UN 1998 revi-sion and the U.S. Census Bureau.

UncertaintyProjections of the size and character-istics of a population at some futuredate are based on assumptions drawnfrom past trends and current theo-

ries. Because the future is unknown, aprojection based on the past is likelyto be wrongthe burning question is:by how much? This is a crucial ques-tion for those who use populationprojections, for example, to meetfuture educational, energy, or pen-sion needs. There is no generallyaccepted approach to characterizingthe uncertainty inherent in all popu-lation projections, but demographersare developing more sophisticated

ways to do this.

A projectionbased on the

past is likely to

be wrongthequestion is, byhow much?


12/44

Approaches to characterizinguncertainty can be grouped into twomain categories: scenarios, used inUN global projections and in manynational projections (see Box 2), andprobabilistic projections, used byIIASA (see Box 3, page 12).

Projecting FertilityFertility has the greatest effect onpopulation growth because of its mul-tiplier effect: Children born today willhave children in the future, and soon. The fertility component of popu-

Population projections according to alternative sce-narios, called variants in some cases, show what thefuture population would be if fertility, mortality, andmigration follows various paths. Some scenarios or

variants are purely hypotheticalsuch as the UnitedNations (UN) constant fertility variant, which pro-

jects world population assuming that fertility levelshold their same level. The UN demographers do notconsider this likely, but it illustrates what would hap-

pen if fertility does not decline at all. The worldpopulation would reach 53 billion by 2100, underthe UN constant fertility assumption, about six timeshigher than projected in the medium scenario.

Other scenarios offer users a choice of more plau-sible projections that they can employ in their own

analyses. Users of population projections sometimesrequire projections that conform to various story-lines. Population projections might form just part ofa scenario of future energy use and greenhouse gasemissions that presuppose particular socioeconomic,technological, or political developments.1

The scenario approach also has several weak-nesses. The most important is that users cannot inter-pret the probability that population will track a

higher or lower scenario. The only differencebetween the high and low scenarios in the UN long-range projections, for example, is the fertility rate(see figure). The UN assumed an average of 2.03children per woman after 2050 for its medium sce-nario, and assigned rates about one-half birth higher

and lower, respectively, for the high and lowscenarios. The UN provides little informa-tion about the likelihood of a particular sce-nario, except that it suggests that both thehigh and low scenarios are unsustainableover the very long run.2 These scenariosproduce a global population that doubles or

is halved every 77 years. Theoretically, theyand would eventually lead to extinction orto implausible crowding. The UN producesintermediate scenarios with more moderaterates of growth or decline and concludesthat future demographic rates will verylikely be bound by these (intermediate) sce-narios if sustainability is to be maintained.

References1. S.R. Gaffin, World Population Projections for

Greenhouse Gas Emission Scenarios, Mitiga-tion and Adaptation Strategies for Global Change3(1998): 133-70; Nebojsa Nakicenovic and RobSwart, eds.,Emissions Scenarios(Cambridge,England: Cambridge University Press, 2000);and G. Gallopin, A. Hammond, P. Raskin, andR. Swart, Branch Points: Global Scenarios andHuman Choice, PoleStar Series ReportNo. 7(Stockholm: Stockhom Environment Institute,1997).

2. United Nations, Long-Range World PopulationProjections: Based on the 1998 RevisionESA/P/

WP.153 (New York: United Nations, 1999): xiii.

Box 2

Using Scenarios to Show Uncertainty

10

UN World Population Projections, 20002100

Note: TFR (total fertility rate) is the average total number of children that would beborn to a woman given current birth rates. These TFRs for the world are derived fromthe values assumed for geographic regions.

The TFR values for the high-medium and low-medium scenarios are between the highand medium, and medium and low values, respectively. The constant fertility scenarioderives from holding constant the TFRs estimated for each region in 1995-2000.

Source: United Nations, Long-Range World Population Projections: Based on the1998 Revision(1999).

Constant fertilityHigh

High-medium

Low-medium

Low

Medium

0

5

10

15

20

2000 2025 2050 2075 2100

Assumed fertility rate (TFR) 20502100

Projection series TFR (Average children per woman)

High 2.51Medium 2.03

Low 1.56Constant 4.18-4.78

Population in billions


13/44

11

lation projections is summarized bythe total fertility rate (TFR), whichestimates the average total number ofchildren a woman will have assumingthat current age-specific birth ratesremain the same throughout herchildbearing years.

In general, the projection of the

TFR reflects an assumption that fertil-ity will eventually stabilize at a specificlevel in a country or region and theassumed path the TFR will follow tothat level. Once fertility reaches thislevel, assuming mortality and migra-tion rates remain the same, the popu-lation age structure will eventuallystabilize as well. The population sizewill change at constant rate. If there isno net migration (the number of in-migrants is cancelled out by the num-ber of out-migrants), and the TFRstabilizes at replacement level (a littlemore than two children per woman,the TFR at which the childbearinggeneration would have just enoughchildren to exactly replace itself), thegrowth rate will eventually be zero.Both the projected pace of fertilitydecline and the assumed eventual fer-tility level are important to determin-ing trends in population size and age

structure. The two factors also inter-act: The lower the assumed eventualfertility level, the more important thepace of fertility decline becomes toprojected population size.11

Demographic TransitionTheoryFor countries currently above replace-ment level fertility, demographic tran-sition theory provides the theoretical

basis for forecasting fertility trends.The concept of demographic transi-tion is a generalization of eventsobserved over the past two centuriesin the more developed countries.While different societies experiencedthe transition in different ways, ingeneral, these societies have graduallyshifted from small, slowly growingpopulations with high mortality andhigh fertility to larger, slowly growingpopulations with low mortality and

low fertility.12

During the transition

itself, population growth acceleratesbecause the decline in death ratesprecedes the decline in birth rates,creating a sudden surplus of birthsover deaths.

