Revising Long-Term National Accounts Statistics of Taiwan

1912-1990: A Comparison of Estimates of

Production Accounts to Expenditure Accounts

Toshiyuki Mizoguchi

Hiroshima University of Economics


    (1) Statistical Data and Previous Contributions

      (1.1) Data
      (1.2) Estimate for the Postwar Period
      (1.3) Previous Estimates for the Pre-1940 Period
      (1.4) Linkage of Pre-and Post-Liberation Data

    (2) Revisd Estimates of Long-Term National Accounts Tables
      (2.1) Nominal Accounts
      (2.2) Real Accounts of Nominal Implicit Dflator
      (2.3) Overview of Our Estimates

    Appendix 1 Longterm Estimates of SNA of Taiwan
      Table 1-1 National Accounts In Current Prices
      Table 1-2 National Accounts In Real Prices
      Table 1-2B National Accounts : Implicite Deflator

    Appendix 2 Title of Taiwan Database
      Table 1-1 National Accounts In Current Prices
      Table 1-2 National Accounts In Real Prices
      Table 1-3 Preliminary Estimate of National Accounts 1903-1911
      Table 2-1 Nominal Production of Agricultural,Forestry and Fishery Industries
      Table 2-1B Comparison of Estimates on Nominal Gross Value Added of Agriculture
      Table 2-2 Real Gross Value Added of Agriculture,Forestry and Fishery Industries
      Table 2-2B Production Index of Agriculture,Forestry and Fishery:S.P.Ho's Index Linked to Index By Taiwanese Government
      Table 2-2C Detailed Production Index of Agricultural, Forestry and Fishery Industry:1946-1990
      Table 2-2D Production Index of Agriculture by S.Ishikawa
      Table 2-3 Background Indicators for Agricultural Production
      Table 2-3B Quantity of Production of Major Agricultural Products
      Table 3-1 Nominal Gross Value Added of Secondary Industry
      Table 3-1B Production Share of Manufacturing Industry
      Table 3-2 Real Gross Value Added of the Secondary Industry
      Table 3-2B Production Index of Manufacturing Insustry
      Table 3-2C Detailed Production Index of Secondary Industry 1946-1990
      Table 3-3 Census of Manufcturing and Commerce
      Table 3-3B Production Index of Major Products
      Table 4-1 Nominal Gross Value Added of the Tertiary Industry
      Table 4-1B Comparison of Our and Wu's Estimates on Tertiary Industry Production
      Table 4-2 Real Gross Value Added of the Tertiary Industry
      Table 5-1 Composition of Nominal Private Consumption Expenditure
      Table 5-2 Real Privete Consumption Expenditure
      Table 6-1 Income and Outlay Account of General Government
      Table 6-1B Composition of Govt.consumption by Purpose
      Table 6-3 Balance of Public Finance of Taiwan before 1945
      Table 6-3B Composition of Espenditure of Govenment General by Purpose Bofore 1945
      Table 7-1 Nominal Gross Capital Formation
      Table 7-2 Real Gross Capotal Formation
      Table 7-3 Capital Stock in Taiwan


The first issue of the Newsletter of the Center of Excellence Project, Hitotsubashi University (COE project) lays out the basic plan for the Asian Historical Statistics Project. While there are numerous details which we will never be able to verify, we will calculate national income statistics (especially flow accounting), and use the results to create and refine a long-term economic statistics database. The area covered by this project is more extensive than Asia in the usual sense: it includes Asian Russia, the People's Republic of China, North and South Korea, Taiwan, the ASEAN countries, India, some Middle Eastern countries and Egypt. The systematic work has been proceeded by work of a number of subgroups by region and economic activity.

The work on Taiwan is more advanced than for other countries or regions, yet even in this case there is certain information which we are still attempting to locate. Therefore, the reader is warned that the research plan outlined below may be revised in the future or replaced by new plans.

