The interactions between human population dynamics and the environment have often been viewed mechanistically. This review elucidates the complexities and contextual specificities of population-environment relationships in a number of domains. It explores the ways in which demographers and other social scientists have sought to understand the relationships among a full range of population dynamics (e.g., population size, growth, density, age and sex composition, migration, urbanization, vital rates) and environmental changes. The chapter briefly reviews a number of the theories for understanding population and the environment and then proceeds to provide a state-of-the-art review of studies that have examined population dynamics and their relationship to five environmental issue areas. The review concludes by relating population-environment research to emerging work on human-environment systems.
Keywords: climate change, coastal and marine environments, land-cover change, land degradation, population dynamics, water resources
Humans have sought to understand the relationship between population dynamics and the environment since the earliest times (1, 2), but it was Thomas Malthus’ Essay on the Principle of Population (3) in 1798 that is credited with launching the study of population and resources as a scientific topic of inquiry. Malthus’ famous hypothesis was that population numbers tend to grow exponentially while food production grows linearly, never quite keeping pace with population and thus resulting in natural “checks” (such as famine) to further growth. Although the subject was periodically taken up again in the ensuring decades, with for example George Perkins Marsh’s classic Man and Nature (1864) (4) and concern over human-induced soil depletion in colonial Africa (5, 6), it was not until the 1960s that significant research interest was rekindled. In 1963, the U.S. National Academy of Sciences published The Growth of World Population (7), a report that reflected scientific concern about the consequences of global population growth, which was then reaching its peak annual rate of two percent. In 1968, Paul Ehrlich published The Population Bomb (8), which focused public attention on the issue of population growth, food production, and the environment. By 1972, the Club of Rome had released its World Model (9), which represented the first computer-based population-environment modeling effort, predicting an “overshoot” of global carrying capacity within 100 years.
Clearly, efforts to understand the relationship between demographic and environmental change are part of a venerable tradition. Yet, by the same token, it is a tradition that has often sought to reduce environmental change to a mere function of population size or growth. Indeed, an overlay of graphs depicting global trends in population, energy consumption, carbon dioxide (CO2) emissions, nitrogen deposition, or land area deforested has often been used to demonstrate the impact that population has on the environment. Although we start from the premise that population dynamics do indeed have an impact on the environment, we also believe that monocausal explanations of environmental change that give a preeminent place to population size and growth suffer from three major deficiencies: They oversimplify a complex reality, they often raise more questions than they answer, and they may in some instances even provide the wrong answers.
As the field of population-environment studies has matured, researchers increasingly have wanted to understand the nuances of the relationship. In the past two decades demographers, geographers, anthropologists, economists, and environmental scientists have sought to answer a more complex set of questions, which include among others: How do specific population changes (in density, composition, or numbers) relate to specific changes in the environment (such as deforestation, climate change, or ambient concentrations of air and water pollutants)? How do environmental conditions and changes, in turn, affect population dynamics? How do intervening variables, such as institutions or markets, mediate the relationship? And how do these relationships vary in time and space? They have sought to answer these questions armed with a host of new tools (geographic information systems, remote sensing, computer-based models, and statistical packages) and with evolving theories on human-environment interactions.
This review explores the ways in which demographers and other social scientists have sought to understand the relationships among a full range of population dynamics (e.g., population size, growth, density, age and sex composition, migration, urbanization, vital rates) and environmental changes. With the exception of the energy subsection, the focus is largely on micro- and mesoscale studies in the developing world. This is not because these dynamics are unimportant in the developed world—on the contrary, per capita environmental impacts are far greater in this region (see the text below on global population and consumption trends)—but rather because this is where much of the research has focused (10). We have surveyed a wide array of literature with an emphasis on peer-reviewed articles from the past decade, but given the veritable explosion in population-environment research, we hasten to add that this review merely provides a sampling of the most salient findings. The chapter begins with a short review of the theories for understanding population and the environment. It then proceeds to provide a state-of-the-art review of studies that have examined population dynamics and their relationship to the following environmental issue areas: land-cover change and deforestation; agricultural land degradation and improvement; abstraction and pollution of water resources; coastal and marine environments; and energy, air pollution, and climate change. In the concluding section, we relate population-environment research to the emerging understanding of complex human-environment systems.
Global Trends in Population and Consumption
At the global level, research has found that the two major drivers of humanity’s ecological footprint are population and consumption (11), so we provide a brief introduction to the status and trends in these two indicators.
The future size of world population is projected on the basis of assumed trends in fertility and mortality. Current world population stands at 6.7 billion people (12). The 2006 revision of the United Nations World Population Prospects presents a medium variant projection by 2050 of 9.2 billion people and still growing, although at a significantly reduced rate. All of the projected growth is expected to occur in the developing world (increasing from 5.4 to 7.9 billion), whereas the developed world is expected to remain unchanged at 1.2 billion. Africa, which has the fastest growing population of the continents, is projected to more than double the number of its inhabitants in the next 43 years—from 965 million to approximately 2 billion. Globally, fertility is assumed to decline to 2.02 births per woman (below replacement) by 2050; it is population momentum arising from a young age structure that will cause global population to continue to grow beyond 2050 (the 2006 revision does not make prognoses about ultimate stabilization). The medium variant is bracketed by a low-variant projection of 7.8 billion (and declining) and a high variant of 10.8 billion (and growing rapidly) by 2050. Fertility in the former is assumed to be half a child lower than the medium variant, and in the latter, it is assumed to be half a child higher.1 As Cohen (2) points out, minor variations in above- or below-replacement fertility can have dramatic long-term consequences for the ultimate global population size; hence, projections are highly conditional, and their sensitivity to the underlying assumptions needs to be properly understood. Finally, the impact of the HIV/AIDS epidemic on future mortality is assumed to attenuate somewhat on the basis of recent declines in prevalence in some countries, increasing antiretroviral drug therapy, and government commitments made under the Millennium Declaration (13).
Consumption trends are somewhat more difficult to predict because they depend more heavily than population projections on global economic conditions, efforts to pursue sustainable development, and potential feedbacks from the environmental systems upon which the global economy depends for resources and sinks. Nevertheless, several indicators of consumption have grown at rates well above population growth in the past century: Global GDP is 20 times higher than it was in 1900, having grown at a rate of 2.7% per annum (14); CO2 emissions have grown at an annual rate of 3.5% since 1900, reaching an all-time high of 100 million metric tons of carbon in 2001 (15); and the ecological footprint, a composite measure of consumption measured in hectares of biologically productive land, grew from 4.5 to 14.1 billion hectares between 1961 and 2003, and it is now 25% more than Earth’s “biocapacity” according to Hails (16). In the case of CO2 emissions and footprints, the per capita impacts of high-income countries are currently 6 to 10 times higher than those in low-income countries. As far as the future is concerned, barring major policy changes or economic downturns, there is no reason to suspect that consumption trends will change significantly in the near term. Long-term projections suggest that economic growth rates will decline past 2050 owing to declining population growth, saturation of consumption, and slower technological change (14).
As in any contested field—and population-environment studies certainly fit this description—a wide array of theories have emerged to describe the relationship among the variables of interest, and each of these theories leads to starkly different conclusions and policy recommendations. Here we review the most prominent theories in the field of population and environment.
The introduction briefly touched on the work of Malthus, whose theory still generates strong reactions 200 years after it was first published. Adherents of Malthus have generally been termed neo-Malthusians. In its simplest form, neo-Malthusianism holds that human populations, because of their tendency to increase exponentially if fertility is unchecked, will ultimately outstrip Earth’s resources, leading to ecological catastrophe. This has been one of the dominant paradigms in the field of population and the environment, but it is one which many social scientists have rejected because of its underlying biological/ecological underpinnings, treating humans in an undifferentiated way from other species that grow beyond the local “carrying capacity.” Neo-Malthusianism has been criticized for overlooking cultural adaptation, technological developments, trade, and institutional arrangements that have allowed human populations to grow beyond their local subsistence base.
Neo-Malthusianism underpins the Club of Rome World Model (mentioned above) (9) and implicitly or explicitly underlies many studies and frameworks. The widely cited IPAT formulation—in which environmental impacts (I) are the product of population (P), affluence (A), and technology (T)—is implicitly framed in neo-Malthusian terms (17), although not all research using the identity is Malthusian in approach (18). IPAT itself has been criticized because it does not account for interactions among the terms (e.g., increasing affluence can lead to more efficient technologies); it omits explicit reference to important variables such as culture and institutions (e.g., social organization); impact is not linearly related to the right side variables (there can be important thresholds); and it can simply lead to wrong conclusions (19).2
The so-called Boserupian hypothesis, named after agricultural economist Esther Boserup, holds that agricultural production increases with population growth owing to the intensification of production (greater labor and capital inputs). Although often depicted as being in opposition to Malthusianism, Malthus himself acknowledged that agricultural output increases with increasing population density (just not fast enough), and Boserup acknowledged that there are situations under which intensification might not take place (20). As Turner & Ali (21) point out, the main difference between the theories of Malthus and Boserup is that Malthus saw technology as being exogenous to the population-resource condition and Boserup sees it as endogenous. Cornucopian theories espoused by some neoclassical economists stand in sharper contrast to neo-Malthunisianism because they posit that human ingenuity (through the increased the supply of more creative people) and market substitution (as certain resources become scarce) will avert future resource crises (22). In this line of thinking, market failures and inappropriate technologies are more responsible for environmental degradation than population size or growth, and natural resources can be substituted by man-made ones.
Political ecology also frequently informs the population-environment literature (23). Many political ecologists see population and environment as linked only insofar as they have a common root cause, e.g., poverty, and that poverty itself stems from economic imbalances between the developed and developing world and within developing countries themselves (e.g., 24). In this view, migrants to deforestation hot spots in frontier areas may be victims of historical inequalities in land access in their country’s core agricultural areas, or they may be responding to global inequalities in which industrialized countries depend on resource extraction from tropical countries to maintain their high standards of living, or both. Whatever the impact of the migrant on the rainforest, it is merely a symptom of more deeply rooted imbalances. Similarly, political ecologists see land degradation as stemming from poor farmer’s lack of access to credit, technology, and land rather than population growth per se.
A number of theories—often subscribed to by demographers—state that population is one of a number of variables that affect the environment and that rapid population growth simply exacerbates other conditions such as bad governance, civil conflict, wars, polluting technologies, or distortionary policies. These include the intermediate (or mediating) variable theory (23) or the holistic approach (25) in which population’s impact on the environment is mediated by social organization, technology, culture, consumption, and values (26, 27). Some also group IPAT in this category because population is only one of the three variables contributing to environmental impacts.
Many theories in the field of population and environment are built on theoretical contributions from a number of fields. A case in point is the vicious circle model (VCM), which attempts to explain sustained high fertility in the face of declining environmental resources (28, 29). In this model, it is hypothesized that there are a number of positive feedback loops that contribute to a downward spiral of population growth, resource depletion, and rising poverty (see the land degradation section). At the simplest level, the model is neo-Malthusian, but it also owes a debt to a number of other theories. First, it builds on the intergenerational wealth flows theory from demography, which holds that high fertility in traditional societies is beneficial to older generations owing to the net flow of wealth from children to parents over the course of their lifetimes (30). It also borrows from a demographic theory that describes fertility as an adjustment to risk, which argues that in situations where financial and insurance markets and government safety nets are poorly developed, children serve as old-age security (31). Finally, it is partially derived from the ecologist Garrett Hardin’s famous (32) “tragedy of the commons,” which holds that as long as incentives exist for each household to privatize open access resources, then there will be a tendency at the societal level to overexploit available resources to the detriment of all users.
It is important to note that population-environment theories may simultaneously operate at different scales, and thus could all conceivably be correct. At the global level, we cannot fully predict what the aggregate impacts of population, affluence, and technology under prevailing social organization will be on the global environment when the world’s population reaches 9 or 10 billion people (3). But many scientists—neo-Malthusian or not—are justifiably concerned with the impact that even the current 6.7 billion people are having on the planet given consumption patterns in the global North and the booming economies of China and India. Meanwhile, at the national level the cornucopian theory may be correct, say, for a country like Denmark, whereas neo-Malthusianism, political ecology, and intermediate variable theories may each illuminate different facets of Haiti’s environmental crisis. Finally, Boserup’s theory of intensification has been found to hold true in the historical experience of many developed countries and in many localized case studies spanning the developing world (33).
Although theory may seem dry and academic, theoretical frameworks can be important guides to action. A good theory helps to develop well-targeted policies. However, bad theory can become the “orthodoxies” that are very difficult to overcome and that underlie government and development agency policies and programs (34, 35). Each of the above theories identifies one or more ultimate causes for environmental degradation, which if remedied would “solve” the problem. In the case of neo-Malthusianism, population growth is the primary problem, and the solution is population programs. In the case of cornucopianism, market failures are the primary problem, and the solution is to fix them. For political ecologists, inequalities at different scales are the main problem, and policies should address those inequalities. Multivariable theories offer few magic bullets but do underscore the need for action on multiple fronts to bring about sustainability. Unfortunately, many theories in the realm of population and the environment have not been subjected to the level of rigorous empirical testing that would allow them to be categorized as robust. This is partly because the linkages are complex and difficult to disentangle. Fortunately for the field as a whole, the picture is beginning to change, and a number of studies at the microlevel have used robust statistical methods and multilevel modeling in order to test theories such as the VCM (36).
We now turn to a review of the five issue areas.
REVIEW BY ENVIRONMENTAL ISSUE AREA
In this section, we review the literature on population-environment interactions in each of five issue areas: land-cover change, agricultural land degradation, water resource management, coastal management, and energy and climate change. We focus largely on peer-reviewed articles published in the past decade with an occasional reference to important earlier work.
Land-Cover Change and Deforestation
The conversion of natural lands to croplands, pastures, urban areas, reservoirs, and other anthropogenic landscapes represents the most visible and pervasive form of human impact on the environment (37). Today, roughly 40% of Earth’s land surface is under agriculture, and 85% has some level of anthropogenic influence (38). Although the world’s population is now 50% urban, urban areas occupy less than three percent of Earth’s surface (39). We can conclude from this that large-scale land-cover change is largely a rural phenomenon, but many of its drivers can be traced to the consumption demands of the swelling urban middle classes (40).