Evidence from all parts of theworld overwhelmingly confirms therelevance of the demographic transi-

tion to todays less developed coun-tries. The transition is well-advancedin all less developed countries, exceptin sub-Saharan Africa, where thebeginnings of a fertility decline arebecoming apparent.13 Fertility isalready below replacement level inseveral less developed countries,including China, Taiwan, and SouthKorea. In many other countries inSoutheast Asia and Latin America,fertility has fallen to levels seen in the

more developed world just a fewdecades ago.The biggest difference between the

transition in more developed coun-tries and less developed countries hasbeen the speed of the mortality andfertility decline. In Europe, NorthAmerica, and Japan, mortality fellslowly for two centuries as food supplystabilized, and housing, sanitation,and health care improved. In con-trast, mortality in most less developed

countries fell over the course of just a

Policies that enhance opportunities for women outside the home are assumedto also favor fertility decline and ultimately slow population growth.

Photo removed forcopyright reasons.


14/44

12

few decades after World War II asWestern medical and public healthtechnology and practice spread tothese regions. Populations are grow-ing much faster in less developedcountries than they did in moredeveloped countries at a comparablestage of the demographic transition.

Demographic transition theory hasbeen and continues to be a guidingprinciple in the study of fertility inless developed countries.14 Demogra-phers have developed many argu-ments about why fertility has declinedin the past and what might drive fur-ther declines in the future. While

One way to communicate the uncer-tainty in population projection resultsis to derive probability distributions forthe projected size and characteristics ofa population by using a range of differ-ent fertility, mortality, and migrationrates. There have been three mainbases for determining the probabilities

associated with vital rates: expert opin-ion, statistical analysis, and analysis oferrors in past projections.

Expert OpinionResearchers at the International Insti-tute for Applied Systems Analysis(IIASA) pioneered a methodology forassessing uncertainty in population pro-

jections based on asking a group ofexperts to give a likely range for futurefertility, mortality, and migration ratesthat is, the vital rates for a given date

would be within the specified range 90percent of the time, or have a 90-per-cent confidence interval.1

IIASA demographers argue that astrength of the method is that it maybe possible to capture socioeconomicchanges and unexpected events thatexperts might take into account butthat other approaches might missbecause they are guided by past events.In addition, this approach may be thebest way to estimate probabilities for

future demographic measures in geo-graphic areas where data on historicaltrends are sparse.

The expert opinion approach hasseveral drawbacksfor example, thetask of deciding who constitutes anexpert will always be problematic, andresearch has shown that experts tendto be too conservative in their expecta-tions for future changes, on average.Demographer Ronald Lee questions

whether experts can meaningfully dis-tinguish between different confidence

levels they may place on estimates offuture vital rates.2 He also argues thatthe method excludes the possibility offluctuations in vital rates that deviatefrom a general trend, which couldunderestimate uncertainty in out-comes. For example, the first proba-bilistic projections based on expert

opinion did not include any scenariosin which fertility starts out high, butends up low, nor any scenarios withbaby booms or busts.

Statistical MethodsStatistical analysis of historical timeseries data can be used either to projectpopulation size directly or to generateprobability distributions for populationsize or vital rates. Lee argues that, unlikemethods based on expert opinion, thesemethods are capable of producing inter-nally consistent probability distributions.

While statistical methods also employexpert judgment, they do not rely on itas much as the expert-based methodused in the IIASA projections.

Statistical analysis methods havebeen applied to some national projec-tions but not to global projections.3

They may be a source of further inno-vation in long-term global projections.

Historical Error Analysis

Population projections made in thepast can be evaluated for how well theyforecast the actual population, andthese errorsthe difference betweenthe projected and actual populationsizecan be used to calculate probabil-ity distributions for new projections. Arecent report by the U.S. NationalResearch Council (NRC) calculatedprobability distributions from theerrors of UN medium scenario projec-tions for 2000 that were made between1957 and 1998. The NRC found the

Box 3

Using Probabilities to Account for Uncertainty

Demographictransition theory

continues to

guide the studyof fertility.


15/44

13

each offers important insights, no sin-gle, simple theory explains the multi-faceted history of demographictransition around the world (see Box4, page 14). Each explanation suffersfrom its own shortcomings, and foreach, exceptions can be found. It isprobably best to think of fertility and

mortality transitions as being drivenby a combination of factors ratherthan a single cause, but determiningthe precise mix of factors at work in aparticular population at a given timeremains an elusive goal.15

The fact that the demographictransition has occurred under so

UN was somewhatmore likely to overes-timate than to under-estimate futurepopulation size at the

world level, althoughthe size of the error

was small. Errors were

much greater for pro-jections of countrypopulations, but theseerrors tended to can-cel out over the longterm at the nationallevel. The averageerror in UN projec-tions for individualcountries varied from4.8 percent for five-

year projections to 17percent error in 30-

year projections,according to the NRCreport. But the reportstates, a statistical review of past accu-racy is an imperfect guide to futureaccuracy.4

These three methods of producingprobabilistic projections are not mutu-ally exclusive. The most recent projec-tions from IIASA combine all threeelements: Expert opinion is used todefine a central path for fertility, mor-

tality, and migration in all worldregions. It is also used, in conjunctionwith historical errors, to define theuncertainty ranges for these values.Time series methods are used to gener-ate paths for each variable that canshow realistic fluctuations over time.

