How can we apply complexity science to urban policy?

Michail Fragkias
Boise State University, USA

In the past few decades, urban researchers and practitioners across the world have jointly enriched our understanding of the new challenges that metropolitan areas in the 21st century are facing. The new challenges for our cities are: (i)  driven in part by global (environmental) change processes and are compounded; that is, their potential impact is severe, especially in places that have not resolved challenges that characterized the past century; (ii) occuring increasingly at the intersection of administratively-defined and bureaucratically-managed urban sectors, such as transportation, water, energy, sanitation, etc.; and, (iii) moving beyond the realm of complicated problems, to that of complex problems; and solutions to the challenges can emerge only if we adopt the viewpoint of complex systems or complex networks.

How exactly complex systems and networks research relates to policymaking can be a murky subject that involves some hand-waiving and is possibly one of the weak links of complexity science. Fundamentally, complex systems involve a set of heterogeneous agents whose behavior is interdependent (and partly a stochastic process) that leads to aggregate properties such as non-ergodicity, phase transitions and tipping points, emergence and universality (Durlauf, 2005). While the ideas and approach are not new, they are gaining in popularity.  However, it not always clear how complexity theory can be translated to complexity practice, or what the theory’s relevance is for societal problems.

My claim is that science meets policy in the area of complexity empirics – i.e., the empirical evaluation of the aforementioned aggregate properties of complex systems. Here, I focus exclusively on the scaling properties of urban systems. Scaling describes how a given quantifiable characteristic of a system depends on the size of the components of the system – for example, how wealth, crime or emissions occurring in a system of cities depends on the size of these cities. Scaling relationships have been identified for a wide array of socio-economic variables and are an indication of generic social mechanisms and properties of social systems at play across the entire urban system. Mechanisms such as networks and flows, nonlinearities and feedback loops integrate complex interactions among the individuals, households, firms, and institutions living, residing and operating in these spaces, leading to emergent phenomena such as scaling laws. Below I use two recently published studies to make a point of the usefulness and potential limits of complex systems thinking for urban policy making (Fragkias et al., 2013, Oliveira et al., 2014).

Both papers explore the scaling of CO2 emissions from cities of the United States. The CO2 data for both papers are obtained from the Vulcan Project (Gurney et al., 2009). The Fragkias et al. study used the 366 Metropolitan Statistical Areas (MSAs) and the 576 Micropolitan Areas, which together constitute the 942 urban ‘core based statistical areas’ (CBSAs) of the United States – the population of urban settlements in the United States. The Oliveira et al. study utilizes the 274 MSAs – the upper tail of the distribution used in Fragkias et al. – but also defines boundaries of cities according to a City Clustering Algorithm. This is an admittedly more “organic” method of defining urban areas as contiguous commercial and residential areas. The population data in Fragkias et al. are calculated from county level data provided by the Department of Commerce’s Bureau of Economic Analysis (BEA). Oliveira et al. extract population data from the Global Rural-Urban Mapping Project (GRUMPv1) at Columbia University.

The choice of a definition for an urban area (and perhaps, sources of population data) is not inconsequential. The studies arrive at the following main findings and conclusions: Fragkias et al. (2013) find a statistically linear scaling relationship between CO2 and urban population: this “suggests that large urban areas in the U.S. are only slightly more emissions efficient than small ones”. Oliveira et al. (2014) identify a superlinear scaling relationship between CO2 emissions and city population; they suggest that “the high productivity of large cities is done at the expense of a proportionally larger amount of emissions compared to small cities.” The authors claim that the difference arises from the fact that an administrative definition of cities produces a different scaling law compared to a functional definition. Furthermore, that the metropolitan statistical area definition is biasing the scaling exponent downwards “due to the overestimation of MSA areas.”

While Oliveira et al. make a good argument for superlinear scaling, the gap in the findings of the two studies could be further widened or narrowed by methodological issues. Oliveira et al. exclude the small urban settlements of the U.S. and there is evidence that including the full distribution of urban settlements in the analysis produces exponents of smaller magnitudes (Fragkias et al., in review). Furthermore, the choice of population size data (GRUMPv1 vs BEA) is unexplored.

Notwithstanding these issues, and given the apparently conflicting messages of the above studies, can we learn anything that can be applied towards policymaking? One answer in the literature has been argued by Louf and Barthelemy (2014). They write: “Faced with these two opposite results, what should one conclude? Our point is that, in the absence of a convincing model that accounts for these differences and how they arise, nothing.”  While they find that scaling analysis is important, data analysis on its own is not enough for concrete conclusions:

It should therefore be obvious that, until they have a satisfactory understanding of the mechanisms responsible for the observed behaviours, scientists should refrain from giving policy advice that might have unforeseen, disastrous consequences. If they choose to do so anyway, policy makers should be wary about what is, at best, a shot in the dark.

While the authors make a good point on the question of interpretability, I have a differing perspective on the policy relevance of the two analyses. Certainly, administrative boundaries and the modifiable areal unit problem (MAUP) is as important for complexity empirics as it is for spatially integrated social science. However, because of that, one may not claim that the results are not policy relevant because of that. There are a few points that a policy maker can consider:

  • Non-trivially, there is a scaling/power law relationship present in CO2 emissions for the U.S. system of cities. This first point is important since policymakers need to be aware of identified scaling and power laws in urban systems even if the findings are sensitive to the choice of geographical scale. Policymakers are often faced with conflicting findings from research conducted in non-complexity oriented research and have learned to implement policies under this reality. So, conflicting results emerging from complexity empirics research should not be a reason for shying away from disseminating our findings.
  • So far, we have no evidence of sublinear scaling for any alternative definition of a city – only linear and superlinear. This point is a finer one since it refers to research results that are not a shot in the dark. The message to policymakers so far should be that given the aforementioned research, larger cities are not more emissions efficient; they may be as emissions efficient or inefficient when compared to smaller cities. The degree of inefficiency depends at minimum on the definition of a city. As scientists, we should always be ready to verify and accept new scientific evidence but until that evidence arrives, we should feel comfortable discussing the extent to which our results can guide policy.
  • Finally, we cannot view the results on CO2 scaling in isolation. We need to consider the findings from the CO2 research alongside other scaling findings regarding wealth, crime, innovation, etc. Recent findings of superlinear scaling in wealth, innovation and crime, linear to superlinear scaling CO2 emissions but sublinear scaling in infrastructure networks, have significant implications for policy when viewed as a whole (either as a test of theory or a collection of a variety of empirical findings). A scaling paper focusing on a single urban property can not be a guide for policymaking since it is effectively “siloed” from other findings; only when viewed as a collection of findings on a variety of urban properties, is scaling analysis relevant for sustainability science and policy.

Researchers have made significant progress in operationalizing complexity empirics for urban sustainability. Major challenges for interpretation remain on the fronts of urban definitions, data selection, potential heterogeneities and forms of temporal dependence. We should not let the ongoing scientific debate become an obstacle in allowing complexity empirics to enter global policymaking agendas.

Rafting

Dr. Michail Fragkias is Assistant Professor at the Department of Economics, Boise State University, USA.
Website
Email

Header Image: London, UK.  Credit: Photographer Unknown/flickr.com

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s