Publication Detail
Dynamic Microsimulation of Heterogeneous Spatial Markets
|
UCD-ITS-RR-01-21 Research Report Urban Land Use and Transportation Center Download PDF |
Suggested Citation:
Abraham, John E. and John D. Hunt (2001) Dynamic Microsimulation of Heterogeneous Spatial Markets. Institute of Transportation Studies, University of California, Davis, Research Report UCD-ITS-RR-01-21
A system for dynamic microsimulation of urban spatial markets is described, where individual economic units (primarily business establishments and households) are simulated. Long term relationships that lead to regular travel are simulated (e.g. workplace/employer, normal shopping locations) together with casual relationships and transactions causing a typical day's travel. The diversity of trades and interactions is represented by forming "parcels" of goods, services, labour or other (non-market) interactions. Each parcel is described by a number of attributes, coded mathematically as integer or real numbers. The attributes represent characteristics of the parcel, including a category, monetary value, size and quality. Certain attributes can correspond to attributes measured in observed data; other numerical attributes can be included as representative of unmeasured diversity. Economic units regularly make "offers", which are proposals to exchange (e.g. to purchase or sell) a particular parcel at a particular location. Examples of such offers include product descriptions in catalogues, help wanted ads, television commercials and social invitations between friends and family. An exchange occurs when such an offer is accepted by another economic unit.
In the proposed framework, economic units use a rational expectation and information search process to develop a set of interactions desired, expected or needed based on their lifestyle. This produces a set of "expected parcels", which are specific transaction expectations that the economic unit has in mind. Each of these leads to an explicit search for a suitable transaction. For each such expected parcel the economic unit searches for a set of similar offers already proposed at various locations. It then makes a choice to accept one of those similar parcels at a location, or to instead publish an offer of its own based on the "expected parcel". There is a continuous supply of offers and the quantity, type and spatial arrangement of outstanding offers at any time influences the choice of whether to accept an existing offer or to publish a new offer.
This process is simulated in four stages for each economic unit: 1) the creation of expected parcels by selecting attributes from distributions, with the parameters of the distributions dependent on - 2 - lifestyle needs, rational expectations and information searches; 2) a search process that establishes a set of offers to consider for each such expected parcel; 3) for each expected parcel, a behavioural choice model between the offers considered and the additional alternative of publishing a corresponding offer of ones own, and 4) the updating of lifestyle and expectations based the economic unit's ability to match its wants and needs through accepting offers or having offers accepted.
Each of the above components can be operationalised using simple models. Lifestyle and expectations (1) can be described as an explicit list of expected parcels, randomly generated based on the economic unit's unique characteristics. The search process (2) can be an equal-probability random selection between all offers in the same category as the expected parcel. For long term relationships the existing situation is also included as one of the alternatives. The choice model (3) can be a nested logit model based on random utility theory. The representative utility of each considered offer can be a linear function of the differences between its attributes and the expected parcel's attributes, and the perceived travel conditions to the offer's location. For long term relationships the existing situation's representative utility can have an additional constant representing the tendency to avoid change. The utility of making an offer can be a constant dependent on the category of transaction. The updating of expectations (4) can be an adjustment of price and quantity expectations for each expected parcel based on the number and utility of the past transactions involving the expected parcel.
The framework moves through time as a discrete event simulation. Components (1) and (4) can be considered in longer (e.g. one-year) time steps, while components (2) and (3) are scheduled for each expected parcel at a specific time within the longer time period.
The framework represents the purposes for all trips in an urban area, and hence all travel, and the supply and demand for all goods, services and social interactions. Thus the framework captures the raison d'être of cities -- to provide a variety of accessible spatial interaction possibilities between complementary actors.
The framework is envisioned as the basis for practical simulation tools for transportation and urban planning policy analysis and government decision making. In such a role, the simulation would calculate travel conditions endogenously based on a transportation network simulation of the route and travel mode choices for each simulated exchange. The foundation of the framework on economic interactions is in contrast with "activity based" transportation planning models, which focus on the activities performed by people over the course of a day without a strong representation of the interactions that occur during those activities. The framework is thus an appropriate tool for economic analysis of transportation policy based on the benefits that occur to firms and households when transportation systems support access to diverse transaction opportunities.
The proposed model can represent the differing effect of policy on economic units or trades categorised by any of the attributes used to represent them. Local microlevel interactions (e.g. - 3 - pedestrian movements and symbiotic relationships between nearby firms) can be represented, as there is no need for zones or zonal averages.
The framework was designed with the intention that it would support a longer-term urban economic model that has a realistic representation of the formation of spatial clusters of complementary activities in urban areas. Such clusters are often highly specialised and spatially very small, with substantial pedestrian movement between the economic units in the cluster. The framework has been designed to allow unlimited spatial detail (without the Order n^2 performance degradation that occurs in zone-pair representations of interactions) and arbitrary diversity in the representation of the unique characteristics of the actors and their desired interactions. Thus the framework can represent the advantages of such micro-level clustering based on the unique characteristics and needs of economic units.
This paper describes the framework and compares the framework with other frameworks to identify its advantages and disadvantages. Issues in applying the framework to different types of firms and households and different categories of transactions are discussed. The paper concludes with recommendations for further research to apply the framework as a calibrated and operational model of a specific urban spatial economy.
In the proposed framework, economic units use a rational expectation and information search process to develop a set of interactions desired, expected or needed based on their lifestyle. This produces a set of "expected parcels", which are specific transaction expectations that the economic unit has in mind. Each of these leads to an explicit search for a suitable transaction. For each such expected parcel the economic unit searches for a set of similar offers already proposed at various locations. It then makes a choice to accept one of those similar parcels at a location, or to instead publish an offer of its own based on the "expected parcel". There is a continuous supply of offers and the quantity, type and spatial arrangement of outstanding offers at any time influences the choice of whether to accept an existing offer or to publish a new offer.
This process is simulated in four stages for each economic unit: 1) the creation of expected parcels by selecting attributes from distributions, with the parameters of the distributions dependent on - 2 - lifestyle needs, rational expectations and information searches; 2) a search process that establishes a set of offers to consider for each such expected parcel; 3) for each expected parcel, a behavioural choice model between the offers considered and the additional alternative of publishing a corresponding offer of ones own, and 4) the updating of lifestyle and expectations based the economic unit's ability to match its wants and needs through accepting offers or having offers accepted.
Each of the above components can be operationalised using simple models. Lifestyle and expectations (1) can be described as an explicit list of expected parcels, randomly generated based on the economic unit's unique characteristics. The search process (2) can be an equal-probability random selection between all offers in the same category as the expected parcel. For long term relationships the existing situation is also included as one of the alternatives. The choice model (3) can be a nested logit model based on random utility theory. The representative utility of each considered offer can be a linear function of the differences between its attributes and the expected parcel's attributes, and the perceived travel conditions to the offer's location. For long term relationships the existing situation's representative utility can have an additional constant representing the tendency to avoid change. The utility of making an offer can be a constant dependent on the category of transaction. The updating of expectations (4) can be an adjustment of price and quantity expectations for each expected parcel based on the number and utility of the past transactions involving the expected parcel.
The framework moves through time as a discrete event simulation. Components (1) and (4) can be considered in longer (e.g. one-year) time steps, while components (2) and (3) are scheduled for each expected parcel at a specific time within the longer time period.
The framework represents the purposes for all trips in an urban area, and hence all travel, and the supply and demand for all goods, services and social interactions. Thus the framework captures the raison d'être of cities -- to provide a variety of accessible spatial interaction possibilities between complementary actors.
The framework is envisioned as the basis for practical simulation tools for transportation and urban planning policy analysis and government decision making. In such a role, the simulation would calculate travel conditions endogenously based on a transportation network simulation of the route and travel mode choices for each simulated exchange. The foundation of the framework on economic interactions is in contrast with "activity based" transportation planning models, which focus on the activities performed by people over the course of a day without a strong representation of the interactions that occur during those activities. The framework is thus an appropriate tool for economic analysis of transportation policy based on the benefits that occur to firms and households when transportation systems support access to diverse transaction opportunities.
The proposed model can represent the differing effect of policy on economic units or trades categorised by any of the attributes used to represent them. Local microlevel interactions (e.g. - 3 - pedestrian movements and symbiotic relationships between nearby firms) can be represented, as there is no need for zones or zonal averages.
The framework was designed with the intention that it would support a longer-term urban economic model that has a realistic representation of the formation of spatial clusters of complementary activities in urban areas. Such clusters are often highly specialised and spatially very small, with substantial pedestrian movement between the economic units in the cluster. The framework has been designed to allow unlimited spatial detail (without the Order n^2 performance degradation that occurs in zone-pair representations of interactions) and arbitrary diversity in the representation of the unique characteristics of the actors and their desired interactions. Thus the framework can represent the advantages of such micro-level clustering based on the unique characteristics and needs of economic units.
This paper describes the framework and compares the framework with other frameworks to identify its advantages and disadvantages. Issues in applying the framework to different types of firms and households and different categories of transactions are discussed. The paper concludes with recommendations for further research to apply the framework as a calibrated and operational model of a specific urban spatial economy.