Production Networks and Business Cycles in China

2.1 Characteristic Facts

2.1.1 Production Network Structure

During China’s rapid economic development, the internal production network has also undergone structural changes. Figure 1(a) illustrates the proportion of intermediate inputs to value-added in the production process from the reform and opening-up in 1978 to the financial crisis in 2008. During this period, the proportion of intermediate inputs to value-added nearly doubled, indicating increasingly close input-output linkages among economic sectors. At the same time, significant structural transformation has occurred within the production network. We categorize all sectors into capital-intensive and labor-intensive sectors based on their capital-labor ratios and illustrate the relative development trends of these two categories in Figure 1(b). The solid black line in the figure represents the relative proportion of intermediate inputs used by the two types of sectors, while the grey line shows the relative proportion of value-added by these sectors, with the relative proportions standardized to 1 in 1978. Since the reform and opening-up, there has been a clear trend of transformation from labor-intensive sectors to capital-intensive sectors, both in terms of value-added and intermediate input usage. Moreover, compared to the structural transformation in value-added, the structural transformation in input-output linkages is even more pronounced, suggesting significant changes within the production network.

Figure 1

The Structure of China’s Production Network and its Dynamic Transformation

2.1.2 Industry Fluctuations and Macroeconomic Fluctuations

Define the following two industry-level weighting coefficients, referred to as the value-added weight and the Domar weight:

(1)  Value-added weight = Added value of sector i Gross value added ; Domar weight = Output of sector i Gross value added 

In traditional horizontal multi-sector economy, there are no input-output linkages between sectors. As a result, the fluctuation of a sector (in percentage terms) is transmitted to the macroeconomy based on its value-added weight. Therefore, only when a sector accounts for a sufficiently large share of the economy, its own fluctuation has a noticeable impact at the macroeconomic level. However, this conclusion no longer holds when input-output linkages exist between sectors. Hulten (1978) pointed that the fluctuation of a sector was transmitted to the macroeconomy based on the Domar weight, rather than the value-added weight. The output of a sector equals its value-added plus the intermediate inputs it uses. Therefore, the Domar weight, building on the value-added weight, takes into account how a sectoral fluctuation is transmitted to other economic sectors through the intermediate input channel. This conclusion is also known as Hulten’s theorem and has been further validated within a general equilibrium framework (Acemoglu et al., 2012; Baqaee and Farhi, 2019).

This explains why some sectors with relatively low value-added shares have a significant impact on China’s macroeconomy. A typical example is the real estate sector. The grey line in Figure 2(a) shows the value-added weight of China’s real estate sector since the reform and opening-up, which has generally remained below 10%. However, the real estate sector has extensive influence on upstream and downstream industrial chains in the Chinese economy. The solid black line in Figure 2(a) shows the Domar weight of the real estate sector, which is more than four times its value-added weight and over five times the average Domar weight of other sectors. According to Hulten’s theorem, shocks to the real estate sector are likely to be significantly amplified through the production network. In Figure 2(b), we illustrate the potential macroeconomic fluctuations caused by real estate sector fluctuations^[1]. First, we use the H-P filter method to calculate the fluctuations of the real estate sector and macroeconomic fluctuations (i.e., deviations from the trend). Next, we calculate the macroeconomic fluctuations driven by real estate sector fluctuations based on value-added weight and Domar weight, as shown by the grey and dashed black lines in Figure 2(b), respectively. Since the real estate sector’s own value-added weight is not large, its fluctuations would not have a significant impact on the macroeconomy if input-output linkages are not considered. However, due to the real estate sector’s extensive upstream and downstream connections within the production network, its impact on macroeconomic fluctuations increases significantly when calculated using the Domar weight.

Figure 2

The Key Node of China’s Production Network: The Real Estate Sector

2.2 Theoretical Model

To understand the relationship between sectoral fluctuations and the macro business cycle, we construct multi-sector real business cycle model with input-output structure in this section. Our goal is to reveal the transmission and amplification process of sectoral shocks within the production network and to understand how the input-output structure and its transformation affect the roles of different industries in the business cycles.

3.1 Empirical Framework

With reference to Sims and Zha (2006) and Chen et al. (2024), the following SVAR model is constructed:

Among them, y˜t=y˜1,t,…,y˜n,tT is an n × 1 column vector, where its i-th component y˜1,t represents the fluctuation of sector i at time t (the proportion that deviates from the trend), calculated using the Hodrick-Prescott (H-P) filter method. B₀ (s_t), B₁ (s_t)...B_p (s_t) are n × n coefficient matrices, and p denotes the lag order of the SVAR model. s_t is a regime-switching variable that takes discrete values over different periods, aiming to capture changes in the transmission paths of shocks caused by variations in the production network structure across periods. ε_t represents structural sectoral shocks. Following De Graeve and Schneider (2023), we introduce a set of heterogeneity restrictions, requiring that sectors more closely connected to the source of the shock exhibit larger impulse responses, thereby identifying ε_t. Existing research has shown that incorporating appropriate heterogeneity constraints can significantly improve shock identification (Kilian and Murphy, 2012; De Graeve and Karas, 2014; Amir-Ahmadi and Drautzburg, 2021). We use the Bayesian Information Criterion (BIC) and Akaike Information Criterion (AIC) to determine the lag order p = 1, and specify the regime-switching variable s_t to take the value 0 before and including 2002, and 1 thereafter. In the subsequent analysis, we refer to the fluctuations in various sectors driven by ε_i,t in Equation (11) as fluctuations driven by sector i.

