December 7, 1995
phone: 510-642-4535
fax: 510-663-8558
email: rlee@demog.berkeley.edu
This paper was prepared for presentation at the Celebration of the 50th Anniversary of INED, held in Paris on October 25-27, 1995. I am grateful to Joshua Goldstein and John Wilmoth for comments on an earlier draft.
It is a great pleasure to return to INED on the happy occasion of its 50th birthday, just 25 years after I arrived here as a post doctoral fellow in the fall of 1970. I have been asked to prepare a paper on the history of the field of demography in the United States over the past 50 years, since World War II. The United States is a nation of immigrants, including many immigrant demographers, which makes it difficult to say which work should be attributed to U.S. demography. I will simply attribute demographic work done in the US to US demography, rather than reminding the reader that Keyfitz, for example, is actually Canadian. The history I give here is impressionistic, drawn from my memory rather than a systematic rereading of the literature. I began my study of demography as a graduate student in 1965, thirty years ago. Inevitably, I know much more about demography in the U.S. during the past thirty years than in the twenty years before, and my account reflects that bias. Also, my training and work have been in demography and economics, and so my account will give more emphasis to economic demography than it is due. Because space is limited, there are many topics I have largely ignored, such as migration.
The remainder of this paper consists of three parts. In the first part, I attempt to analyze the structural features and forces which have given US demography its distinctive flavor. In the second, I describe how substantive research and theories have evolved in a number of areas of research. In the third, I describe how formal demography and demographic methods have evolved.
U.S. Demography has some distinctive features which arise, in part, from certain characteristics of the educational system and from the place of demography in the universities.
Secondary and university students in the U.S. educational system specialize far less than in many other countries, and certainly much less so than in Europe. In particular, a certain level of competence in mathematics and science is required of all students no matter what their disciplinary focus, and university students in the social sciences are usually required to learn a nontrivial amount of statistics. Consequently, students studying demography are reasonably well prepared to handle statistical and mathematical material at least at a basic level. Some students will have a good grounding both in the theory and substance of a social science field and also have strong training in mathematics. I believe that this is less likely to happen in some other countries, where demography students are often recruited from the ranks of students who have earlier chosen the humanities fork rather than the scientific one when the educational road split early in their lives. Some other future demographers choose the science fork in the road and subsequently receive superb training in science and mathematics, but when they come to demography later, they typically have less training in social science. In the U.S. the road does not split, which has contributed to the integration of substantive and theoretical demography with mathematical and statistical modeling that is one of the features of U.S. demography. Perhaps the social theory is less profound, and the mathematics is more basic, but the two are joined.
Another structural feature of academic demography in the U.S. is that it rarely exists as a separate, stand-alone department in our universities. Instead, the teaching of demography is typically incorporated within other departments, usually sociology, but sometimes economics, and occasionally elsewhere. (There are a few exceptions, including my own program at Berkeley, but even these exceptions turn out to be highly interdisciplinary on closer examination.) This embedding of demography within social science programs has important implications for both students and faculty. It means that there is less emphasis on descriptive demography, formal demography, and demographic measurement than elsewhere, and more emphasis on social and economic theories of demographic behavior, and analysis of the social and economic consequences of demographic structure and change. This is true of the content of training programs in the U.S., and is in fact a deterrent to many foreign students who would like to study demography the U.S., but who do not want to spend 85 percent of their study time learning general economics or sociology.
The research programs of the demographers in these universities are also profoundly affected by the embedding of demography within social science departments, because it means that professional reward and advancement depends on doing work that is intelligible and interesting to non-demographer colleagues, and that is integrated with the broader theories and concerns of social science. Within sociology, demography is often viewed as a core field which commands respect. Demography, in such departments, may be quite broadly interpreted to encompass most forms of quantitative sociology, from studies of occupational mobility through sociology of the family or sociology of racial and ethnic groups, with little or no emphasis on fertility, mortality or migration, and little training in formal demography. Within economics the situation is very different. Economic demography is barely recognized as a field within economics, and when it is taught at all, it is often as an area of labor economics. Indeed, to my horror, a recent survey article on economic demography characterized it as a sub-field of labor economics! Yet it is undeniably true that not only the field of economic demography, but also social demography in general, has been energized by contributions from neoclassical labor economists under the intellectual leadership of Gary Becker (1981), and his many followers.
Perhaps because U.S. students and academics do not specialize at an early age, as I discussed earlier, social science research in the U.S. is more highly quantified than elsewhere. In economics, this has meant that economic theory has been highly mathematical. But it has also meant that in both sociology and economics, statistical methods have been very important for empirical research. Indeed the hallmark of social science in the U.S. is the linking of statistical analysis and testing to the mathematical expression of social theory. In the best research, these links are tight; but in any event, this linking is held up as an ideal to which researchers and students aspire. Demographers also strive for this ideal, and this striving has given demography in the U.S. a distinctive character over the post-W.W.II decades, although in recent years this feature of U.S. demography has become far less unusual.
These structural aspects of demography within the U.S. educational system have influenced the character of U.S. demography, as I have just argued. But while they have given it form, there are other forces which have given it direction and shaped it. To these I will now turn.
Most early demographic research was based on official sources of data gathered by the states and the federal government, and particularly census and vital registration data. Much of a demographer's work consisted in assessing the quality of these data and deriving sophisticated demographic measures from them. Sophisticated approaches for the analysis of aggregate data were developed, such as parity specific analysis of fertility which was developed in France by Louis Henry, but also at about the same time in the U.S. by Norman Ryder, who additionally invented demographic translation, a method for relating changes in cohort fertility to changes in period fertility. Such methods were for use with official national statistics.
But over the past fifty years, individual level data have become increasingly available. One important part of this story is the development of the national fertility survey, due in the U.S. to Whelpton, Freedman, Ryder and Westoff (see Freedman et al, 1959, and Ryder and Westoff, 1971), starting in the mid-1950s, and building on an earlier survey carried out in a single city, Indianapolis, in the 1940s. These surveys were originally carried out with the goal of improving population projections by incorporating more detailed information about the family building plans of couples, and trends in their use of contraception to achieve these plans by avoiding unwanted births. The national fertility surveys, which were initially carried out by private researchers working in universities, were eventually taken over by the federal government--first in 1967, and then on a regular basis starting in the early 1970s, albeit with less socio-economic detail. The national fertility surveys also led, towards the end of the 1970s, to the mammoth World Fertility Survey, followed by the Demographic and Health Surveys. These provided a mass of comparable international data on fertility and child mortality, greatly improving our knowledge of the levels and trends in fertility and mortality world wide, but also providing the basis for an analysis of individual fertility and mortality.
