Oil price shocks and China’s consumer and entrepreneur sentiment: a Bayesian structural VAR approach

Abstract

This paper studies the effect of oil price shocks on China’s consumer and entrepreneur sentiment using a novel Bayesian inference structural vector autoregression model. Interestingly, we find that oil supply and demand shocks that raise oil prices have significantly positive effects on both consumer and entrepreneur sentiment. These effects are more significant on entrepreneur sentiment than on consumer sentiment. Furthermore, oil price shocks promote consumer sentiment mainly by increasing their satisfaction with current income and their expectation of future employment. Oil price shocks would change consumers’ saving and consumption decisions but not their plans to buy cars. Meanwhile, the effect of oil price shocks on entrepreneur sentiment differs across different types of enterprises and industries.

1 Introduction

Expectation and confidence factors become more critical as uncertainty increases in world economic development (Lamont and Stein 2006; Baker and Wurgler 2007; McLean and Zhao 2014; Känzig 2021). Stabilizing expectations has become the main task of China’s economic growth under external shocks and the domestic economic slowdown since 2018.¹ It demands that consumers’ and entrepreneurs’ confidence or sentiment be stable when confronting external shocks. In this context, we study how structural international oil price shocks affect China’s consumers’ and entrepreneurs’ sentiment, their evaluation of the current financial situation and expectations about future economic conditions. International oil price shocks provide a proper proxy for external shocks for China because of the frequent booms and busts of oil prices and China’s heavy reliance on oil imports.

The literature shows that the direct effect of oil price shocks on consumers and firms is relatively small and bounded by the share of oil expenditures or costs. Instead, the main transmission channel is the uncertainty effect of oil price shocks, amplified by consumers’ and firms’ expectations. Responding to oil price increases, consumers would depress their purchases of energy-using durable goods, automobiles, for instance, in light of the actual or perceived change in the purchasing power of their disposable income (Hamilton 1988). This direct demand effect is limited to the share of energy expenditures in consumers’ total spending, around 10% in the OECD countries (Herrera et al. 2015).² Oil price shocks also entail a direct cost effect on oil-consuming firms (Rotemberg and Woodford 1996), which is bounded by the cost share of oil in firms’ production (Baumeister and Kilian 2016; Hamilton 2008; Herrera et al. 2015). Besides the direct effect of oil price shocks, consumers reduce consumption out of precautionary saving motivation because oil price shocks possibly reduce their future employment and income due to firms’ responses and sectoral relocation of labors (Davis and Haltiwanger 2001; Edelstein and Kilian 2007, 2009; Ramey and Vine 2011). The reduction in consumption due to uncertainty and precautionary saving, in turn, would cause more unemployment and further decreases in consumption (Hamilton 1988; Baumeister and Kilian 2016). Meanwhile, oil price shocks might lead to an irreversible investment effect that firms postpone their investment in attempting to find out whether the increases in oil prices are transitory or permanent (Bernanke 1983; Lee et al. 2011).³

Regardless of the importance of the uncertainty effect and consumers’ and firms’ expectations, only a few papers study the impact of oil price shocks on investor sentiment and US consumer sentiment. Qadan and Nama (2018) and Ye et al. (2020) study how oil price shocks affect stock market investor sentiment. Edelstein and Kilian (2009) study the response of US consumer sentiment to gasoline price shocks. Wong (2015) investigates whether oil price changes affect the inflation expectations of US consumers, while Güntner and Linsbauer (2018) study how oil supply and demand shocks impact US consumer sentiment.

This paper investigates how structural oil price shocks impact China’s consumer and entrepreneur sentiment. The lens of sentiment can help us understand how China’s economic expectation responds to external shocks and how oil price shocks affect China’s economy since there is no apparent co-movement between oil prices and China’s macroeconomic variables. Consumer and entrepreneur sentiment is proxied by the sentiment indices over (monthly or quarterly) 2001–2019. Consumer and entrepreneur sentiment indices are the leading indicators for the economy’s health, compiled based on survey data by the China Economic Prosperity Monitoring Center of the National Bureau of Statistics of China. When consumer and entrepreneur sentiment indices are high, the economy is likely stable and healthy. By use of the Bayesian inference structural vector autoregressive (VAR) approach recently developed by Baumeister and Hamilton (2019), we consider the impulse response functions of consumer and entrepreneur sentiment to four sources of oil price shocks, namely, oil supply, aggregate demand, consumption demand, and inventory (speculation) demand shocks. On top of the aggregate indices of consumer and entrepreneur sentiment, we also consider the component indices or the sentiment index in different types of enterprises and industries to facilitate our understanding of the specific transmission channels of oil price shocks.

We contribute to the literature in three aspects. First, we focus on the nexus of oil price shocks and sentiment in an oil-importing country, China. To the best of our knowledge, this is likely to be the first paper studying the effect of oil price shocks on China’s consumer sentiment. Unlike the USA, China is highly dependent on oil imports, accounting for 70.8% of crude oil in 2019.⁴ Meanwhile, China has become a vital player in the international oil market since China’s oil consumption has increased steadily over the last decade and reached 13.5% of total world oil consumption in 2018.⁵ The linkage between oil price shocks and sentiment in China might be divergent from the USA, as shown in the literature. For instance, Michelis et al. (2019) show that the impact of oil price shocks on consumption varies between oil-importing and -exporting economies. Second, we consider how oil price shocks affect entrepreneur sentiment on top of consumer sentiment. In doing so, we answer the controversial question of whether the demand side or the supply side is more important in transmitting the impact of oil price shocks through the lens of the responses of non-fundamentals, namely, sentiment. Consumer and entrepreneur sentiment are the leading indicators of consumption and production. The responses of these two variables can reveal the micro mechanism behind the impact of oil price shocks. Third, we adopt the Bayesian inference structural vector autoregression (SVAR) method recently developed by Baumeister and Hamilton (2019) to disentangle the effects of different oil price shocks. Compared with the VAR method by Kilian (2009), the Bayesian inference SVAR method does not impose arbitrary identification assumptions on the underlying structure (identification assumptions). It has the merit of drawing on all available information about the structure and acknowledges this information is imperfect.

We show that oil price increases due to supply disruption, aggregate demand, and consumption demand shocks have a significantly positive effect on consumers’ sentiment in China, mainly by driving up consumers’ expectations of future inflation and employment. Among them, the impact of aggregate demand shock is the most significant. The positive impact of oil price shocks on consumer sentiment seems counterintuitive and conflicting with Güntner and Linsbauer (2018) and Clerides et al. (2021) who find a significantly negative impact of oil price shocks on consumer confidence in the USA and the Euro Area. However, the seemingly conflicting findings might be reconciled by the differences between China and the USA. China has a unique oil product pricing mechanism that prevents full pass-through of increasing international oil prices to China’s consumers. China’s consumers will not likely benefit much from international crude oil price decreases. In contrast, they may benefit from international crude oil price increases, enjoying higher disposable income and better employment conditions. Higher international oil prices can boost China’s output and exports, as shown by Faria et al. (2009), Wei and Guo (2016), Wei (2019), and Azad and Serletis (2022). Our estimation with the data on China’s GDP and exports further confirms the positive role of oil price shocks on China’s GDP and exports.