3.2 Data Source

The sectoral value-added data and input-output matrices used in this paper are sourced from CIP/China KLEMS. The study employs an extended version of CIP 4.0, covering 37 sectors from 1978 to 2018. This database is initially developed by a team led by Professor Wu at the Institute of Economic Research, Hitotsubashi University, Japan, and is currently maintained and updated by the Growth Laboratory at the National School of Development, Peking University. For more detailed information, refer to Wu and Ito (2015) and Wu and Li (2021). To include as many sectors as possible while ensuring the model’s computability, we partially aggregated the original sectors into the following categories: agriculture, energy, metal ores and products, non-metal ores and products, food, tobacco, textiles, wood and furniture, paper and printing, construction and real estate, industrial machinery and equipment, electronic and communication equipment, instruments and office equipment, vehicles and transportation equipment, and electrical equipment, totaling 15 sectors.

3.3 Identification Results

Figure 3 illustrates the macroeconomic fluctuations in China from 1978 to 2018, as well as the fluctuations in individual sectors. The vertical axis is measured in percentages, indicating the magnitude of the deviation from the trend term. The left axis indicates the magnitude of macroeconomic fluctuations, while the right axis shows the fluctuations of individual sectors. The black solid line with circular markers depicts the overall economic fluctuations. Below, we focus on analyzing the two cycles following the establishment of the socialist market economy system in 1992.

Figure 3

China’s Business Cycle

Figure 4(a) presents the macroeconomic fluctuations in China from 1991 to 1996, as well as the fluctuations driven by individual sectors. This period was characterized by an economic expansion, with the construction and real estate sector, as well as labor-intensive manufacturing sectors such as textiles and food, making significant contributions to the prosperity of this era. Quantitative results indicate that the combined shocks from just three sectors—construction and real estate, textiles, and food—are sufficient to explain approximately 69% of the overall economic fluctuations. Between 1991 and 1996, the real estate sector grew at an average annual rate of 9.5%, with an average annual increase of 160 million square meters. Additionally, with the continuous improvement of the market economy system and the support of national foreign trade policies, many labor-intensive sectors that aligned with the comparative advantages of the time, such as textiles and food, experienced rapid development. Beyond these, shocks from most other sectors had almost no impact on macroeconomic fluctuations.

Figure 4

Sources of Economic Fluctuations

The 1997 Asian financial crisis had a negative impact on China’s economy, as shown in Figure 4(b). The construction and real estate sector, as well as labor-intensive manufacturing sectors such as textiles and food, which had driven rapid economic growth in the previous period, also became the dominant sectors contributing to the economic downturn during this time. Under tight monetary policies, real estate, as a key target of regulation, faced significant restrictions. Simultaneously, laborintensive sectors reliant on external demand, such as textiles and apparel, were also hit by the Asian financial crisis. Additionally, since China’s agricultural trade surplus primarily came from Asian regions, the economic setbacks in Asian countries and the decline in China’s export competitiveness led to a downturn in the agricultural sector. Quantitatively, shocks from these dominant sectors accounted for 76% of the economic fluctuations during this period.

From the end of 2001, when China joined the World Trade Organization (WTO), to 2011, the Chinese economy experienced a new round of rapid growth. Figure 4(c) shows that the economic prosperity during this period was driven by multiple sectors, and after the 2008 financial crisis, China’s economic growth model underwent significant changes, with the real estate and construction sectors diverging from other industries in a “K-shaped” trend. Specifically, the global financial crisis profoundly impacted China’s real economy, leading to slowdown in growth across multiple sectors. In response to the crisis, the Chinese government implemented a series of counter-cyclical measures starting in 2008, which not only greatly stimulated the real estate market but also relaxed debt financing restrictions for local governments, providing funding for large-scale infrastructure projects. These measures effectively promoted the development of the construction and real estate sector, helping China’s economy recover and grow from the shadow of the financial crisis. Quantitative analysis shows that from 2009 to 2011, the construction and real estate sector contributed as much as 53% to business fluctuations.

From 2011 to 2018, China’s economy entered a “new normal”, transitioning from high-speed growth of 11% to medium-high growth of 6%. According to the results shown in Figure 4(d), the slowdown during this period was primarily driven by the construction and real estate sector, the metal ores and products sector, the energy sector, and the industrial machinery and equipment manufacturing sector, which collectively contributed approximately 82% of the economic fluctuations. The downturn in the construction and real estate sector during this period was influenced by both changes in supply and demand conditions within its own development cycle and the impact of national policy regulations. During this time, with the implementation of a series of environmental policies, the production of heavy chemical products such as steel and coal significantly declined. It was not until the deepening of supply-side structural reforms in 2016 that the imbalance between supply and demand was fundamentally improved. Simultaneously, after the financial crisis, Western economies entered a period of weakness, leading to a significant decline in effective demand. In summary, the slowdown in China’s economic growth during this period was largely driven by the optimization and adjustment of the sectoral structure.

4.1 Dominant Sectoral Shocks and Macroeconomic Fluctuations

According to theorem 1, the marginal impact of sectoral shocks on real value-added can be decomposed into the effect on the sector itself (within-sector effect) and the effect on other sectors through inter-sectoral input-output linkages (between-sector effect). The decomposition results are shown in Figure 5. In the figure, the black solid line with markers represents actual macroeconomic fluctuations, the black solid line represents macroeconomic fluctuations generated by shocks of the dominant sectors in each period, and the grey line represents macroeconomic fluctuations generated by shocks of the dominant sectors when input-output linkages are shut down, that is, the within-sector effect of dominant sectoral shocks. The results show that shocks of the dominant sectors in each period explain the majority of macroeconomic fluctuations. However, if only the within-sector effect is considered, the contribution of dominant sectoral shocks to economic fluctuations would significantly decrease. This implies that the impact of dominant sectoral shocks on other sectors through input-output linkages is the primary reason that these sectors can dominate the macroeconomic cycle.

Figure 5

Dominant Sectoral Shocks and Macroeconomic Fluctuations

4.2 Network Structure Transformation and Dominant Sectoral Transformation

Based on the characteristic facts of China’s production network structural transformation presented in the second part of this paper, capital-intensive sectors showed a significant upward trend in both their share of value-added and their input-output linkages with other sectors. According to corollary 1, the impact of shocks of capital-intensive sectors on the macroeconomic cycle is likely to increase significantly across business cycles. To validate this hypothesis, Figure 6 illustrates the contributions of capital-intensive and labor-intensive sectors to economic fluctuations during the two business cycles^[1]. In the figure, the grey dashed line with triangle markers represents economic fluctuations driven solely by shocks of laborintensive sectors, while the grey solid line with square markers represents economic fluctuations driven solely by shocks of capital-intensive sectors. Quantitative results show that during the 1992–2002 business cycle, shocks to labor-intensive sectors accounted for 46% of economic fluctuations, whereas shocks of capital-intensive sectors accounted for only 14%, significantly lower than the former. In contrast, during the 2003–2018 business cycle, shocks of capital-intensive sectors explained as much as 61% of economic fluctuations, while shocks of labor-intensive sectors accounted for only about 12%. This demonstrates that, over time, the contribution of labor-intensive sectors to economic fluctuations has declined, while that of capital-intensive sectors has increased.

Figure 6

Network Structure Transformation and Dominant Industry Transformation

Figure 7

Construction and Real Estate Sectoral Shocks and Macroeconomic Fluctuations

4.3 Key Nodes of the Network and Business Cycles

Based on the identification results of sectoral shocks, the shocks of the construction and real estate sector have long contributed significantly to China’s economic fluctuations. Equations (7) and (8) indicate that the impact of such key sectors on macroeconomic fluctuations can be divided into three effects: the within-sector effect generated by the sector’s own shocks, the between-sector effect generated by the sector’s own shocks through input-output linkages, and the passive effect resulting from shocks in upstream and downstream sectors transmitted through input-output linkages to the key sector. Figure 7(a) illustrates the active effect of the construction and real estate sector on the macroeconomy. The black solid line in the figure represents the macroeconomic fluctuations generated by shocks of the construction and real estate sector. It can be observed that the economic fluctuations driven by the construction and real estate sector contributed to more than one-third of macroeconomic fluctuations for most of the time. Next, by shutting down the input-output linkages between the real estate sector and other sectors, we obtain the within-sector effect of its shocks, as shown by the grey line in Figure 7(a). The results show that the between-sector effect is approximately four times that of the within-sector effect, indicating that the primary impact of shocks of the construction and real estate sector on the macroeconomy is mainly reflected in its influence on other sectors. Figure 7(b) illustrates the passive effect of the construction and real estate sector on the macroeconomy. The results reveal that shocks from upstream and downstream sectors, by affecting the construction and real estate sector, also explain nearly one-fifth of macroeconomic fluctuations. When input-output linkages are shut down, shocks from other sectors can no longer be transmitted to the construction and real estate sector, causing the passive effect to disappear. In summary, the characteristic of the real estate sector as a key node in China’s production network is that its impact on the macroeconomy is primarily realized directly or indirectly through input-output linkages.