Featured in Town Planning Review 94.1: Natural experiments in healthy cities research: how can urban planning and design knowledge reinforce the causal inference?

The editors of Town Planning Review (TPR) have selected the following paper as the Featured Article in TPR 94.1.

This paper will be free to access for a limited time:

‘Natural experiments in healthy cities research: how can urban planning and design knowledge reinforce the causal inference?’ by Guibo Sun, Eun Yeong Choe, and Chris Webster

When asked to describe the paper and highlight its importance, the author stated the following:

Our article comments on what makes a strong research design for natural experiments in healthy cities research. Researchers often ask questions about cause and effect: in this field, causes are built environment interventions via urban planning and design practices, such as park renovation, a new bus line or a housing redevelopment programme; effects are individual and community health and well-being.

Recently, discussions about natural experiment have intensified. This may be partially because the Nobel Prize in Economic Sciences in 2021 went to the three scholars who demonstrated how natural experiments could infer causality in real-world problems when randomised controlled experiments are impossible (The Nobel Prize 2021). In healthy cities research, John Snow’s London cholera study is a classic and oft-cited natural experiment. But what is rarely cited is how Snow spent strenuous efforts in an as-if random treatment-control group assignment, in which he used substantive knowledge of water utility organisation in London to validate the group comparison to conclude cholera is a waterborne infectious disease.

There is a growing interest in natural experiments, both in planning and public health (Craig et al., 2012; Craig et al., 2022; Chanam et al., 2022). But designing studies to advance the science of healthy cities and guide evidence-based policy and planning is still fraught with difficulties. The UK Medical Research Council’s broad definition in 2012 allows for studies to be termed natural experiments even if they do not attempt to approach the as-if randomised assignment as a randomised control experiment. Following a contemporary debate in the philosophy of social sciences, our analysis suggested that such a broad definition is not constructive. Defining a study as a natural experiment based on it involving a naturally occurred intervention, has led to a rather chaotic research landscape. Drawing on three well-documented natural experiment research projects that did not attempt as-if-random assignment to treatment and control groups, we note that the causality inferred is hardly convincing.

When moving from association to causality, the key is how we exclude (un)observable confounders. In a strong natural experiment, detailed planning and design knowledge of how the interventions were produced should have a more central role at the research design stage to obviate confounders. Our paper elaborated on how such knowledge can help discover strong natural experiments with as-if random treatment-control assignment, and real-world relevance. We offer a simple conceptual framework for strong natural experiment design in urban planning and policy, and we refer as the LARD principle: ideally, the assignment to treatment or control group should have the strength of Legal Assignment, by strong public regulation or by private contract; and this barrier should have the effect of preserving an as-if Random Distribution of confounders between the treatment and control groups.

We advocate that as natural experiments become more popular in the planning field, the term should specifically refer only to an experimental design-based approach that uses strong research design. Qualitative research in policy evaluation typically comes as an adjunct to help interpret quantitative surveys. We argue that qualitative research should be more central at the experimental design stage, with planning and design domain-specific knowledge being employed ex-ante to reinforce the power of causal inference by improving research design, not just being employed ex-post to interpret survey findings.

– Guibo Sun, Eun Yeong Choe, Chris Webster
Urban Analytics and Interventions Research Lab, Department of Urban Planning and Design, The University of Hong Kong

Follow us for more updates
Sign up to our mailing list
Twitter | Instagram