7.2 Demography and public health

7.2 Demography and public health
Oxford Textbook of Public Health

Demography and public health

Emily Grundy

Global issues
Demographic data and methods of analysis

Population censuses

Vital registration

Cause of death

Other data sources
Analysis of demographic data

Measurement of fertility

Measurement of mortality

Measurement of migration
Population projections
Population dynamics
Population growth

Intrinsic rate of natural increase: stable population theory
Age structure
The demographic transition
Proximate determinants of fertility
The epidemiological transition
Gender differentials
Recent trends in mortality

Disability-free or healthy life expectancy

Implications of recent demographic trends
Chapter References

The health and health-care needs of a population cannot be measured or met without a knowledge of its size and characteristics. Demography is concerned with this essential ‘numbering of the people’ and with understanding population dynamics—how populations change in response to the interplay between fertility, mortality, and migration. This understanding is a prerequisite for making the forecasts about future population size and structure which should underpin health-care planning. Analysis of both the present and the future necessitates a review of the past. The number of very old people in a population, for example, depends on the number of births eight or nine decades earlier and risks of death at successive ages throughout the intervening period. The proportion of very old people depends partly on this numerator but more importantly on the denominator (the size of the population as a whole)—itself a function of reproductive behaviour, mortality, and net migration from yesterday back through time. The number of births in a population depends not just on current patterns of family building, but also on the number of women ‘at risk’ of reproduction—itself a function of past trends in fertility and mortality. Similarly, the number of deaths (and their distribution by cause) is strongly influenced by age structure. For this reason, although life expectancy at birth in the developed world is some 13 years longer than in the less developed world (78 and 65 years respectively), crude death rates—deaths per 1000 population of all ages—are very similar (eight and nine) (World Bank 1999).
Formal or pure demography is largely concerned with answering questions about how populations change and how these changes can be measured. The broader field of population studies embraces the questions of why these changes occur, and with what consequences.
A major theme of this work is the complex interrelationship between population change and human health. This chapter presents information on demographic methods and data sources in the context of their application to health and population issues.
Global issues
There are currently very substantial differences between regions of the world in population characteristics and trends and predominant public health issues. As illustrated in Fig. 1, the size of the world’s population is growing at an unprecedented rate and was estimated to comprise some 6 billion people in October 1999 (UNFPA 1999). While it took an estimated 123 years (from 1804 to 1927) for the world to increase its population from 1 to 2 billion, the increase from 5 to 6 billion was achieved in a tenth of the time (1987 to 1999) (UN 1999). Most of this growth has been in the developing world where currently about 98 per cent of world population increase takes place (Population Reference Bureau 1999; UNFPA 1999). In many countries of sub-Saharan Africa, 40 per cent or more of the population is aged 15 or under and the total population is expected to have more than doubled in size between 1985 and 2015 (Bulatao and Stephens 1992). By contrast, many European countries are expected to experience real falls in population size between 1998 and 2050, by which latter date approaching a quarter of their populations will be aged 65 or more (US Bureau of the Census 1992; UNFPA 1999). These differences in age structure, and associated variations in levels of fertility and mortality, are illustrated for regions of the world in Table 1. Clearly these variations have enormous implications for the health and health-care priorities of the populations concerned. In the developing countries of the world a third of all deaths occur among infants and children aged under 5 years; deaths of those aged 65 or more account for a slightly lower proportion. In the developed world, by contrast, deaths of elderly people aged 65 or more account for 72 per cent of the total and those of children under 5 less than 3 per cent (Fig. 2). As a result, as shown in Fig. 3, only 2 per cent of the world’s deaths among those under 5 years of age occur in the populations of developed regions, compared with 42 per cent of the world’s deaths in elderly age groups.

Fig. 1 World population data and projections: 1950 to 2025. (Data from WHO 1992a; UN 1999.)

Table 1 Age structure, fertility, and life expectancy: world regions and selected countries, 1998 or 1999

Fig. 2 Distribution of deaths by age group in less developed and more developed regions, 1990. (Data from WHO 1992a.)

Fig. 3 Distribution of deaths (%) between more and less developed regions by age, 1990. (Data source: WHO 1992a.)

Closely related to these age variations (and differences in the level of mortality) are differences in the cause structure of death (Preston 1976a). Communicable diseases, maternal and perinatal conditions, and nutritional deficiencies account for 35 per cent of all deaths in low- and middle-income countries compared with only 6 per cent of deaths in high-income states. Conversely, non-communicable diseases are responsible for 54 per cent of deaths in low- and middle-income countries but 87 per cent in high-income ones (WHO 1999). While in parts of the world reproductive health, including family planning and measures to reduce the spread of HIV/AIDS, present the most pressing public health problem, in many developed countries concerns about the growing proportion of old—particularly very old—people and the possible implications of recent changes in family systems predominate.
The watershed that separates populations with high fertility, relatively high mortality, young age structures, and rapid growth from those with low vital rates, older age structures, and slow or no growth is conceptualized as the demographic transition. Identifying and explaining this and the associated profound changes, termed the epidemiological and health transitions, has been described as the central preoccupation of modern demography (Demeny 1972). Before turning to the causes, progress, and consequences of these transformations, the basic methods and materials of demographic analysis must be considered and the issue of population dynamics—how populations change—must be addressed.
Demographic data and methods of analysis
In the seventeenth century John Graunt, a London merchant, used data from the London Bills of Mortality to devise an early life table. This pioneering work has led to him being dubbed the ‘father of modern demography’ (Smith and Kayfitz 1977). However, Graunt was seriously handicapped by the fact that, although he had information on numbers of deaths, he lacked data on the population at risk and could not compute death rates. Essentially all demographic analysis requires data both on the population ‘stock’ and on ‘flows’ in and out—births, deaths, and migration. The traditional sources of information on the former are population censuses and, for the latter, vital registration systems.
Population censuses
Head counts of the population, generally for military or tax purposes, have an ancient history but the first ‘modern’ censuses were undertaken in Scandinavia in the eighteenth century. In England and Wales the first census was undertaken in 1801, although a question on age was not included until 1841. During the nineteenth century censuses spread throughout Europe and are now almost universal. As well as basic questions about age, sex, marital status, and place of residence, data on other characteristics such as employment, education, and housing are often collected. The United Nations, which has done much to ensure that a minimum of roughly comparable questions are included in as many countries as possible, recommends that censuses be conducted at least decennially in years ending in 0 or 1. Recommendations about particular topics are also sometimes made; in the 1990–1991 round, for example, disability was emphasized and questions on this were included in a number of countries, including the United Kingdom.
Censuses have many strengths and are generally the only source of data for small areas or small population subgroups. Although primarily a tool for collecting data on population ‘stock’, censuses have also been used as ways of finding out about vital events, which is particularly useful if other data sources are sparse. In the 1911 and 1971 British censuses, for example, detailed fertility histories were collected from married women. Many countries use censuses to provide data on recent internal migration (through questions on place of residence one or more years earlier) and on immigration (through questions on country of birth and/or date of entry for those born elsewhere). Indirect estimation techniques developed by Brass (1975) and others mean that data from simple questions on number of children who have died, widowhood, and orphanhood included in some developing countries can be used to assess mortality levels (UN 1983). Taking a census requires not only a reasonable administrative infrastructure, but also the co-operation of the population to be enumerated. Some countries, including Germany and The Netherlands, have given up taking censuses because the latter is lacking.
Against the strengths of censuses must be set the huge costs of collecting and processing census data and the major problems involved in ensuring it is of reasonable quality. When censuses are taken, difficulties arising from errors and omissions are common, even in developed countries with a long history of census taking.
Young geographically mobile adults, members of minority ethnic groups, infants, and the very old are the groups most likely to be under-enumerated. In the 1991 British census, for example, an estimated 1.2 million people were missed, including 10 per cent of males in their twenties and 8 per cent of those aged 85 or more (Heady et al. 1994). Groups such as seasonal migrants (including students), naval and military personnel, and people temporarily away from home present particular problems. Not only are they more likely to be missed, but a decision has to be made about whether they should be assigned to their place of usual or legal residence (assuming it can be determined), or counted as belonging to the place of enumeration. The former system is termed de jure, the latter de facto. The issue of assigning people to some place of usual residence is important as often resources are distributed on the basis of population size and characteristics. Moreover, it is essential to try and ensure that demographic events recorded in one system (vital registration) are attributed to the population actually ‘at risk’ of experiencing them. In the developed world, for example, most deaths occur in hospitals which may draw patients from a wide area. If no attempt is made to assign these decedents to the region or locality where they lived prior to hospital admission, areas including large hospitals will appear to have very high mortality rates while in others recorded mortality will be artificially low.
Assessment of the extent of under-enumeration is usually achieved through census validation surveys (surveys of a sample of census addresses in which intensive efforts are made to contact non-respondents and to check information supplied by respondents) and comparisons with population estimates from other data sources. Ensuring near-complete enumeration is only part of the problem; the quality of the data collected is also a major concern.
Age misreporting is one of the most serious problems that must be estimated and allowed for in analyses of census and similar data. In many populations, people may not always know their exact age and some approximation is reported or made by an enumerator. ‘Heaping’ on ages ending in 0 or 5 is a common result. Heaping can be detected by looking at the age distribution and applying various tests of consistency and such data are normally adjusted before publication. More serious problems arise when reported age is based on other characteristics, such as marital status, number of children, or grandparent status, as clearly any analysis of, for example, age at first marriage will be biased.
Overstatement of age by elderly people (particularly older men) is common. Investigations into the causes of reported ‘super-longevity’ in particular populations commonly show that age mis-statement, rather than eating yoghurt for instance, lie behind them (Garson 1991). Age overstatement is not confined to developing countries. Investigations into the numbers of centenarians in the 1971 and 1981 Censuses in England and Wales showed serious overstatement (Thatcher 1992). In the United States, large discrepancies exist between the number of very elderly people in the census and the number estimated from other sources, such as Medicare records (Kestenbaum 1992).
Other characteristics may be ‘mis-stated’ because individuals’ perceptions of their status do not match official classification systems. Thus in England and Wales, it is clear from linked census data that quite high proportions of divorced men revert to describing themselves as single (never married).
These problems are serious enough in themselves; differential age misreporting between census and other sources, such as death certificates, present a further difficulty. Numerator–denominator discrepancies may introduce serious bias into the analysis of mortality at advanced ages, or by characteristics such as occupationally defined social class or marital status (Fox and Goldblatt 1982).
Vital registration
Data on demographic events are needed, as well as data on population characteristics. In the developed world these data are usually drawn from vital registration systems—the compulsory registration of births, marriages, and deaths. In some countries, principally Scandinavian, which maintain population registers, details of geographical moves are also recorded.
Compulsory registration of births and deaths was established in most European countries during the nineteenth century. In England and Wales, for example, civil registration was introduced in 1837. Subsequent improvements to the system included those following the 1874 Births and Deaths Registration Act which made parents legally responsible for registering births and required attending doctors to supply information on cause of death. Other revisions have since been made, for example the inclusion of firstly the mother’s and later the father’s age on the confidential section of birth certificates. Most developed countries now have well-established registration systems with complete, or very near complete, coverage. In the developing world, however, many people have no need for certificates of birth or marriage. Consequently, vital registration systems are frequently seriously incomplete or non-existent, although there are some exceptions and some countries, including India and China, which have sample registration systems for selected areas. By 1998 half the countries of the world had registration systems with at least 90 per cent complete registration. However, as many of the world’s most populous countries are in the other half, over all fewer than a quarter of vital events worldwide are thought to be recorded (World Bank 1999). Even in well-established systems, the fact that registration is undertaken primarily for administrative reasons may mean that demographically relevant details are not recorded. In England and Wales, for example, the number of legitimate children previously borne by the mother is recorded on the confidential section of the birth certificate. Previous births outside marriage, now of growing significance, are not recorded and so true parity cannot be measured from registration data.
The quality of the information supplied and coded is very important. No one registers his or her own death and the information obtained from proxy informants (relatives or, quite frequently, undertakers) about details such as last occupation may be inaccurate. Age, however, is thought to be more reliably recorded on death certificates than in censuses (at least in developed countries), presumably because it is more often checked against other records.
Cause of death
Death certificates are the major source of information on cause of death. In developed countries cause of death is generally certified by a doctor and coded according to the International Classification of Diseases (ICD). There is substantial scope for error and inconsistency at the various stages involved in assigning cause of death. The ICD, which originated from work undertaken by the nineteenth century British medical statistician, William Farr, is now in its tenth revision; each revision has been associated with changes particularly affecting certain causes of death (Alderson 1981). Changes in coding instructions have also been made from time to time (Ashley and Devis 1992). ‘Fashions’ and national preferences also seem to influence assignment of cause of death, as illustrated in a number of classic papers in which case studies of deaths were distributed to doctors in different countries. Place of death may also be important. In Britain an apparent rise in respiratory disease mortality among elderly people in the 1950s and 1960s was found on investigation largely to reflect the increased proportion of deaths occurring in hospital. The junior hospital doctors who filled out the death certificates were much more likely than family doctors to ascribe the deaths of elderly people to bronchopneumonia (OPCS 1981). Changes over time, and variations between countries, in the use of autopsies also affect cause of death assignment.
Elderly people, now the vast majority of decedents in developed countries, are more likely than the young to suffer multiple pathologies and the number of conditions recorded on death certificates has been increasing, although it varies between countries (Alderson 1981; Ashley and Devis 1992). Choice of one over another as the ‘true’ underlying cause of death is bound to be partially arbitrary. Multiple coding of death certificates and analyses by all mentions of a condition—regardless of whether it is designated ‘underlying’—may be more informative in these circumstances and is increasingly being used in the United States (Manton 1982). However, multiple-coded data are not yet available for many other countries.
Variations in death certificate coding reflecting differences in medical knowledge and diagnosis, in the extent to which autopsies are used, in classification systems, and the quality of registration systems are major factors complicating analyses of trends over time or between countries—often potentially the most interesting. The large proportion of deaths assigned to vague and ill-defined conditions such as ‘old age’ or ‘senility’ in historical and contemporary developing countries presents a particular problem. Preston (1976a) and others have shown a general inverse relationship between deaths from these ’causes’ and deaths from heart disease; in short, recorded increases in mortality from cardiovascular disease may partly reflect improvements in certification and registration. In Fig. 4 the proportion of all deaths in England and Wales assigned to circulatory diseases and to ill-defined causes is shown for age groups over 65 from 1911–1915 to 1997. It can be seen that large proportions of deaths among the very old were assigned to ill-defined groups early in the century and that declines in this proportion were associated with increases in the proportion attributed to circulatory diseases.