Fertility surveys were, of course, not the only source of individual level data for demographers. There was a growing number of local and national surveys on many aspects of social and economic behavior, carried out both by private researchers and also increasingly by the federal government. Included among these were surveys on traditional demographic topics such as health and mortality, marriage and divorce, and migration. Even the old census and vital registration data, which after all originated at the individual level, were made available in new ways which avoided problems with protection of confidentiality, as in the so-called PUMS individual level data sets extracted from the census with individual identifiers deleted.
The availability of these new data sets, many of them gathered by demographers themselves, made it possible to probe the motivations of the actors in ways that had been impossible with the older data sets from official sources. Efforts of demographers were redirected from the description of demographic behavior to the explanation of demographic behavior, and the probing of causes for demographic outcomes. Along with this redirection from description to causal analysis went a redirection from analysis of aggregate demographic phenomena to the analysis of private individual demographic events.
But without the parallel development of computer technology, the shift from analysis of aggregate to individual data could not have occurred as it did. Although the development of fertility surveys was well underway before the use of computers became widespread in social science, it would have been impossible for the statistical analysis of large survey data sets to have proceeded as it did without the extensive use of computers. The use of computers spread rapidly in the social sciences in the U.S. after the mid-1960s, and took another quantum leap with the spread of personal computers from the early 1980s. Computers made possible the orderly storage and retrieval of large and complex data sets; they made elaborate statistical analysis routine; and they permitted various kinds of simulation and calculation. The increasing use of multivariate statistical techniques, and causal modeling, particularly with survey data, is surely in part due to the availability of computers. The development of complicated demographic simulation models which combine individual randomness with structural aspects of the reproductive cycle or of households and families would have been quite impossible without the computer. Macro-demographic simulations could have been done with old fashioned calculators, but elaborate macro analyses, such as computable general equilibrium, or CGE, models could also not have been solved without computers.
I believe that the penetration of computer technology in social science occurred considerably earlier in the U.S. than elsewhere, and became widespread earlier, and that this has been another distinguishing feature of U.S. demography.
Of course, the direction of demographic research did not develop in a vacuum; it was influenced by contemporary perceptions of the leading demographic issues. These perceptions were often reflected in federal policy, and by the 1960s, such national policy concerns were expressed through generous programs of research grants administered through the National Institutes of Health. (As an aside, I must mention that again and again, demographic research in the U.S. has benefited from a historical accident which placed its funding under the umbrella of the research funding for health rather than research funding for social science research.)
Through the mechanism of federal funding, policy concerns have indeed had a powerful influence on the direction of demographic research in the U.S. Such topics as the causes and consequences of adolescent fertility, or the efficacy of international family planning programs, have certainly received far more attention in the past three decades than would otherwise have been the case. And the shift in the attention of many leading U.S. demographers to topics in the demography of aging surely is due in part to the vigorous federal funding program in this area.
Yet the imperatives of federal research funding are certainly not decisive. There has been a persistent and urgent demand for research on the economic consequences of rapid population growth in the Third World, and abundant funding for such research, yet very little research has been forthcoming. The reason is that the greatest professional prestige derives from using the advanced econometric techniques on individual level data to probe micro-economic motivations, while the study of consequences of rapid population growth involves old fashioned statistical methods applied to outmoded aggregate data sets.
Furthermore, federal support for demographic research has shown a commendable interest in basic research of scientific value, as judged by the system of peer review; consequently research in such areas as historical demography has also been funded, despite less immediate application to policy issues.
Demography has developed in part following the dictates of its own internal dynamics. One question leads to another; one new method or result points to new applications in other related areas. Stable population theory continues to be mined for new insights.
But in addition to its internal dynamic, demography has been strongly influenced by infusions from sociology and economics. These influences will be discussed later.
Demography in the U.S. has certainly had some distinctive features over the past 50 years. It has been more interdisciplinary, more quantitative, more concerned with behavioral theories from other disciplines, more interested in formal causal models, and less interested in descriptive demography, formal demography, and demographic methods. There are, of course, very significant exceptions to this sketch--much of the influential work at Princeton defies this generalization--yet I think that it is broadly true. In this section I have sought to understand in a general way the forces that have made U.S. demography different from demography elsewhere, and given it shape and direction.
Now it is time to turn to specifics. I will discuss the substance of demographic research in five areas: consequences of population change for economic development in the Third World; analysis of fertility change in the Third World, consequences of demographic change for the U.S.; attempts to explain fertility and other demographic behavior in the U.S.; and population aging.
Soon after the Second World War ended, demographers noted that mortality was beginning to decline rapidly in many Third World populations, and realized that in the expected absence of fertility declines, this would lead to rapid and accelerating population growth. They were concerned that this would render economic growth impossible, and that the absence of economic progress would then perpetuate the conditions which caused high birth rates and rapid population increase. Eventually, a few economic demographers began to address these issues. Classical economists had stressed the centrally important consequences of population pressure against a fixed supply of natural resources, particularly land. However, the apparently limitless ability of technical progress and capital accumulation to expand output following the industrial revolution, and indeed until the 1970s, led economists to view natural resource constraints as unimportant. Instead, investment and capital accumulation, and the generation and transfer of technology, were seen as the keys to economic development. But capital accumulation, too, was believed to be subject to population pressure.
A seminal study by Coale and Hoover (1958), taking India as a case study, stressed the interaction of population growth and capital accumulation, and concluded that high fertility would seriously retard per capita income growth in Third World countries. This study provided the rationale for a number of large scale efforts to reduce fertility in developing countries, including the US program of foreign aid for family planning programs. It also spawned numerous derivative studies which confirmed the results using more elaborate simulation models. Elements of the reasoning are sketched below.
Rapid population growth leads, perhaps with some lag, to rapid growth in the labor force, reinforced in cities by rural-urban migration. If jobs don't increase at the same pace, unemployment grows. Whether or not unemployment is avoided, expansion of the social infrastructure and the capital stock is costly. When population grows more rapidly, a greater proportion of current output must be set aside to create housing, tools, machinery, and schools for new members of the population and labor force. Many forms of investment must increase at a time when higher child dependency may tend to reduce domestic savings rates. If this additional investment does not take place, then new generations will be less well equipped than older ones.
