The previous chapters suggest that to build knowledge of the human causes and consequences of global environmental change, advances in theory, method, data, and research infrastructure are all imperative. This chapter focuses on the needs for data. We begin with three illustrative research problems and then proceed to a systematic account of issues of data availability, data quality, and collection needs.
THREE HYPOTHETICAL RESEARCH PROJECTS
CAUSES OF GLOBAL WARMING: A CROSS-NATIONAL STUDY
Global change researchers want to understand the relative contributions to global warming of population growth, economic growth, technological change, and various aspects of national policy and social and economic structure. One way to shed light on the issue would be through analysis of time-series data on each country in the world over an extended period; data series are available for many variables in many countries. Such a research project ideally would use some or all of the following data:
Atmospheric Gas Concentrations CO2, methane, CFCs, etc.
Greenhouse gas releases Same gases
National population Total, urban-rural split, agricultural-nonagricultural, birth rates, age distribution
Economic activity Gross national product (GNP); production of manufacturing, mining, agricultural, service sectors; income distribution
Energy demand Consumption of coal, oil, gas, nuclear electricity, hydroelectricity, biomass; demand for transportation, space heating, manufacturing process heat, and other major end uses
Technology Transportation subdivided by technology used (e.g., number of passenger miles by automobile, number of ton-miles by train); capacity and fuel efficiency of industrial boilers, steel furnace, etc.
Prices Major energy sources, labor, agricultural land; interest rate for capital investment
Land use Hectares in wetland crops, dryland crops, pasture, forest; number of cattle; dispersion of population around urban centers
Institutions Distribution of land across agricultural users and by land-tenure system; degree of market vs. nonmarket control over markets for energy, land; environmental regulatory style
These data, even when they are desired only on an annual basis and at the national level, are unevenly available. Some, such as national population, are fairly accurate and available, but too infrequent; some, such as GNP, are available and sufficiently frequent, but of highly variable and sometimes unknown accuracy; some, such as energy use subdivided by economic sector, are sufficiently accurate and frequent for some countries (typically, OECD countries), but not available at all for many other countries; and some, such as the degree of market control of the economy, are unavailable because a reliable index does not yet exist. For most of the international data sets that are available, information on data quality is scanty and of uncertain reliability.
HUMAN CONSEQUENCES OF DEPLETION OF STRATOSPHERIC OZONE
Global change researchers want to be able to monitor the possible effects on things humans value of ongoing changes in the level of stratospheric ozone and to attribute the effects to ozone depletion rather than other possible causes. One way to shed light on the issue would be continuing analysis of a time series of data on values that might be affected by increased UV-B radiation at the earth's surface. These include the productivity of crop and forest flora,
human incidence of skin cancers and resistance to infectious disease, and the economic consequences of such changes. Such a research project ideally would use data such as the following:
Health Data at national level or in selected localities on deaths from skin cancer, incidence of skin cancer, and incidence of selected infectious diseases
Agriculture Productivity data on selected crops; vegetation density in forests
Economic Value of selected agricultural products in countries or selected localities; health care expenditures
Stratospheric ozone concentration Measured values from satellite observation for selected regions
UV-B incident radiation Measurements at ground level
Policy responses Enactment of policies to limit CFC releases; implementation of such policies; introduction of education campaigns for avoidance of direct exposure to sun; measures of resultant behavior change
Data on these variables are also uneven. A study of the effects of ozone depletion presents some of the same data problems illustrated by the study on the causes of greenhouse gas emissions, as well as others. For instance, data available summarizing political units (e.g., local or country-wide health data) must be compared with data available summarizing physical areas of map grids (e.g., satellite observations of ozone levels or color of vegetation) and with data collected at selected time points (e.g., incidents of UV-B radiation, surveys of individual beliefs or behaviors). Some data, such as satellite images, are available so frequently that techniques of time sampling must be used. Some data, such as that on human health effects or agricultural productivity, may not be available at the desired geographical level or on the precise effect of interest. Data from diverse methodologies, such as remote sensing, survey research, and health records, are generally not available from the same archives or in mutually intelligible form. And finally, some of the effects of interest are as yet unknown, so the monitoring process must leave room for adding other variables and reanalyzing the data.
