draft - CiteSeerX

W97-10

1/5/98

-DRAFT-

A MULTILEVEL APPROACH TO STUDYING GLOBAL ENVIRONMENTAL CHANGE IN FOREST ECOSYSTEMS

by Emilio Moran Center for the Study of Institutions, Population, and Environmental Change Department of Anthropology Anthropological Center for Training & Research on Global Environmental Change Indiana University, Bloomington Elinor Ostrom Center for the Study of Institutions, Population, and Environmental Change Workshop in Political Theory and Policy Analysis Department of Political Science Indiana University, Bloomington J. C. Randolph Center for the Study of Institutions, Population, and Environmental Change School of Public and Environmental Affairs Indiana University, Bloomington

© 1998 by authors

Support from the National Science Foundation (Grant # SBR-9521918) is gratefully acknowledged. This paper is based on intensive discussions and written commentaries among members of the Center for the Study of Institutions, Population, and Environmental Change, among them David Dodds, Clark Gibson, Glen Green, Charles Schweik, and Catherine Tucker. Comments by George Alter, Dennis Conway, and Kerry Krutilla on an earlier draft are deeply appreciated.

W97-10

1/5/98 A MULTILEVEL APPROACH TO STUDYING GLOBAL ENVIRONMENTAL CHANGE IN FOREST ECOSYSTEMS by Emilio Moran, Elinor Ostrom, and J. C. Randolph

Among the many questions central to the study of the human dimensions of global environmental change, three broad questions surface:

C How are regional and global political and economic processes linked to human behaviors at household and community levels?

C How can macro-scale physical and biological processes observed and modeled at a global scale be linked to meso and micro human organizational and decision-making processes?

C How do institutional arrangements influence the direction and size of the impact of human driving forces, such as population density and transportation networks, on ecosystems and global change? Each of these three questions involve multiple temporal and spatial scales. When is it appropriate to scale up or scale down, how to cycle between scaling up and scaling down, and how to integrate theories developed at a one level to explain phenomena occurring at another or at multiple levels are central issues for most global change research (Pickett and Cadenasso, 1995; Root and Schneider, 1995; Young, 1995). Changes in the extent, composition, structure and function of forest ecosystems are of interest to any research program on global change. These changes are affected by many social, economic, and political factors including some combination of: “population growth, land-hunger, unequitable social conditions, property-rights regimes, misguided government policies, collective action problems, inappropriate technology, international trade relations, economic pressures afflicting debt-burden developing countries, and corruption in the forestry sector” (Andersen et al., 1996: 94; see also Brown and Pearce, 1994). Human-induced causes of both deforestation and afforestation occur at multiple levels. Population growth can occur in one particular community or in a large region. Property rights may be well-defined in one region, but not in another. International trade affects the market price of many commodities but the

1

particular price received by farmers may also depend on government taxes and subsidies or the infrastructure provided in a specific community. Thus, adopting a multilevel approach is essential for understanding human impacts on forests. In this paper, we present an initial effort to elucidate a multilevel theory of human action impacting over time on the extent and composition of forests in the Americas and our strategy to develop methods for such research.

Some Definitions and Assumptions No single measure of forest health or value conveys full information about the condition or rate of change of any particular forest or set of forests. We use the term “forest conditions” to mean a suite of evaluative measures that together provide a composite picture of the current physical, biological, and economic value of a forest as well as the rates of change in these values. Among the measures included in the concept of forest conditions are the density, biomass, basal area, diversity, species, composition, economic and subsistence value, extent, shape, fragmentation, and rates of destruction, growth, removal, and regrowth of forests. These measures are affected by the type of land uses and land cover present in a location. Each of these concepts can be measured in multiple manners.1 In this paper, we do not address the particular measures of dependent variables, but rather how we envision processes that affect this suite of measures. Forest conditions are dependent on both human decisions and on physical and biological processes. Human decisions affect forest conditions in a direct as well as in an indirect manner. Direct actions result in an immediate impact on a particular forest. When a team working for a timber firm harvests a stand of timber, when a subsistence farmer clears a forest patch in order to plant manioc, or when a class of school children plant and protect tree seedlings, specific individuals are directly affecting the extent and shape of a particular forest, its rate of destruction or growth, and perhaps the species composition of that forest. These changes may also affect its economic and subsistence value. Indirect actions affecting forest conditions are mediated by other transactions between the person taking the indirect action and the person