Fig. 4 Percentage of all deaths due to circulatory diseases and ill-defined causes, age groups 65–69 to 80+, England and Wales, 1911–1915 to 1997. (Data from ONS 1998.)

In countries which lack adequate certification and registration systems, data on deaths by cause are seriously limited. To compensate for this, attempts have been made to develop protocols for collecting information from lay informants which can be used to assign cause of death (Garenne and Fontaine 1990; Snow et al. 1992). However, although this approach has been useful in a number of small-scale investigations of population subgroups, its widespread application would be extremely costly.
Other data sources
Sample surveys now represent a major addition or, in some cases an alternative, to conventional demographic data sources. Most developed countries have a range of government-sponsored surveys which provide far more detailed information on, for example, health-related behaviour, family building strategies, or reasons for migration than it would be possible to collect in a census. In the developing world, where other data sources are scarce, surveys of various kinds often present the best source of data on basic demographic parameters. Data quality is potentially much better in a survey than a census, as it is more likely that well-trained interviewers can be used. The World Fertility Survey, an international population research programme launched in 1972 to determine fertility levels throughout the world, and its successor, the Demographic and Health Survey Programme, have been particularly valuable in providing data for a range of countries, including many lacking adequate vital registration systems. Other approaches to data capture include multiround surveys, in which respondents are asked about events since last contact, and dual-record systems which involve two independent data collection systems (one often a multiround survey), the results of which are then combined. This method allows some estimation of missed events to be made, but is expensive. These approaches are described in more detail in most standard demographic textbooks (Shryock et al. 1976; Newell 1988; Pollard et al. 1990; Hinde 1998).
The raw materials of demography relate to individuals’ most personal experiences—sexual activity, family formation, birth control, reproduction, marital breakdown, illness, and death. All of these experiences occur in a social framework which attaches value to some of these behaviours and stigmatizes others. Not surprisingly, respondents in censuses and surveys may be reluctant to disclose non-marital pregnancies, illegal abortions, illegal migration, or deaths of relatives from AIDS. Concealment has also been the policy of some national governments which have treated demographic data as official secrets.
As well as allowing for these personal and political factors, the enormous potential complications arising from people’s uncertainties about age or other ‘basic’ characteristics, uncertain recollections of prior events, and the vast scope for administrative errors of various kinds have to be considered. In this context the demographer’s traditional obsession with data quality becomes understandable. Very real current questions turn on issues of data quality. In the United States, for example, death rates for African-Americans are higher than those for white people until the age of 75 years or so when they appear to ‘cross over’. This has been interpreted as an effect of selective survival and the differential health challenges faced by black and white Americans. As these challenges are greater on average for black than white people it is argued that only the most healthy black people survive to old age and so their mortality thereafter is lower than that of white people (Markides and Keith 1995). However, detailed investigation, including matching with other sources of information on age, has shown that this ‘cross-over’ is almost certainly an artefactual result of differences between ethnic groups in the extent of age misreporting in censuses and on death certificates. Age mis-statement in the census is worse among African-Americans, many of whom were born in the Southern States which did not have birth registration systems until 1920. As a result late age mortality rates in this group have been underestimated (Elo and Preston 1994).
The statistics produced in series like the United Nations demographic yearbooks have their origins in what is or has been done by millions of people, mediated by what is said about these events and experiences, further filtered by how this is recorded, processed, and analysed. Some assessment of data quality is given in the foreword to the United Nations demographic yearbooks, but sometimes users may pay insufficient attention to this. Apart from this series, and a range of other United Nations publications (UNFPA 1998; UN 1999), a number of other organizations produce international reference works. These include the World Health Organization (WHO), which produces the World Health Statistics Annual (WHO 1998) and other publications, the World Bank (World Bank 1999), and the United States Census Bureau for International Research (US Bureau of the Census 1992). The United Nations regional directorates, the Organization for Economic Co-operation and Development (OECD), the Council of Europe, and Eurostat (Eurostat 1998) also produce very helpful series and ‘one-off’ compilations for particular regions.
Analysis of demographic data
A standard array of techniques and measures forms the basis of much demographic analysis, the most common of which are described briefly below. Further detail is supplied in a number of textbooks (Shryock et al. 1976; Newell 1988; Pollard et al. 1990; Hinde 1998). Analysis involves not just the application of a particular technique, but decisions about what units of analysis to use and how to group them. A major distinction of the latter type is between period and cohort analysis. Period analysis deals with events of a particular time period (for example, mortality rates from 1995 to 1999) while cohort analyses follow the experience of individuals through time. Cohorts in this sense are defined as groups of people who have experienced the same significant event at the same time. Thus birth cohorts comprise people born in a particular year or group of years and marriage cohorts those marrying at a particular time. Other important events, such as leaving school, divorce, or retirement, may also be used to define a cohort. The health status of individuals at one point in time is clearly likely to be influenced by past exposures and environments, including experiences very early in life (Barker 1992). For this reason cohort and ‘life-course’ approaches to analysing mortality and other indicators of population health have an intuitive appeal.
Birth cohort and time period are two of the dimensions which locate people in time; the third is age. Duration effects (such as duration of marriage or length of exposure to a particular pathogen) may also be important. Models in which age, period, and cohort effects are considered separately have been used in mortality analyses, but less so in analyses of reproductive behaviour (Hobcraft et al. 1982). Cohort effects may be substantial and, unless allowed for, may mask relationships between, for example, age and various risks or patterns of behaviour. Differences in the smoking behaviour of cohorts, for example, have a major effect on the relationships between age and smoking-related disease observed at different periods.
Other decisions about whether to use individuals, families, households, or geographical areas as units of analysis are often constrained by data availability. Until relatively recently, for example, most census data were only available as aggregate tabulations. The growing availability of microdata—individual-level information—has greatly extended the scope of demographic analysis. Other innovations include the development of sample record linkage systems, such as the Office of National Statistics Longitudinal Study in England and Wales (Fox and Goldblatt 1982). In these datasets individuals’ census records are linked with their vital registration records so numerator–denominator biases in, for example, the analysis of mortality are avoided.
Measurement of fertility
Fertility means the child-bearing performance of a woman, couple, or population. Generally only live births are included. The term fecundity, by contrast, is used to refer to the physiological capability of producing a live-born child. Confusingly, in the French-speaking world the meaning of these terms is reversed, so fécundité in French means fertility in English. A rough idea of fertility may be gained from using census or survey data to calculate child–woman ratios: the ratio of 0- to 4-year-olds to women aged 15 to 49 years. However, the survival of infants (and their mothers) affects these ratios, so they are generally only used if no other data are available.
The simplest measure of fertility commonly used is the crude birth rate—the number of births in a particular year per 1000 population. As the denominator of this includes those not ‘at risk’ of giving birth (men and women outside the reproductive age groups), it is really a ratio rather than a rate. Crude birth rates are influenced by the age structure of the population, but less seriously than crude death rates. In the late 1990s crude birth rates ranged from 10 per 1000 in parts of Europe to 41 per 1000 in sub-Saharan Africa.
Slightly more sophisticated (and demanding of data) is the general fertility ratio—births per 1000 women of reproductive age (generally defined as 15–49 or 15–44 years of age). Where data allow, age-specific fertility rates (births per 1000 women of a particular age or age group) are preferred. These are frequently summarized using the total fertility rate. Where, as is usually the case, period data are used to calculate this, it indicates how many children women in a hypothetical cohort would have if they experienced current age-specific fertility rates throughout their reproductive life. This measure is sometimes explicitly denoted as total period fertility rate. In developed societies a total fertility rate of 2.1 is taken to indicate replacement level fertility as, under this regime, a cohort of women would be succeeded by a cohort of daughters of the same size (after some allowance for mortality and the fact that 105–106 boys are born for every 100 girls). Fertility levels in much of the developed world have been below this level for some 20 years. In 1999 total fertility rates were lowest in southern and eastern Europe and Japan, being 1.2 or lower in Bulgaria, Latvia, Spain, the Czech Republic, and Italy, and between 1.2 and 1.5 in a large number of other countries including Germany, Greece, Russia, Japan, Poland, and Sweden. Total fertility rates are highest in sub-Saharan Africa; in 1999 Niger, Oman, Ethiopia, and Uganda had estimated total fertility rates close to, or above, 7.
One difficulty with the total fertility rate, particularly as a tool for examining trends during periods of change, is that it is affected by changes in the ‘tempo’ rather than the ‘quantum’ of child bearing. If women start delaying their fertility but ‘catch up’ later, there will be a divergence between cohort and period measures, as the latter will be based partly on the behaviour of earlier cohorts whose timing of births was different (Cooper 1991). Similarly, if women have children earlier, total fertility rates will rise, even if eventual family sizes remain unchanged. Crises of various kinds may also produce large tempo effects. In many central and eastern European states fertility rates have fallen precipitously since 1990, but couples may ‘make up’ these births if conditions improve. For this reason, many statistical offices have used cohort measures of fertility as the basis for projections rather than period ones. Although apparently a technical matter, considerable controversy surrounds this issue, particularly in France where it has inspired front-page articles in Le Monde and acrimonious resignations of demographers. This reflects a long-standing French pronatalist tradition and concern about low fertility. In this context it is vital to know whether recent trends in fertility, as measured by total fertility rates, are partly an artefactual result of changes in the timing of births or whether they really indicate a change in final family size.
More sophisticated measures of fertility include parity progression ratios. These indicate the probability of proceeding from one birth to another (for example, what proportion of mothers with two children progress to having a third). Parity progression ratios are normally calculated for cohorts who have completed, or nearly completed, their child bearing. However, it is also possible to use data on births by birth order to derive period, rather than cohort, progression ratios (Feeney and Yu 1987). Summary information on measures of fertility and reproduction are summarized in Box 1 and Box 2.
Box 1 Fertility measures

Fertility: the child-bearing performance of individuals, couples, or populations.
Fecundity: the physiological capability of producing a live birth.
Parity: the number of children previously born alive (or sometimes number of previous confinements) to a woman or couple. Nulliparous women are those who have borne no children.
Crude birth rate (CBR): the ratio of births in a year (other specified period) to the average population in the same year/period (mid-year population), expressed per 1000.

General fertility rate (GFR):births to women aged 15–44/49 in a year/period per 1000 women aged 15–44/49 in the same period.

Age-specific fertility rate (ASFR): number of births to women aged x (or x to x+ n) per 1000 women aged x (or x to x + n) where n is the length of an age interval. ASFRs are frequently calculated for 5-year age groups from 15–19 to 40–44 or 45–49.

