|Home Agenda Registration FAQ Abstracts|
Speaker Abstracts/Position Statements
Andrew A. Beveridge
The use of spatial materials in social science research is growing rapidly. This is largely based upon the more readily available GIS tools along with data with spatial identifiers, especially the United States Census, which provides very high quality electronic maps and data that is linked accurately to those maps. However, the evolution of GIS tools has been somewhat stymied by the cumbersome nature and unreliability of the software in use, as well as difficulties of conflating maps from various sources to the same base. Unlike the evolution of statistical packages, GIS software is neither as robust nor as well developed; nor are there tools or applications that are particularly user friendly. Developing spatial data can be very time consuming. Beginning in 1999 to fill this gap, I proposed the development of a tool to make it fast and easy to access census data visually and numerically. This eventually led to Social Explorer, which is based upon Census data and maps from the National Historical Geographic Information System project. Social Explorer provides demographic information in an easily understood format: interactive thematic data maps, reports that present data for an area in a larger context, and simple tools to visualize complex data patterns across geographies (nation, state, county, census tract) and across time. Extensive data will be available for U.S. counties from 1790 to the present, and for census tracts for each decade from 1910 to the present. Social Explorer includes data from 1940 to 2000 and receives several thousand map requests per day, and supports thousands of individual users per month. Following a brief demo of the application, I will discuss approaches to disseminating spatial materials and analysis to those who know it would be of use to them, but who may not even know what a GIS is or does.
Humanities scholars are beginning to recognize that spatial technologies, especially GIS, offer new ways to understand society and culture, past and present. Much of this effort to date has centered on the development of large spatially enabled databases, but increasingly scholars are using GIS to create new knowledge. This presentation will examine the potential of GIS to reinvigorate and enhance heritage and cultural scholarship and practice. It also will explore whether a spatial approach adds a new tool or a fundamentally new approach to the humanities.
Political science has long recognized the role of social context in politics. Several of the cornerstone theories in subfields from American to Comparative politics to International Relations highlight the role of social interaction and social context. Certainly the idea that geography plays a role in the manifestation of political phenomena has been recognized by many. Although these concepts were fundamental to earlier work, empirical analysis has lagged its theoretical counterpart. While empirical conceptualizations of spatial components exist in some of these works, the analyses did not consider statistical issues arising from endogeneity, inefficiency, and computation, among others. Rigorous examination of these theories may demonstrate that they need to be refined or re-conceptualized, and may have an enormous impact on how we understand and conceptualize power and politics. The juxtaposition of well-developed theories and rapidly developing exploratory "tools" and statistical routines strongly signal untapped potential and the brink of explosive growth. Finally, a review of the literature in political science helps to identify important gaps that are left to be filled in the spatial econometric and computational research.
There is broad agreement in the scientific community that population processes are central to land use and land cover change, both as determinants and consequences. However, because of poorly framed questions, data problems, and division of labor across disciplines, demographers and population scientists have not centrally participated in research on population and land cover change or in debates about climate change or biodiversity loss. This is changing. Of particular note is a flourishing of sustained case studies, often involving a team of researchers from many disciplines. My presentation describes one such case study, based on data from Nang Rong, Thailand, a rural area in the northeast of the country that was a sparsely populated frontier at the middle of the 20th century, experienced dramatic population growth over subsequent decades, but is now an area of net out-migration. Nang Rong serves as a laboratory to explore questions about the expansion of human settlement in frontier areas and its consequences for land use; scale dependency in relationships between population size and distribution, biophysical variables, and patterns of land use; and most recently, household dynamics and land use change at the micro level. I will describe the diverse types of data on which the project draws, including longitudinal panel survey, a time-series of remote images, and a variety of GIS coverages, review the challenges involved with combining these data, and describe how we have used the data to address the substantive questions that have motivated our work, giving most attention to our current work, which uses simulation approaches to model relationships between population and land use/land cover.
While many methods of spatial data analysis have been described, and there is evidence that several disciplines are experiencing a spatial turn, there have been few attempts to identify the fundamental concepts that form the atomic elements of a spatial perspective. I begin by reviewing the various ways in which individuals organize knowledge about space, and then identify a series of fundamental concepts. In some cases these concepts have been the subject of substantive research in spatial cognition, but in other cases the concepts underlie specific analytic techniques. I review the efforts that have been made to analyze the nature of spatial data, and to ask whether fundamental generalizations exist that can underpin both the design of systems for handling spatial data, and the ways in which we characterize human and physical landscapes. The general public now has increasing access to many of the tools of spatial data manipulation and analysis, so education in fundamental spatial concepts is increasingly essential to their informed and responsible use.
The increasing availability of fine-scale land cover/use data is transforming land use change modeling. Thirty years ago, urban economists described urban spatial structure and tested the basic hypothesis of the urban economic model using observations on population density at two points, the city and suburb, and strong assumptions regarding functional form to estimate population density gradients. Now the availability of high-resolution satellite data and highly accurate planimetric data allow researchers to quantify urban land use pattern in a much more comprehensive and spatially explicit manner. However, while a fair amount of research has sought to describe fine-scale changes in urban land use pattern using these data, accompanying development and testing of alternative spatial theories of land use change and in particular, of sprawl, has lagged behind. Urban economists still rely heavily on the basic urban economic model (the so-called monocentric model) to explain urban spatial structure and its evolution, including changes in urban land use patterns. This despite accumulated evidence that the basic model explains a relatively small amount of the variation in urban land use patterns across cities and regions. The increasing availability of fine-scale land use data and growing computational capacities provide an unparalleled opportunity for land use change researchers to not only describe land use patterns in ways that were not previously possible, but also to test conventional theories of urban land use change and develop alternative theories that seek to explain the evolution of urban land use patterns across various spatial and temporal scales. Existing methods, including landscape ecology metrics and spatial econometric models, as well as emerging methods such as agent-based models, provide researchers with a range of spatially explicit techniques that can harness the increasing availability of such data and computing resources to develop and refine spatial theories of urban land use change.
