Skip to Content
Uncategorized

Data Mining Reveals the Extent of China’s Ghost Cities

Overdevelopment in China has created urban regions known as ghost cities that are more or less uninhabited. Nobody knew how bad the problem was until Baidu used its Big Data Lab to find out.

In recent years, China has undergone a period of urban growth that is unprecedented in human history. The number of square kilometers devoted to urban living grew from 8,800 in 1984 to 41,000 in 2010. And that was just the start. China used more concrete between 2011 and 2013 than the U.S. used in the entire 20th century.

Some of this building has been misplaced. In various parts of China, developers have built so much housing so quickly that it has outstripped demand, even in the world’s most populous country. The result is the well-publicized phenomenon of ghost cities—entire urban areas that are more or less deserted.

But much of the reporting on ghost cities is anecdotal or based on unreliable measurements such as a simple count of the number of lights on at night in residential buildings. That’s a particularly inaccurate method, not least because it ignores seasonal variations caused by tourism. Many places are busy during the tourist season but empty during the off-season, and not just in China. So being unable to distinguish these from ghost cities is something of a problem.   

And that raises an interesting question: how bad, really, is the problem of ghost cities in China?

Today we get an answer of sorts thanks to the work of Guanghua Chi at the Big Data Lab at Baidu, China’s version of Google and one of the biggest Web companies on the planet (for an inside look at Baidu’s attempt to compete with the Googles and Amazons of the world, see “A Chinese Internet Giant Starts to Dream”), and a few pals. These guys have used the location data that Baidu gathers about its users to work out exactly where China’s ghost cities lie. And by tracking people over time, the Baidu team can distinguish between ghost cities and towns that are seasonally empty.

Baidu has an extraordinary data base to draw on. Some 700 million people have signed up for the services it offers, a significant proportion of China’s 1.36 billion inhabitants.

Of course, these people are predominantly younger so the data is not broadly representative of Chinese society. However, it does give an idea of urban density and how this varies both in time and across the country at a resolution measured in a few tens of meters.

Baidu tracked its users throughout each day for more than six months in 2014 and 2015 and used a common clustering algorithm to calculate their home location. They then correlated these locations with another data set of known residential areas to work out where they lived. They then calculated the urban density—the number of people living in each 100 meter by 100 meter area.

The Chinese Ministry of Housing and Urban-Rural Development states that a standard urban region with an area of one square kilometer should house 10,000 people. Guanghua and co define a ghost city as an area with half this density.

To find out where these areas lie, they’ve built an algorithm that simply hunts for areas with a density that is less than half the Chinese standard. But they go further and track the density over time to see whether it picks up during the tourist season.

The results make for interesting reading. Not only does the team identify more than 50 ghost cities in China, they are also able to analyze their spatial distribution and how it relates to the surrounding geography and urban setting.

They give as an example the city of Rushan, which is located near the sea and has 21 miles of beautiful coast line that has been heavily developed. The houses here are empty for much of the year but densely populated during the tourist season. This clearly shows Rushan as a tourism center rather than a ghost town.

By contrast, the city of Kangbashi has a clear weekly cycle of population change albeit with very low density of residents. That’s a clear sign that this is a ghost city.

That’s interesting work that properly measures the ghost city phenomenon for the first time. “Instead of just counting the number of homes with light at night in certain residential areas as the indicator of “ghost city,” Baidu big data can count the population precisely, in real time, and in national scale,” say Guanghua and co.

That should help the Chinese government make better planning decisions in the future and should also inform people thinking of moving to these areas. (Baidu has been careful not to rank the ghost cities in this study for fear of influencing property prices there.)

And if anybody wants to explore further, these guys have put their work data on an interactive website, www.bdl.baidu.com/ghostcity, for anybody to use. Enjoy!

Ref: arxiv.org/abs/1510.08505 : “Ghost Cities” Analysis Based on Positioning Data in China

Keep Reading

Most Popular

Large language models can do jaw-dropping things. But nobody knows exactly why.

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

OpenAI teases an amazing new generative video model called Sora

The firm is sharing Sora with a small group of safety testers but the rest of us will have to wait to learn more.

Google’s Gemini is now in everything. Here’s how you can try it out.

Gmail, Docs, and more will now come with Gemini baked in. But Europeans will have to wait before they can download the app.

This baby with a head camera helped teach an AI how kids learn language

A neural network trained on the experiences of a single young child managed to learn one of the core components of language: how to match words to the objects they represent.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at customer-service@technologyreview.com with a list of newsletters you’d like to receive.