Similarly, we find that oil price increases due to supply disruption, aggregate demand, and consumption demand shocks have significantly and persistently positive effects on China’s entrepreneur sentiment. Entrepreneurs in most industries feel more confident in response to positive oil price shocks. Furthermore, the impact of oil price shocks on entrepreneurs’ sentiment is much larger than on consumers’ sentiment. Thus, China’s entrepreneurs are more sensitive to oil price shocks than consumers. This differs from Kilian and Murphy (2014) and Baumeister and Kilian (2016), who show the consumption channel is more important for the USA. The reason might be due to the differences in oil consumption structure and retail oil pricing between the USA and China. In the USA, oil prices are determined by the market, and 75% of oil is consumed by households. In comparison, retail oil prices in China are regulated by the government and a large share of oil is used as a raw material for production.

The remainder of this paper is organized as follows. Section 2 presents the econometric methodology and identification strategy with the data for our empirical analysis described in Sect. 3. Section 4 elaborates on our empirical results of impulse response functions, followed by robustness checks in Sect. 5. Finally, Sect. 6 concludes.

2 Methodology

Kilian (2009) first disentangles three different sources of oil price shocks based on a linear structural VAR model. He shows that demand shocks due to global economic expansion and oil market-specific demand shocks significantly drive up oil prices. In contrast, oil supply disruption has a limited effect on oil prices. Afterward, this framework has been extensively used in studying the relationship between international oil price fluctuations and macroeconomic variables. To identify the structural model, Kilian (2009) and other successive papers assume a triangular or diagonal underlying structure concerning the three shocks, known as the Cholesky identification assumption. However, Baumeister and Hamilton (2019) impose significant doubts about the identifying assumption and propose to draw on all available information about the underlying structure with Bayesian inference based on prior information and the observed data (BHSVAR).⁶ They show that the Cholesky identification approach can be viewed as a special case of the more general Bayesian approach, which incorporates prior information about the underlying structure in Bayesian prior distributions. Considering these improvements to Kilian’s VAR model, we apply the BHSVAR model in studying the impact of oil price shocks on sentiment in China.⁷

The typical dynamic structural model is:

$$\begin{aligned} {{\textbf {A}}}{{y}_{t}}={{\textbf {B}}}{{x}_{t-1}}+{{u}_{t}} \end{aligned}$$

(1)

where $y_t$ is an ($n\times 1$) vector of observed variables, A an ($n\times n$) structural matrix summarizing their contemporaneous structural relations, $x_{t-1}$ a ($(mn+1)\times 1$) vector with a constant and m lags of y $({{{x}'}_{t-1}}=({{{y}'}_{t-1}},{{{y}'}_{t-2}},\ldots ,{{{y}'}_{t-m}},1{)}')$, ${u}_{t}$ a vector of structural errors with diagonal variance matrix ${{\textbf {D}}}$.

According to Baumeister and Hamilton (2019), the prior information about A is represented in the form of a density $p({{\textbf {A}}})$. A more plausible value of A is assigned a larger $p({{\textbf {A}}})$ and $p({{\textbf {A}}})=0$ for any impossible value. The prior information may involve individual elements of A or nonlinear combinations such as elements in ${{\textbf {A}}}^{-1}$. Then, the prior information about D conditional on A is represented as $p({{\textbf {D}}}|{{\textbf {A}}})$ and that of lag structure coefficients B conditional on both A and D is

$$\begin{aligned} p({{\textbf {B}}}|{{\textbf {D}}},{{\textbf {A}}}) \end{aligned}$$

$$\begin{aligned} p({{\textbf {D}}}|{{\textbf {A}}})&=\prod _{i=1}^{n}p({{d}_{ii}}|{{\textbf {A}}}) \end{aligned}$$

(2)

$$\begin{aligned} p({{\textbf {B}}}|{{\textbf {D}}},{{\textbf {A}}})&=\prod _{i=1}^{n}p({{b}_{i}}|{{\textbf {D}}},{{\textbf {A}}}) \end{aligned}$$

(3)

Starting from the prior information $p({{\textbf {A}}},{{\textbf {D}}},{{\textbf {B}}})$ represented by the product of $p({{\textbf {A}}})$ with (2) and (3), the Bayesian inference approach updates the belief. According to the Bayes’s rule, it revises the prior information with the observed data $Y_T$ with sample size T. The posterior distribution then follows:

$$\begin{aligned} p({{\textbf {A}}},{{\textbf {D}}},{{\textbf {B}}}|Y_T) =p({{\textbf {A}}}|Y_T)p({{\textbf {D}}}|{{\textbf {A}}},Y_T)p({{\textbf {B}}}|{{\textbf {A}}},{{\textbf {D}}},Y_T) \end{aligned}$$

This posterior distribution summarizes the researcher’s uncertainty about parameters on observing the sample $Y_T$. However, if the model is under-identified, some uncertainty will remain even if the sample size T is infinite. Therefore, the BHSVAR only partially solves the problems concerning the presumed identifying assumptions in the structural model (Kilian and Zhou 2018). It still needs some identifying assumptions to avoid under-identification and reach accurate results.⁸

As with Baumeister and Hamilton (2019), the structural model in the oil market consists of the following four equations:

$$\begin{aligned} {{q}_{t}}&={{a}_{qy}}{{y}_{t}}+{{a}_{qp}}{{p}_{t}}+{{{{b}'}}_{1}}{{x}_{t-1}}+{{u^*}_{1t}} \end{aligned}$$

(4)

$$\begin{aligned} {{y}_{t}}&={{a}_{yq}}{{q}_{t}}+{{a}_{yp}}{{p}_{t}}+{{{{b}'}}_{2}}{{x}_{t-1}}+{{u^*}_{2t}} \end{aligned}$$

(5)

$$\begin{aligned} {{q}_{t}}&={{\beta }_{qp}}{{p}_{t}}+{{\beta }_{qy}}{{y}_{t}}+\Delta {{i}_{t}}+{{{{b}'}}_{3}}{{x}_{t-1}}+{{u^*}_{3t}} \end{aligned}$$

