At the Fourth General Meeting of the Asian Historical Statistics Project last October, the special session entitled, " Government Statistics and Statistical Organizations " (Part 1) elicited an enthusiastic response from the majority of the participants. The rationale for organizing the session was to facilitate the production of brief overviews of the statistical systems and survey organizations of each country under study. These overviews would be attached as either introductions or appendixes to each of the separate country-specific volumes to be published at the end of the project in order to enhance the understanding of the characteristics and reliability of statistics contained in each of the volumes.
While the session raised numerous issues common to all countries, it also introduced facts which were common knowledge to the specialists of a given area but came as a surprise and discovery for experts of other regions. For example, it was pointed out that in the cases of Thailand and the Philippines, the respective statistical systems were characterized by inadequate sampling design and repeated changes in concepts -a problem common to many other countries. However, it was a surprise for many participants to learn that the statistics of Egypt, a British colony, were heavily influenced by the French system, or that A. L. Bowley, an economist renowned for his Box-diagram analysis, visited India as a statistician to confirm the problems on the spot in order to provide concrete advice on India's statistical system and surveys.
Consequently, after the general conference, many voiced the opinion that a project for a thorough comparative study of the statistical systems of Asian countries was necessary. This past April, the Executive Committee of the Asian Historical Statistics Project decided that among the task groups, one devoted to the study of this particular subject would be launched this year.
With the inauguration of the statistics systems research group, it behooves us to highlight the significance of the group and put forward some objectives so as to establish some common perspectives. In what follows, I offer my views on the research subject and what methods and approaches might be applied.
As implied by the Latin word " status " - the origin of the term " statistics "- statistics had been historically interpreted as a term referring to the research and investigation of the conditions of the state. Nowadays, the interpretation has been broadened, so that statistics are defined as numerical aggregate information about the social infrastructure and the relationship among its components, or as a discipline dealing with such information. However, what has remained unchanged is the central role occupied by states and governments in statistics.
Therefore, it would not be an exaggeration to state that statistics are collected, compiled, and utilized primarily by governments. Historically, statistics began with the compilation of population censuses and trade figures, but along with the expansion and development of government functions, moved on to statistics on agriculture, land taxes, transportation, and education. The same development pattern applied to colonies, since the use of statistics was necessary for the suzerain to rule the colony.
In general, these government statistics were prepared to serve administrative purposes. However, as statistical information became increasingly voluminous and complex, government agencies specializing in statistical work were established. Furthermore, independent statistical organizations were set up to coordinate statistical projects of various ministries and agencies, and to conduct large-scale surveys of their own.
Those statistics that were not covered by the government statistics were supplemented by statistical collection conducted by private firms, affiliated organizations, and / or researchers. Statistics on insurance, railroad passenger traffic, and labor unions represent early examples of this type of statistics.
In addition, the so-called social surveys-surveys of social issues that require immediate treatment-conducted by researchers and social reformers, constituted another important category of survey statistics. Among the most famous of these figures were C. Booth and S. B. Rowntree of Britain, and P. Kellog of the United States. Subsequently, the coverage area of these surveys have expanded rapidly, while becoming vastly more refined in methodology and theory.
Nonetheless, looking at the overall situation for statistics within a nation-state, it still remains a fact that the government statistics in a broad sense have played the dominant role. Particularly after World War I, the importance of statistical information was reaffirmed, and various countries laid out plans to expand and consolidate statistical organizations. Of course, the advent of the Soviet Union and the critical importance of economic statistics in planned economies undeniably exerted great influence in fueling this trend. Moreover, in the 1930's, the theoretical foundations for probability sampling theory were laid down, leading to large-scale national sample surveys by governments.
By around 1940, developed countries realized large scale improvements in their government statistics, not only in population censuses but also national income statistics as well as more detailed sectoral production figures. However, data accumulation methods and the nature and extent of authority possessed by statistical organizations varied greatly from one country to another. These variations stemmed from differences in historical background, culture, industrial structure, and especially from the structure of the administrative framework (which was inextricably intertwined with the first three factors).
Therefore, in comparing statistical systems across countries, we must at the very least address the following three points: (1) To what extent is data accumulation of various administration statistics decentralized over the relevant authorities, or alternatively, to what degree is it centralized through specialized organizations?; (2) In collecting data, how do cooperative and complementary relationships between the central and local governments work? How do central and local organizations share associated costs and supervisory responsibility?; (3) Are sample surveys designed to complement or substitute administration statistics? What is the relative importance of the surveys to the administration statistics? These core issues ought to be addressed since they are common to all countries, and yet significant variations can be observed between countries. For these reasons, a consideration of these three points enables us to acquire information about the characteristics and reliability of statistical systems, and the efficiency of data collection in each country.
Few would disagree that the historical background and culture of a country significantly affect the way data are collected and statistics are prepared. Such influences are generally observed even in population censuses, which are relatively standardized around the world. For instance, the educational levels of enumerators, the literacy rate of respondents, the degree of understanding of and cooperation toward the survey, the reliability of the geographic census units, and complexity of races, religions and languages would all vary considerably from country to country. Accordingly, survey methods (recruitment of enumerators, entry forms, checkup systems, etc.) differ across countries, leading to huge discrepancies in the accuracy of the results.
In case of the population census, however, as a result of the United Nation's frequent attempts at comparable simultaneous censuses after World War II, the standardization of contents, definitions, and methods of survey has been gradually taking place, making international comparison possible to some degree. However, since even censuses have yet to become fully standardized, direct comparisons of the administration statistics of developing countries, riddled with vast differences in development levels of industrial structure and bureaucratic framework, are even more difficult in terms of conceptualization and accuracy.
It seems likely that a considerable amount of time will be needed to make improvements in these areas, but as a supplementary measure, large-scale sample surveys are being steadily carried out in developing countries. Through these surveys, substantive international comparisons have become feasible to a reasonable degree. Especially for developing countries, given the inadequencies of survey networks, costs, quality and quantity of enumerators, and quickness of reports, there are clearly major advantages to adopting national sample surveys rather than complete enumerating surveys dependent on administration statistics organizations.
Moreover, in developing countries, sample surveys are highly effective in obtaining limited specific information used for formulating economic plans. Even in the course of preparing national income statistics based on the United Nation's recommendations, certain portions of relevant information are necessary to derive from sample surveys. Sample surveys possess high levels of effectiveness and utility; therefore, the trend is toward more extensive applications of sample surveys.
Considering the quality and coverage of administration statistics, sample surveys should be better utilized, while more attention should be paid to their significance and their role when carrying out international comparisons of economic statistics. The reason is the following. In sample surveys, not only can we grasp their accuracy (the reciprocal of the sampling errors) but we can also estimate non-sampling errors by reviewing the survey methods and the quality of enumerators. This greatly expedites cross-country comparisons.
A wide range of information will be very important in international comparison of economic statistics, including information about the budget and organizational structure of government statistical organizations specializing in sample surveys, the educational standard of their staff, and the sample size and the relative importance of sample surveys in statistical activities. Such information as a whole is indicative of the quality of sample surveys of a country, and also indirectly reflects connections with administration statistics and their standards.
Consequently, if we were to conduct a comparative study on the reliability of the economic statistics of Asian countries or on the efficiency of their respective statistical systems, the most effective and efficient strategy would be to first select a representative sample survey common to these countries-for instance, surveys on crop-cutting, household expenditure, labor force, etc., and then proceed with the comparative analysis.
We have pointed out that the accuracy of government administration statistics and large-scale sample surveys, and statistical organizations per se, can vary significantly from country to country. Further, there exists a notable difference in statistical systems, reflecting the historical and cultural facets unique to each country. Thus, we emphasized that in international comparisons, individual circumstances of each statistical system must be recognized.
In other words, only on the basis of a full acknowledgment that a statistical system itself is a part of the social system of a country, can we perform international comparative studies of statistical systems. In order to recognize differences among statistical systems in terms of accuracy and discrepancies, the participant observation of researchers at the statistical survey sites in a given country seems to be the best strategy for obtaining powerful insights and rich implications.
(Hitotsubashi University, Institute of Economic Research)
Translated by Hyung Gu Lynn