I. Introduction

Over the past 40 years of reform and opening up, China’s manufacturing industry has achieved considerable development in scale and quality by relying on its comparative advantage in labor. However, with the advancement of science and technology and social changes, China’s demographic dividend has gradually disappeared, and labor prices no longer have a comparative advantage, and faced with practical problems, such as an aging population and overcapacity, factor input has a diminishing output. Therefore, facing bottlenecks in internal development and a turbulent external environment, China’s manufacturing industry is in urgent need of intelligent transformation to achieve a new round of high-speed and high-quality growth.

At present, the research on the traditional determinants of total factor productivity (TFP) is relatively mature and is mostly from the perspectives of R&D, foreign direct investment, and urbanization (Ashraf et al., 2016; Kumar & Kober, 2012; Tsamadias et al., 2019). Some studies have investigated the impact of technological innovation on TFP from the perspectives of Internet technology, information and communications technology (ICT), and industrial robots (Brynjolfsson & Hitt, 1996; Cette et al., 2021; Kallal et al., 2021) but their conclusions are not uniform. Some scholars believe that technological innovation has improved TFP (Venturini, 2022), while others have concluded the opposite (Jorgenson et al., 2008). With the advent of Industry 4.0, intelligent manufacturing continues to penetrate all aspects of production, and intelligence is bound to have a huge impact on production practices. The existing literature has confirmed that intelligence has an important role in the labor force, especially in terms of labor substitution and polarization (Acemoglu & Autor, 2011; Autor et al., 2006; Autor & Salomons, 2018). Intelligence also has a positive role in energy saving, at the same time (Hao & Wu, 2021; Lan & Wen, 2021). Nevertheless, only a few studies have explored the impact of intelligence on TFP. Therefore, this paper explores the effect of intelligence on TFP. We construct and measure intelligence from a systematic perspective to avoid the subjectivity caused by the use of a single indicator in previous studies. Finally, we obtain more objective estimation results, which enrich the literature on TFP determinants.

The contribution of this paper is to select multiple indicators and combine the entropy method to comprehensively measure the level of intelligence in China, and to empirically test the actual impact of intelligence on the TFP of the manufacturing industry, expand related research on intelligence, and provide evidence for the planning and development of China’s manufacturing industry.

II. Data and Method

A. Model

To test the impact of intelligence on the TFP of China’s manufacturing industry, the following model is constructed:


Where i represents the province and t represents the year; TFPit represents the total factor productivity, IDit represents the level of intelligence, Xit is a vector of control variables; μi and vt represent the individual and time fixed effects, respectively, to control other potential determinants of TFP and time trends, and εit represents the random error term.

B. Variables

The explained variable is the manufacturing TFP of each region. This paper follows Wang et al. (2021) and uses the Malmquist index, a non-parametric data envelopment method, to calculate TFP. Among them, based on the index, the specific input indicators select the net value of fixed assets and the average number of all employees in each region over the years, and the output indicator selects the total output value of the manufacturing industry in each region.

The core explanatory variable is the degree of intelligence (ID). We choose comprehensive indicators to describe the degree of regional intelligence. This paper takes the manufacturing industry as an example to explore the impact of intelligence on TFP. Therefore, the selection of indicators needs to integrate the characteristics and actual conditions of the manufacturing industry. Specifically, we investigate intelligence from three levels, namely intelligent technical indicators, intelligent application indicators, and intelligent benefit indicators. As shown in Table 1, the investment in fixed assets of intelligent facilities, the number of patent applications in the electronic communication equipment manufacturing industry, and the profit of the electronic and communication equipment manufacturing industry are selected as the characterization indicators, and the final intelligence score is calculated by using the entropy method, and used as an indicator of the level of intelligence.

Table 1.Intelligent comprehensive indicators
First-level indicators Specific indicators Unit
Intelligent technical Investment in fixed assets of intelligent facilities Million (RMB)
Intelligent application Number of patent applications in the electronics and communication equipment manufacturing industry Item
Smart Benefit Profits of Electronic and Communication Equipment Manufacturing Billion (RMB)

Notes: This table reports the index system for measuring the degree of intelligence. The first column is the first-level index; the second column is the specific quantitative index; the third column is the unit of each variable.

Control variables. (1) Foreign direct investment (FDI): the development of China’s early manufacturing industry was supported by FDI, and, at the same time, it relied on the two-way spillover effect of FDI to obtain technological progress. This affected the overall development of the manufacturing industry. This paper multiplies the FDI value in a given year by the average exchange rate of that year and divides the resulting value by the GDP of that year, to measure FDI. (2) Government fiscal expenditure (GOV): existing studies have shown that GOV plays an important role in industrial development, especially for emerging industries and industries in transition. This paper uses the ratio of government fiscal expenditure to GDP to measure GOV. (3) Human capital (HC): this is measured in this paper using the average years of education. (4) The level of economic development (PGDP): the higher the level of economic development, the more priority will be given to the promotion of related applications of intelligence. We measure the level of regional economic development by the regional GDP per capita. (5) Industrial structure (IND): the industrial structure will affect the development and transformation directions of the regional industry, especially the continuous improvement of the secondary and tertiary industries. We measure the industrial structure as the proportion of the secondary and tertiary industries.

C. Data Sources

The data used in this article are all from China Statistical Yearbook, China Labor Statistical Yearbook, and China High-tech Industry Statistical Yearbook. Our dataset is constructed by considering that the development of China’s ICT industry is relatively lagging, and by considering the continuity and availability of data in the relevant statistical yearbooks. Our dataset is a panel of 30 provinces in China from 2000 to 2019.

III. Empirical Results

This paper uses a double fixed-effect model, which controls for region and time fixed effects, to test the impact of intelligence on China’s manufacturing TFP. Table 2 reports the empirical test results. The results show that the coefficient of intelligence is 0.014 and is statistically significant at the 1% level, indicating that intelligence promotes TFP growth in China’s manufacturing industry. In addition, the regression coefficient of intelligence on TFP does not change in direction and size (the coefficient does not significantly fluctuate), after adding control variables in sequence, indicating that our estimates are robust.

In fact, the continuous development of intelligent technology and technological progress, such as the continuous decline in the price of ICT products, will stimulate the continuous increase of social-related intelligent investment, which will in turn expand per capita capital, and ultimately increase the productivity of society. At the same time, intelligent technology has the attributes of general-purpose technology and, hence, has a spillover effect (Lipsey et al., 2005); its impact on the economy is not limited to a certain industry. Over time, intelligent technology has spread to the whole Chinese industry to varying degrees. In order to adapt to the new technological environment and to changes in production conditions, the relevant organizational structure and business processes of enterprises will have to be adjusted and changed. This kind of change is different from the application of intelligent equipment, which directly affects production efficiency, but indirectly improves production efficiency by changing the intangible characteristics of enterprises, such as product quality, product variety, and service response speed. While improving production efficiency, intelligence has also changed the organizational structure and business processes of enterprises. Different from mechanization, intelligence is not a simple replacement for programmed production tasks, but improves the comprehensive management ability and efficiency of enterprises, and gives enterprises higher dynamic capabilities to optimize their own factor input and improve the marginal output of factors.

Table 2.Benchmark regression results
ID 0.012*** 0.012*** 0.013*** 0.014*** 0.015*** 0.014***
(3.25) (3.30) (3.52) (3.60) (3.79) (3.73)
FDI -0.241 -0.219 -0.250 -0.093 -0.133
(-0.71) (-0.64) (-0.73) (-0.26) (-0.37)
GOV -0.195 -0.213* -0.240* -0.244*
(-1.56) (-1.69) (-1.89) (-1.93)
HC -0.024 -0.018 -0.016
(-1.22) (-0.91) (-0.79)
PGDP -0.047 -0.075**
(-1.60) (-1.99)
IND 0.154
Constant 1.073*** 1.081*** 1.111*** 1.308*** 1.692*** 1.862***
(74.63) (59.34) (42.16) (8.04) (5.84) (5.76)
R-squared 0.115 0.116 0.120 0.123 0.128 0.131

Notes: This table reports the regression results of the effect of intelligence on TFP. In order to obtain robust estimation results, the regression results after gradually adding control variables are specifically reported in the table. Finally, “*”, “**”, and “***” represent 10%, 5%, and 1% significance levels, respectively.

For the regional regression results (Table 3), only the eastern region results are statistically significant. Thus, intelligence promotes TFP growth in the eastern region, but does not impact TFP growth in the central and western regions. The results show that there are regional differences in the effect of intelligence on TFP in China. The reason for the regional heterogeneity may lie in the differences in the level of regional intelligence. Figure 1 shows the trend chart of different regions based on intelligence. It can be seen that the intelligence levels of the three regions are relatively similar in the early stage of the sample. Compared with other regions, the eastern region has better infrastructure, human capital, and economic conditions, which can rapidly promote intelligent equipment and realize intelligent industrial production. The central and western regions gradually lag behind the eastern region in terms of intelligence level, as the Chinese economy develops; the gap in intelligence level between the eastern and central/western regions has become larger over time, explaining why intelligence has no impact on TFP in the central and western regions.

Table 3.Heterogeneous regression results
Variables Nationwide Eastern Central Western
ID 0.014*** 0.024*** 0.003 -0.001
(3.73) (2.94) (0.76) (-0.32)
FDI -0.133 3.676* -0.033 -0.486
(-0.37) (1.93) (-0.15) (-0.77)
GOV -0.244* 0.408 -0.321** -0.061
(-1.93) (1.05) (-2.60) (-0.19)
HC -0.016 -0.066 0.008 0.028
(-0.79) (-1.19) (0.57) (1.41)
PGDP -0.075** 0.127 -0.056* -0.039
(-1.99) (0.95) (-1.80) (-0.71)
IND 0.154 0.410 0.090 0.021
(1.18) (1.02) (0.99) (0.16)
Constant 1.862*** 0.282 1.519*** 1.123**
(5.76) (0.23) (5.39) (2.17)
R-squared 0.131 0.229 0.234 0.438

Notes: This table reports the regression results of the impact of intelligence on TFP in different regions. Specifically, referring to the regional division standards of the National Bureau of Statistics of China, the research samples are divided into eastern, central and western regions. Finally, “*”, “**”, and “***” represent 10%, 5%, and 1% significance levels, respectively.

Figure 1
Figure 1.Trends on the intelligence degree (ID) in different regions.

Source: Calculated from the data shown in Table 1 above.

IV. Conclusion

Intelligent transformation is an inevitable direction for the development of the manufacturing industry. At present, there is limited research on the impact of intelligence on TFP in the context of China. This paper empirically tests whether intelligence has improved the TFP of China’s manufacturing industry, by constructing a provincial panel dataset. The regression results show that the continuous popularization and deepening of intelligence has improved the TFP of China’s manufacturing industry. The results of the regional heterogeneity test show that intelligence only significantly promotes TFP in the more developed eastern region, while it has no impact on TFP in the central and western regions.