The Coale-Hoover study also stressed the costly effects of the age distribution associated with high fertility. In a rapidly growing population, the more recently born generations are larger than older ones, so the population--and the average family--has more children per person of working age. This so-called dependency effect is not in itself a very important factor, but it is thought to have two important indirect consequences. First, it is thought to reduce the savings rate, and thereby reduce the resources available for investment. Evidence on this point is mixed. Second, it is thought to reduce the resources that both families and governments are able to invest in the health and education of each child, and thereby to reduce the labor productivity in the next generation.
The Coale-Hoover study led to many second-generation studies both in the U.S. and abroad, which made the economic model at its core less rigid, and added sectoral detail. The basic structure of the models remained very similar, however.
The Coale-Hoover study and its descendants were challenged on several grounds during the 1960s and 1970s. First, empirical studies provided only mixed support for the view that high fertility reduced savings. Second, the role of capital itself in economic growth was questioned. Empirical studies attributed more importance to other factors such as education and technology. Third, economists pointed to a number of positive effects of population growth. These included stimulating investment demand, breaking down traditional constraints, facilitating collective investments, spurring technological progress, and leading to harder work. Two economists, Boserup (1981) (a Dane) and Simon (1981), argued particularly forcefully for these beneficial influences of population increase. Boserup also stressed that a larger population can more easily bear the costs of providing certain kinds of social infrastructure--transportation, communications, water supply, government, research--for which the need increases less than proportionately with population.
By the 1980s, policy makers were in a state of confusion. Was population growth good? Was it bad? Did it matter at all? Would expenditures on family planning programs lead to more rapid economic development? This uncertainty led to systematic attempts to assess our knowledge, including major projects by the World Bank (1984) and the United States National Academy of Sciences (National Research Council, 1986). These assessments in the mid-1980s revealed a surprising degree of agreement among economists. While few economists accepted the view that population growth was good for development, the consensus was that population growth mattered less than had been thought. Prior work, it was believed, had held too much fixed, failing to appreciate the flexibility of the economic system. Instead, scarcities arising from population growth would lead to price changes, which would in turn generate incentives for corrective action. Responses would include not only minor adjustments, but also larger changes such as the pace and direction of technological progress, and even institutional change. Furthermore, neither natural resources nor physical capital were viewed as critical constraints on development. Instead, human capital was the key. Higher fertility was thought to reduce the parental resources devoted to each child, and to hinder national efforts to increase educational enrollments and improve the quality of formal education.
While economist demographers were concluding that population growth was relatively unimportant, ecologists and environmentalists like Ehrlich (1968) and Hardin (1968) were sounding the population alarm. They pointed out that the biosphere provided essential, although uncounted, inputs to economic activity, and warned that its limits and fragility placed bounds on sustainable levels of economic activity. Similar views were incorporated in a systems analysis model described in the so-called Club of Rome Report (Meadows et al, 1972). This report attracted world wide attention in the early 1970s, with its conclusion that global overshoot and collapse were imminent and inevitable, unless very fundamental changes were made. Catastrophe was inherent in the structure of the socio-economic and ecological relationships, so that detailed testing and estimation of the component parts of the model was unnecessary.
Social scientists soon rejected the Club of Rome approach in favor of careful analysis of each relationship. In the 1980s, however, an onslaught of environmental problems added new urgency to the ecologists' position. Hot summers, drought, acid rain, polluted waters, famine, and holes in the ozone layer seemed to confirm their predictions, leading to heightened concern about population growth.
While some of the ecologists' warnings appeared, on examination, to be wrong or exaggerated, for example, fears of running out of non-renewable natural resources, some of the most important concerns appeared more convincing, particularly those for renewable (but destroyable) resources--air, water, fisheries, land, forest cover, ozone layer, and populations of various species. Most of these are common property resources which lie outside the market economy, and indeed may be used freely, so that price signals and financial incentives provide no guide to their optimal use.
Worries about population growth have now come full circle: from the Classical concern for limited land, to Coale-Hoover's emphasis on physical capital, to more recent emphasis on human capital and the ameliorative influence of competitive markets, and back once again to the natural constraints. This time, however, the concern is for renewable natural resources, most of which fall outside the market. For some, the urgency of population control on ecological grounds is obvious. Others remain skeptical. Research on links between population growth and the environment is underway, but is severely hampered by the difficulties in designing research projects which can shed light on population's specific role in environmental problems. Targetted funding is available to support work in this area, but very little is being done.
Worries about the consequences of rapid population growth after W.W.II led to concern about the prospects for fertility decline in the Third World. In the 1950s, Notestein (1953) built on earlier work by others in France and the US to develop his transition theory, which stressed the influence of industrialization and urbanization on the desired number of births. Major influences in his theory included declining mortality, rising costs of children, rising individualism, and increasing participation by women in the formal labor market. These ideas are still at the core of most efforts to explain Third World fertility forty years later, although the models have become much more formal.
In the 1960s, Leibenstein (1957) advanced an economic theory of fertility in which children were desired for three sources of utility: direct enjoyment of them, their contributions to household production, and support in old age. All but the first of these decline sharply with economic development, leaving only the psychic rewards to child rearing.
These theories paid scant attention to the importance of the technical means of regulating fertility through contraception; since Europe and the US had achieved fertility declines to replacement levels without modern contraceptives, it was felt that only the demand for low fertility really mattered.
In the 1960s, however, there was growing interest in the potential role of organized family planning programs, and Freedman (see Takeshita and Freedman, 1969) began a major prospective study in Taiwan of the effects such programs might have. Initially such work was viewed with great skepticism by other academic demographers, who often thought it superficial and misguided. The European and US fertility declines, predating modern contraceptive methods, seemed to confirm this view. But evidence began to mount from Freedman's and other studies that contraceptive methods, knowledge and attitudes did indeed matter for fertility change.
Easterlin's (1978a) framework for fertility analysis, developed in the 1970s, provided a simple and suggestive way of integrating the role of family planning programs and contraceptive technique with both bio-cultural and socio-economic approaches to fertility theory. Easterlin viewed fertility as resulting from the formally specified interaction of natural fertility, the desired level of fertility, and the extent to which regulation was used to close the gap between the two, which depended on its broadly construed costliness. Variation in any of these three components--demand, supply, and costs of regulation--could change fertility outcomes. Transition theory mainly concerned the demand for children, while ignoring supply and costs of regulation. This was a very useful conceptual framework for relating progress that had been made in various areas of research on fertility, from the biological and cultural to the economic, and it contained important roles for family planning programs as well as socioeconomic change and mortality decline. However, the empirical implementation turned out to be difficult. In the 1980s, Rosenzweig and Schultz (1985) contributed an economic analysis of fertility which shared some of the features of the Easterlin framework, but which could be implemented econometrically.