As with the demographic and development transitions, the world remains divided in various stages of the land-use transition (41) (Figure 1). Although the developed nations have achieved replacement (2.1 births per woman) or below replacement-level fertility, have urbanized, and have economies dominated by service and technology industries, developing nations continue to experience rapid population growth, remain largely rural, and have labor forces concentrated in the primary sector (agriculture and extractive industries).
Land-use transitions. Reprinted from Reference 163 with the permission of Science.
In part because most developed countries largely deforested their lands in past centuries, today most land conversion from natural states to human uses is occurring in the developing world, particularly in the tropics through forest conversion to agriculture. (One exception is the Russian Far East, which is one of the few developed world regions with high rates of primary forest conversion—mostly for logging and not for agricultural lands.) Given the scale of these transformations and their intimate but complex linkages with population dynamics, research on land-use/-cover change (LUCC) and particularly deforestation constitutes a large portion of the population-environment literature. Demographic variables are linked at different scales to this phenomenon (42). But there is disagreement on the impact of population versus other factors, with some studies suggesting that demographic dynamics contribute more than any other process to deforestation (43) and others suggesting the superiority of economic factors (44). Geist & Lambin’s meta-analysis of 152 case studies of tropical deforestation suggests that, although most cases of deforestation are driven at least partially by population growth, population factors almost always operate in concert with political, economic, and ecological processes, and the relative impact of each factor varies depending on the scale of analysis. In this section, we briefly outline how population dynamics affect LUCC through changes in fertility, population structure, and migration as well as how these interactions are largely mediated by scale. We also reference case studies illustrating the sometimes counter-intuitive relationship between population variables and LUCC.
In much of the developing world fertility rates are plummeting, and nowhere have they declined so rapidly as in urban areas, where (apart from sub-Saharan Africa) fertility is at or below the replacement level. Conversely, in most developing countries, the regions of highest fertility also coincide with the most remotely settled lands where the agricultural frontier continues to advance; areas that are both biodiverse and ecologically fragile. This high fertility and associated rapid population growth directly contributes to land conversion in these forest frontier areas. Fertility in remote areas of the tropics is buoyed by a combination of low demand for and supply of contraception (45). In such regions, children constitute an asset to farm families that are often short on labor (30). Furthermore, poor health care access contributes to high rates of child mortality—promoting so-called “insurance births” that guarantee a family a certain number of surviving children (31). Children compensate for land insecurity through income security to parents in their old age (46), and a dearth of education and work opportunities for women also maintains high fertility (47). Positive correlations between fertility and deforestation have been found in studies in Central (48, 49) and South America (50, 51).
Household age and sex composition and life cycle stages are also important factors in frontier LUCC. Although young children divert household labor resources from agriculture, older children contribute labor to the farm or capture public access resources such as firewood, game, and water. The settlement life cycle of farm homesteads also helps to explain when and where forest clearing will occur (52, 53). Immediately following settlement, deforestation is high as land is cleared for subsistence crops (51, 54). A later deforestation pulse may occur as farms move from subsistence to market-oriented crops or expand into livestock. These processes are enabled by children growing old enough to provide labor or capital investments (through, for example, remittances) to the farm household (53).
Despite the high fertility of remote rural populations, migration remains the primary source of population growth in forest frontiers (44). Indeed, at a key point along the forest transitions causal chain, in-migration is a necessary precedent to frontier deforestation. Migration will remain a major driver of frontier forest conversion, often in a leap-frog manner, as more established farm households send younger family members as migrants to the new frontier (55).
Although population dynamics are central to LUCC, in all cases population exerts its influence synergistically with other factors. Demand for agricultural land among small holders directly impacts forest conversion, whereas, owing to market forces, urban and international demand for forest and agricultural products further contribute to LUCC through logging and large-scale agriculture. Political and institutional factors also play an important role in shaping LUCC. For example, government investments in roads, subsidies to the agricultural sector, or land tenure policy can directly influence deforestation rates. Such effects are well researched in the Brazilian Amazon (56–58). Cultural preferences can also affect LUCC, such as the desire for cattle as a status symbol among Central American frontier farmers (59). Thus, intervening variables help explain inconsistencies in population-LUCC dynamics (60).
Changing the scale of analysis reveals examples in which population growth declined yet deforestation accelerated, population growth was accompanied by reforestation, or population growth attended a number of different human-environment responses (60). Examples of this are evident in the literature for Latin America where many nations have experienced declining rural populations but continued deforestation (48). A dramatic example is Ecuador whose Amazon region’s forest canopy is facing rapid attrition owing to growing settlements of frontier farmers, although overall rural population is declining because of falling fertility and rapid urbanization (61). This apparent anomaly is explained by the small populations, which account for a minority of a nation’s rural population, that move to forest frontiers and contribute a disproportionate amount to the nation’s total deforestation. In parts of the Brazilian Amazon, forest conversion has been driven increasingly by exogenous factors, such as the global demand for soybeans, and owing to increasingly mechanized farming, the region has also experienced rural population decline (62). Interestingly, the same association—rural depopulation and continued deforestation in Ecuador and Brazil—results from a completely different causal mechanism in the two cases, highlighting the importance both of scale and place-based effects. Similar scale-dependent phenomena emerge in Asian forest frontiers. Research in Thailand’s northeast suggests, for example, the importance of population factors at finer scales and of biophysical factors at coarser scales for explaining variation in plant biomass levels (63).
Land-cover dynamics are the most evident mark of human occupation of Earth. Links to population are both obvious (without human population presence there is no human impact on forests apart from acid rain) and exceedingly complex, e.g., at what spatial and temporal scales does population interact with political, economic, and social processes to produce LUCC? A challenge for future research is to disentangle the contributions of population and other dynamics across spatial and temporal scales. For example, more research is needed at the mesoscale (subnational) and to build causal chains across spatial scales. A diversity of research methods needs to be combined to improve our understanding of these space-dependent links, including remote sensing, geographic information systems, ecosystem process and multilevel modeling, surveys and interviews, participant observation, and stakeholder analyses.
Agricultural Land Degradation or Improvement
Land-cover change research also considers changes in the quality of land resources as a result of human uses, which is the focus of this section. Perhaps the most contentious debate in the population-environment literature concerns the relationship between increasing population density in subsistence agricultural areas and land degradation or improvement. This is, in part, the result of widely differing estimates regarding the extent of land degradation, with global estimates ranging from 20 to 51 million km2 (64). This section considers arguments and evidence marshaled by two major schools of thought: the vicious circle proponents who believe that increasing population density in the context of high poverty almost inevitably leads to land degradation and the Boserupians who suggest that increasing density leads to intensification of agricultural systems such that yields per unit area (and per capita) are increased (see the theory section, above).
In the VCM, it is hypothesized that there are a number of positive feedback loops that contribute to a downward spiral of resource depletion, growing poverty, and population growth. An elaboration of these linkages can be found elsewhere (29, 65), but in its simplest form, the model describes the following causal connections: poverty leads to high fertility through mechanisms such as a demand for farm labor, insurance births owing to high infant mortality, and women’s low status. High fertility contributes to population growth, which further increases demands for food and resources from an essentially static resource base; the declining per capita resource base reinforces poverty through soil fertility loss, declining yields, and poor environmental sanitation; and poverty, in turn, contributes to land degradation by increasing incentives for short-term exploitation (versus long-term stewardship) and because poor farmers lack access to costly fertilizers and other technologies. The implication of these reinforcing linkages is that, absent intervention, the circle will continue and soil fertility will decline until the land is no longer suitable for crops or pasture.
Economists have been among the major proponents of the VCM. For example, Panayotou (66) and Dasgupta (28, 67) have suggested that children are valued by rural households, in part, because they transform open access resources (forests, fisheries, and rangeland) into household wealth, resulting in the “externalization” of the costs of high fertility. One manifestation is the process of “extensification,” whereby farm households in frontier areas use additional labor to open up new lands for cultivation (68). Thus, household-level responses to resource scarcity can lead to problems at the societal level as each household copes with increased risk and uncertainty by maximizing its number of surviving children.
A number of modeling efforts, such as the Population-Environment-Development-Agriculture model (69) and work by Pascual & Barbier (70) borrow concepts from the VCM hypothesis. Testing of the VCM is difficult, however, because one is searching for a relatively small “resources effect” on fertility when there are at least a score of potentially confounding variables, and testing the direction of causality requires time series data on social and environmental variables, which is quite rare. Economists Filmer & Pritchett (71) found qualified support for the vicious circle hypothesis using detailed data from Pakistan on children’s time use, fire-wood collection activities, and recent fertility. They find that collection activities do absorb a substantial part of household resources and that children’s tasks are often devoted to collection activities. However, they could not establish a “fertility effect” on resource or land degradation. A longitudinal study in the western Chitwan Valley of Nepal (72) found that three measures of local resource depletion—the time to collect fodder, the increase in time required to collect fodder in the prior three years, and household’s dependence on public lands for fodder—were significantly and positively correlated with desired family size, even when controlling for household wealth and numerous other factors found to influence desired fertility. Yet, several other indicators of environmental decline had no significant relationship to either desired fertility or pregnancy outcomes, and the actual relationship to desired fertility depended in part on whether the respondents were men or women. Pascual & Barbier’s (70) modeling of shifting cultivation in the Yucatan found that among poor households, as population density increased, the response was extensification or a reduction in fallow periods, whereas among better-off households, labor was shifted to off-farm employment. Thus, although anecdotal evidence is abundant and development policy-making has been heavily influenced by VCM assumptions, there is only qualified support for the hypothesis in the few existing quantitative studies.
The Boserupian or intensification hypothesis has been tested in a number of studies spanning Africa, Asia, and Latin America. A frequently cited study by Tiffen et al. (6) examined changes in population density and agricultural productivity in Machakos District, Kenya. From 1930 to 1990, the population of Machakos District grew sixfold, from 240,000 to 1.4 million people, with a 1990 population density of 654/km2. The region is mountainous and semiarid (<500 mm rainfall a year), and in the 1930s, it was suffering already from soil erosion (mass wasting and gullies). The region was also isolated from national markets, and there were colonial restrictions on access to certain lands and crops. In the 1950s and 1960s, a new form of terracing was propagated by local work groups, agricultural systems shifted from livestock to intensive farming with emphasis on higher-value crops, feeder roads were built to market towns, and market towns developed with agricultural processing facilities and other small industries. By 1990, the value of agricultural production had doubled on a per capita basis. Many factors led to a positive outcome for this region, including infrastructure development, market growth, private investment, increasing management capacity and skills, self-help groups, food relief during drought, and secure land tenure. This study confirms the basic Boserupian hypothesis: increased food demand, a denser network of social and market interactions, labor-intensive agriculture and economies of scale helped to avert a Malthusian crisis. Yet even in this textbook study, other researchers working in the district found important social differentiation in livelihood improvements, land alienation, and government-imposed limitations on mobility—elements that tend to mar an otherwise rosy picture (73).
Mortimore (74) found similar “success stories” in three dryland areas of West Africa: Kano State in northern Nigeria, the Diourbel Region of Senegal, and the Maradi Department, Niger. Outcomes were assessed in four domains: improved ecosystem management, land investments, productivity, and personal incomes. Taking pains to point out that in none of these regions were indicators under all four domains positive, the author nevertheless found some common ingredients that resulted in improved or stable soil fertility and yields despite rapid population growth and high densities. These ingredients include markets for agricultural produce, physical infrastructure, producer associations, knowledge management, and incentives for investment and income diversification. He concludes that productivity enhancements respond to economic incentives and that the capacity of resource-poor farmers to invest in on-farm improvements should not be underestimated.
In Asia, there have also been successes, thanks largely to success of the “green revolution,” a package of improved seeds and agricultural inputs that resulted in higher yields (75). Turner & Shajaat Ali (21) studied time series data (1950–1986) for 265 households in six villages in Bangladesh. They found support for the induced intensification hypothesis, with yields largely keeping pace with or exceeding population growth despite high population densities (783 persons per km2). Soil conditions in Bangladesh are, on average, much better than in dryland Africa owing to deposits of alluvium during monsoon season flooding and therefore can support far higher densities. They posit that, as smallholders come in contact with the market economy, their redundant production is reduced, and their aspirations increase. Although cropping intensities on average increased significantly (in one village almost tripling), they also found increasing production disparities, with large land holders accounting for most of the surplus production, whereas the growing number of landless suffered shortfalls and malnutrition. They conclude that Bangladesh passed several threshold steps at points along its path towards intensification in which Malthusian outcomes of involution and stagnation might have occurred but were fortunately averted.
As these case studies make clear, population is but one among many factors that influence degradation or intensification. Other variables that are of crucial significance include institutional factors (land tenure regimes, local governance, resource access), market linkages (road networks, crop prices), social conditions (education, inequality of landholdings), and the biophysical environment itself (original soil quality, slopes, climatic conditions). Thus, it would appear that population growth is neither a necessary nor sufficient condition for either declines or improvements in agricultural productivity to occur. Population growth can either operate as a negative factor, increasing pressure on limited arable land, or a positive factor, helping to induce intensification through adoption of improved technologies and higher labor inputs. Where it does which depends on factors in the economic and institutional realms. This conclusion is supported by two ambitious meta-analyses of studies that looked at dryland degradation (or desertification) and agricultural intensification (76, 77). The authors reject both single-factor causation and irreducible complexity but propose instead that a limited number of underlying driving forces, including population, and proximate causes are at work to produce either degradation or intensification.
Although population can perhaps be discounted as the only relevant variable, there is little doubt that rapid population growth in poor rural areas with fragile environments can be a complicating factor in the pursuit of sustainable land use, especially because policies and markets are rarely aligned in such a way as to produce the most favorable results. Furthermore, trends on the basis of past precedents can only be extrapolated with caution, because the exact locations of thresholds in any given system are still largely unknown (21). One important advance for studies in this area will be the development of better maps of soil quality and land degradation with the aid of remote sensing and local soil samples, as at least part of the debate over population’s impact can be explained by differing interpretations of what constitutes degradation and by a paucity of empirical evidence for the relationship.
Abstraction and Pollution of Water Resources
The water cycle ties together life processes. It is fundamental to the biochemistry of living organisms; ecosystems are linked and maintained by water; it drives plant growth; it is habitat to aquatic species; and it is a major pathway of sediment, nutrient, and pollutant transportation in global biogeochemical cycles (78). Population-environment researchers have not dedicated the same level of attention to population dynamics and water resources as they have to research on land-cover change, agricultural systems, or climate change. Yet there are clear relationships between population dynamics and freshwater abstraction for agricultural, domestic, and industrial uses, as well as emission of pollutants into water bodies.