References1. Wolfgang Lutz, ed., The Future Population

of The World: What Can We Assume Today?

(London: Earthscan Publications Ltd.,1996); and Wolfgang Lutz, Warren

Sanderson, and Sergei Scherbov, Expert-Based Probabilistic Projections, inFron-tiers of Population Forecasting, ed. W. Lutz,

J.W. Vaupel, and D.A. Ahlburg, supple-ment to Population and Development Review24 (1998): 139-55.

2. Ronald D. Lee, Probabilistic Approachesto Population Forecasting, inFrontiers ofPopulation Forecasting, ed. W. Lutz, J.W.

Vaupel, and D.A. Ahlburg, supplement to

Population and Development Review24(1998): 156-90.

3. Ronald D. Lee and Shripad Tuljapurkar,Stochastic Population Projections for theUnited States: Beyond High, Medium andLow,Journal of the American StatisticalAssociation89, no. 428 (1994): 1175-89.

4. John Bongaarts and Rodolfo A. Bulatao,eds., Beyond Six Billion: Forecasting theWorlds Population(Washington, DC:National Academy Press, 2000): 51.

0

2

4

6

8

10

12

14

16Population in billions

2000 2025 2050 2075 2100

95%

60%

Probability thatactual population

will be inthis range

Source: W. Lutz, W. Sanderson, and S. Scherbov, Nature412 (Aug. 2,

2001): 543-46. Data provided by IIASA.

IIASA Projections of World Population,20002100


16/44

14

many different conditions and hasbeen driven by multiple causes com-plicates the study of demographic his-tory, but it also lends support to theidea that a transition to lower fertilityis inevitablewhich simplifies the

task of preparing population projec-tions. Presumably, demographersneed not focus on whether a coun-trys fertility will fall from very highlevels, but rather on when, how fast,and to what eventual level.

Box 4Explaining Fertility Decline

The earliest attempts to explain thedemographic transition cited industri-alization and urbanization as the ulti-mate driving forces.1According to thisclassical transition theory, economicmodernization leads to improvementsin health and nutrition that decreasemortality. Modernization also driveschanges in economic and social condi-tions that make children costly to raise

and reduce the benefits of large fami-lies. Eventually, this leads to lower fer-tility. Fertility decline lags mortalitydecline because cultural norms regard-ing reproduction are difficult tochange while improvements in mortal-ity meet little resistance.

The idea that reduced demand forchildren drives fertility decline gainedtheoretical rigor in the 1960s with thedevelopment of a theory based onchanges in determinants of parentsdemand for children. Economist GaryBecker and several collaborators pro-posed a microeconomic model thatdescribed choices parents are assumedto make between numbers of childrenand consumption of material goods atthe household level.2 The modelassumes that fertility falls because, aseconomic development proceeds, par-ents preferences shift toward higherquality children requiring greaterinvestments in education and health,

while increases in womens labor force

participation and wages increase theopportunity costs of raising children. Atthe same time, development leads to adecline in some of the economic bene-fits parents may derive from children,such as household labor, income, andold-age security. Thus, as the net cost ofchildren rises, demand falls.

This framework has been extendedand made more flexible by taking intoaccount sociological aspects. In the1970s, economist Richard Easterlinadded the influence of economic

development on environmental andcultural factors that affect natural fer-tility (what fertility would be in theabsence of regulation) and on the costs(including the psychological, social,and monetary costs) of fertility regula-tion.3 He proposed, for example, thatdevelopment may influence fecundity(the physiological ability to bear chil-dren) or taboos on intercourse while

mothers are breastfeeding, whichcould lead to an initial rise in fertilityas the demographic transition began.In contrast, effects of development onattitudes toward fertility regulation andthe time and money required to learnfamily planning techniques would tendto hasten the transition.

In the 1980s, researchers continuedto struggle to discern which social oreconomic factors are the most impor-tant causes of fertility change. Someexplanations have given much more

weight to sociological over economicfactors. Sociologist Norman Ryderargued, for example, that reproductivedecisions are not based strictly on arational weighting of the consequencesof childbearing, but are strongly influ-enced by cultural and normative con-texts.4Another sociologist anddemographer, Jack Caldwell, elabo-rated a theory that identified a shiftaway from extended family structurestoward the child-centered nuclear fam-

ily as the cause of a reversal in the flowof wealth (money, goods, and services)from children to parents typical in pre-transition societies to the flow of wealthfrom parents to children typical intransition societies.5As children dis-place parents as beneficiaries of thefamily, fertility falls.

The shift in family structure couldbe triggered by economic changes, butalso by the spread of new ideas. In arural agricultural village, for example,a child may provide benefits to the par-

Researchersstruggle to

discern which

social oreconomicfactors drive

fertility decline.


17/44

15

Policies and FertilityDeclineThe role of population policies in thedecline of fertility in less developedcountries over the past severaldecades, and by extension policys

potential role in determining futurefertility levels, is a matter of spiriteddebate. Family planning programshave been a primary policy tool in thepast;16 there are two main points ofview on their effectiveness.

ents through labor that outweigh thecost of having the child. The culturalnorms of the community may justifythis relationship, a situation that willtend to perpetuate high fertility. If cul-tural changes erode the social supportfor relying on children for labor, or ifeconomic development diminishes theimportance of labor-intensive agricul-ture, the benefits of children may no

longer outweigh their costs, removingthe obstacle to fertility decline.