Historical demography has never occupied a central position in US demography, although there has been some important work on New England and on the frontier (Bean et al, 1990). In the late 1960s, however, a major study of the fertility transition in the provinces of Europe was launched under the leadership of Coale (1969). The work was completed in the 80s, and it found very little support for the ideas in classical demographic transition theory: socio-economic variables and mortality change appeared to explain very little of the timing or pace of the transition. Instead, it appeared that the transition occurred in areas at very different levels of economic development, but that differences in fertility behavior seemed to be bounded by the borders of linguistically similar areas. This suggested that cultural factors, coincident with linguistic regions, were the dominant influence on fertility behavior. It was suggested that the knowledge of contraceptive method, and favorable attitudes toward fertility limitation, diffused geographically within culturally and linguistically homogeneous areas. Thus, in Easterlin's terms, the cost of regulation was viewed as the most important factor in fertility decline, rather than the demand for children, that is the number of children desired by the parents, which had been stressed by classical transition theory. Although many demographers questioned this conclusion, and not all research supported it, it was widely accepted. The conclusion loaned support to the idea that fertility decline could be hastened by family planning programs and other forms of government action which would hasten the diffusion of knowledge of contraceptive methods and reduce social barriers to their acceptability.
Controversy about the causes of fertility decline and the role of family planning programs versus economic development remains as active as ever, despite 50 years of research. The Cairo conference appeared to be a triumph of the view that economic development, particularly in so far as it improved the status of women, was the key to fertility. However, the triumph of this view at Cairo was more a political victory than a scientific one, and the questions remain open. A particularly vigorous challenge to family planning was recently launched by Pritchett (1994), generating intense controversy.
Economic consequences of population change in the developed countries has attracted less attention and research than the corresponding question for the Third World. There has been intermittent interest in the Keynesian theory that slowing population growth caused economic depression, but that has largely been rejected. Several other topics have drawn more research on the part of demographers.
Easterlin's (1978b) research stimulated many studies of the effects of the baby boom and baby bust on the earnings of the affected generations, with the conclusion that larger generations do suffer lower wages, higher unemployment, and slower advancement.
Only in the last five or ten years has the U.S. become seriously concerned about the economic consequences of immigration, long after Europeans began studying these matters seriously. Now there have been many studies of the labor market impacts and public costs and benefits of immigrants, as well as studies of the process of assimilation. Economists have been more active in these studies than have demographers.
Population aging and its social and economic consequences has attracted intense research interest in the past decade, fueled by ample federal funding. Studies have considered retirement incentives, saving behavior and wealth, health and disability, the public and private pension systems, intergenerational equity, and so on. Demographers have been active in these areas as well as economists.
During the 1960s, 70s and 80s, fertility in the U.S. was the hottest topic for U.S. demographers. Fertility as a topic combined the deepest social and economic theory, the most pressing policy issues, and the newest techniques and data sources. Fertility as a field was enriched by its relation to such additional factors as marriage, women's labor supply and time use, investment in children's education and upbringing, husband-wife interactions, religion, race, income and wealth, adolescent childbearing, access to contraceptives, and so on.
Easterlin (1978b) offered an exciting theory to explain the baby boom and baby bust, as peak and trough of a self-generating economic demographic cycle in which small generations were economically favored and had high fertility, while the large generations they spawned were economically disadvantaged and had low fertility. Much research was conducted on these ideas, and some parts have been supported; however, the grand theory as a whole now has few believers.
In 1960, Becker published a seminal new economic theory of fertility. After a ten year hiatus, there was a burst of fertility research by labor economists in the 1970s and 80s, under his leadership, with the most widely used synthetic theory due to Willis (1973). Two ideas were emphasized: on the one hand, the influence of the female potential wage rate on her value of time and therefore on the opportunity cost of children and her fertility; on the other hand, the tradeoff between the number of children and the amount that was invested in each one, or "child quality". The influence of this school of fertility theory spread to social demographers, as well.
The national fertility surveys, discussed earlier, and their analysis, were an important part of this surge of fertility research during these decades. Many studies probed the self-expressed motivation and plans of parents, their knowledge and use of contraception, the failure rates of different contraceptives, the effects of grandparental background, the roles of religion, education and income, and so on.
When all is said and done, however, after these three decades of intense research from the 60s through the 80s, we have little hard knowledge of the fundamental causes of fertility behavior. We have no widely accepted explanation of either the baby boom or the baby bust, and, in my view, we have no sound basis for predicting trends in fertility even a few years ahead. We do have a number of very plausible theories, and a great deal of expertise in fertility analysis and survey instrument design.
Over the past fifty years the use of statistical methods in demography has become much more widespread, and new methods have been introduced. Methods now range from ordinary regression analysis through special methods for discrete dependent variables including the analysis of hazards, as well as simultaneous equation techniques and methods for estimating dynamic systems. A great deal of attention has been given over the past twenty years to the problem of unobserved heterogeneity. Despite the importance of this penetration of demography by statistical methods, I will not discuss them except in so far as they are particularly germane to developments in formal demography.
Until the last decade, the central preoccupation of demographers has been fertility analysis, so it is natural to begin with work in this area. Often, the goal has been to create a new measure of fertility which would reveal greater stability in behavior amidst major variations in the conventional measures of fertility. Thus the period Total Fertility Rate removes age distribution variations as a source of change in the Crude Birth Rate; one might expect, then, that the TFR would vary less than the CBR--but in the U.S., the opposite has been the case.
Whelpton pioneered in the development of cohort fertility analysis, and incorporated this approach in his population projections as well. Ryder (1951), the leading US fertility analyst in the post W.W.II period, also stressed cohort fertility measures. But he realized that changes in the timing of fertility by cohorts could have dramatic effects on period fertility, and in the 1950s (Ryder, 1960) he developed analytic machinery which he called "demographic translation" to characterize formally the relation of period to cohort changes in fertility. This is perhaps the most striking analytic contribution to fertility analysis over this period, and the ideas were further developed by Foster (1990). Lee (1980) developed a different approach, in which period fertility emerges from the attempts of cohorts to achieve completed fertility goals that are changing over time. Paralleling the work of Henry in France, Ryder (1951) independently developed parity specific fertility measures. Ryder (1978) also created an accounting framework that related aggregate fertility to the concepts measured by the new U.S. fertility surveys--such things as contraceptive failure rates, desired birth intervals, desired completed number of children, and so on, so that he could derive the implications for aggregate fertility of changes in these measures. This innovative accounting framework should have established a whole new research line in US demography, but in fact it has gone unnoticed.