ECONOMIC AND NONECONOMIC FORCES CAUSING LOSS OF BIOLOGICAL DIVERSITY
The activities most implicated in the loss of biological diversity are influenced by human demographics, technologies, political sys-
tems, economics, traditions, and belief systems. In particular, understanding how individuals, tribes, ethnic groups, and nations regard species and ecosystems and reconciling economic and ecological methods of accounting are integral to solving the biological diversity crisis. Data needed to achieve a comprehensive understanding of the loss of biological diversity might come from studies of:
Intensive versus extensive land use Evolution of intensive versus extensive systems; how human-land relationships affect intensity of use
Markets, laws, and traditional values How economic, legal, and social systems affect which species and ecosystems can be used and which cannot
Consequences of economic measures Effects of gross national product, discount rate, and other measures on biological diversity
Alternative systems of economic valuation Mechanisms that could address adverse effects of neoclassical and Marxian paradigms on living systems; use of common, biologically active elements (e.g., nitrogen, carbon) instead of rare, inert ones (e.g., gold, silver) in accounting systems
Socioeconomic determinants of conservation Comparing effects of different standards of living on society's conservation ethics and different conservation ethics in societies with comparable standards of living
As is the case with the other examples, data on these types of variables are uneven. Some of the independent variables consist of taxonomies of political and economic systems, such as types of markets, legal institutions, or systems of land use. Some of the dependent variables are available on an annual basis (measures of biological diversity, societal standards of living, intensity of land use) while others must be generated from cross-national surveys (comparisons of attitudes toward conservation).
Generally, the data problems raised by such hypothetical studies are the major ones facing quantitative research on the human dimensions of global change. They are, briefly:
availability of and access to existing data;
quality control and interpretability of data;
lack of necessary data for parts of the world;
inadequate time series for some variables;
lack of measures of some key variables; and
incompatibility and incommensurability across data sets in geographical or temporal scope.
Data are available on some social phenomena that are relevant to the human dimensions of global change: for example, trends in population growth, dispersion, and mortality, per capita energy use by nation and region, health and disease statistics, agricultural productivity for nations and regions, various indices of economic and industrial growth and decline, and political events related to the stability of governments. In fact, there is so much statistical information that it was beyond the capacity of the committee to judge its quality or its adequacy for the study of global change. Few if any scholars have a clear sense of the full range of data that are available.
Since the conclusion of World War II, many aspects of social science have been revolutionized by the increasing availability of quantitative data about social phenomena throughout the world and of computers that permit large quantities of data to be analyzed quickly and easily. Traditional administrative and census techniques have been used more intensively to gather more and more fine-grained data and more extensively throughout all areas of the world, and huge archives have been built. The sample survey, which has been greatly refined in the last 50 years, has made it possible to generalize to entire populations from a relatively small, carefully drawn sample and can therefore supplement censuses. Surveys also make it possible to gather data about issues that can be explored only through interviews.
Massive data collection efforts, beginning in the 1930s in the Western countries, have aided in the development of social theory. In some cases, data were collected because theories demonstrated that they would be useful to understanding how social phenomena interacted. National income accounts are a prime example of this interaction between the development of theory and the collection of data. In other cases, theory has allowed exploratory data analysis and the building of inductive theory.
WHAT IS AVAILABLE
The national-level data sets published by the United Nations and other international agencies are global data on human phenomena. After World War II, these agencies developed standards for gathering data, trained statisticians throughout the world, and inaugurated a vast array of serial publications of statistical data. The United Nations and most of its related agencies such as the
International Bank for Reconstruction and Development (World Bank), the International Monetary Fund, the International Labour Organisation, the Food and Agriculture Organization, and the World Health Organization have developed serial publications containing vast quantities of data. Regional organizations, such as the Organisation for Economic Co-operation and Development (OECD), have also developed and published their own statistical series.
Important series include the World Bank's World Tables, the UN's Statistical Yearbooks, Demographic Yearbooks, and World Population Trends and Policies, the OECD's National Accounts and Economic Surveys (annual reviews of member countries' economies). The United Nations Environment Programme and OECD both publish series that deal directly with environmental issues, although the data in them are much more limited than in the economic and demographic series. The quantity of data published by international institutions is huge. The World Bank's World Tables include data on 138 countries for a wide range of economic and social phenomena. The UN's biennial ''Monitoring of Population Trends'' is an excellent global synthesis of national population trends, containing demographic data on over 200 countries. Beyond being available in publications, many of these data are available in machine-readable form to individuals and organizations not affiliated with the international institutions.
In addition, institutions exist that receive, archive, and disseminate social science data. For example, the Inter-university Consortium for Political and Social Research (ICPSR), a consortium of more than 300 member universities that is administratively located at the University of Michigan, manages an archive of more than 2,000 data collections, 28,000 data files, and 190 billion characters of information. The collection includes many important surveys of individual attitudes, as well as machine-readable data sets from the U.S. Census and from international agencies. However, the collection of data sets is neither global (most of the studies are restricted to one country) nor systematic (the archive includes what was sent, rather than all data sets on any particular subject, or data sets on all the topics in a predetermined domain). Other similar archives are the University of Connecticut's Roper Center for Public Opinion Research national sample survey and the Louis Harris Data Center at the University of North Carolina, Chapel Hill.