Other researchers have emphasizedthe role of cultural over socioeconomicfactors. Based on analyses of the fertil-ity transition in Western Europe in the19th and early 20th century, demogra-pher Ron Lesthaeghe argued that dif-ferences in fertility across societiesarose largely from differences in reli-gious beliefs and the degree of secular-ism, materialism, and individualism.6

He proposed that cultural shifts lead-ing to greater individual control overlife goals and the means of achievingthem typically led to reduced fertility.Greater individualism was often associ-ated with a decline in religious beliefsand a growth in materialist values.

In work published in the late 1980s,demographers John Cleland and Chris

Wilson concluded that ideationalchange in general, and the spread ofnew ideas about the feasibility andacceptability of birth control in particu-

lar, was a key driver in fertility declineand likely more important than changesin economic conditions.7 More recently,demographers John Bongaarts andSusan Watkins demonstrated that diffu-sion of ideas and information about lim-iting fertility is important.8 They showedthat fertility transitions typically start inleader countries where development lev-els are relatively high, and then spreadto other countries in the region, oftenbefore the countries have achieved thesame level of development.

References1. Warren S. Thompson, Population, The

American Journal of Sociology34 (1929):959-75; and Frank W. Notestein, Popula-tion: The Long View, inFood for theWorld, ed. T.W. Schultz (Chicago: Univer-sity of Chicago Press, 1945): 36-69.

2. Gary S. Becker, An Economic Analysis ofFertility, inDemographic and EconomicChange in Developed Countries, Universities-

National Bureau Conference Series, no. 11(Princeton: Princeton University Press,1960); Gary S. Becker and H.G. Lewis,On the Interaction Between Quantityand Quality of Children,Journal of Politi-cal Economy82 (1973): 279-88; and Gary S.Becker and Robert Barro, A Reformula-tion of the Economic Theory of Fertility,Quarterly Journal of Economics103 (1988):1-25.

3. Richard A. Easterlin, Towards a Socio-Economic Theory of Fertility, inFertilityand Family Planning: A World View, ed. S.J.Behrman, L. Corsa, and R. Freedman(Ann Arbor, MI: University of MichiganPress, 1969): 127-56; and Richard A. East-erlin, An Economic Framework for Fer-tility Analysis, Studies in Family Planning6(1975): 54-63.

4. Norman B. Ryder, Fertility and FamilyStructure, Population Bulletin of the UnitedNations15 (1983): 15-33.

5. John C. Caldwell, Theory of Fertility Decline(London, Academic Press, 1982).

6. Ron Lesthaeghe, A Century of Demo-

graphic and Cultural Change in WesternEurope: An Exploration of UnderlyingDimensions, Population and DevelopmentReview9, no. 3 (1983): 411-35.

7. John Cleland and Chris Wilson, DemandTheories of the Fertility Transition: AnIconoclastic View, Population Studies41,no. 1 (1987): 5-30.

8. John Bongaarts and Susan C. Watkins,Social Interactions and ContemporaryFertility Transitions, Population and Devel-opment Review22, no. 4 (1996): 639-82.


18/44

16

Proponents argue that programshave had a substantial effect on fertil-ity primarily by reducing unwantedfertilitybirths that occur after awoman has had as many children asshe wants.17 The conventional justifi-cation for using family planning pro-grams to reduce unwanted fertility is

survey data that indicate that manywomen who want to avoid pregnancydo not use contraception. Familyplanning programs, therefore, helpmeet this unmet need for contra-ception by helping couples overcomeobstacles to contraceptive use. Obsta-cles can include limited access to fam-ily planning supplies and services,lack of knowledge about contracep-tives, fear of side effects from specificmethods, disapproval by relatives andothers, and the cost of obtaining con-traceptive supplies.18

In contrast, economist Lant Pritch-ett has argued that unmet need ismuch smaller than commonlyassumed, and that fertility decline isdriven primarily by a decline in thenumber of children women actuallywant (desired fertility) rather than areduction in unwanted fertility.19 Hisconclusion is based on the high corre-

lation between desired fertility andactual fertility (the number of childrenwomen want compared with the num-ber they have), and the lack of correla-tion between actual and unwantedfertility. Pritchett argues that becauselow-fertility countries have low desiredfertility, but do not have especially lowunwanted fertility, the fertility declinemust have been driven by reductionsin desired fertility, not by reducedunwanted childbearing. He also argues

that family planning programs havehad little effect on fertility.Demographer John Bongaarts

concludes that neither view is fullyaccurate.20 He agrees that there is sub-stantial unmet need for contracep-tion, but posits that the unmet need isless important to fertility decline thanmany family planning advocates esti-mate. Family planning advocates tendto include women who want to usefamily planning to delay rather than

prevent their next pregnancy in their

estimates of unmet need. Meeting thefamily planning needs of these womenwill not reduce overall fertility asmuch as family planning aimed atwomen who want to avoid any morepregnancies. Bongaarts concludes,however, that family planning pro-grams historically have had a substan-

tial effect on fertility. He attributes anestimated 43 percent of the fertilitydecline between the early 1960s andlate 1980s to program interventions.