Total (period) fertility rate (TFR/TPFR): the sum of the age-specific fertility rates for all reproductive age groups for a particular period (usually a year), conventionally expressed per woman. The TFR indicates how many children a woman would have if throughout her reproductive life, she had children at the age-specific rates prevalent in the specified year or period.
TFR = å fx
x = 15
where fx is the age-specific fertility rate at age x. If rates for age groups, rather than single years, are used then the sum of the age-specific rates must be multiplied by the number of single ages included in the group (usually 5).
TFR = 5 × å fx
x = 15–19.
Parity progression ratio: the probability of a women of parity xprogressing to parity x + 1.
Sources: various; see Pressat and Wilson (1985) which provides a valuable guide to demographic terms and issues.

Box 2 Reproduction ratios

Gross reproduction rate (GRR): the sum of the age-specific female fertility rates (births of daughters), for all reproductive age groups for a particular period (usually a year) conventionally expressed per woman. The GRR indicates how many daughters a woman would have if, throughout her reproductive life, she had children at the age-specific rates prevalent in the specified year of period. The GRR can be calculated either by summing female age-specific fertility rates (relating to births of daughters rather than all births) or using the formula
GRR = TFR × proportion of female births.
The proportion of female births can be taken as 0.488 (100/205) in the absence of more detailed information.
Net reproduction rate (NRR): the average number of daughters that would be borne, according to specified rates of mortality and of bearing daughters, by a woman subject through life to these rates. The NRR employs the same fertility data as the GRR, but also takes into account the effects of mortality. An NRR of 1.0 indicates that a population’s fertility and mortality levels would result in exact replacement of mothers by daughters.

Marriage patterns have a major influence on fertility, and upsurges and downturns in marriage are associated with lagged upsurges and downturns in births. Where possible, demographers have often preferred to calculate age-specific marital fertility rates (and total fertility rates and other measures) on the grounds that the unmarried population is not ‘at risk’ (or at very reduced risk) of child bearing. Changes in marital fertility indicative of deliberate attempts to limit family size are regarded as one of the defining features of the fertility ‘transition’ (see below) and so distinguishing these from changes due to variations in the ‘at risk’ (married population) has been particularly emphasized. However, in a growing number of populations, rises in non-marital child bearing mean that restricting analyses to marital fertility is no longer appropriate.
Reproduction rates
In the long term, populations will grow if mothers replace themselves with one or more (surviving) daughters and decline if they fail to achieve this. Theoretically, it would also be possible to measure the replacement of fathers by sons, but in practice the difficulties involved in obtaining paternity data make this unfeasible. Reproduction rates thus relate only to female fertility—that is, births of daughters. The gross reproduction rate is derived in exactly the same way as the total fertility rate except that age-specific birth rates based only on births of daughters are used in the calculation. The net reproduction rate makes an allowance for mortality—specifically the chance that a daughter will survive to the age her mother was when she was born. The net reproduction rate cannot be calculated unless both age-specific fertility and mortality data are available (although it can be approximated using the gross reproduction rate and appropriate life table survival data). Changes in either fertility or mortality (or both) will mean a divergence between period measures (based on the experience of a hypothetical cohort) and the experiences of real cohorts.
Measurement of mortality
As for fertility, the simplest measure of mortality is the crude mortality rate, deaths per 1000 population. However, as noted above, this is strongly influenced by age structure, and age- and sex-specific rates, or measures based on them, which are much preferred if data are available to calculate them. As in epidemiology, both direct and indirect standardization are sometimes used to make comparisons between populations with different age and sex structures. Standardized mortality ratios, which are frequently used to compare, for example, mortality in different regions of countries, are calculated using indirect standardization. This involves selecting a set of ‘standard’ age-specific mortality rates, for example those for a national population, and applying these to the numbers of people in the relevant age groups in the subpopulation of interest, for example the population of a particular region. This yields an ‘expected’ number of deaths—the number of deaths there would be in the subpopulation if age-specific death rates were the same as those in the standard population. The ratio of observed to expected deaths, conventionally multiplied by 100, gives the standardized mortality ratio. Thus, for example, a standardized mortality ratio of 124 indicates that mortality in the subpopulation is 24 per cent higher than in the standard population, allowing for age differences. Standardized mortality ratios are useful summary measures of differences in mortality, but they give no indication of the level of mortality.
Age-specific death rates are calculated using the numbers of deaths at age x (or between ages x and x + n) in a particular year as the numerator and the mid-year population of the same age as the denominator. The rate derived is conventionally expressed per 1000 or per 100 000 population. In this calculation, the mid-year population is used as a measure of the average population at risk on the assumption that deaths are evenly distributed throughout the year. However, for some age groups, notably infants, this assumption is invalid. In developed countries deaths in the first 3 days of life may account for half or more of all deaths in the first year of life. Moreover, information on the size of population aged less than 1 year normally comes from birth data (as in nine out of ten years relevant census data will not be available). For these reasons live births in a particular year are conventionally used as the denominator of the infant mortality rate while deaths to infants aged less than 1 year constitute the numerator. Some infants dying in a given year will have been born in the previous year and some born in the year in question will die the following year. This can cause distortions if there are large annual fluctuations in numbers of births (or infant deaths) and often 3-year averages are preferred.
Infant mortality rates were very high in some parts of historical Europe (300 or even 400 deaths per 1000 live births in regions of Russia and Germany at the end of the nineteenth century (van de Walle 1986)). In England and Wales, where declines in infant mortality came later than declines in other age groups, the infant mortality rate at the start of the twentieth century stood at some 140 infant deaths per 1000 live births and a third of all deaths occurred among those under 5 years of age. Infant mortality in contemporary developed countries is now extremely low—fewer than five infant deaths per 1000 live births in Sweden, Norway, Finland, and Singapore. There have also been huge falls in infant mortality in much of the developing world, but rates remain high—well over 100 deaths per 1000 live births in parts of sub-Saharan Africa.
Variations on this kind of scale have a very substantial demographic impact. Infant mortality has also attracted particular research interest because of hypothesized and observed links with fertility behaviour and as an indicator of public health standards and conditions. Particularly in this latter context, perinatal, early and late neonatal and postneonatal mortality rates are often distinguished where data allow (Box 3). A further refinement is to try and distinguish ‘endogenous’ mortality (from congenital malfunctions and birth trauma) from ‘exogenous’ causes which are more amenable to intervention.
Box 3 Mortality measures

Crude death rate: the ratio of deaths in a year (other specified period) to average population in the same year/period (mid-year population), expressed per 1000.

Age-specific mortality rate(ASMR): number of deaths to persons aged x (or x to x + n) per 1000 persons aged x(or to x + n).

Standardized mortality ratio (SMR): the ratio (× 100) of observed to expected deaths in a study population. Expected deaths are calculated by applying a set of standard age-specific mortality rates to the age distribution of the study population. Standardized ratios are only useful for comparisons. They have no intrinsic meaning.

Infant mortality rate (IMR):

This is sometimes decomposed into neonatal mortality rates (deaths of live born infants during the first 4 weeks) and postneonatal mortality (from 4 to 52 weeks). The perinatal mortality rate measures late fetal deaths (stillbirths) and early neonatal deaths relative to live births.

Stillbirths used to refer to deaths of fetuses of 28 or more weeks gestation; however, an earlier threshold of 24 weeks is now more generally used.

Life tables
Life-table analysis is a core demographic technique and life tables provide one of the most powerful tools for analysing mortality and other non-renewable processes.
Life tables are derived from age-specific mortality rates and show the probability of dying (and surviving) between specified ages. They also allow the calculation of various other indicators, including expectation of life. If complete data on the mortality of a birth cohort are available, then a cohort life table may be constructed. However, the use of cohort life tables is obviously only possible retrospectively. More commonly period mortality rates, based on mortality rates at a particular time, are calculated. These life tables show death (and survival) probabilities for a hypothetical cohort with an arbitrary radix (number of babies at the beginning) usually set to 10 000, 100 000 or some other multiple of 100.
Specific notation is used in life-table analysis; this is summarized in Box 4. The basis of the table is a set of probabilities of dying, nqx, where x refers to age at the start of an interval whose length is specified by n. Thus 5q50 refers to the probability of someone alive at 50 dying between age 50 and age 55. The complement of nqx—the probability of surviving—is denoted npx. The (hypothetical) number of survivors at each age is given by lx; thus l0 equals the radix (of 100 000) and l75 the number of survivors at age 75. The number of person years lived in an interval (nLx) and the total number of person years lived after a particular age (Tx) are often not shown in published tables but are steps on the way to the calculation of ex, life expectancy at age x.
Box 4 Life-table measures and notation

x = age attained last birthday
lx = number of survivors at age x, so l65 is the number of people alive at age 65 in the hypothetical life-table population
l0 = the radix of the life table (hypothetical number of babies), usually 100 000
nqx = probability of dying between age x and x + n, so 4q1 is the probability of dying between age 1 and 5 years for a person aged 1
nPx = probability of surviving between ages x and x + n, so 20P65 is the probability of surviving from age 65 to 85 years for a person aged 65
nDx = number of deaths between age x and x + n
nLx = number of person years lived between x and x+ n
Tx = total number of person years lived after age x
e00x = expectation of life at age x, so e00 is expectation of life at birth

