Defining Cities to Measure Sustainability

Michael Batty, Bartlett Professor of Planning at University College London

2030 Policy Endorsement of a Sustainable Future: Implications for Urban Research

For at least a century, there has been a long and increasingly inconclusive debate on how a city is to be defined. This has been made more complicated but the fact that we define populations that live in cities as being urban, in contrast to populations that live in the countryside as rural, and frequently the extent to which our world is dominated by cities or otherwise turns on how ‘urbanised’ a country or region has become. In an era that is often called ‘the urban age’ where big cities are celebrated and the small live in their shadows, it is ever more important to have clear definitions of ‘what is a city’ and ‘what is urbanisation’. These questions have in fact become central to the debate about how important cities are likely to be in the next 100 years.

Parnell, Crankshaw and Acuto (2016) in their discussion of various international policy agendas to secure a sustainable urban future for the medium term until 2030 and beyond, suggest the need for urgent action to clarify these questions. The eight Millennium Development Goals (MDG) defined some 20 years ago are being fast replaced by a broader and deeper set of seventeen Sustainable Development Goals (SDG), one of which is directly focussed on cities and defined to “make cities and human settlements inclusive, safe, resilient and sustainable”. These goals developed in the United Nations and presented to the General Assembly in 2015 are likely to dominate the forthcoming Habitat 3 conference to be held this October in Quito which will be developed around a New Urban Agenda (NUA). Cities have thus become the core vehicles around which to foster progress towards achieving these goals. They now hold the centre of our attention in the quest for a sustainable world, one where most of us will be living in cities by the end of this century. The need to redefine ‘what is a city?’ and ‘what is urbanisation?’ has never been more urgent and as Parnell and her colleagues argue, we are not well prepared for Habitat 3 and the consequences which will flow from this without a much deeper concern for comparability between different cities. We need to engage in a much more serious debate about the nature of our concern – cities – and how they will help us to implement a sustainable future.

Much of this concern has been polarised by the widely broadcast fact that in 2008 we passed a milestone defined as the point where 50 percent of the world’s population had become urbanised. From then on, ever more of us will be living in cities than will not, and it is almost certain that this proportion will steadily rise throughout this century where by 2100 over 90 percent, perhaps 95 percent of us, will be urbanised. This focus on cities although important has been somewhat distorted by a discussion dominated by a concern for the biggest cities. It has also been oriented to the notion that ‘big is better’ and that the economies of urban agglomeration increasingly outweigh the costs of living in ever bigger cities. In fact, by the century’s end, most of us will not be living in the biggest cities but in cities of many different sizes and kinds most of which will be small. The predictions we are able to make suggest that although the biggest cities will continue to grow and form megaurban regions, much bigger than anything mankind has experienced hitherto, the size distribution of cities will continue to become more even. Smaller cities will in fact dominate in a way that has not been articulated very well in the recent past (Batty, 2015). To be a big city of course one has to be a little city first and this asymmetry in urban growth suggests that in world of increasing urbanisation, it is the smallest cities that will get ever more numerous. Although we live in the age of bigness, it is entirely possible that sometime this century we will revert to our past concern for ‘small is beautiful’ to recast Schumacher’s famous mantra which he articulated in his book of the same name as long ago as 1973.

To engage in the debate about a new set of sustainable development goals, we need to begin almost afresh to grapple with the question as to how we define a city. This is Parnell, Crankshaw and Acuto’s argument which they suggest is ever more important but ever more elusive. If by the end of this century we are all living in cities, there is a sense in which the very term city will have outlived its usefulness and then talk of cities may well become redundant. When the world has finally made its transition from non-urban to urban which I and others see as the dominant historical prerogative of the past three centuries, the notion of the city might become superfluous. Will we still talk of cities, when all the world’s a city (Batty, 2011)? I suspect we will for whatever the ambiguities in defining what is a city, there is still the notion that we tend to cluster geographically and these clusters will continue to represent only a tiny proportion of the earth’s habitable surface area. Even if our world becomes entirely global through all pervasive information technologies – which in fact is very much the prospect – we will still retain a sense of identity with respect to place and geography where we are predominantly anchored and this will be a town or city, or a mixture of town and country within a wider sea of urbanisation.

