Urban Density Is Not an Enemy in the Coronavirus Fight: Evidence from China
This page in:
Making a link between a city’s density and its vulnerability to epidemics may seem like an obvious connection. But it may, in fact, be off the mark.
Since the outbreak of the coronavirus (COVID-19) and its spread across the globe, places with high urban population density have seemed to be especially at risk to some observers. A common argument is that high population density makes cities more vulnerable to epidemics because of the possibility of frequent interpersonal contacts. New York City is often cited as a prime example.
This argument may sound straightforward, but on examination, its premise is not well-grounded. If you look around the world, some extremely dense cities, such as Singapore, Seoul, and Shanghai, have outperformed many other less-populated places in combating the coronavirus.
Based on evidence from China, we would like to present a counter argument.
To find out whether or not population density is a key factor in the spread of the coronavirus, we collected data for 284 Chinese cities on two relevant indicators: (i) the number of confirmed coronavirus cases per 10,000 people; and (ii) the population density in the built-up urban area. We presented this data in Figure 1 below.
The reason for choosing Chinese cities is that they have been through the entire circle of the coronavirus outbreak, and the numbers of confirmed cases have stabilized. The 12 cities from Hubei Province, where Wuhan is located, are excluded from this sample because they were subject to a very different level of risk when the virus first hit the country. In addition, confirmed coronavirus cases from inbound international flights are not included either, as they were attributed to the city of the first entry point and, therefore, are irrelevant to the local spread of this contagious disease. The size of the bubble is proportional to the per capita GDP in each city, to illustrate whether city wealth has an impact on response to the coronavirus outbreak.
The evidence we’ve found does not support the argument that density is a key determinant of coronavirus transmission risk. As illustrated by Figure 1, cities with very high population densities such as Shanghai, Beijing, Shenzhen, Tianjin, and Zhuhai have had far fewer confirmed cases per 10,000 people. We notice that the group of dense cities are also the wealthier ones (with bigger bubbles), making them more able to mobilize enough fiscal resources to cope with the coronavirus. This partly explained their low infection rates.
On the contrary, cities with the highest coronavirus infection rates were those with relatively low population densities, in the range between 5,000 to 10,000 people per square kilometer. The higher infection rates can be attributed either to their strong economic connection with Wuhan – which is the case for Wenzhou, with over 180,000 people working in Wuhan despite the long distance between the two cities – or to their geographical proximity, which explains the situation in Xinyang, Zhumadian, Xinyu, and Yueyang, which are close to the provincial border of Hubei.
To further investigate the possible spatial decay effect in the spread of the coronavirus, we made another plot using the infection rate against each city’s distance to Wuhan; see Figure 2 for details. It shows that the chances of infection decline as distance to Wuhan increases. Cities in Xinjiang province (e.g., Karamay, Urumqi, and Turpan) that are over 2,500 kilometers away from Wuhan, have extremely low coronavirus infection rates.
Higher densities, in some cases, can even be a blessing rather than a curse in fighting epidemics. Due to economies of scale, cities often need to meet a certain threshold of population density to offer higher-grade facilities and services to their residents. For instance, in dense urban areas where the coverage of high-speed internet and door-to-door delivery services are conveniently available at competitive prices, it is easier for residents to stay at home and avoid unnecessary contact with others.
The World Bank’s research on urban geography has found that the 3Ds – higher density, shorter distance, and fewer divisions (i.e., better market integration) – are essential for economic development. From an urban resilience point of view, the potential risks of public health emergencies associated with the 3Ds need to be managed carefully, because as urbanization continues, we will be living in a world where people are even closer to each other than before, both spatially and economically. Over time, ensuring well-designed institutions, high-quality infrastructure, and effective interventions (e.g., social distancing and lockdown to flatten the curve of disease transmission) will be the ingredients to making cities stronger against infectious diseases.
Of course, further evidence is needed to draw conclusions on the beneficial impacts of density. But we can already say with confidence that density is not an enemy in the fight against the coronavirus.
Data source: COVID-19 confirmed cases as of April 9, 2020 from Ding Xiang Yuan real-time monitoring platform. https://ncov.dxy.cn/ncovh5/view/pneumonia?from=groupmessage&isappinstalled=0
Population density calculated based on data from China City Construction Yearbook 2018 released by Ministry of Housing and Urban-rural Development. http://www.mohurd.gov.cn/xytj/tjzljsxytjgb/
Per capita GDP data from 2018 were consolidated from the statistical yearbooks of the provinces and distance to Wuhan was calculated using GIS analytic tools.