This indicator of survival is very frequently used and provides a measure of the level of mortality which is very largely independent of the age structure of the population. This makes it more useful than either a standardized mortality ratio (which gives no indication of level) or a crude death rate (which is strongly influenced by age structure). Life expectancy either at birth (e0) or further life expectancy at a particular age, say 65 (e65), is calculated by dividing total person-years lived after age 0 or 65 (T0 or T65) and dividing it by the number of survivors aged 0 (l0) or 65 (l65). Methods of calculation are given in all standard textbooks. The level of infant mortality is a powerful influence on e0 (as so many potential person-years are lost through an infant death). In high mortality populations e0—mean age at death—varies substantially from the median age at death. In Mozambique and Sierra Leone, for example, median age at death in 1990 was only 2, compared with an average (mean) life expectancy of 43 and 38 respectively. In low-mortality populations the correspondence between the two is much closer; in Sweden and Japan in 1990 life expectancy at birth and median age at death were both 78 (World Bank 1993).
Model life tables
Patterns of age-specific death rates show certain similarities whatever the level of mortality. Death rates tend to be higher in infancy than later childhood and rise with age from around the age of puberty. Because of the tendency for death rates at one age to be associated with death rates at other ages in a given population, it is possible to derive hypothetical schedules, called model life tables, describing variations in mortality by age and sex, normally in terms of a limited number of parameters which allow for particular features of the mortality pattern of the population considered. Model life tables are derived from empirical data from countries where these are available. They are extremely useful aids for the estimation of mortality by age in populations with defective data. They are also used (in conjunction with fertility data) to show the outcomes of particular fertility and mortality regimes on, for example, population age structure. All demographic texts give further details of their derivation and application.
Other applications of life table analysis
Life tables are an essential part of much demographic analysis (including, for example, making population projections) and are widely used to analyse events other than death. The oft-quoted figure that one in three marriages in England and Wales will end in divorce is based on life-table analysis of age- and duration-specific divorce probabilities in the mid-1980s (Haskey 1988); more recent work suggests a higher proportion. Life-table methods are also used to measure contraceptive use failure rates and contraceptive use discontinuation rates.
Multiple decrement life tables allow ‘decrements’ from more than one event—for example different causes of death. Cause elimination life tables are also used to identify the ‘pure’ severity of a particular cause of death. Multistate models allow analysis of a range of transitions, particularly those where re-entries into a particular state, such as being married or living in a certain region, are possible. These more sophisticated applications require more detailed data.
Measurement of migration
In many countries migration is the predominant influence on the spatial distribution of the population. In the developing and newly industrialized world, recent rural to urban migration has resulted in the phenomenal growth of cities, often lacking the infrastructure to meet the needs of the expanding population for basic services such as sanitation and power. In Brazil, for example, three-quarters of the population in 1991 lived in urban areas and 36 per cent of the total population were in cities with more than a million inhabitants. In 1965 the equivalent proportions were 56 and 24 per cent respectively (World Bank 1993).
In the older developed world, by contrast, urbanization has been succeeded by ‘counter-urbanization’ involving migration from cities to suburbs or beyond. One result has been a growing concentration of those unable to move (the old and disadvantaged) in inner-city areas.
Measuring migration represents particular difficulties. Some of these arise from problems of definition. The classical definition of internal migration is ‘a permanent or semi-permanent move across an administrative boundary’ (UN 1970). Use of this definition excludes the large volume of movement which takes place within administrative areas—often termed residential mobility—although in many contexts knowing about this may be very important. Use of this definition also means that the extent of migration recorded depends partly on the size of administrative area considered. In a country divided into many small areas a move of over 5 km will count as migration, while in countries divided into few larger ones a move of over 500 km may not. This means that international comparison of internal migration rates is potentially misleading. Even the distinction between international (between country) and internal (within country) migration may be problematic if boundaries are contested or changing. The temporal dimension to migration presents further difficulties. What constitutes permanent or semi-permanent and how should groups such as seasonal migrants be treated?
The reason for defining migration as a move over a boundary is largely a pragmatic one. Often only moves of this kind are recorded; moreover this is the form of data required by the primary users—local administrations. For research purposes, however, analyses of all moves (preferably with an indication of distance moved) may often be preferred. Some countries have registration systems (‘population registers’) in which changes of address or moves between districts are recorded (although with varying completeness and immediacy). More commonly censuses are used to find out about migration. Questions on usual address 1 or 5 years ago allow the proportion of movers in the population to be measured (except for those aged less than 1 or 5 years). These data also allow gross flows—inflows and outflows—between pairs of areas to be measured. Moves, as opposed to movers, are not, however, directly measured as someone moving several times in the reference period cannot be distinguished from someone moving only once. Those leaving an address and later returning to it cannot be identified either. This means that the length of the reference period used is important; the proportion of migrants in the 5 years preceding a census will not equal five times the proportion moving in 1 year before the same census.
In the absence of direct census data, estimates of migration can be made indirectly using the ‘balancing equation’ referred to below. Differences in the size of a population at two points in time (censuses) not accounted for by natural increase or depletion must be due to migration (or errors in the data). If good vital registration data are available, then both births and deaths can be taken into account. If they are lacking, then the survival of groups enumerated in the first of a pair of censuses must be estimated from a life table and the number of expected survivors compared with the number enumerated in the second census (obviously ageing must be allowed for, so the number of 20- to 29-year-olds in the first census will be compared with 30- to 39-year-olds 10 years later). These methods only allow estimation of net migration (balance between in-migration and out-migration). Their major weakness lies in the fact that the residual population balance assumed to be due to migration may in fact reflect differences in the quality of the two censuses considered or errors in the estimates of survival used.
Survey data are also used to measure migration and potentially provide illuminating information on the reasons for, and consequences of, migration. However, as migration over long distances is a relatively rare event, even large samples may yield relatively few such migrants. A similar problem besets samples of international travellers, such as the United Kingdom International Passenger Survey, designed to estimate flows of international migrants through port or border surveys. Tourists and business travellers comprise the vast bulk of people entering or leaving so surveys are an inefficient way of identifying immigrants and emigrants. Unfortunately, other data are often lacking as legal and administrative record systems are frequently concerned with citizenship and right of abode rather than international migration per se (and virtually never with emigration).
At the local level, migration flows may be the predominant influence on population size. Migrants differ from non-migrants so migration has a potentially strong impact on population structure and characteristics. Migration also has a role in determining exposures to new infections. The stresses of migration on the migrant, on those left behind, and possibly on those in the area of destination also indicate that migration is an important element to take account of in health planning. However, both the quality and the timeliness of routinely available migration data may make this difficult.
Population projections
Population projections represent one of the most widely used outputs of demographic analysis. Strictly speaking, a projection simply represents the outcome of applying various assumptions about future fertility, mortality and migration and so differs from a forecast, which implies prediction. However, projections are often treated as forecasts and the degree of uncertainty inherent in them is not always sufficiently acknowledged. The most common method of projection is the component method, based on the balancing equation

where Pt is the population at an end of period, P0 is the population at the beginning of a period, and B, D, I, and E represent births, deaths, immigrations, and emigrations during the same period (that is, net migration). Population subgroups may be similarly defined in terms of entries and exits. Entry to the population aged 75 to 84 years is through ageing (passage from 74 to 75); exit is through further ageing (84 to 85) or death. Although very straightforward, this simple accounting equation is an important one, both methodologically and as a formal reminder of the need to consider past as well as current events.
When making projections, assumptions are made about the three components of change—births, deaths, and migration—and applied to age and sex groups within the initial population to give a projection of future size and structure. To a large extent assumptions are based on recent trends together with other information on, for example, survey data on fertility intentions or (sometimes) models of change in particular causes of death. Forecasting fertility is generally regarded as the most problematic area of population projection; projections of the future population size of the United Kingdom varied widely during the period 1955–1974 when birth rates first rose and then fell. Recently greater attention has also been paid to the errors that have been made in forecasting mortality in developed countries. This has little effect on age groups in which survival is high, but can have quite substantial impacts on forecasts of the number of elderly people (Murphy 1995). At the subnational level migration is an important, and sometimes quite volatile, element which is difficult to predict, especially by those in central statistical offices lacking local knowledge.
International migration is also difficult to deal with because it is affected by policies and events outside the country for which the projection is being made and is often a sensitive political issue. Patterns may vary hugely—witness recent mass movements of refugees. Within the larger countries of Europe, Switzerland has the highest proportion of ‘foreign’ residents—19 per cent in 1997—and 60 per cent of Swiss population growth in the past 15 years has been due to net immigration. In 1993 over two-thirds of (net) immigrants came from former Yugoslavia (UN ECE 1994; Council of Europe 1999). In many other Western European countries, particularly Germany and the United Kingdom, rates of immigration increased during the 1980s and, as in Switzerland, have contributed significantly to population growth (Coleman and Chandola 1999).
Population dynamics
Any population comprises those who have made an entry and not yet exited. When whole populations of defined geographical areas are considered, the only means of entry are birth or immigration and the only means of exit are death or emigration.
Analyses of multiple transitions are rather more complex. The married population, for example, comprises those who have married and not yet died or been widowed or divorced. While being born, dying, or passing from one age to another are unrepeatable events, multiple entries and exits to the state of being married are possible, necessitating the use of multistate models.
Of the three demographic determinants of population size, structure, and growth, fertility is nearly always of much greater importance than either mortality or migration. Every birth represents not just an addition to the current generation of children, but also potentially an exponentially increasing augmentation in the size of future generations. Death carries no such promise of future return, at least in this world. The third demographic determinant—migration—is generally not of significant magnitude to have a major impact on national populations, although there are exceptions. At the subnational level, however, migration may have a very important effect.
For social and biological reasons fertility, mortality, and migration have interactive effects. Decreases in mortality among those with reproductive potential, for example, influence not just the size of the age group affected at the time, but also the size of succeeding generations. Werner (1987) has estimated that, had the 1841 female birth cohort in England and Wales suffered only negligible mortality before the age of 45 years, children born to the cohort would have averaged nearly five per woman as compared with the three per woman (in the original birth cohort) actually achieved.
Declines in male mortality, particularly in populations where large age differences between spouses are common and remarriage of widows is rare, will similarly tend to increase fertility by effectively increasing the proportion of women of reproductive age who are still married. Conversely, reductions in fertility clearly reduce the risk of maternal mortality and may have further positive effects on the survival of mothers, infants, or both. Child bearing at very young ages (under 18 years of age) or old ages (over 35 years of age) and short birth intervals (less than 18 months) all have particular risks (Hobcraft 1991), so a reduction in fertility involving less early or late child bearing and longer birth intervals will have particular benefits. Average age at motherhood also influences rates of population growth. The average age of mothers at the birth of their daughters is termed the mean length of a generation and is generally around 29 years. A shorter interval will mean more rapid generational succession (and so faster population growth), while a longer one will have the opposite effect.
Migration has an effect on both the other demographic parameters because migrants differ from the general population. International migrants are generally young and in good health and often may move from relatively high fertility populations to low fertility populations. As a result immigrants may serve to (temporarily) ‘rejuvenate’ the host population and, at least initially, have higher fertility and lower mortality. In the United Kingdom, for example, in the mid-1980s the total fertility rate for women born in Bangladesh or Pakistan was 5.6, compared with 2.9 for women born in India and 1.8 in the population as a whole (Shaw 1988). In England and Wales, age-standardized mortality ratios for most immigrant groups are below 100 (that is, below average) although the Irish and Scots have above-average mortality. Standardized mortality ratios have also been shown to be raised among second-generation Irish in England, in contrast with the more usual pattern whereby migrant groups in time take on the mortality and morbidity characteristics of the host population (Balarajan and Bulusu 1990).
For these reasons the demographic characteristics of population subgroups largely comprising immigrants and their immediate descendants may vary substantially from those of the population as a whole. The length of the interval since the main period of migration is also important. In Great Britain in 1995 to 1997, for example, over 40 per cent of the population of Bangladeshi origin was aged under 15 and only 5 per cent were aged 60 or over. This reflects both the relatively high fertility of British women born in Bangladesh, but also the fact that half of all immigrants from Bangladesh had migrated since 1975 (Schuman 1999).
Population growth
Population growth is obviously a function of the balance of births and deaths and the extent of net migration. Changes in the size of a population produced by the surplus (or deficit) of births over deaths are termed natural increase (or decrease). A common indicator of growth is the crude rate of natural increase—the difference between the crude birth rate (annual births per 1000 population) and the crude death rate (annual deaths per 1000 population). If net migration is zero, this will be the same as the growth rate of the population—the overall annual change in the population divided by the population size—(conventionally expressed as a percentage). Many populations in Europe have had fertility rates below the level required for long-term replacement for 20 years or more, yet in most cases births still outnumber deaths (exceptions are Italy, Germany, and Sweden) (Council of Europe 1998). This apparent paradox largely reflects the fact that the number of births is a function of the number of potential mothers, as well as of their fertility patterns. If the former is increasing so too may the numbers of births, even if women have fewer children each.
The young age structures of many populations in the developing world mean that these populations have a huge built-in potential for growth. Most sub-Saharan African populations are expected to double in size between 1995 and 2015, despite the devastating impact of HIV/AIDS mortality (UN 1999). Population momentum is the measure which gives the ratio of the ultimate size a given population would achieve to current population size if fertility were to fall immediately to replacement level.
Intrinsic rate of natural increase: stable population theory
Early in the twentieth century Lotka (1907) demonstrated mathematically that a population closed to migration and subject to unchanging age-specific fertility and mortality rates for a long period would eventually have a fixed age structure (in which the proportion in each age group remained unchanged) and would grow at a constant rate. This type of population is called a stable population. The fixed age structure of a stable population is independent of the initial age structure—two very different populations subject to the same unchanging rates for a long period would eventually assume the same structure. A particular variant of a stable population is a stationary population—one in which birth and death rates are constant and in balance and so population growth is zero. The Lx column of the life table is an example of a stationary population. The number of births is fixed (the radix) and the age distribution is also fixed. In non-stationary stable populations the age structure is also fixed but the size of every age group is growing at the same constant rate as the overall population and the number of births. This is called the intrinsic rate of natural increase and is a function of the net reproduction rate and the mean length of a generation (approximated by the mean age of child bearing). Non-stationary stable populations can be calculated by adjusting the Lx values of a particular life table to allow for the intrinsic rate of growth. These are often published in conjunction with model life tables to show the effects of particular (unchanging) fertility and mortality regimes (Coale and Guo 1989).
Although stable and stationary populations are theoretical constructs, real populations at various times have met the model requirements closely enough to allow stable population theory to be used to develop methods for indirectly estimating fertility and mortality in populations lacking adequate directly derived data. Stable population models are also widely used for insurance, pension, and personnel planning.
Age structure
One of the most important results of the work of Lotka and his successors (Coale 1957) was to show theoretically that fertility is the predominant influence on age structure. This has also been demonstrated empirically (Carrier 1962). High fertility populations such as those of contemporary developing countries (Fig. 5) have a pyramid shape with each successive cohort being larger than its predecessor. ‘Old’ populations, such as that of England and Wales (Fig. 6), are more rectangular in structure with a gradual tapering at the top. This difference is the result of sustained downward trends in fertility which reduce the proportion of children in more recently born cohorts in the population and so lead to a corresponding increase in the proportion of older people (survivors of larger cohorts). Historically, and apparently paradoxically, improvements in mortality in those populations which now have high proportions of old people in fact served to offset the trend towards population ageing, as they chiefly benefited the young—and led to increases in the proportions surviving to have children themselves. Population pyramids graphically illustrate both the future and the past of populations. The structure of the Russian, Estonian, and Ukrainian populations, for example, shows the legacy of high male mortality in the Second World War, and high mortality in both sexes during the collectivist period. Wars and other crises affect births as well as deaths. The 1998 populations of Russia and the Ukraine show severe indentations at age 50 to 54 years, reflecting low fertility and high numbers of infant deaths during the period of collectivization, famine, and purges in the 1930s, and further indentation around age 45 years reflecting low fertility during the Second World War (Velkoff and Kinsella 1993). Bulges in population pyramids due to high numbers of births have ‘echo’ effects when members of large cohorts themselves have children.

