Sunday, July 17, 2016

Rickety Livestock Statistics to be Improved

China's Ministry of Agriculture is trying to improve statistics on livestock and poultry. These efforts open a window into the crevices of China's rickety statistical system that is largely untouched by 20th century developments like stratified sampling, weighting, and standard errors. Most outsiders think China's statistical problems are simply a matter of lying by statistical bureaus, but the problems are deep-seated and systemic.

Last month, a training meeting was held in Liaoning Province for personnel from key counties in northern Provinces to teach them how to collect statistics on livestock and poultry, and to ensure that they all use standardized procedures. In his remarks at the meeting, the vice-director of the national livestock station said work on livestock statistical monitoring needs greater attention from leaders, the capacity of personnel needs to be improved, and the work needs to be better-funded. He emphasized that township-level personnel needed the most attention. Attendees heard lectures and were taken to a model county to observe the best practices for collecting, recording, and reporting data on livestock.
Statisticians instruct a village clerk in Shaanxi Province's Zhouzhi County 
how to help farmers China Livestock and Veterinary News
Like many "statistics" in China, livestock statistics are generated by an old-fashioned administrative reporting system that rely on multiple layers of poorly-educated and poorly-paid clerks collecting and reporting reams of numbers they rarely understand. A Ministry of Agriculture document describes the elaborate system for collecting livestock monitoring data that was set up in response to a decree by the State Council in 2007. The system includes annual and quarterly reports of provincial livestock and feed industry data from all provinces; monthly reports from key hog-producing counties; weekly price reports from markets; and another set of monthly reports from a group of villages, companies, and farms that report production, sales, and production cost information. Reporting counties were chosen by ranking them using data from 2006. The document gives general instructions for choosing samples. In villages selected to be hog monitoring sites, officials are to choose one big farm, one medium farm, and a small farm to collect data from.

According to the document, village officials (typically veterinary technicians) collect the numbers and pass them up to their township office who compile the data and report it to the county. The annual and quarterly complete reports are passed up to the province and then to the national office in Beijing. Other monitoring data are transmitted by Internet directly to the Ministry of Agriculture's central database.

An article in a livestock-oriented newspaper last month said the livestock reporting system has "a few urgent problems" that need to be addressed. The problems start at the village level where the task of filling in the report forms is typically an additional task given to the village veterinary technician, the village accountant, or the person in charge of monitoring laborers or settling family disputes. Reporting statistics is not a high priority task for these folks, and they seldom have any understanding of statistical concepts nor of how the statistics are used. The village reporter seldom has face-to-face contacts with livestock producers.  The report forms are too complex and not understood by the person answering the questionnaire nor by the person administering it. Large scale farms pose a new difficulty. The large numbers of animals or poultry make it impossible to easily verify the numbers reported. Biosecurity measures forbid outsiders from entering livestock farms, so there is no way to verify numbers given to survey-takers.

The article recommends simplifying report forms, holding regular training for statistical workers, and making greater efforts to verify and check numbers. The writer thinks that statistical clerks will be less likely to falsify data, make errors, or omit numbers if they are aware of the whole process and the importance of the statistics.

A 2013 article points out similar problems, observing that the system is essentially a census that can't possibly collect accurate information. The area that each village reporter is responsible to is too broad for him or her to conduct face-to-face interviews. The author of the 2013 article gave other reasons why survey reports may not be truthful and accurate. He said livestock report forms are often filled out based on village ledgers, subjective impressions, or phone calls. Most livestock farmers have no production records. The number of animals is typically estimated and farmers may "hide" sales. Some data items are not fully understood.

The livestock reports are filed with township statistical offices which are responsible for all kinds of statistics. The township clerks probably have little or no statistical knowledge of livestock and poultry. They fail to catch inconsistencies or implausible numbers. The writer says there should be more careful auditing of numbers to ensure that carcass weights make sense, that ratios of eggs to chickens and milk to cows are reasonable, and that changes in numbers from the previous year are realistic. He also points out that statisticians with no business experience or sense of responsibility are prone to mistakes.

Last month's training session for county statisticians appears to be designed to remedy some of these problems. The 60 county livestock reporters attending were informed of the overall statistical process, and they visited a model county in Liaoning to learn about its village-township-county reporting process. They learned how to help farmers record expenses and sales, and how to maintain a record system.

Another training session for 50 county livestock statisticians in southern provinces was held in Sichuan Province during May. This session seems to have been more advanced, with emphasis on funding, management, and analysis of statistical data. A number of counties have held training sessions.

Chinese officials have high hopes that "big data" can help them monitor markets in real time, allowing them to achieve their dream of engineering stable markets. The Ministry of Agriculture is planning a new--and equally elaborate--system for collecting livestock industry statistics directly using "big data."  The Ministry plans to set up a pilot system that will collect data from all points in the supply chain: urban and rural consumer markets, storage, transportation, slaughter, and farms. The Ministry will distribute electronic monitoring equipment that will transmit data to a system that will generate automated "early warning" reports, alerting authorities when to intervene in the market. No details are available--this appears to be still on the drawing board and it will take three years to get pilots going. "Big data" presents an interesting prospect but one wonders whether it will be immune to human error and falsification. Couldn't a slaughterhouse find a way to inflate its numbers if it gets subsidies--or turn the machine off when authorities tell them to shut down? It will be a while before big data supplants 20th-century statistical sampling.

In the meantime, all Chinese "statistics" should be taken as suggestions, not precise measures of reality. Various statistics should be cross-checked against one another to see whether they make sense. If Chinese authorities were brave enough to report more of the statistical items they collect to the public in a transparent manner, there would be a greater chance that market analysts and academics could come up with solutions and ideas for improvement in statistics.

1 comment:

Godfree Roberts said...

"China's rickety statistical system". Rickety? Really?
We've got 50 years of rickety statistics that look more conservative every day. A detailed analysis of China's GDP indicates its economy is actually 15% larger than official figures:

Anecdotally, we have Marc Mobius, CEO of the Franklin Templeton Fund:   "I know there's a lot of debate as to whether the numbers are true, whether it's really 6 percent or 7 percent, but our numbers indicate that it is at least that. We think that a lot of the economy is not really being counted because China is being converted from a manufacturing-oriented economy to a service economy."

Academically we have the Federal Reserve and Carsten Holz, dean of Chinese Government statistics studies: The quality of China's GDP statistics. Carsten A. HOLZ

On the regulatory front we have "On the Reliability of Chinese Output Figures". John Fernald, Israel Malkin, and Mark Spiegel, FEDERAL RESERVE BANK OF SAN FRANCISCO