|
POSITION STATEMENT
Spatial externalities arise in a number of interesting contexts, but land use is the most obvious one - and arguably the most pressing one from a policy perspective. Environmental externalities tend to be spatial in nature, but these are only one category of a broader class of spatial spillovers that cause land use/land cover of one parcel to affect the values in differing land uses of neighboring parcels. Since the value of land in alternative uses governs the process of land conversion, spatial spillovers between land parcels can lead to a complex, path dependent process of land use change. This path dependence is only exacerbated by the irreversibility of many types of land conversions.
Local governments in many parts of the U.S. are keenly interested in the underlying causes of land use change, as they struggle with the public finance implications of increasing low density, fragmented exurban sprawl. The amount and spatial pattern of land use is also critical to environmental scientists, because of the implications for bio-complexity, non-point source air and water pollution, and global climate change. Digitized maps of land use pattern serve as the basis for landscape ecology, hydrological engineering, and atmospheric deposition models. The focus of much of my research is the development of methods to predict the spatial pattern of land use change under alternative regulatory and growth scenarios, for use by these "audiences".
Space poses a challenge for economists in a fundamental way. The land use change problem is frequently one in which we are interested in understanding how the location specific decisions of individual economic agents aggregate up to a spatial pattern that is changing over time. Economists are familiar with the concept of aggregating up from an individual to a collective outcome, in the sense that a large number of individual agents making utility or profit maximizing decisions constitutes a market. In understanding how these individual actions aggregate up into market outcomes, the defining factor for economists is the distinction between what is endogenous and what is exogenous at each scale. Our usual way of crossing from one scale to the next is aspatial, though, and understanding spatially distributed aggregate outcomes, especially when spatial externalities exist, is not straightforward.
By far the best known example of an economic spatial model that links individual decisions with a description of aggregate land use pattern is the bid-rent model (or monocentric city model) of urban economics. What was once a robust but parsimonious description of urban spatial structure comes up short when trying to capture the observed complexities in pattern that are currently of chief concern. More complicated versions of the basic model have been developed, but the model's success as a predictor of observed urban land use pattern at a spatially disaggregate scale is limited. This is primarily due to the model's inability to explain the formation and location of urban centers, its reduction of space to a one-dimensional measure of distance from city centers, and its inability to accommodate the types of spatial heterogeneity that are central to the science questions.
The spatial distribution and pattern of land use change is so important to ecologists and other environmental scientists that they have often assumed the role of land use change modelers to meet the pressing need for forecasting tools. Most represent the landscape in terms of pixels or cells, and many model land use change with Markovian or semi-Markovian constructs where the transition probabilities can be made functions of earlier states, sojourn times, and interaction effects among neighboring cells. While complex with respect to the ecology of transition, these models are simplistic with regard to human-induced change - which in many parts of the world is the dominant form. Some landscape ecology models simply extrapolate from past land use change; others include "socio-economic drivers," but the validity of the underlying variable choice and exogeneity assumptions is questionable. More sophisticated state transition models, developed by geographers and spatial statisticians, draw on parallels with thermodynamic systems or use cellular automata models to simulate the complex processes by which global patterns are generated from rules about cell interactions. But again the underlying economic processes are ignored.
Although they are yet to be put to the empirical test, exceptions include the economic agent-based models of interaction that adopt modeling techniques from interacting particle systems. Fujita, Krugman, and Venables (1999) and Anas, Arnott, and Small (1998) review this literature. The common theme of this work is that the pattern of developed land uses is driven by interactions among spatially distributed agents, although the source and specification of these interactions varies. They may arise from transportation costs and pecuniary externalities or directly through agents' preferences over the spatial distribution of other agents through social interactions, knowledge spillovers, or negative externalities. Because these interactions both influence future location decisions and are a function of past location decisions, the spatial distribution of agents across the landscape is endogenously determined. Add to this the durability of most urban development and the result is the evolution of a complex urban spatial structure that is characterized by multiple equilibria and path dependence.
