Monday, October 10, 2022

"Big Data" and Human Element in China's Livestock Statistics

China's National Bureau of Statistics blames its data problems on conspiratorial fraud between the local officials and companies that report data to Beijing. In a May 2020 campaign to crack down on fraudulent statistics an NBS official complained that fraud and fake reporting of data remained rampant in some regions and warned that inspectors are being sent out to local government offices and companies to catch the fraudsters. This was a continuation of a crackdown on statistical fraud and deceit ordered by Xi Jinping in 2021.

In July 2022 statistical inspectors were sent to China's agriculture ministry to root out fraud and falsification, punish perpetrators, and demand that all clerks and accountants take seriously the importance of quality statistics. This follows a visit last year. In August, an inspection team conducted a spot check of livestock statistical reporting in Jilin Province where they checked farm records, investigated the reporting system, and lectured local officials about their responsibilities. 

Statistical inspectors arrive at agriculture ministry. Shall I wear a white shirt or a blue one?

Statisticians are hopeful that developments in "big data" and "artificial intelligence" can cut out the human element in data collection. One example is a recent initiative for companies to set up an automated system to send data directly to Beijing, thus eliminating opportunities for local officials to massage the numbers. 

China's chaotic livestock industry is a longstanding statistical sore spot and a target of "big data" statistical automation. The Jilin livestock statistics investigators got a report on the "Jilin's cow cloud" --a key project--and checked in with a local livestock tech company. Another team visited Shandong Province to catch up on livestock "informatization" efforts, including a demo of the “Shandong province smart livestock industry big data platform” and "big data" capabilities. Jiangsu Province officials held a training seminar for managers of milk collection stations to automate reporting of milk marketing data.

Statistical inspectors lecture a livestock farmer in Jilin Province.
Source: National Animal Husbandry Station.

These "big data" efforts sound great, but progress is still undermined by the human element. Six years ago China's agriculture ministry kicked off a pilot program for an elaborate "consumption-guided whole-industry-chain hog monitoring" system that would take advantage of "big data", cloud computing, and internet-connected devices to collect real-time data from farm records, slaughterhouses, wholesale and retail markets and pork consumers. One of the "scientific" system's stated principles was to "reduce human interference." The system was expected to improve the accuracy, timeliness, and usefulness of pork industry statistics. Officials would use "deep data mining" to identify leading indicators to issue "early warnings" to help business entities make production plans and allow government officials to surgically intervene in the market to smooth out price gyrations. 

The hog statistics pilot launched in 2016 was expected to build the foundation for an improved swine statistical system within three years. Two years later--in 2018--an African swine fever epidemic swept through the country, wiping out at least a quarter of the national herd, severely crimping pork supplies. In early 2019 agricultural officials predicted that pork prices might rise 55 percent. Prices actually went up 120 percent later that year. At the peak of the pork shortage crisis, officials mostly stopped issuing pork statistics altogether as pork prices soared to unprecedented levels.  

Even pig statistics are inherently political in China. In August 2019, Vice Premier Hu Chunhua had ordered officials at all levels to restore pork supplies asap as a "political task." Thus, local and provincial officials had great incentive to hide data that highlighted pork shortages while inflating their pig production statistics to win plaudits from their superiors and demonstrate effects of massive subsidies. Provinces competed to outdo one another in showing big rebounds in hog numbers.

The hog statistics pilot promised to be "consumption driven," including special attention to rural pork  consumption using a 30-year-old agriculture ministry village monitoring survey. However, there has absolutely no discussion or analysis of pork consumption by government, academics or news media during the African swine fever crisis. Such discussion was probably banned to avoid embarrassment since Xi Jinping had announced China would reach "all-round relatively well-off society" status in 2020 and the communist party's centenary was in 2021. Statistics on rural pork consumption would have been especially embarrassing since anecdotal reports indicated that rural consumption plunged while urban consumption took a softer hit during the African swine fever-induced shortage. In 2019, Vice Premier Hu ordered news media to "shape public opinion" on the pork shortage, and Peoples Daily quoted the agriculture ministry as insisting that "the pork supply is assured" at the same time Hu was ordering officials to alleviate the shortage.

Puzzling pork consumption data buried in China Statistical Yearbook.

The hog statistics pilot also aspired to incentivize company participation in the data collection effort by offering "services" to companies, presumably access to statistical reports and analysis. However, there are still no statistical reports on the pork industry that would have any value to companies. Annual and quarterly reports--issued by companies for the benefit of investors--convey far more market information than do government reports. 

We are collecting "big data" inside these cardboard models of giant factories.

Instructions given by a team leader after last month's Shandong livestock informatization inspection reveal some problems in vacuuming up data from various private companies. The team leader ordered local officials to pay close attention to Internet security; to speed up data sharing and eliminate "information silos"; and to devise standardized codes for data items from various companies. 

The Internet security order was probably inspired by July's massive leak of data from a Shanghai Police web site. If government data sites are vulnerable to hackers, companies are not going to "share" their data with statisticians. The economic motivation for companies to compile and analyze "big data" is to gain an advantage over their competitors. Why would companies share valuable data that their competitors might exploit? Wouldn't statisticians in possession of proprietary company data be motivated to engage in insider trading or to accept bribes? Why would it be advantageous for slow-footed government bureaucrats to dictate data standards to fast-moving companies looking to profit from their proprietary data?

Echoing the 2016 hog statistics pilot, last month the leader of the Shandong inspector team ordered local cadres to "develop informatization service capability." In other words, government statisticians in China still have nothing of value to offer companies in exchange for the data they want them to share.

A "big data" platform demo in Shandong Province. Perhaps she's saying, "These pretty pictures based on garbage-in-garbage-out data will definitely upgrade China's livestock industry." Source: National Agricultural Husbandry Station.



No comments:

Post a Comment