Future change in fertility may alsobe affected by public policies thataddress such social and economic fac-tors as womens status, educationaland employment opportunities, andpublic health. Such policies arereceiving increased attention interna-tionally. At the 1994 InternationalConference on Population and Devel-opment (ICPD) in Cairo, 179 coun-tries agreed to a Program of Actionthat marked a fundamental shift inpopulation-related policies away fromdemographic targets and toward anew focus on individual well-being.The Cairo program set a number ofgoals for 2015 that reflected this per-spective. Among the goals were uni-versal access to comprehensive

reproductive health services (includ-ing, but not limited to, family plan-ning); reductions in infant, child, andmaternal mortality; and universalaccess to primary education, with anemphasis on closing the gender gapin education, health, and politicalparticipation.21Although these goalsare not primarily motivated by theirpotential effect on demographictrends, achieving them would likelylead to lower fertility (and lower mor-

tality). Bongaarts estimated, for exam-ple, that eliminating unwantedfertility in less developed countrieswould reduce population in 2100 byabout 2 billion, and that loweringdesired family size in these countrieswould reduce the projected popula-tion by an additional billion.22

In the global projections discussedhere, population policy efforts andeffectiveness are implicitly accountedfor because they are assumed to speed

fertility decline, but population poli-


19/44

17

cies do not explicitly enter the projec-tion process. Measuring the influenceof family planning programs on fertil-ity is difficult, although analysts havequantified program effort and effec-tiveness. Demographers ParkerMauldin and John Ross, for example,took program effort into consideration

in their short-term projections for 37less developed countries, but only inestablishing uncertainty, not in the fer-tility projections themselves.23 Measur-ing the effect of policies that enhancewomens status or promote economicdevelopment on fertility decline iseven more problematic.

Eventual FertilityDemographic transition theory pro-vides the basis for the expectationthat todays high fertility countrieswill experience, or continue to expe-rience, fertility declines. The theoryprovides little guidance, however, onthe long-term average fertility levelthese countries might eventuallyreach. It also has little to offer demog-raphers grappling with the questionof future fertility trends in countriesthat have already completed the tran-

sition to low fertility.Traditionally, many demographersassumed that fertility in all countrieswould eventually stabilize at replace-ment level, leading to stabilization ofpopulation growth. Long-term popula-tion projections reflected this thinkingby setting replacement levelabouttwo children per womanas the levelat which each countrys TFR wouldstabilize. Technically, replacementlevel is reached when each couple has

a daughter who survives to childbear-ing age to have her own children.Because some daughters will diebefore having childrenand becauseslightly less than one-half of all birthsare femalesthe TFR must be justabove 2.0 to maintain replacementlevel. A replacement-level TFR isslightly less than 2.1 children perwoman in more developed countrieswhere mortality rates are low, but is ashigh as 2.6 in Africa and 2.4 in South

Asia where mortality is higher.24

There are two general argumentsin favor of the assumption that fertil-ity will stabilize at replacement levelin the long term. First, replacement-level fertility is a convenient mathe-matical benchmark for demographerspreparing population projectionsalthough it may not be the most

likely outcome. Second, replace-ment-level fertility has been sup-ported by a view that holds thatdemographic rates of a populationare not just the sum of individualbehavior, but also reflect the tendencyof the demographic system to main-tain itself.25 The demographic systemoperates through the interplay of thevital rates and the population agestructure and is assumed to seekhomeostasis, under which birth rateswould equal death rates, and thepopulation would neither grow nordecline. This view interprets thefalling mortality rates that mark theonset of the demographic transitionas a perturbation of a system in bal-ance; birth rates fall as the systeminevitably re-establishes the balancebetween the two rates, and fertilityseeks replacement level.

The idea that low TFRs will eventu-

ally rise to replacement level and sta-bilize has been strongly criticized asassigning a magnetic force to replace-ment level fertility, without anyempirical evidence that TFRs will nat-urally drift to that level.26 Total fertilityhas been below replacement level in20 European countries for at least twodecades, and it is currently below 1.5children per woman in 21 Europeancountries.27 In eastern Germany,northern Italy, and the most urban-

ized regions of the Russian Federa-tion, fertility has been at or below onechild per woman.28 Fertility has alsofallen below replacement level inChina, Thailand, and North andSouth Korea, and several other lessdeveloped countries. By 1995, 45 per-cent of the world lived in countrieswith below-replacement fertility.

There are many arguments thatsupport the idea that fertility willdecline below replacement level in

more populations. These arguments

Traditionally,demographershave assumed

that populationswould eventuallystabilize.


20/44

18

can be grouped under the term indi-viduation, which encompasses theweakening of family ties, character-ized by declining marriage rates and

high divorce rates, the increasingindependence and career orientationof women, and value shifts towardmaterialism and consumerism.29 Indi-viduation, together with increasingdemands and personal expectationsfor the amount of attention, time,and money devoted to children, islikely to result in fewer couples thathave more than one or two childrenand an increasing number of childlesswomen. Demographer Antonio Golini

has speculated that there might be anabsolute lower limit of about 0.7 to0.8 children per woman based on theassumption that between 20 percentand 30 percent of women remainchildless and the rest have just onechild. In principle, this would leaveroom for considerable furtherdecline, but it remains unclearwhether such a limit will be relevantfor national fertility trends.30

While current trends and some

plausible explanations may suggest

that low fertility will continue, there isno compelling theory that can predictreproductive behavior in low-fertilitysocieties. Although fertility typicallycontinues to fall after reachingreplacement level, there is no clearpattern to subsequent fertility trends.In some countries, fertility falls

quickly to very low levels, while in oth-ers it has followed a more gradualslide. In the United States, Sweden,and some other countries, fertilitydeclined well below replacement leveland then rose nearly to replacementlevel again.