In the first part, Major Tables of National Accounts are shown with explanations of the estimates, which are classified into 9 subgroups, including (1) Production of Primary Industries, (2) Production of Secondary Industries, (3) Production of Tertiary Industries, (4) Private Consumption, (5) Government Consumption, (6) Capital Formation, (7) Foreign Trade. The titles of detailed supporting tables are arranged in the second part with the explanations. Data are contained in a floppy desk, which is available in the Office of the COE project.

(1) Statistical Data and Previous Contributions

(1.1) Data

The basic intent of the COE Project is to search for information from the earliest period possible through 1990. Considering amounts of statistical information available, we will divide Taiwan's economic history into the following periods:

1. Before 1885 (Ching Dynasty (1644-1911): Very little statistical material available.

2. 1886-1912 (First period of Japanese occupation): Some statistical data are compiled but not in systematic fashion.

3. 1912-1938 (Middle period of Japanese occupation): Advances in gathering of statistics makes systematization possible.

4. 1939-1950 (Second World War and Chinese Civil War): Some data is kept secret because of wartime policies, and other data is lost because of economic and social turbulence.

5. 1951-1960 (The Taiwanese government begins to collect data): Essential data becomes available.

6. 1961-1990 (The Taiwanese government systematizes data collection): The gathering of data, including censuses and compilation of input-output tables, reaches the standard of an advanced nation.

In the evaluation of the previous estimates of SNA, we should consider these conditions of statistical data.

(1.2) Estimate for the Postwar Period.

The National Accounts Statistics had been published by the Directorate General of Budgets, Accounting & Statistics, Executive Yuan (DGBAS), since the early 1950s based on the 1953 SNA. While there were some problems in the earlier estimates, improvement was remarkable afterwards. In the early 1970s, the DGBAS adopted the 1968 SNA and published new series from 1951. We know the outline of these procedures from the annual reports of SNA.

The official statistics compiled since 1961 are highly respected in Asia for their reliability, and the Statistics Project therefore needs only to adjust the figures to meet its particular guidelines apart from efforts to upgrade capital formation and capital stock statistics, which have created some difficulties in the international comparison. It is necessary to compile some additional estimates for Period 5 (1951-1960), but the official statistics are generally sufficient. The two major sources for the latter two periods are

DGBAS, [National Accounts in Taiwan for 1951-1990 and Preliminary Estimate of National Income in Taiwan for 1992 (NAT-1992)],1992 and

DGBAS, [Statistical Yearbook of the Republic of China (SYRC)],(Annual Publication).

The former provides flow accounting based upon the "1968 System of National Accounts or New System of National Accounts (NSNA)" recommended by the United Nations. It also provides nominal expenditure accounts for 1951-91, as well as real price accounts in 1986 prices for the same period. Nominal production accounts are also provided for 1951-91, but real value accounts are provided only from 1961. These data are compiled in the COE Database with the following notes.

1.While the Taiwanese National Accounts are based on the NSNA, capital formation is estimated making use of various data including financial reports of enterprises. Note that the NSNA proposed to use the commodity flow method in this field.

2.In NAT-1992, the rigid balance is kept between the expenditure and the production accounts in nominal terms. This means that the statistical discrepancy is zero from 1951 to 1990.

3.The statistical discrepancy is also zero in real terms in 1986 prices in NAT-1992. Based on the agreement of the COE project, the reference year for real terms is changed to 1960, so the discrepancies are no longer zero in real terms.

4.In order to estimate the long term real value series,one of the options is to change the base-years every 10 years and compile timeseries data by the link method of indexes. This option will be examined in the future.

The final note is based on the recent trend in the theory of deflators to respect the link type indexes as a proxy of Divisia type deflators as is recommended in the 1993 SNA.

(1.3) Previous Estimates for the Pre-1940 Period

The main data source for periods 2 and 3 (1886-1911 and 1912-1938) is

Government General of Taiwan, [Statistical Annual of the Government General of Taiwan [Taiwan sotokufu tokei sho](in Japanese), available for all years from 1899 to 1942.