Research on human dimensions also requires data on the non-human variables that affect or are affected by human action, such as trace gases and pollutant levels, land cover, precipitation and
frost patterns, incidence of UV radiation, and so forth. The bulk of potentially relevant information far surpasses information on the human dimension itself. Many of the data sets are available from central sources, and there have been some efforts to build data catalogues that would allow researchers to find what they need. However, these efforts are not far along; they typically have little or no input from researchers who might use the archives to study the human dimensions of global change.
There is, however, a trade-off between using existing studies and collecting new data. On one hand, existing data are attractive because the collection costs have already been paid. On the other hand, it may well be the case that the costs of search and assembly of existing data are underestimated and their value in terms of relevance and quality overestimated. These issues must be evaluated in terms of the relative cost-effectiveness of both strategies. The time to make such assessments is before a decision is made to invest in large-scale data assembly or cataloguing (for extant data) or data collection (for new data) efforts. And the assessments should be based on a modest but representative sampling of available materials on key variables for pivotal nations or regions of the world. Issues surrounding such assessments, in terms of articulating criteria for judging quality and costs, are elaborated in some detail in the remaining sections of this chapter.
It is clearly necessary before all else to make existing data available to researchers so that they can find out what can be learned without expensive collection of new data. To do this, steps must be taken to establish a viable information network.
AN INFORMATION NETWORK
Because research on global change is so broad and draws on data, empirical research, and analysis from a wide variety of disciplines and countries, few social or natural scientists are familiar with more than a narrow portion of it. Giving them access to the data they need presents a major task for managing, storing, and distributing scientific information. Meanwhile, the electronic revolution has created a major upheaval in the way information is managed. The speed, density, and cost of storing, processing, and distributing information is changing more rapidly than at any time since the age of Gutenberg. However, our institutions—particularly our libraries—are adjusting to the electronic revolution much too slowly.
Putting these two points together, we believe that intensive development of electronic information management in the social, natural, and environmental sciences holds significant potential for major breakthroughs in empirical work on global change. It could both encourage global change research and help bridge the wide gaps among the disciplines engaged in global change research. By harnessing the electronic revolution, we can form "invisible colleges" of people in different locations and disciplines who work closely together.
An appropriately designed public information system would be a powerful equalizing force in the scientific community. Currently, those who are firmly ensconced in disciplines—especially those at major research universities with amply funded libraries and computer centers—have major advantages over those outside existing centers. A public information network would level the playing field for all researchers—a scientist in Oshkosh or Baton Rouge would have the same access as one in Cambridge or San Francisco. The Inter-university Consortium for Political and Social Research already plays this role for much social science research—thousands of important data sets are available to all member institutions at nominal cost.
It should be noted that a public information system would benefit the entire community of researchers on phenomena related to global environmental change. It would not be limited only to those focusing on the human dimensions of the problem or those whose work concentrates primarily on natural phenomena. To be of any use at all, the information would need to include data on natural or physical changes as recorded, for example, by remote sensing technologies, as well as changes that occur on the many dimensions of human activities that interact with the natural physical changes. (Of course, as noted above, there are huge problems of matching data collected with different conceptual frames and units of analysis.) Such an information system would contribute valuable resources for the interdisciplinary projects needed to advance the field.
The major need at this point is for governmental and private support for the necessary infrastructure for publicly shared data on demographic, economic, political, attitudinal, and natural or physical changes. This includes one-time costs, such as putting material in electronic form and building a network and archival facility to make the material accessible to researchers. The key components include facilities for storage of information; a high-speed network for exchange of information; and local facilities
for receiving and processing data along with generating new knowledge and information. The National Research Council's Committee on Global Change recommended: "... that data be systematically archived, prepared, and disseminated at a central site and made readily accessible to interested researchers" (National Research Council, 1990b:127). While we generally support this recommendation, the committee is divided on whether or not the data and information system should be concentrated in one location. A skeletal description of the needed system might be the following:
Storage Given the economies of scale, there should be a small number of facilities that are information libraries for large data bases, working papers, and documentation in global change. These would not need to be at one central location; specialization should be encouraged. Each center would operate as a public utility, providing services at close to marginal cost with overhead covered by federal subvention or subscriptions. Again, several existing data centers serve as models.
Data transformation Data gathered on different temporal and spatial scales will need to be stored in a form that allows transformation to a variety of spatial or temporal formats. The conceptual and methodological issues involved in storing such data for easy transformation should be addressed. There is a substantial foundation of ongoing work in such areas as epidemiology and geographic information systems that can serve as a basis.
Cataloguing An inventory of archived data should enable potential users to identify sources of data on the variables of interest to them, get critical information on the spatial and temporal characteristics of the data, and select data sets for their particular needs.
Transmission The centers and major research institutions would be connected by a variety of different media. A high-speed network would be desirable for linking researchers and archives together. Telephonic communications would be a backstop for those not on a major data link or can be used for a backup. And of course, physical shipment of tapes or disks involves delays of only a few days.
Local facilities Each research institution would decide how best to configure its local environment and would have the responsibility to determine how best to participate in data networks.
Computation Given progress in development of work stations and in parallel processing, we believe computation is likely to be
the least important part of the problem of better use of information. More important is likely to be the method of connection to the network and the cost of and access to a variety of software.
Some of the elements of such an electronic information network already exist. There is, however, much work to be done in completing the system, particularly with respect to making existing material easily available to remote users. It is noted, however, that expenditures on such a system should not jeopardize needed resources for doing the research. Any electronic data system should facilitate, rather than substitute for, the time-consuming effort of attempting to understand the data, namely, how variables are defined, measured, aggregated, commensurated, and so on. The ultimate goal of these systems is to support the process of doing research. We are especially concerned about the possible temptation to expend valuable resources on the collection of data for its own sake. Data collection and dissemination efforts must be closely coordinated with research designs and analysis plans that can guide decisions about what to collect and how it will be used in analysis.
One piece of important work for research on the human dimensions of global change involves making data from remote sensing instruments intelligible for social scientific purposes. For example, time-series observations of the land area covered by buildings may be a useful proxy for measures of population and economic activity in rural, Third World areas for which direct measures are poor. The data can certainly be updated much more frequently and at much lower cost. A careful analysis should be done of the use of remote sensing for the measurement of social phenomena. Specifically, an effort to identify existing remote-sensing information useful for human dimensions studies and to transform the relevant data into indices of human activity would be a valuable one-time investment in a global change information system.
The Earth Observing System (EOS) is a new remote sensing system for the study of global change created by he National Aeronautics and Space Administration (NASA). EOS will have important new capabilities. Social scientists should be involved in its planning to ensure that the potential of this system for studying the human dimensions of global environmental change is fully realized. In addition, many of the data requirements for research on the human dimensions of global change involve improved methods for ground observation.
We make the following recommendations:
An information network should be designed for global change research (including the human dimensions) that will make catalogues, data bases, data transformation, documentation, scientific literature, and other public information more widely and inexpensively available. On the social science side, the network should include measures of the major driving forces of global change at the lowest available level of aggregation. The decision on how to organize the network should be based on advice from a group widely representative of natural and social scientists, information systems specialists, and librarians and archivists.
As part of the network, we recommend that the federal government should seriously consider establishing a national data center on the human dimensions of global change parallel to the existing national centers for data on climate, oceans, geophysics, and space science. We are aware of the dangers of delegating responsibility for a national data center to an agency that may not be oriented to understanding the data needs of human dimensions research or to promoting a strong research effort in this area. We therefore recommend that an independent advisory committee, composed of researchers working on the human dimensions of global change and including strong representation of social scientists, be set up to oversee the work of any such federal program that may be established.
The federal government should support an effort to examine the utility of existing and forthcoming remote-sensing information as indicators of particular human activities, to validate the most promising remote-sensing indicators against ground data, and to transform the data into indicators and include them in the information network. Whenever possible, this effort should include social scientists in the process of designing and specifying the capabilities of EOS and other remote-sensing systems for measuring global change.
Social scientists should serve on the EOS Science Advisory Panel and on any advisory panels established for the EOS Data and Information System.
COSTS OF DATA
The increasing commercialization of data is a troubling trend and a roadblock to scientific research. More and more, many of the useful data bases are being provided at extremely high cost,
not only by commercial vendors (like Data Resources, Inc.) but also by quasi-governmental agencies (like the IMF or OECD). For example, a complete set of the OECD data costs tens of thousands of dollars annually—far above the marginal cost of providing the data. Clearly, pricing of major data sources to recoup costs of data collection will inhibit global change research that necessarily relies heavily on comparative international data in economics and other areas. This is an area in which the U.S. government could help by encouraging international institutions to make data available at reasonable cost to nonprofit researchers. The committee recommends that the United States government should take steps to prevent pricing of basic data at levels far above the marginal cost of production. The government should try to influence international institutions such as the OECD to do the same.
QUALITY AND INTERPRETABILITY OF DATA
In any research effort, careful attention must be paid to issues of data quality, and efforts to better understand the human dimensions of global change are no exception. But for several reasons the problem of data quality may be more substantial in this area than in other research efforts involving social science. First, the volume of data that will be generated is enormous. NASA estimates that the EOS will generate a terabyte of data a day when the full system, described as "Mission to Planet Earth," is in place. To illustrate the magnitude of this data stream, the daily volume is sometimes referred to as a LOC, meaning Library of Congress, because a terabyte is an amount roughly the size of the library's contents. Managing that volume of data in a way that will make it useful is an enormous problem. But in addition to the management problems, considerable effort will be required to ensure that the data being generated have reasonable levels of reliability and validity.
