2 How Fast Can China Grow?

Wanda Tseng, and Markus Rodlauer
Published Date:
February 2003
  • ShareShare
Show Summary Details
Paul Heytens and Harm Zebregs 

Over the past two decades, China has been one of the fastest-growing countries in the world.1 Growth in real GDP per capita has exceeded 8 percent a year on average, an impressive achievement even by East Asian standards. China’s exceptional growth performance has been spurred by market-oriented structural reforms introduced since the late 1970s, which have also resulted in sizable productivity gains. However, annual GDP growth has declined since 1993, to 7–8 percent in 1999–2001, raising questions about to what extent China’s impressive track record can be sustained in the future. Although most analysts believe some slowdown is inevitable, it is difficult to assess China’s future growth potential given the deep structural reforms under way, the vast scale of the Chinese economy, and the complexities of its growth dynamics.

This chapter examines the dynamics of China’s growth performance and attempts to shed light on how fast the country might grow in the future. In contrast with earlier empirical work, it seeks to identify and estimate the contributions of the sources of total factor productivity (TFP) growth, which explains most of the increase in output growth during the reform period. The chapter shows that China’s growth performance has been underpinned by a large-scale reallocation of agricultural labor to more productive uses and by other productivity-enhancing structural changes.

The main policy implication that emerges is that China’s growth prospects will continue to depend on structural reform. However, many of the earlier reforms had a one-time impact on productivity growth, which suggests that future growth is unlikely to match that of the past two decades. In addition, future structural reforms—involving considerable labor shedding by state-owned enterprises (SOEs) and the shutting down of nonviable enterprises and outdated production capacity—could cause some disruptions in the labor market and slow the absorption of agricultural labor in the short term. Nevertheless, potential output should be able to grow at 7–8 percent a year over the medium term, provided reforms of the SOEs and the financial sector accelerate following China’s entry into the World Trade Organization (WTO).

Structural Reform

Overview of the Reform Period

China’s reform strategy over the past 20 or more years can best be described as incremental. Reforms followed a dual-track approach in which a market track was established in parallel with the preexisting tracks of the centrally planned economy, with the former gradually increasing in importance over time.2 This dual-track approach was initiated in late 1978 with the rapid and comprehensive liberalization of the agriculture sector.

The agricultural reforms quickly allowed a substantial proportion of economic activity and of the labor force to move outside of central planning. The main feature of the agricultural reform was the replacement of the commune-brigade system of collective farming with a system in which communal land was leased to individual peasant households. Although households remained responsible for delivering a portion of their farm output to the state, they were allowed to sell any production above the state procurement quota on the free market. The quota was essentially a lump-sum tax, so that rural households effectively faced market-determined prices in making their production decisions.

The impressive growth of agricultural output resulting from the initial market-oriented reforms also facilitated the liberalization of the industrial and services sectors in rural and urban areas that followed in the 1980s. The growth of the nonstate sector was also allowed to spread well beyond the communes themselves. The regulations governing the registration and supervision of nonstate enterprises were progressively liberalized beginning in 1984, and the result was rapid industrialization, particularly by community-owned enterprises in rural areas, which came to be known as township and village enterprises (TVEs). Thus the dismantling of the communes and the emergence of the TVEs exposed nearly 800 million rural inhabitants (some 80 percent of the population at the time) to market forces and provided a large proportion of them the opportunity to leave agriculture over a relatively brief time span.

Another purpose of the dual-track approach was to integrate China into the global economy. In 1980 four southern coastal cities—Shantou, Shenzhen, Xiamen, and Zhuhai—were designated as special economic zones (SEZs). They were provided certain discretionary powers over taxation and were given autonomy to experiment with new institutional forms, such as foreign-funded enterprises. In addition, SOEs operating in the SEZs were exempted from many elements of the central plan, such as certain labor regulations and the tax code. With the phenomenal initial growth of the SEZs, the privileges they were granted spread quickly to other areas, including Hainan, which became the fifth SEZ in 1988 (Chapter 5).

In contrast to reform in agriculture, the pace of reform in the SOE and financial sectors was initially more gradual. SOEs were not privatized, and indeed, they only became a key focus of reform in the mid-1990s. Instead the SOEs were the target of various attempts to introduce more market-based incentives to improve management and operational efficiency within the framework of state ownership. Beginning in 1984, decision-making power over production, marketing, and investment matters was incrementally devolved to SOE managers. However, it was not until Deng Xiaoping’s celebrated tour of the southern provinces in January 1992 that the authorities formally embraced the view that the market system was not incompatible with socialism, and this change set the stage for a deepening of SOE and financial sector reform.3 The initial focus was on moving small enterprises out of the state sector, placing large enterprises on a commercial footing, and shifting the burden of lending for policy purposes away from the state commercial banks. Following the Fifteenth Party Congress in September 1997, these reforms were expanded and accelerated.

China’s reforms have facilitated a profound transformation of the country’s economic structure and social development:

  • The proportion of the labor force engaged in agriculture has fallen from nearly three-fourths in the late 1970s to less than half, and the proportion of industrial output produced by SOEs declined from nearly 80 percent to about one-quarter over the same period.

  • Total trade (imports plus exports) rose from less than 10 percent of GDP in the late 1970s to just over 40 percent, and foreign direct investment (FDI) inflows have risen from virtually zero at the beginning of the 1980s to over $40 billion a year in recent years.

  • Human development indicators—including life expectancy, literacy, infant mortality, and income per capita—have also improved dramatically. Indeed, the rapid growth and structural change witnessed in China over the past 20 years have delivered perhaps the greatest reduction in poverty in recorded history: 200 million people have ceased to be poor, where poverty is measured by China’s official poverty line; the corresponding figure based on the World Bank’s criterion (living on $1 or less a day) is considerably higher.

Although China’s gradual approach to economic reform has had considerable success, some analysts have pointed out that it has also incurred significant costs (Lardy, 1998). In particular, the slow pace of SOE reform, although arguably necessary to maintain social stability, has contributed to growing SOE losses. These have largely been financed through the banking system and have led to a sharp deterioration in asset quality, reflected in rising nonperforming loans. In the past several years, owing to mounting problems in these areas and lessons learned from the Asian crisis, reform of the SOEs and the financial sector have moved to the top of the authorities’ policy agenda (Chapters 9 and 10).

The Contribution of Structural Reform to Output Growth

Although little empirical work has been done to estimate the contribution of structural reform to growth directly, a number of growth accounting studies have attempted to measure this contribution indirectly.4 Three main channels have been identified:

  • Raising TFP growth directly. The literature agrees on the strong linkage between structural reform and the rapid productivity growth over the past two decades. TFP growth was found to be particularly high following the liberalization of the agricultural sector in the early 1980s, and in the early 1990s after marketoriented reforms were accelerated, and to have been well above that of the prereform period (1952–78).5 Estimates of TFP growth during the reform period range between 2 and 4 percent a year.6

  • Raising aggregate TFP growth through facilitating a more efficient allocation of labor. A reallocation of labor from agriculture to other sectors increases aggregate output if the marginal product of labor is lower in agriculture than in the other sectors. A study by Chow (1993) indicates that there was indeed scope for efficiency gains through labor reallocation at the beginning of the reform period: Chow estimated that, in 1978, the marginal product of labor was only 63 yuan in agriculture but Y 1,027 in industry, Y 452 in construction, Y 739 in transport, and Y 1,809 in commerce. The empirical estimates of the impact of labor reallocation on TFP growth generally fall into a range of ½–2 percentage points, or up to about half the estimated productivity growth.

  • Enhancing the efficiency of capital accumulation and knowledge spillovers. An economy that is more efficient at mobilizing and transforming savings into physical and human capital can realize higher growth rates. The increased marketization and opening of the Chinese economy are widely believed to have improved the efficiency of capital accumulation during the reform period. The rapid increase in FDI is considered to be a particularly important explanatory variable, as is the substantial spillover of technology and managerial know-how from the large number of joint ventures and wholly owned foreign enterprises (Chapters 5 and 6).

Potential Output and Its Components

This section reports the results of several different empirical approaches to identify the main determinants of China’s potential output growth over the period 1970–98.7 Because TFP growth has been identified as a major contributor to the increase in output growth during the reform period, it is important to examine its determinants, in particular, structural reform and labor migration from agriculture to other sectors.

The Hodrick-Prescott Filter

A commonly applied approach to decomposing actual output into its long-run potential and cyclical components is the Hodrick-Prescott (HP) filter. This is a statistical method that does not use any information regarding the determinants of each of the components, but is a useful first approximation of potential output growth. Let Yt and Ŷt be the logarithms of, respectively, actual and potential output at time t; then the cyclical components are given by εt = YtŶt. Hence HP filtering decomposes Yt into Ŷt and εt.

When applied to China’s real GDP, the HP approach reveals that potential output growth picked up strongly during the reform period (Figure 2.1). It peaked at 9.6 percent in 1985–87 and then slowed somewhat in the late 1980s. In the first half of the 1990s, when the reform process was reinvigorated, potential output growth rebounded, peaking at 9.9 percent in 1994–95, but it has since tapered off again.

Figure 2.1.Actual and Potential Output Growth and the Output Gap

(In percent)

Source: Authors’ calculations.

1Calculated using Hodrick-Prescott filtering.

However, the HP approach suffers from what has been called the endpoint problem: future potential output growth may be overestimated if actual output growth was comparatively high (or underestimated if it was comparatively low) at the end of the sample period. In the case of China, this could mean that the slowdown in potential output in the second half of the 1990s was actually more pronounced than the HP filter suggests. This would also imply that the estimated output gap is smaller.

The Production Function Approach

Another approach that is often used to determine potential output is the production function approach, which makes use of information regarding the sources of growth, namely, factor accumulation and TFP growth. The production function approach is described in Appendix II and can be represented as follows:

where At represents TFP, βt is a coefficient vector, and zt is a vector of factor inputs, all measured at time t. Potential output is given by Ŷt = At + βtzt. Often, zt contains only capital and labor, and when constant returns to scale are imposed, the elements in βt must sum to 1.

Neither At nor βt can be directly observed, but there are several ways to calculate TFP and the other parameters of the production function. If constant returns to scale are imposed and the only factor inputs are capital and labor, the parameters can be directly obtained from the national accounts by calculating the share of labor income in GDP; 1 minus this share is then the capital share parameter. The next step is to substitute these parameters into the production function and calculate the Solow residuals: Yt – βtzt = At + εt. These residuals are the sum of TFP and the cyclical components in actual output. The time series of Solow residuals can be decomposed into At and εt through filtering or by regression on a set of variables, xt, that are sources of TFP growth.

Hu and Khan (1996) constructed a time series of labor shares using Chinese national accounts data and a translog production function to calculate TFP growth for the 1953–94 period. Extending their series of labor shares to 1998, we calculated the contributions to potential output growth of capital accumulation, labor force growth, and TFP growth for 1971–98 (Figure 2.2; Table 2.1). In the prereform period, capital accumulation was the main contributor to potential output growth. The contributions of labor force growth and TFP growth were small and negative, respectively. However, TFP growth picked up strongly during the reform period and explains much of the increase in potential output growth during this period. Output gaps were also calculated on the basis of the production function approach, the pattern of which is very similar to that of the output gaps derived with the HP filter.8

Figure 2.2.Potential Output Growth and the Output Gap: Estimates from a Translog Production Function

(In percent)

Source: Authors’ calculations.

1Contribution to potential output growth.

Table 2.1.Contributions to Output Growth

(In percent of GDP)1

Translog production function2
Potential output5.79.29.3
Capital accumulation4.85.24.7
Labor force growth0.71.20.7
TFP growth0.22.84.0
Output gap–0.3–0.20.2
Cobb-Douglas with exogenous TFP growth3
Potential output5.79.29.5
Capital accumulation4.75.66.3
Labor force growth0.71.10.5
TFP growth0.32.52.7
Output gap–0.3–0.1–0.0
Cobb-Douglas with endogenous TFP growth4
Potential output4.99.39.5
Capital accumulation4.85.76.4
Labor force growth0.71.00.5
TFP growth–
Output gap0.5–0.20.0
Actual output growth5.49.19.5
Source: Authors’ regressions.

Period averages.

Based on national accounts-based factor shares from Hu and Khan (1996) supplemented with authors’ estimates for 1995–98.

Based on constant factor shares from Chow (1993).

Based on coefficients from the regression estimated in Table 2.2.

Source: Authors’ regressions.

Period averages.

Based on national accounts-based factor shares from Hu and Khan (1996) supplemented with authors’ estimates for 1995–98.

Based on constant factor shares from Chow (1993).

Based on coefficients from the regression estimated in Table 2.2.

Another way of calculating βt and TFP is by regressing Yt on zt and xt. In this case, direct estimates of At, as a function of xt, and εt are obtained. A simple version of this approach is to estimate

where yt and kt are the logarithms of output per worker and capital per worker, respectively; C is a constant; and t represents a linear trend. In this specification At = C + δt. Chow (1993) and Chow and Li (1999) used this approach and found a capital share, α, of 0.63. This direct estimate of the capital share parameter exceeds that derived from the national accounts, which is estimated to have been 0.56 on average during 1970–98. Although both estimates are high compared with capital share parameters found in industrial countries, they are not outside the range found in other developing countries.9 Nevertheless, the possibility that they are biased upward cannot be ruled out. This may be the result of omitted variables, in particular human capital, or of measurement error in China’s national accounts. The latter tend to overestimate output in the capital-intensive manufacturing sector and underestimate the output of more labor-intensive private and individual enterprises. It is also possible that the high capital share is a legacy of the previous plan system, which was strongly biased toward investment in capital -intensive heavy industries and was perpetuated through the continuation of state control of the banking system during the reform period.

When China’s output growth is decomposed under the assumption that α = 0.63, the contribution of capital accumulation is larger than when the implicit capital shares from the national accounts are used (Figure 2.3). The Solow residuals are therefore also sensitive to the relative magnitudes of the factor shares. In the case of China, a higher capital share reduces the residual and hence the contribution of TFP growth; during the reform period, TFP growth was 2–2¾ percent a year assuming a constant capital share of 0.63, and 3–4 percent assuming the average capital share from the national accounts of 0.56 over 1970–98 (Table 2.1).

Figure 2.3.Potential Output Growth and the Output Gap: Estimates from a Cobb-Douglas Production Function with Exogenous TFP Growth

(In percent)

Source: Authors’ calculations.

1 Contribution to potential output growth.

Although the two approaches yield different conclusions about the level of TFP growth, both indicate that TFP growth picked up during the reform period and made a large contribution to the increase in output growth. However, neither approach can explain what caused the pickup in TFP growth. For this it is necessary to estimate an aggregate production function for the Chinese economy that, besides capital and labor, includes explanatory variables for TFP (Table 2.2). Apart from a trend and a trend dummy, explanatory variables in the regression include a structural reform index and a proxy for labor movements out of the primary (largely agricultural) sector.10 The reform index, discussed in Appendix III, is constructed from four variables—the nonstate share of industrial output, the share of total trade in output, the level of urbanization, and the rate of capital formation—that together capture the impact of structural reform on the economy.

Table 2.2.Estimated Aggregate Production Function for the Chinese Economy

Capital stock per worker20.6430.0709.196
Structural reform index30.2000.0663.035
Percent of labor force in primary sector–0.0190.003–6.677
Trend dummy, 1979–980.0310.0039.690
Summary statistics
R2 = 0.999
Durbin-Watson = 1.83
Source: Authors’ regressions.

The dependent variable is real GDP per worker (in logarithms); the sample period is 1970–98.

At constant prices (in logarithms). Variables are nonstaionary but cointegrated: the time series of the residuals is I(0).

In logarithms.

Source: Authors’ regressions.

The dependent variable is real GDP per worker (in logarithms); the sample period is 1970–98.

At constant prices (in logarithms). Variables are nonstaionary but cointegrated: the time series of the residuals is I(0).

In logarithms.

Table 2.3 shows the estimated contributions to TFP growth of structural reform, labor migration, and technological progress. What stands out is that labor migration has been a dominant factor in TFP growth both before and during the reform period. On an aggregate level, efficiency-enhancing labor migration shows up in TFP, whereas on a more disaggregated level it would be reflected in the contribution of labor to sectoral output growth. Hence, although the contribution of labor force growth has been small at the aggregate level (Figure 2.4), the reallocation of labor across sectors has been a significant contributor to output growth in the manufacturing and services sectors. This source of TFP growth has not yet been exhausted—the proportion of workers in the primary sector, 50 percent in 2000, is still high compared with more advanced economies—but it will certainly become less important in the long run. The contribution of structural reform has been positive and has added, on average, ¾–1 percentage point to TFP growth and hence to output growth.11 The exogenous trend has been negative but has become less so during the reform period, and it might be reflecting aspects of structural reform that the reform index does not capture. These results, together with the estimate of the capital share parameter of 0.64, which is close to Chow’s (1993) estimate of 0.63, suggest that capital accumulation and efficiency improvements from labor reallocation and structural reform have been the main engines of growth in China. Because part of TFP growth was driven by one-time level adjustments, and because capital accumulation will taper off in the future as the Chinese economy matures, potential output growth can be expected to decline over the medium term.

Table 2.3.Contributions to TFP Growth

(In percent of TFP)1

TFP growth–0.532.782.112.812.30
Structural reform0.380.940.760.830.39
Labor migration out of primary sector2.342.011.522.152.08
Exogenous trend–3.25–0.17–0.17–0.17–0.17
Source: Authors’ regressions.

Period averages, based on estimation results in Table 2.2.

Source: Authors’ regressions.

Period averages, based on estimation results in Table 2.2.

Figure 2.4.Potential Output Growth and the Output Gap: Estimates from a Cobb-Douglas Production Function with Endogenous TFP Growth

(In percent)

Source: Authors’ calculations.

1Contribution to potential output growth.

The Growth Outlook

Most analysts, as well as the Chinese authorities themselves, who are projecting average annual growth of 7 percent in the current five-year development plan (2001–05), believe that some slowdown in GDP growth is inevitable. The empirical findings described above provide support for this view. At the same time, however, China is still a relatively poor country and is likely to be able to continue catching up for many years to come. As neoclassical theory suggests, countries such as China that are relatively less developed and distant from the technological frontier in the industrial countries have a capacity for rapid growth if they mobilize and allocate physical and human capital effectively, adapt foreign technology to their factor proportions, and make good use of the opportunities for specialization that come from closer integration with the global economy.

Structural Policies for Future Growth

The discussion thus far in this chapter points to the necessity of intensifying and broadening economic reforms as a necessary condition for sustaining future growth of 7–8 percent a year. The reform areas most critical to China’s growth outlook are the following:

  • Reform of the financial and SOE sectors. Much remains to be done to strengthen the performance and financial health of the SOEs and the commercial banking system. In particular, extensive reforms to upgrade operational efficiency and product quality, to harden budget constraints, and to improve corporate governance in the SOEs are needed in the coming years. Similarly, bank rehabilitation will require far-reaching reforms, including the operational restructuring of the state commercial banks, the elimination of policy lending, and the development of a credit culture. The closely intertwined reforms of the SOEs and the financial sector must deepen if the efficiency of capital allocation in the Chinese economy is to be improved on a durable basis. SOE reform will also lead to scrapping of excess capacity, reflected in temporarily higher depreciation rates and slower output growth, but will result in a pickup in TFP growth over the medium term as the modernization of enterprises progresses and more productive capital is installed. Structural reform is expected to accelerate as a result of WTO accession.

  • Development of the nonstate sector. Realizing the nonstate sector’s potential to absorb workers laid off by SOEs, the authorities are attempting to foster the development of this sector, including by elevating the constitutional role of the private sector to one of parity with the state sector, and by taking some initial steps to improve nonstate firms’ access to bank credit and the stock market. Continued efforts to encourage private sector development will be necessary in the years ahead, for example reducing sectoral barriers to entry and providing better protection of private property rights by deepening the rule of law.

  • Increased factor mobility. The level of urbanization in China is very low, partly because of controls on the movement of labor from rural to urban areas. Some tentative steps have been taken recently to encourage the development of medium-size cities and towns, and the controls on the movement of labor into urban areas have been relaxed. However, China’s labor market will need to become considerably more flexible, and medium-size cities and towns will need to be developed, in order to facilitate the continued reallocation of the rural labor force to more productive uses. In addition to limiting the mobility of labor, government policies have also had a dampening impact on the mobility of capital across provinces (Zhao, 1998). Capital is likely to become more mobile in the future as internal barriers to trade and investment are dismantled following WTO accession and as financial sector reforms proceed.

A Future Growth Scenario

The empirical results of this chapter suggest that future TFP growth and rates of capital accumulation will be below their averages over the past two decades. In addition, the next steps in the reform process, which involve more labor shedding by SOEs and the shutting down of nonviable enterprises and outdated productive capacity, will cause somewhat greater disruption in the short to medium term than did the earlier reforms. Accordingly, the deceleration of TFP witnessed since the mid-1990s is likely to continue over the next several years as reforms, in particular the restructuring of SOEs, intensify. The average rate of investment is also likely to decline somewhat in the next few years as excess capacity is scrapped and the effective rate of depreciation of the capital stock increases.

The direct contribution of labor to output growth, which has been modest throughout the reform period, will become even more so as population growth continues to slow as a result of China’s population control policies. In addition, the further slowing of labor force growth and the rapid aging of China’s population will push up the dependency ratio, particularly after 2010, and cause the saving rate to fall. Labor’s indirect contribution to growth (reflected in TFP growth) through reallocation of workers from rural (agriculture) to urban areas (industry and services) should remain substantial for some time to come, although perhaps not quite as significant as in the past. However, an intensification of reform—in particular, labor shedding by SOEs—will reduce the capacity of urban areas to absorb labor from the countryside in the short and the medium run.

Although the effects of future reform are difficult to quantify, a possible growth scenario based on the estimates in this chapter is summarized in Table 2.4. Under this scenario, TFP growth remains below past levels as a result of a slowing of the outflow of labor out of agriculture (proxied by a slower decline in the primary sector’s share of the labor force), as layoffs increase further as a result of enterprise restructuring. Nevertheless, TFP growth should gradually pick up over the medium term as the positive effects of structural reform begin to prevail. Growth of the capital stock is projected to decline somewhat below historical levels because of a temporarily higher depreciation rate. The growth rate of potential output is projected to be 7 percent over the next few years.

Table 2.4.Contributions to Output Growth, Actual and Projected

(In percent of GDP)1

Potential output9.
Capital accumulation6.
Labor force growth0.
TFP growth2.
Output gap0.4–
Actual output growth10.
Source: Authors’ regressions.

Period averages; the model is the Cobb-Douglas production function with endogenous TFP growth reported in Table 2.1.

Source: Authors’ regressions.

Period averages; the model is the Cobb-Douglas production function with endogenous TFP growth reported in Table 2.1.

Over the medium term, continued structural reform would eventually give new stimulus to TFP growth—not least by facilitating a pickup in the flow of labor out of agriculture into industry and services—and lay the foundation for faster TFP growth in the future, as the experience of other Asian countries suggests. The resulting improvement in efficiency as well as the updating of the capital stock should also sustain faster growth with lower rates of investment than in the past. Accordingly, the growth rate of potential output is projected to rise to 7½ percent on average during the second half of this decade, underpinned by the recovery in TFP growth to almost 1½ percent a year.


Labor migration out of agriculture, structural reform, and the partial reversal of the negative time trend are found to have raised TFP growth significantly during the reform period, and faster TFP growth in turn explains the substantial pickup in GDP growth in comparison with earlier periods. However, because many of these earlier reforms reflected one-time level effects, it is unlikely that future GDP growth will be able to match the levels witnessed during the reform period. China nevertheless remains a relatively poor country, and the scope for continued catching up with the industrial leaders will remain substantial for many years to come.

Capital accumulation will continue to be the largest contributor to economic growth in China. Whether TFP growth will also continue to play a pivotal role depends on the success of the next round of structural reform, and in particular on the progress with enterprise restructuring and banking system reform. Once the scope for further labor migration recedes, the pace of TFP growth will have to be sustained by the adoption of new technologies in the enterprise sector. To fulfill this role, SOEs need to be technologically upgraded and restructured according to market principles. Future technological progress will also require that the potential of China’s rapidly growing nonstate sector and the benefits of opening to the outside world be fully realized. The empirical results suggest that growth in the 7–8 percent range could therefore still be sustained over the next decade, provided the pace of structural reforms is intensified.

Appendix I: Data Sources and Description

Data on the capital stock and real GDP were taken from Chow and Li (1999), who derived their data from source data taken from various issues of the China Statistical Yearbook.

Data on the labor force and the proportion of the labor force employed in the primary sector were taken from the 1999 edition of the China Statistical Yearbook. The labor force series was adjusted for a break in the time series in 1989–90.

Appendix III describes the computation of the structural reform index. The underlying source data on imports and exports, saving and investment, nonstate and total industrial output, and urban and total population were taken from various issues of the China Statistical Yearbook.

Appendix II: Aggregate Production Functions and TFP Growth

The methodology used in this chapter is based on the familiar notion of an aggregate production function:

which relates GDP (Ŷ) to a vector of inputs (z). The production functions employed in this chapter satisfy the properties of a linearly homogeneous neoclassical production function f(·), which

  • exhibits positive and diminishing marginal products with respect to each input zt ∈ z:

  • exhibits constant returns to scale:

  • satisfies the Inada conditions, which state that for each input the marginal product approaches infinity as the input goes to zero, and vice versa:

Equation (1) can easily be expanded to incorporate an index of the level of technology in the economy, so that Ŷ = f(z, A). In addition, the technology index, A, can be a function of time and of a vector, x, of other relevant variables. In this chapter the focus is on two inputs (capital and labor) and TFP. Equation (1) can thus be specified as

where K is capital, L is labor, and A is TFP.

In the remainder of this appendix all variables are defined in logarithms. Next, equation (2), which can be thought of as a potential output relation, is assumed to be related to actual output in the following way:

where Y is the logarithm of actual GDP and ε is a cyclical disturbance term, or the output gap. To calculate TFP growth, potential output growth has to be decomposed into the contributions of capital and labor and a cyclical disturbance term. The first step in this calculation is to take time derivatives of both sides of equation (2) to obtain the growth rate of potential output:

where fQ is the first derivative of f(·) with respect to Q = A, K, L, and a dot over a variable represents a derivative with respect to time. Under the assumption that technological progress is Hicks-neutral, that is, output augmenting, fA = f(·)/A, and equation (3) can be rewritten as

The next step is to assume that factor markets are perfect, which implies that each input is paid its marginal product. Hence KfK/Ŷ and LfL/Ŷ are, respectively, the shares of capital and labor income in GDP. Furthermore, the property of constant returns to scale implies that the sum of both factor income shares equals 1. Depending on the specification of the production function, the factor income shares are constant or a function of relative factor endowments and technological progress.

Two specifications of the neoclassical production function are utilized in this chapter: the translog production function and the loglinearized Cobb-Douglas production function. The translog specification, which is the more general of the two in that it does not impose constant factor shares, has the following form:12

In this translog specification, factor income shares are not constant, and TFP growth is only a function of time t. The translog production function satisfies the property of constant returns to scale if and only if the following parameter restrictions are imposed:

It can be shown that, under the translog specification, the growth rate of the Solow residual (V) takes the following form in discrete time:13

where d is the first-difference operator and αt is the share of capital income at time t. The final step in obtaining the growth rate of TFP is to decompose the Solow residual into TFP and cyclical components, which can be achieved by HP filtering.

In this chapter the labor income shares for the translog specification are obtained from the national accounts. Another way of obtaining the labor income share is through direct estimation of the production function. In this case a log-linear Cobb-Douglas specification is applied, which is assumed to have the following form:

Because of the property of constant returns to scale, equation (4) can be expressed in intensive form as

with yY/L and kK/L. This approach also allows for a direct estimation of TFP growth, as both A and ε are estimated individually. As mentioned above, TFP can be defined as a function of time and a vector of other variables. In the simplest specification A is just a constant, which assumes no TFP growth. Whether this is a valid assumption can be tested by adding a linear time trend, in which case equation (5) becomes

where C is a constant and δ the rate of Hicks-neutral technological progress. Hence in this case, A = C + δt, but the analysis in this paper also considers A = C + δt + x, where x is a vector of variables (labor migration and structural reform) that contribute to TFP.

Appendix III: Measuring the Progress of Structural Reform

In this appendix an index is constructed to measure the progress of structural reform in China. The general approach to constructing the various indices of economic freedom, such as those compiled by the Fraser Institute,14 are useful in this regard. Four indicators, which capture the broad spectrum of structural change in the Chinese economy over the past 20 years, have been selected to construct the index.15 In addition, it is the combination of these variables rather than the variables individually that is considered to be the relevant measure of the progress of structural reform.16 The four indicators are the following:

  • The nonstate share of industrial output. One of the most notable features of the reform process has been the rapid growth of nonstate economic entities: the TVEs during the 1980s, and private (including individual) enterprises during the 1990s. The rising share of nonstate output is also a good proxy for the increased marketization of the economy (that is, the proportion of transactions taking place at market-determined prices) as well as the rapid redeployment of labor from agriculture to industry. The nonstate sector’s share of industrial output grew from just over 20 percent in 1978 to nearly 75 percent in 1999.

  • The share of total trade in output. Another notable feature of the reform period was the opening of the Chinese economy to the world economy following the establishment of the SEZs in 1980. The share of total trade (imports plus exports) in GDP, which rose from less than 10 percent in the late 1970s to just over 40 percent in recent years, is used to proxy the progressive integration of China’s economy with the rest of the world.

  • The level of urbanization. Although the pace of urbanization has not been quite as dramatic as the absorption of agriculture labor by TVEs (which have created over 100 million jobs since 1978), some 80 million people have left the countryside to work in urban areas over the past two decades. China’s level of urbanization (at 30.5 percent in 1998) is still well below the world average of 45 percent (75 percent in industrial countries) and according to the World Bank (1997) is about 13 percent below that in countries with comparable income per capita. Urbanization nevertheless has nearly doubled over the reform period, as controls on labor movements from rural to urban areas have slowly been relaxed.17

  • The rate of capital formation. As noted in the text, the conversion of saving to fixed asset investment has become significantly more efficient during the reform period. However, the efficiency of capital formation is still quite low by international standards, given China’s late start on SOE restructuring and continued state dominance of the banking system. Nevertheless, capital accumulation during the reform period still increased very rapidly because gross saving also grew rapidly.

Measuring each indicator on a zero-to-one scale and assigning an equal (one-fourth) weight to each yields an overall index of structural reform that slightly more than doubles—from 0.25 in 1978 to 0.53 in 1998—over the reform period.18 Although this index points to a profound transformation of China’s economic structure over the reform period, at the same time it suggests there is still considerable scope for further structural change toward a more open and market-oriented economy.


There is considerable debate about the accuracy of China’s official output statistics, and some observers believe annual growth rates are overstated by as much as 2 percentage points. Although many shortcomings are believed to plague the output statistics, most observers have focused on problems with the deflators used to measure real output. See also Box 4.1 of IMF (1998).

Many observers have also emphasized the importance of China’s economic structure at the start of the reform process in the late 1970s—a large population heavily concentrated in low-wage agriculture, creating a situation conducive to labor-intensive, exportled growth like that in other parts of East Asia—as a major explanatory factor in the success of the subsequent reforms.

The position of the SOE sector in the economy had also begun to erode by that time as a result of the emergence of the more dynamic nonstate sector.

Most researchers, for example Chow (1993), Hu and Khan (1996), and Maddison (1998), found little or no evidence of TFP growth during this earlier period.

The variation in the empirical estimates found in the literature stems mainly from differences in the underlying data used, which in some cases result from corrections for the apparent underdeflation of industrial output in the official data, from different methodologies employed to construct time series of the capital stock, and from the use of different capital and labor shares.

Although longer time series are available, the earlier data are not comparable because of the dislocating nature of the events that took place before 1970 (the Great Leap Forward and the Cultural Revolution). See also Chow (1993), Hu and Khan (1996), and Borensztein and Ostry (1996) for a discussion of the problems pertaining to the pre-1970 data. Appendix I describes the sources of the data used in this study.

The output gaps have been calculated by HP-filtering the Solow residuals and therefore may also be affected by the endpoint problem. If the trend of TFP has slowed in recent years, this would not be fully picked up by the HP filter, and the output gap would be overestimated (that is, the actual shortfall of demand would be less than suggested by the calculated output gaps).

In a growth accounting study of seven Latin American economies, Elias (1992) found capital share parameters ranging from 0.45 for Brazil and 0.52 for Chile to 0.66 for Peru and 0.69 for Mexico. Young (1995) found capital shares in East Asian economies between 0.29 for Taiwan Province of China and 0.53 for Singapore.

Because multicollinearity among the explanatory variables could not be ruled out a priori, this was tested for, but no convincing evidence of a significant bias in the estimation results was found.

Note that the reform index measures the result rather than the implementation of structural reform. Hence the relatively low contributions to TFP growth during 1971–78 and 1995–98 reported in Table 2.3 should not be interpreted as if the structural reform effort in the second half of the 1990s were as low as it had been in the prereform era. One interpretation is that the easy structural reforms have been completed and that China has now entered a phase of more difficult reforms that lack an immediate payoff.

See, for example, Hu and Khan (1996) for this derivation.

Although these indices include China, they are very data intensive and have been compiled for only a few years. European Bank for Reconstruction and Development (1996) constructs indices of economic transition for various Eastern European countries. However, this approach assigns numerical grades based on the status of reforms in several sectors in a particular country on the basis of qualitative assessments and thus does not readily translate into a single number, much less a time series.

Such variables (including the various indices of economic freedom themselves) have often been used to proxy the progress of structural reforms in numerous cross-country growth accounting studies, the so-called Barro-type regressions, since the early 1990s.

That is, the variables individually are not expected to be statistically significant or robust in a regression specification, but rather the index combining them is. Aziz and Wescott (1997), for example, argue that it is the package of structural reform policies and not the individual components that matter in transforming economic performance.

Zhang (2000) estimates that for each 1-percentage-point increase in urbanization during the 1990s, an additional 6 million to 7 million rural workers were absorbed in productive employment in urban areas.

Most industrial economies would fall in the 0.7–0.8 range on the basis of the criteria employed in the index.

    Other Resources Citing This Publication