(6)

$$\begin{aligned} \Delta {{i}^*_{t}}&={{\psi ^* }_{1}}{{q}_{t}}+{{\psi ^* }_{2}}{{y}_{t}}+{{\psi ^* }_{3}}{{p}_{t}}+{{{{b}'}}_{4}}{{x}_{t-1}}+{{u^*}_{4t}} \end{aligned}$$

(7)

Equation (4) is the oil supply curve, in which ${a}_{qy}$ is the short-run price elasticity of supply and ${a}_{qp}$ allows for the possibility that economic activity could enter into the supply decision for reasons other than its effect on price. Oil supply is also presumed to be influenced by lagged values of all the variables over the preceding m months or quarters, with ${{{x}'}_{t-1}}=({{{y}'}_{t-1}},{{{y}'}_{t-2}},\ldots ,{{{y}'}_{t-m}},1{)}'$. Equation (5) models the determinants of economic activity, with the contemporaneous effects of oil production and oil prices given by ${a}_{yq}$ and ${a}_{yp}$, respectively. Equation (6) governs oil demand, written here in the inverse form so that $a_{pq}$ is the reciprocal of the short-run price elasticity of demand. Equation (7) acknowledges the possible change in oil inventories in response to oil supply and demand changes. Therefore, ${u^*}_{1t}$, ${u^*}_{2t}$, and ${u^*}_{3t}$ represent shocks to oil supply, economic activity, and oil consumption demand.

Note that Baumeister and Hamilton (2019) impose two zero restrictions that $a_{qy}=0$ and $a_{yq}=0$, assuming that feedback effects between oil production to economic activity are zero. Furthermore, as there is no good data on global oil inventories, an estimate of crude oil inventories is constructed from OECD petroleum product inventories as with Kilian and Murphy (2014):

$$\begin{aligned} \Delta i_t=\chi \Delta ^* i_t \end{aligned}$$

(8)

where $\Delta i_t$ denotes the estimated changes in OECD crude oil inventories, which only represents part of ($\chi$) the total world changes.⁹

To study the impact of oil price shocks from these four sources on sentiment in China, we extend the structural model of Baumeister and Hamilton (2019) by adding the sentiment variable (g) into the model. The extended structural model is specified as:

$$\begin{aligned} \left\{ \begin{matrix} {{q}_{t}}={{a}_{qp}}{{p}_{t}}+{{{{b}'}}_{1}}{{x}_{t-1}}+{{u^*}_{1t}} \\ {{y}_{t}}={{a}_{yp}}{{p}_{t}}+{{{{b}'}}_{2}}{{x}_{t-1}}+{{u^*}_{2t}} \\ {{q}_{t}}={{\beta }_{qp}}{{p}_{t}}+{{\beta }_{qy}}{{y}_{t}}+\chi ^{-1}\Delta {{i}_{t}}+{{{{b}'}}_{3}}{{x}_{t-1}}+{{u^*}_{3t}} \\ \Delta {{i}_{t}}={{\psi }_{1}}{{q}_{t}}+{{\psi }_{2}}{{y}_{t}}+{{\psi }_{3}}{{p}_{t}}+{{{{b}'}}_{4}}{{x}_{t-1}}+\chi {{u}_{4t^*}} \\ {{g}_{t}}={{a}_{qg}}{{q}_{t}}+{{a}_{gy}}{{y}_{t}}+{{a}_{pg}}{{p}_{t}}+{{a}_{ig}}\Delta {{i}_{t}}+{{{{b}'}}_{5}}{{x}_{t-1}}+{{u}_{5t}} \\ \end{matrix} \right. \end{aligned}$$

where $\psi _j=\chi \psi ^*_j$ for $j=1,2,3$. This new system of equations can be expressed in the form of (1) as:

$$\begin{aligned} {{\textbf {y}}}_t=(q_t,y_t,p_t,\Delta i_t, g_t) \end{aligned}$$

$$\begin{aligned} {{\textbf {A}}}=\left[ \begin{matrix} 1 &{} 0 &{} -{{a}_{qp}} &{} 0 &{} 0 \\ 0 &{} 1 &{} -{{a}_{yp}} &{} 0 &{} 0 \\ 1 &{} -{{\beta }_{qy}} &{} -{{\beta }_{qp}} &{} -{{\chi }^{-1}} &{} 0 \\ {{\psi }_{1}} &{} {{\psi }_{2}} &{} {{\psi }_{3}} &{} 1 &{} 0 \\ -{{a}_{qg}} &{} -{{a}_{yg}} &{} -{{a}_{pg}} &{} -{{a}_{ig}} &{} 1 \\ \end{matrix} \right] , {{{{\textbf {u}}}}_{t}}=\left[ \begin{matrix} u_{1t}^{*} \\ u_{2t}^{*} \\ u_{3t}^{*} \\ \chi u_{4t}^{*} \\ u_{5t}^{*} \\ \end{matrix} \right] \end{aligned}$$

(9)

In this structural model, we impose the four zero restrictions in the last column of the coefficient matrix A for identification. For these four zero restrictions to hold, it is necessary that the sentiment variable in China has no contemporary feedback effects on oil market variables. As long as there is no contemporary feedback effect, the four zero restrictions still hold when the sentiment variable in China has lagged feedback effects on oil market variables. These restrictions are likely to hold because not a few studies find the world oil price affects China’s macroeconomy significantly but not the other way around (Chen et al. 2009; Du et al. 2010). China’s factor is proven to have no statistically significant impact on the world oil price (Mu and Ye 2011; Wu and Zhang 2014; Cross and Nguyen 2017).¹⁰

To identify the structural model in Equation (9), we specify the prior information for A, as with Baumeister and Hamilton (2019). We can use the same specification because the parameters in the international oil market are not affected by China’s sentiment indices, at least not contemporarily. As an oil-importing country, China has no substantial influence on international crude oil supply or oil price. Oil inventory in China only takes a small proportion of the world oil inventory, mostly held by OECD countries. Meanwhile, although China’s increasing industrial production is important for global economic activity, steady industrial production growth in China would not alter these parameters. A prior for $\beta _{qp}<0$ is represented by the truncated negative part of a Student $t (c^\beta _{qp},\sigma ^\beta _{qp},\nu ^\beta _{qp})=(-0.1,0.2, 3)$. Similarly, a prior for $a_{qp}>0$ is represented by the truncated positive part of a Student $t (c^a_{qp},\sigma ^a_{qp},\nu ^a_{qp})=(0.1,0.2, 3)$. The prior of the income elasticity of oil demand $\beta _{qy}>0$ also follows the positive part of a Student t-distribution with the location parameter being 0.7, the scale parameter 0.2, and the degree of freedom 3. For the inventory demand equation parameters, $\psi _2$ is set at zero, implying oil inventories depend on income only indirectly through the effects of income on quantity or price. The prior for both $\psi _1$ and $\psi _3$ is unrestricted Student t centered at zero with scale parameter 0.5 and the degree of freedom 3. For the prior parameter D given A, the prior of the reciprocal of structural variance follows the independent gamma distribution. The rest of the parameter settings are also referred to Table 1 in Baumeister and Hamilton (2019, p. 1892).

3 Data

The data on global oil production, global economic activity, real oil prices, and inventory demand are all monthly from January 2001 to October 2019.¹¹ Yet, the data on some sentiment variables are quarterly. In particular, the data on global oil production $q_t$ come from the United States Energy Information Agency (EIA). The raw data of $q_t$ are log-transformed and taken first-order difference to reduce heteroscedasticity. Global economic activity $y_t$ is indicated by the index of world industrial production recently proposed by Baumeister and Hamilton (2019) and Hamilton (2021). Before them, many papers usually indicate global economic activity by Kilian (2009)’s index of global economic activity. It is constructed by the one-way shipping rate of dry cargo ships, reflecting the change in global industrial bulk commodity demand for raw materials (Kilian 2009). We do not use Kilian (2009)’s index because it shows little resemblance to the change in global economic activity since 2009 and has no statistically significant correlation with annual world real GDP growth rates (Hamilton 2021). The world industrial production index offers a much better indicator of monthly world economic activity than measures developed from shipping costs in most cases.¹² International oil price $p_t$ is measured by WTI crude oil spot price from the EIA and Wind database because of its vast influence worldwide, which is also used in many studies. The raw data of WTI crude oil price is first adapted into the constant value with 1984 as the base year by the monthly CPI data of the USA. Then, it is transformed into real oil prices in Chinese currency with the monthly CPI and RMB-dollar exchange rate data from the National Bureau of Statistics of China.¹³ The data on oil inventory demand is also from the EIA website.

We use two categories of sentiment indices, including consumer sentiment index (CSI) and entrepreneurs sentiment index (ESI) in China, considering data availability.¹⁴ First, the monthly consumer sentiment index is compiled by the China Economic Prosperity Monitoring Center of the National Bureau of Statistics of China.¹⁵ It comprehensively reflects consumers’ evaluations of the current economic situation and their expectations on employment and income status in the coming year and whether it is a good time to consume durable goods.¹⁶ The CSI consists of two sub-indices: consumer expectation index (CEI), reflecting consumer’s expectations of future economic conditions, and consumer satisfaction index (CCI), reflecting consumer’s evaluation of the current financial situation. The value of these three indices ranges from 0 to 200. When the index is above 100, consumers hold a positive sentiment, and a negative one otherwise. Figure 1 depicts the trend of WTI oil price and CSI in China, suggesting a close relationship between oil price and consumer’s sentiment, particularly between 2009 and 2015 when both WTI oil price and CSI increase in fluctuations.

Fig. 1

Trend of WTI oil price (unit: $/barrel) and the CSI

On top of the CSI, we use the quarterly data on the diffusion index of urban depositor confidence compiled by the People’s Bank of China.¹⁷ It includes urban depositor sentiment indices about current price (cprice), employment(cemploy), income (cincome), expectation about future price (eprice), employment (eemploy), and income (eincome), percentage of willing to increase savings (savings), percentage of people who are willing to increase investment in securities (invests), and percentage of people who are planning to buy automobiles in next three months (autos).¹⁸ These indices range from 0 to 100. An index above 50 reflects a state of expansion, whereas an index below 50 indicates a state of contraction.

Fig. 2

Trend of WTI oil price (unit: $/barrel) and the ESI

Second, the entrepreneur sentiment index (ESI) is also compiled by the China Economic Prosperity Monitoring Center of China’s National Bureau of Statistics.¹⁹ It reflects entrepreneurs’ confidence in the macroeconomic environment. Based on the entrepreneur’s opinions and expectations of the company’s external market economic environment and macro policies, the index takes a value ranging from 0 to 200, with 100 being the critical value.²⁰ When it is higher than 100, entrepreneurs expect a prosperous state; otherwise, they expect an unfavorable economic condition.

Furthermore, the ESI is categorized by registration types and by industries. Due to data availability and importance in China’s economy, we consider the ESI in three kinds of enterprises by registration types: state-owned enterprises, limited shareholding companies, and private enterprises.²¹ Furthermore, we consider entrepreneurs’ sentiment in nine industries: accommodation and catering, construction, electricity, mining, industry, manufacture, information transmission, computer services, and software, wholesale and retail trade, real estate, and transportation. Figure 2 depicts the trend of WTI oil price and ESI in China, indicating a close relationship between the two variables.

Table 1 summarizes the data sources and period of all variables. Note that the X–12 seasonal adjustment method is applied to all data to eliminate disturbance of seasonal factors. Before estimation, we use the unit root tests for all series and show that all series are stable after the first difference.

Table 1 Variables description and data sources

4 Results

We estimate the impulse response functions of the endogenous variables in the vector ${{\textbf {y}}}_t=(q_t,y_t,p_t,\Delta i_t, g_t)$ to the four structural oil supply and demand shocks, using the structural VAR model specified in (9) and the prior distribution concerning the elements of A.²² For studying how consumer sentiment and entrepreneur sentiment respond to these oil price shocks, we repeatedly estimate the impulse responses of the sentiment index and its sub-indices or component indices in different sectors and types of companies. Following Baumeister and Hamilton (2019), the time lag of the estimations is set as 1 year, namely $m=12$ for monthly data and $m=4$ for quarterly data.²³ We plot the impulse response functions of a standard positive innovation to these four structural oil supply and demand shocks. A positive oil supply shock tends to decrease the real oil price, whereas positive shocks to oil demand will do otherwise. The intervals between two red dotted lines correspond to 68 percent credibility regions within the impulse response functions, and the solid line represents the Bayesian posterior median.²⁴ For a simple comparison of the effects of different shocks, we symmetrically interpret a positive supply shock that lowers oil prices as a supply disruption that raises prices in the following analysis. Accordingly, the sign of responses of sentiment variables is explained in its opposite.²⁵

Before elaborating on our main results, we first discuss the role of supply and demand shocks in the international oil market to validate the robustness of adding the sentiment variable into the BHSVAR model. Let us take the response functions with the CSI as an example.²⁶ Figure 3 shows that our estimations produce almost the same results as Baumeister and Hamilton (2019). A positive oil supply shock that raises oil production (the first row) lowers real oil prices significantly and persistently. In contrast, a positive shock to global economic activity (the second row) and oil consumption demand (the third row) raises oil supply and price. Also, an increase in oil prices due to inventory demand shocks has a persistent and positive effect on inventory and prices.²⁷ Therefore, as with Baumeister and Hamilton (2019), we also show that oil supply and demand shocks are essential for real oil price changes. This finding contrasts with Kilian (2009) and other studies that find the effect of oil supply shocks on real oil prices is only marginal.

