Spatial Regression Models for the Social Sciences. Jun Zhu
lands (Cowen & Jensen, 1998), qualitative (Lewis, 1996) and quantitative (Cardille, Ventura, & Turner, 2001) environmental corridors, growth management factors (Land Information and Computer Graphics Facility, 2000, 2002), and a land developability index (Chi, 2010b).
Population change as a spatial process is also implicitly suggested and considered in theories of sociology (especially urban sociology, rural sociology, and sociological human ecology) and demography (especially spatial demography, rural demography, and applied demography). The spatial process of population change has already been formally incorporated into demographic models and empirical studies (Entwisle, 2007). In addition, researchers often use maps to illustrate spatial patterns and population change with multiple time frames. For a review of the large body of literature on this research, see Entwisle (2007), Fossett (2005), Logan (2012), Reibel (2007), Voss (2007), and the position paper collection Future Directions in Spatial Demography released jointly by University of California, Santa Barbara; Pennsylvania State University; and the National Institutes of Health Advanced Spatial Analysis Training Program.2
2 www.ncgia.ucsb.edu/projects/spatial-demography/docs/All-position-papers.pdf
Rural demographers study population’s spatial dimension, conducting research on population distribution and migration. They argue that migrants prefer somewhat rural or truly “sub”-urban locations within commutable distances of large cities (Brown, Fuguitt, Heaton, & Waseem, 1997; Fuguitt & Zuiches, 1975). Applied demographers often use the idea of neighbors for small-area population estimation and forecasting. For instance, they may adjust populations projected at the municipal level so that they agree with their parent county’s projections; this neighborhood context, however, is different from the spatial effects that we address in this book. Recent population forecasting research has used a modified spatio-temporal regression approach (e.g., Chi & Voss, 2011) and a geographically weighted regression approach (e.g., Chi & Wang, 2017) to formally incorporate spatial effects into the modeling.
Sociological human ecology, or the study of how human beings are affected by the environment in space and time (McKenzie, 1924), also informs sociologists of the spatial distribution of population (Berry & Kasarda, 1977; Frisbie & Kasarda, 1988). Hawley (1950) considers one of human ecology’s main topics to be spatial differentiation within urban systems, whereas Robinson (1950) views human ecology studies as using spatial information rather than individual units. And Logan and Molotch (1987) espouse that the analytical basis for urban systems in human ecology is spatial relations.
Segregation studies, one of the largest bodies of urban sociological research, likewise suggest population distribution has spatial effects (Charles, 2003; Fossett, 2005). There are several theoretical approaches explaining segregation: the spatial assimilation approach claims differences in socioeconomic statuses and associated lifestyles cause it (Galster, 1988), the place stratification approach states discrimination causes it (Alba & Logan, 1993; Massey & Denton, 1993), and the suburbanization explanation argues the suburbanization process leads to segregation (Chi & Parisi, 2011).
The spatial dimension of population dynamics is also studied by neo-Marxists, who mainly focus on population redistribution. They explain that capitalism’s pursuit of profit leads to how cities are structured, how land is used, and how population changes (Hall, 1988; Jaret, 1983). They also argue that the basis of urban development in the United States is capital accumulation (Gordon, 1978; Hill, 1977; Mollenkopf, 1975, 1981). As Hill (1977) explains, because capital accumulation occurs in an environment that is spatially structured, the geographical form and spatial patterning that it takes is (at least provisionally) urbanism.
Although there is a large body of literature on spatial demography, demography has not yet fully integrated spatial statistics and analysis methods (Hugo, Champion, & Lattes, 2003). Sociologists and demographers can benefit from the advances in spatial techniques and the availability of spatial data as these tools become more accessible and well known to develop new theories and ask new demographic and sociological questions (Chi & Zhu, 2008). Sociological and demographic perspectives of population change can be strengthened further as more researchers incorporate the spatial effects of proximity, continuity, and contagion from geography and regional science’s spatially explicit theories into new theories. For a summary of the literature, see Entwisle (2007), Fossett (2005), Reibel (2007), and Voss (2007).
It should be noted that this list of the explicit and implicit spatial theories of population change is far from complete. The implicit spatial theories come mostly from sociological and demographic perspectives, which the authors are familiar with; theories from other social science disciplines regarding population change as a spatial process are not discussed here.
1.3.2 The State of Wisconsin in the United States: The Study Area
Our case study focuses on population changes during the period of 1970 to 2010 at the MCD level in the state of Wisconsin in relation to the spatial and temporal variations of a multitude of explanatory variables. Wisconsin, located in the upper Midwest of the United States, borders Minnesota, Iowa, Illinois, and Michigan, as well as the Mississippi River, Lake Michigan, and Lake Superior (Figure 1.1). Wisconsin contains 34.8 million acres (excluding the Mississippi River and Great Lakes areas in the state), and inland lakes constitute 3 percent of the state’s total surface area (Wisconsin Legislative Reference Bureau, 2005).
Figure 1.1 ⬢ The geography of Wisconsin, United States
Source: Chi and Marcouiller (2011).
Note: The urbanized areas are based on the 2000 Census Urban Areas.
Although Wisconsin became the thirtieth state in 1848, the written history of the state dates to more than 300 years ago, when around 20,000 Native Americans and French explorers were first arriving (Wisconsin Legislative Reference Bureau, 2005). The years between 1824 and 1861 saw the first wave of immigration to Wisconsin, due to lead mining in the southwestern corner of the state (Figure 1.2). As the years passed, immigration continued, and in the early years of its statehood, Wisconsin became a major area for wheat farming. As the wheat industry moved to the northern and western parts of the state after the Civil War, the lumber industry became important for the northern half of the state (1870s to 1890s), as did the dairy industry for the state as a whole (1880s and 1890s). The heavy machinery and brewing industries grew and developed dramatically in Wisconsin until the end of the nineteenth century. By the middle of the twentieth century, the large-scale European immigration had ended. During this time, the lumber industry faded and the brewing industry disappeared temporarily, while the heavy machinery manufacturing, paper, and dairy industries thrived. Tourism has since emerged as a major industry for the entire state, and other industries have been concentrated in its eastern and southeastern areas. In the 1980s, the state grew less than 4 percent (the smallest increase in the state’s history); the 1990s saw 9.6 percent population growth and the 2000s saw 6.0 percent growth.
Figure 1.2 ⬢ Wisconsin population growth since 1840
Source: Decennial censuses, U.S. Census Bureau.
1.3.3 Minor Civil Division (MCD): The Spatial Unit of Analysis
MCDs are designated on the basis of legal entities rather than on population sizes and are recognized in twenty-eight U.S. states. Each MCD (a city, a village, or a town) is a functioning governmental unit with elected officials who provide services and raise revenues. The particular data set to be analyzed in our case study consists of 1,837