Estimating Chinese GDP Using Night-Lights Data

September 14, 2017

China night lights

This is the third post in a three-part series on measuring economic activity in China.

The previous posts in this series covered the challenges in measuring Chinese gross domestic product (GDP) and some alternative methods used in an attempt to obtain better estimates, such as energy consumption and the Li index.

This post examines one other method that some have used to measure Chinese economic activity—one based on the intensity of man-made night lights (luminosity), as discussed in an article in The Regional Economist

“Unlike indexes of human-produced economic data, these data are immune to falsification or misreporting,” wrote Assistant Vice President and Economist Michael Owyang and Senior Research Associate Hannah Shell.


The authors explained that the night-lights data are gathered by Air Force satellites that have been circling the earth 14 times a day since the 1970s, which measure the light intensity emanating from specific geographic pixels.

They cited work by economists J. Vernon Henderson, Adam Storeygard and David N. Weil, who argued that the night-lights data are a good proxy for economic activity because consumption of goods in the evening requires light.1

Owyang and Shell noted that these researchers created a dataset based on information from the satellites and used it to estimate GDP in countries with low-quality data. In particular, the researchers estimated the change over a 14-year period as well as annual changes in economic activity for 188 countries from 1992-2008.

Cumulative Growth

“One way to assess the quality of Chinese economic data is to look at the difference between the growth rate of real GDP reported by the government and the estimated growth from 1992 to 2006 using the night-lights data,” Owyang and Shell said.

They noted that official statistics reported that real GDP growth in China was about 122 percent over that period. In contrast, the night-lights data predicts that growth was 57 percent.

Owyang and Shell wrote: “This sizable gap suggests cumulative Chinese growth over the years could be overstated by as much as 65 percent.” They added that only Myanmar had a larger gap between the official and estimated numbers.

Annual Growth

Regarding the researchers’ estimated annual real GDP growth using night-lights data, Owyang and Shell noted that real GDP growth is consistently overstated, particularly before 1996.

They also looked at the researchers’ estimated real GDP growth using night-lights data in conjunction with China’s long-term growth path based on official data. Owyang and Shell explained that, by including this long-term growth trend, the night-lights data are only informing annual fluctuations from the trend.

While that measure also suggests growth was overstated before 1996, the authors noted that it tracks the official growth rate more closely after 1996. It is also more volatile than official estimates. “This supports the other indexes’ conclusions that quarter-to-quarter fluctuations in Chinese real GDP growth are smoothed, but likely move in the correct direction,” they wrote. 

(For a figure showing these growth estimates using luminosity data, see China’s Economic Data: An Accurate Reflection, or Just Smoke and Mirrors?)


The authors concluded that growth in China was likely overstated during the country’s transition from a command economy to a market economy, which has possibly led to an exaggerated level of output in recent data. If true, that could mean an overstated share of world GDP for China.

“However, while the level of Chinese GDP may remain overstated, both the Li index and estimates from the night-lights data suggest that the recent growth rate numbers for Chinese official data are more reliable,” Owyang and Shell wrote. “They may be subject to collection error and smoothing, but appear to be moving in the correct direction.”

Notes and References

1 Henderson, J. Vernon; Storeygard, Adam; and Weil, David N. “Measuring Economic Growth from Outer Space.” American Economic Review, Vol. 102, No. 2, 2012, pp. 994-1028.

Follow the Series

Additional Resources


Related Topics

This blog offers commentary, analysis and data from our economists and experts. Views expressed are not necessarily those of the St. Louis Fed or Federal Reserve System.

Email Us

Media questions

All other blog-related questions

Back to Top