Darryl J. Holman
Department of Anthropology
Center for Studies in Demography and Ecology
University of Washington

Presented at the NICHD Panel,
"Visions of the Future: A Town Meeting on New Directions in Population Research"
Annual Meeting of the Population Association of America

March 2000

Population studies is a remarkably broad field that borrows tools and theory from many other disciplines. What links us together as demographers, in the traditional sense, is a set of core theories and methods for investigating population dynamics. Beyond this, we find common ground in our focus on the events over the human life course. We are linked together by the tools and approaches that are indigenous to demography. The irreversible nature of time, the long life span that our study subjects have relative to the researcher, and the difficulty of generating complete prospective observations of events over the life course has led population scientists to specialize in extracting the most from sometimes very messy data.

Most of us have two academic lives: one life as a demographer and another within a traditional academic discipline like sociology, economics, geography, public health, political science, or anthropology. I am a biological anthropologist, and as such I consider myself both a biological scientist and a social scientist. My particular interests within anthropology and demography are human birthspacing, fecundity and reproductive aging, as well as the health and mortality of past populations.

My affinity for incorporating biological measurement, mechanism, and theory into demography places me at the margin of mainstream demography. Even so, the field we call "population studies" is interdisciplinary by nature; so that many of us are near the margins. Historically, demography has had a biological element, or at least a biometrical element, to it.¹ After all, most of us really do believe that fertility, mortality, and morbidity have something to do with biology, even if we are not personally interested in it as a scientific question. At the same time, most of us also recognize that social, economic, and political processes can strongly shape and sometimes completely override biology.

The interactions between these human cultural phenomena and our biology in shaping the life course, as well as our health and well-being is a fascinating, fruitful, and relatively unexplored area of scientific inquiry. In what follows, I will discuss some of the exciting interdisciplinary directions and scientific opportunities in population research that combine biology, social and cultural studies, and demography. My focus will be on what has come to be termed "biodemography." I will address five of the six areas listed in the Report to the NACHHD Council (1999): theory and methods, demography and health, family and fertility, sexual behavior and prevention research, and population distribution and movement.

Mathematical theory and methods have always played an important role in demography, particularly in the mathematical (or biometrical) sides of demography. Much of classic demography has been the use of mathematical and statistical models to try to understand biocultural variation in population growth (e.g. Lotka 1907; Pearl 1925), fertility (e.g. Pearl 1933a,b; Gini 1924; Henry 1953a,b; Sheps and Menken 1973), mortality (e.g. Gompertz 1825; Makeham 1860, and Beard 1963), and even population movement (Boyce, Küchemann and Harrison 1971) and marriage (Coale and McNeil 1972).

I am a big fan of this type of etiologic modeling, both as a way of clarifying the sometimes murky waters of theory, as well as for developing measurement models out of theory. More recent extensions of much of the classic work in statistical and mathematical demography have been hampered by the sheer mathematical complexity of the models. As a result, model complexity has become an obstacle to further development and understanding of population dynamics. For example, a distribution for some event that is composed of multiple underlying (and perhaps unobservable) events can be specified as a convolution of the individual underlying distributions. Complexity can grow rapidly, and after some number of nested integrals (perhaps 3 or 4) computing time becomes so slow that iterative maximization techniques are rendered useless. In other words, we have become adept at producing complex equations without actually being able to estimate the parameters of the models given a series of observations.

The development of fast computers has certainly helped this situation, although performing an apparently simple task like numerical integration of three nested integrals is still dauntingly time consuming. Some relatively new methods of stochastic integration have recently been developed, and these appear to be a fundamental breakthrough in computational statistics. Markov chain Monte Carlo (MCMC) and related techniques are rapidly maturing and proving to be powerful tools for statistical inference. The technique has been used in various forms in physics for almost a half century, and in spatial statistics for a quarter of a century (Besag et al. 1995). For the Baysian statistician, MCMC allows computers to achieve solutions that were computationally out of reach only a few years ago. For the likelihoodist, these methods allow us to put together complex mathematical models that capture the most important features of the underlying process we are examining. These methods free us from dependence on standard statistical methods which are typically devoid of any underlying demographic theory. We can now tackle complex multilevel models and specify theoretically interesting underlying distributions for the way in which clusters affect the process under investigation. Or, for example, we gain enormous flexibility in specifying meaningful distributions of unmeasured heterogeneity in our models. With these new methods, we can now go back and dust off some of that old mathematical theory and create useful statistical models from it, and ultimately develop and refine some of the theory, backed up by empirical evaluation. The future looks good for, as Hilborn and Mangel (1997) put it, "confronting our models with data". Computer power will continue to grow exponentially and new statistical techniques should allow us to more easily put together models in which we can directly interpret the underlying parameters, rather than simply hunting for significant p-values.

