Territories are defined by boundaries at scales ranging from countries and indeed alliances of countries) to neighbourhoods via regions and cities. These may be government or administrative boundaries, some formal, some less so; or they may be socially defined as in gang territories in cities. Much data relates to territories; some policies are defined by them – catchment areas of schools or health facilities for example. It is at this point that we start to see difficulties. Local government boundaries usually will not coincide with the functional city region; and in the case of catchment boundaries, some will be crossed unless there is some administrative ‘forcing’. So as well as defining territories, we need to consider flows both within but especially between them. Formally, we can call territories ‘zones’, and flows are then between origin zones and destination zones. If the zones are countries, then the flows are trade and migration; if zones within a city region, then the flows may be journeys to work, to retail or other facilities.
It is then convenient to make a distinction between the social and political roles of territories and how we make best use of them in analysis and research. In the former case, much administration is rooted in the government areas and they have significant roles in social identity – ‘I am a Yorkshire-man or -woman’, ‘I am Italian’, and so on; in the latter case, these territories don’t typically suit our purposes though we are often prisoners of administrative data and associated classifications.
So how do we make the best of it for our analysis? A part of the answer is always to make use of the flow data. In the case of functional city regions, the whole region can be divided into smaller zones and origin-destination flows (O-D matrices technically) can be analysed, first to identify ‘centres’ and then perhaps a hierarchy of centres. (Google ‘Nystuen and Dacey, 1961’ for one way to do this systematically.) It is then possible, for example, to define a city region as a ‘travel to work area’ – a TTWA – as in the UK Census. Note, however, that there will always be an element of arbitrariness in this: what is the cut-off – the percentage of flows from an origin zone into a centre – that determines whether that origin is in a particular TTWA or not?
In analysis terms, I would argue that the use of flow data is always critical. Very few territories – zones at any scale – are self-contained. And the flows across territorial boundaries, as well as the richer sets of O-D flows, are often very interesting. An obvious example is imports and exports across a national boundary from which the ‘balance of payments’ can be calculated – saying something about the health of an economy. In this case, the data exists (for many countries) but in the case of cities, it doesn’t and yet the balance of payments for a city (however defined) is a critical measure of economic health. (There is a big research challenge there.)
It is helpful to point to some contrasts in both administration and analysis when flows are not taken into account, and then to consider what can be done about this. There are many instances when catchments are defined as areas inside fixed boundaries – even when they are not defined by government. Companies, for example, might have CMAs – customer market areas; primary schools might draw a catchment boundary on a map giving priority to ‘nearness’ but trying to ensure that they get the correct number of pupils. In some traditional urban and regional analysis – in the still influential Christaller central place theory for example – market areas are defined around centres – in Christaller’s case nested in a hierarchy. This makes intuitive sense, but has no analytical precision because the market areas are not self-contained. As it happens, there is a solution!
Think of a map of facilities – shopping centres, hospitals, schools or whatever – and for each, add to the map a ‘star’ of the origins of users, with each link being given a width to represent that number of users. For each facility, that star is the catchment population. And it all adds up properly: the sum of all the catchment populations equals the population of the region. This, of course, represents the situation as it actually is and is fine for retail analysis for example. It is also fine for the analysis of the location of health facilities. It may be less good for primary schools that are seeking to define an admissions policy.
A particular application of the ‘catchment population’ concept is in the calculation of performance indicators. If cost of delivery per capita is an important indicator, then this can be calculated as the cost of running the facility divided by the catchment population. It is clearly vital that there is a good measure of catchment population. In this case, the ‘star’ is better than the ‘territory’. But the concept can be applied the other way round. Focus on the population of a small zone within a city and then build a reverse star: link to the facilities serving that zone, each link weighted by what is delivered. What you then have is a measure of effective delivery and by dividing by the zonal population, you have a per capita measure. (An alternative, and related, measure is ‘accessibility’.) This may sound unimportant, but consider, say, supermarkets and dentists. On a catchment population basis, any one of these facilities may be performing well. On a delivery basis to a population, analysis will turn up areas that are ‘supermarket deserts’ (usually where poorer people live – those who would like access to the cheaper food) or have poor access to dental treatment – even though the facilities themselves are perfectly efficient.
So what do we learn from this: that we have to work with territories, because they are administratively important and may provide the most data; but we should always, where at all possible, make use of all the associated flows, many of which cross territorial boundaries, and then calculate useable concepts like catchment populations and delivery indicators ‘properly’.