In 1912, the Annual shifted from using vertical writing to horizontal writing, at which time the content of the statistics was improved and expanded. This means that a detailed statistics database can be compiled only for the period from 1912, although data is available for certain areas such as agriculture and foreign trade for as far back as 1890. Taiwanese and Japanese researchers have already made extensive use of this statistical material to produce some important findings. Two important works based on the SNA, both including national income statistics on the production accounts, are the appendix of

Lee, Teng-hue, "Intersectoral Flows in the Economic Development of Taiwan, 1895-1960," Ph.D. Dissertation, Cornell University, 1966, and

"HSING Estimates," both written in the 1960s and cited in Ho, Samuel [Economic Development of Taiwan, 1860-1970], Yale University Press, 1978.

However, it is not quite clear what methods were used to compile the estimates (we are informed by Taiwanese scholars that Lee's appendix was published in Chinese, but the author has not had a chance to access the article). It should be noted that Ho's volume provides an appendix of the excellent works which arranges and overviews Taiwan's historical statistics.

Another statistical research project was conducted in cooperation with Japanese researchers, primarily from Hitotsubashi University, in the late 1960s. The results appeared in the following two works:

Miyohei Shinohara and Shigeru Ishikawa (eds.), "Taiwan's Economic Growth" [Taiwan no keizai seisho], Institute of Developing Economies, 1971 (in Japanese) and

Toshiyuki Mizoguchi "Economic Growth in Taiwan and Korea" [Taiwan Chosen no keizai seicho], Iwanami, 1975 (in Japanese).

The research projects examined historical data not only on the production but also on the expenditure side of the SNA. These were in turn compiled in a major collection:

Mataji Umemura and Toshiyuki Mizoguchi (eds.), "Basic Economic Statistics of Former Japanese Colonies, 1895-1938: Estimates and Findings" [Kyu Nihon shokuminchi keizai tokei, suikei to bunseki], Toyo Keizai Shimposha, 1988 (in Japanese).

This collection included estimates of the expenditure accounts of SNA. While the preliminary estimates of production accounts are also shown, the statistical discrepancy remains relatively large.

(1.4) Linkage of Pre- and Post-Liberation Data

The above three works cover only the period of Japanese occupation, but Mizoguchi attempts to compile long-term statistics beginning from around the time of Liberation from Japanese control and link them to the earlier time series in his article

Toshiyuki Mizoguchi, "Long-term National Accounts of Taiwan and Korea [ Taiwan・Kankoku no kokumin keizai choki keiretsu no suikei" ] in Mizoguchi ( editor), "Analysis of the Long-term (including prewar and postwar) Development of Japan, Taiwan, and Korea" ([Nihon・Taiwan・Kankoku no chokihatten (senzen・ sengo ofukumu) no bunseki ], Institute of Economic Research, Hitotsubashi University, Mizoguchi Research Office, 1988 (in Japanese).

The work compares the pre-liberation expenditure data to the post-liberation data, in real as well as in nominal terms, using price data. This meant that two benchmark years were used to create a deflator, so that two periods were linked through the "indirect method" of a real base.

What really stimulated this type of research was the long-term statistical time series analysis completed at the Department of Economics at Taiwan University in the early 1990s. Among the multifaceted research, national economic statistics were led off by

Wu, Tsong-Min, "An Estimation of Taiwan's Gross Domestic Product: 1910-1950 (1910 Nian zhi 1950 Nian Taiwan Diqiu Guonei Sheng-chan Maoer zhi Guji)", "Taiwan Economic Review " 19-2, June 1991, (in Chinese).