Another set of tools that has been increasingly used in demography are methods borrowed from biology and the biomedical community—tools like hormone assays, measurements of immune function, and sensitive assays for nutritional status. There are many potential application of biomedical tools to further our understanding of human health and the life course, limited only by our acceptance of the tools and our willingness to apply them to demographic questions. I am not proposing that demographers use these tools for low-level physiological investigation. Instead, we can use these tools in the very tradition of demography as a means of illuminating and understanding another level of heterogeneity and how it influences the life course. Several examples will help illustrate the point.

New assays and other methods are now available that provide a more detailed measure of nutritional status and health, such as tools that assess immune system competence by a rapid and nearly non-invasive test. Shell-Duncan (1993) studied childhood health and morbidity among the Turkana pastoralists in Kenya. She used this relatively simple immunological test as one biological measure of health—immune system competence. Using these tests, she was able to identify the frequency with which Turkana children's immune systems are suppressed, and study subsequent health outcomes. The results, summarized in Figure 1, reveal considerable age-related differences (from 28 to 100 percent) in the frequency of immune system suppression. The method illuminates an important source of heterogeneity in child morbidity and mortality at a level that is inaccessible to traditional demographic methods.

Mark Flinn, a cultural anthropologist at the University of Missouri, has spent a decade studying family dynamics, household stress, and child health in a Caribbean population. For example, he has examined behaviors and interactions among household members, levels of emotional and physical stress and the resulting health outcomes for children. In order to measure some of the underlying biological patterns of health and stress, he has collected well over 20,000 saliva samples from the children he studies. From each sample, he has measured glucocorticoid levels as an indicator of stress. The picture that emerges from his studies is that individuals respond differently to the stresses of life events. Furthermore, he has shown that individual responses to social environment can have important health consequences.

Figure 2

shows individual response to some stressful life events in three children, observed over differing temporal scales.

A common finding in a number of studies investigating cortisol and stress is that different individuals react differently to similar stressful events. An individual's response to stress may provide long-term information about that individual’s health; indeed, it may prove to be a sensitive predictor of sexual and non-sexual risk-taking behaviors (Halpern et al. 2000).

For a number of years I have been working with James Wood, Kenneth Campbell and Kathleen O'Connor using hormonal measures to examine biological determinants of human fertility, birthspacing, and reproductive aging. In one study (Holman 1996; O'Connor et al. 1998), we collected twice weekly interviews and urine samples from about 500 Bangladeshi women over the course of a year in order to study fecundability and fetal loss². The use of highly sensitive pregnancy assays allowed us to detect most pregnancies before the first missed menses. Our statistical model was developed from a two-point frailty model to statistically capture the most biologically important component of pregnancy loss: an increase in chromosomal abnormalities by maternal age. The combination of theory embodied as a parametric model, sensitive and specific endocrine measures, and careful attention to the nature of the data allowed us to estimate for the first time age-specific total fecundability and total fetal loss. Somewhat surprisingly, our results suggested that total fecundability is quite high and changes very little from the mid-20s until about age 40. What demographers have traditionally interpreted to be an age-related decline in apparent fecundability is, in fact, almost entirely a result of early fetal loss.

The point of this example is that we could not have measured these two most fundamental parameters of demography (total fecundability and total fetal loss) without a little bit of biological understanding, some simple biomedical tools, a population-based (i.e. non-clinical) sample, and the demographer's analytical tools.

Tools like endocrine assays have been used by demographers to investigate sexual behavior (as well as non-sexual behaviors) for a number of years (e.g. Udry 1994). I want to emphasize a point, which has been made by many others. Sexual behavior is an area where we really should think deeply about the underlying biology and how our biology influences our behavior. Humans, like all sexually reproducing species, have biological instincts to ensure the successful production of later generations. Biological ideas are almost certainly necessary for a thorough understanding of age and sex-specific sexual behaviors.

