Fulfilling Kuznets' Dream

Comments on the Asian Historical Statistics Project

Harry T. Oshima

The papers at the recent Taiwan Historical Statistics Workshop (held in February 1997 at Sano Hall) were very interesting and thought-provoking. I would like to offer a few comments on the project based on the presentations and a number of the articles in your newsletters.

Significance of the Project

Simon Kuznets once pointed out that economic growth is a long-term phenomenon, not just a four or five decade process. It is difficult, for instance, to understand post-World War II economic performance unless one knows what happened in the previous half century or so. Thus without data on GDP and other economic factors for the prewar decades, one does not know whether East Asia's growth rates will continue into the next century. What were the causes and conditions underlying the pre-World War II era performance and how do they compare with those of the postwar era's growth performance? If these causes had to do with institutions, technology, capital formation, and labor force behavior in the postwar period, what differences were there from the prewar decades? Was it political forces in the prewar Philippines which would account for the country's comparatively high growth in that era? Does the explanation for the relatively slow postwar growth of South Asia (compared to Southeast Asia) lie in the former's pre-World War II performance? Are there explanations in the prewar experience of China which might shed light on the recent burst of growth in that country?

Probably Kuznets had these and other thoughts in mind when he visited India, Thailand, Indonesia, and others to see what possibilities existed for working out pre-World War II estimates of GDP. However, he did not find enough economists and statisticians in Asian countries to work out prewar estimates, and eventually gave up the idea. Instead he recommended to US foundations that they train social scientists, and especially economists. The Rockefeller Foundation acted upon this recommendation in establishing the University Development Program. The Program commenced its activities in Japan and was then extended to the Philippines, Thailand, Indonesia, and later, through the Ford Foundation, to India and Bangladesh.

Kuznets found enough good economists in Japan, especially at Hitotsubashi University, to start up the project to compile long-term historical estimates for Japan. The result was the publication of the fourteen-volume historical statistics (LTES), and the development of the university as a center of expertise on empirical studies. It is fitting that Hitotsubashi is now taking the initiative to fulfill Kuznets' dream, especially since the University Development Program and other initiatives have enabled a number of countries in Asia to develop large and competent corps of economists to perform the task. Of course, such a project cannot be accomplished in a mere five years, but it will be well worth the time and money invested.

In the remainder of the paper, let me suggest some tasks that the Project may want to consider.

Task (1): A Careful Study of the First Postwar Decade

Although I agree with Prof. Toshihiko Kawagoe ("Estimating Long-term Economic Statistics for Agricultural Products," Newsletter No. 4, January 1997) that the main efforts of the Statistical Project should be devoted to compiling data for the pre-World War II decades, work on the first postwar decade would also be very useful because the figures from this decade will serve as benchmarks for extrapolation back to the prewar era.

While working on my book, Economic Growth in Monsoon Asia (University of Tokyo Press, 1987), I looked into some pre-World War II estimates of national income and related data. One problem I had to wrestle with was assessing the quality of the 1950s (and sometimes 1960s) estimates as benchmarks from which the prewar estimates are to be projected backward. As in the paper by Pierre van der Eng ("Gauging Growth: Development of National Accounting in Indonesia," Newsletter No. 4, January 1997), I found very little information as to the reliability of the 1950s (or even 1960s) figures. They were suspect because for most countries the data base of the estimates was scanty. And even in the case of population censuses there were many limitations since the censuses were conducted by inexperienced personnel. And similarly with the quality of the national income staff in most of these countries. I recall in 1959 while at Hitotsubashi being asked to lecture to the national income staff in Tokyo on the UNSNA, with particular attention to the estimation of each item.

Therefore, if the estimates for the 1950s (or 1960s) are to serve as benchmarks for the prewar, we need to be able to assess the quality of the early postwar estimates. It may be necessary for the project to give detailed information on the methods and concepts of the estimates, and even in some cases, where there is an opportunity, to improve them.

And if it is true, as Pierre points out, that "later publications of Indonesia's national accounts" have become "noticeably sparse in explanations," shouldn't an attempt be made to interview or contact the statistical agencies and get information on how even the later estimates were constructed? After all, if there is an upward bias in GDP growth rates in the postwar decades (as I suspect in many cases), this comparison with prewar growth estimates may be misleading.

For several countries the earliest postwar estimates of GDP may need adjustments because of certain deficiencies. For example, Taiwan did not conduct its first population census until 1956. Previous population estimates were based on household registration statistics. Similarly, South Korea's first postwar census was conducted in 1949. Previous censuses (up to the one in 1944) were conducted for both North and South Korea, and I am not sure how the figures for the two countries can be disaggregated since there was considerable movement of population between the two as a result of the Korean War. In Indonesia, no census was taken for 1940 or 1950 because of World War II and the War for Independence (1945-49). Consequently, the first census was not conducted until 1961. It is surprising that many countries (Thailand, Philippines, Malaysia, Singapore, Sri Lanka, and China) did not conduct censuses in 1950 or 1951, although a few did so in 1946 or 1947, or in 1953 or 1955.

Further, agricultural, commercial, and industrial censuses were not conducted in most countries until the 1950s. The Philippines was exceptional in that under the influence of the US it conducted censuses (or surveys) of agriculture, industry, and commerce in 1948. Hence, one must look into the nature of the sources of data on production, wages, prices, etc., for nearly all the countries of Asia, especially those in South and Southeast Asia. In monsoon Asia the small firms predominate in production in agriculture, industry, and commerce and other services, but data on the production and wages of small firms are difficult to obtain. In agriculture, although major crops are covered for tax purposes, minor crops are not. I have discussed these issues in my "National Income Statistics of Underdeveloped Countries," Journal of the American Statistical Association, Vol. 52, pp. 162ff. (June 1957) and in Philippine Economic Journal, No. 8, Second Semester, 1965, Vol. IV, No. 2, speculating that due to lack of data, the estimates of GDP by industrial origin may be underestimated for the 1950s, possibly producing an upward bias in estimates of growth rates from the 1950s to the 1960s.

Task (2): Coverage of Non-economic Data

The Asian Historical Statistics Project indeed covers economic statistics very well, but thought should also be given to using the census information to investigate other areas of social science concern. For example, it might be possible to analyze the prewar class structure regarding landlords, working farmers, tenants, landless laborers, tribes, and so on. Changes in areas such as educational attainments and literacy might also be investigated. These data may throw additional light on matters such as social and political stability.

It is too much to expect that estimates on income distribution can be made, as there was no probability sampling during the prewar era. But there may be a way to work out measures that would roughly indicate distribution. In order to work out some distributive data historically, for instance, I would suggest that the project work out coefficients of variation as follows:

[1] Rank each industry by average income per person, or by dividing real production originating in each industry by the number of persons engaged therein: e.g. agriculture, service, manufacturing, construction, mining, government, utilities; and compute the coefficient of variations (C.V.) of average (real) income over time.

[2] Rank each region (province, state) by average income per capita (obtained by dividing total regional product by population of the region), and compute C.V.

[3] Rank each occupation by average income (shown by the earliest data for each occupation (possibly 1950 or 1960)), e.g., professionals, managers, clerks, laborers, workers, etc.; then compute C.V.

[4] Compute average of C.V. of above.

Unfortunately, it is unlikely that data for the prewar decades is available for some of the above for many countries. Nevertheless, the results could prove valuable even if they can be computed for only a few countries. The Taiwan Workshop: A Comment

Let me end this essay with a few remarks on the February workshop on Taiwan's long-run economic development. With the time series data collected on GDP and employment, the papers at the Taiwan Workshop could have worked out product per worker for the economy as a whole. These productivity figures may be revealing as to factors responsible for the growth of efficiency in the prewar decades. And if possible estimates of productivity for the agricultural, industrial, and service sectors may show where the efficiency originated. Of course, there are problems in the estimation because of multi-occupations (e.g., a farmer may be engaged in non-agricultural work in the dry seasons of the monsoon), so that some attempt should be made to adjust the data. Although the workshop also included a paper on structural changes in manufacturing, it will also be important to get the larger picture of structural changes as between agriculture, industry, and services.

Incidentally, product per worker data may provide a rough idea of how reliable the GDP and employment estimates may be. If you get negative productivity for long periods, or very high productivity, product per worker data may indicate problems with the estimates.

In addition, based on the estimates worked out by Mizoguchi and Terasaki on capital formation, it should be possible to push the capital stock data for the 1950s back into the prewar decades. This will make it possible to work out total factor productivity which, despite its limitations (as pointed out by Abramovitz and Nelson), could go beyond labor productivity to reveal factors underlying the growth of efficiency in the prewar decades. This data could then be compared with those of the post-World War II decades.

East-West Center, Honolulu, Hawaii