Saturday, December 31, 2011

Fighting Falsification of Statistics

China Economic and Industry News reports that China's National Bureau of Statistics (NBS) is setting up a new direct reporting system designed to eliminate the chronic problem of padding statistics by local officials. The announced reform is a needed and welcome change, but it shines a light on fundamental problems in the Chinese statistical system and makes one wonder if there is any value in the reams of numbers published by Beijing every year.

NBS is introducing a new direct statistical reporting system designed to cut out the opportunities for intermediary organizations to falsify data in the traditional bottom-up reporting system. In the new system companies will report data directly to a national statistical data center or to a provincial data center accredited by the national statistics bureau.

Most Chinese statistics are based on a bottom-up reporting system. Local statistical departments collect data from companies, compile the data and report them up to higher-level departments. As numbers are passed up from level to level they tend to be inflated (in the case of GDP) or otherwise altered. It is well-known that GDP estimates reported by NBS in Beijing are routinely adjusted downward. It is common for the national GDP growth rate to be lower than all the provincial estimates.

The article explains that local officials have an incentive to pad the statistics since they are judged on the basis of economic measures like GDP growth. Local officials provide support to certain industries and therefore have incentive to manipulate statistics to show that their strategic plans are succeeding. Consequently, officials routinely manipulate data as they are passed from one level to the next. Statisticians in Beijing have long, contentious meetings to decide what to do with the false data sent to them by the provinces. Obviously, the statistics reported by Beijing are not objective data, but rather manufactured numbers that emerge from countless excel file manipulations and discussions around long wooden tables in Beijing.

The article says that people view statistics as tools of government officials to be used to trumpet their successes at meetings. The inaccuracy of statistics makes it difficult to make accurate forecasts and undermines the credibility of statistical departments.

The new direct system relies on direct reporting of data by 600,000 companies. Participating companies will be companies in manufacturing and services industries with output or sales exceeding a certain threshold and construction companies recognized by the government. The big companies are said to create most of the value and the most important effects on the economy. The direct reporting system will not include the millions of small companies, entrepreneurs and vendors. The article doesn't explain how these small operators will be measured and tracked.

According to the Economic and Industry News article, most economic and industry statistics are produced by bottom-up reporting systems. NBS in Beijing only directly estimates data on commodity prices and household income.

There has been a similar shift away from bottom-up reporting in agricultural statistics. The first agricultural census conducted in the 1990s was the first time a national unified count of farmers, land, animals was conducted. The results confirmed that statistics overstated the number of livestock and understated the amount of farmland by a substantial margin. Since then, NBS has been switching to survey-based approaches to estimating crop and livestock output. The current system is a hybrid of surveys and reporting systems. Village- and farm-level surveys of grain yields, changes in land use, and hog inventories and sales in major production areas are combined with a bottom-up reporting from minor production areas, state farms, and companies. National estimates are made by expanding survey estimates based on census data.

However, even the survey sampling may be subject to local influence and lack of expertise. Surveys are carried out by teams of local statisticians and officials. Who watches them to ensure they don't pick the "best" villages and farmers for the sample? Do they take samples from random plots in the fields, or from plots known to give the highest yields? Suppose a plot chosen for the sample is decimated by drought in a given year. Do enumerators record a "0" for the output from that plot or do they substitute another plot that produced a "reasonable" yield?

Ma Jiantang, the current director of the Statistics Bureau, seems intent on improving statistics. (See a report on agricultural statistics from earlier this year.) However, the problems with the Chinese statistical system are fundamental and deeply entrenched. Statistics are still politicized at all levels. At the national statisticians' meeting where the new reforms were announced, the statisticians were ordered to implement the State Council's instructions on statistical work and the spirit of the 3rd, 4th, 5th, and 6th plenums of the 17th Party Congress and economic work meeting. The meeting was attended by leaders of the communist party's Central Discipline Commission, the Party Organization Department, the Supervision Department, the Audit Department and the Liberation Army's Strategic Planning Department.

Statistics have always been manipulated in China. The modern discipline of statistics relies on completely random selection of samples and use of consistent rules for estimates. The interference of either intentional manipulation or well-intentioned judgment calls by enumerators and reporters can undermine the process.

For historical perspective, we post here some comments on Chinese population statistics from A Cycle in Cathay, a book written in 1895 by W.A.P. Martin, an American who lived in China for 60 years in the 1800s. Martin's brief two-page report reveals practices and manipulation of statistics in the 1800s quite similar to those used today.

Martin (p. 460): "Owing to imperfection in their mode of enumeration, strict accuracy is not to be expected. A governor of a province will sometimes add what he supposes to be the probable increment to an old census, instead of taking the trouble to make a new one."

Martin goes on to cite an instance when Chinese authorities reported to western foreign ministers that China's population was 215 million when it was believed to be closer to 400 million. According to a Dr. Dudgeon quoted by Martin, the Chinese officials deliberately reduced the population figures by one-third. The purpose of deflating the figure was to discourage foreign countries from sending missionaries. According to Martin, the following year Chinese authorities, having seen "no abatement of missionary immigration," added the missing one-third back into the population figures they reported.

No comments: