Printer Friendly

Post-reform industrial productivity performance of China: new evidence from the 1985 industrial census data.

I. INTRODUCTION During the last decade China implemented a series of economic reforms that followed what many observers agreed was a period of stagnation in industrial productivity.(1) The first wave of Chinese economic reforms (1978-79)--following the economic opening to the West under the "Four Modernizations"--began after the Third Plenum of the Eleventh Central Committee of the Communist Party of China announced a new policy of economic modernization in December 1978. A major goal of the new policy was to improve productivity. Toward this end, the Party would (1) allow the formation of private enterprises; (2) permit state-owned and collective enterprises to retain a portion of their profits; (3) devolve a greater degree of decision-making to factory managers and drastically reduce the scope of planning; (4) introduce material incentives such as bonuses to labor; and (5) place increased reliance on markets for inter-industry resource allocation. It is important to note that at this time, in the early 1980s, real reform was limited to the agricultural sector, as discussed in Perkins [1988]. The second wave of reforms occurred in 1983-84 with the announcement of sweeping changes for the urban industrial sector. These reforms included four major measures: (1) a reduction in the number of leadership positions in enterprises; (2) a further expansion of enterprise managers' authority; (3) the substitution of an income tax for remission of profits to the state; and (4) removal of the ceiling on bonuses. Naturally, it is important to know whether or not these reforms improved Chinese industrial productivity. Since the early 1980s both Chinese and Western economists have conducted many empirical studies on China's post- reform economic performance. However, the results from these studies are far from conclusive. For example, the World Bank [1983; 1985], Field [1984], Yeh [1984], Chen [1986], Chen and Sang [1986], Pan [1986], and Tidrick [1986] have reported that input accumulation accounts for nearly all of output increases in industry as a whole in the early 1980s. In particular, Tidrick [1986] found that "China's performance has been extremely disappointing": total factor productivity in the state-owned industrial sector declined at an annual rate ranging from -0.10 percent to -1.20 percent for the period 1978-83. In contrast, more recent studies, such as those by Kuan et al. [1988], and Jefferson [1988], have found that China's post-reform industrial productivity was significantly improved. Most notably, Kuan et al. [1988] found that Chinese industrial productivity in the state-owned sector grew at an average annual rate ranging from 2.7 to 3.1 percent for 1980-84, and from 15.3 to 18.2 percent for 1984-85. This difference in the findings reported in the literature is striking and requires further investigation. One explanation is that the various studies are based on different data sets, time periods, models, and data construction methods. For example, Tidrick [1986] used aggregate time-series data for state-owned industry for the period 1952-83, while Kuan et al. [1988] used aggregate time-series data including only independent accounting units within the state-owned industrial sector for the period 1953-85. Perkins [1988] reviewed China's overall post-reform economic performance using data at the national level. He found that, during 1976-85, total factor productivity in China grew at an annual rate of 3.79 percent, accounting for over 40 percent of China's net material product. He concluded that "reform and productivity growth thus led the way to higher overall growth" and that agriculture and the collectively organized small-scale industrial enterprises "play an important role in the accelerated productivity growth of the reform periods." Perkins, however, hastened to point out that his conclusion is "based on impressionistic evidence and is in no sense definitive" and suggested that disaggregated data on outputs and inputs by sector or individual industries are required to identify those sectors that accounted for most of the rise in productivity (Perkins [1988, 628]). In this study, we use the newest and most comprehensive data available on Chinese industrial activities, the National Industrial Census of China [1988]. This data set contains information on inputs, outputs, and other variables by branch of industry and by type of enterprise for the years 1980, 1984, and 1985, and thereby provides a unique opportunity to conduct a more complete productivity analysis than could be done before. First, as Perkins pointed out, these disaggregated data allow us to identify those industries, sectors, and types of enterprise that contributed most to the overall post- reform economic performance. Second, the three years of cross-sectional data not only permit us to estimate total factor productivity growth for each industry using industry specific weights, but they also allow the estimation of output elasticities based on production functions of various functional forms. This in turn allows us to examine the sensitivity of the estimated productivity growth to the magnitude of the weights, elasticities, functional forms, and other assumptions underlying the competing factor productivity models. Finally, the three years 1980, 1984, and 1985 conveniently encompass the two waves of reform (1978-79 and 1983-84), allowing us to evaluate the performance of each set of reforms. With the new data we estimate total factor productivity growth for thirty- nine individual industries and three types of enterprise (state-owned, collective, and others) for the periods 1980-84 and 1984-85, using both the value-added model and the gross-output model.(2) For each model we estimate six variants and use the results to test for the sensitivity of the estimated productivity growth rates to different specifications. Finally, we use regression analysis to examine factors explaining differences in productivity growth across industries. Several interesting results emerge from the analysis. We find that our growth estimates are strongly affected by output specifications. Specifically, our results show evidence that the value-added model yields biased estimates of productivity growth. For Chinese industries, it substantially overestimates both productivity growth and its decline. The results of the gross-output model appear to be robust and reasonable. Based on these results, we first find that Chinese industries, in particular those in manufacturing, experienced sharp increases in factor productivity growth in the 1984-85 period as compared to the 1980-84 period. This indicates that the economy responded quickly to the 1983-84 reforms. Second, we find that total factor productivity growth is uniformly greater for collective and private enterprises than for state enterprises during 1980-84. Moreover, private enterprises continued to grow at a rapid rate and outperform both state-owned and collective enterprises in 1984-85. Third, our regression results indicate that the proportion of technical employees has significant positive effects on productivity growth in the Chinese industrial sector. Finally, there is evidence that retained profits have a positive impact on productivity growth; however, bonuses to labor appear to have a negative effect. In the next section, we discuss our productivity growth model. A brief discussion of the data and calculations is given in section III. Section IV presents the results. Section V examines the sources of productivity growth across industries and provides evidence on productivity differences by ownership. Section VI offers suggestions for further work. II. MODEL SPECIFICATIONS The Gross-Output Model The conventional methodology to measure factor productivity is based on a production function. For a three-input model, a general production function can be written as(3) (1) Q(t) = A(t)f[K(t),L(t),M(t)], where Q(t), K(t), L(t), M(t) are output, capital, labor and materials inputs, and A(t) is an index of Hicks-neutral technical change or total factor productivity at time t. Differentiating (1) with respect to t, and with some algebraic manipulation, we can derive the following basic productivity growth model where are the growth rates of output, total factor productivity, capital, labor, and materials at time t, and [W.sub.qk], [W.sub.ql], [W.sub.qm] are output elasticities of capital, labor, and materials. The Value-Added Model Previous aggregate productivity studies, including those of the Chinese economy, often used the value-added model rather than the gross-output model.(4) The value-added model, which allows the estimation of factor productivity growth without the inclusion of the materials input, can be written as where denote the growth rates of value added, capital, labor and productivity; [w.sub.vk] and [w.sub.vl] denote output elasticities of capital and labor. Given the two models, which one is more appropriate for productivity analysis? The gross-output model has a less restrictive formulation of inputs. Consequently, productivity growth rates estimated using the gross-output model should reflect the true factor productivity growth because gross output is likely to approximate theoretical output. Under the assumption of constant returns to scale, Baily [1986] has derived the following relationship between the theoretically correct factor productivity growth rate, and that estimated using a value-added model, where and r, and [w.sub.m] denote the materials to output ratio (M/Q), the growth rate of r, and the share of materials (i.e., [w.sub.m] = [p.sub.m]M/[p.sub.q]Q, where [p.sub.m] and [p.sub.q] are the prices of materials and output, respectively). Among other things, equation (4) implies: (1) If the materials to output ratio, r, is fixed so that then and are proportional to each other. In this case, overestimates both the growth and decline of. For example, if [w.sub.m] = .50 and, then using equation (4) we find, which is two times as large as. (2) If r and the relative price of materials, [p.sub.m]/[p.sub.q], change over time, then could be smaller, greater, or equal to the theoretically correct productivity growth rate, depending on the sign and the magnitude of r, [w.sub.m], and. For example, during the period when both r and [p.sub.m]/[p.sub.q] increase so that [w.sub.m] [is greater than] r and, and hence [Beta] [is greater than] 0, then overestimates the growth rate of true total factor productivity when. By the same token, will overestimate a decline in productivity growth when [Beta] is negative and is negative or equal to zero. Thus, theoretically, the value-added model yields systematically biased productivity growth estimates even in the case of constant materials-output ratios (i.e.,). For the Chinese industrial sector as a whole, using actual data we find that [Beta] is negative.(5) It follows that if the data are accurate we expect that all the value-added-models will (i) amplify productivity declines when, (ii) underestimate productivity growth when and its absolute value is equal or somewhat greater than that of [Beta], and (iii) overestimate productivity growth if the absolute value of is much greater than that of [Beta]. In practice, however, the value-added model is often required when the analysis is undertaken at a highly aggregate level such as the whole economy or an entire manufacturing sector. The usual argument for this is that double-counting problems are rife in aggregate gross output measures. Since outputs of an industry can be purchased and used as inputs by another industry for assembly into final goods, value added is a more appropriate measure of output because intermediate outputs are netted out. Thus, most researchers agree that for analysis at the micro level, such as establishments where there is little intra-industry trade and gross output should reflect theoretical output, the gross-output model is the correct one. For analysis at highly aggregate levels where the double-counting problem is severe, the value-added model is preferred. However, as Baily [1986] notes, for analysis of individual industries there is no clear basis for using the gross-output