A little bit later, in the 1960s and 1970s, Coale and his collaborators at Princeton, particularly Trussell (Coale and Trussell, 1974), developed several new methods and models for fertility analysis, all of which have seen wide use. The Princeton Fertility Indices, Ig, Im, If, are indirectly standardized measures of fertility based on Hutterite fertility age profiles, for use when one has numbers of births by marital status, and the age distribution of women by marital status, but no data on births by age (Coale, 1969). The measures were developed for use in historical studies of European fertility at the provincial level during the late 19th and early 20th centuries. More innovative, analytically focused, and controversial were the so called "big M--little m" measures of fertility. Big M indicates the level of the underlying natural fertility in a population, while little m measures the extent of fertility control. The strategy is based on Henry's observations about natural fertility, and the concentration of fertility control in the later years of childbearing. These measures have been used extensively to infer the extent of fertility regulation in the absence of direct evidence. Sometimes too much was asked of the methods, and too much read into the measures, and they have come under heavy criticism in recent years, by Trussell among others. Coale and Trussell (1974) also teamed up to construct a much used set of model fertility schedules. All in all, the Coale-Trussell contributions to fertility analysis represent the high point of that particular approach based on model-based curve fitting motivated by demographic theory.
Work in France by Henry on modeling the reproductive cycle, in conjunction with his work on historical demography and natural fertility, led to important work in the U.S. on biometric models of the reproductive cycle, and associated statistical inference. The key work was by Sheps and Menken (1973), but there were other important contributions as well. After a decade of refinement of these models and estimates in the 1970s, Bongaarts (1978) developed his greatly simplified fertility indices which provided a decomposition of the total fertility rate into the more important contributions from the proximate determinants of fertility: marriage, contraception, abortion, lactation. His indices were relatively easily estimated from data available from fertility surveys, and they quickly came into wide use and remain widely used today.
In the 1960s and 70s there was controversy and uncertainty over whether fertility had begun to decline in parts of the Third World with flawed and incomplete data. Brass, a Scottish demographer, pioneered in the development of methods for evaluating the available data and deriving estimates from them. His seminal work stimulated a great deal of research in the U.S. extending his approach, by Coale, Preston, Trussell, Goldman and Hill (an Englishman), culminating in Manual X of the United Nations (United Nations, 1983). The Brass methods drew on the full repertory of formal demography, from stable and quasi-stable population models (with pioneering work by Bourgeois-Pichat in France), through model schedules for fertility, mortality and nuptiality, to novel methods for exploiting census and survey data on children ever born and surviving. Preston and Coale (1982) developed the so-called "variable r" approach for analyzing unstable populations, an approach with important applications to estimation with bad data. After the mid-80s, however, as the methods were more fully developed and as better data became available from the World Fertility Survey and DHS, work in this area tapered off and interest in it waned. An unfortunate side effect has been a diminution of interest and training in formal demography in general in the U.S.
The life table remained the principal analytic tool for studies of mortality throughout most of this period. Model life tables were a very useful innovation, developed first at the United Nations; however, the Coale-Demeny (1966) system, with its associated stable populations, developed at Princeton in the 1960s, has seen very wide use and acceptance in the U.S. and abroad.
The multi-state life table, deriving from work by Rogers on interregional demography, has been a useful generalization of the basic model. It has been used to study many interrelated demographic processes, such as healthiness, sickness, disability, and death: in each case, rather than one probability of transition from life to death, there are multiple probabilities of transition. Expected years of life in each state can be calculated.
With the conventional life table, if we wish to examine the influence of some characteristic such as marital status or education on mortality, we have to calculate a separate life table for a sub-population with each characteristic. Not only is this tedious, but worse, one quickly runs out of data when trying to consider the joint effect of two or more characteristics. Hazard models, imported from the field of statistics in the 1970s, permit multivariate models to be estimated for the force of mortality or for other transitions, which is a considerable advance. They are used for the analysis of individual level outcome data, and therefore fit well with the increasing availability of data of this kind. They provide an efficient means for dealing with differing lengths of exposure and problems of selection and truncation. The shape of the hazard by duration can either be left completely free to be determined by the data, or functional forms can be imposed on it. Of course, these models bring their own sets of simplifying assumptions, such as the widely used proportionality of response, which are not always appropriate and are sometimes uncritically accepted. Hazard models, also known as event history analysis, are now widely used for all kinds of demographic analysis.
Hazard model analysis focused attention on the shape of the hazard by duration; for example, how did the probability of conception vary with delay since the previous birth? Or how did the risk of mortality vary with age? During the 1970s, some demographers (Vaupel, Manton, Trussell and Heckman, for example; see Vaupel et al, 1979) began to realize that the duration dependence of a hazard might reflect a true change in risks as a function of duration or age, but it might also reflect selection in a heterogeneous population. If the risk of conception begins to fall a few months after birth, is that a physiological effect, or is it because the most fecund women are selected out of the observed population at higher durations due to pregnancy? The problem arises when the population is heterogeneous in ways that the analyst cannot directly observe, and therefore is called the problem of unobserved heterogeneity. This phenomenon is pervasive in demography, and it leads to many other problems besides that of estimation of true duration dependence. Demographers, statisticians and econometricians have worked on methods to deal with it in various contexts, but most of these involve strong assumptions.
Demographers increasingly wished to study and understand mortality at a deeper level then description. Preston (1980), in a series of widely influential studies, examined changes over the course of the mortality transition in the cause-structure of disease, and also showed that measured aspects of economic development accounted for less than half of the gains in life expectancy in the decades around the middle of this century. Following on the success of Easterlin's (1978a) and Bongaarts' (1978) frameworks for analyzing fertility change, mortality specialists sought to develop comparable frameworks for mortality and morbidity; the most widely used of these is Mosely and Chen's (1984), but none achieved the clarity or widespread acceptance of the fertility frameworks.
Manton (Woodbury and Manton, 1977) and his collaborators took another approach. In the 1980s, they put forward a stochastic model in which mortality was a function of risk factors, and the risk factors evolved over time according to a specified set of interlinked quadratic equations, which also depended on age. These models could, in principle, be estimated from suitable micro-data and then, because of their recursive nature, the fitted models could be used to predict future morbidity and mortality. The models took demographers several steps toward the physiological processes governing senescence, disease, disability and death, and also provided estimates of the ways in which chosen behaviors influenced that evolution.