2

taking the related direct actions. The construction of dams, highways, and other types of physical infrastructure that directly affect forest conditions and their uses typically result in significant indirect effects. When individuals decide to build or purchase houses and furniture, to buy stock in certain multinational corporations, or even to buy coffee, bananas, or particular brands of ice cream, they indirectly affect forest conditions in forests that may be located nearby or across the globe from where they live. Changes in population density or its spatial distribution also indirectly affect forest conditions. Whether these changes affect forest conditions in a positive or negative manner depends on how the opportunities faced in a local setting and the choices of particular sets of individuals as mediated by institutions and demographic factors (Tiffen, Mortimore, and Gichuki, 1993; Fox, 1993). Forest conditions are simultaneously affected by physical factors such as solar radiation, temperature, and precipitation, the chemical composition of the atmosphere, and topography, as well as by a wide diversity of biological phenomena such as seed dispersal, competition, and herbivory. These physical and biological factors affect forest conditions at multiple spatial and temporal scales, as do anthropogenic changes. In analyzing human decisions as they impact on forest conditions, we assume that: 1. Human decisions occur within tiers of decision-making units that extend from an individual to international organizations. 2. Within all tiers of decision making, fallible individuals make decisions that are intended to increase net benefits to themselves and potentially to others. 3. Individuals learn from their experiences and from culturally transmitted experiences. 4. Human decisions at all tiers are affected by the cultural values of the individuals, the resources they possess, the information they obtain, the incentives and disincentives they face, the internal learning and choice processes used, and the time horizon invoked. 5. Decisions at any one tier affect the information, incentives, and time horizon (and, perhaps the cultural values, resources, and internal choice processes) of others at that tier, at present and future time periods. 6. Decisions at any one tier frequently— but not necessarily— affect the cultural values, resources, information, incentives, internal choice processes, and time horizon invoked at other tiers. Thus,

3

human choice is interdependent within tiers and across time and space. Impacts may be horizontal, upward and downward. 7. Decisions at any one tier directly or indirectly affect forest conditions measured or aggregated at other tiers. 8. Physical and biological process also affect the information, the incentives, and the time horizon that are used in human choice as well as being affected by human choice.

Human Choice In the second, third, and fourth assumptions listed above, we self-consciously adopt a conception of human choice that views fallible humans as seeking goals perceived to be of net positive benefit to themselves, and potentially others, but limited in their success by the information they possess, the incentives they face, the time horizon they use, and their culturally constrained learning and choice-making processes. This conception is consistent with a large body of theoretical work in the bounded rationality and anthropological tradition (Cancian, 1972; Johnson, 1972; Simon, 1985, 1997). All individuals making choices are viewed as fallible learners— whether they are high-level officials with Ph.D. degrees and years of experience or landless peasants with no education trying to eke out an subsistence for their families. Whether fallible learners make good choices for themselves and for others depends not only on their values, but also on the information they possess and the internal learning and choice processes they have developed. Education and experience affect the levels and type of information and the heuristics and/or modes of analysis used in making decisions, in learning about the structure of the world, and in understanding how actions are linked to outcomes. Knowing who to trust, who is a group member, and who has shown group solidarity affects the strategies that individuals adopt including positive or negative reciprocity. Rules related to information flow (e.g., who must be informed about actions taken and forest conditions) as well as to the payoffs received by participants, also affect what individuals learn about the situations they face. All individuals may or may not have the relevant information they need to make effective decisions in a particular setting. Officials may or may not have the relevant information and appropriate modes of analysis needed to determine sustainable development practices for all forest 4

ecosystems present in many governmental jurisdictions. One would expect a higher probability of success when officials are making decisions that affect a limited range of forest ecosystems with which they are already deeply familiar as a result of years of prior experience. Similarly, a landless family may or may not have the relevant information and appropriate modes of learning and analysis needed to determine whether migration from one region to another would improve their life’s chances. Because it is extremely difficult to understand enough about particular decision makers that affect forest conditions through direct and indirect actions, we utilize a general conception of human behavior in our theoretical work. Consequently, we focus most of our attention on the structure of a linked series of nested situations rather than on the particular characteristics of specific individuals (see Popper, 1967).2 We can then draw on a variety of models of individual choice ranging from agents using very rough heuristics and having limited access to relevant information all the way to “hyper-rational” individuals who have complete information relevant to a situation and who have learned how to undertake analysis so as to maximize summed, quantitative goals (Ostrom, 1998). We view all of these models as potentially useful in an effort to understand human choice in complex, interdependent worlds. In those settings which are very tightly constrained and where objectives can be measured quantitatively, hyper-rational (i.e., “ideal”) models of human choices that affect environmental conditions are extremely useful baseline models (see, for example, Ostrom, Gardner, and Walker, 1994). In modeling global change phenomena, however, it will frequently be necessary to use models that make many fewer assumptions about the internal knowledge and calculation capabilities of individuals including cellular automata models (Park and Wagner, 1996), and agent models (Gimblett, 1997). In many instances, individual choices are aggregated within corporate bodies such as private firms, government agencies, local communities governed by elders, and nongovernmental organizations. A frequent technique to simplify analysis is to treat such corporate actors as if they were single decision makers. We adopt this convention in many of our analyses. We examine the internal incentive structures for some corporate actors so as to better understand the likely actions to be taken by such actors. In our 5

efforts to measure the structure of situations in which individuals make decisions that impact on forest conditions, we focus on those aspects of structure which have repeatedly been shown to affect the type of decisions that individuals make. In the next section, we briefly describe the variables that we utilize to analyze a wide diversity of situations.

The Structure of Action Situations When individuals make choices that either directly or indirectly affect forest conditions, they do so in the context of other individuals who are holding positions of various types (land owner, squatter, governmental forest manager, purchasing agent for a timber company, etc.). Each individual in these positions has available to them an array of actions they may take (clear land, plant with selected species, harvest timber, sell or buy commodities, etc.) within a specified time period. In some cases, individuals must gain the agreement of others, thus have less direct control over the actions they can take. These actions produce outcomes, such as the amount of forested land, nature and amount of commercial crop production, or conversion of forest to subsistence agriculture or pasture, etc. Individuals have diverse levels of information about their own actions and those of others (and, even who else is part of the interaction), about the perceived benefits and costs of different actions, and about how actions are linked to produce outcomes for participants. Each of the italicized words are the universal elements that institutional analysts use in analyzing how the structure of these situations affect individual actions and how these actions jointly produce outcomes (see Ostrom, 1986, 1997a). In addition, one needs to know whether the participants in such situations are facing a decision that is unlikely to be repeated, such as migration from urban slums to the frontier, or will be repeated indefinitely, such as buying and selling commodities at a local farmers’market. Situational analysis is at the core of explanations offered by analysts for the actions and outcomes observed in diverse settings. One tries first to identify the structure of a situation— its participants in positions who choose with more or less control actions that are linked to outcomes in light of the

6

information they possess and the likely costs and benefits they face. When the structure of the situation is such that everyone can be better off if they select some strategies rather than others— a pure coordination problem— explaining why some groups achieve better outcomes than others is challenging. Most of the relevant problems related to human uses of forests are not this benign. Many of the situations of interest to this research program will involve some aspects of “social dilemmas” and “collective action problems.” Collective action problems exist whenever there are substantial gains to be achieved if a group of individuals coordinates their strategies while at the same time each individual may find it advantageous to seek out independent strategies that leave them better off if everyone else coordinates. Such problems are frequently referred to as “social dilemmas” (Dawes, 1980; Ostrom, 1998). Having identified the structure of a situation, the analyst next tries to understand how this structure is affected by physical and biological variables, by the structure of the surrounding community and economy, and by the specific rules in use. The expected benefits of planting tree crops or harvesting mature trees, for example, depend on (1) the soils, climate, and location of the decision maker, (2) the resulting topography, climate, and soils; (3) the shared values held by those considered to be part of the decision-maker’s community; and (4) the rules in use specifying what may or may not be planted or harvested. Because the variables that affect the structure of a situation are scale dependent, as one moves across levels of action, the relevant physical and biological world, human community, and set of relevant rules changes. While situations are normally analyzed separately, they are linked together in complex chains in the ongoing world. The linkages among diverse situations are particularly important to understand when studying processes of deforestation or reforestation because the outcomes of one situation tend to spill out or spill in to many other situations. Consequently, it is essential to understand how they are linked across space and within particular time frames as well as across spatial and temporal scales.

Tiers of Decision-Making Units and Their Direct and Indirect Impacts

7

While the number of tiers of decision-making units that may affect forest conditions in a particular location may be quite numerous, we will broadly group them into five tiers: 1. International conferences, international regimes, multinational corporations, and international NGOs (nongovernmental organizations) making decisions that affect commodity and land use distributions as well as rules related to commodities, land use, and forest conditions. 2. National governments, powerful groups, for-profit firms, and NGOs making decisions that affect commodities and land use distributions as well as rules related to commodities, land use, and forest conditions. 3. Regional governments, powerful groups, for-profit firms, and NGOs making decisions that affect commodities and land use distributions as well as rules related to commodities, land use, and forest conditions.3 4. User groups, settlements, and small businesses making decisions that affect commodities and land use distributions as well as rules related to commodities, land use, and forest conditions. 5. Individuals and households making decisions about location, fertility, migration, planting, harvesting, and other relevant activities. Decisions made in situations located at any of these five tiers may have a horizontal impact on decisions made in situations located at the same level or at any of the other four levels. For example, the adoption of a particular international treaty, such as the Montreal Protocol limiting the production of chlorofluorocarbons (CFCs), affects the decisions of multinational corporations such as DuPont and the Imperial Chemical Industries regarding which chemicals to produce (see Oye and Maxwell, 1995). Decisions made at this “global level” affect decisions at all other levels, such as individual choice of a refrigerant. At the 1992 United Nations Conference on Environment and Development (UNCED) held in Rio de Janeiro, decisions were made concerning the importance of protecting the rights of indigenous peoples to their traditional lands. These provisions have been invoked in many national and regional governmental decisions that then impact on decisions at a community and individual level. In our research and in this paper, we focus primarily on the last three tiers of decision making, both in terms of developing theoretical analyses and in conducting empirical research examining specific hypotheses. All research programs have to make fundamental decisions about what variables they will attempt to explain and what variables they will take as givens. We will not, for example, try to explain the 8

underlying geological forces that have created the shape and composition of a particular terrain in an area. Nor will we try to explain broad economic trends or the kinds of values shared in a community. We will consider all of these variables as exogenous variables explaining other variables. The rules that are used by individuals in any particular setting will be treated as both exogenous variables at one tier of action and as endogenous variables at the tier where individuals make such rules (see Snidal, 1995). In the fifth tier— the individual level— we will focus empirical research on decisions related to primary processes that have a direct impact on forest conditions along with physical and biological characteristics. The physical characteristics of a location include variables such as temperature, precipitation, soil fertility, slope, aspect, and elevation (Moran et al., 1997; Tucker, Brondizio, and Moran, 1997; de Silva-Forsberg and Fearnside, 1997). The biological characteristics include the structure and functions of the species that live in a location. We will attempt to explain why households characterized by variables such as income, land tenure, family size and composition, and land ownership patterns make particular patterns of decisions that affect forest conditions (see Durham, 1995; Bilsborrow and Ogendo, 1992). In its simplest form, decisions made at this level can be represented by Figure 1. Among the many relevant decisions made at this tier are decisions about family size, amount of land cleared, speed of rotation among agricultural plots, use of commercial or organic fertilizers, planting of trees, choice of firewood, charcoal or other sources of fuel, or possible migration to a city to earn wage labor income (Dodds, 1997a, 1997b). Some of the hypotheses that we plan to examine at this tier include: * Individuals who must struggle to meet minimum subsistence needs will emphasize short-term benefits rather than long-term sustainability in land use decisions. * Individuals will over-harvest from jointly owned forests when forest products are scarce and when rules defining use are poorly specified and/or unenforced. * Individuals with secure land tenure are more likely to choose sustainable forms of land use (e.g., soil conservation methods in agriculture) and forest use (e.g., rotation of harvesting among management units) than individuals with insecure tenure.

9

* The more dependent households are on forest products for their livelihood, the more likely they are to make long-term investments in sustainable practices. * As the income of a household rises, household dependence on forest resources will decline. * Educational effects on individual decisions about forest resources tends to be bimodal: both highly educated and uneducated (in the sense of formal education; however, they may be very knowledgeable of resources) individuals will make decisions that have less impacts on forest resources and will be generally more conservative of those resources. [Figure 1 about here] But individual decisions rarely occur in isolation. Individuals live in locations that are characterized by social, economic, biological, and physical opportunities and constraints. Thus, in order to explain decisions made at the fifth tier, one needs to understand what is happening at the fourth tier— the community level (Tucker, 1997; Ostrom, 1997b). This more complex picture is represented in Figure 2. [Figure 2 about here] In this multilevel schema, community characteristics include such variables as the history of a location, its geographic location in relation to large urban centers or major roads, socioeconomic factors, demographic trends, cultural norms shared by most community residents, degree of cultural heterogeneity, types of educational opportunities available, number and types of local business and philanthropic organizations, and structure of local government. Further, the kind of shared information present in a community about the value of the forest as a generator of ecological services, about various agroforestry technologies, and about their political efficacy and authority also affect decisions made at this level. We will be particularly interested in how such groups overcome a wide diversity of collective action problems associated with the governance and management of forest resources. Some groups at this level will invest heavily in forms of social capital which lead toward the development of common norms or enforced rules affecting decisions made at the household level. Others will not. The relevant decisions made at a community level may include at least the following:

C What land is designated to be harvested (or otherwise managed) by units larger than individuals? C Who is authorized to use this land? 10

C Who can participate in making rules about use and investment in this land? C What kinds of harvesting and investment rules have been made? C What kind of harvesting and investment decisions are actually made? C Who monitors conformance with these rules and what is their legal status? C What kinds of sanctions are imposed for violations of these rules? At this level, decisions are made by groups of individuals rather than single individuals. We will attempt to understand both the impact of the structure of community organization(s) on the decisions made by groups related to primary processes as well as their investments or lack of investment in methods to overcome collective action problems. Some of the relevant hypotheses we will examine at this level include: * When forest resources are perceived to be valuable and scarce in a location, communities are more likely to design, monitor, and enforce rules limiting forest use. * Forests under a form of secure tenure that restricts use will be characterized by better forest conditions. * The value of forest products and labor depend on the geographic location of a community in relationship to urban centers and transportation networks. Local decisions are thus influenced not only by local characteristics but by how a local community is situated in a geo-political and economic milieu. * Individuals will invest resources (time and financial) in improving forest conditions where others in a community also have a strong commitment to contribute. * Where communities have been subject to substantial and rapid in migration, there is an increased probability of breakdown of local institutions related to resource use, greater contestation over access to resources, and degraded condition of resources. * A greater community investment in developing local rules related to forests is more likely to occur where (1) communities already have other kinds of local institutions, (2) interactions occur with other communities that govern and manage their forest resources, (3) the preferences related to types of forest use are relatively homogeneous, and (4) where there has not been a history of regional or national government expropriation of forest lands. * The impact of the size of a community on decisions affecting forest conditions will tend to be curvilinear: very small communities will have difficulty generating the resources needed to monitor and enforce rules and thus protecting forest conditions. Transaction costs and conflict are likely to be very high in larger communities. These relationships are also affected by the heterogeneity of interests, the distance of a community from a forest, and the level of interaction in a community. 11

At the third tier— the regional level— we find fewer actors who are taking direct action that impacts on particular forests. We find many more individuals and groups whose actions have an indirect impact because they affect the structure of incentives of those who are operating at the fourth and fifth tiers. Here we are interested again in explaining the kinds of actions taken and particularly the kinds of rules and policies adopted since they have a major impact on the fourth and fifth tiers. Some of the hypotheses we can explore by adding this level of analysis include: * The presence of larger, commercial firms with regional or larger operations, will be more likely to clear large areas of land to gain access to its commercial value. Thus, the impact of larger firms on forest resources will be greater than in locations with only smaller firms. * Larger, commercial firms concerned with acquiring land for commercial development, mining, or industrial purposes will have more impacts on forest resources than will firms focused on the acquisition, use, and sale of forest products. * If governments or NGOs intervene frequently in the institutions of a community, there is a lower probability that the community will create institutions regarding resource use. * If governments or NGOs provide institutions to help overcome collective action problems regarding a resource, there is a greater likelihood that a community will construct institutions regarding that resource. * If governments or NGOs do not enforce their policies, communities will ignore them and treat a forest resource as open access. * The more valuable a forest resource is, the more community members will defy government policies that seek to limit community use. At the first and second tiers— the international and national levels— many decisions are made that impact on the primary processes occurring at the fourth and fifth tiers. We will primarily use these as important exogenous variables rather than trying to explain international or national level decision making. If one adds these three tiers to a simplified picture of how decisions impact on forest conditions, one arrives at the highly simplified (but more complex) schema represented in Figure 3. [Figure 3 about here] There are many different types of decisions involved in the processes that affect forest conditions; thus, a single theory or model of this complex set of human decisions is not realistic. Individual studies will

12

usually examine a subset of decision processes. Drawing on the concepts of hierarchy theory in ecology, we will develop our explanations based on what processes are occurring at a focal level (level 0) by understanding mechanisms at the level below (level -1), and the constraints and opportunities of the level above (level +1) (see Allen and Hoekstra, 1992; Gibson, Ostrom, and Ahn, 1997). When focusing on the individual as a focal level, the level below becomes the internal structure and process of individual choice. The level above becomes the community tier of decision making. In order to explain decisions at this level, one has to use the assumptions about human behavior listed above to generate and test predictions about the likely behavior of individuals. Constraints and opportunities on household decision making come directly from the characteristics of the community in which an individual or a household is located. Further constraints and opportunities may come from the regional, national, and international levels as these impact on a community. Examples of hypotheses that link these levels are: * Individuals employed by firms are likely to have two perspectives: one reflecting their place of residence and a second reflecting the nature and location of their employment, whereas self-employed individuals or those engaged in subsistence activities will have primarily a single perspective on forest use. * Individual decisions regarding family formation, fertility and migration are affected by the comparative advantages represented by where they live as contrasted to opportunities available in other regions of a country or even in other countries. * If national politicians gain substantial resources (rent) from concessions to commercial forestry enterprises, they are less likely to enforce the rights of local users. If one shifts up to the community level as a focal level, the level below becomes individual characteristics and decisions while the level above becomes the decisions made by regional organizations and authorities (with national and inter-national organizations and authorities also impacting in a more indirect manner). Understanding the impact of national policies on the actions of individuals and communities is extremely important. Heath and Binswanger (1996: 65) argue, for example, that population pressure is an important factor influencing the sustainability of natural resources, but population pressure “exercises a much less critical impact than the overall policy framework.” As Moran (1993) has documented, credit policies for cattle ranchers plus government investments in roads fueled deforestation in Brazil during the 13

1970s and 1980s. Since the particular processes that affect deforestation may vary from location to location, simply calling for better national policies may be ineffective — or worse, in some locations — while highly efficacious in others. Thus, examining how mixtures of policies affect outcomes will be an important objective of this research (see also Bilsborrow, 1987; Turner, 1997). Since few research programs have adopted a multilevel approach to human decision making as it affects forest conditions, we face many theoretical, methodological, and empirical puzzles in this effort.

Processes, Measurement, Aggregation, and Comparisons Human decisions and physical and biological processes occur at many spatial scales that we group into local, landscape, regional, and continental scales. At each level there are processes that can be measured at that level and potentially at other levels through aggregation, and there are comparisons that can be made to test hypotheses about causal processes. At a local scale, the processes involve: * individual decisions within households, communities, or divisions of larger decision-making units as impacted by decisions of others at the same or higher tiers; and * small-scale biological and physical processes such as species composition and recruitment, soil nutrients, topography, solar radiation, and water availability as these are impacted by biological and physical processes occurring at larger scales. To understand these processes, measurements must be made at a local scale for an adequate sample of sites in a location. This sampling must include similar sites representing the typical patterns of forest conditions present at the location. In many cases there may be additional, adjacent sites that are “outliers,” which represent interesting and potentially important forest habitats of interest to the local population, or where the outcome of human actions has left a very different imprint upon the forest. From these samples of sites, useful comparisons can be made across households, farms, and forest stands; across human

14

communities; and across categories of human action such as migrant populations, communal tenure systems, and forest conservation. Local-level processes provide a rich array of variables to explore how decisions about forest uses are made. Differences in the age and gender composition of households structure how much labor is available within a household and thus how much labor can be used to substitute for capital, and how constrained such labor might be by other competing obligations. Wealth, in turn, can either substitute for labor, or it can become a force on its own to bring about rapid change. In traditional communities, wealthy families may be conservative in their use of wealth to transform forests because they are more concerned with maintaining their wealth and status, while in upwardly mobile and frontier communities, wealth can lead to rapid environmental change because initial wealth is seen as a means of further capital accumulation to ensure long-term wealth and status. Local land tenure is one of the most important factors affect aspects of local land use. The security of tenure may influence short-term versus long-term strategies of land use. But equally important is the history of this tenure and the perceived security of its status. Where land rights have been contested, and where violence has been used to obtain those rights, it is more likely that strategies will be short-term in nature due to the doubts of the local community in their ability to guarantee that the products of their longterm investments of labor and capital will be vested in their descendants. Religion and ethnicity may, in some cases, be very important to the decision-making processes of local people. Whether forests are seen as embodying, or being the home of, spirits can greatly influence how a community uses its forest resources. Religion and ethnicity embody values and traditions that often include valuation of natural resources and attitudes toward proper use. These may not always be followed, but provide a normative set of rules that local people take into account and that may act as a brake on other incentives present in their lives to act in contradiction to such values. At a landscape level, the processes involve: * decisions by large private and public actors as impacted by decisions of others at other tiers; 15

* the aggregation of decisions by individual households and communities; and * landscape level biological and physical processes affected by species composition, climate, and geomorphology. To understand these processes, two strategies may be followed: aggregating the site data to a landscape level, or collecting data sets at the landscape level itself. Using remotely sensed data (aerial photos, satellite digital data, radar imagery), it is possible to obtain considerable landscape level information about forest types, their areal extent, and the spatial and temporal distribution of land cover, topography, biomass, water availability, and transportation networks. The first step in data collection at this level is an unsupervised classification of land cover to determine the types of land cover present using the digital reflectance values from satellite imagery such as the Landsat Thematic Mapper and by rectification to reflectance values (Green and Schweik, 1997). These data can be further enhanced and refined with the use of aerial photos and radar imagery. These regional data can be an excellent guide to fieldwork, help design sampling of sites, and ensuring that the local and community site-level studies are truly representative of the landscape and the land cover change processes of interest. When combined with sitespecific forest mensuration, field-level geocorrection, and ground truthing using a global position system (GPS), the level of accuracy of the higher tier of analysis is increased by incorporating more detailed information both spatially and temporally. For example, if a landscape is 60% tropical deciduous forests and 40% is tropical moist forests, field samples can be selected to represent these proportionalities, or if 95% of the tropical deciduous forests are heavily used by humans, but only 20% of the tropical moist forests are, the sampling strategy can be further modified to understand why the communities focus more on deciduous rather than moist forests in their land use. At the site-specific level, soil analysis can determine differences in texture or nutrient availability. Forest inventory can determine species composition, tree size, age, and biomass, as well as local topographic effects. Finally, individual and household surveys could determine if there is an association of moist forests with dangerous spirits and richer hunting resources, or the ownership of the moist forests by a powerful family or clan.

16

At a national or regional level, the processes involve: * decisions by private and public actors at this tier of decision making as impacted by decisions of others at other tiers; * the aggregation of decisions by individual households, communities, local governments, private firms at lower tiers; and * large, often continental scale, biological and physical processes operating from seasonal to geological time scales. To understand these processes, some measurements made at a local and landscape level can be aggregated to a national or regional level for some purposes, but other measurements with coarser resolution may not be necessary for national or regional levels. Useful comparisons can be made across nations, regions or decision-making units located within different countries or regions. At this level, population data are reasonably available, although variable in quality. The decadal census provides a comprehensive timeseries source of information at the county, regional and national levels, not only on household composition but on spatial distribution, major economic activities, institutions, and ethnicity. Linking the national census to remotely-sensed imagery is currently an interest of a number of investigators (e.g., Wood and Skole, 1998; Wood and Perz, 1996).

Conclusions Studying the human dimensions of global environmental change presents numerous challenges that reliance on purely discipline-derived theory and method cannot begin to address. This paper provides a starting set of propositions addressing five tiers or levels of analysis that are of interest to studies of global change. As the paper has attempted to demonstrate, each tier elucidates different types of human interaction and gives priority to different variables in explaining the measurable variables at that scale of aggregation or disaggregation. Temporal variability, too, varies across levels in its magnitude and rates of change. Generational change is more likely to explain processes at higher levels of aggregation than at the level of the household, although even at this level, concerns with production and reproduction across generations may play significant roles. 17