One argument against assumingthat total fertility will remain very lowin these countries is that the TFR isaffected by changes in the timing ofbirths even if the actual number ofbirths women have over their lifetimedoes not change. Since the mean ageof childbearing has been increasingin many industrialized countries overthe past several decades, part of thedecline in TFR has been due to thistiming effect and not to a change inthe completed fertility of women.Demographers John Bongaarts andGriffith Feeney argue that the TFR islikely to increase in the future once

the mean age of childbearing stopsrising, as happened in the 1980s inthe United States when fertility roseto its current, near-replacementlevel.31An additional argumentagainst continued very low fertility isthat in surveys conducted in much ofEurope, women consistently say theywant about two children.32 There aremany reasons why women may fail toreach this target (career plans,divorce, or infertility, for example),

but this finding suggests that fertilitymay be unlikely to remain extremelylow, especially if societies make it eas-ier for women to combine careersand childbearing.

Even if Europes low fertility levelsmask a pent-up demand for morechildren, however, the TFR in Euro-pean countries may not rise toreplacement level unless the youngerwomen who are currently postponingbirths recuperate much of this

delayed fertility at older ages.33

This

Number of children

2.5

2.0

1.8

1.6

2.6

2.2

2.1

2.3

2.1

1.8

1.5

2.4

2.2

2.01.9

Austria France Italy United Kingdom

Woman's year of birth 1935 1945 1955 1965

Figure 3Completed Fertility for European Women, SelectedCountries and Birth Cohorts

Note: Completed fertility for each birth cohort refers to the average number of children women hadby age 45.

Source: Council of Europe, Recent Demographic Developments in Europe 1999(1999): 78.


21/44

19

would mean a reversal of recenttrends in cohort fertilitythe totalnumber of children produced bywomen from a given birth cohort.Cohort fertility was already belowreplacement level in most Europeancountries for women born between1945 and 1965 (see Figure 3).34

Feedbacks: EnvironmentalChange and FertilityIt is well known that changes in fertil-ity, through fertilitys effects on popu-lation size, growth rate, and structure,can influence environmental condi-tions. Changes in the environmentcan, in turn, affect fertility. If suchfeedback loops are strong enough, itwould be important to consider themwhen projecting future populationgrowth. Historically, environmentalchange has affected fertility mainlythough its impact on agriculture andfood supply. In 16th- and 17th-centuryEngland, for example, a prolongedcool period was associated with adecline in grain yield, fertility, lifeexpectancy, and population growth,while the average age at marriage andnet out-migration increased.35 Simi-

larly, in China and Western Europe,periods of warmer temperaturesbetween the 13th and 19th centurieshave been linked to simultaneousincreases in population growth rates.

Links between environmentalchange, agriculture, and fertility canbe mediated by a number of factors.When facing an extended drought,for example, men in an agriculturalcommunity may leave their wives andcommunity to seek work in adjoining

agricultural regions or in cities.36

Couples may delay marriage becausethey lack financial assets or housing.Fears of inadequate food supply mayalso induce changes in attitudes. Thedeterioration of natural resources inEthiopia since the 1980s may havepushed womens preferences towardlater marriage and smaller family sizesas well as encouraged greater use offamily planning services.37

In each of these cases, however,

environmental factors had relatively

modest affects on fertility levels.Moreover, studies of historical peri-ods are not always relevant to con-temporary conditions when manyeconomic, social, and technologicalfactors have changed. For these rea-sons, long-term projections do notexplicitly take into account environ-

mental feedbacks on fertility.38

Future Fertility LevelsWhat do the historical records andcurrent theories suggest about fertilitytrends in the future? Demographersat the major projection institutionshave slightly different interpretations,which yield slightly different results.The differences are greater for spe-cific countries and small regions thanfor the world as a whole.

All major global population projec-tion series assume that the transitionfrom higher to lower fertility will con-tinue throughout the world. Projec-tions vary in the pace of decline andin the ultimate fertility level.

The UN has historically assumedin its medium scenarios that fertilitywould level off at replacement level.In countries that had already dipped

below that level, the UN invariablyforecast TFRs to rise back up to about2.1 children per woman. But as TFRsfell below replacement in more andmore countriesincluding China,North and South Korea, and Thai-landand sank to previously unimag-inable low levels in Germany, Italy,Spain, and other more developedcountries, the UN and other groupschanged their strategy. In their 1998and 2000 revisions ofWorld Population

Prospects, the UN assumed that coun-tries in which TFR is already belowreplacement level would remainbelow replacement level until 2050.39

For the long-term projections,however, the medium-scenario fertilityin the low-fertility countries isassumed to rise to replacement levelbetween 2050 and 2075, dependingon the region. The UN appears tohave assumed replacement level TFRin the long run to establish for

Europe a benchmark scenario in


22/44

20

which population ultimately stabilizes,not because it is judged to be themost likely scenario. The projectionis described as representing a con-ceptual dividing line between long-range future population increase andlong-range population decline.40

Projections prepared by the U.S.

Census Bureau carry the assumptionthat eventual fertility will be belowreplacement level in a number ofcountries. IIASA adopted this assump-tion for all regions of the world in itscentral scenario.

In countries with fertility above 2.1children per woman in 1990-1995, theUN maintains its historical assump-tion that fertility will undergo asmooth decline to replacement leveland remain constant thereafter. Thedate that a countrys fertility reachesreplacement level is chosen basedmainly on the current level of andrecent trends in fertility and on com-parisons with similar countries. TheU.S. Census Bureau also assumes thatfertility will eventually level off atabout two births per woman, whileIIASA assumes that, in the long run,fertility will decline below replace-ment level.

Projecting MortalityMortality projections are based onprojecting life expectancy at birththat is, the average number of years achild born in a given year can expectto live if current age-specific mortalitylevels continued in the future. Lifeexpectancy (like the total fertilityrate) measures the situation at a given

period of time; it does not reflect theactual experience of an individual.Nonetheless, life expectancy providesa useful summary of the mortalityrates for each age and sex group in apopulation at a particular time.

Projections of mortality must spec-ify how the distribution of mortalityover different age and sex groups maychange over time. Changes in mortal-ity at different ages have differentconsequences for population growth

and age structure. When child and

infant mortality decline, for example,a greater proportion of babies willsurvive to adulthood to have theirown children and contribute tofuture growth. Mortality declinesamong the older population have amore short-term effect on populationgrowth because the survivors are

already past reproductive age.

Conceptual Basis forProjectionsUncertainties about future changes inlife expectancy are quite different inhigh- and low-mortality countries.Low-mortality countries, primarily inthe more developed regions, haveseen their life expectancies increaseto levels once considered a biologicalupper limit to the human life span.Future improvements depend mainlyon whether or not such a limit existsand, if it does exist, how soon it mightbe reached. In less developed coun-tries where mortality remains high,future life expectancy will be deter-mined by the efficiency of localhealth services, the spread of tradi-tional diseases such as malaria andtuberculosis, and new diseases such as

HIV/AIDS, as well as living standardsand educational levels. The gap in lifeexpectancy between more developedcountries and less developed coun-tries has narrowed over the past 50years, and is likely to narrow furtherunless the AIDS epidemic stallsprogress in a significant number ofless developed countries.

In more developed countries,mortality is concentrated at old ages,so uncertainty about future life

expectancy is based on uncertaintyabout future death rates among theelderly. Death rates have been declin-ing steadily for this age group, butthere is a range of opinions on howlong this trend can continue.

One point of view is that lifeexpectancy in more developed coun-tries is unlikely to increase wellbeyond 85 years from its current levelof about 75 years. Some argue thatthis age represents an intrinsic

(genetically determined) limit to the

When infant andchild mortality

decline, more

babies survive toadulthood tohave their own

children.


23/44

21

human life span.41 Improvements inmortality that do occur are likely toincrease an individuals chances ofsurviving to the maximum life span,but not to extend the maximumitself. Other researchers argue thatwhile the intrinsic limit may be modi-fiable, in practical terms it is unlikely

to be exceeded without medicalbreakthroughs.42 This view is basedon calculations showing that increas-ing life expectancy to 85 years wouldrequire dramatic reductions in mor-tality rates, particularly among theelderly. Following this line of reason-ing, complete elimination of deathsfrom diseases such as heart disease,cancer, and diabeteswhich accountfor a large proportion of deathsamong the elderlywould notextend average life expectancybeyond 90 years. Only breakthroughsin controlling the fundamental rateof aging could achieve substantiallylonger life expectancies.

Other researchers hold thatreduced mortality among the oldestages could produce substantialimprovements in life expectancy. Datafrom several more developed coun-tries show that death rates at old ages

have been falling over the past severaldecades, and this improvement hasbeen accelerating, not decelerating aswould be expected if a limit werebeing approached.43 In attempting tounderstand this trend, researchers areinvestigating the evolutionary basisfor aging.44 Evolutionary biologistsand biodemographers theorize thatsenescencethe degeneration of cel-lular processes over timeis an inad-vertent consequence of sexual

reproduction. Genes responsible forlethal diseases that usually affect peo-ple when they are past childbearingage tend to evade the influence ofnatural selection because, unlikegenes associated with diseases earlierin life, these genes are passed onbefore they are expressed.45 Thusmortality rates inevitably rise after thereproductive period. Intriguingly,increases in mortality decelerate atolder ages, not only in humans but in

several other species as well.46

No sin-

gle evolutionary theory satisfactorilyexplains this empirical finding.

The likelihood that biological orpractical obstacles to overcoming thisgenetic legacy will be surmounted inthe foreseeable future remains anopen question. If they are, a signifi-cant increase in life expectancy could

have a large impact on projectedpopulation. In a hypothetical case, iflife expectancy were to increase to150 years over the next two centuries,global population would stabilize at alevel twice as high as it would if lifeexpectancy did not exceed 85 years.47

In most less developed countries,possible limits to the life span are notas relevant to projections because lifeexpectancies are lower and mortalityis not as concentrated at the oldestages. Life expectancy in less devel-oped countries increased from about40 in the 1950s to just over 60 in thelate 1990s, a remarkable achievementdriven mainly by reductions in mortal-ity from communicable diseases.Regional progress was variable, withthe slowest gains in sub-SaharanAfrica, where average life expectancyis just over 50, and the fastest inChina, where life expectancy reached

68 in the 1990s. Projecting mortalityin less developed countries is difficultbecause of the relative scarcity andpoor quality of data on current andpast trends. In addition, the futurecourse of the HIV/AIDS epidemiccould substantially affect mortalityin many countries, especially in sub-Saharan Africa where HIV prevalencerates are especially high.

Effects of HIV/AIDSHIV/AIDS brings premature death tothe most economically active and pro-ductive population groups, and isimposing an enormous economic andsocial toll on the African continent. Inaddition, HIV/AIDS has slowed, andin some cases reversed, the impressivegains in life expectancy in the lessdeveloped countries over the past sev-eral decades. Sub-Saharan Africa hasbeen most affected. In Botswana, for

example, life expectancy has dropped


24/44

22

from about 63 years in the late 1980sto 44 years in the late 1990s. Zim-babwe has seen life expectancy fallfrom 57 to 43 years over the sameperiod, according to the UN.

