Scaling and Demographic Issues in Global Change Research: The Great Plains, 1880-1990

Myron P. Gutmann

University of Texas at Austin

Austin, Texas

Gutmann discussed ways of looking at questions regarding the correct scale at which to study the relationship between population and environment with the goal of demonstrating some appropriate strategies for understanding scaling in studies of population and environment. This research is part of an interdisciplinary project directed at examining the relationships between population, land use, and the environment on the Great Plains. The basic unit of analysis in this work is the county. In addition, Gutmann and colleagues are also gathering data at other scales. For a small number of recent demographic variables, they have data for census tracts, a much finer resolution than the county because the average census tract has a population of about 6,000 persons. Census tract data are readily available in digital format only for 1980 and 1990. They are also interviewing about 150 farm families in the Great Plains, and doing very detailed historical research about a number of communities.

The relationship between population and environment is being analyzed in a way that characterizes both population and environment as independent and dependent variables. The fundamental analytic premise is that neither population nor environment is always the driving force. Environment does not always shape population, nor do changes in the population always shape changes in the environment. In this sense, population and environment demonstrate a kind of causal recursiveness, where change in the first produces change in the second in one time period, but those changes later produce new changes in the former.

Major environmental episodes, such as the drought and the Dust Bowl of the 1930s, demonstrate the recursiveness the researchers believe existed. A stylized pattern of recursive causality that may have existed is shown below in Figure 1.11, which links the major environmental episodes that might be found in the standard history of the Great Plains.

The Great Plains have been occupied and exploited by humans for more than 10,000 years. Europeans entered the Plains in the sixteenth century, but they had little or no impact until the end of the seventeenth century. In the twenty-five years following 1850, Anglo and Indian hunters devastated the bison and the U. S. government confined all Indians in the Great Plains to reservations.

Environment does not always shape population, nor do changes in the population always shape changes in the environment.

Figure 1.11

A stylized representation of how Great Plains history is recursive

In the second half of the nineteenth century, American stock raisers found a region that was lush, a consequence of high rainfall and the elimination of bison and Indian horses, competitors for grass. The late nineteenth century brought new systems of agriculture to the Plains. Good rainfall and a boom in grain prices caused by World War I led farmers to prosper from roughly 1912 to 1920. Grain boom profits led farmers to plow up native grasses and leave the land vulnerable to wind and drought.

The grain boom ended in the mid-1920s. It was followed by a long drought that lasted from 1932 to 1941 on the Southern Plains, and from 1928 to 1936 on the Northern Plains. The drought brought dust storms and economic collapse, leaving families with failed crops, dead cattle, and eroded soil to desert their farms. The drought also led to massive Federal government efforts to change the environment, society and economy.

In the conventional story-telling of the development of the Great Plains, the plow-up of the grasslands for wheat, combined with the drought of the 1930s, provoked the disastrous dust storms and social dislocation of that time period. While all might not agree that those were the only causes, or that the greatest areas of wheat farming suffered the worst drought and dust storms, there was a causal relationship. Plowed land is a much greater source of blowing dust than uncultivated grassland. Humans responded to the problems of the 1930s. Some migrated out of the region, although the 1930s were not the period of the greatest out-migration from the agricultural areas of the Great Plains (the 1950s were). Those who remained changed their farming techniques between the 1930s and 1970s.

Scales of Analysis

The data used in this project came at a variety of scales, and these have a strong influence on the research. Demographic data constitute the primary source for much of the analysis. While there are many kinds of demographic data available, population census data constitute the most important source in this work. The U. S. population census data are published at higher levels of aggregation than the individual, such as the city, county, metropolitan area, or state. The smallest levels of aggregation for which there are published data for recent censuses are the block, block group, or census tract. (Census blocks are small areas bounded on all sides by visible features such as roads, streams and railroad tracks, and by invisible boundaries such as city, township, and county limits, property lines, and short, imaginary extensions of streets. A block group is a cluster of blocks having the same first digit of their three-digit identifying numbers within a census tract. Census tracts are small, relatively permanent statistical subdivisions of a county and are delineated for all metropolitan areas and other densely populated counties.) There are few data at levels smaller than the county before 1940.

Individual-level data are available for many censuses, but these data are in the form of public use samples, stripped of most characteristics that might permit a person to be identified. In 1990, for example, the smallest identified geographic unit must have 100,000 persons, many more than a single rural county. It is impossible to link these population data effectively to environmental or land use data.

The variety of scales at which demographic data are available leads to the question: what is the appropriate scale at which to conduct the analysis?

The variety of scales at which demographic data are available leads to the question: what is the appropriate scale at which to conduct the analysis? Figure 1.12 presents data about population density in the Great Plains counties located in Texas. Figure 1.13 presents population density in those same counties, but this time, divided into census tracts instead. Dividing the region into census tracts shows that the rural population is much less dense than it appears at the scale of the county, while the population living in towns is more dense. If one were analyzing the role of human population density on levels of air or water pollution, looking at census tract densities would be much more productive than looking at county densities.

Figure 1.12 (above)

Population Density on the Texas Plains, 1990

(counties as unit of analysis)

---

Figure 1.13 (above)

Population Density on the Texas Plains, 1990

(census tracts as unit of analysis)

Census Land Use Data constitute the second major source of data for this project. The U. S. has undertaken a census of agriculture at regular intervals since 1850. These censuses of agriculture were performed every ten years from 1850 until 1920, and generally every five years since then. Data at the level of the individual farm are scant for censuses taken after 1880. The Census Bureau has always published these data at the level of the county and state, and more recently at the level of the zip code. The publications are extremely rich. In recent years, the published tables (and their digital analogs) include thousands of cells of data for each county. Users need to face the trade-off between rich detail in the number of variables reported and the lack of detail in terms of identifying land use and farm activity in an area smaller than a county.

