Statistics: Seeking truth from (tonnes of) facts
If in nothing else, China is extremely rich in data. Each year, agencies under the National Bureau of Statistics (NBS) and the statistical departments of line ministries collate a vast array of figures on matters ranging from the average distance of road freight journeys (58 kilometres) to the number of notarised documents authenticating photocopies (12,526 in 1999, apparently, representing 0.2% of all notarised documents). An increasing proportion of this information is published. Government statisticians speak of a steady transition from gathering data 'administratively' to reliance on modern survey methods; and suggest that an equally fundamental shift is taking place in their own role - from state functionaries who serve (and inform) only the state to public servants who serve and inform the wider public. But some critics claim that in important areas the data is seriously distorted -- either by deliberate manipulation or simply because the system is not yet capable of producing accurate figures. Other observers point to data gaps and raise concerns about the quality of analysis and interpretation.
Notoriously, China under Maoism substituted political fantasy for statistical diligence: no claim was too outlandish to pass muster, and many careers were built on wild exaggerations of agricultural and industrial output, which further undermined the capacity of central planners to foresee or respond to the famine of the early 1960s. During the Cultural Revolution, government officials now admit, data collection and analysis virtually ceased altogether, and some reports suggest that a great deal of archived material was burnt. The re-cranking of the state statistical bureaucracy during the 'reform' period starting in the late 1970s was a return to historical form - for China, of course, has a history of detailed and meticulous record keeping stretching back many centuries before the compilation of the Domesday Book - but it also faced the new challenge of building a system that would be appropriate to the demands of a 'socialist market economy.'
One aspect of this is dissemination of data to inform a wider circle of decision makers than just the top government leadership, and the last decade has certainly seen explosive growth in the volume of material published. The 1984 Statistical Yearbook, containing basic information on national accounts, production, population, employment, income distribution and livelihoods, health, education and social welfare weighed in at 586 grammes. The 2000 yearbook, running to 890 printed pages, weighs 2.26 kilos, and is only the tip of the published iceberg. Last year saw also the publication for the first time of a Social and Economic Statistical Outline (1115 pp), a Rural Township and Small Towns Statistical Outline (712 pp), Western Regions Rural Statistical Materials (207 pp), a Rural Statistical Yearbook (480 pp), Rural Household Survey Report (252 pp) and Rural Poverty Supervision Report (142 pp) - with a combined weight of nine kilos. For those who cannot wait on annual publishing cycles, the National Bureau of Statistics now also publishes China Monthly Economic Indicators (132 pp), containing much the same kind of information that international financial newsletters sell at many times the NBS asking price. Since the early 1980s, provincial statistical bureaus have also started publishing their own Provincial Statistical Yearbooks; and some counties publish County Yearbooks (although, particularly in poorer areas, this information often remains 'internal'). Finally, government line ministries also produce yearbooks reporting on their own domains, from Education to Industry. Altogether, therefore, the statistical regime's published output could probably be measured in tonnes. To lighten the load on bookshelves, the NBS is beginning to make information databases available on-line (www.stats.gov.cn); but with such a plethora of material to manage the project shows early and familiar signs of having to run ever faster just to stand still.
So much for the quantity of numbers. Where do they come from, and what of their quality?
In answer to the first question, the state remains not only the main source of information put into the public realm, but also at least nominally in control of all information gathering. Procedures have, however, been introduced to licence private research firms to conduct surveys (with special stipulations for 'foreign related' research); and minjian zuzhi ('popular organisations' - generally used to denote government initiated 'NGOs') have also been cautiously encouraged to undertake their own research, under the 'guidance' of relevant authorities. Line ministries may collect information within their jurisdiction, but are required to notify the NBS, and lodge copies of all findings. Surveys that cut across ministerial boundaries are supposed to be submitted to the NBS for approval. Thus, all statistical research in China is expected to be either directly implemented, 'guided' or at least authorised by government. The bald phrase genju tongji ('according to statistics'), liberally sprinkled in newspaper reports and official documents, is still generally taken as synonymous with 'according to official statistics.' All 'facts' that are to be regarded as such must be endorsed by government.