This work assembles statistical estimates on the production accounts of the SNA for 1910-1951 using both nominal and real bases. The figures on expenditures for nominal gross domestic production, when compared to those compiled by Mizoguchi and Noriyuki Nojima (cited in Umemura and Mizoguchi (1988) ), are rather high but the difference is within the 15% range. On the other hand, when a real base is employed, Wu's estimated figures are rather lower, though the difference is certainly not decisive. This research has two distinctive features:

1.The "direct method" was adopted to link the timeseries of estimates for the period of economic chaos, including hyper-inflation, which occurred in 1945-51, at and just after the liberation.

2.Their production side estimates were much more reliable than past ones, and could be compared with Mizoguchi and Nojima's expenditure accounts of the SNA.

The task of compiling estimates was greatly advanced by Evaluation Committee of National Income Statistics, DGBAS,Executive Yuan, "Data for Examinations on Estimate of Gross Domestic Products, 1941-50 (Minguo 30 Nian zhi 39 Nian Guonei Shengchan Moer)" (unpublished materials), 1995 (in Chinese). This research used the NSNA method to calculate production accounts for 1946-1951, and might be regarded as "quasi-official materials." The revised version of this quasi-official estimates was published in

Kuo, Fang-Yao, Zho-Ying Cui, Ming-Zi Lin and Jung-Yi Dong (1997), "Estimates of Gross Domestic Products in Taiwan Area, 1938-1950 (Minguo 26 Nian zhi 39 Nian Taiwan Diqu Kuonei Shangchan Maoer Zhi Tuigu)", "Paper presented for the seminar on Economic Conditions of 1940s Taiwan held in the National Taiwan University in March, 1997", (in Chinese) (to be published in "Taiwan Economic Review") .

Let us call these figures as the "DGBAS Estimates" in this paper. While figures were very similar to their previous estimates, the new version have advantage in the coverage of estimated period. These sources give figures for gross national product for 1946 which are lower than those calculated by Wu. By investigating reasons for the discrepancies and revising Wu's previous figures, we may be able to reduce the discrepancy between Wu's previous estimates and those of Mizoguchi and Nojima.

We are been strongly encouraged by the progress made so far. The next task of the Statistics Project with regard to Taiwan will be to use the official and the quasi-official time series of the DGBAS, the expenditures tables of Mizoguchi and Nojima, and the production accounts tables of Wu to compile a long-term timeseries which will incorporate both expenditures accounts and production accounts of the SNA.

(2) Revisd Estimates of Long-Term National Accounts Tables

(2.1) Nominal Accounts

The final object of the COE project is to estimate flow accounts in the NSNA. Especially important are the "Gross Domestic Product and Expenditure Accounts" of the "Consolidated Accounts for the Nation" and the "Gross Domestic Product by Industry". These accounts are composed of nominal accounts, real accounts and the system of implicit deflators. However, owing to the lack of data for the pre-liberation period of Taiwan, a simplified system of tables is adopted.

1.We have not completed the estimates on the distribution of GDP in the former accounts, which are left as our future work. The expenditure accounts are directly balanced to the latter accounts.

2.The industrial classification in the NSNA is too detailed for the pre-liberation estimates. Following Wu's work, we classify industries into Primary Industry (agriculture, forestry and fishery), Secondary Industry (manufacturing, mining, construction and public utilities) and the Tertiary Industry (other industries, government service and nonprofit institutions).

3.In order to calculate GDP from the total of gross value added, import duties minus imputed service charges should be added to the latter. These items are added to the production of Tertiary Industry. Thus the production of the banking industry is evaluated as nearly zero.

The outlines of our approaches to compiling estimates are as follows. For nominal accounts use the DGBAS's official statistics for 1951-1990 production and expenditures accounts. Remember that the statistical discrepancy between these two accounts is zero in the official estimates. Wait for the DGBAS's 1941-1950 quasi-official production accounts materials to become available for public use (in the meantime, preliminary testing will continue). Refer to previous contributions including Wu's recent works on the production accounts for the pre-liberation period as our previous estimates are weak in this area because our efforts were mainly concentrated on the expenditure accounts.

For the primary industry production and gross value added we use Wu's estimates because these are consistent with the DGBAS's figures for 1941-1950. Note that Wu's estimates for agricultural production are a little larger than the JCRR (the Joint Committee for Rural Reconstruction) estimates which have been generally accepted as the most reliable data. Wu's estimates on secondary industry seem to be revised in two respects. First, his value added ratios are much higher than the DGBAS's data and our previous estimates. Second, Wu does not make adjustments on the "downward biases" of the production figures in the Statistical Annual of Government General of Taiwan, which were pointed out by Shinohara (see his article in Shinohara-Ishikawa (1971)) . The 1941 DGBAS figure is used as the benchmark data, and is linked to the nominal production index of manufacturing industry by Shinohara. Wu's nominal estimates are used for production and gross value added for both construction and public utility industries.

Some additional works are needed for the tertiary industry. Wu gives reliable data on the transport and communication industry. Wu's estimates on government services seems to be underestimated, so new figures are calculated from budgetary data. Other tertiary industry production is estimated by an indirect method based on the regression analysis in Wu's work. While we respect the results as one means of checking data, we try to estimate them by a direct method.

It is very fortunate for us that the DGBAS quasi-official estimates give us 1941 data by detailed industry classification for tertiary industry. These are adopted as the bench mark figures.

1.The nominal production index of the transport and communication industries is calculated from Wu's estimates and linked to the benchmark figure.

2.New estimates are calculated from budgetary data by ourselves, to be consistent with the quasi-official estimates for 1941.

3.To extrapolate the value added of the commerce industry, needed is the timeseries of amounts of trade. The quasi-official report gives us 1941 figures by domestic wholesale, retail and foreign trade. For pre-liberation period, timeseries data are available for foreign trade. The retail trade could be broadly estimated by linking the bench mark data to the nominal index of private consumption from the expenditure accounts. The wholesale data are estimated by the nominal index of total of nominal production amounts of the primary and the secondary industry after the deduction of exports.

4.Regarding the production of the banking, insurance and real estates industries, as noted above, the value of the financial industry is nearly zero if we subtract the imputed rent. The value of the real estates industry is obtained by linking the nominal index of housing expenditures of private consumption to the benchmark figure. This depends on a broad hypothesis that the rent of business use moves over time in a similar pattern to that for private housings.

5.We have little information on the production of the service industry other than benchmark value. But the service for the household sector would occupy the major portion of the service production in pre-liberation period. A proxy nominal service production can be calculated from the household expenditure on education, medical and other personal services. This index is adopted to link the benchmark figure though the writer agrees that the approximation is too broad to be a final figure.

6.The import tax including adjustment items is obtained from the foreign trade reports.

Combining these estimates, we can obtain the gross value added of tertiary industry. The production accounts are shown in the latter part of Table 1-1 of Appendix 1 of this paper.

Regarding the nominal expenditure accounts, the DGBSA's official estimates are adopted for the 1951-1990 period. We will make estimates of only one portion of the nominal expenditures accounts for the 1941-1949 period using the DGBAS's report; i.e. the government consumption expenditure, imports and exports are obtained. Estimates for consolidated accounts will not be conducted during this period, because the the scarcity of data prevents application of the commodity flow method for private consumption and capital formation.

The figures of Mizoguchi and Nojima, slightly adjusted, will be used for the 1912-1938 expenditures accounts. Supplementary outside materials will be used to obtain the preliminary estimates for the 1939-41 period.

1.The data is available for government consumption and foreign trade. So consistent timeseries are obtained for these items from 1912 (or if necessary 1890) to 1990.

2.The preliminary figures of private consumption are calculated by linking the 1938-1941 nominal wage index to the 1938 private consumption.

3.The government capital formation from budgetary data is used to extrapolate the 1938 capital formation data.

4.The 1941 expenditure accounts are calculated to compare the 1941 production accounts by the DGBAS.

Because 2 and 3 are expected to be replaced by the proceeding works by the present Historical, only the results for 1. are compiled in our database as is shown in Table 1-1. According to our preliminary tests, information available through the DGBAS can be used to directly link nominal figures for government consumption, imports, and exports to the pre-liberation SNA timeseries. It is difficult to obtain statistics on private consumption and gross domestic capital formation for 1942-50, but it is not impossible to directly link nominal figures for the pre-1941 figures to post-liberation data.

Regarding Mizoguchi-Nojima estimates on the expenditure accounts, explanations on the method of estimation is given in the text and other previous publications, so only summarized notes are given here.

1.The accounts follow the NSNA system as far as possible.

2.The private consumption and the gross capital formation are estimated by the commodity flow method.

3.The government consumption is estimated from budgetary data of the Government General and the local governments. Note that the expenditure for the Japanese military in Taiwan belongs to the Central Government of Japan which is not included in Taiwan's accounts.

4.The increase in stocks is not estimated as an independent item, but included as a part of other expenditure items.

5.The exports and imports in this database include commodity and service trade although Mizoguchi-Nojima data cover only the commodity trade. The figures for the service trade are taken from Yamamoto's estimates shown in Umemura-Mizoguchi (1989).

It is important to test the reliability of data by the statistical discrepancies between production and expenditure accounts for the pre-liberation period. First, it is confirmed that the discrepancy is relatively small in 1941 between DGBAS production accounts and our preliminary estimate on the expenditure accounts. Further ratios of discrepancy to GDP are shown for our estimates for the pre-1938 period in Table 1-1. For the 27 years from 1912 to 1938, the distribution of ratios are as follows:

from under -10% - 5% 0% 5% over

to -10% - 5% 0% 5% 10% 10%

number of years 0 6 6 8 6 1

These suggest that GDP's estimated from the production accounts exceed slightly GDP's from the expenditure accounts, and that the statistical discrepancy ratios are not too large in their absolute values. It is safe to say our nominal national accounts shown in Table 1-1 could be used as a long timeseries although some detailed revisions are expected in the future.

(2.2) Real Accounts and Implicit Deflator

As the case of nominal accounts, four series are available for real SNA of Taiwan. They are:

1.The real accounts in 1986 prices on expenditure by the DGBAS from 1951 to 1990 and those on production from 1961 to 1990.

2.The quasi-official estimates on the production accounts in 1951 prices by the DGBAS for the 1941-1951 period.

3.Mizoguchi-Nojima's estimates on expenditure accounts in 1934-36 average prices prices for the 1912-1938 period (the preliminary data are also available from 1903 to 1911).

4.Wu's estimates on production accounts in 1951 prices for the 1910-1951 period.

Note that official implicit deflators are not published for the 1951-1960 production accounts.

Our task is to unify these series and test their reliability. For the post-liberation period, the official estimates by the DGBAS are adopted. To fill the lack of official estimates on the 1951-60 real production by industry, we took the following approximation:

1.The 1961 real production of primary industry is linked by the production index of agriculture by the JCRR.

2.The 1961 real production of secondary industry is linked by the production index of mining and manufacturing by the DGBAS.

3.The real production of tertiary industry is estimated by reducing the real production of primary and secondary industry from the real GDP from the expenditure accounts.

Item 3 depends on the system of NSNA of DGBAS which defines statistical discrepancy as zero not only in the nominal accounts but also in the real accounts. These works are preliminary in their nature and should be replaced if the DGBAS publishes the official figures for the real production accounts.

The Historical Statistics Project requests the members to take 1960 as the reference year of implicit deflators. Because nominal and real accounts give us the system of implicit deflators, we can switch the reference year from 1986 to 1960 as is shown in Table 1-2B. The revised real accounts in Table 1-2 are calculated by these deflators. Note that the statistical discrepancy is no longer zero.

The pre-liberation real accounts are initially calculated in 1934-36 average prices. For the expenditure accounts, Mizoguchi-Nojima estimates are used. The real production of primary and secondary industries are obtained by the deflation of nominal values by the deflators of each industry shown in Wu's estimates after the switch of reference year. The real production of tertiary industry is obtained by the sum of 5 sub-industries shown in the nominal accounts, of which deflators are defined as follows:

transportation and communication: Wu's deflator, wholesale and retail: GDP deflator in the expenditure accounts, finance and real estates after the deduction of imputed rent: deflator for accommodation in the private consumption, government service: deflator for government consumption, and other service: deflator for services in private consumption.

The deflator for tertiary industry is the ratio of nominal values to real values.

To link these real series of the pre-liberation period to those of the post-liberation, we will rely not on indirect methods as we have in Mizoguchi (1988), but shift to the direct method of using the Taiwanese government's official statistics on price indicators, which also cover the period of economic chaos (1941-51). By this approach, the real series are obtained for three industries in the production accounts, and for government consumption and foreign trade in the expenditure accounts. The semi-direct method is used for the linkage for personal consumption and gross capital formation; i.e. the nominal data are adjusted by deflators obtained from timeseries of price indices. The results are shown in Table 1-2 and Table 1-2B of Appendix 1 of this paper.

In comparison with the nominal accounts, statistical discrepancy GDP ratios are large in the real accounts. This reflects the 'biases' by changes of relative prices. To distribute smoothly these biases into both deflators and real values, we now examine changing the base year of deflators every ten years and to linking them. This depends on the idea of Divisia index although the preliminary results are shown in the next section.

(2.3) Overview of Our Estimates

Let us explain the tests we have been conducting. Table I shows a portion of the tests. Nominal data used in the same chart are calculated on the basis of New Taiwan Dollars, so 1 New Taiwan Dollar = 40,000 Old Taiwan Dollars or Taiwan Yen (note that the pre-1940 and post-1940 units are different).

Table I

The information in Table I can serve as the basis for any number of analyses. For example, a sharp decline can be found after 1960 for the share of primary industry, which was the major production sector under the Japanese occupation. The share of government consumption increase in the post-liberation period because the defense expenses were financed by the Central Government of Japan in the pre-liberation period. While the foreign trade had been in surplus before the Second World War,the trade balance was nearly zero in the post-liberation period. Further examinations are left for future works.

Next if we look at Table II and then consider the relationship between :

nominal GDP = deflator * real GDP

real GDP = population * real per person

we can then analyze the growth rate of nominal gross domestic product by breaking it down into the real GDP, GDP deflator, and population growth, and find average growth rates for different periods. Taiwan's pre-World War U growth rate was high by international standards but because population growth was also high, there was little improvement in living standards. In contrast, rapid economic growth from 1960 brought a large increase in per capita GDP. In addition, we can infer that the economic chaos in the 1940s slowed real GDP growth and increased the GDP deflator by an even greater rate.

Nominal GDP and its Composition

1915 1925 1935 1951 1960 1970 1980 1990
Nominal GDP in 1000 yuan in billions of yuan
5.33 17.49 23.33 12.33 70.04 226.8 1491 4222
Composition (%) in Expenditure Accounts
Private consumption 82.49 70.65 64.79 72.42 67.87 56.28 51.49 54.52
Government consump. 6.00 5.62 6.08 17.85 19.25 18.25 15.91 17.62
Capital formation 8.07 12.95 18.62 14.43 19.96 25.52 33.80 22.40
Export 40.68 40.34 44.37 10.20 14.00 30.31 52.53 53.72
Import 35.59 29.57 33.86 14.89 21.08 30.36 53.27 42.24
Composition (%) in Production Accounts (excluding discrepancies)
Primary industry 38.49 43.87 38.25 32.28 28.54 15.47 7.68 4.13
Secondary industry 21.08 22.50 27.60 21.33 26.87 36.83 45.75 42.53
Tertiary 40.42 33.63 34.15 46.38 44.59 47.71 46.57 53.34

Table II

It is well known that the constant price system is too rigid to be applied for long-term historical statistics. It is true for our case. Because the relative prices changed remarkably through the economic development of Taiwan, the growth rate varies depending on the selection of the reference year. It is a good idea to apply the yearly linkage type indices for the implicit deflator as a proxy of Divisia type index suggested in the 1993 SNA, but time consuming works are necessary to complete the idea. To escape from the dilemma, let us propose a simplified form of link index method. Suppose an aggregated value (for example GDP) is composed of a number of subgroups (for example, three industrial groups). To save work, let us adjust the effects of relative price changes between subgroups supposing the effects within groups are relatively small. This assumption would be accepted in the normal economic conditions. Based on this assumption, let us change base years of deflators every 10 years, and calculate the aggregated values. The long-term series of real aggregated values can be obtained by the linkage of of these series. Table III shows our preliminary results. The GDP's in 1960 constant prices (Series A) are compared with those adjusting the effects of relative price changes between three industrial groups (Series B) and those between 5 expenditure groups -- private consumption, government consumption, gross capital formation, export and import -- (Series C). Because data are not available from 1938 to 1950 for Series C, we interpolate them using the deflators for Series B.

A Summary of Historical

Annual Growth Rates of GDP

Period nominal deflator real Population
1912-20 7.19 10.40 2.41 1.13
1920-30 4.22 -0.43 4.95 2.22
1930-40 7.50 4.76 2.17 2.64
1940-45 29.05 49.67 -12.09 2.73
1945-50 843.55 671.01 17.38 2.67
1950-60 22.85 11.70 10.00 3.68
1960-70 13.76 3.44 10.00 2.57
1970-80 20.72 9.36 10.39 2.01
1980-90 10.97 2.85 7.88 1.29

Table III

(3) Summary of Statistical Database

The estimates of National Accounts mentioned above are the synthetic systems of various estimates. Each category will include basic data necessary to compiling SNA estimates. For example, the gross value added of the primary industry based on the data on agricultural production, input of fertilizer, number of agricultural laborers, etc.

In our database for the National Accounts estimates, they are classified in a three-stage structure. These most basic data were compiled in our First stage database for private use. They were united to items which are essential in the calculation of national accounts. These are compiled in the Second stage structure, and available to the public. The secondary database is composed of nine sections:

Section 2 : Activities Indicators of Primary Industries,

3 : Activities Indicators of Secondary Industries,

4 : Activities Indicators of Tertiary Industries,

5 : Activities Indicators of Private Consumption,

6 : Activities Indicators of Government Finance,

7 : Statistical Data on Capital Formation,

8 : Foreign Trade Indicators,

9 : Finance and Price Statistics, and

10 : Population and Labor Statistics.

Section 1 is composed of National Accounts. The titles of series from Section 2 to 7 compiled in these sections are shown in Appendix 2 of this paper while the series of Sections 8.9 and 10 are to be published by other members of our project. Timeseries are identified by three numbers indicated as (a, b, c): they are

a : the number of the section,

b : the number of tables (two figures; e.g. 1B)

the first digit 1: nominal values,

2: real values,

3: reference indicators, and

c : the serial number.

These data series are available in the Excel95 format either by a Floppy desk or the Down Road in the internet through the office of the COE Project after January, 1998.The detailed explanation on the method of estimation on each item will be given in the forthcoming manual(in Japanese).

Comparison of Constant Price Method with

Semi-Divisia Method in the Calculation of real GDP

(millions of yuan)
year 1912 1920 1930 1940 1951 1960 1970 1980 1990
Series A 9655 12779 19056 24835 30920 62507 161708 434620 928295
Series B 8517 8181 14768 20699 27922 62507 172291 466411 995772
Series C 6918 9497 13945 - 30413 62507 161612 383009 716235