A presentation at the 14th International Conference on Computers in Urban Planning and Urban Management (CUPUM) which was held at MIT in Cambridge, Massachusetts USA on July 7-10, 2015
Urban models are important tools for planners in their capacity to offer in-sight into future urban growth. However, the majority of urban models overlook the role of developers’ behaviour in capturing the growth of ur-ban residential spaces. This paper redresses this gap by embedding the spa-tial consequences of privately-driven urban residential development on se-lection of potential land and their impact on land prices within an agent-based model. Jakarta (JMA), Indonesia forms the case study context. Re-sults from the model highlight the emergence of new urban areas on the JMA’s fringe that are strongly tied to land value. The model offers poten-tial to offer new insights into the relationship between land cover and land prices and the role of developers’ decisions in shaping the expansion of residential areas.
Modelling the spatial decisions of private developers: A case study of Jakarta Metropolitan Area, Indonesia
1. A case study of Jakarta Metropolitan Area, Indonesia
Modelling the spatial decisions of
private developers:
Agung Wahyudi, Yan Liu, Jonathan Corcoran
School of Geography, Planning and Environmental Management (GPEM)
The University of Queensland
AUSTRALIA
2. Agung Wahyudi S42817150
Why model the residential developer behaviour?
• Urban models aim to provide:
• a better understanding of the mechanisms underpinning
urban system and;
• Simulate the impact of a spatial policies on urban form
• There has been various urban models used to simulate urban
growth and urban form.
• Models accounting for residential developers behaviour are
largely missing from these models.
• In particular, the representation of profit-driven
developers on searching for the most profitable areas,
and the impact of this behaviour on land cover and land
value.
3. Agung Wahyudi S42817150
The spatial context of this study
• Covers 6800 km2,
• Population: 28 million people
(~10% of Indonesian). MEGAcity
• Jakarta (the capital of Indonesia)
as the economic core.
4. Agung Wahyudi S42817150
Factors impacting developer decisions
• Developers consider
• multi-dimension factors,
• various actors, as well as
• details on the factor intrinsic to the site.
• In the model, we selected few factors to localize and to focus the
impact of these factors on urban form.
• In the model, factors intrinsic to the site :
• Proximity of the targeted area from the city centre (CBD),
• Proximity to the (toll) roads, and
• Land value.
• Factors intrinsic to the developer :
• Access to capital affects the decision (incl loan),
• Number of developers in the area.
5. Agung Wahyudi S42817150
Conceptual framework – potential profit
hypothetical curve
• Blue line: Raw land
price before
installment of
infrastructure varies
according to the
distance from city
centre
• Red line: Road
construction improves
site accessibility
• Green line: Added
value increase equally
as much as ∼30-
100% of cost
• Purple line: Selling
price 30%
• Monocentric
assumption still
applicable in JMASource: adapted from Winarso 2000
Landvalue
6. Agung Wahyudi S42817150
Method #1:
The modules in agent-based urban growth model
• The model has three modules:
1. environment,
2. agent, and
3. interaction
• The environment module represents
the properties owned by the land; land
cover type, land price, distance to
CBD, distance to road
• The agent represents properties
exhibited by the developers; the
capital, number of developers
Environment Agent
8. Agung Wahyudi S42817150
Method #3:
The model interface
• The interface offers views on land
cover, land value, distance to
CBD, or distance to road.
• Developers’ parameters: number
of developers, the searching
radius a, initial capital, and loan.
• Monitoring panels (on the right).
9. Agung Wahyudi S42817150
Results #1:
Urban land cover growth
• The movement of the developers: toward the outer
most of study area.
• The developers capture the potential economic
opportunity on the outer most area.
t=0 t=0+n
10. Agung Wahyudi S42817150
Results #2:
Land value changes
• Land value in newly developed areas
increase (∼300%) as developers construct
the infrastructure in these sites.
• Increasing land values prevent other
developers moving into these areas.
t=0 t=0+n
11. Agung Wahyudi S42817150
Tentative conclusions and future work
• The model offers insight into how new residential areas and land
values across JMA is shaped by the developer behaviour.
• Deviation of the simulated urban pattern from the observed
coverage suggests that the profit oriented (profit maximization)
seems only one dimension of the decision to develop.
• Marketing
• Land banking
• Limitations:
1. Simple initial set of variables,
2. No temporal dimension.
• Future work:
1. A typology of developers,
2. Incorporate notion of land banking, and
3. Investigate inclusion of a temporal dimension.
In the presentation. I must show why this study is important. Three points: The spatial scale of the study, nowhere in the world has ever studied before. The impact of suct urban development to environment sustainability, and affecting million of lives in the area, and the magnitude even greater because JMA is a economic centre of Indonesia, and not only environment sustanaibility but also on the urban development sustainability where the proportion of urban-non urban is guaranteed to still support non-urban activities, green life.
Stressed that the model is the first step to simulate the proposed (modified) conceptual framework . It is a preliminary work.
Script”
In today’s presentation, I will explain the work of my self, my supervisors Yan Liu, and Jonathan Corcoran. We are from the school of geography, planning, and environment management, university of Queensland in Brisbane, Australia.
Gap:
Embedding developers behaviour within an agent based urban growth model
Megacity developing world context
== script ===
I start this presentation by briefly explain why the study is important in the context of urban planning. We all now with various utilities of urban models, the two most prominent uses of urban models are –first– helping urban planners to better understand the mechanism of urban system and secondly to simulate the impact of the implementation of a spatial policy. [explain further with example if necessary]
Since the introduction of computer as an aid in urban modelling, there has been a healthy grow of urban model in the literature, investigating the expansion of urban areas as well as the impact on bio-diversity, environment, and physical changes on land.
The model that account for residential developers however has been lacking, in particular the one that specifically investigate the supply side of land market that is the behaviour of private residential developers in their effort to gain maximum profit from non-urban to urban land conversion.
And I need to add that this
The largest in Indonesia, in terms of population and areas
Issue on collaboration municipalities 9 municipalities
Australia population is 23 million in 2015
Cover the area equalling to Brisbane (6000 km2)
Entire population of Australia, packed in a city of Brisbane Area in size
US context
=script==
And I need to add that this study was done using a data from Jakarta Metropolitan Area, Indonesia. It is a megacity in term of its population, equalling
===Script===
We will look first on factors that drive the developers decision in the selection of potential site for a new residential developments
We can pull the factors from visiting the earliest land theory proposed by Alonso (bid-rent-theory). In his theory – I simplify – , the concentric shape of a city with different use according to willingness to pay the land rent is caused by the different level of importance on the accessibility of the site. In other words, who wants to get closer to the city centre should pay more the cost than the one who doesn’t necessarily need to be as close as possible to the city centre. The theory assumes that the city has single market activities in the centre (monocentric function).
Now, if we look on the current situation, the empirical studies held in UK, North America (US Canada) (Goldberg empirically surveyed the developers in Canada) and in Developing countries like Indonesia (Winarso in Indonesia) are in general in line with the proposed theory by Alonso, in that accessibility (access to the city centre) is still considered as the primary factors in determining the location. Of course, other factors are considered by the developers. These factors such as the general spatial planning policy adapted in the country (or even down into the municipality), the macro-economic situation of the country to predict the expected cost recovery and benefit will meet with the projected timeframe, (financier , government, household composition), to details (micro factors) intrinsic to the targeted sites for example the slope, the size , and the availability of the site to be converted into urban area.
In the model, we tried to select as few as possible the number of factors to localize the impact on the urban form. As this is a the first-stage of urban modelling in the larger project, the aim is to study the impact of the factors and we were less concerned with including others factors which might be relevant for the urban growth model. We will leave that for the next phase of urban model.
The factors that we selected are the proximity of the targeted area from the city centre, the proximity to the toll roads, and land values. The selection was based on previous study conducted in JMA by Haryo Winarso using questionnaire distributed to the developers to rank the importance of the factors according to their best knowledge on the land market in JMA. The findings by Winarso was also supported by our previous study that investigated the shape of urban form according to its proximity to the CBD and toll road. The study suggests that the urban area JMA forms a monocentric pattern (radial as distance to CBD) and the expansion of new urban area largely occur in the close proximity of the toll road (6km).
The land value has been mentioned in the study by Winarso, ??? ???. With construction cost generally equal everywhere, the only component in cost-benefit analysis that matter from the developer point of view. It contributes to 15-25% of overall land-house selling cost (Gillen Fisher) and spatially different from one site to another is land value, thus ensuring the lowest land value is of primary concern for the developer.
Another factors intrinsic to the developer are: access to the capital, be it from its own money or from loan. We were not going into detail on that site, but at maximum 70% of loan can be obtained from developers’ own money.
And the last is the number of the developers.
We depicted the conceptual cost analysis with factors adopted from classical land theory of Alonso , adaptation of Winarso , with modification on the addition of expected profit area.
We are not trying to represent all the factors but to only focus on the distance to city centre and represent others factors using price per hectare. For instance, we aknowledge variation as much as 6-fold of land in the first 2 km from the toll road than land locates in far away.
The curves simply gives the impression that classical land theory to some extent applicable in this study (monocentric) and heavily weighted to city centre.
The Idea is that raw land price, imagining all factors equal (ceteris paribus)
Road construction improves site accessibility but increase cost to the developers
=Script=
The relationship among the factors considered in the model could be represented in the potential profit hypothetical curve, which was our conceptual framework for the study.
These curves closely resemble the bid-land theory by Alonso, in that the x-axis is the distance to the city centre and the y-axis in the land price. The adaptation from Alonso bid-rent theory in these curves was on the potential profit obtained by the developers after stages of development process. The development process was considered by the developer on their cost-analysis.
The cost analysis is a composite of factors affecting the land price.
On the lowest curve (the blue line) there is a raw land price. That is the price of land with everything else equal (ceteres paribus) and only the distance to CBD is matter.
Then we have here the site improvement, by the developers, such the construction of basic infrastructure such as connecting road from the nearest main (toll) road to the area, the cost of the improvement is someway not equal because at certain point of the distance (how we could prove this) the cost is increasing.
The land value increase as the infrastructure is ready. The selling price however follow the raw land price?? Because the selling price follow the demand and willingness to pay from the customers which is low, in the far distance from the city.
Overall, these curves are purely hypothetical and could only be suggest the relation among factors in a simplified market. Variation does occur, and the exact quantification and relation among factors will need further empirical studies which are beyond the intention of this study.
How you get the data
Land cover: Landsat image classification (1994). 300 meters in resolution.
Land value: Combination of National Land Agency + reports from newspaper (standardized [0 1])
Distance to CBD: from GIS buffer analysis
Distance to toll road: from GIS buffer analysis
The interaction represent the decision of the developers impacting the properties of land ; will be explained on the next slide
=script==
We are moving the component of the model. The model was constructed under the agent-based modelling platform and has three inter-related modules ; the environment, the agent, and the interaction module. Each module represents different parameters in the urban system.
The environment represent factors that inherent in the site such as the proximity to the CBD and toll road, land value, and land cover.
The agent represents the behaviour of the developers. Specifically, the amount of capital secured by the developers to initiate a development project, and the number of agent in the area (this is actually a global parameter in the model)
The model was constructed with NetLogo because it is the agent-based modelling system that allows users with less experience in programming to grasp the language in a short time and test the model and change the parameters’ values straight away.
Before we present The interaction module, the workflow of the model is represented in this diagram.
The developers start by finding the area with the lowest land values.
The prime and foremost important factor for the developers is the land value. If the land price is within or lower of their (--%) of their budget, then stay on the land, otherwise, leave the land and find another area.