Senior NBS staff complain, however, that the public and mass media use statistics sloppily and very frequently fail to quote the source. This in itself suggests recognition that sources of public information are becoming more diverse, as market research companies, statistical departments of large enterprises, parastatal organisations such as the high-profile Consumers Association, and some more autonomous organisations begin to compile and publish their own data. No Chinese agency would yet openly challenge figures that the government has published but, as information sources diversify, the realm of statistical fact is likely to become increasingly contested.
Independent interpretation of official data is already often trenchant. For example, writing on the website version of Tianya magazine (www.tianya.com.cn), Chinese Academy of Social Sciences researcher, Yang Fan, argues that official figures showing steadily increasing rural incomes are largely illusory, because these do not represent cash gains but are locked up in stored grain (which farmers cannot sell) and agricultural machinery. Over the last twelve years, Mr. Yang calculates, the real per capita income of 600 million rural people has risen by an average of only CNY 7 (USD 85 cents) per year. Official figures for the period show per capita rural income rises in the range of CNY 50 - 350 each year.
Chinese academic researchers still generally find it difficult to conduct independent quantitative research and rely heavily on analysis of official data. But, despite the proliferation of published data, scholars do not always have access to information that they want. NBS household survey teams, for example, undertake extensive, continuous surveys of sample households, and these are widely thought to provide high quality, accurate data. A rural survey team oversees a sample of 68,000 households in 7,000 villages of 857 counties. The sample households (which are rotated every four years) keep daily accounts of all their income and expenditures, down to every bar of soap and packet of salt, and this is reported monthly to trained surveyors. An urban survey team conducts a parallel exercise with 36,000 households in 226 cities. These ongoing exercises generate a huge amount of data, only a fraction of which is published. Chinese scholars who want to look at the raw data to draw their own conclusions may, however, negotiate with the survey teams for private access to the information. Seasoned Chinese social scientists have told chinabrief that it is quite common for researchers to pay the NBS for releasing information to them
Turning to the quality of official data, there is little doubt that many of the approximately 70,000 staff employed in the NBS hierarchy are dedicated professionals committed to improving the accuracy and relevance of data they generate. The work undertaken by the urban and rural survey teams are one means of doing so; and the statistical system as a whole is moving towards greater reliance on modern sampling and survey methods, and away from patently flawed routine administrative reporting systems.
'Adminstrative reporting' is the process whereby officials in individual ministerial hierarchies are charged with collecting and passing data up to higher levels. Its inherent weakness is that at all levels the officials responsible for collecting the data will inevitably tend to see the figures they present as management performance indicators: if, for example, they report a rise in productivity, or in the number of children attending school, this will reflect well on them. This leads to a phenomenon known in China as 'adding water'. Township officials may massage the figures a bit, and these are then adjusted again by the county authorities, and so on upwards through the command chain. Central government has repeatedly denounced this practice, and a 1996 amendment to the 1983 Statistics Law expanded provisions to prevent and punish fraudulent reporting or alteration of statistics, including an article explicitly prohibiting government leaders from taking reprisals against statisticians who refuse to falsify figures at their behest. The emphasis given to the issue, and the need for an amendment to underline it, is probably indicative of the extent of the problem. However, government's propensity for issuing high profile targets, and the political weight that attaches to these, must significantly increase the pressure on officials to report successes.
In many fields, figures issued by the NBS remain wholly or largely based on administrative reporting supplemented by surveys undertaken either by NBS teams or by line ministries (often with NBS technical support). Although government capacity for designing and carrying out modern statistical surveys is generally considered to be advancing rapidly (and is already substantially greater than that of many other large, developing countries), surveys can nonetheless vary widely in quality and their use is not itself a universal panacea for inaccuracy. A health sector specialist from one international organisation points out, for example, that survey results can be influenced by perceptions of what the survey is for. If local officials know that a survey is a precursor to an international aid project, they may be tempted to distort the results in a way that exaggerates the problem in their area in order to improve their chances of inclusion in the project. Surveys that are not tied to project activities may, on the other hand, again be seen as management performance indicators, and create pressure for unduly positive reporting. 'Comprehensive' surveys using very large samples are most prone to these kinds of distortion.