Human settlement is heavily predicated upon the availability of water. A map of global population distributions closely tracks annual rainwater runoff, with lower densities in the most arid regions and as well as the most water abundant, such as the Amazon and Congo Basins. Whereas the former areas are water constrained for agriculture, in the latter areas, year-round rainfall in excess of 2000 mm has rendered these environments less favorable for agriculture (owing to soil leaching and oxidation) and more favorable for human and livestock diseases.
At the global level, irrigation water for agriculture is the biggest single user (about 70% of water use), followed by industry (23%) and domestic uses (8%) (79). If “green water” is added to the mix (water that feeds rainfed crops), then crop production far and away outstrips other water uses. As demand for food increases with growing populations and changing tastes (including growing demand for animal versus vegetable protein with its far greater demands for water), it is expected that water diversions for agriculture will only increase. Today, humanity is estimated to use 26% of terrestrial evapotranspiration and 54% of accessible runoff (80). Falkenmark & Widstrand (81) established benchmarks for water stress of between 1000 and 1700 m3 per person, water scarcity of between 500 and 1000 m3 per person, and absolute scarcity of less than 500 m3 per person. Northern and southern Africa and the Middle East already suffer absolute scarcity. As population grows and water resources remain more or less constant, many countries in the rest of Africa are projected to fall below 1000 m3 per person (82).
Perhaps because such water resources are hidden underground, groundwater resource depletion could potentially remove some agricultural areas from the map. Although it is well known that some Arab countries rely on fossil water for wheat production, less recognized is that 70% of Chinese and 45% of U.S. irrigation is based on nonrenewable water resources (C. Vorosmarty, EM Douglas, personal communication). Groundwater levels in India have been dropping for more than a decade owing to the unregulated tapping of aquifers (83), rendering some semi-arid regions vulnerable to shortages. A study in Karnataka State, India, identified a major shift from surface to groundwater use in the past decades and found that groundwater use is highly inequitable; large farmers possessing 12–16 ha of land make up only 8% of all farmers but consume 90% of groundwater (84). In the lower delta of the Ganges-Brahmaputra Basin, upstream diversions at the Farakka Barrage, rather than local demands for irrigation water, appear to be causing dry season groundwater deficits and intrusion of the saline front, illustrating how complex basin-wide hydrological connections complicate the attribution of population impacts (85).
Other studies at the local level reveal a similarly complicated picture. Research in the Mwanza region of western Tanzania finds that accessible runoff varies significantly across a relatively small area and that population density closely tracks available water (86). Migrants to towns were generally less likely to have access to water from standpipes and more likely to rely on unimproved wells. Rural-urban migration is not correlated to relative water scarcity in places of origin but rather to proximity to roads and to towns. The researchers conclude that high fertility—a traditional adaptation to peak labor demands during the short cropping season—increases the problems of water access and supply maintenance in agricultural and domestic spheres. But they also note that gloomy prognoses about future water shortages often fail to acknowledge that large portions of developing country populations never have had the kind of access to water, or levels of consumption, deemed necessary by international bodies.
In the Pangani Basin of northeastern Tanzania, a complex set of factors is leading to water conflicts (87). Population is one factor: Owing to high fertility and migration, rural population is doubling every 20 years, and the population of towns is doubling every 10 years. But other factors include water extraction and land alienation for export flower production and protected areas, growth and mobility of livestock herds, declining summer runoff from glaciers on Mount Kilimanjaro owing to global warming, and hydroelectricity generation. The greatest conflict is between farmers and pastoralists, as farmers progressively moved into areas previously considered too marginal for agriculture and pastoralists were squeezed by restrictions on grazing areas owing to newly established protected areas. In recent years, the pressure on land has led to stresses on water and other resources, leading to heavy out-migration from the basin.
Researchers in the densely populated Sao Paulo State in Brazil examined water resources in the Piracicaba and Capivari River Basins within the Campinas Administrative Region (AR) (88). Campinas is Brazil’s fourteenth largest city, as well as its third largest industrial center, and an important agricultural region as well. The Metropolitan Region of Campinas (the 19 core municipalities of the AR) saw high, though declining, average annual population growth rates during the 1970–2000 period: 6.5% (1970–1980), 3.5% (1980–1991); and 2.5% (1991–2000). The authors find that problems in the form of urban growth and the patterns of population distribution during these three decades have accentuated water quality problems because the rapidity and low density of growth meant that water supply and sanitation infrastructure could not keep up. By mid-1995 only 5% of waste was treated before reentering streams, and large reaches of the Piracicaba and Capivari River Basin tributaries were deemed of poor quality. Water supply infrastructure (mostly surface reservoirs as groundwater is scarce) did not kept pace with population growth, and the situation was reported as critical as of the mid-1990s. In response, state water basin agencies are applying some institutional solutions such as fees for water withdrawals and restrictions on residential development, as well as some technical ones, particularly the treatment of waste waters.
In summary, as in other areas, the relationship between population dynamics and water resources is complex. At the aggregate level, other things being equal, population growth most assuredly does reduce per capita water availability. It is in this light that the Global International Waters Assessment listed population growth first in a series of root causes of the “global water crisis” (89). Yet there is more to population change than growth alone, and rarely are other factors equal, so the specific impacts of population dynamics on water often come down to a complex array of place-specific factors that relate to economic and climatic changes, agricultural and industrial technologies, sewage treatment, and institutional mechanisms, to name but a few. One of the challenges to research in this area is the common property nature of water resources, and another challenge is caused by rapid regulatory changes as water resources become scarcer, which alters the institutional context. The field could use more basin or watershed studies to understand how variables such as population and climate change may affect future water availability and required institutional responses (90). Basin-level population-development-environment modeling would also help understand and resolve competition between urban and rural water uses as the world becomes more urbanized (91).
Coastal and Marine Environments
From the earliest times, the preponderance of global economic activity has been concentrated in the coastal zone (92), with settlements often growing on the continental margins to take advantage of overseas trade and easy access to the resources of the rural hinterlands. As a result, the coastal zone has attracted large and growing populations, with much of their growth attributable to migration rather than natural increase (93). Today, 10% of the world’s population lives at less than 10 m above sea level (even though this area only accounts for 2.2% of the world’s land area), and coastal zones have higher population densities than any other ecologically defined zone in the world (39, 94). Coastal and marine environments are very important for human health and well-being, and they are also quite vulnerable to anthropogenic impacts. Yet, until recently most population-environment research has focused on terrestrial ecosystems, possibly because the human “footprint” on coastal and marine ecosystems is harder to discern.
Not surprisingly, over half of the world’s coastlines are at significant risk from development-related activities (95), and the potential (and realized) environmental damage is substantial. Population growth is often named as the driver of coastal and marine environmental problems, whereas proximate causes can be traced to specific practices (96). A recent study highlights how the Kuna population (an indigenous population in Caribbean Panama) has practiced coral mining and land-filling for decades in response to population growth (97). Since 1970, live coral cover declined 79%, and at the same time, the Kuna population increased by 62%. The Kuna gradually enlarged their island landmass to adjust for their growing population by building coral walls out into the water and then filling in the enclosed areas with corals, sea-grass, and sand. In addition to direct loss of coral reef, consequences include coastal erosion and a local increase in sea level. This example provides a clear and direct link between population growth and coastal degradation.
Population growth can lead to many other coastal and marine environmental disturbances. For instance, tropical mangroves are being converted to fish and shrimp aquaculture farms, which undermines coastal protection and decreases natural habitat that many fish species use for reproduction. Expanding coastal cities undermine natural protection from storms and hurricanes as well as increase pollution and runoff. Additionally, untreated sewage and agricultural runoff continue to be a worldwide problem. Although listed as a driver, like other issues, the impact of population size and growth depends on many other factors such as the sensitivity of coastal systems to stress, local institutions, and global markets. For example, demand for shrimp is the ultimate driver of mangrove loss, and sewage treatment systems and no-till agriculture could significantly reduce nutrient loading in coastal areas.
The relationship between human activities and environmental impacts are hard to assess and regulate in coastal and marine environments because the environmental resources are almost always governed by common property resource (CPR) management systems, whereas terrestrial environments are generally managed by the government or private sector. CPR management systems may be especially vulnerable to disruption caused by in-migration or urbanization. However, the social and economic context largely determine whether in-migration and population pressure disrupt the CPR system and thus cause environmental degradation (98–100). Thus, a significant recurring theme in this research is that the social and economic context in which the population is changing as well as when, how, and with whom people interact is more important in determining the impact on the environment than simply demographic change (101, 102).
Studies in developing countries on migration and the marine environment have focused on a mediating variables approach, such as how technology, local knowledge, social institutions of kinship or marriage, and markets mediate the role of population in resource extraction and consequent environmental degradation or enhancement. For example, some work has hypothesized that migrants misuse resource extraction technologies, which leads to environmental degradation (103). In a coastal Brazilian population, technological change imposed by outsiders who lacked knowledge of the ecological and social context of the community contributed to decreased ecological resilience (104), and rapid in-migration and technological changes in sea cucumber fishing techniques in the Galapagos led to a collapse in the sea cucumber industry (105). In both cases, the results seem to be a function of the migrants’ limited local knowledge as well as expansionist attitudes and short-term time horizons for profiting from the extraction of coastal and marine resources.
Thus, population-environment researchers have begun to incorporate other social theories such as social capital and migrant incorporation to understand when population pressures do not necessarily degrade the environment (106). Most studies have found that, in systems with strong land tenure or social capital, migrants do not disrupt the environment and are able to develop local knowledge that mitigates environmental impacts (107–109). A case study in the Solomon Islands contests the notion that sea tenure regimes are weakened by in-migration and population growth. Rather, potentially negative impacts of population pressure on the environment are diminished significantly with greater reciprocal ties among close kin or neighbors (110, 111). Similarly, intermarriage between a migrant and a nonmigrant in Sulawesi, Indonesia, has been shown to mitigate the otherwise negative association between migrant households and coral reef degradation (106).
Migration has been the most studied component of population dynamics in coastal and marine environments. Yet, urbanization and tourism are other primary human drivers affecting coastal ecosystem quality (112, 113). Fourteen of the worlds largest 17 cities are located along a coast; this affects freshwater flows to coastal estuaries, sewage emissions, and ecological processes at the land-sea interface (114). Also, without careful planning in anticipation of a growing tourist market, cultural and ecological resources may be over-exploited, resulting in unsustainable development, as is the case in Turkey (115).
Human impacts on coastal and marine environments are not a simple function of population size or density. As the aforementioned studies suggest, technology, knowledge systems, social cohesion, common property systems, migrant incorporation, and the economic and ecological context in which these interactions take place all play an important role in population and environment research, especially in developing countries. Nonetheless, coastal and marine environments continue to be among the most threatened ecosystems in the world, owing in part to the sheer scale of detrimental human activities associated with urbanization along the coasts, continued population growth, and a growing number of tourists in search of coastal amenities.
An unresolved issue in this area of research—as in the case of LUCC research—is how to spatially and temporally link populations and human activity to a specific environmental outcome. This is especially difficult in marine and coastal ecosystems because environmental boundaries are fluid. Also important is the impact of local and global consumption on marine and coastal environments. For instance, per capita consumption of seafood is high in many traditionally seafaring countries even though population sizes are low (116), and specialized tastes for rare species can have dramatic impacts on fish stocks (117). Further research is needed to assess how population-environment linkages in marine and coastal areas are influenced by the global food trade connecting consumers and producers from opposite sides of the world.
Energy, Air Pollution, and Climate Change
Even when they are connected to the electric grid, some two billion poor people in the developing world still largely rely on biomass to meet their energy needs. That leaves approximately 4.7 billion people with more energy-intensive lifestyles who consume, with little help from the world’s poorest, the energy equivalent of 77 trillion barrels of oil a year (118).3 More than 80% of global energy consumption is derived from fossil fuels (119), and it is this dependence on fossil energy that is responsible for the release of the greenhouse gases and airborne pollutants that are altering atmospheric composition and processes on a global scale. As concern mounts over the health impacts of urban air quality (particularly in developing countries) and the potential adverse effects of climate change across multiple systems and sectors, population-environment researchers have paid particular attention to understanding the demographic drivers of energy consumption. Although it is clear that there are vast differences in consumption levels (per capita energy consumption in the United States is 48 times what it is in Bangladesh and 4.7 times the world average), it would be wrong to suggest that population variables are irrelevant. Hence, we review a number of empirical studies that examine population-energy linkages in a systematic and quantitative manner.4
In studies of energy consumption researchers have found that it is more appropriate to use the household rather than individuals as the unit of analysis because a large portion of energy consumption related to space conditioning (heating and air conditioning), transportation, and appliance use is shared by household members. This sharing results in significant economies of scale, with large households generally showing lower per capita energy use than small ones (29, 120). Energy studies have identified a range of household characteristics as key determinants of travel patterns (121–123) and of other types of residential energy demand, such as for heating, cooking, and operating domestic appliances (124–127). In a pioneering study, MacKellar et al. (128) used the IPAT identity to decompose the annual increase in energy consumption of the more developed regions during the period 1970–1990. They found that, because growth in the number of households outpaces population growth owing to trends in fertility, divorce, and ageing, growth in household numbers accounted for 41% of the total increase in energy consumption, whereas population growth accounted for only 18%. However, this study did not take into account the lower energy requirements of smaller households, so it likely exaggerated the contribution of the growth in household numbers to energy use.
O’Neill & Chen (129) drew on household survey data to quantify the influence of household size, age, and composition on residential and transportation energy use in the United States and found that changes in household size have caused 14% of the increase in per capita energy use over the past several decades. Lenzen et al. (130) assessed the importance of various demographic characteristics on household energy demand in Australia, Brazil, Demark, India, and Japan, and they found similar patterns across countries: The average age of residents is positively associated with per capita energy consumption, whereas household size and urban location are negatively associated. To explore the importance of adopting adequate demographic variables in understanding transport-related energy consumption, Prskawetz et al. (131) combined a cross-sectional analysis of car use in Austria with detailed population/household projections and tested the sensitivity of projections of future car use across a wide range of households by size, age, and sex of householder and the number of adults and children. They found that car use will likely increase by 20% in the period 1996–2046 if the number of households is the unit of analysis. However, it will only increase by 3% if one applies a composition that differentiates households by size, age, and sex of the householders. Therefore, household characteristics can impact aggregate demand for car use via differences in demand across households as well as likely changes in household composition.