Then, cost analysis starts. Calculate the road development cost (connecting road from toll road to the area), site clearing cost, and compare them with the expected selling price. If the margin is positive (benefit) the continue to acquired and build the land, otherwise (loose) leave the area.
If YES develop; then change land type from non-urban to urban including the area surrounding it (until reaching the optimum size of a residential complex – 3 cell = 1 hectare)
Then update the land price.
The next developers who visit the area recognized that land price has increased due to the development by previous developers, and thus less likely to acquire the land (land value reason), and find another place to start development.
Eventually all the developers gone into the maximum land values where profit could not be reached and the model stop.
In the conference paper, one of the reviewer asked about whether the uncertainty was taken into account in the model. We did not report systematically in the paper, but we do implement uncertainty in the model by adding the normalized random error on the perceived information on land values by the developers.
=Script===
This is the interface of agent based urban model of JMA in NetLogo
On the left panel there are parameters in the model that can be changed by the users.
These parameters include the number of agent (According to Winarso in JMA there were no more than 50 developers (large) that works in JMA. In the model we only allow maximum 10 developers)
Searching radius by the developers
Initial capital owned by the developers
Loan with maximum of loan is 70% of the initial developers
On the right hand side, there are panels monitoring the size of land cover classes that change during the simulation.
How would you answer, the striking difference in the location of the new urban development from the model?
Validation of model has two approaches (i) model which its parameters closely represent the system being modelled, (ii) matching the model output with the variable in the system (Rand 2003).
“Conceptually true”: This model is “conceptually true” as the developers agent react as we designed/expected (move to cheapest land first, and make a decision to buy or to leave according to their budget). And this is our main aim. We are not targeting “spatially true pattern”.
==script==
OK, now we see the result of running the model until the developers could not find any land affordable from their capital point of view and/or has been assessed by other developers
The movement of developers in finding the potential area for development was motivated by profit maximisation in that the lowest land value is a priority.
It is thus the result on land cover changes shows that the potential area for development according to the developers with the profit maximisation in mind was on the outer most of the area – as you see here as black nodes – I
The agent spend their capital to develop the area until it could not establish new development (non-urban to urban). Then it dies.
==script==
The result on land value shows an increasing land value after the development – as expected -.
The land value increase – according to empirical studies – 15-30% (Check number)
The increasing land values – according to profit curves shown previously – were caused by the construction of infrastructure, increased added value, and selling price set by the developers.
The increasing land values prevent other developers to move into these areas, because cost to convert these areas is already too high beyond its capital, and the agent look for another sites.
Other concern: Site improvement possible more than just the construction of connecting road, but also the marketing image. By constructing near the CBD, selling point increase. From the expense spend by the developer, construction of road might be less critical (occupy very few in the proportion of the overall budget)
Strategy in the conceptual framework might well adopt in very large developers (not known as, trying to unlimit the budget seems not changing the overall pattern)
Benefit from the land conversion has not been included in the model
(acquired land but for development in the near future)
=Script==
At the moment, we could not establish the accuracy analysis on the result because (i) the model does not account for the temporal dimension (ii) the target was not to mimic the pattern of urban areas on certain period.
But the visual check comparing the simulation results and land cover map from subsequent years (>1994) reveals different in that the new urban area of the simulated maps come on the outer most of JMA whilst in the land cover map, it is located in the near CBD.
Deviation from the simulation and observed pattern might suggest that the profit oriented is not only the single dimension considered by the developers.
In the current version of the model, it is indeed a simple model in that the variable that closely inherent in the developer behaviour for instance profit gained and re-investment were not taken into account.
The model set no account on the temporal dimension because ??
A future work on the subsequent version of the model will look on the typology of the developers, to distinguish the developers based on the capital secured for the development, and to incorporate the decision of the developers whether to build the land immediately or for later use (land banking), and also to investigate further on the inclusion of a temporal dimension.
In the spirit of using social media to get/socialize/introduce your research profile, I provide my twitter account. Feel free to discuss