For the NBS, pursuit of accuracy and objectivity are relatively straightforward institutional objectives, but line ministries and local governments may be influenced by other considerations. The designation of official 'poverty counties' is a case in point. As this brings access to special central and provincial government funding, candidate areas naturally have a vested interest in qualifying. In poor areas, officials often seem uncomfortably torn between projecting their success in meeting poverty alleviation targets and emphasising that they are still needy. (In chinabrief's experience there is often an uncanny resemblance between quoted numbers of remaining poor in any given county, and the latest pronouncement from central government about the proportion of poor nationwide. It would be surprising if the rate of progress against poverty were even throughout all areas of the country). Line ministries are often essentially in a very similar position: anxious to demonstrate competence in meeting economic and social development objectives on the one hand, but also needing to advance their claims for budgetary allocations by showing areas of need that remain to be addressed.
The NBS is thus in the extraordinarily difficult position of having to make an independent assessment of the reliability of data coming from different sources (which may involve reconciling conflicting figures, as in the - probably unusually extreme - case of the Qinghai population statistics presented in the supporting article linked to page 1). This raises issues not only of the accuracy of the numbers, but also of what is counted in the first place, and how it is defined.
Measuring unemployment is a telling example. According to a ministry profile published on the Chicago based www.chinaonline.com, during the 1990s the NBS tried to introduce a way of measuring unemployment that is consistent with International Labour Organisation definition, but the effort was blocked by the Ministry of Labour, which was presumably concerned to conceal, or at least soften with euphemisms, the full extent of the problem.
A dramatic example of the way that different measurement techniques can affect results was provided after the World Resources Institute published its 1995 report Who Will Feed China?, which argued that shortage of arable land would result in grain deficits that world markets could not meet. China responded with new figures for its area of farmed land, based on satellite imaging, which showed a substantially larger cultivated area than originally reported (132 million hecatres, as opposed to 95 million) - showing that per hectare productivity was much lower than originally implied, and therefore that there was greater potential for future productivity gains. (This debate is now further complicated by the policy of 'ecological construction' and the reforestation of farmed land.)
International organisations working in China raise many other important issues of definition and measurement in the compilation of statistics, ranging from the way that non-performing bank loans are represented (in a way, at present, that is thought to considerably understate the amount of unrecoverable debt) to the way that 'urban' and 'rural' populations are classified - and the closely related issue of how 'floating' migrant populations are treated statistically.
At the same time, a significant amount of international effort is expended in encouraging ministries to develop new indicators that might provide greater insight into their areas of responsibility. For example, UNESCO and a consortium of donor agencies have been working for several years with the Ministry of Education on the development of 'quality indicators' for basic education, to measure not just the numbers of children entering or completing school but also their learning achievement while they are there. Unicef has worked with the NBS urban and rural survey teams to incorporate data about children in the information the teams gather.
The government of Sweden has recently completed a project, implemented with the Swedish national statistical agency, to support the development of a gender perspective in Chinese statistical work. Much of the copious data gathered by government remains gender blind, with information on key areas such as employment, education and health rarely disaggregated by sex. The Swedish funded project, building on previous cooperation between the NBS and Unifem, began to address this by helping to specify and compile statistics gathered by NBS and inter ministerial National Working Committee for Children and Women for a monitoring report on the 1995-2000 National Plan of Action for the Development of Women. This involved the selection of more than 100 indicators, covering political participation, education, employment, health and legal protection, as the basis for nationwide data collection with appropriate training at provincial and sub-provincial level. To generalise the engendering of statistics throughout China's statistical system would doubtless require many years, but this project appears to have represented a significant step in that direction.