The effect of HIV/AIDS on popu-lation growth and age structure is sig-nificant in countries with the highest

prevalence. The UN estimates in its2000 projection series that for the35 African countries in which itadjusts its projections to account forHIV/AIDS, population size will be onaverage 10 percent lower in 2015than it would be without any deathsfrom AIDS. In the nine most affectedcountries, AIDS mortality lowers theprojected 2015 population by nearly18 percent. An independent studyshows that population size may welldecline in Botswana where the HIVprevalence rate is estimated at morethan 30 percent of adults.48 More-over, the age structure will becomeseverely distorted by AIDS deaths,which will have long-term effects onpopulation growth. AIDS orphans,rising health expenditures, and aworsening health status of the laborforce are likely to present majormacroeconomic problems in addition

to immense human suffering.The ultimate impact of HIV/AIDSon the population of Africa as awhole will be moderate if exception-ally high HIV prevalence rates arelimited to Botswana and a few othercountries in South and East Africa. Ifprevalence rates increase in othersub-Saharan regions, HIV/AIDS willhave a significant impact on popula-tion dynamics of the entire continent.HIV/AIDS could also affect popula-

tion growth in Asia and Latin Amer-ica and other world regions wherethe virus has spread.

Environmental Feedbacksand HealthEnvironmental change has hadimportant direct and indirect effectson mortality in the past. Climatechange, for example, probably con-tributed to the collapse of the Classic

Maya culture in the Yucatan in A.D.

800-1000 and the decline of theEaster Island civilization in the 18thand 19th centuries.49When thesemassive disruptions occurred, how-ever, the populations had extremelylimited technical capacity to respondto change; the relevance of theseancient occurrences to future envi-

ronmental change is unclear.The most frequently discussed

possibilities for future effects centeron the idea of carrying capacity (themaximum number of people that theEarth can support) and the potentialhealth impacts of climate change.Currently, however, population pro-jections do not take explicit accountof possible environmental feedbackson mortality, based on the belief thatthey are unlikely to be an importantdeterminant of future mortalitytrends.50

The concept of carrying capacityhas its roots in ecology and the popu-lation biology of nonhuman species.Simple models of population growththat assume a limit to population sizegive rise to a logisticor S-shapedgrowth pattern, in which populationsize increases quickly at first, thenmore slowly as it approaches its ulti-

mate limit. There is a long history ofestimates of the Earths human carry-ing capacity, based mainly on the ideathat a growing population will eventu-ally trigger an increase in death ratesas it pushes up against the limit ofthe planet to provide the resourcesnecessary to support life. Proposedlimits have been based on a widerange of factors, including suppliesof energy, food, water, and mineralresources, as well as disease and

biological diversity. No consensus onthe human carrying capacity hasemerged; on the contrary, the rangeof estimates has widened over time.51

Carrying capacity is not consideredin long-term population projectionsfor at least three reasons. First, thereis no agreement on what the limitingfactors to population growth mightbe. Any proposed limit relevant toprojections over the next century ortwo would depend primarily on

which factor or factors were assumed


25/44

23

to be limiting, as well as on howthinly any one factor had to be spreadto begin to exert its limiting influ-ence. While food is often taken as alimiting factor, for example, the maxi-mum population that could be fedwould depend on, among otherthings, the typical diet, agricultural

productivity (which would depend ontechnology, agricultural research, irri-gation, and other factors), the allow-able fraction of land usable foragriculture, and so on. In addition, afactor that may be scarce in oneregion may be available in excess inanother and, therefore, inter-regionaltrade might overcome limits in partic-ular areas.

Second, even if the relevant factorscould be agreed on, it may be too dif-ficult to project the future evolutionof those factors for use in populationprojections.52 Future agricultural sys-tems, energy supplies, and water avail-ability are difficult to foresee in theirown right, and there is no consensusin these areas to which demographersmight turn. Third, even if these fac-tors could be reliably predicted, theireffects are mediated through eco-nomic, political, and cultural systems

in ways that are not possible to quan-tify with confidence.Although no long-term projections

routinely take carrying capacity intoaccount, some researchers haveargued that it may be worth consider-ing these limiting factors. Limitingfactors might be especially relevantwhen projecting populations overlong time horizons, in particular loca-tions where resources are especiallylimiting and potential for trade is low,

or when analyzing the relationshipbetween demographic factors andspecific environmental constraints.53

IIASA demographer Wolfgang Lutzand colleagues examined the potentialdemographic consequences of anassumed carrying capacity of 2.5 bil-lion for sub-Saharan Africa to illustratehow such an exercise might be carriedout. They demonstrated that if war,famine, disease, or some other catas-trophe increased mortality by 20 per-

cent but left fertility unchanged at a

high level, the population of theregion would regain its 20 percent losswithin 10 to 15 years. The rate of

demographic recovery depends on theage and sex structure of the mortalityreduction as well as on assumptionsregarding fertility, so that incorporat-ing carrying capacities into populationprojections requires a fairly detailedaccounting of the effects of a catastro-phe on demographic variables.

In addition, some projections takeenvironmental feedbacks into accountindirectly. The IIASA methodology fordeveloping probabilistic projections

implicitly includes a small possibilityof a substantial increase in future mor-tality, allowing for the possibility ofnegative feedbacks from environmen-tal changes such as global warming.

Environmental effects on mortalityshort of a large-scale catastrophe havereceived increasing attention recently,especially those that might be drivenby future climate change. Climatechange could cause infectious dis-eases to spread to new populations.54

An increase in such severe weather as

Concern about increasing HIV/AI

World Population Futures (Sept 2001)

Documents