The fields of sociobiology and behavioral ecology draw upon biological ideas of evolution and adaptation in order to understand social and behavioral phenomenon. Many social scientists have rejected (and continue to reject) these ideas. Even so, in recent years, demographers have begun to embrace some of these ideas for understanding fertility.³

The theories that come out of behavioral ecology and sociobiology provide a larger context within which to understand sexual behavior, reproductive behavior, and investment in children, and yet they may never provide detailed practical information, or prescriptions for social or public health interventions. By analogy, evolutionary theory, which has long been the foundation for the biomedical sciences, has provided a powerful framework for understanding general concepts such as the evolution of disease virulence, antibiotic resistance, human senescence, and dietary preference (Williams and Nesse, 1991). Notwithstanding this important role in explaining aspects of human health and disease, evolutionary theory provides little in the way of guidance for medical cures and interventions.⁴ Likewise, evolutionary theory can help us develop a broad and unified understanding of sexual, reproductive and parenting behaviors but is unlikely to suggest practical interventions aimed at changing these behaviors.

Studies of non-industrialized populations have always played, and I hope will continue to play, an important role of demography. Ultimately our most powerful theories must be able to explain population processes in non-Western settings. These populations act as natural experiments outside the original context in which the theory was developed.⁵

One such future opportunity arises out of the tragedy of HIV infection, particularly in those parts of Africa where a substantial fraction of the adult population has been infected. This is the kind of natural experiment we would prefer never happened. A news item in Science reporting on a recent infectious disease conference (Cohen 2000) points out that infection rates among adults in some countries are from 20 to over 32 percent. The consequences on all aspects of the life course in these countries is likely to be enormous.

Population scientists have contributed a number of innovative tools for understanding the spread of this epidemic, and I hope these efforts continue and intensify. This tragedy will also provide opportunity to enrich our understanding of many aspects of health, economics, society, and the human life course. For example, Bloom and Canning (2000) have proposed that health plays a leading role in driving development and economic growth rather than economic growth driving improvements in health. This and other theories will have important testing grounds in parts of Africa, with the ultimate outcome being a richer understanding of population processes and better health for future generations.

Likewise, as humans conglomerate in urban centers of unprecedented size and density, demographers have a "natural experiment" in which to study the life course under demographic extremes (Cohen 1995).

It is an exciting time for population studies. We have many interesting questions, better tools, and more data with which to address the questions. Good training and resources in population studies will inevitably lead to new methodological and theoretical breakthroughs, followed by a better understanding of the human life course; ultimately, this will lead to population and health interventions that are better informed by science.

I thank Ken Campbell, Robert Jones, Martina Morris, Ken Weiss, and Jim Wood for sharing with me their ideas about anthropology and demography. A number of people kindly agreed to in-depth discussions about the topics in this paper including Kathleen O'Connor, Bettina Shell-Duncan, Charlie Hirschman, Bob Plotnick, Jim Wood, Geoff Kushnick, Matthew Steele, Mary Schenk, and Bob Jones. I thank them for sharing their insights. Inaccuracies, errors, improper grammar, and bad ideas are mine alone.

¹ Many of the pioneers of mathematical demography are also recognized as fundamental contributors in fields of biology. Raymond Pearl founded the journal Human Biology in 1929, which continues publication today. Alfred J. Lotka, from who we get stable population theory, is considered a pioneer in physical biology and ecology.

² In fact, the study was similar in design to Chen et al. (1974) who also used endocrine assays and specimens collected once a month.

³ This was particularly noticeable at the 2000 meetings of the Population Association of America meetings, where a number of sessions included papers along these lines.

⁴ In fact, many medical interventions that we consider necessary are not consistent with long-term evolutionary strategies. One example is the use of antibiotics, which inherently selects for resistant strains of bacteria.

⁵ Sometimes work in non-industrialized populations simply allows us to control for particular behaviors or particular biases. A good example of this is measuring fecundability in non-contracepting populations in order to avoid serious biases that would otherwise result (Wood 1994).

Beard RE (1963) A theory of mortality based on actuarial, biological and medical considerations. In Proceedings of the International Population Conference. London: International Union for the Scientific Study of Population.

Besag J, Green P, Higdon D, Mengersen K (1995) Bayesian computation and stochastic systems. Statistical Science 10(1):3-66.