During the last decade, a new field has developed at the intersection of demography and biology. Like Manton's work, it seeks to understand the biological states and processes underlying health and mortality. A central question is whether there are biological limits to life expectancy, limits sometimes argued to be in the mid-80s for sexes combined, not so far from levels already reached in some sub-populations. A related question is the relative importance of genetic variation, environmental variation, and behavior in determining health and mortality. Vaupel and Carey (Carey et al, 1992) have addressed these questions by large scale studies of non-human populations of fruit flies, finding that death rates level off or decline in extreme old age, contrary to Gompertz predictions. Other studies examine the mortality of twins, or analyze the influence of specific genes on longevity. Theoretical studies consider the implications of evolutionary theory for human longevity.
Just as the Manton approach modeled the way in which risk factors and disease processes influenced morbidity and mortality, so the micro-economic approach to health and mortality modeled the way in which individuals made decisions about investing in their own health and that of their children. This approach reconciles socioeconomic theory with the intermediate or proximate variable type analysis which incorporates biological factors. In the 1970s Grossman (1972) developed the concept of the demand for health within the kind of human capital and home production framework developed by Becker for analyzing demographic behavior. Health is produced from inputs of time, goods and medical inputs according to a health production function, which also varies according to intrinsic characteristics of the individual which may or may not be observed (that is, there may be unobserved heterogeneity). Given the health production function, the demand for health by the individual then leads to a derived demand for the inputs to the health production function. The basic strategy is very similar to the one later proposed by Rosenzweig and Schultz for fertility analysis.
Until recently, there was a very wide consensus that teen childbearing had devastating effects on the futures of both the young mother and the child. Indeed, it was agreed that adolescent childbearing was a key point at which government policy might intervene to break the intergenerational cycle of poverty in the U.S. Abundant empirical studies, as well as common sense, appeared to provide strong support for this view. Controlling for many observed characteristics of the mothers and their parents, teen childbearing was associated with poor outcomes for both mothers and their children. Recent research, however, has questioned whether there might not be unobserved characteristics of the mothers which both make them more likely to give birth early, and also make them and their children less likely to be economically successful in later life. Studies have exploited natural experiments to control for this unobserved heterogeneity. Pairs of sisters, one giving birth as a teenager and the other in her twenties, effectively control for family background, and reveal that outcomes for the sisters and their children differ by only about a third as much as the earlier studies had suggested. Similar results are found when outcomes are compared for pregnant teenage girls who had miscarriages and those who did not, or those who had twin births and those who had singleton births. It has become clear that straightforward multivariate statistical analysis does not adequately control for the effects of unobserved heterogeneity in many contexts, sometimes leading to incorrect conclusions.
In the 1950s and 60s, demographers such as Goodman and Keyfitz (see Goodman et al, 1974) used stable population theory to calculate the distribution of certain kinds of kin relations in a population, work to which LeBras in France later contributed as well. Results were limited to biological relationships, such the proportion of 10 year old children who had a surviving grandmother. Paralleling these formal developments, sociologists such as Elder developed an approach emphasizing the ordering of demographic events over the life course, and the distribution of the life cycle years lived across differing family status categories. This approach has been very important in sociological studies of the family. In the 1980s, the multistate life table first developed by Rogers (1966) for migration analysis was applied to family relationships, which permitted social relationships to be studied as well as biological ones: what was the expected number of years of living with a divorced mother? But this approach was limited by the necessary assumption of fixed probabilities of transitions, and therefore could not be applied to a changing demographic regime. The household microsimulation approach, first developed in the US by Orcutt for economic simulation, and then taken up by demographers. Hammel and Wachter in the early 1970s (see Wachter et al, 1978) developed the first closed model which permitted a demographically flexible and socially rich study of changing kin relations and household membership in an unstable population. One particular problem in the formal demography of the family has attracted the attention of mathematical demographers in the US: the so-called two sex problem of how to reconcile the independently derived dynamics of the male and female sub-populations, when they must be closely linked by the sex ratio at birth. Empirical studies by Schoen have helped our understanding of the working of the marriage market, and after many formal efforts over the past forty years, Pollak's (1990) axiomatic approach offers fresh theoretical insights.
As described earlier, Whelpton pioneered in the cohort analysis of fertility in the U.S., and built his well-crafted but unsuccessful population projections on this approach. The early fertility surveys were conducted in part to provide information on fertility intentions which, it was hoped, would improve forecasts. But it was found that individual intentions were quite changeable over time, and did not coincide well with subsequent childbearing. Aggregated intentions did a better job of matching subsequent fertility for a few years, but even so, little was gained. Furthermore, some studies showed aggregate intentions to change much as aggregate fertility did, so that one would have to project changes in the intentions--and why not just project fertility directly? Easterlin's (1978b) work on demographic cycles led to some hope that a new baby boom could be predicted, but the predicted upturn did not occur. Nor did other social or economic models of fertility change hold promise for improving forecasting. Becker's New Home Economics model, for example, seems to suggest that fertility would continue to decline indefinitely as husbands' incomes and women's wages rose--due to the substitution of quality for quantity, and to the rising time cost of children.
Since the ability of demographers to predict the future appeared to be severely limited, others set about trying to quantify our ignorance, so that the unwary would be warned. In the early 1970s, Stoto (1983) and Keyfitz developed a method for deriving standard errors for forecasts based on the analysis of past forecasting performance for organizations like the United Nations and the U.S. Census Bureau. Other demographers sought to evaluate uncertainty by analyzing the process of demographic renewal. One approach was to view demographic rates as probabilities at the individual level, but it was quickly learned that this kind of uncertainty vanished in large populations. Another approach, pioneered by Sykes (1966) in the late 1960s, viewed the vital rates driving population change as stochastic processes and sought to quantify the inherent uncertainty in population growth as a consequence. In the 1970s and 1980s statistical time series methods were used to model fertility for forecasting purposes, but the main result was to depress demographers with the extent of our uncertainty and the rapidity with which it grew over the forecast horizon. In the 1980s and 1990s, some demographers have attempted to combine time series models of vital rates with a stochastic Leslie matrix to develop probabilistic forecasts of population itself and its age distributions. It is still too early to know whether this line of work will turn out to be of practical use, but as one who works in this area myself, I am optimistic (Lee and Tuljapurkar, 1994).