In studying demographic impacts (via energy consumption) on air pollution, scientists have identified a number of important factors that jointly determine pollutant emissions, including the familiar elements of the IPAT identity—population, affluence, and technology as reflected in energy and emissions intensities (132). Selden et al. (133) analyzed the reduction of U.S. major air pollution emissions from 1970 to 1990 and found that changes in economic scale, economic composition, energy mix, energy intensity, and emissions intensity all played important roles. In quantifying the impacts of population on air pollution, researchers have reached different conclusions depending on which pollutants are under study, in which locations, at what scale, and for which time periods. For instance, a study of California counties shows that population size significantly contributes to the increase of the reactive organic gases NOx and CO and has little impact on PM10 and SOx, which are derived more from production activities (134). Population size shows no significant relation to ground-level ozone because ozone is very difficult to measure at specific sites owing to its nature as a diffuse secondary pollutant (135). In research using national-level data, researchers found an almost linear positive correlation between population size and CO2 emissions (128, 132, 136, 137) and an inverted U-shaped curve for SO2 (136). However, a more recent study of Canadian provinces over the period 1970–2000 suggests that population size has an inverted U-shaped curve with CO2 emissions as well, which is at odds with previous literature investigating these variables for other regions and time periods (138). The different patterns of impacts may reflect the nature of complicated interactions between different pollutants and regional geographic/climatic conditions (139, 140), income, and technological levels (139, 141).
The same inconsistencies in the relationship between population size and emissions of various pollutants are in evidence when examining other population-related variables. Cramer (134) in his study of California counties and Cole & Neumayer (136) in their cross-national studies found that other variables such as the percent of population that are migrants, age composition, household size, and level of urbanization have the same basic relationship as overall population size on emission levels of each of the pollutants they studied. However, caution should be used in interpreting these results because the studies only cover short time periods (10 to 20 years) in which there were only small changes in the demographic variables.
Because of the complexity of population interactions as well as political issues, population issues were not considered in formulation of the Kyoto Protocol (142) and have also been largely excluded from the Intergovernmental Panel on Climate Change (IPCC) assessment reports (143), although population projections are an integral part of the Special Report on Emissions Scenarios (SRES) (144). The original emissions scenarios were constructed in 1996 using population projections with a base year of 1990. Although the projections used in the SRES were largely consistent with actual population sizes for the 1990–2005 period, the projections to 2050 and beyond were higher than more recent projections (see the text, above, on global trends in population and consumption) (11, 145, 146). Therefore, even though the 1996 scenarios continue to serve as a primary basis for assessing future climate change and possible response strategies, the Fourth Assessment Report of the IPCC is based on slightly lower population projections than the Third Assessment Report under the A2 scenario, which describes an economically divided world with slow technological progress and high population growth. Consideration of demographic factors beyond population size, such as changes in age structure, urbanization, and living arrangements, which as discussed above are important in modeling future energy use, are not accounted in the SRES population assumptions. Making progress in this area requires a better understanding of the scope for future demographic change as well as methods for including demographic heterogeneity within energy-economic growth models used for emissions scenario development.
Simultaneous and consistent projections of population, urbanization, and households are a challenging demographic tasks (147). Recently, Dalton et al. (148) introduced heterogeneous households into a general equilibrium population-environment-technology model of the U.S. economy. Because different types of households have unique demands for goods, capital stock, and labor supply, and these characteristics have direct and indirect implications for energy demand, they were incorporated into cohorts by age groups (or “dynasties”). These dynamics and other relationships implied by household projections create nonlinear interacting effects that influence each dynasty’s future saving and consumption decisions. Their research shows that including age heterogeneity among U.S. households reduces emissions by almost 40% in the low-population scenarios by year 2050, and effects of aging on emissions can be as large as, or larger than, effects of technical change in some cases. Those effects are believed to be much larger for the developing world, where more significant demographic changes such as population growth, aging, household nuclearization, and urbanization are occurring.
One of the reasons natural scientists have found population to be so appealing as a human dimension of environmental change is that data are readily available (in contrast to other human variables such as values, culture, and institutions), projections are reasonably reliable (149), and population can be treated in models in a manner that is analogous to all the other quantitative variables. This has promoted something of a reductionist view of population-environment interactions. Fortunately, a growing number of natural scientists are beginning to appreciate that humans interact with the environment in more ways than their raw numbers often imply. Populations are composed of people who collectively form societies, and people and societies cannot easily be reduced to food and material demands that result in some aggregate impact on the environment.5 This makes human societies at once messy for modeling and fascinating to study. The new understanding builds on the concept of coupled human-environment systems, which are more than the sum of their parts (150, 151).
In the human-environment system, the impacts are not unidirectional but reciprocal. For example, the environmental change impacts on morbidity and mortality are a growing area of interest, and some have sought to close the circle by looking at how environmentally induced mortality may affect population projections (2). There is also growing research on the health impacts of landscape or climatic changes on humans, in the one instance through the creation of mosquito breeding habitats that contribute to malaria (152), and in the other through heat stress or famine (153). Research on the human-environment system also takes advantage of new data sources (remote sensing, biophysical data, as well as georeferenced household surveys), new technologies (high-powered computers, geographic information systems, spatial statistics), and new models (agent-based, multilevel, and spatially explicit modeling). Much of the research reviewed in this chapter has sought to deconstruct population into its component parts and to understand how human social institutions in all their complexity (e.g., markets, policies, communities) mediate the impact of population variables on the use of resources, waste generation, and environmental impacts. Thus, they could be said to fit into this growing understanding of the human-environment system.
Much population-environment research, whether at the local or global scales, is motivated by a broader concern for sustainability. Underlying some of the research, and contributing to some of the controversy, has been a concern for distributional justice in two forms: that the 5.4 billion citizens of developing countries might be able to raise their living standards and hence their consumption levels from their previously low levels and that the costs of biodiversity conservation and climate change adaptation not be unfairly borne by the poorest. Whether research proves that population dynamics have a dominant or negligible effect on environmental outcomes in each of the domains we surveyed, it is still left to human societies to address these inequities in consumption and costs as well as to seek long-term solutions. Here, research on culture, consumption, values, institutions, and alternative industrial and food systems will add to what is known about the demographic dimension as societies seek to transition to sustainable systems (10, 154).
Although we have sought to objectively review the literature rather than take a normative stance concerning the environmental impacts associated with population dynamics, at the global scale there is no question that humanity faces significant challenges in the coming decades owing to the scale and pace of changes in human numbers, population distribution, and consumption patterns. To quote Cohen’s definitive study on the global carrying capacity, “The Earth’s human population has entered and rapidly moves deeper into a poorly charted zone where limits on human population size and well-being have been anticipated and may be encountered” (2, p. 11). In recent decades, scientists have increasingly warned of the potential to reach the upper limits of the planet’s productive, absorptive, and recuperative capacities (155). A challenge for micro- and mesoscale researchers is to understand how changes at the local and national scale relate to global-scale changes and how, in turn, their research can inform policies and programs at these lower scales that will attenuate environmental impacts at all levels.
There is more to population dynamics than population size and growth. Recent research has illuminated the ways in which a number of population variables–age and sex composition, household demographics, and the elements of the population balancing equation (fertility, mortality and migration)–are related to environmental change.
Most demographers and many other social scientists subscribe to a mediating variable theory, which states that population dynamics affect the environment through other variables such as culture, consumption levels, institutions, and technology.
Across the environmental issues covered in our review, population dynamics usually act in concert with other significant factors such as local institutions, policies, markets, and cultural change. Teasing out the relative contribution of each factor can often be difficult.
The scale of analysis can significantly affect findings concerning the role of population dynamics in environmental change.
Evidence for the impacts of population on land and resource degradation has been mixed in part because time series data at appropriate scales and measurements of appropriate variables are rare and because the population “signal” is often difficult to isolate from other signals.
Both freshwater resources and coastal and marine ecosystems are often managed as common property resources (CPRs); hence levels of resource degradation or depletion depend more on the existence of effective management systems than on population variables per se.
In research on population and energy use, the household has been found to be a more useful unit of analysis than the individual, and population-environment researchers have made major strides in understanding how household size, composition, and income are related (via energy use) to environmental impacts.
Emerging understanding of complex human-environment systems is informing work in the area of population and the environment, and vice versa.
Greater exploration of the linkages between micro- (farm or household level) and macroscale (global) processes manifested at meso- (subnational) scales in population-environment research across the different issue areas is needed.
Careful microscale longitudinal studies measuring population variables, household consumption, biophysical variables, institutional arrangements, and technologies employed over time should be conducted.
Given the environmental footprints of urban areas on rural hinterlands, one unresolved issue relates to the impact of population spatial distribution. For example, what would environmental impacts be if the same population were spread more evenly across the landscape rather than concentrated in urban areas?
Population-environment researchers could contribute to better understanding current consumption levels and the effects of future aspirations of the growing middle classes of Asia and Latin America as they relate to the sustainability transition.
Advances in demographic modeling are needed to develop a new population/household model with moderate data requirements, manageable complexity, explicit representation of demographic events, and output that includes sufficient information for population-environment studies.
A new generation of IPAT modeling is needed that explicitly accounts for the interactions among the IPAT terms, including the reciprocal impacts of environmental changes on population dynamics, and that is made part of integrated assessment modeling.
Future research could explore the increase in human mobility and collapse of geographical space as it affects population-environment relationships.
For most people, thinking about health and health care is a very personal issue. Assuring the health of the public, however, goes beyond focusing on the health status of individuals; it requires a population health approach. As noted in Chapter 1, America's health status does not match the nation's substantial health investments. The work of assuring the nation's health also faces dramatic change, systemic problems, and challenging societal norms and influences. Given these issues, the committee believes that it is necessary to transform national health policy, which traditionally has been grounded in a concern for personal health services and biomedical research that benefits the individual. Such repositioning will affirm and expand existing commitments to reflect a broader perspective. Approaching health from a population perspective commits the nation to understanding and acting on the full array of factors that affect health.
To best address the social, economic, and cultural environments at national, state, and local levels, the nation's efforts must involve more than just the traditional sectors—the governmental public health agencies and the health care delivery system. As has been outlined in the preceding pages, what is needed is the creation of an effective intersectoral public health system. Furthermore, the efforts of the public health system must be supported by political will—which comes from elected officials who commit resources and influence based on evidence—and by “healthy” public policy—which comes from governmental agencies that consider health effects in developing agriculture, education, commerce, labor, transportation, and foreign policy.
This chapter describes the rationale behind a transformed approach to addressing population health problems. This approach identifies key determinants of the nation's health and presents evidence for their consideration in developing effective national strategies to assure population health and support the development of a public health system that blends the strengths and resources of diverse sectors and partners (IOM, 1997).
A POPULATION PERSPECTIVE
For nations to improve the health of their populations, some have cogently argued, they need to move beyond clinical interventions with high-risk groups. This concept was best articulated by Rose (1992), who noted that “medical thinking has been largely concerned with the needs of sick individuals.” Although this reflects an important mission for medicine and health care, it is a limited one that does little to prevent people from becoming sick in the first place, and it typically has disregarded issues related to disparities in access to and quality of preventive and treatment services. Personal health care is only one, and perhaps the least powerful, of several types of determinants of health, among which are also included genetic, behavioral, social, and environmental factors (IOM, 2000; McGinnis et al., 2002). To modify these, the nation and the intersectoral public health system must identify and exploit the full potential of new options and strategies for health policy and action.
Three realities are central to the development of effective population-based prevention strategies. First, disease risk is currently conceived of as a continuum rather than a dichotomy. There is no clear division between risk for disease and no risk for disease with regard to levels of blood pressure, cholesterol, alcohol consumption, tobacco consumption, physical activity, diet and weight, lead exposure, and other risk factors. In fact, recommended cutoff points for management or treatment of many of these risk factors have changed dramatically and in a downward direction over time (e.g., guidelines for control of “hypertension” and cholesterol), in acknowledgment of the increased risk associated with common moderately elevated levels of a given risk factor. This continuum of risk is also apparent for many social and environmental conditions as well (e.g., socioeconomic status, social isolation, work stress, and environmental exposures). Any population model of prevention should be built on the recognition that there are degrees of risk rather than just two extremes of exposure (i.e., risk and no risk).
The second reality is that most often only a small percentage of any population is at the extremes of high or low risk. The majority of people fall in the middle of the distribution of risk. Rose (1981, 1992) observed that exposure of a large number of people to a small risk can yield a more absolute number of cases of a condition than exposure of a small number of people to a high risk. This relationship argues for the development of strategies that focus on the modification of risk for the entire population rather than for specific high-risk individuals. Rose (1981) termed the preventive approach the “prevention paradox” because it brings large benefits to the community but offers little to each participating individual. In other words, such strategies would move the entire distribution of risk to lower levels to achieve maximal population gains.
The third reality, provided by Rose's (1992) population perspective, is that an individual's risk of illness cannot be considered in isolation from the disease risk for the population to which he or she belongs. Thus, someone in the United States is more likely to die prematurely from a heart attack than someone living in Japan, because the population distribution of high cholesterol in the United States as a whole is higher than the distribution in Japan (i.e., on a graph of the distribution of cholesterol levels in a population, the U.S. mean is shifted to the right of the Japanese mean). Applying the population perspective to a health measure means asking why a population has the existing distribution of a particular risk, in addition to asking why a particular individual got sick (Rose, 1992). This is critical, because the greatest improvements in a population's health are likely to derive from interventions based on the first question. Because the majority of cases of illness arise within the bulk of the population outside the extremes of risk, prevention strategies must be applicable to a broad base of the population. American society experienced this approach to disease prevention and health promotion in the early twentieth century, when measures were taken to promote sanitation and food and water safety (CDC, 1999b), and in more recent policies on seat belt use, unleaded gasoline, vaccination, and water fluoridation, some of which are discussed later in this chapter.
The committee recognizes that achieving the goal of improving population health requires balancing of the strategies aimed at shifting the distribution of risk with other approaches. The committee does, however, endorse a much wider examination, and ultimately the development, of new population-based strategies. Three graphs illustrate different models for risk reduction (see Figure 2–1).
Models for risk reduction. SOURCE: Data for current distribution from Schwartz and Woloshin, 1999.