Second, much of the data will consist of measurements on variables that have not been given much methodological attention in the past. Considerable effort has been devoted to the measurement of fertility rates, gross national product, public opinion, etc. In contrast, relatively little attention has been given to reliable and valid ways of measuring deforestation rates, energy efficiency, social movement mobilization, and so on. Certainly, experience in developing reliable and valid measures can transfer across concepts, but applying general methodological knowledge to the specifics of global change research will require a substantial effort.
Third, many variables will be measured with new technologies. Of particular importance is the potential for making use of data from satellites and other forms of remote sensing. In the physical and biological sciences there is some tradition of using remotely sensed data, but it is a novelty to the social sciences. Very little effort has been devoted to developing methodologies that would make such data useful for understanding the human dimensions of global change. Indeed, as we note elsewhere, the social sciences, with the exception of geography, have paid little attention to space as a concept and will have to develop new theory and methodology to accommodate global change phenomena that have a strong spatial component. It is especially important that social scientists become involved in the planning of remote sensing systems and the derived data bases. If social science concerns are not introduced early on, information that could be of great value in better understanding the human dimensions of global change may not be collected and archived in a form that optimizes its utility.
Fourth, much of the data of greatest interest will have to be collected on a regional, international, or global basis. The problems of aggregation from smaller to larger units and of assessing comparability across different local or national measurements will be important. There is considerable experience in dealing with these problems in the fields of economics, demography, and comparative politics. But that experience will have to be translated so as to be useful in measuring the very different variables important to global change.
Fifth, the conceptualization and measurement of many key variables must be interdisciplinary. In most cases sound conceptualization and measurement must span several social science disciplines; in many cases it will involve the physical and biological sciences as well. For example, a land use taxonomy must be sensitive to the biological and physical characteristics of a particular land use, such as species diversity that typically accompanies it, its albedo, and its percolation rate. But the taxonomy must also be sensitive to the economic function of the land use, its political regulation, and its cultural meaning. It may be that several taxonomies are required, but if so they must be coordinated to produce useful data.
Finally, although quality and interpretability is of the utmost importance for research on the human dimensions, attention must also be given to issues of cost-effectiveness. New large-scale data collections are costly enterprises. Just as some research projects would be given priority over others, so too would some variables
take precedence in data collection efforts. It will be necessary to develop criteria for determining where to place investments in data-collection efforts. In part, this is a problem of developing an accounting system similar to that described in Chapter 3. It is also a problem of estimating the quality of data available on the high-priority variables. For these reasons, it is important to include, as part of the assessment of quality, the likely costs involved in obtaining the data. Importance, quality, availability, and cost-effectiveness are all criteria for judging the feasibility of mounting large data-collection efforts.
PROBLEMS OF RELIABILITY AND VALIDITY
Threats to validity and reliability are not unique to data dealing with global change, so we can draw on the more general methodological literature on techniques to improve measurement. In some cases that literature is quite substantial; in other cases it is not. The issues noted above suggest that existing methods can be a source of valuable insights but will require some translation and modification as they are applied to global change. With this is mind, we believe that there are four major issues that should be given special attention.
First, the raw numbers used to generate a measurement may themselves be inaccurate because of errors in survey responses, official records, and so on. In some cases these errors are random and simply reduce reliability; in others they are systematic and thus introduce bias and reduce validity. The errors may be caused by the happenstance of human error and imperfect memory or by intentional obfuscation and deception. Decades of experience with official statistics, survey procedures, censuses, and so on have generated a rich repertoire of methodological tools for identifying and minimizing these sources of error (Bollen, 1989).
Second, there are problems of sampling and coverage. Much of statistical analysis is directed toward understanding sampling error, but nonreporting and other sources of bias remain a very difficult problem. The biases introduced by poor coverage are likely to be substantial in research on global change. It is the affluent industrialized nations that generally have the most sophisticated statistical systems and the richest data bases, while the poorer nations often have no estimates for many key variables. For example, estimated annual emissions of nitrogen oxides are available for 19 industrial nations, but for only 3 developing nations (World Resources Institute, 1990: Table 24.6). The
same problem applies within a nation or region. The middle class and the urban are more likely to be included than the poor or rural in surveys and other data collection efforts. In the 1950s and 1960s, when concern with population growth was at the center of the international research agenda, the problems of coverage for basic demographic data such as fertility rates and the prevalence of contraception were similar to those that exist now for key global change variables. Careful efforts over several decades have greatly improved coverage as well as the quality of measurement. We hope a similar effort can be undertaken to improve understanding of the human dimensions of global change.
A related issue arises when it is necessary to estimate values on key variables for some countries or regions. Estimation may be required because of a lack of coverage. It may also be necessary because the reported data are suspect. In many cases, data are supplied by a local or national body that may have a strong political interest in the values reported and may distort measurements either intentionally or by using methods that are biased. In other cases, the lack of infrastructure may prevent the local or national body from providing reliable and valid data. Whatever the reason, if the reported data are badly flawed, it is advisable to use multiple methods in order to cross-check results. Care must be taken to carefully evaluate these estimation procedures. Again, the history of demography and economics since World War II suggests that it is possible to develop sound estimation and validation procedures to supplement official reports and primary data. Of course, the ability to develop estimates and checks depends on the character of the variables being estimated. For many important demographic and economic variables, the logic of accounting and internal consistency can be applied to estimation. This should also apply to some of the key variables in global change.
A fourth problem arises because different jurisdictions may use different definitions and thus produce noncomparable data. Variables must be defined taking account of theoretical and practical considerations, and both theory and practical context will vary from place to place. Again, demographers and economists have considerable experience dealing with problems of comparability.
RESEARCH ON MEASUREMENT PROBLEMS
We begin by raising the general point that methodology can be improved by careful, focused research. We suggest that projects intended to enhance knowledge of the human dimensions of glob-
al change should be assessed not only on the quality of the methodology used but also on their likely contribution to advancing methodology. We also suggest that some projects that are intended primarily to enhance methodology should be supported.
For many variables important to the study of global change, there is a long chain of aggregation and integration that leads from data collected in the field to published reports or data bases. The procedures used at any step of the process may degrade reliability and introduce bias, but those problems may not be apparent in the final data table. It is easy to find published values of gross national product per capita, rate of population growth, or CO2 emissions for most nations of the world, but it is more difficult to document the full chain of research that produces those numbers for each nation. Much cross-national research simply accepts the published figures at face value, ignoring serious problems that may be hidden in their production. A few case studies that trace the production of data on key variables would be very helpful. Much methodological information seems to be available only in the minds of official statisticians or in a fugitive literature of internal documents, manuals, and memos. By examining the production process, we will develop a better sense of potential threats to validity and reliability and thus will be in a better position to recommend improved methodology and offer appropriate cautions regarding the use of existing data. (See Cook and Campbell, 1979, for a checklist of common threats to validity, including measurement and data analysis problems.)
Modern methodology also moves beyond simple dichotomies of data quality such as ''good'' and "poor" or "reliable" and "unreliable" to procedures for both identifying the reliability of indicators and for aggregating them into more reliable indices by taking explicit account of measurement error. For example, Bollen's (1980) work on political democracy has both identified the character of errors in existing measures and developed a more reliable indicator that has seen extensive subsequent use. There are long traditions in psychology (measurement theory), sociology (path analysis), and econometrics (errors in variables procedures) addressing these issues. In the last decade they have been unified into a powerful set of techniques that can advance our ability to make appropriate use of flawed data (Bollen, 1989). Again, we would suggest that a few targeted studies using these techniques to analyze multiple indicators would greatly enhance our understanding of data quality and suggest methods for improving both conceptualization and measurement.
Problems of limited validity and reliability in the measurement of key variables are not unique to the social sciences. Physical and biological science data sets of the sort used for the study of global change are fraught with the same problems that have had the attention of social scientists for decades. Social scientists have developed many procedures for dealing with such data and for improving data quality. Analyses and data collection efforts in the physical and biological sciences could benefit from application of some of them. In particular, the statistical procedures developed for multiple indicators, for analysis in the presence of missing data, and for analysis in the presence of unusual but possibly valid observations could easily be transferred to the physical and biological sciences. So could some of the technologies for improving the reliability and validity of data collection.
DATA COLLECTION NEEDS
To what extent do existing raw data suffice for the needs of global change research? Data needs tend to be very researcher-specific, so one person's experience may be quite different from another's. On the whole, data relating to many critical variables concerning global change exist in some form or another, although (as already noted) the data are sometimes of very poor quality and often of unknown quality.
Data that are not currently available will undoubtedly be required for the study of the human dimensions of global environmental change. What they are will become clearer as the inventory of existing data is completed and as the studies proceed. Even at this early stage, however, it is apparent that there are significant lacunae. Environmental quality, valuation of nonmarket activities, and leisure-time budgets have been areas in which data are poor relative to the plentiful data available for the study of financial markets or labor markets. Data on the land market, however, are very sparse. Broad-based analyses of land values, which are needed for the study of the human impact of a rise in sea level, require such data. Baseline data on public and elite attitudes toward global change on a worldwide comparable basis and on the enactment and implementation of environmental policies also do not exist.
Survey data in particular tend to be sparse, partly because surveys have been expensive and highly vulnerable to the budgeteer's ax. Improved survey data in the areas of energy use, perceived environmental quality, and recreation and time-use patterns could
help improve our understanding. In addition, surveys in developing and socialist countries are particularly important for comparison with the responses of individuals in developed capitalist countries.
Undoubtedly, additional kinds of new data will have to be gathered. It is important, however, that this be done carefully, deliberately, and generally in the context of actual studies. There are important synergies between the development of theory and the collection of data, so premature data collection may yield data of little value for future research.
Since the study of the human dimensions of global environmental change aims at understanding dynamic processes, it is crucial that baseline data be available so that change can be observed. Special priority should be devoted to the collection of baseline data on key measures, including indicators of the major driving forces of global change for which data do not already exist.
Governmental and intergovernmental agencies will be collecting the bulk of social data, as they do now. They may also be developing new indices, for instance, supplements to national accounts for the costs of resource use and depletion and of environmental services and degradation. In both these roles, it will be important to ensure that the new data fit easily with existing time-series data. It is therefore important that the scholarly and governmental communities work together constructively. To more carefully assess needs for new data, we recommend that an inventory of data and survey needs in selected areas be developed and examined to determine whether expanded data collections in these areas are needed. The areas should include the following, among others:
land use and food production
consumption of energy and materials
The inventory should be developed through consultation among social and natural scientists, governmental and nongovernmental statistical agencies, and data base management and archiving specialists associated with the global change data network. It should be updated periodically to take into account new data needs uncovered by recent research.
This project would benefit from oversight provided by an inter-
disciplinary advisory group consisting of leading researchers as well as representatives of the key agency sponsors of research on the human dimensions of global change. The advisory group would be responsible for planning the project's design and implementation.
ANALYTICAL DATA AND ACCOUNTING
Studies of global change have been hampered by a number of problems with analytical data, that is, social indices or aggregates that take raw data and put it in a form useful for analysts. Examples include data on gross national product, indices of inequality of land or income distribution, average energy prices, and so forth. The major issues have been inadequate data outside the advanced countries and excessively narrow accounting systems for all countries.
It is well known that reliable social and environmental data have been sparse outside the OECD countries. This difficulty may be of little consequence for national social, economic, or environmental policies, but it is significant for analyses at the global level. Because poor countries often do have not the resources or statistical personnel to collect and process data, their social, demographic, economic, and environmental data are usually of poor quality. Often the best data on some environmental variables come from satellite studies. In addition, the poor quality of data from socialist countries has been compounded by antiquated accounting standards and politically motivated distortion of data. A major effort to improve the accounts of socialist and low-income countries would clearly benefit both the countries and global change researchers.
Existing systems of national income accounts omit many sectors that are important for global change research. Three critical omissions are nonmarket use of time (which is important, for example, where climate change affects recreation); use of depletable natural resources (such as oil and gas); and environmental spillovers and abatement activities (such as the costs of oil spills or the changes in amenity values of unpriced resources). Research has moved ahead sporadically on some of these topics over the past 20 years, although little progress has been made in the United States. Researchers in Germany, Japan, and other countries have made major contributions to augmenting the national accounts, while some attempts have been made to see how such changes might affect the accounts of a developing country such as Indonesia. This work needs to be undertaken either through developing aug-
mented accounts in the Commerce Department or by supporting research in this area. In the area of analytical data; we make two recommendations: The United States should support major improvements in data, accounting practices, and national accounts for developing countries and current or former socialist countries.
The United States should develop augmented national accounts to include nonmarket productive activity, depletion of natural resources, and the value of environmental spillovers and resources. These factors would contribute to more accurate indicators of the gross national product.
Research on the human dimensions of global change poses an analytical dilemma. On one hand, understanding of causes, responses, and consequences depends on analyses of data collected at a global level of analysis. On the other hand, any attempt to assemble data for all the nations of the world would be prohibitively expensive and impractical since data of high quality probably exist only for a small set of nations. What is needed is an analytical strategy that would serve to economize collection, dissemination, and analysis while not forfeiting our ability to derive implications for the globe. One approach is to develop a systematic sampling design for national-level data collections.
An appropriate sampling of nations would include the large and pivotal environmental countries such as the United States, the Soviet Union, China, India, and Brazil; two or three countries from Europe; and the remaining countries selected from a pool of nations for which data resources have been found to be adequate for analysis. This approach entails an evaluation of resources (and costs) prior to selecting countries for inclusion in the sample. For the countries chosen, there would be national-level data collections bearing on the relevant institutional structures, assessment of environmental variables beyond what can be inferred from satellite data, trends in public opinion, and others as illustrated by the lists presented earlier in this chapter. It is true, however, that the nation is not the only relevant sampling unit. Other territorial sampling units may be appropriate, including regions and cultures that cross the borders of recognized polities. Any sampling frame could be stratified on the basis of classifications drawn from satellite measurement as well as types of social and economic institutions, land tenure patterns, population densities, level of technological development, and so on. (A variety of alter-
native sampling designs are presented in Kish, 1965, and in the earlier work by Hansen et al., 1953; these sources remain the classic treatments of this subject.)