Other data at different scales are also used in the study. These include soils data (varying polygons), and weather data (weather stations are points). The researchers also get greater detail from the two kinds of small-scale research they are doing. These parts of the research project involve interviews with farm families, plus detailed historical research. In the future, remotely sensed data, either satellite imagery or aerial photographs may be added to the study as well.

Advantages and Disadvantages of Different Scales

The major advantage of working at a small scale is that environmental conditions and environmental change are geographically precise, and should be measured that way. That task is difficult because most demographic data are not available at this resolution, especially for the past. The same holds for agricultural data. One can, of course, disaggregate, taking smaller areas than counties and assigning the characteristics of counties to them. Such a strategy, while environmentally detailed, risks a false sense of precision, in both description and statistical analysis.

The advantage of working at a large scale, such as counties, is that virtually all data can be converted to that level of analysis. There is no false sense of precision. The disadvantage of working at the county scale is that there is a considerable loss of accuracy if one assumes that the characteristics of a unit the size of a county apply to all its components. This error might lead one to conclude from this analysis that what happens at a unit the size of a county also happens to each of the individual persons, families, and towns within its administrative borders.

Large scales of analysis, such as the county units used in this study, are appropriate to the extent that they are homogeneous. If they are not homogeneous, then it is necessary to consider whether by aggregating smaller units one loses precision or introduces distortions. Looked at the other way, small units are attractive only so long as they capture the diversity of the environment or the diversity of the human population, and if the data really exist to support them.

Migration Analysis: An Example

Many descriptions of the demographic consequences of climate and environment in the Great Plains discuss the role played by migration as a mechanism for adjustment. The often-told story of the drought of the 1930s emphasizes migration. This history stresses the fact that residents of the driest parts of the Great Plains left the region to go elsewhere when their efforts at farming failed. In a recent paper, Gutmann and colleagues have attempted to test the hypothesis that weather played an important role in determining which counties had the greatest net migration away from the region in the decades from 1930 to 1990. Because the demographic and economic variables are available only at the county level, they perform their analysis at that level.

The major advantage of working at a small scale is that environmental conditions and environmental change are geographically precise, and should be measured that way.

The general findings of the research are as follows: the most important determinants of migration are the extent to which the county has an agricultural economy, and the extent to which there are meaningful alternatives to agricultural employment within the county. Thus, counties with a large proportion employed in agriculture had relatively large out-migration, while counties with a large proportion having a college education had relatively large in-migration. The environmental variables are not always significant, but two groups of those variables are nonetheless worth summarizing. First, counties with relatively greater drought in the 1930s did have more net out-migration than counties with less drought. Second, counties with relatively high elevations had relatively high in-migration during the 1960s and 1970s, during the first phase of the development of a mountain recreation economy in the western Great Plains.

This analysis of the determinants of county net migration works because net migration can only be measured at a large level of aggregation. There are no individual "net migrants," only the record of the counties that migrants have entered and left. There are still perils in this strategy. The environmental variables may be aggregated to a level that may not be homogeneous. Even the demographic units can be a concern. If the urban status of the county is important, care must be taken in cases where the urban area is only part of the county.

Even if the analysis of the impact of economy, society, and environment on migration in the Great Plains at the level of the county is successful, one must also ask if it can be scaled up to larger units, and whether it can be scaled down to the level of individuals, families, and communities. On the scaling down side, the analysis is not yet complete. Nevertheless, the analysis of net migration at the county level appears to reflect the same conclusions as the preliminary research into the experiences of individuals and families.

Scaling up from the county to larger units of analysis poses a number of problems. Could the analysis of county net migration be scaled up to much larger aggregations of county-sized units than the Great Plains region, such as the United States, the North American Continent, or the globe? In other words, can county-level net migration be predicted for areas larger than the Great Plains? The answer appears to be yes, at least for the United States, and perhaps for all of North America.

Can county-level net migration be predicted for areas larger than the Great Plains? The answer appears to be yes, at least for the United States, and perhaps for all of North America.

The success of regional and national analyses of county net migration in response to a stable or changing environment does not necessarily mean that these results can be used to estimate conditions elsewhere in the world or in other time periods. Human responses to drought varied by time period and may have varied by land use. Drought produced out-migration in the 1930s, but it did not produce out-migration in other time periods. Rather, the out-migration from the counties of the Great Plains in the 1940s, 1950s, and 1960s was largely independent of differences in weather conditions, and more likely the result of forces pulling the people of the Great Plains to the growing industrial cities of the Midwest and California. If we attempt to project these findings forward for the United States, or more problematically for the whole world, it can only work if one is imaginative enough to foresee economic and social trends.

The migration results of this study indicate that it is difficult, but perhaps not impossible, to jump from one scale to another in work about population and environment. The key is the ability to predict the complex behavior of humans in the future.

The migration results of this study indicate that it is difficult, but perhaps not impossible, to jump from one scale to another in work about population and environment . The key is the ability to predict the complex behavior of humans in the future. Gutmann and colleagues have learned that there are scientifically measurable relationships between environment and human migratory behavior. That is important on its own. They have also learned that humans do not respond to the same environmental change or condition in the same way under all conditions or at all times, even in a single semi-arid region such as the Great Plains. This means that we cannot yet predict how this or any other population will react to similar changes in the future. What we can do now is describe some of the limits on how humans will react, and continue to do research on past and present human populations in order to improve our understanding in the future.