What is certain is that the New Urban Agenda which will be built on the post-2015 Sustainable Development Goals needs clear agreement about the kinds of science that will be needed to progress these goals and this in turn needs a much clearer definition of cities than we have had hitherto. We need to cover all cities of whatever size and this suggests that we should be devising and mobilising a science of cities that covers all geographical and social scales. We need clear definitions of our objects of interest – of our concern – if we are to assemble consistent data and to devise consistent indicators to measure progress towards agreed goals of sustainability. Moreover we need consistency so that we can ensure that the world-wide effort is comprehensive and inclusive. There is a real danger of over concentration on the biggest cities which are the most charismatic, exciting and diverse, but are by no means the places where most of the world’s population will reside over the next 100 years. The present focus on assembling meetings to progress the agenda which are dominated by the mayors of the biggest cities – world cities – and by the notion that the biggest cities are like nation states in their own right is not helpful in fashioning this New Urban Agenda. The focus on ‘leaving no city behind’ which Parnell and her colleagues so clearly identify, needs to become central to the new agenda.

I will note kind of science that Parnell and colleagues call for in a moment but it is important to impress still further the difficulties that we face in defining cities as the objects of our attention. There is a view that is not inconsistent with the notion that the concept of a city has outlived its usefulness, and this is argued quite forcibly by academics such as Brenner and Schmid (2014). In their paper ‘The ‘Urban Age’ in Question’, they argue that a city centric focus is ‘empirically flawed’ in that cities are merely statistical artefacts and that the idea of an urban age is a ‘chaotic conception’ or bad abstraction using the terminology of the realist philosophers. Scott and Storper (2015) in an equally vibrant and challenging paper ‘The Nature of Cities’ argue that despite the fragmentation of the field of urban studies over the last 25 or more years, cities are still definable and they invoke a particularly mechanistic realisation of cities as being the outcome of “two main processes, namely the dynamics of agglomeration/polarization, and the unfolding of an associated nexus of locations, land uses and urban interactions”. Somewhat ironically despite their rather traditional definition, they do not refer at all to the long-standing ideas that form part of location theory or even urban economics in articulating this image of the city, despite the emerging synthesis of many of these ideas with new conceptions from physics that are fast converging on a new paradigm which is loosely called a new science of cities (Batty, 2013).

To generate an improved understanding of the urban age and of a world where everyone lives in cities, we need to fashion a science that is consistent with the great array of problems, issues and idiosyncratic features that we find in all cities. No science can progress without there being agreement about its objects of concern for the starting point of our understanding must be observations of the urban condition about which there is no ambiguity. In this quest, we need to be absolutely clear as to how we define a city but this must be set against our knowledge that different definitions are required for different purposes, and that their definition pertains to one point in time and may change as cities evolve and as society becomes more urbanised. To address the sustainable development goals, and the new urban agenda, the time to begin this is now for it is absolutely essential to construct indicators of city performance that enable us to compare the extent to which individual cities are moving towards these goals.

This probably means that a definition of cities is needed that focuses on their geographical or spatial extent and therein lies the problem. We know that as we aggregate spatial data, our interpretations of this data can change, sometimes quite radically to the point where our conclusions about the relative meaning and value of some phenomena can be turned upside down. This problem has been known by spatial analysts for a long time. It generates a so-called ‘ecological fallacy’ that was first exposed more than half a century ago by Robinson (1950) when he found that the correlation between the level of immigration and the degree of illiteracy at the US state level was negative but when considered at a much finer level of spatial granularity became positive, thus reversing the conclusion as to causal effect. More recently Openshaw (1983) phrased the problem as the modifiable areal unit problem, meaning that as one manipulated the boundaries of a set of units such as cities, the distribution of the data within the system of cities could change in such a way that conclusions as the form of the distribution could be reversed. In short, for one definition of city boundaries, one could draw different conclusions as to the performance of the cities in question with respect to a different set of boundaries for the same set of cities. This is another form of ecological fallacy and was observed even earlier by Gehlke and Biehl (1934).