The focus used in this paper, on forest ecosystems, provides a natural comparative laboratory for examining the complex interactions between human actions and environmental outcomes and their feedbacks. Human decisions, at any one of the five levels of analysis, play a central role in shaping the condition of forests. These decisions while rational, are constrained by cultural values, the quantity and quality of the information available to them, the incentives offered, and the learning and choice making routines that they have been socialized into using. These individual decisions occur within situations where these same individuals act as members of institutions at one (e.g., family) or multiple levels (family and city environmental commission) of aggregation wherein synchrony and/or conflict may be present in the incentives, values, and culture of the respective institutions within which the decision occurs. Our efforts to understand why households, communities, and regions make the decisions that they do with regard to the quantity and quality of their forest ecosystems have led to a set of methods, approaches to measurement, and for testing a number of propositions. These methods have come from the social, biological and physical sciences. They have been brought together in the interest not only of addressing the questions of interest to this research but as an experiment in how these sciences' methods can be adjusted and integrated to work in as parsimonious a fashion as possible in addressing the complex processes encompassed by global environmental change occurring at levels from local to global. One of the tools for accomplishing this goal is to link the traditional tools of survey research and the national census with the earth science tools of remotely-sensed data and geographic information systems. The latter offer a technical set of procedures for examining change in land cover at whatever scale of resolution is offered by the data acquisition instrument (anywhere from one meter to several hundred meters), often available with multiple data reacquisitions which facilitate time-series analysis, and from which human action can be inferred if the resolution is adequate enough for observing the impact of those actions at a given scale of analysis. Resolutions of 10 to 30 meters are now widely available. Within the next two years, resolutions of one to three meters are expected, facilitating still more the analysis of human action from remote platforms. 18

In short, the combined use of earth science and social science tools for the analysis of human decisions about the use of natural resources offers the potential not only for addressing practical problems presented by global change but also as a powerful tool for testing hypotheses and generating theories at any number of scales of aggregation. This is one of the greatest challenges offered by the human dimensions of global change research agenda. The approach proposed in this paper is one effort to move this agenda forward in both theoretical and methodological terms.

19

Notes

1. These measures are defined in many standard textbooks on forest mensuration and forest ecology. 2. This does not preclude our studying particular decision makers in some of our work in our effort to understand the structure of specific situations as they affect the values, information, time horizon, and internal processes of learning and making decisions. Just as we intend to scale up and scale down in our analysis of forest ecosystems, we also will shift our focus from time to time from the external structure of situations to understanding how such structures are viewed by particular individuals and why they made the particular decisions that they did.

3. In some countries there will be at least two if not more “regional” governments. In the United States, for example, county governments are frequently large enough to be considered regional as well as state governments.

20

Bibliography

Allen Timothy F.H., and Thomas W. Hoekstra. 1992. Toward a Unified Ecology. New York: Columbia University Press. Andersen, Lykke E., et al. 1996. Report on Amazon Deforestation. San Diego: University of San Diego, Department of Economics, Discussion paper 96-40. Bilsborrow, Richard E. 1987. “Population Pressure and Agriculture Development in Developing Countries: A Conceptual Framework and Recent Evidence.” World Development 15(2):183-203. Bilsborrow, Richard E., and H.W.O. Okoth Ogendo. 1992. “Population Driven Changes in Land Use in Developing Countries.” Ambio 21(1) (Feb.): 37-45. Brown, Katrina, and David W. Pearce, eds. 1994. The Causes of Tropical Deforestation. Vancouver, British Columbia: UBC Press. Cancian, F. 1972. Change and Uncertainty in a Peasant Economy. Stanford, CA: Stanford University Press. Dawes, Robyn M. 1980. “Social Dilemmas.” Annual Review of Psychology 31:169-93. Dodds, David. 1997a. “A Multilevel Analysis of Population Growth in Eastern Honduras.” Bloomington: Working paper. Indiana University, Center for the Study of Institutions, Population, and Environmental Change. Dodds, David. 1997b. “Lobster in the Rain Forest: The Political Ecology of Miskoto Wage Labor and Agricultural Deforestation.” Working paper. Bloomington: Indiana University, Center for the Study of Institutions, Population, and Environmental Change. Durham, William H. 1995. “Political Ecology and Environmental Destruction in Latin America.” In The Social Causes of Environmental Destruction in Latin America, ed. Michael Painter and William H. Durham, 249-64. Ann Arbor: University of Michigan Press. Fox, Jefferson. 1993. “Forest Resources in a Nepali Village in 1980 and 1990: The Positive Influence of Population Growth.” Mountain Research and Development 13(1):89-98. Gibson, Clark, Elinor Ostrom, and Toh-Kyeong Ahn. 1997. “Scaling Issues in the Social Sciences. A Report for the International Human Dimensions Program.” Working paper. Bloomington: Indiana University, Center for the Study of Institutions, Population, and Environmental Change. Gimblett, Randy. 1997. “RBSim - Recreation Behavior Simulator.” Tucson: University of Arizona, School of Renewable Natural Resources. Green, Glen. 1997. “Making Satellite Image Radiometric Calibration Palatable: Guidelines for the Global Community.” Working paper. Bloomington: Indiana University, Center for the Study of Institutions, Population, and Environmental Change.