Cross-national data collections and analyses are at an intermediate level, between global-level data and intensive local studies. Operating at this level of analysis has the advantage of providing greater richness of information than what is likely to be available at a global level while affording better coverage of key variables than what is usually available from a scattering of episodic case studies. It is also a level at which such critical processes as policy formulation and implementation, discussed in Chapter 4, occur. Therefore, we suggest that research on the human dimensions of global change utilize sampling strategies that would take advantage of the richness of available information without forfeiting coverage.
INTENSIVE LOCAL DATA COLLECTION
As we have noted in Chapters 3 and 4, data analysis at the global and national levels can yield only part of the necessary understanding. Important human causes of and responses to global change occur at the local and individual levels and require study at those levels. To relate activities at those levels to their global effects requires understanding of how and why human activities and responses differ from one locale to another; this understanding, in turn, requires detailed knowledge of the locales. These considerations provide a rationale for the intensive study of a relatively small number of locales, selected so that they can be compared with each other to yield the desired knowledge.
The identification and study of locales for intensive study requires rather detailed information and assessment about the relationships between environmental and social variables. Typically, the larger the spatial and temporal scale of analysis, the more difficult to achieve and manage the requisite level of detail; yet the smaller the scale, the less significant the case in absolute terms and the less transferrable the results to other situations. Studies in locales, despite their limited generality, can offer valuable insight into comparative human-environmental relations if they are carefully selected. They can generate more complex hypotheses by testing the initial assumption that similar environmental and socioeconomic circumstances generate similar patterns of environmental change and of response to change. To identify such similarities, detailed analysis will require common
protocols that allow for comparability among circumstances. Studies of locales can also help build capabilities to monitor at the local level environmental changes and their consequences that are also of interest at the global level. And they can be used to develop measures that can then be applied in other areas of the world.
Analysis of local causes or responses to environmental change is complicated by the interconnections between locales. The importance of physical impacts received from other regions (e.g., acid rain) or from the global system (e.g., global climatic change) and of driving forces originating elsewhere, such as commodity demand, means that analysis cannot be successfully conducted only within the bounds of the region in question. Locales that import resources from elsewhere are resistant to local environmental changes but vulnerable to changes occurring in the supply areas. Analysis must also address these broader spatial linkages, the difficulties of doing so notwithstanding.
These considerations complicate but do not reduce the importance of the intensive study of locales. Such studies are worthwhile not only for the substantive reasons noted, but also because they have the potential to bring together interdisciplinary research groups in continuing interaction—a much-needed attribute of effective global change research, as we noted in Chapters 3 and 4.
A similar idea is put forth in the recent report of the National Research Council's Committee on Global Change, Research Strategies for the U.S. Global Change Research Program. They suggest that regional resource sites be established "where relevant studies from the natural and human sciences can be conducted in concert, and their data sets pooled" (National Research Council, 1990c:122). This approach would insure against a lack of coordination characteristic of projects in which, for example, "ecological data are collected at one site, tropospheric chemistry data at another, and human activity data at a third" (p. 122).
But how should locales be selected for intensive study? Because of the importance of comparative research on locales, a haphazard process would be a terrible waste of the potential for knowledge.
One strategy for selecting locales for detailed study is to choose critical zones. By this term, we mean locales in which human-induced degradation in the physical or biological conditions of the area are contributing (or are expected soon to contribute) to a loss of things the occupants value (for alternative definitions, see Price, 1989). Such zones are interesting as microcosms of some possible undesirable futures of global environmental change. The study of
such areas would allow determination of the factors contributing to environmental changes of particular interest and to the various responses to such changes at the local level. It would also generate more rapid results than studies of slower-changing or larger areas, because with more rapid change, relationships are more apparent. It would be appropriate to begin with relatively modest pilot projects that alert investigators to possible problems that can be dealt with prior to developing the infrastructures needed for the research on patterns of social, economic, and ecological change in the chosen locales (see also National Research Council, 1990b).
It would also be possible to conduct intensive studies of sustainable zones, that is, areas characterized by presumably sustainable use of natural resources. Such studies would best be comparative between similar or nearby regions with different patterns of resource use, one presumably more sustainable than the other. The logic is similar to that for critical zones.
Other criteria for selecting locales might also be defensible. We recommend that intensive study of locales should be included in the global change research program. The locales and the programs to study them be selected according to criteria including:
groups of locales should be chosen together on grounds of similarity and difference that would illuminate important global change questions (e.g., similar natural environment but different political systems);
social scientists and natural scientists should work together on the same ongoing research;
projects should preferably contribute new methods or measures that could be applied elsewhere; and
projects should employ measures for the locales taken from global data archives and return quantitative data on the locales to the same archives.
It should be noted, however, that the suggested model for doing such research is collaborative field work in the anthropological tradition: researchers would remain at the site only until completion of the project; they would not establish a center at the field site. We are not recommending that permanent institutions be developed as, for example, in the mold of the biological field station.