1. INTRODUCTION

Both the size and rapidity of the growth performance of the Chinese economy have attracted a great deal of attention from economists and politicians alike. The key issue under scrutiny is the long-run real performance of Chinese industry, since it is widely believed among specialists that the recent Chinese industrial growth has been greatly exaggerated (Maddison, 1998; Ren, 1997; Woo, 1996; Keidel, 1992).

Despite tremendous efforts made since the late-1980s by the Chinese State Statistical Bureau (SSB) to shift from the Soviet-style material product system (MPS) to the standard system of national accounts (SNA), the Bureau has not yet to achieve an acceptable level of accuracy in measuring and assessing China's post-war industrial performance. This is mainly due to methodological inertia stemming from China's long use of statistical practices associated with the MPS, practices which are fundamentally different from those used under the SNA.¹⁾ Recent studies (Wu, 1993; SSB-IER, 1997) have shown that attempts to reconstruct China's post-war national accounts based on the SNA framework are strongly influenced by the MPS, due to the extreme difficulties associated with eliminating the effects on output valuation of price distortions, insufficient statistical coverage, and improper data collection systems.²⁾

One way to get around these effects is to construct an independent output growth index based on physical output rather than value output. Such a quantitative index approach has been applied to countries with a MPS-type output accounting system: this quantitative index method is perhaps the most systematic and comprehensive method of checking the plausibility and consistency of the official data. This approach was also used by economists in the West to measure and assess China's economic performance in the 1950s (see Liu and Yeh, 1965; Chao, 1965).

2. THE UPWARD-BIAS HYPOTHESES

It is argued that one of the main factors for possible upward biases in Chinese official output measurement is a 'comparable price' system that was copied from the Soviet Union in the 1950s. The SSB 'comparable price'-based GDP deflator is currently assembled from five administrative 'constant prices' (i.e. 1952, 1957, 1970, 1980 and 1990 constant prices), derived from average prices of 'representative products' of each industry for a particular period. There is however no detailed information on how these representative products were chosen and how the average prices were computed over time. Enterprises at the basic accounting level are required to report their output in both the given 'constant prices' and current prices. This approach does not have the same meaning as 'constant prices' in western statistical usage. Many researchers believe that the use of 'comparable prices' to construct the GDP deflator has understated inflation and exaggerated growth because they do not adequately reflect price increases (Maddison 1998; Woo 1996; Keidel 1992).

Even Chinese official statisticians have doubted the reliability of this approach. A study carried out by the Industrial Division of SSB Hunan Province Branch (1989) constructed an independent industrial index for Hunan and found that the annual real growth based on this index (9.2 per cent) was systematically lower than that based on the 'comparable price' approach (13.5 per cent) for the period 1983-87.

The following points may be considered likely factors generating upward biases in China's output statistics, thereby overestimating China's actual economic performance:

a) reporting output in current prices as in constant prices by non-state owned enterprises because: 1) they might not have begun their business at the time when the prevailing constant prices were set, 2) of convenience and 3) of ignorance;

b) the complexity of turning new products that did not exist in the base year (the year for the prevailing constant prices) into something equivalent in the base year prices;

c) the coverage bias towards (low) state listed prices and insufficient coverage of the (high) prices of other transactions that are mainly influenced by market;

d) the tradition of not writing off unsaleable products;

e) the tradition of imposing low prices on some producer goods which inflated profits;

f) enterprises' incentives to oblige their supervising bureaux by overreporting output since high growth performance can be interpreted as evidence of superior management ability from the upper management level;

g) political incentives of local officials to ignore or even encourage the misconduct of accounting practices by enterprises favouring high growth performance;

h) possible improvement in statistical coverage over time.

3. METHODOLOGY ³⁾

The basic methodology used in this study is based on Laspeyres' quantity index using base-period weights. The year 1987 was selected as the base period mainly because it was covered by China's first SNA-type input-output table and the data from the table were available when this study began. It should be noted that although more input-output tables have been compiled since this study began, we still argue that 1987 is a better choice than later periods. It is true that compared to prices in the pre-reform period, prices in the post-reform period were generally better indicators for factor costs. However, although prices in the late 1990s were much more market-determined than those in the 1980s, they may not necessarily be better indicators than those of 1987 for the output of the 1950s and the 1960s, as the real factor costs might have changed significantly over such a long period.

There are three major steps involved in converting the calculations of aggregate commodity-level quantity data to the standard branch level. The first step is to obtain 1987-based quantity indices at the industry (or commodity group) level that exactly match the industry classification of the 1987 input-output table. The second step is to extrapolate gross value added (GVA) in the 1987 prices at the industry level to other years using the estimated quantity indices. The extrapolation involved assumes that the GVA to GVO ratio remains constant over time, which implies 1) technology remains the same over time - essentially as given by the 1987 input-output table; and 2) the 1987 price weights (input prices versus output prices) are assumed to hold constant over the entire period in question (note the correction effects of 1987 prices to the previously distorted output valuation). The last step is to aggregate the estimated industry-level GVA into the standard branch and use this as the basis on which to calculate the growth index for each branch.

ANUAL COMPOUND GROWTH RATE OF GROSS VALUE ADDED (GVA) IN CHINESE INDUSTRY BY BRANCH, SELECTED PERIODS IN 1949-1997

(IN PERCENTAGE)

Source : Official results are from SSB and SSB-IER (1997, Table A10-note 1997figures are directly adopted from SSB).

4. DATA

There are mainly two types of data that are used in this study: output in quantity and producer prices. Data on the physical output of each industrial product were obtained from various volumes of China Industrial Economic Statistical Yearbook published by the Department of Industry and Transportation Statistics (DITS) of SSB. Products included in the Yearbook are therefore termed CIES items in this study. The number of CIES items is around 200 though it varies slightly each year. In this study, most of the CIES items for 1995 have also been crosschecked using the information from China's Third National Industrial Census.