Meanwhile, with so many complex issues of measurement and interpretation, assessing the accuracy of Chinese government statistics is no easy task. A great deal of useful and almost certainly reliable data is contained in the fine print, particularly that generated by the urban and rural household survey teams. (Figures on average household income and expenditures are widely credited, as are as those for per capita ownership of bicycles, sewing machines, TVs etc). And a great deal can be learned from the details: tracking township hospital occupancy rates, for example, may reveal more about the state of health services than mainstream health indicators. The more politically sensitive, headline figures and social and economic indicators probably need to be treated with more caution. These include statistics on poverty, unemployment, birth and mortality rates and the most widely quoted indicator of all - gross domestic product growth.
An article by University of Pittsburgh economist, Tom Rawski, in the current China Economic Quarterly (Volume 5, Issue 1), not only raises doubts about the 7-8% growth figures posted over the last two years, but goes as far as to suggest that the economy might even have contracted over this period.
Since the early 1990s, the NBS has been trying to bring the country's system of national accounting in line with international standards set by the United Nations. (The task has not been made any easier by the fact that the UN recently changed its guidelines). Professor Rawski accepts that the NBS is trying to do as good a job as it can, but believes that since 1998, with an economic downturn precipitated by the South East Asian financial crisis, statisticians have once again come under personal pressure to cook the books, in order to gratify the leadership's demands for more growth at all costs.
He supports this conclusion by citing political drives for higher growth and published remarks of statisticians reporting pressures upon them, but also by highlighting apparent inconsistencies in a range of official data. Notably, energy consumption fell substantially in 1998 and 1999, which is unlikely to be consistent with increased output; industrial output allegedly rose despite many major producers reporting declining outputs; steel consumption and cement rose much more modestly than the alleged rise in investment spending, and a reported 1998 increase in farm output looked highly unlikely in view of the disastrous floods of that year.
Professor Rawski also believes the government's financial stimulus packages may have accounted for one or two growth percentage points. In sum, 'A plausible guess would place real GDP growth for 1998 between -2% and +2%'; and 1999 could have been even worse. What's more, 'A return to statistical normalcy may require convergence of actual economic growth with the "politically correct" levels of 7% to 8%.'
Whether or not Professor Rawksi's arguments are valid, international investors clearly have a strong interest in 'statistical normalcy' in China, and a great deal of international effort has been invested in various kinds of capacity building programme. In addition to those mentioned above, ongoing technical assistance, exchange and training programmes have been funded by Canada (with an emphasis on management issues, enhancement of national accounting and infrastructure); France (with a recent emphasis on establishing a statistical business register, improving retail trade statistics and economic forecasting and modelling); Germany (with a recent emphasis on industrial and transport statistics); Italy (technical assistance and equipment for the 1997 agricultural census); Japan (training courses at the Japanese government Statistical Training Institute); the Republic of Korea (mainly bilateral meetings and ad-hoc exchange), and the European Community (with an emphasis on China-EU trade statistics, labour markets and the informal economy). The Asian Development Bank, International Monetary Fund and World Bank have also been involved in numerous technical assistance projects and advisory missions.
There is no doubt that greater global engagement and integration, including membership of international organisations with requirements to exchange relevant statistical information, greatly increases both the demand on China for reliable data and its access to international reporting standards. But achieving standards of reliability and accuracy is ultimately a domestic affair. In part this is dependent on the political will of the leadership, but it may also in part depend on demand from domestic constituencies for greater freedom of information. Increased access to non-government sources of information, including overseas sources, and the release by government of greater quantities of information over the last few years, on subjects ranging from air quality to women's literacy, may have served to stoke debate and feed that demand.
It is no longer particularly surprising, for example, to find an op-ed column in the official China Daily (Government Should Be More Open, 28/03/01) calling for 'more access to government information' in order to 'help further the development of the economy and trade, better protect consumers' rights and improve government functions' - not to mention it being 'a basic obligation of the government in order to guarantee citizens' right to know.' Quite so.
About the article:
China Monthly Economic Indicators can be ordered from firstname.lastname@example.org (within China), email@example.com (Hong Kong) or firstname.lastname@example.org (USA).
Professor Tom Rawski's website contains several articles on China's statistical system: