Where are China’s ghost cities?

This article is published in collaboration with The Washington Post.

For decades, China has been engaged in a building boom of a scale that is hard to wrap your mind around. In the last three decades, 260 million people have moved from the countryside to Chinese cities — equivalent to around 80 percent of the population of the U.S. To make room for all of those people, the size of China’s built-up urban areas nearly quintupled between 1984 and 2010.

Much of that development has benefited people’s lives, but some has not. In a breathless rush to boost growth and development, some urban areas have built vast, unused real estate projects — China’s infamous “ghost cities.” These eerie, shining developments are complete except for one thing: people to live in them.

China’s ghost cities have sparked a lot of debate over the last few years. Some argue that the developments are evidence of the waste in top-down planning, or the result of too much cheap funding for businesses. Some blame the lack of other good places for average people to invest their money, or the desire of local officials to make a quick buck — land sales generate a lot of revenue for China’s local governments.

Others say the idea of ghost cities has been overblown. They espouse a “build it and they will come” philosophy, pointing out that, with time, some ghost cities fill up and turn into vibrant communities.

It’s been hard to evaluate these claims, since most of the research on ghost cities has been anecdotal. Even the most rigorous research methods leave a lot to be desired — for example, investment research firms sending poor junior employees out to remote locations to count how many lights are turned on in buildings at night.

Now new research from Baidu, one of China’s biggest technology companies, provides one of the first systematic looks at Chinese ghost cities. Researchers from Baidu’s Big Data Lab and Peking University in Beijing used the kind of location data gathered by mobile phones and GPS receivers to track how people moved in and out suspected ghost cities, in real time and on a national scale, over a period of six months. You can see the interactive project here.

Vacant housing areas. Image from “Ghost Cities” Analysis Based on Positioning Data in China. Guanghua Chia, Yu Liu, Haishan Wu

Google has been blocked in China for years, and Baidu dominates the market in terms of search, mobile maps and other offerings. That gave the researchers a huge data base to work with —  770 million users, a hefty chunk of China’s 1.36 billion people.

To identify potential ghost cities, the researchers created an algorithm that identifies urban areas with a relatively spare population. They define a ghost city as an urban region with a population of fewer than 5,000 people per square kilometer – about half the density recommended by the Chinese Ministry of Housing and Urban-Rural Development.

The map below shows 20 of 50 cities with large vacant housing areas that they tracked. (The researchers don’t disclose all 50 cities, in order to avoid influencing real estate prices, they say.) But most of them are medium-sized cities in China’s East, located mostly in a city’s periphery or in a newly developed area.

“Ghost Cities” Analysis Based on Positioning Data in China. Guanghua Chia, Yu Liu, Haishan Wu

Here the researchers ran into another problem: Perhaps some of these were not ghost cities, but tourist towns. As in other places around the world, tourist destinations are largely vacant at certain times of the year, but can fill up at other times. The report illustrates the difference with a tale of two cities: Rushan, a seaside development in Shandong province on China’s east coast, and Kangbashi, a suburb of Ordos, a coal- and resource-rich area in Inner Mongolia, in China’s north.

To weed out the tourist destinations, the researchers tracked how population density varied over time in large vacant housing areas. You can see the patterns they measured in the graphic below. The population of Kangbashi, shown by the blue line, swells during the week, and falls during the weekend and holidays – a clear sign that people are working there but living elsewhere, the researchers say. Rushan, shown by the green line, sees its population spike during national holidays – periods where Kangbashi empties out.

“Ghost Cities” Analysis Based on Positioning Data in China. Guanghua Chia, Yu Liu, Haishan Wu

According to the researchers, these patterns indicate that Kangbashi fits the profile of a sparsely inhabited “ghost city,” while Rushan is better described as a tourist town. Of the 50 cities, they say 26 ended up looking like tourist sites, while 24 were ghost cities.

This kind of data could offer a lot of lessons for the Chinese government as it tries to evaluate the quality, rather than just the quantity, of its economic growth. Too often in China, local officials have sought to give the appearance of development, sometimes without that much concern for achieving the real thing.

Haishan Wu, one of the researchers, says the data can help local governments make better planning decisions in the future, as well as understand where to improve basic infrastructure, which can help turn a ghost city into a functioning one. Wu cites the example of the new city of Zhengzhou, which was once widely reported as a ghost city but is now attracting more people.

The data could also offer plenty of fodder for savvy real estate investors, as well as lessons for China’s urban planners. According to the researchers, the main takeaway is that, to be a functioning community, a city needs jobs, industry, entertainment, healthcare and schools. A city is so much more than a group of vacant apartments.

This article has been updated to include Baidu’s comments. 

Publication does not imply endorsement of views by the World Economic Forum.

To keep up with the Agenda subscribe to our weekly newsletter.

Author: Ana Swanson is a reporter for Wonkblog.

Image: Newly planted trees line a new street in front of residential buildings under construction. REUTERS/David Gray.