Measuring and explaining productivity growth in China

Abstract

Europe was the past, the US is the present and a China-dominated Asia the future of the global economy. (Martin Wolf, 2003)

1. Introduction

At the end of the first decade of the 21st century China's economy has overtaken Japan and become only second to that of the United States. While the economies of the industrialized nations have encountered great difficulties as a result of the global financial crisis, China's rapid recovery from the economic downturn has attracted worldwide attention. Not only is the capability of the ‘China Model’ impressive in dealing with the global financial crisis and stimulating economic growth but China's intermediate-term growth outlook also remains remarkably strong in terms of investment opportunities and potential for technical progress and efficiency improvement (e.g. Zheng, Hu, and Bigsten 2009). Recently, the productivity performance of China's economy was brought into the spotlight thanks to a well-publicized report that placed China on the top of the international TFP growth ranking during 1990–2008 (Economist 2009). Moreover, it was reported that China may catch up with Japan in terms of research and development expenditure in 2011. While China's model of development is increasingly appealing to less-developed countries and nations in economic transition, the growing importance of China's economy and its potential as a centre for business and institutional innovation is affecting the way that developed nations think about the current world economic order (e.g., Zakaria 2008; Jacques 2009). Although investment will stay as the major source of China's growth, whether the trend of economic development will continue depends to a large extent on the sustainability of China's productivity growth. It is therefore important to understand the forces behind China's remarkable productivity performance in some areas and shortcomings in others.

The outstanding performance of China's economy is characterized by extensive participation in the world market, a private sector playing a major role in the creation of employment, and some key industries dominated by state firms. Some argue that China's high economic growth is not just because of heavy investment, but also due to the ‘world's fastest productivity gains’ (Economist 2009). Participation in the world market is the decision to open the country up to the outside world. It has resulted in the introduction of numerous advanced technologies and business practices from the west, and hence raised productivity. The rapid economic growth in China would have not been possible had there not been the tremendous development of a vibrant Chinese business sector. China's businesses have experienced an early stage of gradual reform in the state sector, followed by the dramatic expansion of the rural industries and a surge in the privatization of state and collective firms. While China was becoming the most attractive destination for foreign direct investment in the process, Chinese firms in recent years have started to acquire foreign companies and attempted to build up their own brand names. To make the presence of the Chinese firms sustainable internationally, innovation and productivity improvement are crucial to their continued success abroad.

China's industrialization process has basically followed the historical path of the developed nations. However, the limitation of this traditional approach is that it exerts enormous pressure on the natural environment. China's further development is thus constrained by its limited natural resources and by the fact it is facing the daunting task of meeting environmental challenges, while the current world economic recession and fierce international competition are putting ever greater pressure on China to upgrade its industries. This situation raises concerns as to the extent to which productivity growth will continue to contribute to economic growth in the foreseeable future.

The complexity and enormous scale of China's economy require methodologies that can be used for systemic investigations of productivity performance. This special issue encourages empirical studies within the neo-classical framework of Solow, which is still considered as the core of modern growth theory in the literature. While imposing a basic structure on a macro economic model of a general equilibrium nature, the Solow framework can be applicable to both well-developed market economies and planned economies. To accommodate significant differences in total factor productivity across countries and regions, efforts in the exploration of its cross-sectional implications are also widely appreciated.

In recent years, there seems to have been a consensus between scholars of growth empirics and applied productivity analysts that improving productivity estimations and better analysis to identify determinants of the productivity performance are the two areas to which major research efforts should be devoted. One direction we suggest is the study of the ‘knowledge production function’. This approach is attractive in an empirical sense in that the techniques of applied productivity analysis might be effective in investigations on the determinants of innovation and economic growth.

In many circumstances, the environmental issues can be studied under the standard framework of applied productivity analysis, which is based on the well-founded economic theory of production. Applications of environmentally related productivity models have been increasing. Experiences obtained so far are mixed and deserve further evaluations.

2. Growth empirics and productivity analysis

Comprehensive investigation of growth empirics under the framework of the neo-classical economic theory with reference to China often concerns the issues of general interests, such as growth determinants and convergence, which are instrumental to the explanation of high living standards in industrialized nations and the large international variations in per capita income. The case of China is particularly interesting due to its characteristic approach to economic reform and its phenomenal growth performance during the last three decades. A better understanding of the issues involved would contribute greatly to our knowledge of economic development and to studies of economies in transition.

The complexity and enormous scale of China's economy require methodologies that can enable researchers to link theory and reality, identify controversial issues, and cast doubt on conventional wisdom. A popular empirical methodology employed in the literature was developed in the neoclassical tradition of Solow (1956), which is ‘still at the heart of modern growth theory’ (McAdam and Allsopp 2007). While imposing a basic structure on a macro-economic model of a general equilibrium nature, some believe that the methodology of Mankiw, Romer, and Weil (1992, henceforth MRW) can in principle be used to evaluate not only the Solow model but arguably other candidate growth models as well (Bernake and Gurkaynak 2001). It was acknowledged by the literature that the MRW analysis has been further strengthened due to the extension of Islam (1995) and others to panel data.

Although neoclassical growth theory was conceived for understanding growth in industrial countries and should be mainly applicable to a planned economy or a well-developed market economy (Solow 2001), efforts in the exploration of its cross-sectional implications have proven rather fruitful. Especially when the MRW analysis is combined with the panel data approach, the process of identifying the individual country/region effect keeps reminding researchers that there might be a balance between neoclassical growth empirics and development economics (Islam 1995). The original MRW analysis emphasizes the role of investment, education, and political stability in explaining the large differences in international income, while the panel data approach can also accommodate significant differences in total factor productivity (TFP) across-countries and regions.

The neoclassical growth model has been rather useful for systemic investigations of developing economies that experience rapid economic growth (Zheng, Hu, and Bigsten 2009). The level of development, differences in production structure, and the capacity to absorb new technologies and improve efficiency are the main factors that can be attributed to the miraculous growth performance of the Chinese economy during the reform period. The literature as a whole gets involved in all three dimensions, although to different degrees. Several studies have been involved in the test of convergence across Chinese provinces, for example, in Gundlach (1997) and Yao and Zhang (2001). The importance of the convergence concept lies in the fact that it reflects a major consequence of the structural assumptions imposed on the supply side of the Solow growth model. The model predicts in general that poor countries/regions are capable of growing faster than rich countries/regions if the right conditions are met. Other studies took up the issue of openness and regional growth, and some paid special attention to trade and foreign direct investment (Madariaga and Poncet 2007; Liu, Song, and Romilly 1997). Both factors have contributed to the rapid economic growth in China since it can now benefit from a vast pool of new technologies from the outside world. Within this framework, methodologies were also developed to address the issue of structural change (Dekle and Vandenbroucke 2006; Jiang 2010), which has been regarded as another major suspect behind China's economic success, especially in the early reform years. The literature has so far devoted major attention to the supply side of the economy but it would be desirable to include an analysis taking account of the demand side. This is important because although there exists ample growth potential for China, the balance between saving/investment and consumption has yet to be determined if growth is to be sustained for the intermediate term (Zheng, Hu, and Bigsten 2009). A growth model with the household sector should open up avenues for in-depth analysis of transitional dynamics when household behaviour alters.

While the strength of the MRW analysis lies in the solid structure imposed on the neoclassical growth framework, an empirical methodology that is capable of accommodating issues outside the Solow model, and its systemic treatment of the factors associated with China's economic reform and development process, the literature appears to have put much emphasis on the convergence aspect of the empirical evidence and consequently the quantitative aspect of the panel data results have not been fully explored. If the forces of the convergence and productivity difference work in the opposite directions, then the effort to bring openness and structural change into the analyses should not only produce results of faster convergence but may also explain differences in total factor productivity across the regions to a certain extent, especially when considering that the openness factor measured with trade and FDI should present distributions in a geographical fashion (from the east coast to the west areas). A complete analysis of the panel data results should include an analysis of the individual dummies that take care of the total factor productivity differences across provinces as attempted in Islam (1995) and discussed further in Islam (1999). It would be preferable in future studies to know quantitatively whether the result of convergence dominates individual effects using some growth accounting methods.

Some studies using this approach gives the impression that the results were much in favour of convergence, and the use of the panel data approach only reinforces the traditional view that saving and education explain the most variation in income differences across countries and regions alike. However, the panel data procedure controls for more sources of differences in the steady state level of income, but the cost is that the obtained convergence results are thus rendered less meaningful. Moreover, the weaknesses with the panel data approach to address the issue of productivity differences are that the specification of an aggregate production function is restricted to the simple Cobb-Douglas form; some parameter values including the factor share parameter are assumed to be homogeneous; and productivity estimates are subject to the pitfalls of certain econometric methods (Islam 1999).

In fact, recent studies indicate that differences in productivity are better explanations of cross-country income differences than the differences in capital accumulation (Islam 2003a, 2003b). For example, Easterly and Levine (2001) noted that even after physical and human capital accumulation were accounted for, TFP seemed to explain large differences in the level and growth rate of GDP per capita across countries. Several authors have suggested that TFP should be the focus of growth research (Prescott 1998; Easterly and Levine 2001; Klenow 2001).

The issue of productivity dynamics can also be particularly interesting for China (Sakamoto and Islam 2008). A study in Tsui (2007) indicates that TFP and factor inputs were found to have different and sometimes opposing effects on inter-provincial inequality in the pre- and post-reform period. The increase in inequality in the pre-reform period was due to the dominant contribution of TFP over that of physical capital accumulation, while ‘the increase in the 1990s is mainly driven by the skewed distribution of investments in favour of the richer coastal provinces reinforced by the increasing contribution of TFP’. Thus, the convergence results in the literature should be interpreted with caution as far as policy implications are concerned. Traditional analysis of growth emphasizes the role of capital accumulation since it is the transitional dynamics of production factors that drive growth, but excessive investment may lead to growth patterns that may not be sustainable even in the intermediate term. Moreover, in discussions on growth strategies for developing nations in the literature, a major implication of the Solow growth model (Solow 1957) that seems to have often been forgotten is that ‘sustained growth in a country's per capita income can only occur if there is a rise in total factor productivity’ (Krugman 1994).

Zheng and Hu (2006) report that China's growth patterns had been rather extensive in the second half of the 1990s, relying primarily on capital accumulation rather than total factor productivity growth, according to their findings from a panel of provincial data for the period of 1979–2001. Aggregate studies of the Chinese economy also show that total factor productivity slowed during the last decade or so (OECD 2005, and Islam, Dai, and Sakamoto 2006) and the exceptionally high growth have been driven through high growth in capital stock, resulting in a series of imbalances in the macro economy (Zheng, Bigsten, and Hu 2009).

While the policy implications of this large literature were not crystal clear, some studies nevertheless went on to investigate the factors that may affect total factor productivity by augmenting the Solow growth model with variables that capture openness and structural change. Although the approach used to augment the Solow growth framework is rather common in the literature, it does model the productivity determinants in a way that may be consistent with China's growth experience. China's productivity growth before the mid-1990s might have been driven mainly through one-time dramatic improvements in policies. Several studies have predicted that if China does not keep up its reform momentum, its productivity and perhaps per capita income growth might slow down. This phenomenon is related to the concept of ‘level effect’, which is not sustainable in the long run under the framework of the Solow growth model, while sustained productivity growth is referred to as the ‘growth effect’ in the terminology of growth modelling. Augmenting the Solow model with additional variables as done in the original MRW analysis with human capital will only produce a level effect. The end result is to suggest a growth strategy based on the expansion of inputs since policy changes may only have ‘level effect’, or to recommend stronger policy measures and even bolder reforms of the economic and political system.

3. The knowledge production function approach

The 1980s were marked with the emergence of endogenous growth literature and increasing interest in the impact of R&D expenditure on productivity performance of firms and industries, while studies on productive efficiency experienced an explosive expansion. Conceptually, the three branches of the literature are well connected. The endogenous growth model studies the effects of knowledge accumulation on the productivity growth of the aggregate economy in the long run; the studies on the innovation–productivity link focus on the empirical estimation of the return to R&D; and the literature of productive efficiency specializes in models that measure technical efficiency and technical progress.

Although the size of the literature that estimates productive efficiency is huge, empirical investigations on the determinants of the productivity performance based on these estimates appear to be difficult and the relationships between explanatory and explained variables are not well structured. The endogenous growth models paid more attention to mathematical formulations, but the structures employed so far are often restrictive and simplistic, as in for example Romer (1990), Aghion and Howitt (1992), and Jones (1995). It is therefore important to sort out the potential microstructures that can be formulated regarding the relationship between innovative inputs and total factor productivity.

The basic idea here is to sketch a production function framework that can be established when considering the relationship between innovation and productivity. The concept of interests can be naturally referred to as the ‘knowledge production function’ as in Romer (1990). However, the formal presentation was inspired to a large extent by the notion of the ‘technology function’ in Phelps (1966).

Assuming that the production unit is the firm, its production function, with y as output and x as input, takes the standard Hicks-neutral form as follows:

and more specifically with reference to the growth literature,

where Y stands for the ordinary output, K and L_Y refer to capital and labour, and A is the level of technology or useful knowledge stock that can be employed in the production process. Although the specification was used in the macro context, it is equally valid for firm level analysis since what is of interest here is the functional form or the microstructure of the knowledge production function. Following Romer (1990), the knowledge production is measured as the change in knowledge stock , and the basic knowledge production function can be expressed as follows.1

or more specifically as in Phelps (1966),

where K_A and L_A stand for technology capital or research capital stock and research labour input respectively, and L = L_A + L_Y is the labour force of the firm. Generally, we assume the knowledge production function with properties as follows.

• A1. The knowledge production function is monotone in inputs.

• A2. The input set of the knowledge production is convex.

• A3. The researcher is the essential input, i.e., .

• A4. The input set is non-empty and closed.

• A5. The knowledge production function is finite, non-negative, real valued, and single valued for all non-negative and finite inputs.

And note that the knowledge production function f(•) is related to the ‘technology function’ of Phelps (1966) through the relationship as follows:

Moreover, Phelps (1966) suggested the dynamic properties required of a valid ‘technology function’, i.e. the knowledge production function in stock form, A(t) as follows.

• B1. Diminishing returns.

• B2. Diminishing marginal rate of substitution.

• B3. The marginal effectiveness of current research is an increasing function of the level of technology recently attained (technical progress in research).

• B4. Exponential growth of researchers will produce an exponential increase of the level of technology.

Owing to the dynamic nature of the knowledge production function, not all specifications are feasible for theoretical analyses and empirical estimations. Therefore, one possible specification is a Hicks-neutral form as follows:

which makes the knowledge level multiplicatively separable from the rest of the function. Interestingly, is simply the ‘effective research function’ that we have seen in Phelps (1966). This form allows separate analysis of instead of dealing with the as a whole. This property is very important since the analysis of the former is much simpler.

When approaching the issues involved from the perspective of the standard production theory, the analytical advantage is that the basic properties required of the ordinary production function can be employed to infer about the microstructure of the knowledge accumulation process. This approach is also attractive in an empirical sense in that the standard techniques of applied productivity analysis can be effective in investigations on the determinants of productive efficiency and total factor productivity growth.

For example, one may proceed in the direction that Solow (2001, 2007) suggested and echoed in Islam (2003b, 2008), treating TFP as the left-hand side variable in regression analysis. In fact, in the applied productivity analysis literature, there is already a popular two-stage procedure in which productivity indexes were calculated in the first stage and regression analysis follows in the second stage (e.g. Zheng, Liu, and Bigsten 2003; Castellacci and Zheng 2010). In this framework, a Translog production function can be specified in the test of the specific knowledge production functions.

A second virtue of this procedure is that TFP – technological progress in the broad sense – can be decomposed into its components. Islam suspects that the influence of human capital on growth can also be better examined with such a two-stage analysis. There seems to have been a consensus between scholars of growth empirics and applied productivity analysts that improving productivity estimations and better design of the two-stage analysis to identify determinants of the productivity are ‘the two lines in which future research on this topic can fruitfully proceed’ (Islam 2003b).

Using the popular DEA-based Malmquist productivity indexes as an example, it can be demonstrate that when they are used in regression analysis the set of linear equations involved can be treated as a system. With reference to the special structure of the knowledge production function, the regression equations can be further specified as a dynamic system. The properties of the Malmquist Index regression equations provide rich microstructures for the relationship between productivity growth, productivity growth components, and the determinants of productivity growth. Knowledge of these structures will be very helpful in applied work. Its dynamic properties may help interpret empirical results and facilitate efficient econometric estimations.

4. Measuring technical and environmental efficiency

For China to develop an environment-friendly economy in the short term it will require a major breakthrough in renewable energy. However, before commercially viable technologies for wind and solar power become available, China will have to rely on coal for electricity generation and for its non-satiable industrial demand. A more immediate challenge is thus the environmental pollution resulting from the consumption of fossil fuels.

In recent years, considerable efforts have been made in the studies of China's pollution problems in the literature. As with the majority of the literature, many researchers carry out the estimation of efficiency through the DEA methodology, a non-parametric programming technique developed by Charnes, Cooper, and Rhodes (1978). Since the mid-1990s, the directional-distance-function based DEA approach has been developed to measure the environmental impact of emissions in productivity analysis. Recently, some studies also applied this approach to Chinese data in relation to environmental issues involved in coal-fired power generations (e.g. Xie and Ren 2006). In spite of these efforts the usefulness of this relatively new methodology is still not very clear. These models often encounter difficulties when measuring the impact of the emissions.

Treatment of undesired output in DEA analysis varies in the literature. Some of the literature simply includes the undesirable output as input to the production process. Alternatively, one can apply a directional distance function approach (Chambers, Chung, and Fare 1996; Chung, Fare, and Grosskopf 1997; Hu et al. 2008). The model can treat undesired output in output space. However, in practice, both approaches were found to ‘have a weaker discriminating power’ because their efficiency measures are radial, i.e. the models seek simultaneous expansion of outputs or contraction of inputs by the same proportion towards the production frontier. Fare and Lovell (1978) point out that ‘there is no reason to measure technical efficiency radially, even for homothetic technologies’. Several non-radial measures have been proposed in the literature (e.g. Banker and Morey 1986; Thanassoulis and Dyson 1992). Because non-radial models often require the use of subjective weights to construct an aggregate measure of overall efficiency for individual decision making units (Zhu 1996), difficulties may arise in applied work if these weights are not available (Zhou et al. 2007).

Thus, when undesirable outputs are involved, sometimes researchers may not only want to work with efficiency measures that are directional but also non-radial with no subjective weights attached. For example, in the case of coal-fired power generation in China, very often the technical efficiency of ordinary inputs, such as the utilization of labour and capital, is not a serious problem, while the extent of SO₂ emission varies across the power plants. This is because some firms may sacrifice environmental considerations for technical efficiency when there have been power shortages. When both ordinary input and undesired output such as SO₂ emission are considered simultaneously, the priority is sometimes to reduce pollution, but some attention should also be given to the use of ordinary inputs. This amounts to developing models that can measure technical and environmental efficiency in different directions with no subjective weights attached.

5. The special issue

As China's economic development is reaching its critical mass, the productivity issues researchers are concerned with are increasingly different from what they were just a few years back. A scientific and technological takeoff that appeared remote and hard to achieve ten years ago seems now to be more likely in the decades ahead thanks to China's continuous open door policies and its constant endeavour to catch up with the advanced economies. Inspired with this fascinating while challenging prospect, a cohort of researchers stationed in Sweden, Norway, Australia, the UK, Japan, South Korea, and mainland China have joined forces to make possible the 2011 special issue on Chinese productivity.

The first contribution by Wu investigates the extent to which growth in China's industries has been driven by productivity change, evaluates the relative productivity performance of enterprises with different ownership types, and inquires whether productivity across China's key regions exhibits convergence or divergence. The second article by Fukao and his collaborators focuses on stock-market listed companies in China, Japan, South Korea, and Taiwan. The article examines productivity catch-up at the firm level and the role of absorptive capacity for technological catch-up, and finds that policies to raise the technology level of national frontier firms are beneficial for all firms in one country. Jiang in the third article addresses the issue of regional TFP. By building a model of international technological spillovers and using a nonlinear regression method, this article examines the effects of the regional economic environment, openness, and technology diffusion on China's regional TFP growth, and shows that regional openness has a significantly positive effect on regional TFP growth.

Sutherland et al. in the fourth paper discuss productivity performance in Chinese business groups, and evaluate some of the positive and negative attributes of Chinese groups. The article shows that accompanying improvements in productivity, some Chinese business groups are taking their first steps towards developing pyramidal structures, which could have important implications for longer-term productivity growth in China's business groups. Li and co-authors attempt to investigate the situation with agriculture productivity in China in the fifth paper. The article finds that Chinese agriculture experienced significant productivity growth in the last few decades with considerably varying growth rates across the subsectors, and since the 1990s Chinese agriculture experienced a great technological progress and yet a considerable efficiency loss. The final article, from Førsund et al., introduces the short-run production function approach, which used to be popular in Nordic countries. In order to show how the production function concept is the key to understanding industry dynamics, the article presents an empirical application on data for Chinese coal-fired electricity generation plants.

This special issue is a second attempt by JCEBS after the 2008 special issue in a single volume to focus on the productivity issues involved in Chinese economic development and business studies. We hope that the effort will encourage and bring forward more in-depth studies and applications of innovative methodologies in the near future.

Acknowledgement

I thank Jinchuan Shi and Yuxiang Jiang for hosting and providing financial support to the 2008 International Workshop on Chinese Productivity at the School of Economics, Zhejiang University; I am particularly impressed with their generosity and hospitality at the workshop. Financial support from the Center for Research of Private Economy, Zhejiang University is also highly appreciated. I am also grateful to Finn Førsund and Lennart Hjalmarsson for their support of the special issue from the Nordic perspective. Special thanks also go to Xiaming Liu who as editor-in-chief of JCEBS supported both special issues on Chinese productivity without the slightest reservation. I am heavily indebted to a number of scholars who were involved at different stages during the preparation of the special issue, including Angang Hu, Robert Gordon, Gary Jefferson, Yanrui Wu, Yuxin Zheng, Rouen Ren, Xiaoxuan Liu, Yanqing Jiang, Carsten Holz, Hiro Izushi, Jin-Li Hu, Hans C. Blomqvist, Song Han, Lúcia Lima Rodrigues, Maggie Fu, Satoshi Honma, and Joe Hughes.

Additional information

Notes on contributors

Jinghai Zheng†

†The author is also guest research fellow at Center for China Studies, Tsinghua University, China.

Notes

†The author is also guest research fellow at Center for China Studies, Tsinghua University, China.

Note

1. Phelps (1966) uses the term ‘technology function’ in referring to the relationship between research inputs and research outcomes; Gomulka (1970) mentioned ‘the production function of innovations’; and Jones (1999, 2002, 2005) prefers to phrase the relationship as ‘the idea production function’.