Note: Selected cities do not include Hong Kong SAR, China; Macau SAR, China and Taiwan, China. Number of COVID-19 confirmed cases did not include those without symptoms.
Related links:
- Subscribe to our Sustainable Cities newsletter
- Follow @WBG_Cities on Twitter
?
good to have current information and thought
Thank you for an interesting and insightful article.
Although it feels good to see support in favor of density, we still need more evidences to establish link between a city's density and its vulnerability to epidemics. Singapore, Seoul, and Shanghai, the examples cited in the article (all are resourceful cities!), have certainly performed well in combating the coronavirus, but it might be the response measures adopted by the cities (e.g., lockdown, mass testing etc.) rather than density that helped these cities outperform other (less dense) cities.
Even in the case of China, excluding cities from Hubei Province raises a question if dense cities are more susceptible to the 'first' outbreak of epidemics.
The article finds high coronavirus infection rates in cities with relatively low population densities in China's case. However, the high infection rates are attributed, in the article itself, to the economic connection with or geographical proximity to Wuhan. So it is not clear how low densities should/could be blamed for the underperformance to control infection rates.
Undoubtedly high densities have significant benefits primarily due to economies of scale. But above some threshold, high densities might bring more negative than positive externalities. For instance, it would be difficult to maintain social distancing in densely populated apartments where the use of elevators is inevitable. Same applies for crowded marketplaces, which are common in many dense cities.
More studies/discussions (including cases outside China) are needed to understand the relationship between a city's density and its vulnerability to epidemics. But the authors of this article deserve appreciation for contributing to this discussion.
Thank you for your excellent comments. In fact, our key argument here is that we should avoid making simplistic attribution of COVID-19 infection rates to population density in cities, which has been a common misunderstanding. We surely did not blame low densities for the underperformance to control infection rates. As you rightly pointed out, indeed many other factors matter, such as the cities’ fiscal resources, management capacity to cope with emergencies, distances and economic connections to the epicenter, configurations of the built environment, and even cultural reasons---people are more used to wearing masks in some countries. The next blog in this series will touch on hotspots and the built environment in cities and please stay tuned.
The pandemic spread can't be explained by a single factor. China's highly restrictive strategy of fighting the pandemic could have moderated the effect of popukation density (PD). When people can't get out even of their apartments, PD effects are reduced. Looking beyond Wuhan and Hubei makes no sense as these were quarantined. Focus on Wuhan vs NY City, where the respective PDs are 3200 vs 27000.
urban density may in fact be an enemy. I think that one missing key component to this analysis is time.
Can you get for each city the number of days each one took to get to 1, 10 and 50 cases since the first China's case? If you can try to plot this 3 graphs vs density.
I think that it could give us a fair comparison about cities reactions over time and their density.
I agree that is not a matter of two variables correlation but i don't think that we should discard that they can be part of a multivar correlation problem. If that would be the case we actually would conclude that in density is an enemy.
Thank you for bringing up the factor of time. The factor of time would have made sense if there were instances of cities where the speed by which the epidemic spread would have resulted in crashing the medical system and hence contributed to an exacerbation of the situation and an exponential growth in infections (this was the case in Wuhan for instance, but, as explained in the blog, the 12 cities in Hubei province were excluded because the statistics on incidence were not reliable – as the first hit cities, medical authorities were not sure what they were dealing with, and accounting of cases did not follow the standardized methodology).
It is clear by now that population density is one of many factors that determines viral spread. But there are so many problems with your confident assertion that "density is not an enemy in the fight against the coronavirus."
First you used statistics from China without question (even though Chinese authorities were revising death statistics as recently as 3 days before your report). Then you excluded the 12 cities worst-hit cities from Hubei province to fit your data model. Then, you ignore that lockdown measures implemented in China are impractical elsewhere (for instance, a well-resourced network of community associations directed by the Communist party is absent in most other countries). But most disappointing, you've not examined the problems with opposing views. Instead, you confidently contradict decades, if not centuries, of data-based scientific insight on the basis of your analysis.
You are part of the urban planning team at the World Bank -- your views are taken seriously. Please retest your views against opposing viewpoints. For instance, research papers on the medical research publication site PubMed that contradict your viewpoint. Or please talk to Jed Kolko, chief economist with the jobs website Indeed, whose recent analysis found that across the US, denser counties are seeing higher rates of coronavirus infections and deaths from COVID-19, even after taking into account factors such as weather, race and age. And Fernando Rodriguez, a public health professor at Madrid's Autonomous University, who has been quoted as saying, "[Spanish] cities are built vertically, there is a lot of density and this can also facilitate the transmission of the virus".
Thank you for your interest in this topic, and for sharing with us studies that may lead to different conclusions. We very much welcome the different viewpoints. In fact, our key argument here is that we should avoid making simplistic attribution of COVID-19 infection rates to population density in cities, which has been a common misunderstanding.
The reason why we used data from Chinese cities was because they have been through a relatively complete cycle of COVID-19 outbreak and the numbers of confirmed cases started to stabilize when the analysis was conducted. While in other parts of the world, it is still a developing situation and it was too soon to draw any conclusion yet as outbreaks haven’t stabilized, nor is it clear the peaks have been reached. The reason for excluding data from the 12 cities from Hubei Province is that at the early stage of COVID-19 outbreak in the province, there were no standard criteria for diagnosis of this disease unknown to all of us, and because shortage in testing kits and medical workers at that time were important variables that could explain the discrepancy between the actual infection situation and reported numbers. Hubei were using its own reporting standard in the early days, and made a few adjustments afterwards, which makes the accumulated numbers of confirmed cases not comparable with those from other cities in China that have adopted standardized methods for diagnosis and reporting. Hence the rationale to exclude the 12 cities in Hubei, rather than any attempt to fit the model. You’ll note the R2 is not high enough, so there’s clearly more variables to control for.
Also, as we mentioned in the note of the Figures, confirmed cases in China did not include people who test positive but have no symptoms. This makes cross-country comparison impossible. Not to mention there’s few other countries where the situation has stabilized and thus you can compare (needless to say this will change in the coming months); even Singapore, the situation has changed dramatically in the past 2-3 weeks. Note that we also reviewed relevant literature on the relationship between population density and epidemics. Investigations of possible links between population density and the propagation and magnitude of epidemics have so far proven inconclusive (Li, Richmond and Roehner, 2018).
Finally, what is clear is that there is a lot of nuance even when we talk of a city’s population density. In a city like NYC, there’s major differences in the incidence of cases between boroughs, which suggest the quality of infrastructure and service delivery including public health, the ability to enforce social distancing policies, and the overall situation of households’ vulnerability appear to be important explanatory variables. We are sure you read Jed Kolko’s work that suggests the same. Also note that places with similar population density like Mumbai and NYC have very different built environments (compare the vertical density of Manhattan with the horizontal density of Dharavi) and thus will have a very different ability to put in place social distancing policies.
In summary, we do not believe that population density is the enemy in this, but the poor management of the density is. You may want to explore our other blog entitled “Cities, Contagion and the Coronavirus” https://blogs.worldbank.org/sustainablecities/cities-crowding-and-coron…, which provides a methodology for the identification of hotspots of transmission and contagion risk based on population density, the built up area, and the quality of infrastructure within cities.
Thank you for your article. I wonder if the result would change if more socio-economic aspects are considered. I'm thinking of density of homelessness, density of poverty, a concentration of uneducated population. I don't know, of course, whether such data are available for Chinese cities.
Thanks for sharing. This is definitely an interesting study_ under the current pandemic situation, it makes us to rethink about the urban density.
In the article, it seems that density is not an enemy in the fight of the pandemic, however, other main factors need to be weigh into the statistic studies_ the demographic composition of the top high density cities in China and the time when the pandemic occurs.
The makeup of the population for the top high density cities in China, including Beijing, Shanghai, and Shenzhen, have a unique demographic structure. A huge number that contributes to these cities’ populations are migrant workers _ they left the city and went home for the Chinese new year at that time. Then the quarantine starts. People are staying in their hometown other than in the high density cities where they work and actually counted as part of the resident population. So the actual number of population in those top high density cities, at the time when it happens, are much less than the statistics in the yearbooks and normal time. Therefore it seems those cities have fewer confirmed cases per capita, might not be the reality using the population per the yearbook.
Along with the fact that higher density urbanized cities would offer higher grade infrastructure, health care facility and capacity, etc. , it might be too soon to draw the conclusion that density is in our favor facing a pandemic spreading rapidly and widely like this.
Very relevant to have these insights in a systematic way before jumping to conclusions especially where full benefits of density are not understood, notwithstanding its role in sustainable growth
An interesting analysis . Considering grossdensity of built up area may not be correct. Net density of smaller residential spatial units like zone/sector is more relevant. Even within a city the prevalence of corona infection is lntense at certain locations called hot spots. Possibly these are the high density units. Considering density in isolation also May not be adequate. The level of physical infrastructure like water supply, sanitation and housing standards are also major factors.
Is high use of public transport a factor causing more infections?
Thanks for your comments. You may want to check our blog related to this very subject, which you can view here: https://blogs.worldbank.org/sustainablecities/cities-crowding-and-coron…
Interesting study and article