Fig. 3

Impulse response function for the five-variable model described in (9) with the consumer sentiment index being the last variable. Note: Solid lines: Bayesian posterior median; dotted red lines: 68 percent posterior credible sets

4.1 oil price shocks and consumer’s sentiment

Consider first the role of oil price shocks on China’s consumer sentiment from different sources. The corresponding impulse response functions in Fig. 4 indicate how China’s consumers adjust their evaluation of the current financial situation and their expectations about future economic conditions in response to a standard positive oil price shock from oil supply, aggregate demand, oil consumption demand shock, and inventory demand shock.

Fig. 4

Impulse response functions of the consumer sentiment index and its two component indices to one-standard-positive-deviation structural oil market shocks

First, the responses in the first column in Fig. 4 suggest that an exogenous disruption of global oil production, which raises real oil prices, promotes China’s consumer sentiment. This promoting effect is significant almost over the whole horizon. The CSI increases up to over 0.4 index points in seven to eight months and maintains this level. Furthermore, since the CSI is constructed based on the two sub-indices, CEI and CCI, oil price shocks can only affect the CSI by influencing the CEI or the CCI. Responding to oil price increases due to oil supply disruption shocks, the CEI increases up to 0.6 index points in the seventh month and fluctuates at over 0.4 index points in the rest months. However, the response of the CCI, although positively signed, is smaller and only marginally significant after 1 year. Therefore, oil price increases due to oil supply disruption shocks boost China’s consumer confidence mainly by raising their expectations of future economic conditions (CEI).

Second, in response to a positive aggregate demand shock that also increases real oil price (the second column), the CSI initially increases up to 1.0 index points in seven to eight months and then fades into 0.5 index points for the rest months. As such, after an initial bout of optimism, consumers tune down their optimistic sentiment but still maintain a cautious optimism concerning the rising oil price driven by aggregate demand shocks. Similarly, consumers’ optimism is driven mainly by increasing their expectations rather than satisfaction. The increase in the CEI is larger and more persistent than in the CCI. The CEI increases up to 1.2 index points in the eighth month and reaches a stable point of 0.8 index points. However, the long-lasting effect on the CCI is only 0.2 index points and becomes insignificant after nine months.

Third, a standard positive oil consumption demand shock (the third column) also persistently positively affects the CSI. This positive effect of oil consumption demand shock is smaller than oil supply and aggregate demand shocks. The response of the CSI to oil consumption demand shock is marginally significant for the entire horizon, but it is at most 0.15 index points. In the same vein, consumers’ expectation is the main transmission channel through which oil consumption demand shock affects China’s consumer sentiment. The CEI increases to a peak of 0.23 index points and reaches a stable point of around 0.1 index points. On the contrary, the response of the CCI is relatively smaller and transient.

Finally, the role of oil inventory demand shocks (the fourth column) in consumer sentiment is limited. The response of the CSI, although positive-signed, is significant for only two months. Moreover, the responses of the CCI and the CEI are almost the same. Thus, oil inventory demand shocks have no lasting effect on China’s consumer sentiment.

As a result, our analysis above reveals that oil price increases caused by either supply or demand shocks have a persistently positive effect on China’s consumer sentiment. However, the magnitude of the positive effects differs across different sources of shocks. The impact of aggregate demand shocks is more significant than supply disruption shocks followed by oil consumption demand shocks. The effect of inventory demand shocks is limited. Therefore, it is crucial to disentangle different sources of oil price shocks in the oil-price-sentiment nexus. Furthermore, the positive impact of oil price shocks on China’s consumer sentiment is mainly transmitted through increasing consumers’ expectations rather than satisfaction. This aligns with the literature that emphasizes the indirect effect of oil price shocks.

Nevertheless, the positive impact of oil price shocks on consumer sentiment seems counterintuitive and conflicting with the findings in the USA and the Euro Area. For instance, Edelstein and Kilian (2009) find a significantly negative impact of rising energy prices on US consumer confidence. Oil price increases due to sudden oil supply disruption or demand shocks have a significant adverse effect on the US and the Euro Area consumer sentiment (Güntner and Linsbauer 2018; Clerides et al. 2021).

The seemingly conflicting findings might be reconciled by the differences between China and the USA. First, China’s unique oil products pricing mechanism may prevent the full pass-through of increasing international oil prices to domestic consumers (Aastveit et al. 2015). The retail prices of gasoline and diesel in China are regulated by the government.²⁸ According to the current regulation policy, the maximal retail prices of gasoline and diesel are determined and adjusted based on international crude oil prices, considering domestic average processing costs, taxes, reasonable circulation costs, and appropriate profits.²⁹Oil prices are adjusted every ten days based on international crude oil prices. If the change in oil prices is lower than 50 yuan per ton, no adjustment is made, but it is accumulated or offset in the subsequent adjustment. Under such a pricing mechanism, the pass-through effects of oil price shocks on China’s inflation are incomplete and decrease along the price chain (Chen et al. 2020). As shown in Fig. 5, the fluctuation of China’s retail price of oil #97 is much smaller than the WTI crude oil price.

Fig. 5

Historical trend of crude oil price (unit: $/barrel ) and the retail price of gas #97 (unit: RMB/ton ) in China

Second, under the pricing mechanism, international crude oil price decreases do not necessarily mean lower refined oil prices. China’s refined oil price is not likely to drop much even when the international crude oil price plunges. Thus, China’s consumers will not likely benefit much from international crude oil price decreases. On the contrary, Chinese households might lose income because of a lower crude oil price, especially those working in the oil and other energy sectors. Low international crude oil prices would put much pressure on energy prices, lowering the profits in these sectors. Furthermore, when oil price drops and investment related to energy sectors is shelved, their impact reaches much further than the energy sector, such as steel and industrial equipment sectors.

Third, China’s households will likely benefit from international crude oil price increases. They may enjoy higher disposable income and better employment conditions because a higher oil price can boost China’s output and exports. According to Azad and Serletis (2022), oil price uncertainty has a positive and statistically significant effect on China’s real output. Meanwhile, oil price shocks can significantly increase China’s exports (Faria et al. 2009; Wei 2019). Moreover, China’s energy structure in China is coal-dominated. High international crude prices may increase the profits of the oil and energy-related sectors and the disposable income of households working in these sectors.

To analyze further how structural oil supply and demand shocks are transmitted to consumer sentiment, we take a closer look at the impulse response functions of the urban depositor confidence indices. Each column of Fig. 6 plots the impulse responses to a typical positive oil supply, aggregate demand, oil consumption demand shock, and inventory demand shock, respectively. Note that the data on these indices are quarterly, and the horizon includes 12 quarterly, namely 3 years.

Fig. 6

Impulse response functions of urban depositor confidence to one-standard-positive-deviation structural oil market shocks

While we have already shown that oil supply disruption shocks positively affect China’s consumer sentiment, the first column of Fig. 6 further confirms this result. The responses of most urban depositor confidence indices are also positive. Among them, the responses of cincome and eemploy are the most prominent and (marginally) significant in the whole horizon, for which cincome and eemploy increase up to 4 and 3 index points, respectively. The response of the future price expectation index (eprice) is positive but becomes insignificant after six quarters. Hence, oil price increases due to oil supply disruption shocks boost depositors’ confidence mainly by improving their satisfaction concerning current income and their expectations about future employment conditions. Oil price increases due to supply disruption would increase consumers’ inflation expectations, but this inflation-increasing effect is temporary. Moreover, oil supply disruption shocks do not materially affect consumers’ behaviors. The responses of autos, invests, and savings are marginal and statistically insignificant. China’s consumers are not likely to postpone their plans to buy automobiles or adjust their decisions on savings and investment because of oil supply disruption shocks.

In response to a positive aggregate demand shock, the CSI increases to 1.0 index points and then maintains at 0.5 index points. The second column of Fig. 6 further shows that the optimism among consumers comes mainly from their increased satisfaction with current income and better expectations of future income. Responding to aggregate demand shocks, cincome and eincome continually increase to 10 and 5 index points, respectively. Meanwhile, the response of eprice is positive and persistently significant for the whole horizon. Consumers expect that oil price increases due to aggregate demand shocks would cause inflation. With the expectation of inflation, more consumers prefer to decrease their investments and savings, for which the responses of both savings and invests are negative. In particular, the response of savings is persistent and significant, up to 5 index points. Still, similar to the case of oil supply shocks, the response of autos to aggregate demand shocks is also insignificant. Therefore, consumers feel better off in response to aggregate demand shocks. They expect to earn more and save less in a boom of the economy with inflations. Again, oil price increases due to aggregate demand shocks would not alter consumers’ plans to purchase automobiles.

The third column plots the responses of urban depositor confidence indices to a standard positive oil consumption demand shock. The responses of these indices also verify the slight optimism among consumers entailed by oil consumption demand shocks. Responding to oil consumption demand shocks, consumers believe their income will increase while current and future employment conditions improve. The responses of cemploy, cincome, and eemploy are positive and significant for at least 1.5 years, reaching a peak of 1, 1, and 0.5 index points, respectively. Due to oil price increases caused by oil consumption demand shocks, consumers temporarily lower their satisfaction with the current price but expect future prices to be more stable. The response of cprice is negative but only significant for one quarter. The response of eprice becomes significantly negative after three quarters and reaches a stable point of -2 index points. In this case, consumers would save more but invest less, for which the index of savings increases by over two index points while the index of invests continually decreases by five index points. This finding partially coincides with Edelstein and Kilian (2009), who show that oil consumption demand shocks would entail consumers’ precautionary saving motivation. Note further that the effect of oil consumption demand shocks on autos is also insignificant.

Finally, the responses of most urban confidence indices to oil inventory demand shocks (the fourth column) are statistically insignificant. This further indicates the limited impact of oil inventory demand shocks on consumer sentiment. Only the response of invests is persistently negative, where invests decreases by over ten index points. In responding to international oil market speculative (inventory demand) shocks, consumers would divest their investment in financial markets.

In brief, oil price shocks from different sources affect consumer sentiment through different channels. First, oil supply and consumption demand shocks mainly affect consumers’ satisfaction with current income and their expectations of future employment. Aggregate demand shocks increase consumers’ confidence in their current and future income. This result partially coincides with Davis and Haltiwanger (2001) and Karaki (2018), who also reveal the critical role of oil price changes on employment changes. Second, despite oil price increases due to these three shocks, only aggregate demand shocks would raise consumers’ inflation expectations. Third, the role of these shocks on consumers’ behaviors also differs. Oil supply shocks do not have much impact, which is consistent with Alsalman and Karaki (2019), who also show that the effect of oil supply shocks on consumption is limited. Aggregate demand shocks would induce consumers to save less but spend more. In contrast, oil consumption demand shocks would increase consumers’ willingness for more savings.

Interestingly, oil price shocks due to all these four shocks have no significant impact on consumers’ plans to buy cars in the next three months, which is in contrast to the finding of Güntner and Linsbauer (2018) in the USA. It also rejects the common argument of operation cost that oil price increases tend to decrease consumers’ expenditure on energy-intensive durable goods such as automobiles (Hamilton 2008). The reason might be that the car demand in China is relatively rigid because China’s car ownership is still low. In 2018, US car ownership was 790 per thousand people, whereas this number is only 140 in China.³⁰ Responding to oil price increases, consumers will not postpone or give up their purchase plans but shift to small-displacement cars. The oil price for China’s consumers is relatively small compared to the cost of purchasing a car. The share of oil consumption expenditure in consumers’ total spending is much lower in China than in the USA. Liu (2020) also shows that oil price increases do not significantly affect the sales of automobiles in China.

4.2 oil price shocks and entrepreneur sentiment

Figure 7 depicts the impulse response functions of the entrepreneur sentiment index to oil price shocks from the four sources. We highlight three findings.

First, as in the case of consumer sentiment, oil price increases from oil supply, aggregate demand, or consumption demand shocks have persistently positive effects on China’s entrepreneur sentiment. The responses of the ESI to these three oil price shocks are significantly positive for the whole horizon. A positive inventory demand shock has only a transiently positive impact on China’s entrepreneur sentiment. The response of the ESI is positive but only significant for less than one quarter. Intuitively, it is easy to understand the positive role of aggregate demand shocks on entrepreneurs’ confidence. The benefit of global economic activity booming dominates the associated cost effect due to oil price increases. The positive impact of oil supply disruption and consumption demand shocks indicate further that the cost effect of oil price shocks is not a dominant factor in the mind of China’s entrepreneurs, which coincides with the argument of Baumeister and Kilian (2016) in the USA. Indeed, China’s energy-related sectors would benefit rather than suffer from a high oil price due to the coal-dominated energy structure in China.³¹ Other sectors might also benefit from the increasing exports to other countries, especially the US and European countries, that can influence the world oil price (Du et al. 2010). Wei and Guo (2016) also show that oil price shocks positively affect China’s output and exports. We also estimate using the data on China’s GDP and exports and further confirm the positive role of these oil price shocks, although the magnitude and duration of the impact differ.³²

Second, the effect of aggregate demand shocks followed by supply shocks is much more significant than that of oil consumption demand shocks. Responding to oil supply and aggregate demand shocks, the ESI increases up to 8 and 12 index points, respectively. In contrast, the maximal effect of oil consumption demand shocks on the ESI is 2.5 index points. Moreover, the response of the ESI to oil consumption demand shocks is only significant for 2 years. Therefore, China’s entrepreneurs are more sensitive to oil price shocks due to changes in fundamentals, supply disruption, or global economic activity booming than those from exogenous demand increases.

Third, compared with the case of consumer sentiment in Fig. 3, the effect of oil price shocks on entrepreneur sentiment is much larger.³³ For instance, in response to a standard positive aggregate demand shock, the CSI increases up to 1 index point, whereas the ESI rises to 12 index points. As such, China’s entrepreneurs respond to oil price shocks more significantly than consumers. The supply channel in the sense of entrepreneur sentiment is more important than the demand channel in consumer sentiment, by which oil price shocks affect China’s sentiment and macroeconomy. In this sense, the relative importance of supply and demand channels in China is contrary to that in the USA, where Baumeister and Kilian (2016) indicate that the demand channel is more important. The reason might be due to the differences in oil consumption structure and retail oil pricing between the USA and China. In the USA, oil prices are determined by the market, and 75% of oil is consumed by households. In comparison, retail oil prices in China are regulated by the government and a large share of oil is used as a raw material for production.

Fig. 7

Impulse response functions of the entrepreneur sentiment index to one-standard-positive-deviation structural oil market shocks

Nevertheless, without further analysis, we can only conjecture the possible transmitting channels through which structural oil supply and demand shocks affect China’s entrepreneur sentiment. Therefore, we investigate how entrepreneurs in different enterprises and industries respond to these shocks.

State-owned, shareholding, and private companies Figure 8 illustrates the responses of entrepreneur sentiment in state-owned enterprises (ESI_SOEs), limited shareholding companies (ESI_share), and private enterprises (ESI_private). The results confirm the positive effect of oil price shocks on entrepreneur sentiment. However, the magnitude and significance of this positive effect differ. Among them, the response of ESI_share is the most prominent across the four shocks. For instance, responding to either a standard oil supply or aggregate demand innovation, the ESI_share increases up to 12 index points, larger than the increases in the ESI_SOEs and ESI_private. Thus, entrepreneurs in limited shareholding companies feel more optimistic about these oil price shocks. It coincides with the fact that most oil-producing enterprises in China, notably the three major oil companies, are limited shareholding companies.³⁴ They will likely earn more profits and income when international oil prices increase.

Fig. 8

Impulse response functions of China’s ESI in different types of enterprises to one-standard-positive-deviation structural oil market shocks

Meanwhile, entrepreneurs in SOEs are also optimistic about oil price increases due to the three shocks because the response of the ESI_SOEs is almost the same as that of the ESI. Oil supply and aggregate demand shocks induce consistent and significant increases in the ESI_SOEs up to over ten index points for every standard innovation. In contrast, oil consumption demand shocks cause a relatively smaller and temporary improvement. The reason might be that most SOEs are distributed upstream of the industrial chain. When oil price increases, they can pass the increasing cost to the downstream and improve their profits by raising retail oil prices.

In contrast, only the increase in oil price driven by aggregate demand shocks could persistently positively affect ESI_private. The fluctuation of oil prices due to oil supply disruption or consumption speculation demand shocks does not have persistent impacts. The response of ESI_share is statistically significant for less than one quarter and six quarters, respectively. Intuitively, private business owners would benefit significantly from the increased demand for their products in the case of positive aggregate demand shocks, which dominate the possible cost effect. However, when oil price shocks are purely from the oil market, either supply or consumption demand shocks, the benefits of these shocks are limited. Private enterprises, primarily distributed downstream of the industrial chain, would still have to bear the expense of high oil prices as the cost effect passes through the industrial chain with lags. Therefore, entrepreneurs in private enterprises would be prudent after an initial boom of confidence.

Therefore, entrepreneurs in all three enterprises feel optimistic about oil price increases due to aggregate demand shocks. However, only entrepreneurs in state-owned enterprises and shareholding companies hold a persistently positive attitude to supply disruption and consumption demand shocks.

Different industries Figure 9 reveals heterogeneous responses of China’s entrepreneur sentiment in nine industries to oil price shocks due to the four shocks. The heterogeneous responses indicate the importance of industry characteristics in the nexus as with Davis and Haltiwanger (2001). Narayan and Sharma (2011) also show that oil price affects the returns of firms differently depending on their sectoral location. Since China’s economic structure, energy consumption structure, and oil import dependence are different from the USA, these industry characteristics might play a different role in China.

First, an oil supply disruption shock that increases oil prices (the first column) has a persistently and significantly positive impact on the ESI of all industries except for the electricity industry. Among them, the positive effect on the ESI_mining is the most significant, for which one standard supply innovation entails a sharp increase up to over 30 index points. This result is intuitive as the mining industry can benefit from a higher oil price, particularly oil mining and refinery enterprises. Also, as a sub-industry of the mining industry, the coal industry might make a larger profit in the case of high oil prices. The persistently positive effect on ESI_Build, ESI_transp, and ESI_retail provides further evidence against the cost effect argument in China because a higher oil price would increase the cost of these industries. Interestingly, although the information-related industry is not closely related to the oil sector in the industry chain, the entrepreneurs in this industry feel more confident because they may benefit from a better international trade market due to a higher oil price. In contrast, the response of the ESI in the electricity industry (ESI_ely) to oil supply shocks is marginal and insignificant. Entrepreneurs in this industry are generally neutral to oil supply disruption. Therefore, oil supply disruption shocks entail optimism among China’s entrepreneurs in most industries.

Fig. 9

Impulse response functions of China’s ESI in different sectors to one-standard-positive-deviation structural oil market shocks

Second, as with oil supply shocks, aggregate demand shocks (the second column) also cause a persistently significant positive effect on the ESI in all industries except for the electricity industry. The associated cost effect of oil price increases is completely dominated. The positive effect of aggregate demand shocks is even larger. Notably, the responses of the ESI in the accommodation and catering manufacturing industries are more significant, in which ESI_acomm and ESI_manuf consistently increase up to a peak of over 20 and 15 index points, respectively. These two industries will likely benefit from the increased demand caused by aggregate demand shocks. On the other hand, the entrepreneurs in the real estate industry become increasingly pessimistic after three quarters, in which ESI_estate decreases over 20 index points. One possible explanation is that oil price increases from aggregate demand shocks would divest the investment in the real estate industry to other sectors, such as energy and manufacturing.

Third, in response to an oil consumption demand shock (the third column), the ESI in all industries reacts less dramatically, indicating again that oil consumption demand shocks are less crucial than supply and aggregate demand shocks in the mind of China’s entrepreneurs. The responses of the ESI in all industries are positive, but five of them become insignificant after around 1.5 years. This is consistent with the response of the ESI. The optimism caused by oil consumption demand shocks in these five industries is not supported by consistently increasing demands related to aggregate demand shocks. This argument is supported by the insignificant response of the ESI in the accommodation and catering industry, which is most sensitive to aggregate demand variations in the economy.

Finally, the impact of oil inventory demand shocks (the fourth column) on entrepreneur sentiment is short-lived and marginally significant in most industries, which is also in line with the response of the ESI. Therefore, inventory demand shock has no persistent impact on the entrepreneur sentiment in most industries.

In short, the responses of entrepreneur sentiment in different industries confirm the significant positive effect of oil supply, aggregate demand, and consumption demand shocks on China’s entrepreneur sentiment. As with the literature, the cost effect of these shocks is unimportant, which might be dominated by the increased demands for goods and services in a global economic activity boom. The effects of oil supply and aggregate demand shocks are more important than oil consumption market demand shocks across industries, whereas the role of inventory demand shocks is limited. However, not all entrepreneurs feel better about these oil price shocks. The entrepreneur sentiment in the real estate industry is negatively affected by aggregate demand shock, possibly because these shocks would divest the investment in this industry to other sectors.

5 Robustness check

To check the robustness of our results, we estimate three different specifications.³⁵ First, we use Kilian’s VAR model alternatively without oil inventory demand shock and find that our main results remain qualitatively unchanged. The results in Fig. 10 depict the responses of the CSI and ESI to negative supply and positive demand shocks that increase oil prices. Qualitatively similar, oil price increases driven by either supply disruption or aggregate demand shocks can significantly improve consumers’ and entrepreneurs’ confidence. The effect of oil market-specific demand shock is comparatively smaller and less significant, particularly on the CSI.³⁶ Furthermore, oil price shocks have a more significant effect on entrepreneur sentiment than on consumer sentiment.

Fig. 10

Robustness check with Kilian’s model. Note that the two dotted lines represent 95% error bands and two dashed lines 68% error bands

Second, we replicate our analysis using a shorter and longer time lag rather than choosing a lag of 1 year ($m=12$ for monthly data, $m=4$ for quarterly data) following the existing literature. When the time lag becomes half a year ($m=6$ for monthly data, $m=2$ for quarterly data) or 2 years ($m=24$ for monthly data, $m=8$ for quarterly data), the impulse response functions of consumer sentiment are the same, and those of entrepreneur sentiment are qualitatively unaffected. Note also that when the time lag becomes half a year ($n=2$ quarter), the role of supply shock is statistically insignificant. However, it is significant for the whole horizon when $n=4$ or $n=8$ quarters. As such, a sufficiently high lag order is necessary for capturing the effect of oil supply shocks (Fig. 11).

Fig. 11

Robustness check with different time lags

Last, we replace the WTI oil price with the Brent oil price and conduct the same estimations. Figure 12 indicates that our main results are virtually unaffected, both qualitatively and quantitatively.

Fig. 12

Robustness check with Brent oil price

6 Conclusions

The economic consequences of oil price shocks may be passed through either the demand side by affecting consumers’ consumption or the supply side by affecting firms’ production decisions. However, most of the literature focuses on the effects of oil price shocks on GDP, inflation, and stock market return. It does not consider how oil price shocks affect consumers’ consumption and firms’ production, particularly the relative importance of these two channels. Meanwhile, the literature agrees that the uncertainty or indirect effect of oil price shocks is more important than its direct cost effect. However, few papers study how economic agents assess and respond to oil price shocks. Motivated by the gap in the literature and by China’s increasingly important role in the global oil market, we study the impact of structural oil price shocks from four sources on consumer and entrepreneur sentiment, which are the leading indicators of consumption and production of the economy.

We find that oil price increases due to supply shock, aggregate demand, and consumption demand shocks can significantly boost consumers’ and entrepreneurs’ confidence. Moreover, the confidence-boosting effect of aggregate demand shocks is the most prominent. This result provides a micro foundation for the positive relationship between oil price and output in China revealed in the literature, for instance, Du et al. (2010). However, it is contrary to Güntner and Linsbauer (2018), who show a negative effect of oil price shock on US consumer sentiment. Thus, the nexus of oil price shocks and sentiment in China differs from that in the USA. This indicates the importance of economic and energy consumption structures in the nexus of oil price shocks and sentiment. The significant and persistent impact of oil supply shocks also contrasts with the minor and temporary effect in Güntner and Linsbauer (2018). In responding to oil supply shocks, China’s consumers will not “expect these shocks to be only temporary” but expect their income to increase and future employment conditions to improve. The different results are not due to model specifications between Kilian (2009) and Baumeister and Hamilton (2019) since we can replicate the same results using Kilian’s model.

We show that the role of oil price shocks on entrepreneur sentiment is much more significant than that on consumer sentiment across different sources of shocks in China, suggesting that China’s entrepreneurs are more sensitive to these oil price shocks than consumers. Accordingly, the economic consequences of oil price shocks are more important on the supply side in China’s economy. This finding differs from the case of the USA, where Baumeister and Kilian (2016) argue that the demand side is more crucial. This may be explained by the differences in oil consumption structure and retail oil pricing between the USA and China.

Furthermore, oil price shocks affect China’s consumer sentiment mainly by influencing their expectations of future employment conditions and income. Consumers would save less and spend more in response to aggregate demand shocks but save more in the face of oil consumption shocks. China’s consumers will likely keep their plans to buy cars when the oil price increases due to these shocks. Moreover, oil price shocks from different sources entail a diverse effect across various enterprises and industries in China. The impact on the entrepreneur sentiment in limited shareholding companies is more significant than that in private and state-owned enterprises because limited shareholding companies dominate China’s energy sector. Corresponding with this, the response of the entrepreneur sentiment in the mining industry is the most prominent.

One significant implication of our findings is that policymakers should watch out for the decrease rather than the rise of international oil prices, resulting in a persistent reduction in consumers’ and entrepreneurs’ confidence. Furthermore, the effects of oil price shocks on entrepreneurs’ confidence vary across different types of enterprises and industries. Policymakers should also tailor their policies to key enterprises and industries most affected by oil price fluctuations, such as the mining and transportation industry. Our study is limited to the effect of oil price shocks on sentiment in China. It would be interesting to check these results further by using the microdata of consumers’ expenditures and entrepreneurs’ investments and capacity utilization.