Modelling Top-Down Policy

• Source Code for Location Model Change Access (543 KB)
• Shapefile (19 KB)

Unlike ‘traditional’ urban models, ABM explores the interaction of individuals and how this interaction leads to the development of more macro structures. The question therefore is ‘How do the agents react to outside change, for instance change introduced by governments?’ One of the central tenets of the location model developed in this chapter is accessibility defined as the ease of movement between places. As in traditional urban models, it is assumed as accessibility increases in an area, the areas’ overall attractiveness (both as an origin and as a destination) increases, and this will affect its land-use. Therefore the ability to alter the accessibility of an area can be considered as a top-down policy such as the introduction of a new rail line. By altering the accessibility of an area or multiple areas, one can explore how the agents in the system interact to such a change through the way that change alters the competitive balance of locations.

A simple experiment was therefore carried out where a monocentric area is first used as in previous simulations in this chapter. However, over time, a new centre is added and connected to the original centre by a series of high accessibility zones. One could consider this to be the development of a new town with a high speed rail link connecting the two centres. This can be accomplished through modifying the shapefile used in previous simulations, specifically adding extra zones as highlighted in Figures 1 and 2. Figure 1 highlights the initial starting conditions with agents spread throughout the most accessible areas. Initially the accessibility scores were as in previous simulations (e.g. the Effects of Space and Income). On the other hand unlike the previous simulations after 100 iterations, the accessibility scores were changed in certain zones as highlighted in Figure 2 so that the development of say a high speed rail network could be simulated, for example. The reason for changing the accessibility scores at 100 iterations was the fact that previous simulations had highlighted that by 100 iterations, the spatial distribution of agents and land-use patterns had stabilised. Thus one could explore how the system reacts to such a change after a steady state had been reached.

While the accessibility score values of each zone could have been altered using the GUI of the model, to aid in our ability to run multiple simulations, a variation of the location model was created (see LocModelChangeAccess model for more details). The model itself is the same as the one used in Varying Income. But at the start of iteration 101, the model changes the accessibility scores of zones 8, 9, 10, 11 and 12 to correspond to those shown in Figure 2. This was achieved by adding a few lines of code in the pre-Step method of the LocationModel Class as shown below.

if (this.STEP_NUM == 100){

Iterator environmentIterator = urbanAgentList.iterator();

while(environmentIterator.hasNext()){

EnvironmentAgent currEnvironmentAgent= (EnvironmentAgent)environmentIterator.next();

if(currEnvironmentAgent.getWARD_ID().equals("new")){

currEnvironmentAgent.accesWork=75.00;

}

if(currEnvironmentAgent.getWARD_ID().equals("newCentre")){

currEnvironmentAgent.accesWork=101.00;

}

}

}

The getWard_ID relates to information pertaining to the information in the EnvironmentAgent which is read in from the shapefile used to create it.

The model was run several times, and within each run, agents of different types and groups had different income and space preferences (as in Varying Income). Additionally the small areas and local neighbourhoods were set to 75m. The remainder of this section will highlight how the spatial distribution of agents changes as accessibility scores within the system are altered.

Figure 1: The resident and employer distribution at the start.

Figure 2: Change in accessibility scores over time.

As with previous simulations in this chapter, over the first 100 iterations of the model, the zones with the lowest accessibility scores all become empty by 100 iterations with the agents concentrating in the zones with the four highest accessibility scores. However, changing the accessibility of certain zones within the system has a large effect on the distribution of agents. For example, Figure 3 traces the average total number of agents of both type resident and employer in different zones between the start and end of the simulations. It should be noted that the time period between 100 and 110 iterations has been expanded to highlight how the change in accessibility alters the number of agents in different zones.

The number of employers in zones 1 and 2 (i.e. the most accessible areas) between 10 and 100 iterations remain roughly constant (Figure 3 and see the animations below). After accessibility within certain zones has increased, their number decreases over time, as they move to other accessible areas (e.g. zone 12). Similarly, residents are moving from less accessible areas (e.g. zone 4) to more accessible areas (e.g. zone 12). Figure 4 shows in more detail what has occurred in areas with small populations of residents and employers. For example, residents are leaving zone 8 and moving to other accessible areas (e.g. zones 12, 11 and 10). However for residents moving to zone 12, over time as employers enter this zone, these residents are forced out. This is a result of residents initially being able to afford the area, but over time, as more agents move in, the price rises (animations of this change can be seen below).