Bloom DE and Canning D (2000) The health and wealth of nations. Science 287:1207-9.

Boyce AJ, Küchemann CF, Harrison GA (1971) Population structure and movement patterns. In Brass W (ed.) Biological Aspects of Demography. London: Taylor and Frances, pp. 1-10.

Chen LC, Ahmed S, Gesch M, Mosley WH (1974) A prospective study of birth interval dynamics in rural Bangladesh. Population Studies 28:277-97.

Coale AJ, McNeil DR (1972) The distribution by age of the frequency of first marriage in a female cohort. JASA 67:743-9.

Cohen J (2000) AIDS researchers look to Africa for new insights. Science 287:942-3.

Cohen JE (1995) How Many People Can the Earth Support? New York:W. W. Norton.

Flinn MV (1999) Family environment, stress, and health during childhood. In: Panter-Brick C& Worthman C (eds) Hormones, Health, and Behavior. Cambridge: Cambridge University Press, pp. 105-138.

Gini C (1924) Premières recherches sur la fécondabilité de la femme. Proceedings of the International Mathematical Congress 2:889-92.

Gompertz B (1825) On the nature of the function expressive of the law of human mortality and on a new mode of determining the value of life contingencies. Philosophical transactions of the Royal Society 115:513-85.

Halpern CT, Campbell B, Agnew CR, Thompson VD (2000) Associations between stress reactivity and sexual and non-sexual risk-taking in adolescent males. Paper presented at the Annual meetings of the Population Association of America Meetings, Los Angeles, 23 March.

Henry L (1953a) La Population Canadienne au Début du XVIIIème Siècle. Paris: Presses Universtaires de France.

Henry L (1953b) Fondements théoriques des mesures de la fécondité naturelle. Revue de l'Institut International de Statistique 21:135-151.

Hillborn R and Mangel M (1997) The Ecological Detective: Confronting Models with Data. Princeton, New Jersey: Princeton University Press.

Holman DJ (1996) Total Fecundability and Fetal Loss in Rural Bangladesh. Doctoral Dissertation, The Pennsylvania State University.

Lotka AJ (1907) Mode of growth of material aggregates. American Journal of Science 24:199-216.

Makeham WM (1860) On the law of mortality. Journal of the Institute of Actuaries 13:325-58.

O'Connor KA, Holman DJ and Wood JW (1998) Declining fecundity and ovarian aging in natural fertility populations. Maturitas 30:127-36.

Pearl R (1925) The Biology of Population Growth. New York: Alfred A. Knopf.

Pearl (1933a) Factors in human fertility and their statistical evaluation. Lancet 225:607-11.

Pearl (1933b) On the frequency of the use of contraceptive methods and their effectiveness as used by a sample of American women. Bulletin de l'Institut International de Statistique 27:208-24.

Report to the NACHHD Council (1999), Demographic and Behavioral Sciences Branch (NICHD), National Institutes of Health.

Shell-Duncan B (1993) Cell-mediated immunocompetence among nomadic Turkana children. American Journal of Human Biology 5:225-35.

Sheps MC, Menken JA (1973) Mathematical Models of Conception and Birth. Chicago: University of Chicago Press.

Udry JR (1994) The nature of gender Demography 31(4):561-73.

Williams GC, Nesse RM (1991) The dawn of Darwinian medicine. The Quarterly Review of Biology. 66:1-22.

Wood JW (1994) Dynamics of Human Reproduction: Biology, Biometry, Demography. Hawthorne, NY: Aldine de Gruyter.

Figure 1. Percent of Turkana children whose immune system was incompetent (anergic) by age. Redrawn from Shell-Duncan (1993).

Figure 2. Longitudinal monitoring cortisol levels on different chronological scales as a tool for investigating stress response among children. Top panel: Hourly sampling of a 12-year-old male showing an elevation of cortisol associated with work-load stress. Middle panel: Twice-daily sampling of a 13-year-old girl showing changes in cortisol associated with the absence of a caretaking grandmother. Bottom panel: Twice-daily samples from a male over a 7-year period showing the change in cortisol associated with the absence of his father. Redrawn from Flinn (1999).

Figure 3. Maternal age-specific total fecundability and total pregnancy loss in Bangladeshi women (Holman 1996; O'Connor et al. 1998).