Asymmetric tail risk contagion across China’s automotive industrial chain: a study based on input–output network

Abstract

The input-output network of an industrial chain provides a channel for risk transmission. Using Smooth-Transition Vector Autoregression model (STVAR) and Diebold-Yilmaz directional connectedness measures, we explore tail risk (extreme risk) contagion across China’s automotive industrial chain. We find significant spillover effects that are asymmetric in different phases of China’s business cycle, monetary cycle, and policy uncertainty. When China’s economy is in a recession, under a monetary expansion, or at a high level of policy uncertainty, the total risk spillover across the chain is higher. We also find apparent risk spillover from the financial services industries to the automotive industrial chain as China’s economy is in a recession or a monetary expansion period. Still, a reverse spillover is found as policy uncertainty is at a high level. Meanwhile, the direction of risk propagation across the automotive industrial chain may change with the transition in the economic state or policy uncertainty state.

KEYWORDS:

• Tail risk
• input–output linkage
• asymmetric contagion
• China’s automotive industrial chain

JEL Classifications:

• C34
• E32
• E42
• F13
• L62

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 The data for some variables in our research, such as the manufacturing PMI index, is only available from January 2005.

2 The stock index of China’s insurance industry starts from January 2007, so we set its CoVaRs between January 2005 and November 2006 constant, equaling the CoVaR of January 2007.

3 We collect data on China’s manufacturing PMI, CCI, MSR, and IOR from the WIND database of China.

4 The Trade Policy Uncertainty Index (based on mainland newspapers) was made by Davis et al. (2019). It can be found at www.policyuncertainty.com. The Exchange Rate and Capital Account Policy Uncertainty Index (based on the full sample of 114 mainland newspapers) constructed by Huang and Luk (2019) is available at http://economicpolicyuncertaintyinchina.weebly.com.

9 We use two composite indicators to check whether the probability of the economy being in recession matches the recession dates in China’s economy. One is the Macroeconomic Climate Index (MCI), published by China’s National Bureau of Statistics. The other one is the Composite Leading Indicator (CLI), designed by the OECD. Based on MCI, we find 77 months are in recession, occupying about 38.9% of the time in our sample period. Based on CLI, we find 76 months in our sample period are identified as recession dates, which occupy about 38.4% of the total 198 sample months. In our results, the number of months that the transition function F(z_t) of PMI or CCI is smaller than 0.5 is 74, occupying about 37.4% of the total. According to MCI and CLI, the likelihood that the economy is in a recession roughly corresponds with the recession dates in China’s economy.

10 In the NBER’s convention for measuring the duration of a recession, the first month of the recession is the month following the peak, and the last month is the month of the trough.

5 Based on the Schwarz information criterion, we choose two lags. The estimated smoothness parameter γ and threshold parameter c, defined in Equation 8, are 7.7 and −0.4 for PMI and 6.3 and -0.6 for CCI, respectively.

6 We check the robustness of the results with respect to the different ordering of endogenous variables. In our STVAR model, there are 9 endogenous variables, so there are 9!  = 362880 possible orderings. For computational considerations, we divide all orderings into three subgroups according to the position of the state variable, i.e., the state variable is put at first, middle, and last, respectively. Then, ten different orderings are randomly chosen from each subgroup. Throughout, we find that ‘risk spillover from one industry to the other', ‘total risk spillover’ and ‘net risk spillover' are completely unchanged.

7 Based on the Schwarz information criterion, we choose two lags. The estimated smoothness parameter γ and threshold parameter c, defined in Equation 8, are 5.4 and 0.7 MSR, and 6.6 and -0.4 for IOR, respectively.

8 Based on the Schwarz information criterion, we choose two lags. The estimated smoothness parameter γ and threshold parameter c, defined in Equation 8, are 3.7 and 0.2 for TPU and 5.2 and 0.9 for ECPU, respectively.

Additional information

Funding

This work was supported by National Natural Science Foundation of China: [Grant Number 72171100]; Humanities and Social Science Fund of Ministry of Education of China: [Grant Number 18YJA790037].