Figure 5 traces a typical simulation between 100 and 110 iterations. It highlights the fact that the area does not just change in one time period but evolves over time. This is the result of restricting agents’ movement (see Movement in the Location Model). As in previous simulations, it shows how employer agents of high income focus on the most accessible areas while residents who require the most space, are pushed out of the new more accessible areas. The change in accessibility also has an effect on whether or not an agent is searching (see Figure 6). To reiterate, an agent is classed as searching if in post-Step part of the model after evaluating its local neighbourhood, it can no longer afford their desired amount of space in the area (as described in sections Local Searching and Steps within the Location Model). Previous sections have highlighted the fact that not all agents ever become ‘settled’ for by changing the accessibility scores, the system enters a period of instability as highlighted in Figure 6. This is a result of agents moving to new areas and competing for space. However over time, the number of agents searching decreases and becomes lower than when there was just one area of high accessibility.

Figure 3: Change in total numbers of employers and residents in different zones over time for the whole system.

Figure 4:Change in total numbers of employers and residents in different zones over time for the Inner zones.

Figure 5: A typical distribution of agents between 100 and 110 iterations when accessibility has been changed.

Figure 6: The affect changing accessibility of an area alters the number of agents searching.

The fewer number of agents searching after a period of instability when the accessibility within the system has changed can be explained by the agents having more choice in where to locate. They are no longer concentrating in one area of high accessibility but in several. There are always a certain number of agents which are classed as searching, as agents are forced to move. This movement therefore affects the prices in small areas.

The fewer number of agents searching after a period of instability when the accessibility within the system has changed can be explained by the agents having more choice in where to locate. They are no longer concentrating in one area of high accessibility but in several. There are always a certain number of agents which are classed as searching, as agents are forced to move. This movement therefore affects the prices in small areas. Table 1 shows how the system becomes more stable over time. It traces the increases and decreases in the average number of agents in different zones over time with the negative sign representing an area losing agents from the previous time period. Time 0 shows the initial number of agents in different zones. When the model is first initialised, there are large numbers of agents who are searching (Figure 6) and this is reflected in certain zones loosing agents, specifically the least accessible ones (e.g. 10, 11 and 6) and other zones gaining agents (e.g. zones 1 and 3). These agents focus their location, searching on more accessible zones causing the prices to change and therefore unsettling other agents. For example, the number of residents in zones 1 and 2 decreases as employers enter these zones. However by 100 iterations, this movement between zones is relatively small (Table 1) and the number of agents searching has also stabilised (Figure 6). Nonetheless over time as accessibility in zones 12, 11, 10 and 9 increases, agents are once again moving away from less attractive zones (e.g. zone 4). These zones were previously unoccupied causing competition for space and therefore forcing the price to increase in these zones. However over time, the movement between the zones decreases as residents are being forced out of the most accessible zones by the employers who have moved in. As before, accessibility increases, the number of agents moving between zones decreases overtime, and fewer agents are classed as searching.

Table 1: Change in the average number of agents in different zones over time.

Figure 7 shows a typical distribution of agents at 100 and 200 iterations which have already been highlighted above as relatively stable conditions before and after accessibility has been altered (which is supported by the animations below). The distribution of agents at 200 iterations is similar to that seen by iteration 110 in Figure 5 and remains roughly constant with further iteration of the model. However as noted above, over time fewer agents are searching and occupy more zones within the system. In none of the simulations did agents occupy the most outer zone (i.e. 7, the least accessible). Agents did occupy zones 13, 6 and 5 which after zone 7, are the least accessible areas. When agents occupied these zones, they did so near the boundary with high accessible areas as highlighted in Figure 7. This can be explained by the use of small areas when the agent is locating (as described in the The Choice of Location section in the Location Model). These small areas would overlap the more accessible and least accessible zones, thus making them more attractive to agents.

The relationship between accessibility and agents’ income and space requirements is similar to that of other simulations, with commerce employers out bidding all other groups for the most accessible areas. At the same time, industrial residents can live in more accessible areas than other residents due to their demand for space being less than other residents as highlighted in Figure 7.

This simulation has highlighted how an area can change over time by the introduction of a new centre although this change did not occur all at once. The agents within the system slowly adapted themselves to the changes introduced by increasing accessibility in different areas. This evolutionary pattern would not have been possible with the static theories developed earlier. For example in Alonso’s (1964) theory, all agents would have changed positions through changes in the associated bid-rent curves at one point in time. This model demonstrates how competition for space over time results in the system evolving to a nearly stable distribution of employers and residents. Additionally this simulation has highlighted how the underlying space can be changed to make it more attractive illustrating how agents react to this change.

Figure 7: A typical distribution of agents at 100 and 200 iterations when accessibility has changed. A: whole system, B: Inner zones.

A sample animation highlighting how the spatial distribution of agents changes over time as accessibility values are changed1.

Download zip file of images from model runs.

1 Note: the areas that are black which normally represent empty areas are a result of the way the layers are drawn; OpenMap does not allow Islands within polygons to be drawn, if there is an Island it fills it in.

Reference

Alonso, W. (1964), Location and Land Use: Toward a General Theory of Land Rent, Harvard University Press, Cambridge, MA.

[TOP]