There are starting points that we can take from ‘systems thinking’ and theory development (‘evolvere theoria et intellectum’). Add ‘methods’ (including data – ‘Nullius in verba’) to this and this becomes the STM first step.
- S: define the system of interest, dealing with the various dimensions of scale etc
- T: what kinds of theory, or understanding, can be brought to bear?
- M: what kinds of methods are available to operationalise the theory, to build a model?
This is essentially analytical: how does the system of interest work? How has it evolved? What is its future? This approach will almost certainly force an interdisciplinary perspective and within that, force some choices. For example, statistics or mathematics? Econometrics or mathematical economics? We should also flag a connection to Brian Arthur’s (see ‘combinatorial evolution’ entry) ideas on the evolution of technology – applied to research. He would argue that our system of interest in practice can be broken down into a hierarchy of subsystems, and that innovation is likely to come from lower levels in the hierarchy. This was, in his case, technological innovation but it seems to me that this is applicable to research as well.
Then if applicable, we have also seen that a second step is to ask questions about policy and planning in relation to the system of interest – the PDA step:
- P: what is the policy? (That is, what are the objectives for the future?) Should we develop a plan – another ‘P’?
- D: can we design – that is ‘invent’ – possible plans?
A: we then have to test alternative plans by, say, running a model and to analyse – evaluate – them. Ideally, the analysis would off a range of indicators, perhaps using Sen’s capability framework or offering a full cost-benefit analysis (CBA). A policy or a plan is, informal terms, the specification of exogenous variables that can then be fed into a model-based analysis.
These six steps form an important starting point that usually demands much thought and time. Note the links: the STM is essentially the means of analysis in the PDA. It may be thought that the research territory is in some sense pure analysis but most urban systems of interest have real-world challenges associated with them, and these are worth thinking about. Some ideas of research problems should emerge from this initial thinking. Some problems will arise from the challenges of model building, some from real on-the-ground problems. Examples follow.
- Demographic models are usually built for an aggregate scale. Could they be developed for small zones – say for each of the 626 wards in London?
- While there may be pretty good data on birth and death rates, migration proves much more difficult. First there are definitional problems: when is a move a migration – long distance? – and when is it residential relocation?
- If we want to build an input-output model for a city, then unlike the case at the national level, there will be no data on imports and exports – so there is a research challenge to estimate these.
- There is then an economic analogue of the demography question: what would an input-output model for a small zone – say a London ward – look like? This could be used to provide a topology of zone (neighbourhood) types.
- In the UK at the present time there is, in aggregate at the national level, a housing shortage. An STM description might focus on cities, or even small zones within cities. How does the housing shortage manifest itself: differential prices? What can be done about it? This last is a policy and planning question: alternative policies and plans could be explored – the PD part of PDA – and then evaluated – the A part.
- What is the likely future of retail – relative sizes of centres, the impact of internet shopping etc?
- How can ‘parental choice’ in relation to schools be made to work (without large numbers feeling very dissatisfied with second, third or fourth choices)?
- Can we, should we, aim to do anything about road congestion?
- Does responding to climate change at the urban scale involve shorter trips and higher densities? If so, how can this be brought about – the D-question? If not, why not?!
- Can we speculate about the future of work in an informed way – taking account of the possibilities of ‘hollowing out’ through automation?
- And so on……!
Research questions can be posed and this framework should help. The examples indicated are real and ambitious, and it is right that we should aim to be ambitious. However, given the resources that any of us have at our disposal, the research plan also has to be feasible. There are different ways of achieving feasibility – probably two poles of a spectrum: either, narrow down the task to a small part of the bigger question; or stay with the bigger question and try to break into it – test ideas on a ‘proof-of-concept basis? The first of these is the more conservative, and can be valuable; and is probably the most popular with undergraduates doing dissertations or postgraduate students – and indeed their supervisors. It is lower risk, but potentially less interesting!
We can then add a further set of basic principles – offering topics for thought and discussion once the STM-PDA analysis is done at least in a preliminary way.
- Try to be comprehensive, at least to capture as much of the inevitable interdependence in your system of interest as is feasible.
- Review different approaches – e.g. to model building – and integrate where possible. There are some good opportunities for spin-off research outcomes in this kind of territory.
- This of applying good ideas more widely. I was well served in the use of the entropy concept in my early research days: having started in transport modelling, because I always wanted to build a comprehensive model, I could apply the concept to other subsystems and (with Martyn Senior) find a way of making an economic residential location model optimally sub-optimal!
- The ‘more widely’ also applies to other disciplines. Modelling techniques that work in a contemporary situation, for example, can be applied to historical periods – even ancient history and archaeology (entry to come!).
- There is usually much work to do on linking data from different sources and making it fit the definitions of your system of interest. Models can also be used for estimating missing data and for making samples comprehensive.
Alan Wilson, April 2015