The problem is very easy to define with respect to every city that we know. For example, were I to define the performance of London as pertaining only to its historic core – the so-called square mile of ‘The City’ as it is known administratively – then I would probably compile data on its economic growth rate which showed that it was amongst the top performers for attracting income and jobs of any of the cities in western Europe during the last five years. Were I to widen this definition to embrace London’s inner areas, my conclusion might change quite radically in comparison to similar definitions of cities in Europe and as I continued to aggregate and change the boundary of the city to the point where we included all the metropolitan area, my conclusions as to its performance would continue to change. It is thus crucial to agree on city definitions if we are to measure progress towards sustainable development but the problem goes beyond the geographical. We must be absolutely clear about the way we categorise and define different groups within the population for as we change these typologies, indicators describing their attributes will also change in meaning. To an extent what we are saying is that whenever we look at cities which are intrinsically ill defined, we must be very specific about their definition, and sensitized to the relativities that will continue to plague their definition for such is the nature of the phenomena we are dealing with. In short what we see when we look at cities is dictated by the lens through which we view them. This is a kind of quantum effect for cities that we must learn to live with continually. There are few absolutes when it comes to cities.

It is not easy to embrace more fluid viewpoints such as that suggested by Brenner and Schmid (2014) which changes the focus away from measuring anything geographic within the city. Perhaps as Boone et al. (2014) suggest, we might move to a situation where we measure the urban-ness of a place on a continuum from completely rural to completely urban but at some point, thresholds must be established which determine whether a place is urban or not. In striving to measure progress towards sustainable development, at some point comparisons must be made and lines of demarcation between the urban and the non-urban and the city and the non-city must be drawn. In short as Parnell and her colleagues argue we must embrace the urban and define our indicators as being city-centric. Stable definitions of cities are thus needed that depend on the kinds of indicators that we develop to measure SDGs and these must pick up sufficient detail to measure inequality as well as poverty and the ‘interdependence of social, economic and environmental values’ across different scales and temporal periods. Moreover the spatial scale must be relevant to the data that needs to be collected to first develop appropriate definitions of cities as geographical systems and second to enable the best measure of performance of cities to be evaluated. Our city definitions must be inclusive of all kinds of cities – north and south, developed and developing – and there is little doubt that the data problems that we face are enormous. This is despite there being new methods for generating data at comparable scales, particularly from remote sensing, crowdsourcing and a host of developments in the way we are beginning to use information technologies to sense and detect our own presence and footprint. These problems of definition will continue to dominate this new science, but at least we can now begin to see this as intrinsic to the urban problem. In addressing the NUA and moving towards the targets implied by the SDG, we need to wear these definitional dilemmas on our sleeves and make the discussion as transparent as possible.


Batty, M. (2011) Commentary: When all the World’s a City, Environment and Planning A 43, 765–772.

Batty, M. (2013) The New Science of Cities, The MIT Press, Cambridge, MA.

Batty, M. (2015) Editorial: Cities in a Completely Urbanised World, Environment and Planning B 42, 381 – 383.

Boone, C. G., Redman, C. L., et al. (2014) Reconceptualizing Land for Sustainable Urbanity, In K. C. Seto and A. Reenberg (Editors) Rethinking Global Land Use in an Urban Era, The MIT Press, Cambridge, MA, 313-330.

Brenner, N. and Schmid, C. (2014) The “Urban Age” in Question, International Journal of Urban and Regional Research 38, 731-755.

Gehlke, C. E. and Biehl, K. (1934) Certain Effects of Grouping upon the Size of the Correlation Coefficient in Census Tract Material, Journal of the American Statistical Association 29, 169–170.

Openshaw, S. (1983) The Modifiable Areal Unit Problem, CATMOG 38, Geo Books, Norwich, UK.

Parnell, S., Crankshaw, O, and Acuto, M. (2016) 2030 Policy Endorsement of a Sustainable Future: Implications for Urban Research, Discussion Paper, Prepared for the ESRC Urban Transformations Programme.

Robinson, W. S. (1950) Ecological Correlations and the Behavior of Individuals, American Sociological Review 15, 351–357.

Schumacher, E. F. (1973) Small Is Beautiful: A Study of Economics As If People Mattered, Harper Collins, New York.

Scott, A. J. and Storper, M. (2015) The Nature of Cities: The Scope and Limits of Urban Theory, International Journal of Urban and Regional Research 39, 1-15.