These agent-based models are of importance because they provide a means of deriving aggregate spatial patterns from the microeconomic behavior of atomistic, but interdependent, agents. However, these models have tended to be abstract, ignoring the many heterogeneous features of the landscape that are likely to influence location decisions (e.g. roads, zoning, environmental amenities/disamenities). This simplification allows for tractable analytical models that demonstrate the potential role of interactions, but has inhibited empirical research. Empirical applications of these models require not only spatially explicit data but the resolution of serious identification problems. The effects of endogenous interactions are difficult to separate from spatially correlated exogenous landscape features, which may evoke land use patterns that are observationally equivalent. Similar challenges have been outlined in a separate literature on empirical models of social interactions, most notably by Manski (1995, 2000).
Because it is difficult to measure such interactions, separating these effects from unobserved exogenous heterogeneity is possible only for limited cases. One such case arose in the context of work by Irwin (1998) and Irwin and Bockstael (2000). Drawing upon the agent-based interaction models discussed earlier, we developed a model in which exogenous features create attracting effects (e.g. central city, road, public services) among developed land parcels and interactions among land use agents create net repelling effects. Empirical evidence of a negative interaction effect among land parcels in a residential use was econometrically identified. This model offers an explanation of the fragmented residential development pattern found in urban fringe areas. One of the more interesting questions that arises from this type of work is the conceptual connection between Manski's identification problem, set in a spatial context, and the difficulty of distinguishing between spatial autocorrelation and spatially dependent variables in spatial econometrics.
This discussion has assumed a modeling strategy based entirely on micro-level agents making interrelated economic decisions on individual land parcels. The strategy provokes two types of challenges. First, are parcel-level data on economic variables sufficiently prevalent as to make estimation of such models possible? Such data are available in Maryland where the Maryland Office of Planning has, since 1996, digitized the centroids of all land parcels in the tax assessment data base. Similar digitization programs are only beginning elsewhere, and only in some of the more affluent states. Even for Maryland, however, sufficient historical data to estimate cross-section time-series models are difficult to obtain, and in states where these activities are only beginning it may be many years before enough "history" has been compiled. Together with remote sensing researchers, we are investigating to what extent Landsat data can supplement or be substituted for parcel level data, given that Landsat data are broadly available for the US for the last 15 to 20 years but represent the landscape as a mosaic of cells rather than parcels. The second challenge is more fundamental: are agent-based, parcel-level models the right modeling strategy? Alternatives exist that, although based on theories of individual behavior require data on at a more aggregate level. These include, but may not be limited to, large-scale urban simulation models based on transportation zones (e.g. Anas and Kim, 1996) and locational-equilibrium sorting models (Epple and Sieg, 1999).
|
|
REFERENCES
Anas, A., R. Arnott, and K. Small (1998), 'Urban Spatial Structure,' Journal of Economic Literature, 36, 1426-1464.
Anas, A. and I. Kim (1996), 'General Equilibrium Models of Polycentric Urban Land Use with Endogenous Congestion and Job Agglomeration,' Journal of Urban Economics, 40(2), 232-256.
Epple, D. and J. Sieg (1999), 'Estimating Equilibrium Models of Local Jurisdictions', Journal of Political Economy, 107(4), 645-681.
Fujita, M., P. Krugman, and A. Venables (1999), The Spatial Economy: Cities, Regions, and International Trade, Cambridge, Mass: M.I.T. Press.
Irwin, E. (1998), Do Spatial Externalities Matter? The Role of Endogenous Interactions in the Evolution of Land Use Pattern. Ph. D dissertation. Department of Agricultural and Resource Economics, University of Maryland.
Irwin, E. and N. Bockstael (2000) 'Interacting agents, spatial externalities and the evolution of land use change,' In review.
Manski, C. (1995), Identification Problems in the Social Sciences. Cambridge, MA: Harvard Univ. Press.
Manski, C. (2000), 'Economic Analysis of Social Interactions,' Journal of Economic Perspectives, 14, 15-136.
|