over the valued-added approach. Following this line of argument, we apply both the gross-output and valued-added models in our analysis. Comparisons of these models will serve as a sensitivity test of the estimated results from the two competing specifications. III. ESTIMATION, SENSITIVITY TESTS, AND DATA SOURCES Estimation Procedure For estimation, we need data on output (Q or V), capital (K), labor (L), materials (M) and output elasticities (w). While data on Q, K, L, and M are available, output elasticities (w) are unobserved and must be estimated. Ideally, the elasticities should be estimated using a general model of production. However, we have only three observations for each industry (one observation for each of the years 1980, 1984, and 1985), and it is not possible to estimate separate industry output elasticities. The lack of data obliges us to use output shares as proxies for output elasticities in calculating productivity growth rates for individual industries. We recognize that the use of output shares can lead to biased estimates of industry productivity growth rates because output elasticities can be identified with output shares only when markets are competitive. Chinese markets are far from competitive, so it is virtually certain that output shares are biased estimates of output elasticities and the resulting productivity growth rates are also biased. We note, however, that though estimates of factor productivity growth based on output are biased, their trend should be insignificantly affected. This is because the trends are primarily affected by the percentage change, not the magnitude, of output elasticities, except for some extreme cases.(6) In particular, if output elasticities are stable and the bias is approximately constant, then it can be shown that the error in the change in productivity growth equals the bias multiplied by the change in the growth rates of inputs. For example, if the bias in the estimate of output elasticity for each input is 10 percent and the change in the growth rate of each input is 10 percent, then in a three- factor model the error in the change in productivity equals 3 percent. Because the period under study is relatively short (1980-85) and the reforms were gradually implemented, we do not expect substantial changes in output elasticities. Sensitivity Tests While it is not possible to estimate output elasticities for each individual industry using production functions, we can estimate the average elasticity for each industrial sector using a cross-sectional production function. These estimated average elasticities can be used to calculate average productivity growth rates which, in turn, can be compared to productivity estimates developed using output shares as proxies for output elasticities. For a three-factor gross-output model, the translog production function can be written as (5) lnQ = [[Alpha].sub.0] + [[Alpha].sub.K]lnK + [[Alpha].sub.L]lnL + [[Alpha].sub.M]lnM + 0.5[[Alpha].sub.KK][(InK).sup.2] + [[Alpha].sub.LL][(lnL).sup.2] + 0.5[[Alpha].sub.MM][(lnM).sup.2] + [[Alpha].sub.KL](lnK lnL) + [[Alpha].sub.KM](lnK lnM) + [[Alpha].sub.LM]lnL lnM. For the translog function to be well behaved, symmetry and homogeneity are usually imposed. Symmetry is imposed by the restriction [[Alpha].sub.ij] = [[Alpha].sub.ji]. Homogeneity of degree [Lambda] is imposed by setting (6.a) [summation of] [[Alpha].sub.i] = [Lambda], and [summation of] [[Alpha].sub.ij] = [summation of] [[Alpha].sub.ji] = 0, i, j = K, L, M. Homogeneity of degree one is imposed by setting (6.b) [summation of] [[Alpha].sub.i] = 1, and [summation of] [[Alpha].sub.ij] = [summation of] [[Alpha].sub.ji] = 0, i, j = K, L, M. Because the output elasticity of the ith input equals [Delta]lnq/[Delta]ln[x.sub.i], estimating equation (4) yields a set of parameter estimates (i.e., and i, j = K, L, M) that allow us to calculate output elasticities using the following equation, (7) [Delta]lnQ/[Delta]ln[X.sub.i] = [S.sub.i] = 1/[Lambda][[[Alpha].sub.i] + [summation of] [[Alpha].sub.ij]ln[X.sub.i]], i, j = K, L, M. The translog function reduces to a Cobb-Douglas function if the following restrictions are imposed: (8) [[Alpha].sub.ij] = [[Alpha].sub.ji] = 0, i, j = K, L, M. It follows from this specification that the output elasticities are simply [[Alpha].sub.K], [[Alpha].sub.L], and [[Alpha].sub.M]. Moreover, the Cobb-Douglas function is homogeneous of degree [Lambda] if [[Alpha].sub.K] + [[Alpha].sub.L] + [[Alpha].sub.M] = [Lambda] (constant returns to scale if [Lambda] = 1; increasing returns to scale if [Lambda] [is greater than] 1; decreasing returns to scale if [Lambda] [is less than] 1). For the sensitivity analysis, we estimate twelve competing models listed in Table I. Six of the twelve models are based on the gross-output specification, and the remaining six models are based on the value-added specification. This yields twelve different sets of estimated output elasticities which are then used to calculate twelve sets of total factor productivity growth rates. These results permit comparisons of the effects of the following factors on the estimated productivity growth rates and their intertemporal change: (i) alternative output specifications; (ii) different output elasticity estimates; and (iii) the imposition of constant returns to scale on the production technology. Data and Sources The data are taken from the People's Republic of China's National Industrial Census [1988], which reports information on individual branches of industry for three years, 1980, 1984 and 1985. These data were collected from large and medium enterprises, which account for about half of China's industrial output. Output Measures. Output is measured as gross output and value added. Data on gross output for each industry are available in both current and 1980 constant yuans. Thus, we are able to derive an output price deflator by dividing value of gross output in current yuans by value of gross output in 1980 constant yuans. Data on value added are available only in current yuans. To obtain "real" value added we divide value added in current yuans by the derived output price deflators.

[TABULAR DATA OMITTED] Capital Input. Data on fixed assets (capital stock, KS) used in production are directly available in the data set. Fixed assets are calculated by adding each year's investment, [I.sub.t], to the sum of assets from previous years, K[S.sub.t-1], less depreciation. That is, (9) K[S.sub.t] = (1 - [Delta])K[S.sub.t-1] + [I.sub.t], where [Delta] is the depreciation rate. Thus, the capital stock is calculated based on the widely used perpetual inventory method. However, there are problems with the data reported in the National Industrial Census. In particular, each year's investment is valued at current prices, and K[S.sub.t] and K[S.sub.t-1] are valued at original purchase prices. This means that changes in prices of capital goods will lead to a biased measure of capital stock. To obtain data on "real" capital stock we need data on prices of capital goods, but these data are not available. For this reason, we develop PK, an output-weighted price index, using prices of five capital and machinery goods producing industries as a deflator for the capital stock. The deflator is calculated by dividing gross output for these five industries in current prices by their gross output in 1980 constant prices.(7) Thus, our real capital stock, KR, is calculated as (10) K[R.sub.t] = [K.sub.t][S.sub.t]/P[K.sub.t]. It is generally agreed the appropriate measure of capital input for production and productivity analyses is capital services rather than capital stock; e.g., see Jorgenson and Griliches, [1972], and Berndt and Wood, [1975]. Because data to construct a series of service prices of capital (and hence value of capital services) are not available, we assume that the value of capital services equals the value of capital depreciation.(8) Thus, our proxy for capital services, K, is calculated as (11) [K.sub.t] = [[Delta].sub.t]K[R.sub.t.], where [[Delta].sub.t] is calculated by dividing total expenditure on capital depreciation by total values of capital stock in year t. Labor Input. Data on the total number of employees for each industry are available in the data set. However, this total number includes production-related workers as well as workers providing employee services such as education, health care, and related activities. We include only production-related workers in our measure of labor input by subtracting the number of workers providing services from total number of employees. Materials. Data on materials used in production are available only in current yuans. Because data on prices of materials are not available, we use the materials price deflators developed by Wang [1990]. Real materials input is calculated by dividing the value of materials used in production in current yuans by the Wang deflators. Output Shares of Inputs. Labor's output share is calculated by dividing total labor compensation (including wages, WA, social welfare and security expenditures, SW, and non-production capital services, [k.sub.o]) by gross output, Q, (in the gross-output models) and by value added, V, (in the value- added models).(9) That is, (12.a) [S.sub.qlt] = (W[A.sub.t] + S[W.sub.t] + [k.sub.ot])/[Q.sub.t] and (12.b) [S.sub.vlt] = (W[A.sub.t] + S[W.sub.t] + [k.sub.ot])/[V.sub.t]. Materials' output share is obtained by dividing the total value of materials, (VM), by the total value of gross output, VQ, (13) [S.sub.qmt] = V[M.sub.t]/V[Q.sub.t]. Capital's output share is calculated in different ways, depending on each particular model. For models in which we did not impose constant returns to scale, we calculated the capital share independently by (14.a) [S.sub.qkt] = (TAX + PROFIT)/V[Q.sub.t] and (14.b) [S.sub.vkt] = (TAX + PROFIT)/V[A.sub.t]. This calculation is based on the assumption that total returns to capital equal the sum of taxes and profits. In the constant-returns-to-scale models, we assume that output shares of inputs sum to unity and the share of capital is a residual. That is, (15.a) [S.sub.qkt] = 1 - [S.sub.qlt] - [S.sub.qmt] and (15.b) [S.sub.vkt] = 1 - [S.sub.vlt]. After obtaining [S.sub.qit] and [S.sub.vjt] (i = K, L, M, j = K, L), we use them to approximate the weights, [w.sub.qit] and [w.sub.vit], in the productivity equations as follows: (16.a) [w.sub.qit] = 1/2 ([S.sub.qit] + [S.sub.qit-1]) and (16.b) [w.sub.vit] = 1/2 ([S.sub.vit] + [S.sub.vit-1]). Details on the data and variable measurements are fully discussed in McGuckin et al. [1990] and can be obtained upon request. IV. THE RESULTS Sensitivity Analysis Table II reports output-weighted means of productivity growth rates and their intertemporal changes based on the twelve models. Three aspects of Table II are worth highlighting. First, we find that all six value-added models yield greater productivity declines (e.g., non-manufacturing, 1980-85) and larger productivity growth (e.g., manufacturing, 1984-85) than those obtained from the comparable gross-output models. These results are a striking confirmation of the expected biases represented in equation (4). Thus, it is evident that the two output specifications lead to substantially different estimates of productivity growth, and the value-added model overestimates its growth and decline. Second, we find that productivity growth estimates are qualitatively unaffected by use of different methods of obtaining output shares and output elasticities. For example, the results from all six gross-output models (I.a -III.b) show that productivity in manufacturing increased and that in non-manufacturing declined throughout 1980-85. Quantitatively, the differences are rather small. The estimated productivity growth rates based on different sets of output shares and elasticities lie in a narrow range. For example, the annual rate of productivity growth in non-manufacturing estimated by the six gross-output models ranges from -0.20 to -0.59 percent during 1980-84, whereas the corresponding estimates for manufacturing range from 0.19 to 0.74 percent. Given the fact that the output shares and elasticities used in the estimation of productivity growth differ substantially in magnitudes, these results are remarkable. (The output shares and elasticities are summarized in an appendix available from the authors upon request.) Finally, the impact of the constant-returns-to-scale restriction on the estimated productivity growth rates is minimal. In some cases, the estimate obtained from the constant-returns-to-scale model is identical to that obtained from the nonconstant-returns-to-scale model. Thus, it appears that for Chinese industries, constant-returns-to-scale can be imposed on the production function without causing severe biases in the model parameter estimates.(10)
Estimated Total Factor Productivity Growth Rates
(Output-Weighted Means)
Models 1980-1984 1984-1985 Change
I.a -.0048 -.0011 .0037
I.b -.0059 -.0049 .0010
II.a -.0038 -.0098 -.0060
II.b -.0020 -.0094 -.0074
III.a -.0026 -.0092 -.0066
III.b -.0026 -.0091 -.0065
IV.a -.0125 -.0119 .0006
IV.b -.0144 -.0137 .0014
V.a -.0133 -.0246 -.0113
V.b -.0124 -.0222 -.0098
VI.a -.0085 -.0189 -.0104
VI.b -.0087 -.0194 -.0107
I.a .0056 .0228 .0172
I.b .0019 .0201 .0182
II.a .0058 .0187 .0129
II.b .0058 .0194 .0136
III.a .0072 .0161 .0063
III.b .0074 .0165 .0091
IV.a -.0274 .0282 .0556
IV.b -.0279 .0289 .0568
V.a -.0275 .0280 .0555
V.b -.0220 .0333 .0553
VI.a -.0111 .0390 .0501
VI.b -.0116 .0384 .0449

In summary, the results based on the twelve models show that productivity growth estimates are very sensitive to output specifications and that the gross-output model is preferred. However, the use of output shares as proxies for output elasticities and imposing constant returns to scale do not appear to lead to important differences in productivity estimates. Total Factor Productivity Growth Estimates Because the results from the six gross-output models are similar we report and discuss only the results for individual industries obtained from model I.a. (Complete results are given in an appendix, Tables B1-6 available from the authors.) Table III exhibits the distribution of productivity and output growth across industries for the periods 1980-84 and 1984-85. For non- manufacturing industries, we find that in the years 1980-84, only three industries had positive productivity growth with an average annual rate of about 0.77 percent. The remaining eight industries experienced a productivity decline with an average annual rate of about -3.50 percent. Among these, the coke industry suffered the worst productivity decline, with a negative growth rate of -8.24 percent. In 1984-85, the number of non-manufacturing industries with positive growth increased to five with an average annual growth rate of 3.30 percent. The non-ferrous ore mining industry showed the most impressive improvement, with a productivity growth rate of 11.56 percent. The remaining six industries experienced negative growth at an average rate of -5.5 percent. The ferrous industry showed the worst productivity decline with a rate of -15.20 percent. Manufacturing industries appear to perform better. We find that sixteen of the twenty-eight manufacturing industries (57 percent) had positive productivity growth in 1980-84 with an average annual rate of 2.80 percent. The communications equipment industry had the greatest growth with a rate of 8.24 percent. The remaining twelve industries experienced a productivity decline averaging 2.05 percent per year. The "other" industry showed the worst decline (-5.30 percent per year). In the 1984-85 period, the number of manufacturing industries with positive growth increased to nineteen (68 percent). Among these, the fibers industry experienced the most rapid productivity growth, averaging about 14 percent in 1984-85 compared to 4.85 percent in 1980-84. If one judges productivity performance by looking at the change in productivity growth between 1980-84 and 1984-85, then twenty-seven of the thirty-nine industries had positive changes. Of the remaining twelve industries with negative changes, two industries (plastics and communications equipment) showed positive productivity growth in 1984-85, but at a decreasing rate. Thus by 1985, only three industries in non-manufacturing (coal, ferrous metal ore mining, and salt mining) and seven in manufacturing (food, tobacco, textiles, furniture, petroleum refinery, nonmetal work, and ferrous metal melting) experienced worsening productivity performance. In contrast to productivity growth, output grew for both manufacturing industries and non-manufacturing throughout the entire period 1980-85. However, output in manufacturing grew much faster than that in non- manufacturing (from 5.71 percent in 1980-84 to 9.57 percent in 1984-85 for manufacturing and from 0.98 percent to 1.79 percent for non-manufacturing). Table IV shows that in the 1980-84 period, output growth in both sectors is attributable to growth in capital and materials. The contributions of labor and productivity changes are close to zero. In comparison, manufacturing productivity grew by 2.28 percent, accounting for 24 percent of output growth during 1984-85. However, non-manufacturing productivity continued to decline in 1984-85. At the total industry level, the results indicate an apparent improvement in productivity growth with a rate of 2.17 percent in 1984-85 compared to 0.09 percent during the 1980-84 period. While we cannot directly compare our results with those reported in some published studies due to differences in units of analysis, models, data used, and periods examined, they are consistent with a number of previous aggregate studies. For example, Field [1984], Yeh [1984], Chen [1986], Chen and Sang [1986] and Pan [1986] reported that input accumulation accounts for nearly all the output increase in industry as a whole in the early 1980s. Our results, however, do differ from studies suggesting that Chinese industrial productivity increased drastically during the post-reform period. In particular, Kuan et al. [1988] reported that the average annual growth rate of total factor productivity in Chinese industries ranged from 2.7 to 3.1 percent for 1980-84 and from 15.3 to 18.2 percent for 1984-85.(11) There are several reasons why our results might differ from theirs, but the major reason is that the studies use two different data sets. We use census data for medium and large enterprises for three years: 1980, 1984, and 1985. Our data cover all three types of enterprises (i.e., state, collective, and private). In contrast, Kuan et al. used aggregate time-series data (1952-85) for independent accounting units within the state-owned industrial sector only, but including enterprises of all sizes (i.e., small, medium, and large). While these independent accounting enterprises are owned by the state, they are as equally independent from the state as collective and other enterprises in terms of decision-making, signing contracts, and seeking profits.(12) Thus, while the high estimates by Kuan et al. appear striking, they are not so surprising when compared to our 1984-85 estimates for the private sector. Our estimates based on the valued-added models (which were also used by Kuan et al.) range from about 10 percent to more than 13 percent, which are not far from Kuan et al.'s estimates. In any event, both studies show that the Chinese industrial sector experienced a clear improvement in factor productivity growth during 1984-85.
Contribution to Growth in Output: 1980-85(a)
(Results obtained from Model I.a)
Sector/Year(a) Capital Growth Labor Growth Materials
1980-1984 .0094 .0008 .0044
1984-1985 .0079 .0007 .0105
1980-1984 .0231 .0022 .0262
1984-1985 .0194 .0015 .0521
1980-1984 .0325 .0300 .0305
1984-1985 .0272 .0021 .0626
Source: Appendix Table B1 (available from authors).
a Entries are output weighted by mean of individual industry

V. SOURCES OF TOTAL FACTOR PRODUCTIVITY GROWTH Our analysis of the sources of economic growth is divided into two parts. First, we look at the effect of enterprise type on observed productivity growth. In the second part we examine other factors affecting productivity growth using cross-section analysis at the industry level. Enterprise Organization The recent emphasis on decentralization of economic decision-making in China follows widespread disillusionment with centralized control of production in China and around the world. We expect that collective and private enterprises would obtain greater growth rates in productivity than those in the state sector because the state sector has adopted the reforms much more slowly than the collective and private sectors. Because data by type of enterprise are available only at the total industry level, we must use output shares as proxies for output elasticities in our calculation of productivity growth for the three types of enterprises. An alternative procedure is to use the output elasticities estimated in the cross-section models specified in Table I to calculate total factor productivity growth rates for the three types of enterprise. A weakness of this procedure compared to using the output shares is that it requires the assumption that all enterprise types have identical output elasticities. This assumption is unlikely to be correct. For this reason, we report in Table V only the results obtained from the gross-output model with calculated output shares used as the weights (model I.a).(13) Table V shows that total factor productivity in all three types of enterprise grew throughout the 1980-85 period. However, productivity growth in the collective and private sectors is more than four times greater than that observed for the state sector in 1980-84. Most strikingly, during 1984-85, while productivity in the state and collective sectors grew at a similar rate (about 1.5 percent), productivity in private enterprises continued to grow rapidly at a rate of 3.26 percent. These results suggest that collective and private enterprises have been able to increase their productivity in the era of economic reforms far better than state-owned enterprises. In light of the advantages that state-owned enterprises are purported to have--access to state-allocated inputs at lower prices and relatively better manufacturing facilities--these findings are striking.
Productivity Growth Rates by Type of Enterprise
(Results obtained from Model I.a)
Growth Rates and
Enterprise Type Period Changes
Output Growth(*) 1980-1984 1984-1985
Total .0673 .1155 .0482
State .0651 .1118 .0467
Collectives .1766 .2505 .0739
Private .1496 .2418 .0922
Productivity Growth(*)
Total .0060 .0166 .0106
State .0052 .0152 .0100
Collective .0387 .0168 -.0219
Private .0280 .0326 .0046
Weighted Capital Growth(*)
Total .0280 .0346 .0066
State .0278 .0346 .0068
Collective .0399 .0411 .0012
Private .0372 .0339 -.0033
Weighted Labor Growth(*)
Total .0030 .0021 -.0009
State .0029 .0021 -.0008
Collective .0053 .0027 -.0026
Private .0051 .0020 -.0031
Weighted Materials Growth(*)
Total .0303 .0621 .0318
State .0292 .0599 .0307
Collective .0928 .1899 .0971
Private .0793 .1733 .0940
* Data for state, collective, and other enterprises are
available only at the aggregate level and add up to the total
of forty-three industries. Thus, the figures for total in this
table are not equal to those reported in Tables III and IV
which use only thirty-nine industries.

Inter-Industry Differences In examining factors linked to observed differences in the estimated productivity growth rates across industries, we focus on two types of variables. The first includes the proportion of engineers and technical employees to total employment (ENG/L) and the number of computers per employee (COMP/L). We hypothesize that if technical employees and computers are efficiently employed, then industries that employ a high proportion of technical employees and large numbers of computers per employee would obtain high productivity growth. We note, however, that if these resources are not efficiently allocated and employed, then increases in these factors do not necessarily lead to increases in productivity. Thus, the coefficients of (ENG/L) and (COMP/L) could be either positive, zero or negative. The second type of variable includes two variables associated with material incentives to enterprises. The first is the percentage change in profits retained by the enterprise taken as a proportion of capital assets (% Prof/K). The second is the percentage change in bonus wages per employee (% B/L). Because retained profits and bonus wages, introduced by the reform, were intended to reward performance, we expect these variables to have a positive effect on productivity growth. However, according to a number of economists, in state-owned industries bonuses were often provided indiscriminately to all workers regardless of productivity. Similarly, retained profits are directly linked to total profits which, in turn, are often affected by factors extraneous to the firms, in particular prices.(14) This indicates that the coefficients of the profit and bonus variables could be negative. To test the above hypotheses, we run regressions in which the productivity growth rate is the dependent variable, and (ENG/L), (COMP/L), (% B/L), and (% Prof/K) are independent variables. Note that this analysis does not determine causality. For example, the simple regression analysis we use is not able to distinguish between the hypothesis that bonus payments lead to higher productivity growth and the hypothesis that bonuses are a reward for past productivity. With this caveat we proceed to the regression results. Table VI reports OLS estimates of the regressions. The dependent variable, productivity growth rate, is estimated using the twelve models. We note that we excluded the computer variable from our final regressions because of serious collinearity between the computer and engineer variable (r = 0.85). The coefficient of each of the two variables is positive and significant when either variable is included in the regressions alone; but when both are included in the regressions, the coefficient of the computer variable becomes negative and insignificant. This suggests that collinearity exists between the (ENG/L) and (COMP/L) variables. Table VI shows that all the coefficients for (ENG/L) and (PROF/K) are positive and significant, while that for the (% B/L) variable is significantly negative in most regressions. These estimates are robust in that they are invariant with respect to output specifications and the magnitudes of the estimated output elasticities. These results strongly suggest that allowing enterprises to retain a portion of their total profits is an effective measure to promote productivity growth. However, the current Chinese bonus system appears to have a negative effect on productivity growth. This result appears to be consistent with the observation that bonuses were often distributed equally among all workers regardless of productivity. Finally, an increase in the proportion of technical employees appears to have a significant, positive effect on productivity growth. While this result is reasonable, it should be interpreted with caution because our labor input data are not adjusted for labor quality. The positive correlation between productivity growth and the (ENG/L) variable could be a consequence of the quality unadjusted labor data.(15) VI. SUMMARY AND CONCLUDING REMARKS This study uses twelve competing models to estimate and analyze Chinese industrial productivity growth and identify its sources by industry and by type of enterprises for the post-reform period 1980-85. We use data extracted from the new National Industrial Census of China. This data set makes it possible to examine economic growth and determine its sources at a level of detail not previously possible. While the data set is valuable, as with most data sets it is far from perfect. One limitation is that it does not contain information on input prices. As a result, real inputs cannot be measured accurately. Most importantly, the data set does not provide the data necessary for constructing an accurate and theoretically sound measure of capital services. Instead, it provides data on capital stocks valued at original purchase prices. This inaccurate measure of capital stocks would lead to biased estimates of growth rates and the output share of capital input. Equally important, the lack of information on the relationship of labor compensation and the allocations of capital and other inputs to production and non- production in China makes it difficult to estimate the sources of productivity growth. In spite of data problems, our results are relatively robust. We find that our estimated productivity growth rates are strongly affected by output specification and that the value-added model yields biased estimates of productivity growth. However, the impact of differing values of output elasticities is rather small. Several important results emerge when examining the results from the theoretically correct gross-output models. First, the estimates obtained from all models show that total factor productivity growth strongly increased in manufacturing and declined in non-manufacturing. Second, the results show substantially higher productivity growth by collective and private enterprises relative to state-owned enterprises in the 1980-84 period. However, during 1984-85, while state-owned enterprises began to approach a productivity growth rate equal to that of collective enterprises, private enterprises continued to lead with a rate that was two times greater than those in other enterprises. Thus, the evidence suggests that private enterprises were are able to increase their productivity in the era of economic reforms far better than were state-owned and collective enterprises. Third, we find evidence that enterprises employing more technical employees were likely to obtain increases in productivity growth. We also find that retained profits are positively correlated with productivity growth. Finally, bonus payments to labor, as currently applied, appear to have a negative effect. While the above conclusions are drawn with a certain degree of confidence, we emphasize that they are by no means definite. This is due in part to limitations in the data used, and the constructed variables, such as capital input, may contain measurement errors which would lead to biased productivity growth estimates. These limitations suggest several areas for additional research. One important area is to collect more data at the level of individual industries to estimate separate industry output elasticities. Analysis at a more disaggregate level, as well as use of provincial data from China's industrial census, should help in this regard. Similarly, further research on pricing and capital valuation is needed. International price comparisons, such as those undertaken by the United Nations' International Comparisons Project, might be used to revalue capital and output or at least to provide a basis for assessing the bias in the factor share estimates. 1. Among these are Rawski [1984], Lardy [1989], and Perkins [1988]. Because these reforms are amply discussed elsewhere (e.g., see Field [1984], Tidrick and Chen [1987], Naughton [1986], Byrd [1987], and Wu and Reynolds [1988]) it would be redundant to discuss then in detail here. Instead, we provide a summary of the major reform measures applied in China during 1978-85. 2. Collective enterprises are owned by "collectives," including counties, cities, towns, neighborhoods, etc. State enterprises are owned and operated by government departments, army units, scientific research institutes, etc. 3. This methodology traces back to the pioneering work of Solow [1957] and later is used in many productivity studies such as those in Denison [1967], Griliches and Jorgenson [1967], and Lieberman, Lau, and Williams [1990]. 4. See Perkins [1988], and Kuan et al. [1988]. 5. The materials-output ratio for the Chinese industrial sector declined throughout 1980-85 with values of .5850, .5847, and .5428 for 1980, 1984, and 1985. 6. In his study, Perkins states that only an implausibly high estimate for the output share of capital (i.e, .07 or .08) would alter his conclusions about the post-reform productivity performance of the Chinese economy. In fact, he found that "even such a high elasticity, if it existed earlier as well, would not change the conclusion that the rates of growth in productivity were higher after the reform than before" [1988, 629]. 7. These five industries are (1) engineering, (2) traffic transport equipment, (3) electric equipment and appliances, (4) electronic and telecommunications equipment, and (5) instruments and meters. 8. We note that depreciation expenditures may be arbitrary and depend on accounting practices, but our calculations show that the depreciation rates for Chinese industries are rather stable during 1980-84. An alternative approach is to assume that capital services are proportional to the stock of capital and use capital stock as a proxy for capital input. However, this will not alter our results because, as shown in equation (11), our measure of capital input, [K.sub.t], is proportional to the capital stock, K[R.sub.t]. 9. Non-production capital includes that providing services to employees and their families such as living quarters, schools, nurseries, hospitals, etc. Because data on the value of non-production related capital services, [k.sub.o], are not available, we use depreciation expenditures on non-production related capital as a proxy for [k.sub.o]. 10. We have also calculated total factor productivity growth rates and their intertemporal changes for each of the thirty-nine industries using the twelve competing models. However, the results for individual industries do not alter our conclusions regarding the sensitivity of the estimated productivity growth rates based on the weighted means reported in Table II. Because of space limits, we do not report the detailed results here, but interested readers can obtain them from the authors upon request. 11. These estimates are based on their revised data, Table IV, p. 583. 12. An enterprise is classified as an independent accounting unit if (1) it has an independent administrative organization, (2) is able to account independently for its profits and losses, and (3) is allowed to sign contracts with other units and to open an independent bank account. 13. Despite our concern about the assumption of equal elasticities across enterprise types, we estimated total factor productivity growth rates for the three types of enterprises based on the twelve models presented in Table I. The results are consistent with our earlier findings based on data for thirty-nine industries, suggesting that calculated output shares are preferred for this analysis. 14. These points are made by Balassa [1987] and Ma [1983]. 15. Unadjusted labor data could lead to our underestimating the labor input and its growth rate and, therefore, overestimating productivity growth. In this case, the estimated productivity growth rates would be positively correlated with increases in the proportion of high-quality labor (i.e., engineer and technical employees). We are indebted to an anonymous referee for this argument. Unfortunately, data on the number of engineers and technical employees are available for 1985 only. Therefore, we are unable to use this information to adjust labor quality for all three years. REFERENCES Baily, M. N. "Productivity Growth and Materials Use in U.S. Manufacturing." Quarterly Journal of Economics, February 1986, 185-95. Balassa, B. "China's Economic Reforms in a Comparative Perspective." Journal of Comparative Economics, September 1987, 410-26. Berndt, E. R., and D. O. Wood. "Technology, Prices, and the Derived Demand for Energy." Review of Economic and Statistics, August 1975, 259-68. Byrd, W. A. "The Impact of the Two-Tier Plan/Market System in Chinese Industry." Journal of Comparative Economics, September 1987, 295-308. Chen, S. "Sources of Economic Growth: Changing Input Structure in Chinese Industry Since the 1970s," in Theory and Practice of Evaluating Technical Progress, edited by Z. Wang. Beijing: Kexue Jishu Wenxian Chubanshe, 1986, 171-89. Chen, S., and G. Sang. "The Problem of Total Factor Productivity in Economic Growth Analysis of China's Industrial Sector Since the 1970s," in Theory and Practice of Evaluating Technical Progress, edited by Z. Wang. Beijing: Kexue Jishu Wenxian Chubanshe, 1986, 57-71. Denison, E. F. Why Growth Rates Differ: Post-War Experience in Nine Western Countries. Washington D.C.: The Brookings Institute, 1967. -----. "Final Comments." Survey of Current Business, May 1972, 95-109. Field, R. M. "Changes in Chinese Industry Since 1978." The China Quarterly, December 1984, 742-61. Griliches, Z., and D. W. Jorgenson. "The Explanation of Productivity Change." Review of Economic Studies, July 1967, 249-83. Jefferson, G. H. "A Decade of Reform in China: Pitfalls and Lessons in Evaluating Industrial Productivity." Paper presented at the ASSA Annual Meeting, December 1988. Jorgenson, D. W., and Z. Griliches. "Issues in Growth Accounting: A Reply to Edward F. Denison." Survey of Current Business, May 1972, 65-94. Kuan, E., J. Chen, H. Wang, Y. Zheng, G. Jefferson, and T. Rawski. "Productivity Changes in Chinese Industry, 1953-1985." Journal of Comparative Economics, December 1988, 570-91. Lardy, N. R. "Technical Change and Economic Reform in China: A Tale of Two Sectors." Unpublished paper, March 1989. Lieberman, M. B., L. J. Lau, and M. K. Williams. "Firm-Level Productivity and Management Influence: A Comparison of U.S. and Japanese Automobile Producers." Management Science, October 1990, 1193-215. Ma, H. New Strategy for China's Economy. Beijing: New World Press, 1983. McGuckin, R. H., S. V. Nguyen, J. R. Taylor, and C. A. Waite. "Multifactor Productivity in Chinese Industry: 1980-85," in An Industrial Census of China. Beijing: The Secretariat of the International Seminar of Analysis on China's 1985 Industrial Census, March 1990, 113-38. McMillan, J., J. Walley, and L. Zhu. "The Impact of China's Economic Reforms on Agricultural Productivity Growth." Journal of Political Economy, August 1989, 781-807. Naughton, B. "Finance and Planning Reforms in Industry," in China's Economy Looks Toward the Year 2000, vol. 1, Washington, D.C.: U.S. Congress Joint Economic Committee, 1986, 604-29. -----. "Macroeconomic Policy and Response in the Chinese Economy: The Impact of Reform Process." Journal of Comparative Economics, September 1987, 334-53. Pan, G. "Scientific and Technological Progress in the Development of the Chemical Industry," in Theory and Practice
COPYRIGHT 1993 Western Economic Association International
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1993 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:McGuckin, Robert H.; Nguyen, Sang V.
Publication:Economic Inquiry
Date:Jul 1, 1993
Previous Article:Job security and job search in more than one labor market.
Next Article:Promotions as work incentives.

Terms of use | Copyright © 2017 Farlex, Inc. | Feedback | For webmasters