Fig. 5 Population pyramids for a less developed country (Egypt, 1992). (Source: United Nations Demographic Yearbook 1994.)

Fig. 6 Population pyramids for developed countries (England and Wales, 1991). (Source:United Nations Demographic Yearbook 1994.)

Although fertility has the greatest potential impact on age structure and population growth, in some circumstances mortality may become a more important influence. Many populations in developed countries now have fertility at or below the level required for long-term replacement, average expectations of life at birth of 75 years or more, and near universal survival to the end of the (female) reproductive span. In these demographic conditions, further improvements in mortality have the greatest impact at old ages and further population ageing occurs from the apex, rather than, or as well as from, the base of the population pyramid. Mortality changes are now the main motor of the further ageing of the populations of a number of developed countries with already old age structures (Preston et al. 1989). Recent changes in mortality at older ages have been quite substantial in a number of developed countries. Table 2 shows life-table survivorship based on period mortality data from England and Wales in the early 1970s and mid-1990s. The mortality rates of the latter period imply survival to age 85 years for 40 per cent of women. Moreover, changes in survivorship between the two periods were greater at older ages than at younger, and mortality at young ages is now so low that there is little scope for demographically important further change.

Table 2 Life table estimates of survivorship in England and Wales, 1970 to 1972 and 1994 to 1996

Population ageing is now a major issue in the developed world and a subdiscipline of demography focusing on the demography of ageing has emerged. Population ageing is also rising high on the agenda in a number of recently industrialized and some developing countries. In these latter populations the pace of demographic change has been much faster than occurred historically in the Western world. As a result, the pace of population ageing has also accelerated. The proportion of the Japanese population aged 65 years of age or over increased from 7 per cent in 1970 to nearly 16 per cent in 1997. In France a similar increase took 130 years to achieve. The origins of these age structure changes lie in the demographic transition.
The demographic transition
Towards the end of the nineteenth century (earlier in France) both birth and death rates started falling in a number of Western countries. Long-term trends in fertility and in the proportion of women in certain age groups ever married in England and Wales are shown in Table 3. It can be seen that between 1871–1875 and 1911–1915 the total fertility rate dropped from 4.8 to 2.8; by the early 1930s it was below replacement level, a development which was viewed with alarm and led to the first Royal Commission on Population. Although modern methods of contraception were lacking, it was clear that this huge drop in fertility was the result of the deliberate limitation of family size. Half of all couples married in the 1870s had six or more children compared with 12 per cent of couples married in 1911/1915 (Coleman and Salt 1992). Expectation of life at birth meanwhile increased by some 15 years between the end of the nineteenth century and the early 1930s. Gains were greatest at young ages and greater for women than for men.

Table 3 Long-term trends in fertility (TFRs) and marriage (percentage of 20–24 and 35–39 year old women ever-married), England and Wales

Scholars attempting to understand these profound changes sought to relate changes in demographic regimes to changes in the economic and social environment and so originated the theory of the demographic transition. The ‘classical’ view propounded by Notestein (1945) and others was that in ‘traditional’ societies fertility and mortality are both high and roughly in balance. Change is driven by economic advance which results in lower mortality. Fertility initially remains high, resulting in a rapid period of population growth. After this lag, however, fertility also falls in response to falling mortality and the erosion of ‘traditional’ pronatalist values.
This classical view has since been considerably modified. The work of Coale and his collaborators, in an ambitious project to track the transition in historical Europe, suggested that no economic ‘threshold’ for fertility decline could be identified (Knodel and van de Walle 1979). Moreover, in some populations, notably France, fertility fell before any change in mortality. Regional analyses showed that falls in infant mortality, assumed to be a particularly important stimulus to fertility decline, sometimes followed rather than preceded changes in fertility (van de Walle 1986). Indeed, Woods et al. (1989), on the basis of data from England and Wales, have argued convincingly that the direction of this relationship may lie the other way and that declines in fertility led to reductions in infant mortality, rather than vice versa.
Other historical and contemporary demographers have also shown that in both historical England and Wales and a number of populations in the near contemporary developing world, fertility rose before it fell (Wrigley and Schofield 1981; Dyson and Murphy 1985).
Both the role of mortality decline as a trigger and the dominance of economic change have both been questioned (Cleland and Wilson 1987). Caldwell (1976) has argued that it is not so much modernization but ‘Westernization’ involving increased emphasis on the nuclear family and a change in intergenerational wealth flows (resulting in the costs of children coming to outweigh their potential benefits) that is important.
The huge amount of research on the historical demographic transition in Europe, while interesting, may seem of limited relevance to contemporary problems. However, much of the fuel for this research and debate has come from postwar fears about enormous population growth in the developing world. By 1950 significant mortality declines had been achieved or initiated throughout the world. In China, for example, expectation of life at birth increased from 43 in 1960 to 72 in 1999. Even in sub-Saharan Africa where mortality remains unacceptably high, a gain of nearly 10 years—from 43 to 52—was achieved between 1950 and 1990 although, as discussed below, mortality rates in parts of Africa are expected to rise again as a result of the HIV/AIDS epidemic (World Bank 1993; UN 1999). In this context it seemed imperative to discover the causes of fertility decline and use this knowledge to accelerate fertility ‘adjustment’ to falling mortality. Was ‘development the best contraceptive’ as concluded at the stormy 1974 World Population Conference? Could change be achieved through intensive family planning programmes, as the Taiwanese experience seemed to suggest? Or was some combination of these and other factors the key to fertility transition? Studies of societies in which the fertility transition had occurred seemed to offer the best prospect of an answer to these questions. While simple answers to complex questions are rarely forthcoming, Coale (1973) identified three factors which he considered a prerequisite for fertility decline in contemporary populations. These were that (a) potential parents must think it acceptable to balance the advantages and disadvantages of another child, (b) some advantage must be gained from reduced fertility, and (c) effective techniques of fertility control must be available.
Recent work in developing countries has shown that substantial fertility declines have occurred in a number of countries with only a limited amount of development, such as Sri Lanka, Thailand, and China. In 1981, for example, Sri Lanka, with a per capita gross national product of US$500, had a total fertility rate of 2.5 (which by 1998 had fallen to 2.1) (World Bank 1993; WHO 1999). Common factors identified in some poor and largely rural populations where fertility has fallen significantly since 1950 are well-established education systems, improvements in health care, and some form of extrafamilial welfare. The education of women has been identified as of particular importance as a factor associated with both falling fertility and improved infant survival although, as might be imagined, these relationships are complex (Cleland 1990). Female education and empowerment were recognized as key policy objectives at the 1994 International Conference on Population and Development (ICPD) (UN 1995). At this meeting, and subsequent international consensus conferences, agreed goals to be met by 2015 included universal access to basic education and a closing of the gender gap in education.
The United Nations (UNECE/UNFPA 1992) has classified countries into those where the fertility transition occurred prior to 1950, those experiencing significant declines between 1950 and 1990, and those with high fertility and, by 1990, little downturn. In 1990, 33 per cent of the world’s population lived in countries, nearly all in the developed world, in the first group. The second group included China, India, much of the rest of non-Arab Asia, most of North Africa and Latin America, South Africa, Turkey, and Albania, and accounted for 60 per cent of the world’s population. The remaining 17 per cent of the global population lived in predominantly very poor countries in sub-Saharan Africa, Pakistan, Middle-Eastern nations such as Iran and Iraq, and a few Latin American countries including Bolivia and Guatemala. Even in this latter group there are now clear signs of change and it seems that by the middle of this century fertility transition will be complete or well advanced throughout the world. However, even in populations where changes have occurred, fertility is in many cases still well above-replacement level. In the late 1990s, 56 per cent of the world’s population lived in countries with above-replacement fertility (UN 1999). This, coupled with the youthful age structures of most of the developing world, means that further rapid population growth is inevitable, but current estimates of the extent of this growth are much lower than those made in the recent past.
Proximate determinants of fertility
One of the positive contributions of research into the fertility transition has been improved understanding of biosocial influences on reproduction. A huge range of social, economic, cultural, and psychological factors may influence decisions about family building strategies and family size. However, literacy rates and measures of per capita gross national product do not beget or bear children and none of these variables can have the slightest effect unless translated into patterns of behaviour or physiological characteristics that influence the risks of conception or delivery.
Conversely, other patterns of behaviour with potentially important influences on fertility may be adopted with little or no thought to these consequences. In short, social, economic, and other factors which influence fertility can only do so through the proximate determinants—the biological and behavioural factors which have a direct influence. Davis and Blake (1956), in a classic paper, distinguished a series of ‘intermediate fertility variables’: factors influencing the chance of exposure to risk of pregnancy (marriage and coital frequency), factors influencing risk of pregnancy (such as contraception), and factors influencing pregnancy outcome (spontaneous and induced abortion). The most influential refinement of this work is the Bongaarts decomposition model (Bongaarts 1978). In this four elements chiefly responsible for observed fertility variations were identified:

the proportion of women married (exposed to risk)

contraceptive use

induced abortion

postpartum non-susceptibility to conception (largely determined by breast-feeding practice).
The total fertility rate is dependent on the interactive effect of these variables and hypothetical maximum fertility. In modern ‘post-transition’ populations, fertility decisions are normally couple (or woman) based and are implemented through contraception and abortion (although variations in marriage and partnership are also important). In non-contracepting populations biosocial factors, notably marriage patterns, breast-feeding practices, sexual frequency, and, in some populations, the prevalence of infertility, have been, or are, of major importance.
Entry into marriage, or more generally any sexual union, is important because it marks entry to what Menken (1989) has termed the social reproductive span which nearly always comes later, often much later, than menarche and is terminated by the end of marriage. Fecundity—the potential for bearing children—decreases after the third decade, more sharply after the age of 35; in most non-contracepting populations the average age at last birth is around 40 years, several years earlier than average age at menopause. Social factors, as well as biological ones, are important influences. Sexual activity may cease before menopause as a result of widowhood or separation. In some African populations continued child bearing after becoming a grandmother is disapproved of, leading to ‘terminal abstinence’.
For those within the effective reproductive span—biologically capable of child bearing and in a sexual union—overall fertility is largely a function of length of intervals between births. The length of postpartum amenorrhoea is strongly influenced by the extent of breast feeding. Among non-breast-feeding women, average duration of amenorrhoea is only 1.5 to 2 months, compared with 18 months or more in rural Bangladesh where breast feeding is protracted. Moreover, in some populations sexual activity is proscribed for breast-feeding mothers and so the period of postpartum infecundability extends beyond the period of amenorrhoea.
Longer birth intervals and increased breast feeding also have positive effects on infant and child health. In Kenya infants born within 18 months of the birth of a previous child are twice as likely to die as those born after a longer interval; in Egypt the risks are trebled (Hobcraft 1991). Overall child mortality (deaths before the age of 5 years) might be reduced by as much as 30 per cent in some countries if closely spaced births were delayed (World Bank 1993).
The epidemiological transition
Transitions from relatively high to low mortality regimes have in all populations been associated with transformations in the age, cause, and sex structure of death (Preston 1976a). Omran (1971) coined the phrase ‘epidemiological transition’ to describe this process. Changes in the response of societies to health and disease processes also need consideration. The term ‘health transition’ has been proposed as one which embraces both these phenomena.
Substantial falls in death rates from infectious and parasitic diseases, bronchitis, influenza, pneumonia, diarrhoeal diseases, and maternal mortality are all the hallmark of the epidemiological transition. In England and Wales over half of the gain in life expectancy at birth between 1871 and 1911 was due to reduced infectious disease mortality. Some 20 per cent of the total gain—equivalent to 2 and 2.5 years ‘extra’ life expectancy for men and women respectively—was due to falls in death rates from respiratory tuberculosis (Casselli 1991). The decline in these causes of death, from which the young benefited more than the old, and women more than men, meant that deaths at older ages accounted for a larger share of all deaths. Chronic degenerative diseases, notably circulatory diseases and cancers, are the predominant causes of death in low mortality populations in all of which women live longer than men. As the risks of degenerative disease are strongly age-related, relatively more people are exposed to these risks in populations with old or ageing structures.
Life-table estimates based on current cause specific mortality data (WHO 1992b) imply that about half of all male infants in the developed world will eventually die from diseases of the circulatory system (varying from 32 per cent in France to 59 per cent in Bulgaria). Malignant neoplasms will, if current patterns persist, account for a further 14 per cent (Bulgaria) to 30 per cent (France), and deaths from diseases of the respiratory system between 5 and 15 per cent. Fewer than 1 per cent, or in Japan and some Eastern European countries about 1.5 per cent, will eventually die from infections and parasitic diseases. Among women the chances of eventual death from malignant neoplasms are generally lower and those from other diseases correspondingly higher.
The relative contribution of various eighteenth- and nineteenth-century developments in promoting the historical epidemiological transition in the West remains a matter of debate. Improved nutrition, better housing and living conditions, public sanitation schemes, and specific public health initiatives, such as smallpox inoculation and vaccination, all have their particular adherents (Coleman and Salt 1992). A common thread linking most of these factors is their relationship to overall social and economic development and improvements in standards of living. During the twentieth century, however, developments in medical technology and pest control offered the potential for ‘exogenous’ mortality decline less dependent on a particular country’s level of income and development. The relationship between per capita income and life expectancy has shifted upward (Preston 1976b). In 1900, for example, life expectancy in the United States was about 49 and income per capita was about US$4800 (1991 prices). In 1990 that income per capita was associated with a life expectancy of about 71 years (World Bank 1993). The WHO has recently analysed associations between changes in infant mortality and changes in average income during 1952 to 1992 in a range of low- and middle-income countries. They concluded that, had this relationship remained constant, falls in infant mortality over the period would only have been a third of the falls in fact observed. This report suggested that access to health technology and improved education were the keys to health improvements while the effects of economic growth were relatively weak (WHO 1999).
One result of growing scientific knowledge is the greater potential importance of public policy and education. Historical research suggests that in the nineteenth century differences in child mortality by level of mother’s education or standard of living were slight, while in contemporary developing countries they are marked. Countries such as Sri Lanka which have vigorously pursued public health programmes and achieved high levels of education, have attained particularly marked gains in life expectancy. In the developed world differentials in health-related behaviour are also strongly associated with income and education. In England, for example, some 40 per cent of men in households headed by an unskilled worker smoke cigarettes, compared with 12 per cent of men from professional groups (Prescott-Clarke and Primatesta 1997). It has also been argued that societal factors—such as the extent of income inequalities and degree of social cohesion—may account for some of the differences between developed countries in the overall level of mortality (Wilkinson 1996).
The process of the epidemiological transition (or at least the initial phases) is now complete or under way in much of the world. Non-communicable causes of death now predominate not just in the developed world, but also in Latin America and East and Southeast Asia and are projected to become increasingly important elsewhere.
Gender differentials
Changes in gender differentials in mortality are a major component of the epidemiological transition. Figure 7 shows sex ratios in mortality rates by age for England and Wales in 1901 and 1997. It can be seen that the extent of female advantage was considerably greater in 1997 and most marked in young adulthood and late middle age. Although the former peak is more pronounced, the latter (and continuing differential in old age) is much more important demographically, as death rates are much higher at these older ages. In north-western European populations, such as those in England and Wales, The Netherlands, Denmark, Sweden, Norway, and Switzerland, about half the gender difference in life expectancy at birth is due to differences among those aged 65 years or more. In much of Eastern Europe, where sex differentials in mortality are greatest, the age pattern is rather different. Countries such as Hungary and Poland show a smaller decline from the 20- to 24-year-old peak in sex ratios of mortality and a greater proportion of the overall difference in life expectancy at birth is due to gender differences among 35- to 44-year-olds than is the case in Western Europe (UN Secretariat 1988; Velkoff and Kinsella 1993).

Fig. 7 Sex ratios of death rates (males/females) by age in England and Wales, 1901 and 1997. (Data from UN 1988; ONS 1997.)

In all developed and the great majority of developing countries, female life expectancy is now greater than male (Fig. 8), but there is considerable variation in the extent of this difference. Low-mortality countries generally have larger sex differences in life expectancies than high-mortality populations, reflecting the association between falls in mortality and an increasing female advantage. Waldron (1985), in a review of this topic, pointed to changes in the intrahousehold allocation of resources, declines in causes of death specifically or primarily affecting women (such as maternal mortality and respiratory tuberculosis), gender differences in health-related behaviour and in exposure to occupational hazards, and the possibly greater susceptibility of men to stresses associated with socio-economic changes, as causal factors.

Fig. 8 Female life expectancy at birth (e0) and male disadvantage in e0, selected countries, 1995–2000. (Data from UN 1999.)

In a few developing countries, such as Bangladesh, Pakistan, and Iran, the female advantage in mortality is either very slight or non-existent. The low status of females and consequent inequitable distribution of resources is partly thought to account for this. A more extreme example of discrimination against females—female infanticide—has been alleged to account for the biologically implausible sex ratio of infants and children in China. This allegation is highly controversial and may reflect girls being ‘hidden’ from officials by couples anxious for the chance of having a son (Zeng et al. 1993). While the extent of sex-selective infanticide is disputed, recent trends in sex ratios at birth in countries with strong preferences for male offspring and small families, such as China and the Republic of Korea, are clearly indicative of sex-selective abortion. The future demographic and social implications of this are a growing area of concern (Zeng et al. 1993; Coale and Bannister 1999).
In Western developed countries a large proportion of the sex differential in life expectancy at birth is due to differences in ischaemic heart disease (well over a third in Britain, Sweden, and Finland), lung cancer (5 to 15 per cent), and accidents and violence (10 to 20 per cent) (UN Secretariat 1988). The twentieth-century epidemic of smoking and attendant smoking-related diseases is undoubtedly a major cause of gender differences in death. In some populations, including England, Wales, Italy, and the United States, there are recent signs of a reduction in the extent of female advantage in life expectancy (Kinsella and Gist 1998; UN Secretariat 1988). This may partially reflect the narrowing of differentials in exposure to tobacco as cohorts including larger proportions of female smokers reach later life. In the United States, age-adjusted death rates from respiratory cancer increased by over 400 per cent for white females aged 65 years or more between 1960 and 1986, while among white males the increase was 115 per cent (Furner et al. 1993). However, in other countries with very high overall life expectancy such as Japan, France, and Germany, gains in female longevity continue to outpace those of males.
Particularly in Eastern Europe, alcohol is also a factor contributing to gender differentials in mortality. As discussed further below, in contrast to the experience of many Western developed nations, male mortality has stagnated or deteriorated in much of Eastern Europe in recent decades; in Russia the crude death rate increased by over a third between 1989 and 1993 (Shapiro 1994) and male life expectancy at birth dropped to 56 in 1994 (WHO 1998). In the former Soviet Union, however, there was a sharp drop in mortality, particularly among men, in 1986 to 1987 (unfortunately, since reversed). This has been attributed to measures introduced in May 1985 to reduce alcohol consumption (Virganskaya and Dmitriev 1992).
Gender differences in mortality are generally discussed in terms of female advantage. However, this advantage has some negative consequences. The elderly population in nearly all countries of the world is predominantly female. In some populations, notably those of the former Soviet Union, the legacy of heavy male war-related mortality, together with the general trend towards differential mortality, has resulted in a particular preponderance of females in elderly age groups. Widowhood is consequently high. Other populations with very different demographic patterns but with large age differences between spouses, notably Bangladesh, also have high proportions of widows. Apart from the economic and social disadvantages which often result from widowhood, older women experience more disability than older men and spend proportionately less of their life free of disability (Robine et al. 1998). This seems to reflect women’s greater risk of disability from musculoskeletal disorders and (related) longer survival than men after onset of disability (Manton et al. 1995).
As a result of population ageing, the epidemiological transition, and recent declines in mortality rates at older ages, the issue of measuring and assessing trends in morbidity, and their relationship with mortality, is now a topic of major concern.
Recent trends in mortality
In much of the developed world rapid gains in life expectancy in the first half of the twentieth century were followed by a period of stagnation in adult male mortality rates in the 1950s or 1960s. It was assumed by some analysts of the period that further major falls in mortality were unlikely, either because endogenous mortality rates from degenerative diseases were inextricably associated with urbanization or industrialization, or because life expectancy was close to biological limits (Bourgeois-Pichat 1952). However, during the 1970s both male and female mortality began to fall again in many developed countries (excluding Eastern Europe). These gains were greatest in older age groups (where the scope for further reduction was greatest). As a result, changes in death rates at older ages have come to play an increasingly important role in both population ageing and in overall mortality change. Myers (1995), in a detailed examination of changes in six developed countries, found that in five of them the proportion of overall life expectancy increase in the 1980s, due to gains among those aged 65 years or more, was over 40 per cent for males and nearly 60 per cent for females. Table 4 shows that in Japan 48 per cent of the female and 31 per cent of the male gain in life expectancy at birth during 1985 to 1990 was due to falls in mortality among those aged 75 or over. In 1955 to 1960, by contrast, half of female and two-thirds of male gains were due to falls in child mortality, while changes among those aged 65 years or over had no or a negative effect on life expectancy. Improvements in the mortality of older people have been just as, or more, evident among the ‘older old’ as among those in their sixties or seventies (Manton and Vaupel 1995). Indeed painstaking analyses of verified and corrected data show remarkable falls in mortality at very advanced ages leading to an ‘explosion’ of centenarians (Thatcher 1999). In England and Wales, for example, there were some 5523 centenarians in 1996 compared with about 100 in 1911 (the total population over this period increased by 44 per cent).

Table 4 Contribution of mortality reduction in each broad age group to the increase in life expectancy, Japan 1955–1990

Recent gains in life expectancy in many developing countries have also been substantial. Between 1978 and 1998 average life expectancy at birth increased by 10 years or more in Bangladesh, India, and Indonesia (WHO 1999). In other regions, however, recent changes have been less benign. Since 1990, mortality has increased in over 30 countries of the world (World Bank 1999). In Russia and other ‘transition’ countries of the former USSR levels of mortality have stagnated or deteriorated reflecting the collapse of previous support systems, extensive alcohol abuse, and resurgence of infectious diseases, including diphtheria. Even more grim is the situation in those parts of sub-Saharan Africa most severely affected by the HIV/AIDS epidemic.
In the nine African countries with an adult HIV prevalence of 10 per cent or more, average life expectancy in 1995 to 2000 is estimated to be 48 years, compared with the 58 years that would have been expected in the absence of AIDS (UN 1999). In Botswana a quarter of the adult population has HIV/AIDS. Such high rates of morbidity and premature mortality have enormous social and economic implications for the well being of future, as well as current, generations. In the 19 African countries considered in a recent United States Agency for International Development study (USAID 1997), an estimated 16 per cent of the population under 15 years of age—up to 40 million children—will be orphaned by AIDS in 2010.
Disability-free or healthy life expectancy
Ruzicka and Kane (1990) are among those who have argued that one important consequence of the epidemiological transition has been a decline in the usefulness of conventional indicators of the health of a population, such as life expectancy and cause-specific mortality. Many of the most common chronic conditions reported by the growing proportions of older people, such as musculoskeletal and sensory impairments, may have serious implications for health status but are not directly life-threatening and do not feature as prominent recorded causes of death. If, as some have argued, recent reductions in mortality are due partly to the prolongation of the process of dying rather than an extension of healthy life, then life expectancy may be becoming a less valid summary indicator of a population’s health. In response to these concerns, increasing attention has been paid to disaggregating life expectancy into ‘healthy’ and ‘unhealthy’ or ‘disabled’ components. Estimates of healthy life expectancy now exist for over 30 countries and an international network has been established to promote standard methodologies and harmonization of data (Robine et al. 1992). Most existing estimates have been derived from cross-sectional morbidity or disability prevalence data used in conjunction with age-specific mortality data. More sophisticated (and data-demanding) multistate approaches which allow transitions both to and from disabled states have also been pursued (Rogers et al. 1990). One result of this work has been to demonstrate formally the greater disability experienced by women.
A slightly different approach to the measurement of population health has been adopted by the World Bank which, together with the WHO, has developed the disability-adjusted life year as a way of quantifying the full loss of healthy life (World Bank 1993). This ambitious exercise involved calculating the years of life lost by each death (defined as the difference between actual age of death and expectation of life at that age in a low-mortality population) and then making a further adjustment to allow for disability according to duration of the condition, with a weighting for severity. The disability data used in this project were drawn from community surveys and expert opinion and both deaths and disability were cause-specified (a process also involving the use of expert opinion where other data were lacking).
Further development of this approach is in progress and doubtless revisions of some of the detailed results may be suggested as other data become available. However, the estimation of disability-adjusted life year data represents a major step forward in the assessment of the burden of disease in the world’s diverse population. As can be seen in Figs 9 and 10, this measure graphically illustrates the different problems faced by various world regions. In parts of the developing world, notably sub-Saharan Africa, loss of disability-adjusted life years is greatest among the young. In the developed market economies, by contrast, the health of the older population represents the major public health challenge.

Fig. 9 Disability-adjusted life years lost per 1000 population by world region, 1991. (Data source: World Bank 1993.)

Fig. 10 Distribution of disability-adjusted life years lost by age group and world region, 1991. (Data source: World Bank 1993.)

Implications of recent demographic trends
In the developed world there have been substantial variations in post-transition fertility levels and trends. Many developed countries experienced a post-Second World War ‘baby boom’ during the 1950s and early 1960s. In Britain this peaked in 1964 when the total fertility rate reached 2.9. The baby boom was more pronounced in the United States where the total fertility rate in the late 1950s exceeded 3.6. Japan, by contrast, had a very short baby boom (1947 to 1949). The baby boom was followed by a ‘baby bust’ period in which fertility declined to very low levels in the 1970s. In most North and European countries fertility rates continued to fall in the 1970s but have since stabilized. Recent falls in fertility have been most pronounced in southern and eastern Europe (and in Ireland) and in general, northern European fertility levels are now higher than southern European ones. Fertility rates in the United States are slightly higher, being close to replacement level rather than below it. Although there have been recent increases in the population of women remaining childless in some developed countries, more important demographically has been the shift towards smaller families. In Portugal, for example, a third of all births in 1965 were to mothers who already had at least three children; by 1995 only 5 per cent of births were fourth or higher order (Council of Europe 1999). The reasons for these recent trends and for the very low fertility now prevalent in much of the developed world remain a matter of lively debate. Economic changes, including the increased labour force participation of women, attitudinal shifts, and advances in the availability of effective birth control have all been proposed as predominant influences. All are probably important, and all may have interactive effects (Becker 1981; Murphy 1993).
The very low fertility in most of Europe and the rest of the developed world (including Japan and the newly-industrialized Southeast Asian nations), the continued ageing of the population, and recent changes in patterns of family organization have raised concerns in a number of countries about the implications of these demographic trends (Allen and Perkins 1995). On the one hand, a group deemed to have substantial support needs (the old) is growing, while on the other, it is feared that the ability or willingness of younger generations to meet these needs may be diminishing.
Currently, and even more so in the recent past, the elderly populations of some developed countries, particularly in north-west Europe, include large proportions who lack children. In Ireland in the late 1970s, for example, a third of elderly people were childless—a legacy of the extreme variant of the European marriage system which prevailed until relatively recently. In the short term, many developed countries will see decreases in the proportion of childless old people (as the parents of the baby boom attain elder status). Moreover, the recent narrowing of sex differentials in mortality observed in some countries has resulted in a delay in the age of widowhood and increase in the proportion of older women still married. Increases in the extent of childlessness and celibacy among postwar birth cohorts and rises in divorce, however, suggest a rather different long-term future. In a number of European countries, a fifth of those born in 1955 are projected to remain childless; in later cohorts this proportion may be even higher (Roussel 1994). In the United States some 60 per cent of recently contracted marriages may end in divorce and 45 per cent of children will experience the break up of their parents’ marriage by the age of 18 years (Bumpass 1990). In Britain current divorce rates imply that over 40 per cent of marriages will end this way (Haskey 1996). Rates of entry into marriage are also diminishing; in Britain and a number of other countries first partnerships are now more often formed through cohabitation than by formal marriage (Kiernan 1999).
In a number of European countries over a quarter, or even half, of all births now occur outside marriage. A large proportion of these (particularly in Scandinavia) are born to couples who have not legally formalized their union rather than to single parents. Union breakdown (divorce) is still the major cause of lone parenthood. Lone parent families now account for 10 per cent or more of all families with children in many parts of Europe.
In Britain the most recent estimates show that in 1991 nearly a fifth of dependent children lived in lone-parent families, compared with fewer than 8 per cent in 1971. In some areas of London lone-parent families account for over 40 per cent of all families with dependent children (Haskey 1994). In the United States, the proportion of children living in mother-only families increased from 8 per cent in 1960 to 23 per cent in 1992 (Da Vanzo and Rahman 1993). Such families suffer a number of disadvantages, many of which may have implications for the health of the children raised in them.
Increases in divorce, lone parenthood, and the participation of married women in the labour force are frequently cited as constraints on the availability of younger generations to provide support for their elders. While the evidence to support these hypothesized relationships is rather weaker than often assumed (Sundström 1994), recent changes in patterns of family formation and dissolution have been interpreted by some as indicative of a major reorientation from familial to individual goals and aspirations representing a ‘second demographic transition’ or ‘cultural shift’ (van da Kaa 1987; Inglehart 1990). If so, this has far-reaching implications for inter- and intragenerational patterns of support, including socialization and rearing of children and support for older people.
Demographic changes in fertility and mortality patterns determine not only the ‘macro’ composition and size of populations but also kinship patterns and the proportion of life spent in certain types of familial relationship. Lower fertility implies a reduced ‘supply’ of children, siblings, cousins, aunts, uncles, nephews, and nieces. Lower mortality, however, means an increase in the availability of older-generation relatives and an increase in the shared life enjoyed by adult children and their parents and grandparents. In Britain today half of all 50-year-olds still have a surviving parent and three-quarters of adults are members of families including at least three living generations linked in direct descent (Grundy et al. 1999).
Whether recent changes in partnership, parenthood, and household characteristics (including large increases in the proportions living alone) represent a second demographic transition (van da Kaa 1987) or the continuation of changes attendant on the first one (Bumpass 1990) is still a matter of debate. If current low levels of fertility persist, the populations of more developed countries will start to decline in size in the twenty-first century, unless negative natural increase is offset by immigration. In most developed countries, between 10 and 18 per cent of the population are aged 65 years or over, and 4 to 7 per cent are aged 75 years or over. In the early decades of this century those over 65 years of age will account for nearly a quarter—and those over 75 years of age 10 per cent or more—of most European populations. In Japan, projections suggest that over 15 per cent of the population will be aged at least 75 years (and 27 per cent at least 65 years old) by the year 2025. Even if the health status of the older population improves over time (owing to improvements in the health legacy of later cohorts), the impact of these demographic changes on demand for health care, particularly long-term care, is likely to be substantial.
The transition to lower fertility now seems to have been initiated in most of the world. However, fertility levels are still high, the population young, and population growth rapid in regions of the developing world particularly sub-Saharan Africa. Despite fears that the pace of decline in infant and child mortality might have slowed in the 1980s, research based principally on the Demographic and Health Survey Programme suggests that this was not the case and that in Northern Africa, Latin America, and Asia the pace of decline continued or increased during the 1980s and 1990s (Cleland et al. 1992; UNFPA 1998).
Continuing improvement in both child and adult mortality is projected for the developing world. Particularly in sub-Saharan Africa, these projections have been revised downwards to take account of AIDS. In the nine African countries with the highest seroprevalence rates, average life expectancy at birth in 2010 to 2015 is projected to reach only 47 years, instead of 64 years in the absence of AIDS—a loss of 17 years (UN 1999).
One result of the differing demographic trends in developed and developing countries is the decreasing demographic significance of the former. In 1950 a third of the world’s population lived in more developed regions, by 1990 this proportion had fallen to 23 per cent, and by 2200 it is projected to be only 10 per cent (UN 1999). Among the many consequences of this shift is an increase in the potential impact of migration from less to more developed countries on the latter.
While there are major differences in demographic indicators and the composition of the population, there is perhaps less divergence in trends between the less and more developed world than has been the case in earlier periods. Moreover, a number of population-related issues are now firmly established as areas of global concern. The ICPD programme of action included recommendations on reproductive health, gender equity, and the health of older people relevant to both developed and developing regions of the world. The consequences of environmental degradation and pollution, persistent or in some cases growing differentials in the health status of different socio-economic groups, and rising numbers of refugees and changing family patterns are similarly issues of global health significance, all of which arise from, interact with, or affect populations. Measuring these trends and assessing their effect on health and demand for health care requires an understanding of population dynamics and population-based measures, and suitable demographic data.
Chapter References
Alderson, M. (1981). International mortality statistics. Macmillan, London.
Allen, I. and Perkins, E. (1995) Demographic influences on the future of family care. In The future of family care (ed. I. Allen and E. Perkins). HMSO, London.
Ashley, J. and Devis, T. (1992). Death certification from the point of view of the epidemiologist. Population Trends, 67, 22–8.
Balarajan, R. and Bulusu, L. (1990). Mortality among immigrants in England and Wales, 1979–83. In Mortality and geography: a review in the mid 1980s (ed. M. Britton), pp. 103–21. HMSO, London.
Barker, D.J.P. (1992). Fetal and infant origins of adult diseases. BMJ Publications, London.
Becker, G.S. (1981). A treatise on the family. Harvard University Press, Cambridge, MA.
Bongaarts, J. (1978). A framework for analysing the proximate determinants of fertility. Population and Development Review, 4, 105–32.
Bourgeois-Pichat, J. (1952). Essai sur la mortalité ‘biologique’ de l’homme Population, 3, 381–94.
Brass, W. (1975). Methods for estimating fertility and mortality from limited and defective data. University of North Carolina, Laboratories for Population Statistics, Chapel Hill, NC.
Bulatao, R.A. and Stephens, P.W. (1992). Global estimates and projections of mortality by cause, 1970–2015. World Bank Policy Research Working Papers (Population Health and Nutrition), WPS1007. World Bank, Washington, DC.
Bumpass, L.L. (1990). What’s happening to the family? Interactions between demographic and institutional change. Demography, 27, 483–98.
Caldwell, J.C. (1976). Toward a restatement of demographic transition theory. Population and Development Review, 2, 321–66.
Carrier, N. (1962). Demographic aspects of the ageing of the population. In Society, problems and methods of study (ed. A.T. Welford, M. Argyle, D.V. Glass, and J.W. Morris). Routledge and Kegan Paul, London.
Casselli, G. (1991). Health transition and cause specific mortality. In The decline of mortality in Europe (ed. R. Schofield, D. Reher, and A. Bideau), pp. 68–96. Clarendon Press, Oxford.
Cleland, J. (1990). Maternal education and child survival: further evidence and explanations. In What do we know about health transition? The cultural, social and behavioural determinants of health (ed. J. Caldwell et al.), pp. 400–19. Health Transition Centre, Australian National University, Canberra.
Cleland, J. and Wilson, C. (1987). Demand theories of the fertility transition. Population Studies, 41, 5–30.
Cleland, J., Bicego, G., and Fegan, G. (1992). Socioeconomic inequalities in childhood mortality: the 1970s to the 1980s. Health Transition Review, 2, 1–8.
Coale, A. (1957). The effects of changes in mortality and fertility on age composition. Milbank Memorial Fund Quarterly, 34, 79–114.
Coale, A. (1973). The demographic transition. In International Population Conference 1973. International Union for the Scientific Study of Population, Liège.
Coale, A. and Bannister, J. (1999). Five decades of missing females in China. Demography, 31, 459–79.
Coale, A. and Guo, G. (1989). Revised regional model life tables at very low levels of mortality, Population Index, 55, 613–43.