These hypothetical models assume etiological links exist among all exposures and disease outcomes. Figure 2–1a shows the effects of an intervention aimed at reducing the risk of those in the highest-risk category. In this example, people with the highest body mass index (BMI)1 are at in creased risk for cardiovascular heart disease and a plethora of chronic illnesses. Intervening medically, for example, to decrease risk (by lowering levels of obesity, as measured by BMI) ultimately decreases the proportion of the population with the highest BMIs. Such measures among very high-risk individuals may even be endorsed in cases where the “intervention” itself carries a substantial risk of poor outcome or side effects. However, use of such an intervention would be acceptable only in those whose medical risk was very high. Moreover, interventions in high-risk groups may have a limited effect on population outcomes because the greater proportion of those with moderate risk levels may ultimately translate into more chronic disease or other poor health outcomes.
Figure 2–1b illustrates Rose's classic model whereby the greatest benefit is achieved by shifting the entire distribution of risk to a lower level of risk. Because most people are in categories of moderately elevated risk as opposed to very high risk, this strategy offers the greatest benefit in terms of population-attributable risk, assuming that the intervention itself carries little or no risk. The hypothetical example shows what might occur if social policies or other population-wide measures were adopted to promote small decreases in weight in the general population. The committee embraces this kind of model of disease prevention in the case of policies such as seat belt regulation and the reduction of lead levels in gasoline.
The final hypothetical model (Figure 2–1c), although not discussed by Rose explicitly, illustrates a reduction in the distributions of those at highest and lowest risk with no change in the distribution of those with a mean level of risk. This model is appropriate for illustrating phenomena relating to inequality, where redistribution of some good (e.g., income, education, housing, or health care) reduces inequality without necessarily changing the mean of the distribution of that good. One hypothetical example is the association between low income and poor health. In many cases, there is a curvilinear association between these goods and health outcomes, with decreased health gains experienced by those at the upper bounds of the distribution. For example, data on income suggest that there are large differences in the health gains achieved per dollar earned for those at the lower end of the income distribution and fewer differences in the health gains achieved per dollar earned for those at the upper end. Thus, the curvilinear association, if it were a causal one, would suggest that substantial gains in population-level health outcomes may be achieved by a redistribution of some resources without actual changes in the means.
These graphs help to illustrate three different strategies for improving the health of the population. The nation has often endorsed the first strategy without a critical examination of the other two, especially the second one. The American public has grown accustomed to seeing differences in exposures to risk, both environmental and behavioral, and disparities in health outcomes. Acknowledging these gradients fully will help develop true population-based intervention strategies and help the partners who collaborate to assure the public's health move to take effective actions and make effective policies.
Understanding and ultimately improving a population's health rest not only on understanding this population perspective but also on understanding the ecology of health and the interconnectedness of the biological, behavioral, physical, and socioenvironmental domains. In some ways, conventional public health models (e.g., the agent–host–environment triad) have long emphasized an ecological understanding of disease prevention. Enormous gains in the control and eradication of infectious diseases rested upon a deep understanding of the ecology of specific agents and the power of environmental interventions rather than individual or behavioral interventions to control disease. For example, in areas where sanitation and water purification are poor, individual behaviors, such as hand washing and boiling of water, are emphasized to reduce the spread of disease. However, when environmental controls become feasible, it is easy to move to a more “upstream”2 intervention (like municipal water purification) to improve health. The last several decades of research have resulted in a deeper understanding not only of the physical dimensions of the environment that are toxic but also of a broad range of related conditions in the social environment that are factors in creating poor health. These social determinants challenge the discipline of public health to more fully incorporate them.
Over the past decade, several models have been developed to illustrate the determinants of health and the ecological nature of health (e.g., see Dahlgren and Whitehead , Evans and Stoddart , and Appendix A). Many of these models have been developed in the United Kingdom, Canada, and Scandinavia, where population approaches have started to shape governmental and public health policies. The committee has built on the Dahlgren-Whitehead model—which also guided the Independent Inquiry into Inequalities in Health in the United Kingdom—modifying it to reflect special issues of relevance in the United States (see Figure 2–2). This figure serves as a useful heuristic to help us think about the multiple determinants of population health. It may, for instance, help to illustrate how the health sector, which includes governmental public health agencies and the health care delivery system, must work with other sectors of government such as education, labor, economic development, and agriculture to create “healthy” public policy. Furthermore, the governmental sector needs to work in partnership with nongovernmental sectors such as academia, the media, business, community-based organizations and communities themselves to create the intersectoral model of the public health system first alluded to in the 1988 Institute of Medicine (IOM) report and established in this report as critical to effective health action.
A guide to thinking about the determinants of population health. NOTES: Adapted from Dahlgren and Whitehead, 1991. The dotted lines between levels of the model denote interaction effects between and among the various levels of health determinants (Worthman, (more...)
Most models of health determinants identify macro-level conditions and policies (social, economic, cultural, and environmental) as potent forces in shaping midlevel (working conditions, housing) and proximate (behavioral, biological) determinants of health. Macro-level or upstream determinants (such as policies and societal norms) and micro-level determinants (such as sex or the virulence of a disease agent) interact along complex and dynamic pathways to produce health at a population level. As mentioned above, exposures at the environmental level may have a greater influence on population health than individual vulnerabilities, although at an individual level, personal characteristics including genetic predispositions interact with the environment to produce disease. For instance, smoking is a complex biobehavioral activity with both significant genetic heritability and nongenetic, environmental influences, and many studies have shown an interaction between smoking and specific genes in determining the risk of developing cardiovascular disease and cancers. It is also important to note that developmental and historical conditions change over time at both a societal level (e.g., demographic changes) and an individual level (e.g., life course issues) and that disease itself evolves as agents change in virulence.
In the pages that follow, the committee provides a concise discussion of the key determinants that constitute the ecology of health, including environmental and social determinants, and elaborates in more detail on the social influences on health. This decision was made in recognition of a longer history in studying the ways in which environment shapes population health.
THE PHYSICAL ENVIRONMENT AS A DETERMINANT OF HEALTH
At least since the time of Hippocrates' essay “Air, Water and Places,” written in 400 B.C.E., humans have been aware of the many connections between health and the environment. Improved water, food, and milk sanitation, reduced physical crowding, improved nutrition, and central heating with cleaner fuels were the developments most responsible for the great advances in public health achieved during the twentieth century. These advantages of a developed nation are taken for granted, but in fact, they could deteriorate without adequate support of the governmental public health infrastructure.
Environmental health problems, historically local in their effects and short in duration, have changed dramatically within the last 25 years. Today's problems are also persistent and global. Together, global warming, population growth, habitat destruction, loss of green space, and resource depletion have produced a widely acknowledged environmental crisis (NRC, 1999). These long-term environmental problems are not amenable to quick technical fixes, and their resolution will require community and societal engagement. At the local and community levels, environmental issues are equally complex and are also related to a range of socioeconomic factors. A brief look at some of the evidence on environmental determinants of health may help shed some light on why health is not equally shared.
The importance of “place” to health status became increasingly clear in the last decades of the twentieth century. The places in which people work and live have an enormous impact on their health. The characteristics of place include the social and economic environments, as well as the natural environment (e.g., air, water) and the built environment, which may include transportation, buildings, green spaces, roads, and other infrastructure (IOM, 2001b). Environmental hazards in workplaces and communities may range from tobacco smoke to pesticides to toxic housing. Rural areas may present increased health risks from pesticides and other environmental exposures, whereas some environmental threats to health can occur because of urban living conditions.
More than three-quarters of Americans live in urban areas (Bureau of the Census, 1993). Although rural Americans experience certain health-related disadvantages (e.g., health care access issues due to transportation and availability) (Slifkin et al., 2000; NCHS, 2001), some of the health effects of the inner city (i.e., decay and crime) are often dramatic and may be related to broader social issues. The “urban health penalty”—the “greater prevalence of a large number of health problems and risk factors in cities than in suburbs and rural areas” (Leviton et al., 2000: 863)—has been frequently discussed and studied (Lawrence, 1999; Freudenberg, 2000; Geronimus, 2000). A variety of political, socioeconomic, and environmental factors shape the health status of cities and their residents by influencing “health behaviors such as exercise, diet, sexual behavior, alcohol and substance use” (Freudenberg, 2000: 837). The negative environmental aspects of urban living—toxic buildings, proximity to industrial parks, and a lack of parks or green spaces, among others—likely affect those who are already at an economic and social disadvantage because of the concentration of such negative aspects in specific pockets of poverty and deprivation (Lawrence, 1999; Maantay, 2001; Williams and Collins, 2001). Urban dwellers may experience higher levels of air pollution, which is associated with higher levels of cardiovascular and respiratory disease (Hoek et al., 2001; Ibald-Mulli et al., 2001; Peters et al., 2001). People who live in aging buildings and in crowded and unsanitary conditions may also experience increased levels of lead in their blood, as well as asthma and allergies (Pertowski, 1994; Pew Environmental Health Commission, 2000; CDC, 2001a). These examples illustrate some of the profound effects of the physical environment on health. The places where people live may expose them to harmful factors.
Methylmercury: A Case Study
The case of methylmercury as an environmental pollutant illustrates the potentially dramatic effects of the physical environment on health. Environmental toxins are a specific form of environmental hazard, caused in most cases by industrial enterprises, and the adverse effects of such toxins on the nervous system have been well documented. High levels of exposure to certain environmental pollutants are known to cause acute effects including convulsions, paralysis, coma, and death. The effects of lead on health and development have been documented for decades, and policy action regarding leaded gasoline and lead-based paints has been taken, with positive effects on child health. However, there is growing concern about emerging evidence that other ubiquitous pollutants such as polychlorinated biphenyls (PCBs) and mercury may cause behavioral problems and affect mood and social adjustment. The adverse impacts of exposure to these pollutants may be most profound during fetal development and early childhood. Amidst growing national concern about developmental disabilities, exposure to mercury in the environment represents an emerging and preventable environmental health threat.
The National Research Council (NRC) report Toxicological Effects of Methylmercury (NRC, 2000) examined the evidence of adverse health impacts resulting from exposure to mercury, focusing on consumption of seafood contaminated by releases to the environment. Fossil fuel combustion represents the major source of mercury released to the environment. The deposition of mercury on the land and in surface waters results in conversion to forms that accumulate in the food chain. This bioaccumulation can result in very high concentrations of mercury in some fish, which are the main source of exposure for the population. The developing brain is particularly sensitive to the adverse effects of mercury exposure. Prenatal exposures may interfere with the growth and development of neurons and cause irreversible damage to the nervous system. Infants whose mothers were exposed to high levels in poisoning episodes in Minamata, Japan, and in Iraq were born with severe disabilities, including mental retardation, cerebral palsy, blindness, and deafness (EPA, 1997; NRC, 2000). More recently, epidemiological studies of lower-level exposure from maternal fish consumption have raised concerns about subtle neurodevelopmental deficits.
The NRC report concluded that the evidence of developmental neurotoxic effects from mercury exposure is strong and called for revision of the Environmental Protection Agency (EPA) reference dose that provides public health guidance on acceptable population exposure levels. This conclusion was based on epidemiological studies of low-level chronic exposure from seafood consumption. The population at risk consists of women of childbearing age and their children. Frequent consumers, particularly of fish that tend to accumulate high levels of mercury, may be exposing their unborn children to levels of mercury in the range that has been shown to be associated with developmental deficits. Based upon the available data on fish consumption, the NRC committee estimated that as many as 60,000 newborns may be at risk for adverse neurodevelopmental effects from in utero exposure to mercury. Recently, the Centers for Disease Control and Prevention (CDC) released the first National Exposure Report, which provided dramatic confirmation of the emerging threat of mercury. Ten percent of a national sample of women of childbearing age had mercury levels in their blood within 1/10 of potentially hazardous levels, indicating a narrow margin of safety for many women (CDC, 2001c).
Currently, 40 states have issued fish consumption advisories to reduce exposure to mercury. EPA and the Food and Drug Administration (FDA) have also recently revised their guidance concerning consumption of fish species that have been shown to have high levels of mercury. Ultimately, the threat of mercury can be most effectively reduced through control of the sources of pollution. However, control of sources from the burning of fossil fuels may be decades away. In the meantime, prevention of adverse public health impacts from mercury will require a partnership among health care providers, public health agencies, and others.
The example of methylmercury clearly illustrates the serious impact of just one environmental risk factor. The influences of many other environmental risk factors on health have not been fully documented, and evidence of the influence of environmental factors for some health conditions like asthma is rapidly accumulating (Trust for America's Health, 2001). The association between certain chronic diseases and environmental causes is devastatingly clear, yet knowledge about the scope of environmental health risks and their impact on the public's health is limited. Most states do not track environmental risk factors like pesticides and other hazards or most chronic diseases (such as asthma) and birth defects (Pew Environmental Health Commission, 2001). Certainly, a significant amount of work remains to be done to address the physical environment's powerful influence on health status. A great deal about health determinants in the built and natural environments has been learned in recent decades, but much more is yet to be examined.
THE SOCIAL DETERMINANTS OF HEALTH
Most recently, social epidemiologists and other researchers have focused on identifying the social equivalents of leaded gasoline and environmental tobacco smoke. Among the greatest advances in understanding the factors that shape population health over the last two decades, and clearly since the last Institute of Medicine (IOM, 1988) report on the health of the public, has been the identification of social and behavioral conditions that influence morbidity, mortality, and functioning.
The evidence amassed strongly and consistently points to the importance of these conditions as significant determinants of population health. Because they also feature prominently in the committee's determinants-of-health model, the evidence related to four conditions whose importance is robustly supported is reviewed here: (1) socioeconomic position, (2) race and ethnicity, (3) social networks and social support, and (4) work conditions. Additionally, we discuss the evidence related to a fifth condition that has been and that still is the subject of great interest as well as controversy: ecological-level influences, namely, economic inequality and social capital.3 The present analysis reviews key evidence related to these five conditions that has been presented more extensively in Health and Behavior (IOM, 2001).
Socioeconomic Status and Health
A strong and consistent finding of epidemiological research is that there are health differences among socioeconomic groups. Lower mortality, morbidity, and disability rates among socioeconomically advantaged people have been observed for hundreds of years; and in recent decades, these observations have been replicated using various indicators of socioeconomic status (SES) and multiple disease outcomes (Syme and Berkman, 1976; Kaplan and Keil, 1993). SES is defined in terms of education, income, and occupation. Furthermore, educational differentials in mortality have increased in the United States over the past three decades, leading to a growing inequality, even though mortality rates have dropped for all groups (Feldman et al., 1989; Pappas et al., 1993; Tyroler et al., 1993).
Although it may be measured as level of education or income, SES is a complex phenomenon often based on indicators of relationships to work (occupational position or ranking), social class or status, and access to power. From a policy perspective as well as an etiological perspective, it is important to understand which of the components is critical—for instance, if education is found to be important, the policies that may be implemented would differ from the policies needed if income was found to be the most influential factor. In fact, most research has not tested such competing hypotheses directly, so in the examples that follow, these have not been disaggregated, although the indicators used in each study are explicitly identified.
Several major studies have ascertained that education, income, and occupation, as indicators of SES, are associated with mortality and with mortality due to certain causes. The National Longitudinal Mortality Study found that mortality was strongly associated with all three measures of SES (Rogot et al., 1992; Sorlie et al., 1992, 1995) (see Box 2–1).
Linking SES to Health: Findings from the National Longitudinal Mortality Study. Age-adjusted death rates for white men and women ages 25 to 64 with 0 to 4 total years of education that were 66 and 44 percent higher, respectively, than those for men and (more...)
The Multiple Risk Factor Intervention Trial followed 320,909 white and African-American men for 16 years (Davey Smith et al., 1996a, 1996b) and found that the median family income in one's zip code of residence was predictive of death from a variety of causes. Heart disease, the leading cause of death in the United States, provides a strong example of the association between SES and mortality. Research has documented the relationship between SES and cardiovascular disease (NCHS, 1992; Kaplan and Keil, 1993), and the British Whitehall longitudinal study of civil servants found that those in the lowest grades of employment were at the highest risk for heart disease (Marmot et al., 1991).
A striking finding that emerges from analyses of occupation- and area-based income measures is the graded and continuous nature of the association between socioeconomic position and mortality, with differences persisting well into the middle socioeconomic ranges (Davey Smith et al., 1990; Blane et al., 1997; Macintyre et al., 1998). For example, in the Whitehall studies (Davey Smith et al., 1990; Marmot et al., 1991), the individuals in each employment grade had worse health and a higher rate of mortality than those in the grade above.
Although many of the studies that focused on occupation-, education-, or area-level SES showed a gradient that is virtually linear, studies that focus on income often show somewhat different results. For example, in work by Backlund and colleagues (1996), the association between (increasing) income and (decreasing) mortality is clearly curvilinear, with the decline in the mortality rate with increasing income greatest among those in groups earning less than $25,000 per year but with the decline with increasing income being much less among those earning between $25,000 and $60,000 per year. This curvilinear relationship suggests diminishing returns of income as one approaches the highest income categories, although some association may persist. This curvilinear association between income and health is what lays the framework for findings that more egalitarian societies (i.e., those with a less steep differential between the richest and the poorest) have better average health, because a dollar at the bottom “buys” more health than a dollar at the top. Whether SES has a linear or curvilinear relationship with health has enormous implications for understanding both the etiologic associations and the policy implications of this research. In either case, however, it is important to note that a “threshold” model focused exclusively on the very poorest segments and ignoring others near the bottom and the working poor will not address the relatively poor population health outcomes for the U.S. population as a whole. The major reason for this is because there are groups in the moderate-risk categories of working poor and working class who contribute disproportionately large numbers to death rates and poor health outcomes.
SES is linked to health status through multiple pathways (such as distribution of health care, psychosocial condition, toxic physical environments, and health-related behaviors), but these relationships have not yet been fully elucidated. It is also likely that some degree of reverse causation influences the strength of these associations. Studies in which education rather than income or occupation is used as an indicator of SES are stronger in this regard since most people are not influenced by serious chronic diseases related to cardiovascular disease, stroke, or cancer in ways that inhibit their level of educational attainment in their adolescence and early twenties. Furthermore, although many studies have included a broad range of covariates in their multivariable analyses, it is of course possible that unobserved attributes account for some observed disparities. There is ample evidence that SES is strongly related to access to and the quality of preventive care, ambulatory care, and high-technology procedures (Kaplan and Keil, 1993); but health care appears to account for a small percentage of the variation in health status among different SES groups. It has been argued that differential access to health care programs and services is not entirely responsible for socioeconomic differentials in health (Wilkinson, 1996), because causes of death that apparently are not amenable to medical care show socioeconomic gradients similar to those for potentially treatable causes (Mackenbach et al., 1989; Davey Smith et al., 1996a). Furthermore, similar gradients persist in countries with universal coverage, such as the United Kingdom.
Despite the past century's great advances in sanitation, which have contributed to the sharp increase in life expectancy observed among all socioeconomic groups, the socioeconomic gradient in health status persists. It has been proposed, and to some extent documented, that the gap in health status by SES may still be attributable to the effects of crowded and unsanitary housing, air and water pollution, environmental toxins, an inadequate food supply, poor working conditions, and other such deficits that have historically affected and that still disproportionately affect those in the lower socioeconomic strata (USPHS, 1979; Williams, 1990; Adler et al., 1994; Sargent et al., 1995; McLoyd, 1998). Studies that incorporate assessments of material deprivation and aspects of the physical environment will be important to explicate these important potential pathways.
Considerable evidence links low SES to adverse psychosocial conditions. People in lower socioeconomic positions are not only more materially disadvantaged, but also have higher levels of job and financial insecurity; experience more unemployment, work injuries, lack of control, and other social and environmental stressors; report fewer social supports; and more frequently, have a cynically hostile or fatalistic outlook (Berkman and Syme, 1979; Karasek and Theorell, 1990; Adler et al., 1994; Heaney et al., 1994; Bosma et al., 1997).
There is most often, especially in the United States, a striking and consistent association between SES and risk-related health behaviors such as cigarette smoking, physical inactivity, a less nutritious diet, and heavy alcohol consumption. This patterned behavioral response has led Link and Phelan (1995) to speak of situations that place people “at risk of risks.” Understanding why “poor people behave poorly” (Lynch et al., 1997) requires recognition that specific behaviors formerly attributed exclusively to individual choice have been found to be influenced by the social context. The social environment influences behavior by shaping norms: enforcing patterns of social control (which can be health promoting or health damaging); providing or denying opportunities to engage in particular behaviors; and reducing or producing stress, for which engaging in specific behaviors (such as smoking) might be an effective short-term coping strategy (Berkman and Kawachi, 2000). Both physical and social environments place constraints on individual choice. Over time, those with more economic and social resources have tended to adopt health-promoting behaviors and reduce risky behaviors at a faster rate than those with fewer economic resources.
Socioeconomic disparities in health in the United States are large, are persistent, and appear to be increasing over recent decades, despite the general improvements in many health outcomes. The most advantaged American men and women experience levels of longevity that are the highest in the world. However, less advantaged groups experience levels of health comparable to those of average men and women in developing nations of Africa and Asia or to Americans about half a century ago (Berkman and Lochner, 2002). Furthermore, these wide disparities coupled with the large numbers of people in these least-advantaged groups contribute to the low overall health ranking of the United States among developed, industrialized nations. A major opportunity for us to improve the health of the U.S. population rests on our capacity to either reduce the numbers of the most disadvantaged men, women, and children in the highest risk categories or to reduce their risks for poor health.
Racial and Ethnic Disparities in Health
A substantial body of research documents the relationship between racial and ethnic disparities and differences in health status. Numerous studies have shown that minority populations may experience burdens of disease and health risk at disproportionate rates because of complex and poorly understood interactions among socioeconomic, psychosocial, behavioral, and health care-related factors (NCHS, 1998; DHHS, 2000; IOM, 2002). Although Americans in general experienced substantial improvements in life expectancy at all ages throughout the twentieth century, substantial gaps in life expectancy, morbidity, and functional status remain between white and minority populations. Life expectancy at birth for African Americans in 1990 was the same as that for whites in 1950. Even after controlling for income, African-American men and women have lower life expectancies than white men and women at every income level (for example, see Geronimus et al.  and Anderson et al. ). When indicators of SES are considered, these differences, which are often substantial across a diversity of health outcomes, are commonly reduced but remain significant. Few studies have adequately controlled for SES in terms of the inclusion of economic indicators of wealth, homeownership, or other sources of income. Although these indicators should be included, they are unlikely to reduce disparities between African Americans and whites because data suggest that there are even greater disparities in wealth (all assets) than in household income between these two groups (Ostrove et al., 1999). This phenomenon has led researchers to investigate the health effects of discrimination itself. Aspects of discrimination might influence health through any number of mechanisms, including SES. However, conceptualizing discrimination (whether it applies to racial or ethnic minorities, women, homosexuals, or groups of different ages) as a stressful experience that can influence disease processes through a number of potential pathways is a major advance in scientific thinking over the past decade (Krieger and Sidney, 1996). Additionally, although many disparities are measured across broad racial and ethnic classifications, there is significant health status differentiation or “hidden heterogeneity” within, for instance, Asian-American and Pacific Islander populations (NCHS, 1998). The acknowledgment of disparities itself may generalize or aggregate groups that are highly heterogeneous because of variations ranging from the date of immigration and level of acculturation to genetic, social, and cultural differences (Williams and Collins, 1995; Korenbrot and Moss, 2000).
African Americans and other minority populations experience worse health from infancy to old age. Although the national infant mortality rate has decreased over the years to about 7 per 1,000, the rate among African-American infants is nearly twice as high, 14 per 1,000, and that among American Indians is 9.3 per 1,000, whereas it is 5.8 per 1,000 among whites (NCHS, 2002).
Rates of illness such as asthma are much higher among African Americans than among whites, as are levels of obesity, diabetes, and other cardiovascular risk factors that are often established in adolescence and young adulthood. For example, the prevalence of obesity among African Americans is 29.3 percent and that among Hispanics is 21.5 percent, whereas it is 18.5 percent among whites (CDC, 2002). In 2000, the rate of diabetes-related mortality in non-Hispanic African Americans was 49.4 (per 100,000), whereas it was 32.4 in Hispanics and 20.8 in non-Hispanic whites (CDC, 2001b). Rates of death due to HIV/AIDS are 31.9 among African Americans and 3.7 among whites (CDC, 2000).
Some of the racial and ethnic differences in health status may be associated with the fact that minority populations often encounter the health care system in very different ways in terms of both access and quality of care (Fiscella et al., 2000). For a variety of reasons—both structural (having to do with the health care system itself) and financial or cultural—racial and ethnic minorities encounter barriers to health care that often result in less than optimal care and worse outcomes (Carlisle et al., 1997; Epstein and Ayanian, 2001; IOM, 2002). For example, many studies have concluded that African-American patients are significantly less likely than white patients to receive certain revascularization procedures to treat coronary artery disease (Epstein and Ayanian, 2001). Barriers to care may include linguistic differences, a lack of insurance or difficulties with payment, immigration status, social issues such as trust and some pervasive but subtle forms of racism and discrimination, and even logistical problems related to distance and transportation (Thomas, 2001; IOM, 2002). African-American and Hispanic children are more likely to be uninsured than white children and are less likely to have a usual source of health care (Weinick and Krauss, 2000). Recent research indicates that disparities in access persist even after controlling for socioeconomic circumstances and health insurance coverage status (Roetzheim et al., 1999; Weinick and Krauss, 2000). Among other disparities in health care, African Americans have been shown to be less likely to receive certain diagnostic testing; adequate pain medication; early-stage diagnoses of cancer; dialysis as initial treatment for end-stage renal disease, placement on a kidney transplant waiting list, or a kidney transplant; and preventive rather than acute asthma control measures (IOM, 2002). Hispanics are also likely to experience similarly unequal access to health care services (IOM, 2002). With regard to treatment for HIV infection, once tested, HIV-infected African Americans are less likely to receive antiretroviral and related therapies (IOM, 2002). This is in the context of the fact that HIV infection is spreading more rapidly among African Americans and Hispanics than among whites.
Although many studies indicate that certain racial differences in health persist among people of similar SES, it is also true that many minority groups are likely to be poorer and more disadvantaged than whites. This overlap along both racial and economic lines creates a kind of “double jeopardy,” which is associated with substantially increasing risks for poor health. In terms of the association between poverty and minority status, in 1998, for instance, 10 percent of non-Hispanic white children lived in poverty, whereas 36.4 percent of African-American children and 33.6 percent of Hispanic children lived in poverty (CDC, 2000). When health outcomes are examined by level of education of the mother, family income, and ethnicity and race, enormous differences emerge between the least-advantaged African-American children and the most advantaged white children. For instance, among African-American children living below the poverty line, 22 percent have elevated blood lead levels, whereas 6 percent of African-American children in high-income families and slightly more than 2 percent of white children in high-income families have elevated blood lead levels. These patterns are persistent and are seen for other outcomes such as low birth weight and hospitalizations for asthma (NCHS, 1998). Such pronounced disparities have led to a presidential initiative targeted at ethnic and racial health disparities in six specific areas (White House, 1998; Office of Minority Health, 2000). Also, the elimination of health disparities is a goal of Healthy People 2010 (DHHS, 2000).
Social Connectedness and Health
The association between social connectedness and health has received much attention in recent years. Concepts of social connectedness relate to social integration at the broadest level, social networks, social support, and loneliness. Social connectedness may be conceptualized as a societal characteristic related to civic trust and social capital. This area-level experience is discussed in a later section. This section reviews the evidence that the structure of social ties is related to health outcomes and discusses pathways that may link such social experiences to health. People form ties to others the moment they are born. The survival of newborns depends upon their attachment to and nurturance by others over an extended period of time (Baumeister and Leary, 1995). The need to belong does not stop in infancy, but rather, affiliation and nurturing social relationships are essential for physical and psychological well-being throughout life.
Over the past 20 years, 13 large prospective cohort studies in the United States, Scandinavia, and Japan have shown that people who are isolated or disconnected from others are at increased risk of dying prematurely from various causes, including heart disease, cerebrovascular disease, cancer, and respiratory and gastrointestinal conditions (Berkman and Syme, 1979; Blazer, 1982; House et al., 1982, 1988; Welin et al., 1985; Schoenbach et al., 1986; Orth-Gomer and Johnson, 1987; Cohen, 1988; Kaplan et al., 1988; Seeman et al., 1988, 1993; Sugisawa et al., 1994; Seeman, 1996; Pennix et al., 1997). Studies of large cohorts of people enrolled in health maintenance organizations or occupational cohorts also report that social integration is critical to survival, although it may not be as critical an influence on the onset of disease (Vogt et al., 1992; Kawachi et al., 1996).
Powerful epidemiological evidence supports the notion that social support, especially intimate ties and the emotional support provided by them, is associated with increased survival and a better prognosis among people with serious cardiovascular disease (Orth-Gomer et al., 1988; Berkman et al., 1992; Case et al., 1992; Williams et al., 1992) and strokes (Friedland and McColl, 1987; Colantonio et al., 1992, 1993; Glass et al., 1993; Morris et al., 1993). The lack of social support, expressed in terms of conflict or loss of intimate ties, is also associated with health outcomes and risk factors such as neuroendocrine changes in women (Kiecolt-Glaser et al., 1997), high blood pressure (Ewart et al., 1991), elevated plasma catecholamine concentrations (Malarkey et al., 1994), and autonomic activation (Levenson et al., 1993). Caregivers of relatives with progressive dementia are characterized by impaired wound healing (Kiecolt-Glaser et al., 1995, 1998). Social conflicts have been shown to increase susceptibility to infection (Cohen et al., 1998).
Several studies have recently shown that older men and women with high levels of social engagement and networks have slower rates of cognitive decline (Bassuk et al., 1999; Fratiglioni et al., 2000) and better survival independent of physical activity (Glass et al., 2000). The pathways by which social networks might influence health are multiple and include pathways related to health behaviors, health care, access to material resources such as jobs, and direct physiological responses leading to disease development and prognosis. For instance, evidence suggests that, in general, social network size or connectedness is inversely related to risk-related behaviors. People who are socially isolated are more likely to engage in such behaviors as tobacco and alcohol consumption, to be physically inactive, and to be overweight (Berkman and Glass, 2000). Behavioral pathways such as these do not appear to account for a large part of the association between social isolation and poor health, but they are important to consider. It is important to note that networks themselves have generally been shown to exert powerful influences on the behavior of both adolescents and adults, so that networks can either promote health or increase risk depending on the norms of the networks themselves.
Experimental work with animals and humans indicates that social isolation can have a direct effect on physiologic function and subsequent diseases. Animals that are isolated in adulthood, that experience maternal separation, or that are not nurtured in infancy develop more atherosclerosis; have poor, inefficient, or exaggerated neuroendocrine responses; and may have higher levels of immunosuppression (Nerem, 1980; Shively et al., 1989; Suomi, 1991; Meaney et al., 1996). Among humans and primates, those who lack affiliation and strong social networks have been shown to be more likely to develop colds, have stronger stress responses in terms of neuroendocrine reactions and higher levels of cardiovascular reactivity, and have altered immune responses (Glaser et al., 1992, 1999; Kirschbaum et al., 1995; Cohen et al., 1997; Sapolsky et al., 1997; Roy et al., 1998; Cacioppo et al., 2000). There is limited research on whether access to material goods and resources is a mechanism through which social networks might influence health, and this is an important area for investigation. We do know, however, that networks have the capacity to provide informational and instrumental support effectively. Although much of the research in this area examines the effects of close relationships and social support, there is also evidence that weak social ties may also have indirect positive effects on health and well-being. For instance, a classic investigation of how people find jobs suggests that weak ties to others may be more helpful in enabling people to find jobs, providing access to one of the most critical life opportunities. Whereas one's close friends and relatives (who are likely to belong to the same social circles) may often provide redundant information, weak social ties (e.g., a friend of a friend) may allow individuals to tap into new sets of information (Granovetter, 1995). Instrumental and informational support, two critical components of the support paradigm, relate to help with practical matters such as grocery shopping; rides to the doctor; and information about health care, behavior, and risk. Finally, many of the observational data linking social connectedness to health outcomes do not permit us to rule out issues of reverse causation or the possibility that some unobserved condition explains these associations. More experimental work is needed to answer these questions completely. Much of the experimental work cited here supports the concept that social isolation increases the risk for poor health. However, a recent clinical trial, Enhancing Recovery in Coronary Heart Disease, aimed at improving social support to reduce mortality and reinfarction among subjects after myocardial infarction, found no effect (NIH, 2001). Developing both clinical and population-based experimental studies is the next step in this work.
A large body of evidence accumulated over the last two decades consistently points to the importance of social connectedness, and incorporation of this evidence would involve the inclusion of nurturing community and social networks. As we think of broad social determinants of health that could be influenced to improve health, social connections may be one example that has the support of a number of sectors. Because social relationships influence health through such a myriad of pathways, broad health improvements may be facilitated by considering and enacting policies that support social connections.
Work-Related Conditions and Health
Two decades of research show that the workplace not only generates adverse health effects due to economic circumstances such as downsizing and unemployment or to work conditions such as job demands, control, latitude, and threatened job loss (Karasek and Theorell, 1990), but also generates protective health effects such as social ties that may help counteract the physical and mental adverse effects of work stressors (Buunk and Verhoeven, 1991). The “demand–control” model was developed to describe the psychosocial work environment (Karasek and Theorell, 1990), and other empirical studies have tested the predictive validity of the model with respect to physical health, for instance, by examining the effects of reward relative to effort (Sigerist, 1996).
It has been hypothesized that job strain (the combination of a psychologically demanding workplace and a low level of job control) leads to adverse health outcomes, and findings show that job control is an important component of health-promoting work environments (Johnson et al., 1996; North et al., 1996; Bosma et al., 1997, 1998; Theorell et al., 1998). Schnall and colleagues (1994) found that lower levels of job control (the opportunity to use and develop skills and to exert authority over workplace decisions) were predictive of adverse cardiovascular disease outcomes in 17 of 25 studies, whereas high psychological demands of work had similarly negative effects in only 8 of 23 studies.
The links between unemployment and health have been investigated by European researchers and, to a somewhat more limited extent, U.S. researchers. Although longitudinal studies of European populations have demonstrated a significant relationship between unemployment and higher standardized mortality ratios (SMRs), even after adjusting for age and social status (Moser et al., 1984, 1986, 1987; Costa and Segnan, 1987; Iversen et al., 1987; Martikainen, 1990; Kasl and Jones, 2000; Stefansson, 1991), U.S. data based on the U.S. National Longitudinal Mortality Study (Sorlie and Rogot, 1990) have shown no significant association between age, education, and income-adjusted SMRs and unemployment for either men or women. However, other U.S. epidemiological findings associate unemployment or risk of job loss with health conditions such as depression and engagement in negative health behaviors such as substance abuse, poor diet, and inactivity (Dooley et al., 1996). Analysis of panel data from the U.S. Epidemiologic Catchment Area study suggested that the 1-year incidence of clinically significant alcohol abuse was greater among those who had been laid off than among those who had not (Catalano et al., 1993). Examination of cases of job loss due to factory closures is important because worker characteristics in such cases have no effect on the loss of jobs. Morris and Cook (1991) reviewed longitudinal studies of factory closures and found that the job loss experience exerts a negative effect on physical health.
The impact of threatened job loss has received increased attention recently. European studies found negative effects on health because of threatened job loss or organizational change, although there were no significant differential trends in weight, blood pressure, or blood glucose over time.
The Whitehall II cohort of British civil servants (Ferrie et al., 1995, 1998) found that white-collar workers under threat of major organizational change (elimination or transfer to the private sector) may experience adverse changes in self-rated health, long-standing illness, sleep patterns, and number of physical symptoms and may experience minor psychiatric morbidity. Longitudinal data on male Swedish shipyard workers threatened with job loss and on stably employed controls (Mattiasson et al., 1990) showed that serum cholesterol concentrations increased significantly among the former group. In a study of Finnish government workers (Vahtera et al., 1997), downsizing was associated with increased medically certified sick leave. Among American automobile workers (Heaney et al., 1994), extended periods of job insecurity were associated with increased physical symptoms. However, workers who remain in an organization after a downsizing do not experience a decline in well-being, despite an increase in work demands (Parker et al., 1997). Contrary to work conditions related to involuntary job loss, retirement does not appear to have negative health consequences (Moen, 1996; Kasl and Jones, 2000).
Ecological-Level Influences: The Importance of Place in Population Health
Social characteristics of individuals are closely related to health. Among the most important findings to emerge from public health research over recent years is the extent to which characteristics of areas exert independent effects on health. This ecological4 approach has been rediscovered and is now embedded in a multilevel framework. The major idea is that characteristics of places—neighborhoods, schools, work sites, and even nations— carry with them health risks for the individuals who live in those environments. The health risk conferred by these places is above and beyond the risk that individuals carry with them. Thus, we might view characteristics of physical environments (e.g., parks and buildings) as well as social environments (e.g., levels of inequality and civic trust) as truly properties of places, not individuals. In this section, the committee reviews evidence related to two aspects of places—economic inequality and social capital— that are assessed at an ecological level to examine their effects on health. These findings are relatively new and undoubtedly will be refined with further research. Economic inequality may exert an effect on health in addition to the effect of individual income on health. Such an effect may be particularly robust for people in the United States who are at the lower ends of the distribution.
The United States is among the richest countries in the world, yet it is also one of the most (and increasingly) unequal in terms of the distribution of its wealth as measured by a wide and growing gap between the best-off and the worst-off quintiles (Weinberg, 1996; Jencks, 2002) (see Box 2–2). At a national level, the hypothesis linking income inequalities and health would predict that two countries with the same average income but different income distributions would experience different patterns of mortality, with the country with the more even distribution having a longer life expectancy overall. Cross-national studies initially supported an association between income equality and population longevity, but more recent research, which includes newer and more accurate data for more countries, suggests that the area-level effects of inequality across nations may not hold over time (Lynch et al., 2001; Gravelle et al., 2002; Rodgers, 2002). Recent studies have shown the cross-national correlation between economic inequality and mortality to be very weak or virtually nonexistent (Kunst et al., 1998). Furthermore, in several countries (Canada, for example), inequalities at the level of provinces or neighborhoods within cities often have been found to be not significant in terms of health status. In the United States, however, data are more consistent in supporting the area-level effect of inequality net of individual effect. For example, Kaplan and colleagues (1996) and Kennedy and colleagues (1996) independently found that the degree of household income inequality in the 50 states was associated with the state-level variation in total mortality, as well as with the state-level variations in infant mortality and rates of death from coronary heart disease, cancer, and homicide. The findings persisted after controlling for urban–rural proportion and for health behavior variables such as cigarette smoking rates.
Income Inequality in the United States. In 1968, the wealthiest 20 percent of U.S. households earned an average of $73,754, whereas the poorest 20 percent of households earned $7,202 (Atkinson et al., 1995). In 1994, the inflation-adjusted average income (more...)
Lynch and colleagues (1998) observed a relationship between income inequality and mortality at the level of U.S. metropolitan areas. Although income inequality is strongly correlated with poverty (R = 0.73), the adverse effect of income inequality on health outcomes does not appear to be explained entirely by the fact that places that exhibit income inequality have greater concentrations of poor people, who in turn have a higher risk of mortality (compositional effects). There is also evidence of a contextual effect of income inequality directly on individual health (Wilkinson, 1992; Kennedy et al., 1998; Soobader and LeClere, 1999). Kennedy and colleagues (1998) reported that people residing in states with the greatest income inequality were 1.25 times more likely to report being in fair or poor health than were those living in the most egalitarian states. The effect of income inequality was statistically significant and independent of absolute income levels.
These findings pose the challenge of explaining why the effects of inequality are more significant and conclusive in the United States than in other developed nations. Some (Kawachi and Kennedy, 1997) have argued that inequality is associated with a lack of investment in education, development, and social services and is also related to weak civic and social bonds—or a lack of trust—between people (Wilkinson, 1996; Kawachi et al., 1997; Kawachi and Berkman, 2000). Some countries buffer the effects of inequality with stronger social service programs. Investigators have argued that U.S. analyses have not adequately considered other state-level or country-level social and demographic factors (e.g., racial composition) that may not be “downstream” in the causal chain linking inequality to health (Deaton and Lubotsky, 2001). These questions remain challenges to a new field. However, it is important to note that these studies are all examining the contextual or area-level effects of inequality, net of individual or “compositional” effects. No one has disputed the strong and consistent effects of SES on individual health. New research on area-level efforts related to neighborhoods, work sites, and states and even across countries poses considerable methodological challenges (Deaton, 2002). Nonetheless, such research holds great potential to help us understand the ways in which both the social and the physical (built and natural) environments may affect health and behavior.
Social participation and integration can also be conceived of as both individual and societal characteristics (Kawachi and Kennedy, 1997). Some investigators have started to conceptualize these dimensions at an ecological or group level. At the group level, a socially cohesive society, or one in which most citizens are socially integrated, is one that is endowed with stocks of “social capital,” which consists partly of moral resources such as trust between citizens and norms of reciprocity. This has led investigators to examine the area-level effects of these domains. Particular interest has been focused on the relationship between social capital and health. At a group level, more socially integrated societies seem to have lower rates of crime, suicide, and mortality from all causes and a better overall quality of life (Wilkinson, 1996; Kawachi and Kennedy, 1997; Kawachi and Berkman, 2000).
Kawachi and colleagues (1997) analyzed social capital indicators across the United States in relation to state-level death rates. The per capita density of membership in voluntary groups was inversely correlated with age-adjusted mortality from all causes. Density of civic association, group membership, and levels of interpersonal trust (i.e., percentage of citizens endorsing the expectation that altruistic behaviors will be returned in kind at some future time) were also associated with lower mortality. Kawachi and colleagues (1999) also carried out a multilevel study of the relationship between the above indicators of state-level social capital and individual self-rated health. A strength of this study was the availability of information on individual medical and behavioral confounding variables, including health insurance coverage, cigarette smoking, and being overweight, and on sociodemographic characteristics, such as household income, education, and whether one lived alone. Even after adjustment for these variables, people residing in states with low levels of social capital were more likely to report fair or poor health. The odds ratio for fair or poor health in association with living in areas with the lowest levels of interpersonal trust as opposed to the highest levels of interpersonal trust was 1.41.
There are several plausible mechanisms by which social cohesion might influence health through contextual effects. At the neighborhood level, social capital might influence health behaviors by promoting the more rapid diffusion of health information. Sampson and colleagues (1997) provide evidence that ”collective efficacy,” or the extent to which neighbors are willing to exert social control over deviant behavior, plays an important role in preventing crime and delinquency. Neighborhood social capital also could affect health by increasing access to local services and amenities (Sampson et al., 1997). Finally, neighborhood social capital could influence health through direct psychosocial pathways by providing social support and acting as the source of self-esteem and mutual respect.
Although there has been a great deal of interest in these area-level studies of social capital, there has also been a fair amount of skepticism regarding their validity. Several social scientists (Portes and Landolt, 1996; Sandefur and Laumann, 1998; Durlauf, 1999) have voiced concerns about the ambiguity of the concept, the potential for social capital to lead to undesirable outcomes related to the exclusion of certain groups, and insufficient attention to the determinants of social capital itself or the causal patterning between it and other social conditions. Future studies will be strengthened with the addition of items tapping the conceptual richness of the domain of social capital and the capacity to distinguish it from other closely related constructs of social networks and SES.
POPULATION-LEVEL PREVENTIVE INTERVENTIONS
The evidence presented and discussed in this chapter aims to demonstrate that taking into account the environmental and social determinants of health is essential to creating effective population-level interventions for health improvement. Health risk is related to a complex of social, economic, and political factors that both surpass and powerfully interact with “downstream” elements such as individual behaviors, biological traits, and access to health care services. There have been few empirical tests of population-based approaches to health promotion that focused on risk-related social conditions, but in an effort to understand how such approaches might work, several examples are presented to illustrate the effectiveness and efficiency (e.g., cost) of population-based interventions to prevent disease and promote health.
Preventive interventions at the population level may be classified as universal, selective, and indicated, borrowing the classification developed by Gordon (IOM, 1994b). A universal measure is one that would be desirable for everyone in an eligible population. It would focus on shifting the entire population distribution rather than on targeting only relatively high-risk individuals, as illustrated earlier in this chapter. It would likely involve an agreed-on public policy requiring broad-based public understanding and political support. A selective preventive measure is one that is desirable only when an individual is a member of a subgroup of the population whose risk of becoming ill is above average. These are the more traditional population-oriented public health education interventions targeted toward the high-risk segments of the population. Finally, an indicated preventive measure is one that is applicable to persons who, on examination, manifest a risk factor, condition, or abnormality that identifies them individually as being at high risk for the future development of a disease. This type of intervention, usually provided in the context of clinical practice, deals only with individuals diagnosed with a disease, not with the nameless statistical subset of a population as in selective preventive measures. For example, a universal preventive measure for heart disease could include the provision of general advice to consume a diet low in fat accompanied by a regulatory policy requiring food labeling. A selective intervention could include a program focusing on diet and behavioral changes for overweight individuals who do not exercise regularly, and an indicated preventive measure might include antihypertensive medication for those diagnosed as hypertensive.
Although many studies have looked at the effectiveness of preventive measures, few have studied universal, population-level strategies. In some cases, however, such as tobacco use prevention and automobile-related injury prevention population-based strategies (e.g., laws) have been used successfully, largely because of recognition of the broad determinants of health. Results of these interventions indicate that, at least in some cases, a population-level strategy or, to use Gordon's classification, a universal measure may be more optimal and cost-effective than interventions targeted further downstream (i.e., at the individual level). Acting on the most upstream level of determinants of health typically means the level of national policy. This may help shift national norms and values that lead to the passage, adoption, and ultimately, success of the respective legislation, as in the case of seat belt legislation, which has steadily and gradually normalized this behavior across America, or tobacco policy, which has curbed the use of tobacco (e.g., through changes in the social landscape of outdoor advertising and sanctions on smoking in the workplace and public places such as restaurants). Alternately, upstream policy interventions may also refer to modifying the broader, social determinants of health such as income (e.g., through the provision of earned income tax credits and minimum wage increases), education, and social connectedness.
Policy making at the national, state, and local levels has the potential to positively shape population health by addressing specific elements of the determinants of health, such as inadequate housing, unavailability of family-friendly social and work policies, lack of public transportation, lack of safe public spaces, and so forth. The health consequences of contaminated water and leaded gasoline have been well elaborated. It is now time to determine their social equivalents—elements of the social environment that influence health status—and take action to shape them in support of population health. Such action may focus not only on education, decent housing, and a living wage but also on the political choices that move the broad (social and other) determinants of health in a positive direction. For example, certain health care disparities (e.g., disparities in access and quality) are created in part by political choices and by allowing public and private insurance programs to limit coverage for preventive health care and for conditions related to mental health, substance abuse, and oral health.
Seat Belt Laws
Federal legislation has been an important strategy in reducing motor vehicle injuries (IOM, 1998). Between 1966 and 1970, highway safety acts authorized the federal government to set safety standards for new vehicles and equipment (e.g., standard safety belts for all automobile occupants) and to develop a coordinated national highway safety program, established in 1970 as the National Highway and Traffic Safety Administration.
A number of early studies found seat belts to be cost-effective (Warner, 1982). A more recent report outlines the benefits of safety belts based on medical and financial information from the Crash Outcome Data Evaluation System. A 1996 report to Congress revealed that safety belts are highly effective in reducing morbidity and mortality and in decreasing the severity of injuries (e.g., the inpatient charge for unbelted accident victims was 55 percent greater than the charge for those who wore seat belts) (NHTSA, 1996). Other evidence suggests smaller, although still significant, differences between injuries experienced by belted and unbelted accident victims. National averages indicate that for each occupant involved in a crash, medical costs average $2,930 for restrained riders and $5,630 for unrestrained riders (in 1995 dollars) (Miller et al., 1998).
Although there are ethnic differences in seat belt use, rates of seat belt use are higher in states that implement and enforce restraint laws (Davis et al., 2001; Schiff and Becker, 2002). Seat belt laws have had a significant impact on modifying behavior and thus decreasing risk and improving health outcomes across the population. Such strong effects for any single piece of legislation illustrate that this orientation can be effective. A meta-analysis of research regarding the effectiveness of interventions to increase the use of seat belts found that both primary seat belt laws (a motorist can be stopped for not wearing a seat belt) and secondary seat belt laws (a motorist stopped for other reasons can also be cited for not wearing a seat belt), as well as enhanced enforcement policies (e.g., more officers and checkpoints for seat belts), are effective (Dinh-Zarr, 2001). An economic analysis found that the benefit–cost ratio for passage of a seat belt law was $260 per new user. The benefit–cost ratio per quality-adjusted life year showed that all three interventions offered net cost savings (Miller, 2001). The case of seat belt legislation demonstrates that such upstream or population-level measures aiming to prevent disease and disability may be effective in transforming social norms and ultimately changing behavior.
The Case of Tobacco
Tobacco prevention and cessation efforts have offered many lessons about the links between behavior and disease and how to intervene effectively to improve population health. CDC described the “antismoking campaign” dating from the first Surgeon General's report as one of the major public health successes of the second half of the twentieth century (Warner, 2000). Effective antismoking campaigns are generally comprehensive, multidimensional interventions involving several aspects of prevention and control. One of the most important lessons learned from the tobacco experience is that the social context or social environment serves as a potent force in shaping smoking behavior. Therefore, measures such as creating educational and information-filled environments (from counteradvertising to truthful labeling and Surgeon General's warnings) and enacting regulations to restrict smoking in buildings or public spaces and to control tobacco marketing and sales (to minors) have been effective in changing smoking behavior.
School-based antismoking interventions constitute an effective prevention strategy, although one that is resource intensive (and thus not sufficiently accessible), given the state-of-the-art programs needed to assure success (DHHS, 1994; Warner, 2000). The Growing Up Tobacco Free report (IOM, 1994a) has detailed three categories of tobacco prevention strategies: information dissemination approaches, effective education approaches, and social influence approaches. According to that report, the former two strategies fail to address the relationship between the acquisition of knowledge and behavior and the addictive nature of tobacco, nor do they address the social context in which smoking occurs, which often involves peer pressure and norms about use. The third strategy (the social influence approach) was developed to address the deficiencies of earlier strategies, and in a meta-analysis of 143 adolescent drug use prevention programs, it was found that peer programs based on this strategy had the greatest effects on all outcome measures for the average school-based population (Tobler, 1986). Multiple levels of influence and multiple determinants affect the uptake of smoking, from individual characteristics and behaviors to population-level advertising and availability. Also, as noted earlier in the context of generic population health improvement, upstream approaches, including action at the community or population level, may be more cost-effective than downstream approaches directed at specific individuals (Corbett, 2001). Such measures, it seems, make use of the characteristics of social networks and relationships that may be used as elements to further protect health. Although recognizing the importance of approaches that go beyond individuals and their behavior, use of a social influences strategy may not work well if it is used alone, as in the case of the Hutchinson Smoking Prevention Project, a long-term randomized trial that used a school-based social influences approach and concluded that it lacked long-term effectiveness to deter smoking (Peterson et al., 2000). According to Warner (2000), success is more likely when a broad array of multisectoral, multilevel, upstream interventions is used.
Such upstream measures include taxation of cigarettes, which appears to affect tobacco consumption among youth and adults and both the initiation and the cessation of tobacco-smoking (Chaloupka et al., 2002). It also appears to have a more powerful impact among lower-income groups than higher-income groups, an important concept because most of the educational interventions are believed to be more effective among more highly educated groups (Evans and Farelly, 1998; Warner, 2000). Overall, on average it appears that an approximately 10 percent increase in the price of cigarettes will produce a 4 percent reduction in demand. Policies based on taxation have the potential to have a significant impact on smoking rates nationwide (Warner, 2000), although the effectiveness of taxation on teenagers becomes somewhat attenuated in adulthood, underscoring the need for interventions on several fronts (Glied, 2002). An added benefit of tax increases is the production of revenue while the level of consumption declines. Cigarette taxation is an excellent example of a cost-effective preventive strategy aimed at the entire population. It is serving to broaden the face of policy interventions in public health, leading to consideration of taxation in relation to alcohol consumption, unhealthy diets, and even gun control. It is feasible that such economic disincentives could affect a broad range of other social and behavioral conditions as well.
In a recent effort to develop a model for evaluating the outcomes of youth-targeted tobacco prevention and control programs, the Social Science Research Center at Mississippi State University developed a Social Climate Survey. Baseline data were obtained in 1999, and national data were added to the model in 2000 for comparison purposes. The Social Climate Survey measures beliefs, norms, and practices related to tobacco use, sale, taxation, and regulation in seven social institutional areas: family and friendship groups, education, government and political order, work, health and medical care, recreation/leisure/sports, and mass communication or culture. Data from the surveys have indicated that implementation of a comprehensive, multidimensional approach to tobacco prevention and cessation will affect the social climate in which decisions regarding tobacco use are made (McMillen et al., 2001).
Some of the most striking examples of the environmental embeddedness and social implications of health risk come from the study of alcohol and tobacco product advertising. For example, a study of billboards in Chicago was compared to census data and demonstrated that the density of advertising for alcohol was five times higher in poor and minority urban wards than in other geographic areas and that the density of advertising for tobacco products was three times higher (Hackbarth et al., 1995). An observational study of tobacco billboard advertising in St. Louis used geographic information systems and found that advertisements were concentrated in low-income and minority neighborhoods, as well as in close proximity to public school property (Luke et al., 2000). Such data support recent efforts to control tobacco advertising, adding this to the arsenal of taxation and limits on smoking in public places.
THE PUBLIC HEALTH SYSTEM IN ACTION: A SCENARIO
In this section, the committee uses the specific scenario of a risk to population health—namely, obesity—to present and discuss the contributions that communities, the health care delivery system, employers and business, the media, academia, and the governmental public health infrastructure can make to improve health. Food and eating are integral to the life of individuals and communities. Families, friends, and neighbors gather around meals; and both the process of eating and the food itself are heavily imbued with cultural, social, and even emotional meaning. This helps explain why eating and nutrition perfectly exemplify the influence of multiple determinants on health. The development of obesity itself is influenced by multiple determinants of health, from the genetic to the social and environmental, and the public health partners must consider these many dimensions in formulating their responses.
Obesity: Magnitude and Future Trends
The problems of overweight and obesity in America have reached epidemic proportions and threaten the health and quality of life of millions of adults and children. According to CDC's 1999 National Health and Nutrition Examination Survey, more than 61 percent of adults are either overweight or obese and at least 13 percent of children ages 6 to 11 and 14 percent of adolescents ages 12 to 19 are overweight. The prevalence of obesity among adults has grown by nearly 20 percent over the past 30 years, and the number of children who are overweight has tripled in the same period (see Figure 2–3) (CDC, 2001a).
Obesity trends among U.S. adults: Behavioral Risk Factor Surveillance System (BRFSS), 1991, 1995, 2000. NOTE: Obesity is BMI ≥ 30, or –30 lbs overweight for 5′4″ woman.
Obesity is a growing concern because it poses a higher risk and results in a higher incidence of health conditions such as diabetes, cardiovascular disease, stroke hypertension, osteoarthritis, and certain cancers than other risk factors (NIH, 1998; Allison et al., 1999; Must et al., 1999; Williamson, 1999; Tataranni and Bogardus, 2001; Tuomilehto et al., 2001).
The human and economic costs are impossible to ignore. Every year an estimated 300,000 U.S. adults die of causes that may be attributed to obesity; in addition, others suffer from chronic disease and an impaired quality of life (Allison et al., 1999; Mokdad et al., 2000). According to another estimate, as many as 309,000 to 582,000 deaths in 1990 were associated with poor diet and inadequate physical activity (McGinnis and Foege, 1993). Should the number of overweight and obese Americans continue to grow at its current rate, obesity will surpass tobacco as the most preventable cause of death and illness in the United States. Additionally, the estimated direct and indirect costs associated with obesity are $100 billion annually (Wolf and Colditz, 1998), and this figure does not take into consideration the cost of treating the uninsured or the personal impact of obesity on quality of life.
The fact that the prevalence and incidence of type II diabetes mellitus (see Figure 2–4) have increased exemplifies the association between weight and health. The rise in the rate of obesity during the past decade has been paralleled by a 25 percent increase in the rate of type II diabetes (Harris et al., 1998). The rapid and large rate of increase in obesity among children is especially alarming, given that childhood obesity is clearly associated with obesity in adulthood and subsequent health problems. For instance, until recently, type I diabetes was the most prevalent type of diabetes among children. However, new studies have shown that the rate of type II diabetes is increasing dramatically and that 85 percent of children with type II diabetes are either overweight or obese (ADA, 2000). Approximately 63 percent of the total direct costs associated with obesity are related to type II diabetes (Wolf and Colditz, 1998).