Population forecasting is the practical side of the study of population dynamics. But there have also been extensive theoretical studies, going back at least to the 1950s. Coale (1972) and Keyfitz (1968) carried out many illuminating analyses of the rate at which populations subject to fixed vital rates converged to the stable age distribution, and the fluctuations in births, growth rates and age distributions which occurred in the course of this transition. Keyfitz's concept of population momentum was a powerful focus for some research in this area. Lopez (1961) established the Weak Convergence Theorem, that different populations subject to identically varying vital rates would converge to the same age distribution. Work on convergence to the stable state continued in the 1970s, as Cohen (1976) established stochastic convergence and in the 1980s as Wachter (1984) deepened Coale's (1972) analysis of Lotka's roots. More recently, a series of studies by Schoen and Kim (1991) has developed methods for studying the dynamics of unstable populations undergoing transitions. Lee (1974) showed how the population renewal equation could be run in reverse to estimate vital rates from numbers of demographic events, a method which has proved useful in demographic history, and which has been further analyzed by Wachter (1986).
While the foregoing work viewed population as evolving according only to its own internal laws of motion, another formal approach viewed population renewal as subject to economic and environmental constraint. The theoretical roots of this approach are distinctly Malthusian, but with an emphasis on age distributional aspects of constraints as elaborated by Easterlin (1978). This work looked for an explanation of the baby boom and baby bust in the effects of a large cohort or number of workers on fertility. Independent studies by Lee (1974), Frauenthal (1975), and Samuelson (1976) in the early 1970s developed a variety of models, but their full dynamic implications were not appreciated until deeper mathematical studies by Tuljapurkar (1990) and Wachter (1991) examined the possibilities of limit cycles and chaos.
Work based on stable population theory has continued throughout this period, particularly by Keyfitz (1968, 1977) who published several landmark books in mathematical demography full of interesting new insights and applications, including studies of pensions, population momentum, demographic estimation based on age distributions, rates of advancement, and so on. In the 70s, Arthur and McNicoll (1978) pioneered in integrating stable population theory with models of economic growth, and showed how Samuelson's (1958) seminal work on intergenerational trade and transfers in stable populations could be placed in a more realistic demographic context. Subsequent work by Preston (1982), Willis (1988) and Lee (1994) has further developed the mathematical demography of intergenerational relations.
It would be foolish to attempt to summarize what is already a summary. Instead it would be natural to conclude this story with an attempt to sketch which the likely directions in which demographic research in the US might go in the future. Fortunately, this is the task assigned to Samuel Preston, who as a leader in much of the best demographic research of the past three decades, and as a demographer who continues to work at the frontier, is well placed to undertake this task.
References
Arthur, Brian W. and Geoffrey McNicoll (1978) "Samuelson, Population and Intergenerational Transfers," International Economic Review, vol. 19, no. 1 (February), pp. 241-246.
Bean, Lee, Geraldine Mineau and Douglas Anderton (1990) Fertility Change on the American Frontier: Adaptation and Innovation (University of California Press, Berkeley).
Becker, Gary S. (1960) 'An Economic Analysis of Fertility', in Universities-National Bureau of Economic Research (ed.) Demographic and Economic Change in Developed Countries (Princeton: Princeton University Press) pp.209-231.
Gary Becker, A Treatise on the Family (Cambridge: Harvard University Presss).
Bongaarts, John (1978) "A Framework for the Proximate Determinants of Fertility," Population and Development Review v.4: 105-132.
Boserup, Ester (1981) Population and Technological Change: A Study of Long-Term Trends (Chicago: University of Chicago Press).
Butz, William and Michael Ward (1979) "The Emergence of U.S. Countercyclical Fertility", American Economic Review 69:318-328.
Carey, J.R., P. Liedo, D. Orozco and J.Vaupel (1992) "Slowing of Mortality Rates at Older Ages in Large Medfly Cohorts," Science v.258, pp.457-461.
Coale, Ansley (1972) The Growth and Structure of Human Populations: A Mathematical Investigation (Princeton, Princeton University Press).
Coale, Ansley and James Trussell (1974) "Model Fertility Schedules: Variation in the Age Structure of Childbearing in Human Populations," Population Index v.40 n.2 (April) pp.185-258.
Coale, Ansley and Paul Demeny (1966) Regional Model Life Tables and Stable Populations (New York: Academic Press).
Coale, Ansley J. (1969) "The Decline of Fertility in Europe From the French Revolution to World War II," in S.J. Behrman, Leslie Corsa and Ronald Freedman, eds., Fertility and Family Planning: A World View (Ann Arbor: the University of Michigan Press), pp.3-24.
Coale, Ansley J. and Edgar M. Hoover (1958) Population Growth and Economic Development in Low Income Countries (Princeton: Princeton University Press).
Cohen, Joel (1976) "Ergodicity of Age Structure in Populations with Markovian Vital Rates, I. Countable States," Journal of the American Statistical Association, 71:335-339.
Easterlin, Richard A. (1978a) "The Economics and Sociology of Fertility: A Synthesis," in C. Tilly, ed., Historical Studies of Changing Fertility (Princeton: Princeton University Press).
Easterlin, Richard A. (1978b) "What will 1984 be like? The socioeconomic implications of recent twists in age structure". Demography, 15:397-432.
Ehrlich, Paul (1968) The Population Bomb (Ballantine Books, New York).
Foster, Andrew (1990) "Cohort Analysis and Demographic Translation: A Comparative Study of Recent Trends in Age Specific Fertility Rates from Europe and North America," Population Studies v.44 n.2 (July) pp.287-316.
Frauenthal, James (1975) "A Dynamic Model for Human Population Growth", Theoretical Population Biology 8:64-73.
Freedman, Ronald, P.K. Whelpton and Arthur Campbell (1959) Family Planning, Sterility and Population Growth, McGraw-Hill, New York.
Goodman, Leo, Nathan Keyfitz and Thomas Pullum (1974) "Family Formation and the Frequency of Various Kinship Relationships," Theoretical Population Biology 8:376-381.
Grossman, Michael (1972) The Demand for Health: A Theoretical and Empirical Investigation (Columbia University Press, New York)
Hardin, Garret (1968) "The Tragedy of the Commons," Science v.162 pp.1243-1248.
Henry, Louis (1972) On the Measurement of Human Fertility (translated and edited by Mindel C. Sheps and Evelyne Lapierre-Adamcyk; Elsevier Publishing Co.)
Keyfitz, Nathan (1968) Introduction to the Mathematics of Population (Addison-Wesley Publishing Company, Reading).
Keyfitz, Nathan (1977) Applied Mathematical Demography (Wiley).