In the past 10 years, historical GIS has been a robust and rapidly developing field, spawning national projects of enterprise scale, digital publications and websites focused on particular research domains, books and articles making use of spatial analysis, and special journal issues about the field. HGIS practitioners, implicitly or explicitly, have devoted themselves to the admirable goal of centering spatiality in their analysis, and identifying how to abstract spatial change over time. I want to frame the problem the other way around. What if, instead of thinking about geography (as it is transformed over time), we inquire instead about history (which has a spatial element)? In some cases, the results would be similar. The Religious Atlas of China, for instance, would still include a dataset of spatially referenced temples. However, if it is to reflect the interests of most scholars of religion, it needs to find a way to model religious experiences that embody more complex and more ambiguous geography: the extent and spread of particular sects and teachings, the locations of events associated with the expansion and contraction of religious practices, the networks of masters and disciples. For most cultural and social history, and for almost all kinds of history prior to the nineteenth century, GIS as constituted only goes so far. The solution is perhaps to think of GIS as only one a part of a toolkit for the digital and quantitative humanities. In addition to GIS, for many kinds of humanities projects (let us say monsoon trade routes in the Medieval Indian Ocean world), digital gazetteers need to be used to capture complex changes in place names, names in multiple languages, and relationships among places in a hierarchy, particular for those which cannot be precisely georeferenced. Moreover, in many cases, historical GIS applications are best envisioned as only one component in Cultural Atlas systems that also include images, historical maps, documents, and search tools. Finally, temporal modeling and timeline visualization for historical narrative is an under-researched area. Events and the relationships among them need to be identified, associated with often complex geographies, formally modeled, and the results made visual, not only using maps, but network models, timelines, and animations besides. Finally, even for scholars interested in GIS as such, the capacity to handle uncertainty and ambiguity needs to be substantially improved. This problem is often raised, but has not yet been addressed satisfactorily. From my perspective as a medieval world historian, it is clear that most interesting research problems in the humanities have a spatial component, but only a range of them can be asked using the current methods of the historical GIS community. In my comments at this workshop, I will consider where we can go next to expand the range and sophistication of our questions and answers with reference to all the digital tools at our disposal. In this vision, rather than practicing historical GIS as such, scholars may move toward (to use a term coined at the Computer Applications in Archaeology conference last March), a spatially informed digital humanities.
In my treatment of the various subtexts of violence, the role of place is ever-present. I believe that communities provide the key social context that serves to promote or inhibit the behaviors of local members (and visitors) of an area and offers the most promising approach to understanding where violence occurs. Ecological models of crime (using spatial regression) have clearly demonstrated that crime patterns can not be explained by the socio-economic characteristics of place alone. Instead, there appear to be particular social processes or mechanisms (i.e., neighborhood effects) that manifest themselves in such a way that crimes in one location influence the levels and patterns of crimes in nearby or connected places. These findings serve as a constant reminder that space matters, thereby refuting the notion that neighborhoods, regardless of areal unit of analysis, are analytically independent and that ecological models of crime need to consider the ways in which the observable outcomes in one neighborhood are dependent upon the actions and activities occurring in other areas. Though demonstrating that space matters is relatively easy, identifying and modeling the social processes responsible for the how and why space matters and influences is not. To date, the majority of spatial studies of crime have employed an inductive approach wherein the first step is to demonstrate the presence of spatial dependence and then post-hoc explanations are fashioned to explain this dependence. Drawing upon the modeling of network autocorrelation within the social influence literature, I believe that a deductive approach, wherein specific social processes are posited, measured and modeled a priori, can help us unlock the black box of neighborhood effects. I will use the example of gang violence to demonstrate how the conceptualization of communities as nodes in a larger spatial network, explicitly tied to one another through the social structure of gang networks, provides important insights into the spatial distribution of crime. Through careful construction of alternative spatial weights matrices (W), I explore whether the spatial distribution of gang violence is best explained by dependence among spatially adjacent areas or by dependence among socially linked areas. Using gang violence as an illustrative example, results show that the model explicitly considering the socio-spatial dimensions of gang rivalries outperforms the model in which actions are influenced solely by what occurs in neighboring areas. This finding demonstrates the utility of explicitly measuring the underlying mechanisms often assumed to account for neighborhood effects.
We study through structural economic models incorporating geography the temporal and spatial dynamics of entrepreneurship and intermediation in Thailand in the high growth period 1986 to 1996. Using biannual census village data as well as the geo-locations of more than 3,000 of those villages, we are able to study in each of four provinces the spatial patterns of the data as well as the interplay of that data with the predictions across time and space of a spatially specified binary occupation choice model and a spatially specified financial deepening model, both representative of the development literature. We compute and numerically simulate the models, from the given 1986 initial conditions to the end-of-sample 1996 year, focusing on the differential patterns of entrepreneurial activity, wealth and credit access over time and space. Specifically, we use space to estimate key cost parameters that dictate occupation choice, and we conclude that costs of occupational shifting increase with geographic distance from agglomeration markets. We analyze spatially prediction errors for goodness of fit and anomalies. The resulting simulations over a ten-year timeline are superiorin matching actual data than if costs had no spatial component. Our spatial analysis of financial deepening identifies an apparent policy distortionrelated to the diverse behavior of private and public financial sector providers that is perversely related to geographic agglomeration access. More generally, the work is indicative of further research possibilities that combine numerical methods, computation, and estimation with GIS platforms.