The exercise of intra-industry weighting requires detailed producer price data that are mainly obtained from The Post-War Chinese Industrial Price Data now available at IER, Hitotsubashi University and Price Yearbook of China. Gross value added (GVA) data by industry are obtained from China 1987 Input-Output Table, called CIOT industry in this study. CIOT industries are also grouped in to ISIC branches.

Not every CIES item can be used in constructing quantity index. At the first stage of data processing, items that were less representative or were difficult to group with other items, and in these significantly incomplete time series were dropped. As a result, 123 of the 198 items were selected. Most of the selected CIES items were the time series from 1949 to 1997. However, some of them started later than 1949, either because they were new products only available after 1949, or because they were not included by official statistics in the earlier period, or the data was lost. Some series were broken for some reason. For those items that were in incomplete time series but should not be dropped, interpolation and/or extrapolation were used to reconstruct the time series by following either trend of the item itself or the trend of other relevant items obtained by regression.

5. RESULTS (SELECTED) AND DISCUSSION

Results at the branch level for selected periods are reported in the table below, along with comparisons of official estimates. They support our argument that official statistics have significantly exaggerated China's industrial growth performance. The overestimation by official data is about 1.9 per cent per annum for the pre-reform period 1952-78 and 3.5 per cent per annum for the post-reform period 1978-97. The question is whether these results will hold up under careful scrutiny.

Our assumption is that the 1987 price structure is likely to be more appropriate than those of both earlier and later years. Furthermore, the 1987 measure of gross value added is based on the SNA approach that characterised the 1987 input-output table. However, it is possible that there have been changes in the aggregate input-output ratio for industry in the 49 years covered. Obviously, if the ratio of value added to gross output has been falling over time, our estimates will still overstate the growth of value added; conversely, if the ratio has been rising, our estimates will understate it. More importantly, if the ratio has indeed been falling, the overestimation by this study will be more for the post-1987 period than the pre-1987 period: this in turn would mean that China's real industrial growth after 1987 is even lower than the levels indicated by our estimates.

Unfortunately, there is no adequate input-output table prior to that of 1987, hence it is impossible to estimate the extent to which the 1987 weights are biased. However, there are input-output tables for 1990 and 1992. An analysis of the aggregate ratios in these sources reveals a significant fall in the ratio of value added to gross output of the industrial sector between 1987 and 1990, and between 1990 and 1992, down from 34.5 per cent in 1987 to 30.3 per cent in 1990 and 28.6 per cent in 1992. If this increasing input ratio was characteristic for the period of 1949 to 1994 as a whole, then our fixed 1987 weights will have definitely overstated China's industrial growth.

Research on China and East Europe's centrally-planned economies has shown that their value added ratios in the late 1980s were significantly lower than in Western economies. These centrally-planned economies made wasteful use of raw materials and other intermediate inputs because the price system did not encourage efficiency. Inputs of steel and energy were characteristically higher in these countries than in the West. Apart from waste of inputs, these countries also tended to accumulate large unmarketable inventories. Moreover, the ratios of these economies fell a good deal from pre-war levels to the 1990s. Research suggests that the Chinese economy was less wasteful of inputs than other former centrally-planned economies, but more wasteful than industrialized market economies.

There is another reason for thinking that the use of 1987 weights may exaggerate industrial growth. In general, smaller enterprises have lower value added ratios than larger ones. Since the size of the average enterprise has fallen in the process of economic reform mainly due to the fast development of labour-intensive rural enterprises, we can assume that there was a corresponding decline in the value added ratio for the industrial sector due to this reason.

(The Hong Kong Polytechnic University)

References

Chao, Kang (1965), The Rate and Pattern of Industrial Growth in Communist China, University of Michigan, Ann Arbor.

Keidel, A., 1992, How Badly do China's National Accounts Underestimate China's GNP? Rock Creek Research Inc. E-8042.

Liu, Ta-Chung and Yeh, Kung-Chia (1965), The Economy of the Chinese Mainland: National Income and Economic Development, 1933-1959, Princeton.

Maddison, Angus (1998), Chinese Economic Performance in the Long Run, OECD Development Centre, Paris.

Ren, Ruoen (1997), China's Economic Performance in An International Perspective, OECD Development Centre, Paris.

SSB-IER, The Historical National Accounts of the People's Republic of China, 1952-1995, State Statistical Bureau of the People's Republic of China and Institute of Economic Research Hitotsubashi University, September 1997.

Woo, Wing Thye (1996), Chinese Economic Growth: Sources and Prospects, paper presented at Economics Discipline Seminar, RSPAS, Australian National University.

Wu, Harry X. (1993), "The 'Real' Chinese Gross Domestic Product (GDP) for the Pre-reform Period 1952-77", Review of Income and Wealth, Series 39, No.1.

Note

1)See the author's paper of the IER (Hitotsubashi) Discussion Paper series (forthcoming, 1999) for a detailed comment on the difference between MPS and SNA with mathematical expression.

2)Although the account on a SNA basis have recently been pushed back to 1952, the new estimates by SSB and IER (Hitotsubashi) (1997) are not convincing, because the underlying growth trend of the newly estimated industrial GDP is virtually the same as that of the old NMP estimates. A test for the relationship between the SSB-IER newly estimated industrial GDP and the old industrial NMP based on the equation ln GDPt =a+ƒÀln NMPt has found an almost identical growth trend of the two series with an adjusted of R2 0.9997.

3)Full mathematical expression of the methodology is given in the author's IER (Hitotsubashi) Discussion Paper (forthcoming 1999).