Lee, Ronald "Aiming at a Moving Target: Period Fertility and Changing Reproductive Goals," Population Studies, vol. 34, no. 2 (July, 1980), pp. 205-226.
Lee, Ronald "Estimating Series of Vital Rates and Age Structures from Baptisms and Burials: A New Technique, with Applications to Preindustrial England," Population Studies 28, no. 3 (November 1974), pp. 495-512.
Lee, Ronald "The Formal Demography of Population Aging, Transfers, and the Economic Life Cycle," in Linda Martin and Samuel Preston, eds.; The Demography of Aging (National Academy Press, 1994) pp.8-49.
Lee, Ronald and Shripad Tuljapurkar, "Stochastic Population Projections for the U.S.: Beyond High, Medium and Low," Journal of the American Statistical Association, December, 1994.
Lee, Ronald D. (1974). 'The Formal Dynamics of Controlled Populations and the Echo, the Boom and the Bust', Demography 11(4):563-585.
Leibenstein, Harvey (1957) Economic Backwardness and Economic Growth (John Wiley and Sons, New York).
Lopez, Alvaro (1961) Problems in Stable Population Theory (Princeton New Jersey, Office of Population Research).
Meadows, Donella, Dennis Meadows, Jorgen Randers and William Behrens III (1972) The Limits to Growth: A Report for the Club of Rome's Project on the Predicament of Mankind (Universe Books, New York).
Mosely, Henry and Lincoln Chen (1984) "An Analytical Framework for the Study of Child Survival in Developing Countries," Population and Development Review A Supplement to Vol.10: 25-49.
National Research Council (1986) Population and Economic Development: Policy Issues (National Academy Press, Washington D.C.).
Notestein, Frank W. (1953) "Economic Problems of Population Change," Eighth International Conference of Agricultural Economists, Proceedings (London, Oxford University Press), pp.13-31.
Paul R. Ehrlich (1969) The Population Bomb (Ballantine Books, New York)
Pollak, Robert (1990) "Two-Sex Demographic Models," Journal of Political Economy v.98 n.2 (April) pp.399-420.
Preston, Samuel (1982) "Relations Between Individual Life Cycles and Population Characteristics", American Sociological Review v.7, n.2 (April) pp. 253-264.
Preston, Samuel and Ansley Coale (1982) "Age Structure, Growth, Attrition and Accession: A New Synthesis," Population Index (Summer) 50(2), p.214-226.
Preston, Samuel H. (1980) 'Causes and Consequences of Mortality Declines in Less Developed Countries During the Twentieth Century', in R. A. Easterlin (ed.) Population and Economic Change in Developing Countries (ch. 5) (Chicago: University of Chicago Press) pp. 289-360.
Pritchett, Lant H. (1994) "Desired Fertility and the Impact of Population Policies," Population and Development Review, v.20 n.1 (March) pp.1-55.
Rogers, Andrei (1966) "The Multiregional Matrix Growth Operator and the Stable Interregional Age Structure," Demography, 3 pp.537-544.
Rosenzweig, Mark and T. Paul Schultz, "The Demand For and Supply Of Births: Fertility and Its Life Cycle Consequences," American Economic Review (December 1985), v.75, 992-1015.
Ryder, Norman (1951) The Cohort Approach (University Microfilms, Ann Arbor), republished under the same title by Arno Press, New York, in 1981.
Ryder, Norman (1960) "The Structure and Tempo of Current Fertility," in Universities- Demographic and Economic Change in Developed Countries (Princeton: Princeton University Press).
Ryder, Norman (1978) "A Model of Fertility by Planning Status," Demography v.15, n.4 (November), pp.433-458.
Ryder, Norman and Charles Westoff (1971) Reproduction in the United States 1965 (Princeton: Princeton University Press).
Samuelson, Paul (1958) "An Exact Consumption-Loan Model of Interest With or Without the Social Contrivance of Money," Journal of Political Economy v.66 n.6 pp.467-482.
Samuelson, Paul (1976). An Economist's Non-Linear Model of Self-Generated Fertility Waves, Population Studies 30(2):243-248.
Schoen, Robert and Young Kim (1991) "Movement Towards Stability as a Fundamental Principal of Population Dynamics," Demography v.28 pp.455-466.
Sheps, Mindel and Jane Menken (1973) Mathematical Models of Conception and Birth (Chicago: University of Chicago Press).
Simon, Julian (1981) The Ultimate Resource (Princeton University Press, Princeton).
Stoto, Michael (1983) "The Accuracy of Population Projections," Journal of the American Statistical Association v.78 n.381 (July) pp.13-20.
Sykes, Zenas (1966) The Variance of Population Projections University Microfilms, Ann Arbor, Michigan.
Takeshita, John and Ronald Freedman (1969) Family Planning in Taiwan: An Experiment in Social Change (Princeton: Princeton University Press).
Tuljapurkar, Shripad (1990) Population Dynamics in Variable Environments (Springer-Verlag, Lecture Notes in Biomathematics n.85).
United Nations/National Research Council (1983) Manual X: Indirect Techniques for Demographic Estimation (United Nations, New York).
Vaupel, James, Kenneth Manton and Eric Stallard (1979) "The Impact of Heterogeneity in Individual Frailty on the Dynamics of Mortality," Demography v.16 n.3, pp.439-454.
Wachter, Kenneth W. (1984) "Lotka's Roots Under Rescalings," Proceedings of the National Academy of Sciences v.81 pp.3600-3604.
Wachter, Kenneth W. (1986) "Ergodicity and Inverse Projection," Population Studies v.40 pp.275-287.
Wachter, Kenneth W. (1991) "Elusive Cycles: Are there Dynamically Possible Lee-Easterlin Models for U.S. Births?", Population Studies v.45, :109-135.
Wachter, Kenneth W., Eugene A. Hammel and Peter Laslett (1978) Statistical Studies of Historical Social Structure (Academic Press, New York).
Willis, Robert (1988) "Life Cycles, Institutions and Population Growth: A Theory of the Equilibrium Interest Rate in an Overlapping-generations Model," in Ronald Lee, W. Brian Arthur and Gerry Rodgers eds., Economics of Changing Age Distributions in Developed Countries (Oxford University Press), pp.106-138.
Willis, Robert (1973) 'A New Approach to the Economic Theory of Fertility Behavior', Journal of Political Economy, v.81 n.2:S14-S64.
Woodbury, M.A. and Kenneth Manton (1977) "A Random Walk Model of Human Mortality and Aging," Theoretical Population Biology 11:37-48.
World Bank (1984) World Development Report (Oxford University Press, London).