Daily Management Review

Big Data: the young blood of statistical economics


The economic growth of the last 30 years was largely the result of advances in information technology. The computer and Internet revolution made possible controlling all stages of production and transportation of goods across the globe. The Earth, in the words of economist Thomas Friedman, has become flat. The next revolutionary step is from "flat" to a fully transparent economy. And apparently, it will happen thanks to Big data. The term first appeared in the journal Nature on 3 September 2008, written by its editor Clifford Lynch.

Many economic processes are difficult to understand, not only in theory but also in practice. Statistics is often useless here. The second economy of the world - China - grew by 6.9% of GDP in 2015, as the authorities claim. However, many analysts doubt the reliability of these data. 

The relatively developed countries give more confidence, yet there is another problem - the data is often greatly delayed. For example, the final estimate of GDP for the past year in the United States (their statistics in many respects is a model) appears only three months after the year’s end.

There are other tools for a more rapid assessment of the situation. This is a variety of surveys on the current state of affairs in the various sectors. Monthly surveys of purchasing managers in industry and services (Purchasing Managers Indices - PMI) are the most popular. Managers need to assess the current situation at the plant over the previous month using several parameters. Index below 50 indicates a decrease in business activity, above 50 - growth. The tool is fast (final PMI for the month comes immediately after it), but not very accurate. For example, there are just 400 managers responsible for the entire industry in the US. However, the US PMI index has been calculated since 1948, and has a reputation of a reliable tool. Other countries, where the PMI was introduced just recently, trust it less.

Another solution is as unexpected as original: the economy can be monitored from space. In 2015, the Californian startup Spaceknow launched its own index - China Satellite Manufacturing Index (China SMI), which monitors the industry in China by analyzing satellite images. The company singled out the country’s 6 thousand largest enterprises. Using constantly updated pictures (now there are 2.2 billion of them), Spaceknows monitor all visual changes at these sites, and then interpret them in terms of economic activity. There are many signs of activity: construction activity, dynamics of smoke emissions from factories or mines stockpiles rising, vans and other vehicles filling the parking near the company, the volume of commodity and raw material resources, the traffic of goods and so on.

All data are summarized in the index, which is very reminiscent of the classic PMI. China SMI’s latest data for March (48.2) is significantly worse than the official PMI (50,2).

"We provide an independent view on the Chinese economy on a completely new methodology, - says Spaceknow’s CEO Pavel Machalek. - We do not conduct surveys, our index is automated and completely objective." Machalek acknowledged that he had never been to China, but it is not important for the analysis since it’s done automatically. It is understandable why the company has chosen China. Firstly, the country is famous for its unreliable statistics, and secondly, the state of the economy is very important for the whole world. 
Spaceknow is one of a whole galaxy of companies engaged in the data’s analysis in relation to geography (geolocation). The company took advantage of the sharp reduction in price of start-ups and, at the same time, miniaturization of commercial satellites. Three years ago, launch of a CubeSat-like satellite (a cube with edge of 10 cm and weight of a little more than a kilogram) into orbit was worth about $ 1 million. Now, it can be done for less than $ 100 thousand. It's all in the price war, started by Elon Musk (up to 2013th, the commercial launch market had been dominated by European Arianespace and the Russian-American International Launch Services). Spaceknow does not have its own satellites, but uses a sharp decline in the prices of their launch - vendors of the pictures are selling them increasingly cheaper and even offer photographs free of charge.

A breakthrough even more important is the growth of computing power and development of data processing algorithms. These two made interpretation of huge amounts of data (Big Data) possible. It is not enough to simply maintain petabytes of information (usually information is classified as Big Data from this threshold). One need to teach a computer to isolate relevant information from an array, and interpret it correctly. 

Geolocation and Big Data are applicable not only in macroeconomic statistics. Recently, the laboratory of the Chinese search engine Baidu Big Data Lab tried to figure out how many "ghost towns" are in China. People often think that there is a lot of them, yet the assessment of the vacant buildings differ greatly: from 720 million to 6 billion square meters. Using signals from mobile phones and devices with a GPS-receivers (several billion locate points per day), Baidu Big Data Lab has monitored Chinese cities from September 2014 to April 2015. The results were published in part - probably so as not to frighten the public. Only 20 randomly towns selected from the top-50 "ghost towns" got in Baidu’s list. All of them had specific areas of empty buildings. The publication said the company did not want to influence the sale of housing in these "ghosts", so it had not ranked them.

It is still not quite clear whether the Big Data technologies and geolocation will press up questionnaires methods in statistics, traditional ways to solve logistical problems and assessment of many processes in the economy. However, we can already see that these technologies are moving to analysis of the country or even global level. The revolution is just beginning. 

source: fortune.com