21

Heath, John, and Hans Binswanger. 1996. “Natural Resource Degradation Effects of Poverty and Population Growth are Large Policy-Induced: The Case of Colombia.” Environment and Development Economics 1(1):65-83. Johnson, A. 1972. “Individuality and Experimentation in Traditional Agriculture.” Human Ecology 1:14959. Lambin, E. 1994. “Modelling Deforestation Processes: A Review. Trees Report.” Ispra, Italy: Institute for Remote Sensing Applications. Moran, Emilio F. 1993. “Deforestation and Land Use in the Brazilian Amazon.” Human Ecology 21(1):121. Moran, Emilio F., et al. 1997. “Effects of Soil Fertility and Land Use on Forest Succession in Amazônia.” Working paper. Bloomington: Indiana University, Center for the Study of Institutions, Population, and Environmental Change. Ostrom, Elinor. 1986. “An Agenda for the Study of Institutions.” Public Choice 48:3-25. Ostrom, Elinor. 1997a. “Institutional Rational Choice: An Assessment of the IAD Framework.” In Theories of the Policy Process, ed. Paul Sabatier. Boulder, CO: Westview Press, forthcoming. Ostrom, Elinor. 1997b. “What Makes for Successful Institutions to Govern Common-Pool Resources.” Working paper. Bloomington: Indiana University, Center for the Study of Institutions, Population, and Environmental Change. Ostrom, Elinor, Roy Gardner, and James M. Walker. 1994. Rules, Games, and Common-Pool Resources. Ann Arbor: University of Michigan Press. Oye, Kenneth, and James Maxwell. 1995. “Self-interest and Environmental Management.” In Local Commons and Global Interdependence: Heterogeneity and Cooperation in Two Domain, ed. Robert Keohane and Elinor Ostrom, 191-221. London: Sage. Park, Soohong, and Daniel F. Wagner. 1996. “Cellular Automata as Analytical Engines in Geographic Information Systems.” Bloomington: Indiana University, Center for the Study of Institutions, Population, and Environmental Change. Pickett, S.T.A., and M. L. Cadenasso. 1995. “Landscape Ecology: Spatial Heterogeneity in Ecological Systems.” Science 269 (July 21): 331-34. Popper, Karl R. 1967. "Rationality and the Status of the Rationality Principle." In Le Fondements Philosophiques des Systemes Economiques Textes de Jacques Rueff et Essais Rediges en son honneur, ed. E. M. Claassen, 145-50. Paris: Payot. Root, Terry L., and Stephen. H. Schneider. 1995. “Ecology and Climate: Research Strategies and Implications.” Science 269 (July 21): 334-39.

22

de Silva-Forsberg, Maria Clara, and Philip M. Fearnside. 1997. “Brazilian Amazonian Caboclo Agriculture: Effect of Fallow Periods on Maize Yield.” Working paper. Bloomington: Indiana University, Center for the Study of Institutions, Population, and Environmental Change. Simon, Herbert A. 1985. “Human Nature in Politics: The Dialogue of Psychology with Political Science.” American Political Science Review 79(2) (June): 293-304. Simon, Herbert A. 1997. Models of Bounded Rationality: Empirically Grounded Economic Reason. Cambridge, MA: MIT Press. Snidal, Duncan. 1995. “The Politics of Scope: Endogenous Actors, Heterogeneity and Institutions.” In Local Commons and Global Interdependence: Heterogeneity and Cooperation in Two Domain, ed. Robert Keohane and Elinor Ostrom, 47-70. London: Sage. Tiffen, Mary, Michael Mortimore, and F. N. Gichuki. 1993. More People. Less Erosion: Environmental Recovery in Kenya. New York: John Wiley. Tucker, Catherine M. 1997. “What’s the Difference between Private and Communal Forests? Preliminary Indications from a Honduran Community.” Working paper. Bloomington: Indiana University, Center for the Study of Institutions, Population and Environmental Change. Tucker, Joanna, Eduardo Brondizio, and Emilio Moran. 1997. “Rates of Forest Regrowth in Eastern Amazonia: A Comparison of Altamira and Bragantina Regions, Para State, Brazil.” Working paper. Bloomington: Indiana University, Center for the Study of Institutions, Population, and Environmental Change. Turner, Paul. 1997. “Constitutional Orders and Deforestation: An Analysis of the Humid Tropics.” Chapters from Ph.D. Dissertation to be completed during the spring of 1998. Wood, Charles, and Steve Perz. 1996. “Population and Land Use Change in the Amazon.” In Population Growth and Environmental Issues, ed. S. S. Ramphal and Steven Sindit, 95-108. Westport, CT: Praeger. Wood, Charles, and D. Skole. 1998. “Linking Satellite, Census, and Survey Data to Study Deforestation in the Brazilian Amazon.” In People and Pixels, ed. D. Liverman, E. Moran, R. Rindfuss, and P. Stern. Washington, D.C.: National Academy Press, forthcoming. Young, Oran. 1995. “The Problem of Scale in Human/Environment Relationships.” In Local Commons and Global Interdependence: Heterogeneity and Cooperation in Two Domains, Robert Keohane and Elinor Ostrom, 27-45. London: Sage.

23

Figures 1, 2, and 3 (Framemaker 4)

24

25

26