Coale, A. and Guo, G. (1990). New regional model life tables at high expectation of life (addendum to 1989 paper). Population Index, 56, 27–41.
Coleman, D. and Chandola, T. (1999). Britain’s place in Europe’s population. In Changing Britain: families and households in the 1990s (ed. S. McRae), pp. 37–67. Oxford University Press.
Coleman, D. and Salt, J. (1992). The British population: patterns, trends and processes. Oxford University Press.
Cooper, J. (1991). The divergence between period and cohort measures of fertility. Population Trends, 63, 19–21.
Council of Europe (1998). Recent demographic developments in Europe. Council of Europe Publishing, Strasbourg.
Da Vanzo, J. and Rahman, M.O. (1993). American families: trends and correlates. Population Index, 59, 350–86.
Davis, K. and Blake, J. (1956). Social structure and fertility: an analytic framework. Economic Development and Cultural Change, 4, 211–35.
Demeny, P. (1972). Early fertility decline in Austria-Hungary: a lesson on demographic transition. In Population and social change (ed. D.V. Glass and R. Revelle), pp. 153–72. Edward Arnold, London.
Dyson, T. and Murphy, M. (1985). The onset of demographic transition. Population and Development Review, 11, 299–340.
Elo, I.T. and Preston, S.H. (1994). Examining African-American mortality from incomplete data. Demography, 31, 427–58.
Eurostat (1998). Demographic statistics 1981. Eurostat, Luxembourg.
Feeney, G. and Yu, J. (1987). Period parity progression measures of fertility in China. Population Studies, 41, 77–102.
Fox, A.J. and Goldblatt, P.O. (1982). Longitudinal study: socio-demographic mortality differentials. HMSO, London.
Furner, S.E., Maurer, J., and Rosenberg, H. (1993) Mortality. In Health data on older Americans United States, 1992 (ed. J.F. Van Nostrand, S.E. Furner, and R. Suzman), pp. 77–112. Vital Health Statistics 3(27). National Center for Health Statistics, US Department of Health and Human Services, Hyattsville, MD.
Garenne, M. and Fontaine, O. (1990). Assessing probable causes of death using a standardized questionnaire: a study in rural Senegal. In Measurement and analysis of mortality: new approaches (ed. J. Vallin, S. D’Souza, and A. Palloni), pp. 123–42. Clarendon Press, Oxford.
Garson, L.K. (1991). The centenarian question: old age mortality in the Soviet Union 1897–1970. Population Studies, 45, 265–78.
Grundy, E.M.D. (1996). Ageing in Europe. In Europe’s population in the 1990s (ed. D. Coleman). Oxford University Press.
Grundy, E., Murphy, M., and Shelton, N. (1999). Looking beyond the household: intergenerational perspectives on living kin and contacts with kin in Great Britain. Population Trends, 97, 19–27.
Haskey, J. (1988). Trends in marriage and divorce and cohort analyses of the proportion of marriages ending in divorce. Population Trends, 54, 34–7.
Haskey, J. (1990). The ethnic minority populations of Great Britain: estimates by ethnic group and country of birth. Population Trends, 60, 35–8.
Haskey, J. (1994). Estimated numbers of one parent families and their prevalence in Great Britain in 1991. Population Trends, 78, 5–19.
Haskey, J. (1996). The proportion of married couples who divorce: past patterns and current prospects. Population Trends, 83, 25–36.
Heady, P., Smith, S., and Avery, V. (1994). 1991 census validation survey: coverage report. HMSO, London.
Hinde, A. (1998). Demographic methods. Edward Arnold, London.
Hobcraft, J. (1991). Child spacing and child mortality. Paper presented at the DHS World Conference, Washington DC, 5–7 August. Results summarized in DHS World Conference Executive Summary (December 1991, Ann K. Blanc). DHS, Columbia, MD.
Hobcraft, J., Menken, J., and Preston, S. (1982). Age, period and cohort effects in demography: a review. Population Index, 48, 4–43.
Inglehart, R. (1990). Culture shift in advanced industrial society. Princeton University Press.
Kestenbaum, B. (1992). A description of the extreme aged population based on improved Medicare enrolment data. Demography, 29, 411–26.
Kiernan, K. (1999). Cohabitation in western Europe. Population Trends, 96, 25–32.
Kinsella, K. and Gist, Y. (1998). Gender and aging—international brief. IB/98–2. US Bureau of the Census, Washington, DC.
Knodel, J. and van de Walle, E. (1979). Lessons from the past: policy implications of historical fertility studies. Population and Development Review, 5, 217–45.
Kono, S. (1994). Demography and population ageing in Japan. In Ageing in Japan. JARC, Tokyo.
Lotka, A. (1907). Relation between birth and death rates. Science, 26, 21–2.
Manton, K.G. (1982). Changing concepts of morbidity and mortality in the elderly population. Milbank Memorial Fund Quarterly, Health and Society, 60, 183–244.
Manton K.G. and Vaupel, J.W. (1995). Survival after the age of 80 in the United States, Sweden, France, England and Japan. New England Journal of Medicine, 333, 1232–5.
Markides, K.S. and Keith, V.M. (1995). Race, aging and health in the United States. Reviews in Clinical Gerontology, 5, 339–45.
Menken, J. (1989). Proximate determinants of fertility and mortality: a review of recent findings. In Demography as an interdiscipline (ed. J.H. Stycos). Transaction Publishers, New Brunswick, NJ.
Murphy, M. (1993). The contraceptive pill and women’s employment as factors in fertility change in Britain 1963–1980: a challenge to the conventional view. Population Studies, 47, 221–44.
Murphy, M.J. (1995). Methods for forecasting mortality and their performance. Reviews in Clinical Gerontology, 5, 217–27.
Myers, G. (1995). Comparative study of mortality trends among older persons in developed countries. In Health and mortality among elderly populations (ed. G. Casselli and A. Lopez). Clarendon Press, Oxford.
Newell, C. (1988). Methods and models in demography. Belhaven Press, London.
Notestein, F.W. (1945). Population: the long view. In Food for the world (ed. T.W. Schulz), pp. 36–57. University of Chicago Press.
Omran, A.R. (1971). The epidemiologic transition: a theory of the epidemiology of population change. Milbank Memorial Fund Quarterly, 49, 509–38.
ONS (Office for National Statistics) (1998). Mortality statistics—cause. Series DH2. HMSO, London.
ONS (Office of National Statistics) (1999a). Table 1.6 Population Trends, p. 97. HMSO, London.
ONS (Office of National Statistics) (1999b). Marriage and divorce statistics. Series FM2. HMSO, London.
ONS (Office of National Statistics) (1999c). Mortality statistics—general. Series DH1. HMSO, London.
OPCS (Office of Population Censuses and Surveys) (1981). Trends in respiratory mortality 1951–1971. Series DHI No.7. HMSO, London.
OPCS (Office of Population Censuses and Surveys) (1987). Birth statistics 1837–1983. Series FM1 No.13. HMSO, London.
OPCS (Office of Population Censuses and Surveys) (1992). 1990 mortality statistics. Series DH1. HMSO, London.
Pollard, A.H., Yusuf, F., and Pollard, G.N. (1990). Demographic techniques (3rd edn). Pergamon Press, Sydney.
Population Reference Bureau (1999). World Population Data Sheet. Population Reference Bureau, New York.
Prescott-Clarke, P. and Primatesta, P. (1997). Health survey for England 1995 Vol. 1: Findings. HMSO, London.
Pressat, R. and Wilson, C. (1985). The dictionary of demography. Basil Blackwell, Oxford.
Preston, S.H. (1976a). Mortality patterns in national populations. Academic Press, New York.
Preston, S.H. (1976b). The changing relation between mortality and level of economic development. Population Studies, 29, 231–48.
Preston, S.H., Himes, C.L., and Eggers, M. (1989). Demographic conditions responsible for population aging. Demography, 26, 691–704.
Robine, J.M., Blanchet, M., and Dowd, J.E. (ed.) (1992). Health expectancy: First Workshop of the Healthy Life Expectancy Network (REVES). HMSO, London.
Robine, J.M., Mormiche, P., and Sermet, C. (1998). Examination of the causes and of the mechanisms of the increase in disability-free life expectancy. Journal of Aging and Health, 10, 171–91.
Rogers, A., Rogers, R.G., and Belanger, A. (1990). Longer life but worse health? Measurement and dynamics. Gerontologist, 30, 640–9.
Roussel, L. (1994). Fertility and family. In European Population Conference Proceedings, Vol. I. UN Economic Commission for Europe/Council of Europe/UN Population Fund. United Nations, New York.
Ruzicka, L. and Kane, P. (1990). Health transition: the course of morbidity and mortality. In What we know about health transition: the cultural, social and behavioural determinants of health, Vol. 1 (ed. J. Caldwell et al.), pp. 1–26. Health Transition Centre, Australian National University, Canberra.
Schuman, J. (1999). The ethnic minority populations of Great Britain—latest estimates. Population Trends,96, 33–43.
Shapiro, J. (1994). The Russian mortality crisis and its causes. In Russian economic reform in jeopardy (ed. A. Aslund). Pinter, London.
Shaw, C. (1988). Components of growth in the ethnic minority population. Population Trends, 52, 26–30.
Shryock, H.S., Siegel, J.S., and Stockwell, E.G. (1976). The methods and materials of demography (abridged edition). Academic Press, New York.
Smith, D. and Keyfitz, N. (ed.) (1977). Mathematical demography: selected papers. Springer-Verlag, New York.
Snow R.W., Armstrong, J.R., Forest, D., et al. (1992). Childhood deaths in Africa: uses and limitations of verbal autopsies. Lancet, 340, 351–5.
Sundström, G. (1994). Care by families: an overview of trends. In Caring for frail elderly people. OECD, Paris.
Thatcher, A.R. (1992). Trends in numbers and mortality at high ages in England and Wales. Population Studies, 46, 411–26.
Thatcher, R. (1999). The demography of centenarians in England and Wales. Population Trends, 96, 5–12.
UN (United Nations) (1970). Methods of measuring internal migration. Population Studies, No. 47. UN, New York.
UN (United Nations) (1983). Indirect techniques for demographic estimation, Manual X. UN, New York.
UN (United Nations) (1991). World population prospects 1990. UN, New York.
UN (United Nations) (1993). World population prospects: the 1992 revision. UN, New York.
UN (United Nations) (1995). Population and development. Volume 1:Programme of action adopted at the International Conference on Population and Development, Cairo, 5–13 September 1994. UN, New York.
UN (United Nations) (1999). World population prospects: the 1998 revision. UN, New York.
UNECE (United Nations Economic Commission for Europe) (1994). International migration bulletin No. 5. UNECE, Geneva.
UNECE/UNFPA (United Nations Economic Commission for Europe and United Nations Population Fund) (1992). Changing population age structures: demographic and economic consequences and implications. UN, Geneva.
UNFPA (United Nations Population Fund) (1994). Population: putting people first. UN DPI/1519/SOC/CON. UN, New York.
UNFPA (United Nations Population Fund) (1998). The state of world population 1998. UNFPA, New York.
UNFPA (United Nations Population Fund) (1999). The state of world population 1999. UNFPA, New York.
United Nations Secretariat (1988). Sex differentials in life expectancy and mortality in developed countries: an analysis by age groups and causes of death from recent and historical data. Population Bulletin of the United Nations No. 25. UN, New York.
US Agency for International Development (1997). Children on the brink: strategies to support children isolated by HIV/AIDS. Executive report. USAID, Washington DC.
US Bureau of the Census (1992). An aging world II. International Population Reports, pp. 25, 92–3. US Government Printing Office, Washington, DC.
van da Kaa, D.J. (1987). Europe’s second demographic transition. Population Bulletin, 42.
van de Walle, F. (1986). Infant mortality and the European demographic transition. In The decline of fertility in Europe (ed. A.J. Coale and S.C. Watkins), pp. 201–33. Princeton University Press.
Velkoff, V.A. and Kinsella, K. (1993). Ageing in Eastern Europe and the former Soviet Union. US Bureau of the Census, Center for International Research. US Government Printing Office, Washington, DC.
Virganskaya, I.M. and Dmitriev, V.I. (1992). Some problems of medicodemographic development in the former USSR. World Health Statistics Quarterly, 46, 4–13.
Waldron, I. (1985). What do we know about causes of sex differences in mortality? A review of the literature. Population Bulletin of the United Nations, 18.
Werner, B. (1987). Fertility statistics from birth registrations in England and Wales 1837–1987. Population Trends, 48, 4–10.
WHO (World Health Organization) (1992a). Global health situation and projections, estimates. Division of Epidemiological Surveillance and Health Situation and Trend Assessment, WHO, Geneva.
WHO (World Health Organization) (1992b). World health statistics annual. WHO, Geneva.
WHO (World Health Organization) (1998). World health statistics annual. WHO, Geneva.
WHO (World Health Organization) (1999). The world health report 1999. WHO, Geneva.
Wilkinson, R.G. (1996). Unhealthy societies: the afflications of inequality. Routledge, London.
Woods, R.I., Watterson, P.A., and Woodward, J.H. (1989). The causes of rapid infant mortality decline in England and Wales, 1861–1921 (Part II). Population Studies, 43, 113–32.
World Bank (1993). World development report 1993: investing in health, world development indicators. Oxford University Press, New York.
World Bank (1999). World development indicators 1999. World Bank, Washington, DC.
Wrigley, E.A. and Schofield, R.S. (1981). The population history of England 1541–1871: a reconstruction. Edward Arnold, London.
Zeng, Y., Tu, P., Gu, B., Xu, Y., Li, B., and Li, Y. (1993). Causes and implications of the recent increase in the reported sex ratio at birth in China. Population and Development Review, 19, 283–302.


2 comments on “7.2 Demography and public health


  2. knowledge is power,i want to have a good understanding

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: