Price Spillover Effects in U.S.-China Cotton and Cotton Yarn Futures Markets Under Emergency Events

Abstract

As a strategic material second only to grain, cotton serves both as a vital agricultural commodity and a key industrial crop. With the increasing frequency of global shocks and the deepening financialization of commodity markets, price linkages among major international cotton futures markets have strengthened. Consequently, in addition to fundamental supply and demand factors, cross-border price transmission has become a significant determinant of cotton pricing. This study employs daily closing prices of China’s cotton futures, cotton yarn futures, and U.S. cotton futures from 1 September 2017 to 31 March 2025 to examine the spillover effects among these three futures markets using time series models. The results reveal that U.S. cotton futures have dominated the Chinese cotton-related futures markets even prior to the onset of trade tensions, with strong domestic market comovements. However, both the U.S.-China trade war and the COVID-19 pandemic significantly weakened price co-movements while intensifying volatility spillovers. Although these external shocks enhanced the relative independence of China’s cotton yarn futures and modestly increased China’s pricing influence, U.S. cotton futures have consistently maintained their central role in price discovery.

1. Introduction

Cotton serves as a critical commodity integral to national strategic industries and socioeconomic well-being, bridging the agricultural and manufacturing sectors while playing a highly significant role in daily life and economic activities [1]. China dominates global cotton production, accounting for 26% of worldwide output (121 million bales in 2024, USDA Foreign Agricultural Service), while also leading in consumption and imports. According to data from the National Bureau of Statistics of China, China’s cotton consumption reached 7.69 million metric tons, and its import volume stood at 2.618 million metric tons in 2024. The United States, ranking as the second-largest exporter (2558 thousand metric tons), trails only Brazil (2680 thousand metric tons). U.S.-China cotton trade exemplifies deep interdependence: America serves as China’s critical cotton supplier while importing substantial textile products from China, creating bidirectional market linkages. U.S. imports of cotton textile products totaled 17.17 billion square meters, of which China supplied 3.738 billion square meters, constituting 21.77% of total American cotton textile imports in 2024. The highly pronounced complementary strengths between China and the United States in the cotton textile sector offer an exceptionally favorable basis for establishing enduring and stable bilateral cooperation [2].

In 2018, the United States imposed a 25% tariff on over $34 billion worth of goods imported from China, prompting an immediate countermeasure from China, which imposed an equivalent 25% tariff on a range of U.S. imports. This marked the beginning of a prolonged trade conflict between the two countries. Over the course of the dispute, there were several rounds of escalation and temporary ceasefires, as well as multiple failed negotiations, all of which had profound impacts on both countries’ foreign trade and associated industries [3,4,5]. As key commodities in international trade, cotton and cotton textiles were severely affected. The cotton supply chain was exposed to intensified disruptions, and global supply–demand dynamics shifted, resulting in heightened price volatility [6]. In 2020, the outbreak of the COVID-19 pandemic brought unprecedented disruption to economies and societies worldwide. China was particularly hard hit, with public health challenges paralleled by significant economic setbacks across sectors, including commerce, tourism, transportation, and healthcare. The impact was widespread and multifaceted, causing direct and indirect economic losses that were difficult to quantify. The cotton and textile supply chain, deeply integrated into global trade, faced extraordinary challenges under the pressure of the pandemic [7]. In January 2020, during the Chinese Spring Festival, the sudden outbreak of the pandemic—an unexpected “black swan” event—led to the rapid spread of the virus. Despite the swift issuance of warnings and containment measures by Chinese authorities, the speed of transmission exceeded expectations. This triggered immediate market reactions: commodity price indices fell across the board, and cotton prices declined sharply in response. However, the prompt and effective public health measures implemented in China led to a continuous decline in confirmed cases, gradually restoring market confidence and injecting momentum into the national economic recovery. By the end of 2020, as the domestic pandemic was largely brought under control, China’s economy began to recover. Meanwhile, the pandemic accelerated globally, creating widespread anxiety. Despite this, the textile industry in China began to resume production, and the cotton market showed signs of stabilization and even partial rebound. Nevertheless, due to the ongoing pandemic, a large number of international orders were canceled, resulting in renewed market stress. As the global economy entered a prolonged downturn, China faced growing external uncertainties. Domestic demand remained weak due to persistent containment measures, industrial supply chains experienced instability, and enterprises faced serious challenges in maintaining normal operations, including excess inventories and declining sales of related products. In 2021, China’s effective pandemic control resulted in a rapid rebound in the textile and apparel markets. A surge in global orders drove a sharp increase in cotton demand. Combined with a reduction in planting area and adverse weather conditions, cotton output declined while prices soared. Although the “Xinjiang cotton” controversy exerted some downward pressure on the market, the decline was short-lived, and prices quickly stabilized and resumed an upward trend. In the first half of 2022, high temperatures kept cotton prices elevated. However, as the risk of a global recession deepened, panic selling swept through U.S. equity and commodity markets. Alongside weakening demand for cotton fabrics, international cotton prices plummeted. Domestic and foreign textile enterprises faced shortages of orders and excessive inventory levels, and cotton demand remained subdued. The production of Xinjiang cotton and its downstream products also came under constraint, contributing to pronounced shifts in China’s cotton futures market. In summary, the prices of cotton futures in both China and the United States have experienced significant volatility, driven by a complex interplay of factors including U.S.-China trade tensions and the COVID-19 pandemic. These events have intensified market risks and disrupted global price-setting mechanisms. Moreover, the price linkage between Chinese and U.S. cotton and cotton yarn futures markets has exhibited dynamic and evolving patterns in response to these external shocks.

Macroscopically, futures markets mitigate price volatility risks across industries, enhancing macroeconomic stability through risk transfer mechanisms [8]. Their price transparency establishes benchmark references for global trade, enabling nations to integrate domestic and international markets. Microeconomically, futures facilitate price discovery, empowering supply chain actors to optimize production planning and resource allocation [9]. Crucially, price fluctuations propagate beyond immediate underlying markets, generating cross-market spillovers [10]. This dynamic compels market participants to analyze multi-market correlations and transmission mechanisms rather than isolated price movements.

Cotton futures contracts were first introduced at the New York Cotton Exchange in the 19th century. After over a century of development, they now rank among the most significant contracts on the New York Mercantile Exchange (NYMEX), wielding substantial global influence and commanding dominant pricing authority in the cotton trade [11]. With over 80% of U.S. cotton production exported, this pricing power secures competitive advantages in international trade. Price fluctuations in U.S. cotton futures profoundly impact global production, processing, and distribution networks, while transmitting volatility risks worldwide [12]. Although launched later, China’s cotton futures market has achieved sustained growth in trading volume alongside stabilized and standardized operations. By 2024, its trading activity had surpassed U.S. volumes by 4.5-fold, attaining considerable international influence. The market further matured with the 2017 debut of cotton yarn futures and the 2019 launch of cotton options, establishing a comprehensive derivatives suite that provides diversified risk management tools for textile enterprises and enhances price discovery efficiency. Analyzing price transmission and volatility spillovers among Chinese cotton futures, cotton yarn futures, and U.S. cotton futures—particularly across varying periods and emergency scenarios—serves three critical purposes: initially evaluating the operational maturity of China’s cotton derivatives complex, subsequently benchmarking pricing efficiency in U.S.-China cotton markets and ultimately informing risk mitigation strategies against international price shocks.

Extensive research has been conducted on price spillover effects among agricultural commodity futures. From a research perspective, the existing literature can be broadly categorized into two dimensions: horizontal linkages across products or regions, and vertical linkages along the agricultural supply chain.

In studies of horizontal linkages, scholars have investigated price comovements and spillover effects across agricultural futures markets in countries such as China and the United States. For instance, Holder and Pace analyzed time series data on corn and soybean futures from the U.S. and Japan, concluding that U.S. futures dominate the price transmission process in both markets [13]. Liu and Tang examined the edible oil futures markets of China, the U.S., and Canada, finding that futures prices in the U.S. and Canada exert greater influence, while China’s futures prices play a relatively passive role internationally [14]. Yang and Xu showed that U.S. soybean-related futures (soybeans, soybean meal, and soybean oil) act as the main source of both mean and volatility spillovers in relation to Chinese soybean futures [15]. Li and Hayes compared the soybean futures markets of the U.S., Brazil, and China, determining that the U.S. market plays a leading role in guiding price changes in the other two markets over the long term [16]. As for cross-regional spillovers in the cotton futures market, Ge et al. analyzed the U.S. and Chinese cotton futures markets and found a long-term cointegration relationship and similar volatility characteristics [17]. Wang et al., using an ARCH model, identified bidirectional spillovers between the two markets, with the U.S. market exerting a stronger influence on China [18]. Wang et al. further confirmed that U.S. cotton futures have consistently played a dominant role, exhibiting a unidirectional guiding effect on Chinese cotton futures in both short- and long-term analyses [19]. In terms of vertical linkages along the agricultural supply chain, Yan et al. compared the soybean industry chains in China and the U.S., finding mean and volatility spillovers in both markets. They noted that China’s soybean futures exhibit a more complete price transmission path along the supply chain, while both countries’ soybean oil and soybean meal futures display bidirectional volatility spillovers [20]. Huang and Wang found that soybean and soybean oil futures unidirectionally influence soybean meal futures prices [21]. In the corn industry chain, Ling and Kong found evidence of bidirectional spillovers between corn and starch futures, with increasing interdependence over time [22]. In the hog industry chain, Fu et al. identified multiple stages of price transmission and a close relationship between hog and grain prices, suggesting price signals propagate along the chain [23]. In the edible oil futures market, Liu and Tang found spillovers from soybean oil to palm oil futures, and from palm oil to rapeseed oil futures, with soybean oil exhibiting the strongest influence. In terms of risk transmission, both soybean oil and palm oil futures were found to transfer risk to rapeseed oil futures [14].

Some researchers have examined spillover effects across several commodity markets. Zuo studied how corn futures influence other agricultural and sideline product markets, asserting corn’s central role within the broader grain system [24]. Xiao and Zhang argued that crude oil prices are closely related to agricultural product prices, with oil price fluctuations significantly influencing markets like corn [25]. Yan and Zhu conducted a risk spillover analysis among agricultural, metal, chemical, and energy futures, finding that volatility spillovers between agricultural futures and the other sectors remain relatively weak [26].

Specific to the cotton market, Zhu and Si observed a strong but asymmetric relationship between spot prices of cotton and cotton yarn [27]. Ding and Xiao analyzed the cotton and wool markets and found that both mean and volatility spillovers varied under different policy regimes, indicating time-varying linkages between cotton and wool prices [28].

In addition, some scholars have examined the impact of unexpected events on the spillover effects among agricultural commodity markets. Such events can influence futures prices through various channels, including industrial linkages [29], market price expectations [30], and speculative behavior [31]. Ji et al. argued that the U.S.-China trade war reduced the degree of connectedness between the two countries’ soybean-related futures markets, weakening the influence of U.S. soybean futures prices on their Chinese counterparts [32]. Similar findings were reported by Yang and Xu, who also observed a decline in spillover intensity from U.S. to Chinese soybean futures during the trade conflict [15]. Jiang et al. contended that price spillovers are not unidirectional from the U.S. to China; Chinese agricultural commodity prices can also affect the U.S. market. For example, during the global financial crisis, they observed a significant spillover from China to the U.S., although the reverse impact remained stronger [33]. Liu et al., in the context of the U.S.-China trade tensions, found that the influence of U.S. soybean prices on China’s soybean futures had weakened due to escalating trade frictions [34]. Liu suggested that, as a response to the trade war, China adopted a more diversified trade strategy and strengthened cooperation with a broader group of countries—particularly those involved in the Belt and Road Initiative—thereby increasing the pricing influence between China’s agricultural products and those from its new trading partners [35]. Similarly, Wu et al. found that tariff policies implemented by both countries led to reduced market interdependence, weakening the influence of U.S. cotton prices on the Chinese market [36]. Qiao and Han discovered that the outbreak of COVID-19 significantly amplified both lower- and upper-tail contagion in commodity markets. Their findings revealed pronounced clustering characteristics, where tail-risk contagion was stronger within commodity classes. Agricultural commodities exhibited lower contagion levels compared to metals and energy, with soft commodities, in particular, offering considerable diversification benefits to investors [37]. Chen et al. observed that during periods of major crises—including the global financial crisis, the COVID-19 pandemic, and the oil market collapse—the dynamic correlation between crude oil and agricultural commodities intensified [38]. Most recently, Gao et al. investigated the effects of the trade war and COVID-19 on soybean futures in China and the United States, concluding that such exogenous shocks exert substantial influence on market dominance and price transmission between the two countries [39].

Research on the spillover effects of agricultural product prices boasts a lengthy history, encompassing diverse research directions and an expanding scope of subjects. Price changes in a single commodity can propagate through various channels, including futures-spot markets, industrial chains, and cross-regional markets. Existing literature consistently indicates that futures prices frequently assume a dominant role in price discovery, with their fluctuations typically serving as precursors to future spot price trends. Consequently, research on futures market pricing holds considerable significance.

As global trade deepens, agricultural commodity markets across nations have become increasingly interconnected, leading to richer and more sophisticated research on their price transmission relationships. China’s cotton futures market was established relatively early, with predominant research focus directed toward either Sino-US cotton futures markets or comparisons of the price discovery functions between cotton futures and spot markets within these two countries. Comparatively less attention has been paid to investigating the spillover effects and information transmission efficiency specifically among the Chinese cotton futures market, the Chinese cotton yarn futures market, and the US cotton futures market. Meanwhile, the frequent occurrence of unforeseen events in recent years has intensified interest in understanding how such shocks alter spillover effects between markets. Despite the ongoing impacts of events like the US-China trade friction and the COVID-19 pandemic, research examining concurrent shifts in price spillover effects between the agricultural futures markets of China and the US in the wake of such events remains scarce.

Therefore, this study extends the existing literature by measuring the price spillover effects among cotton futures and cotton yarn futures markets in China and the United States. Further, the sample is segmented into distinct periods to conduct a comparative analysis of the evolving characteristics of mean and volatility spillover effects across these three markets—Chinese cotton futures, Chinese cotton yarn futures, and US cotton futures—under the influence of different event contexts. This investigation aims to elucidate how the cotton futures markets in the two nations respond to unforeseen shocks and to assess the role played by the relatively newly listed Chinese cotton yarn futures market within this system and its spillover dynamics, thereby providing pertinent insights and perspectives.

The subsequent sections of the paper are structured as follows: Section 2 delineates the methodological framework and variable selection employed in this study. Section 3 presents a comprehensive analysis of the empirical results. Section 4 concludes with key findings and discusses their theoretical and practical implications.

2. Methods and Data Description

2.1. Methods

2.1.1. Granger Causality Test

The Granger causality test determines whether a variable is influenced by lagged values of other variables. If such influence exists, Granger causality is inferred between the variables. The test specification is formalized as follows:

$$y_t={\displaystyle {\sum }_{i=1}^q{\alpha }_i{x}_{t-i}+{\displaystyle {\sum }_{j=1}^q{\beta }_j{y}_{t-j}}+u_{1t}}$$

(1)

$$x_t={\displaystyle {\sum }_{i=1}^s{\gamma }_i{x}_{t-i}+{\displaystyle {\sum }_{j=1}^s{\delta }_j{y}_{t-j}}}+u_{2t}$$

(2)

The null hypothesis for Equation (1) is denoted as $H_0:{\alpha }_1={\alpha }_2=\cdots ={\alpha }_q=0$.

The null hypothesis for Equation (2) is denoted as $H_0:{\delta }_1={\delta }_2=\cdots ={\delta }_s=0$.

If the lagged coefficient ${\alpha }_i$ for the variable $x_{t-i}$ in Equation (1) is statistically significant at the specified level, rejection of the null hypothesis $H_0$ implies that x Granger-causes y. If the lagged coefficient ${\delta }_j$ for the variable $y_{t-j}$ in Equation (2) is statistically significant at the specified level, rejection of the null hypothesis $H_0$ implies that y Granger-causes x.

2.1.2. Vector Autoregression (VAR) Model

The vector autoregression (VAR) framework models each endogenous variable as a function of lagged values of all endogenous variables within the system, enabling comprehensive analysis of cross-variable temporal influences. This approach incorporates both a variable’s historical effects on its current state and lagged impacts from other variables. For the price returns of three futures markets, the VAR model is formalized in Equation (3):

$$\left[\begin{array}{c}{R}_{1,t}\\ R_{2,t}\\ R_{3,t}\end{array}\right]=\left[\begin{array}{c}{\alpha }_1\\ {\alpha }_2\\ {\alpha }_3\end{array}\right]+{\displaystyle {\sum }_{k=1}^p\left[\begin{array}{ccc}{a}_{11,k}& a_{12,k}& a_{13,k}\\ a_{21,k}& a_{22,k}& a_{23,k}\\ a_{31,k}& a_{32,k}& a_{33,k}\end{array}\right]}\times \left[\begin{array}{c}{R}_{1,t-k}\\ R_{2,t-k}\\ R_{3,t-k}\end{array}\right]+\left[\begin{array}{c}{\epsilon }_{1,t}\\ {\epsilon }_{2,t}\\ {\epsilon }_{3,t}\end{array}\right]$$

(3)

$R_{1,t}$, $R_{2,t}$, $R_{3,t}$ represent the return series at time t for China’s cotton futures, China’s cotton yarn futures, and U.S. cotton futures, respectively. The variable p denotes the optimal lag order selected for the VAR model, and the vector α represents the constant terms. Matrix $A_k$ is a 3 × 3 coefficient matrix, where each diagonal element $a_{ii}$ reflects the autoregressive impact of a variable on itself, and each off-diagonal element $a_{ij}(i\ne j)$ captures the information transmission between different series. ${\epsilon }_{1,t}$, ${\epsilon }_{2,t}$, ${\epsilon }_{3,t}$ refer to the vectors composed of the residuals from the three mean equations, which are considered the sources of market volatility or risk. It is generally not meaningful to interpret the economic significance of each individual parameter estimate in the VAR model. Rather, the primary purpose of constructing the VAR model is to perform impulse response analysis and variance decomposition, in order to evaluate the dynamic impacts of one variable on another—both contemporaneously and with lags—and to quantify each variable’s contribution to the overall variation in the system.

2.1.3. Impulse Response Analysis

Impulse response analysis examines the effect of a shock to one endogenous variable on all other endogenous variables within the system. The impulse response function traces the reaction of each dependent variable to shocks in the explanatory variables over time. If the system is stable, the impact of the shock will gradually diminish and eventually converge to zero.

Taking a bivariate system as an example, suppose that at time t = 0, a shock occurs to $R_{1,t}$ through $R_0=\left[\begin{array}{c}{\epsilon }_{10}\\ {\epsilon }_{20}\end{array}\right]=\left[\begin{array}{c}1\\ 0\end{array}\right]$, and all subsequent error terms are equal to $\left[\begin{array}{c}0\\ 0\end{array}\right]$. Under these conditions, $R_1=\left[\begin{array}{cc}{b}_{11}& b_{12}\\ b_{21}& b_{22}\end{array}\right]\left[\begin{array}{c}1\\ 0\end{array}\right]=\left[\begin{array}{c}{b}_{11}\\ b_{21}\end{array}\right]$, $R_2=\left[\begin{array}{cc}{b}_{11}& b_{12}\\ b_{21}& b_{22}\end{array}\right]\left[\begin{array}{c}{b}_{11}\\ b_{21}\end{array}\right]=\left[\begin{array}{c}{b}_{11}^2+b_{12}{b}_{21}\\ b_{21}{b}_{11}+b_{22}{b}_{21}\end{array}\right]$. The term $b_{21}{b}_{11}+b_{22}{b}_{21}$ is referred to as the response of $R_{2,t}$ to a shock in $R_{1,t}$. Similarly, the impulse responses of other variables in the system can be derived accordingly.

2.1.4. Variance Decomposition

Variance decomposition utilizes market information to quantify the proportion of the variance in the model’s stochastic disturbance term that can be attributed to different sources of shocks. This technique allows us to assess the extent to which a specific piece of information contributes to the volatility of all endogenous variables in the system. By applying variance decomposition to the return series of the three markets, it is possible to evaluate the relative contribution of lagged shocks from each market to the current variability in their own and others’ returns. This analysis helps identify which market exerts a greater influence over time and which is more susceptible to price movements in other markets across different forecast horizons.

2.1.5. BEKK-GARCH Model

This study employs a trivariate BEKK-GARCH model to examine the volatility spillover effects among the U.S. and Chinese cotton and cotton yarn futures markets. A key advantage of this model lies in its ability to incorporate constraints that ensure the positive definiteness of the conditional variance-covariance matrix. Moreover, it allows for the measurement of the dynamic interdependence of volatilities across futures markets to a certain extent. The conditional variance equation of the BEKK-GARCH model is specified as follows:

$$H_t=C’C+A’({\epsilon }_{t-1}\epsilon {’}_{t-1})A+B’H_{t-1}B$$

(4)

Matrix $H_t$ denotes the conditional variance-covariance matrix at time t, matrix A represents the coefficient matrix for the ARCH terms, capturing the impact of past error terms on the current conditional variance. The ARCH coefficients reflect the short-term volatility clustering in the series and measure the effect of transitory shocks. Matrix B corresponds to the coefficient matrix for the GARCH terms, indicating the influence of past conditional variances on the current conditional variance. These coefficients characterize the persistence of volatility in the series and quantify the impact of long-term fluctuations. Matrix C is the constant parameter matrix. The specific forms of these matrices are defined as follows:

$$A=\left[\begin{array}{ccc}{a}_{11,t}& a_{12,t}& a_{13,t}\\ a_{21,t}& a_{22,t}& a_{23,t}\\ a_{31,t}& a_{32,t}& a_{33,t}\end{array}\right],B=\left[\begin{array}{ccc}{b}_{11,t}& b_{12,t}& b_{13,t}\\ b_{21,t}& b_{22,t}& b_{23,t}\\ b_{31,t}& b_{32,t}& b_{33,t}\end{array}\right],C=\left[\begin{array}{ccc}{c}_{11,t}& 0& 0\\ c_{21,t}& c_{22,t}& 0\\ c_{31,t}& c_{32,t}& c_{33,t}\end{array}\right]$$

$a_{ij}(i=j)$ represents the ARCH effect within the respective return series of China’s cotton futures, China’s cotton yarn futures, and U.S. cotton futures, $a_{ij}(i\ne j)$ captures the ARCH-type volatility spillover effect from one market to another among the three cotton-related futures markets. $b_{ij}(i=j)$ denotes the GARCH effect within each market, while $b_{ij}(i\ne j)$ indicates the GARCH-type volatility spillover between two different cotton futures markets.

In conjunction with the Wald test, we assess the presence or absence of statistically significant volatility spillover effects between each pair of the three markets—China’s cotton futures, China’s cotton yarn futures, and U.S. cotton futures—through a set of three hypotheses.

Hypothesis 1.

The null hypothesis is $a_{12}=b_{12}=0$, which implies that there is no unidirectional volatility spillover effect from China’s cotton futures to China’s cotton yarn futures. $a_{21}=b_{21}=0$, which implies that there is no unidirectional volatility spillover effect from China’s cotton yarn futures to China’s cotton futures. $a_{12}=b_{12}=a_{21}=b_{21}=0$, which implies that there is no bidirectional volatility spillover effect between China’s cotton futures and cotton yarn futures.

Hypothesis 2.

The null hypothesis is $a_{13}=b_{13}=0$, which implies that there is no unidirectional volatility spillover effect from China’s cotton futures to U.S. cotton futures. $a_{31}=b_{31}=0$, which implies that there is no unidirectional volatility spillover effect from U.S. cotton futures to China’s cotton futures. $a_{13}=b_{13}=a_{31}=b_{31}=0$, which implies that there is no bidirectional volatility spillover effect between China’s cotton futures and U.S. cotton futures.

Hypothesis 3.

The null hypothesis is $a_{23}=b_{23}=0$, which implies that there is no unidirectional volatility spillover effect from China’s cotton yarn futures to U.S. cotton futures. $a_{32}=b_{32}=0$, which implies that there is no unidirectional volatility spillover effect from U.S. cotton futures to China’s cotton yarn futures. $a_{23}=b_{23}=a_{32}=b_{32}=0$, which implies that there is no bidirectional volatility spillover effect between China’s cotton yarn futures and U.S. cotton futures.

2.2. Variable Selection

U.S. cotton futures were listed earlier than Chinese cotton yarn futures. Due to the availability of data, we selected daily trading data from 1 September 2017, to 31 March 2025. After removing unmatched samples caused by differences in financial holidays between China and the United States, a total of 1775 observations were retained. The U.S. cotton futures prices are represented by the closing prices of the lead contracts traded on the Intercontinental Exchange (ICE), while Chinese cotton futures prices are based on the closing prices of the lead contracts listed on the Zhengzhou Commodity Exchange. Similarly, Chinese cotton yarn futures prices are derived from the closing prices of the lead contracts on the Zhengzhou Commodity Exchange.

Due to differences in trading units between the U.S. and Chinese markets, price data were converted using the exchange rate of 1 U.S. cent per pound equals 22.0462 USD per metric ton. Subsequently, all prices were converted to USD per metric ton using the daily central parity exchange rate of the Chinese yuan against the U.S. dollar, as published by the State Administration of Foreign Exchange (SAFE). All futures price data were obtained from the WIND database. Finally, we computed the price returns for the three time series using the following formula:

$$R_t=100\times (\ln P_t-\ln P_{t-1})$$

(5)

3. Analysis of Research Results

3.1. Descriptive Analysis

We begin with a descriptive statistical analysis of the three time series, as presented in

Table 1. Descriptive statistics.

Variable	CF	CY	CT
Sample size	1774	1774	1774
Mean	−0.011737	−0.013902	−0.003986
Median	0.006400	−0.013733	0.011657
Maximum	7.274970	7.588547	8.476624
Minimum	−7.939304	−9.114855	−11.23252
Standard deviation	1.296746	1.225465	1.729732
Skewness	−0.210446	−0.210543	−0.329006
Kurtosis	8.210807	9.815430	5.844558

. Negative mean returns across all three futures markets indicate overall downward pressure during the sample period. The median daily return of Chinese cotton yarn futures remains persistently negative (−0.013733), confirming stable bearish momentum, whereas positive median returns for both Chinese (0.006400) and U.S. cotton futures (0.011657) indicate profitable outcomes on more than half of trading days. The significant divergence in median values—with the U.S. market exhibiting 81.8% greater median returns—reveals fundamentally stronger resilience in daily performance. In terms of the range of extreme values, the return volatility interval for Chinese cotton futures is [−7.939304, 7.274970], for Chinese cotton yarn futures is [−9.114855, 7.588547], and for U.S. cotton futures is [−11.23252, 8.476624]. Among them, U.S. cotton futures exhibit the widest range of return fluctuations. Regarding standard deviation, U.S. cotton futures have a standard deviation of 1.729732, which is higher than that of Chinese cotton futures (1.296746) and Chinese cotton yarn futures (1.225465), indicating that the U.S. cotton futures market demonstrates the greatest return volatility and, consequently, the highest market risk. In contrast, the return volatilities of Chinese cotton and cotton yarn futures are lower and relatively similar. In terms of kurtosis and skewness, all three return series exhibit kurtosis values greater than 3 and negative skewness, suggesting that the return distributions are left-skewed and leptokurtic, reflecting a characteristic of sharp peaks and fat tails.

3.2. Multiple Breakpoint Tests

This study employs the Bai-Perron L + 1 vs. L sequential structural breakpoint test under the Ordinary Least Squares (OLS) framework to detect breakpoints in the price series of Chinese cotton futures, Chinese cotton yarn futures, and U.S. cotton futures over the period from 1 September 2017 to 31 March 2025. The identified breakpoints are then analyzed in relation to corresponding real-world events, as shown in

Table 2. Bai-Perron L + 1 vs. L sequential structural breakpoint test results.

Break Test	F-Statistic	Scaled F-Statistic	Critical Value **
0 vs. 1 *	534.4376	1603.313	13.98
1 vs. 2 *	265.3075	795.9224	15.72
2 vs. 3 *	124.9469	374.8407	16.83
3 vs. 4 *	89.83734	269.5120	17.61
4 vs. 5	0.000000	0.000000	18.14
Sequential F-statistic determined breaks:			4
Significant F-statistic largest breaks:			4
Estimated break dates:
1: 2023/6/05
2: 2021/8/30, 2023/6/05
3: 2020/8/03, 2021/9/02, 2023/6/05
4: 2018/10/10, 2020/8/04, 2021/9/03, 2023/6/05
5: 2018/10/10, 2020/3/19, 2021/4/23, 2022/6/23, 2023/7/28

* Significant at the 0.05 level. ** Bai-Perron (Econometric Journal, 2003) [40] critical values.

. The test results reveal four structural breaks occurring on 10 October 2018, 4 August 2020, 3 September 2021, and 5 June 2023.

In relation to actual events, the China-U.S. trade conflict officially began on 22 March 2018, when then-President Donald Trump signed a memorandum imposing tariffs on Chinese imports. In response, China announced retaliatory measures on 23 March marking the formal outbreak of the trade war. During its initial stage, the first round of U.S. tariffs did not target cotton or textile-related products, so the direct impact on China’s cotton and cotton yarn markets was relatively limited. Consequently, the effect on price co-movements among the futures markets was also minimal.

However, as the trade war escalated, its influence on cotton-related markets became increasingly pronounced. In June 2018, China imposed retaliatory tariffs on key U.S. imports, including cotton, a strategic raw material in the textile supply chain. Given that U.S. cotton exports to China were relatively small at the time, the immediate market price impact remained modest. On 24 August 2018, the United States launched a new round of tariffs on Chinese imports, imposing a 25% levy on $16 billion worth of goods, including a substantial volume of cotton textile products. This marked a turning point at which the trade dispute began to significantly affect the price linkage across cotton markets.

The COVID-19 pandemic, which emerged in early 2020 and rapidly spread worldwide, had a profound impact on all sectors. By August 2020, China had adopted proactive containment measures and entered a phase of normalized epidemic control. In contrast, the United States faced considerable losses due to a delayed and passive response. As vaccines were developed and became widely available, the global impact of the pandemic began to recede by September 2021. On 6 May 2023, the World Health Organization officially declared that COVID-19 no longer constituted a “Public Health Emergency of International Concern”.

Based on the four identified structural breakpoints, the sample period is divided into five distinct phases for analysis:

• Phase I: Pre-trade war and pre-pandemic period
• Phase II: Period influenced by the China-U.S. trade conflict
• Phase III: Period dominated by the impact of the COVID-19 pandemic
• Phase IV: Period of trade tension de-escalation and normalized pandemic control
• Phase V: Period in which the pandemic’s impact had largely subsided

3.3. Stationary Tests

Since regression analysis involving non-stationary time series data may lead to spurious results, we first conduct the Augmented Dickey–Fuller (ADF) unit root test to assess the stationarity of the return series for the three markets across the five defined phases. As shown in

Table 3. Stationary test results.

Phase	Variable	T-Statistic	Prob.	Conclusion
I	CF	−16.4968	0.0000	Stationary
	CY	−16.3041	0.0000	Stationary
	CT	−16.5778	0.0000	Stationary
II	CF	−19.2500	0.0000	Stationary
	CY	−19.3399	0.0000	Stationary
	CT	−22.8987	0.0000	Stationary
III	CF	−20.4325	0.0000	Stationary
	CY	−18.5191	0.0000	Stationary
	CT	−20.1707	0.0000	Stationary
IV	CF	−19.9309	0.0000	Stationary
	CY	−19.6787	0.0000	Stationary
	CT	−18.3702	0.0000	Stationary
V	CF	−20.4701	0.0000	Stationary
	CY	−21.2364	0.0000	Stationary
	CT	−19.5606	0.0000	Stationary

, the ADF test results indicate that all return series are stationary at the 1% significance level in each of the five phases, thereby validating the suitability of these data for subsequent analysis.

3.4. Granger Causality Tests

Before conducting the Granger causality tests, the optimal lag lengths for each of the five phases were determined using information criteria, resulting in lag orders of 1, 1, 2, 1, and 1, respectively. Granger causality tests were then performed to assess the directional price relationships among the three markets across the five phases. The results are reported in

Table 4. Results of the Granger Causality Test.

Phase	Variable	Null Hypothesis	Chi-sq	Prob.	Conclusion
I	CF & CY	CF does not Granger-cause CY	2.1160	0.1458	Fail to reject
		CY does not Granger-cause CF	0.7694	0.3804	Fail to reject
	CF & CT	CF does not Granger-cause CT	0.0035	0.9526	Fail to reject
		CT does not Granger-cause CF	38.2892	0.0000	Reject
	CT & CY	CT does not Granger-cause CY	12.0231	0.0005	Reject
		CY does not Granger-cause CT	0.0245	0.8755	Fail to reject
II	CF & CY	CF does not Granger-cause CY	0.0898	0.7644	Fail to reject
		CY does not Granger-cause CF	1.7693	0.1835	Fail to reject
	CF & CT	CF does not Granger-cause CT	2.7161	0.0999	Reject
		CT does not Granger-cause CF	47.0128	0.0000	Reject
	CT & CY	CT does not Granger-cause CY	46.6076	0.0000	Reject
		CY does not Granger-cause CT	1.6515	0.1988	Fail to reject
III	CF & CY	CF does not Granger-cause CY	6.5466	0.0379	Reject
		CY does not Granger-cause CF	5.3653	0.0684	Reject
	CF & CT	CF does not Granger-cause CT	1.9543	0.3764	Fail to reject
		CT does not Granger-cause CF	63.6262	0.0000	Reject
	CT & CY	CT does not Granger-cause CY	34.0040	0.0000	Reject
		CY does not Granger-cause CT	2.2279	0.3283	Fail to reject
IV	CF & CY	CF does not Granger-cause CY	0.0000	0.9975	Fail to reject
		CY does not Granger-cause CF	0.0016	0.9685	Fail to reject
	CF & CT	CF does not Granger-cause CT	0.0597	0.8070	Fail to reject
		CT does not Granger-cause CF	53.5387	0.0000	Reject
	CT & CY	CT does not Granger-cause CY	44.3634	0.0000	Reject
		CY does not Granger-cause CT	0.5338	0.4650	Fail to reject
V	CF & CY	CF does not Granger-cause CY	0.5025	0.4784	Fail to reject
		CY does not Granger-cause CF	0.6159	0.4326	Fail to reject
	CF & CT	CF does not Granger-cause CT	0.0149	0.9029	Fail to reject
		CT does not Granger-cause CF	28.8731	0.0000	Reject
	CT & CY	CT does not Granger-cause CY	21.6715	0.0000	Reject
		CY does not Granger-cause CT	0.1747	0.6760	Fail to reject

.

In Phase I, there is no Granger causality between Chinese cotton futures and cotton yarn futures. This suggests that during the initial period following the listing of cotton yarn futures in China, market participation remained limited, liquidity was relatively low, and pricing linkages with cotton futures had yet to form. Moreover, neither Chinese cotton futures nor cotton yarn futures are Granger-caused by U.S. cotton futures; however, U.S. cotton futures are found to Granger-cause both Chinese cotton and cotton yarn futures. This unidirectional causality indicates that China’s cotton futures markets remain in a passive position in relation to U.S. cotton futures, lacking significant influence and leaving considerable room for development.

In Phase II, the previously unidirectional relationship in Phase I—where U.S. cotton futures led Chinese cotton futures—evolves into a bidirectional Granger causality. A plausible explanation lies in the effects of the China-U.S. trade conflict: following the imposition of tariffs on Chinese cotton by the U.S., China adopted countermeasures such as rotating its national cotton reserves and shifting imports toward Brazilian cotton, thereby diminishing the price-setting authority of U.S. cotton in global markets.

In Phase III, the prolonged trade conflict compounded by the impact of the COVID-19 pandemic strengthened the linkage between Chinese cotton and cotton yarn futures, resulting in a bidirectional Granger causality. This could be attributed, on the one hand, to frequent disruptions in logistics caused by the pandemic, which created cotton shortages and pushed up yarn production costs, thereby reinforcing the price transmission from cotton to yarn. On the other hand, mass cancelations of global orders prompted Chinese yarn producers to reduce raw material inventories, leading to inventory pressure in the yarn segment that was transmitted upstream to the cotton market through production cuts. Additionally, heightened market panic during the pandemic triggered a surge in risk-hedging demand, significantly boosting the liquidity of cotton yarn futures. As hedging became widespread throughout the supply chain, cotton yarn futures shifted from a marginal to a central pricing role within the industry. During this phase, U.S. cotton futures continued to Granger-cause both Chinese cotton and cotton yarn futures. However, the reverse causality from Chinese cotton futures to U.S. cotton futures disappeared. This asymmetry reflects the collapse of end-user demand in China’s textile sector due to the pandemic, which eroded China’s pricing power as the world’s largest cotton consumer. Simultaneously, disruptions in global supply chains prompted China to rely more heavily on domestic reserve cotton, further weakening the ability of domestic demand shocks to transmit to international markets, leaving Chinese futures markets subject to the unilateral influence of the financialized U.S. cotton market.

In Phases IV and V, as the effects of the pandemic and the trade conflict gradually subsided, the bidirectional Granger causality between Chinese cotton and cotton yarn futures once again disappeared. This may be attributable to the Chinese government’s temporary interventions via supply stabilization and price control policies, as well as market-based differentiation in risk premiums between the two commodity types. These factors further weakened the traditional pricing linkages along the cotton-textile industrial chain, temporarily nullifying mutual price discovery mechanisms. Additionally, liquidity in the cotton yarn futures market declined during these periods, diminishing its pricing functionality and impairing its role in effective price transmission. Notably, U.S. cotton futures consistently remained a Granger-causal driver of both Chinese cotton and cotton yarn futures throughout all phases, suggesting that despite the shocks of the COVID-19 pandemic and the China-U.S. trade conflict, China’s cotton-related futures markets have yet to establish significant global pricing influence. The U.S. continues to dominate price leadership in the international cotton market.

3.5. Mean Spillover Analysis Based on a Vector Autoregression (VAR) Model

A Vector Autoregression (VAR) model is constructed to examine the influence of both the lagged values of each variable and those of other related variables on future outcomes. This approach provides the foundation for subsequent analysis using impulse response functions and forecast error variance decomposition, which allows for an assessment of the magnitude, direction, and duration of dynamic responses to exogenous shocks within the system. The VAR model facilitates the quantification of inter-variable relationships and enables a structural interpretation of the dynamic interactions among variables, thereby allowing us to identify the direction and intensity of mean spillover effects across the three futures markets.

First, the optimal lag length p for the VAR model is determined using the Akaike Information Criterion (AIC) and the Schwarz Criterion (SC). For all five phases, the optimal lag order is found to be p = 1. Second, stability diagnostics are conducted for each VAR specification. Specifically, the inverse roots of the characteristic equations for all five models lie within the unit circle, indicating that the estimated parameters are stable and that the models possess valid predictive capabilities. This confirms the appropriateness of proceeding with further dynamic analysis. Finally, impulse response analysis and variance decomposition are carried out to investigate the transmission of shocks and the relative contributions of each variable to system fluctuations.

Figure 1, Figure 2, Figure 3, Figure 4 and Figure 5 illustrate the magnitude and duration of the responses among the three futures markets across five distinct phases under the impulse response analysis framework. The solid line represents the impulse response function of the response variable, the dashed lines denote the 95% confidence intervals. These figures capture how price shocks originating in one market affect the others over time.

As shown in Figure 1, during Phase I, the response of Chinese cotton futures to shocks from cotton yarn futures occurs between periods 1 and 3, peaking in period 2, though the overall impact remains relatively modest. In contrast, the response to shocks from U.S. cotton futures is both faster and more pronounced. Chinese cotton yarn futures exhibit a strong response to Chinese cotton futures, with the largest effect appearing in period 1. The response to U.S. cotton futures peaks in period 2. U.S. cotton futures display a relatively large response to Chinese cotton futures, but their response to Chinese cotton yarn futures remains weak throughout.

Overall, the results suggest a bidirectional influence between Chinese cotton and cotton yarn futures; however, the influence of cotton yarn futures is markedly weaker. Notably, only Chinese cotton futures exert a limited influence on U.S. cotton futures, while U.S. cotton futures exert significant influence on both Chinese cotton and cotton yarn futures—with a particularly strong impact on the Chinese cotton futures market. These findings are broadly consistent with real-world market dynamics.

As shown in Figure 2, during Phase II, the response of Chinese cotton futures to shocks from cotton yarn futures remains largely unchanged compared to the previous phase. However, its response to shocks from U.S. cotton futures becomes more pronounced, peaking in period 2. This increased sensitivity may be attributed to the stringent grading standards and advanced testing methods used for U.S. cotton, particularly its “no foreign fiber” quality, which continues to be highly favored by Chinese textile manufacturers. Consequently, despite the ongoing China-U.S. trade conflict, China remains heavily reliant on imports of U.S. cotton, thereby intensifying the influence of U.S. cotton prices on China’s cotton futures market.

Chinese cotton yarn futures also exhibit a significantly stronger response to both domestic and U.S. cotton futures, with the peak impact occurring in period 1. This suggests that as the cotton yarn futures market became more mature and the trade tensions escalated, the linkage between cotton yarn and cotton futures markets strengthened. However, the influence of Chinese cotton yarn futures on U.S. cotton futures remains negligible. This section is not mandatory but may be added if there are patents resulting from the work reported in this manuscript.

As illustrated in Figure 3, during Phase III, the response of Chinese cotton futures to cotton yarn futures strengthened compared to the previous phase, while the response of cotton yarn futures to cotton futures weakened. Notably, prior to the COVID-19 pandemic, U.S. cotton futures primarily followed the pricing signals of Chinese cotton futures. However, the outbreak disrupted global supply chains and prompted a shift in the U.S. market’s pricing anchor. On the one hand, China’s strict pandemic containment measures led to a decline in cotton import demand, thereby weakening the influence of Chinese cotton futures on U.S. cotton futures. On the other hand, international logistics bottlenecks and China’s “domestic circulation” policy reinforced the status of cotton yarn as a final consumer product. As textile manufacturers in Europe and the United States were forced to increase local procurement in response to unstable supplies of Chinese cotton yarn, U.S. cotton futures began to reflect domestic demand expectations through cotton yarn prices. Consequently, the responsiveness of U.S. cotton futures to shocks in the cotton yarn futures market increased during this period.

As shown in Figure 4, during Phase IV, the bidirectional response relationship between Chinese cotton and cotton yarn futures once again shifted compared to the previous phase. Both markets exhibited similar patterns in terms of the magnitude and duration of their responses to shocks from U.S. cotton futures, indicating that cotton yarn had become a key component in the evolving dynamics between Chinese and U.S. cotton markets. This development is partly attributable to the fact that, in addition to importing large volumes of raw cotton, China also imports cotton yarn from the international market to meet the needs of its textile industry. When the relationship between the Chinese and U.S. cotton markets reaches a relative equilibrium, attention naturally shifts toward the pricing of cotton yarn, making it a focal point for both enterprises and investors.

Meanwhile, the responsiveness of U.S. cotton futures to Chinese cotton futures further intensified, reaching a value of 0.6 in the first period. This trend reflects the diminishing impact of the pandemic and the shifting focus of the China-U.S. trade tensions. It also corresponds with a series of domestic initiatives launched by China, including improvements to infrastructure in major cotton-producing regions, the enhancement of quality control throughout the cotton supply chain, targeted financial support at the regional level, and the implementation of centralized government procurement and subsidy programs. These policy interventions have collectively strengthened the resilience and stability of China’s cotton industry, restored market confidence, and effectively elevated China’s pricing influence in the global cotton market.

In Phase V, compared to the previous stage, Chinese cotton futures began to exhibit a clear response to cotton yarn futures prices, where no such response had been observed before. A plausible explanation is that the full retreat of pandemic-related disruptions led to a substantive recovery in end-user demand, revitalizing the cotton yarn market. This recovery enabled yarn producers to resume inventory restocking and cost pass-through practices, thereby enhancing the market influence of cotton yarn futures. Improved market liquidity further strengthened the price signaling capacity of cotton yarn futures, contributing to their increased impact on cotton futures pricing.

Conversely, the responsiveness of Chinese cotton yarn futures to Chinese cotton futures prices disappeared during this phase. This shift likely reflects a fundamental transformation in the pricing dynamics of the cotton yarn market. With the rebound in downstream consumption and the restored ability of producers to manage inventories and transfer costs, the price of cotton yarn has become increasingly driven by its own supply–demand fundamentals rather than upstream cotton costs. As a result, cotton yarn futures began to operate under an independent pricing logic, with greater market autonomy.

Additionally, the responsiveness of U.S. cotton futures to Chinese cotton futures declined markedly. This trend may be attributed to the acceleration of offshore relocation within China’s textile industry since 2023, which led to a reduction in China’s cotton import volumes and, consequently, a diminished global pricing influence of Chinese cotton futures. Furthermore, the release of state reserve cotton has increasingly aligned Chinese cotton futures prices with domestic policy adjustments rather than global market fundamentals, reducing the sensitivity of U.S. cotton futures to pricing signals from China.

Variance decomposition analysis is employed to determine the contribution rates of different factors to the dynamics of Chinese cotton futures, Chinese cotton yarn futures, and U.S. cotton futures markets, thereby assessing the relative importance of each market in explaining forecast error variance.

Table 5. Variance decomposition results of the three futures markets in phase I.

Period	Variance Decomposition of CF			Variance Decomposition of CY			Variance Decomposition of CT
	CF	CY	CT	CF	CY	CT	CF	CY	CT
1	100.000	0.000	0.000	28.887	71.113	0.000	9.905	0.154	89.941
2	86.771	0.431	12.798	28.105	67.554	4.341	9.925	0.164	89.910
3	86.560	0.467	12.973	28.111	67.545	4.344	9.925	0.165	89.910
4	86.555	0.469	12.975	28.110	67.542	4.348	9.925	0.165	89.910
5	86.555	0.469	12.976	28.110	67.542	4.348	9.925	0.165	89.910
6	86.555	0.469	12.976	28.110	67.542	4.348	9.925	0.165	89.910
7	86.555	0.469	12.976	28.110	67.542	4.348	9.925	0.165	89.910
8	86.555	0.469	12.976	28.110	67.542	4.348	9.925	0.165	89.910

,

Table 6. Variance decomposition results of the three futures markets in phase II.

Period	Variance Decomposition of CF			Variance Decomposition of CY			Variance Decomposition of CT
	CF	CY	CT	CF	CY	CT	CF	CY	CT
1	100.000	0.000	0.000	66.631	33.369	0.000	11.530	1.582	86.888
2	89.649	1.046	9.305	60.683	29.992	9.325	11.348	2.135	86.517
3	89.419	1.172	9.408	60.532	29.971	9.498	11.337	2.182	86.480
4	89.409	1.180	9.411	60.524	29.973	9.504	11.337	2.185	86.478
5	89.408	1.181	9.411	60.523	29.973	9.504	11.337	2.185	86.478
6	89.408	1.181	9.411	60.523	29.973	9.504	11.337	2.185	86.478
7	89.408	1.181	9.411	60.523	29.973	9.504	11.337	2.185	86.478
8	89.408	1.181	9.411	60.523	29.973	9.504	11.337	2.185	86.478

,

Table 7. Variance decomposition results of the three futures markets in phase III.

Period	Variance Decomposition of CF			Variance Decomposition of CY			Variance Decomposition of CT
	CF	CY	CT	CF	CY	CT	CF	CY	CT
1	100.000	0.000	0.000	58.102	41.898	0.000	8.352	1.298	90.349
2	83.210	2.351	14.439	52.860	38.430	8.710	8.268	1.676	90.056
3	83.190	2.359	14.451	52.864	38.445	8.691	8.172	2.023	89.805
4	82.610	2.343	15.048	52.730	38.356	8.914	8.161	2.023	89.815
5	82.560	2.356	15.084	52.712	38.349	8.940	8.162	2.037	89.801
6	82.551	2.360	15.088	52.710	38.350	8.940	8.162	2.037	89.801
7	82.549	2.360	15.091	52.709	38.350	8.941	8.162	2.037	89.801
8	84.857	2.361	15.091	52.709	38.350	8.941	8.162	2.037	89.801

,

Table 8. Variance decomposition results of the three futures markets in Phase IV.

Period	Variance Decomposition of CF			Variance Decomposition of CY			Variance Decomposition of CT
	CF	CY	CT	CF	CY	CT	CF	CY	CT
1	100.000	0.000	0.000	56.808	42.777	0.414	7.150	0.000	92.850
2	88.129	0.000	11.871	51.197	38.668	10.135	7.275	0.131	92.594
3	88.113	0.018	11.869	51.192	38.676	10.132	7.276	0.131	92.594
4	88.109	0.018	11.873	51.190	38.675	10.135	7.276	0.131	92.594
5	88.109	0.018	11.873	51.190	38.675	10.135	7.276	0.131	92.594
6	88.109	0.018	11.873	51.190	38.675	10.135	7.276	0.131	92.594
7	88.109	0.018	11.873	51.190	38.675	10.135	7.276	0.131	92.594
8	88.109	0.018	11.873	51.190	38.675	10.135	7.276	0.131	92.594

and

Table 9. Variance decomposition results of the three futures markets in Phase V.

Period	Variance Decomposition of CF			Variance Decomposition of CY			Variance Decomposition of CT
	CF	CY	CT	CF	CY	CT	CF	CY	CT
1	100.000	0.000	0.000	57.022	42.978	0.000	9.896	0.079	90.025
2	93.906	0.088	6.006	54.372	41.048	4.580	10.099	0.113	89.788
3	93.906	0.088	6.006	54.373	41.043	4.583	10.098	0.113	89.789
4	93.905	0.088	6.006	54.373	41.042	4.584	10.098	0.113	89.789
5	93.905	0.088	6.006	54.373	41.042	4.584	10.098	0.113	89.789
6	93.905	0.088	6.006	54.373	41.042	4.584	10.098	0.113	89.789
7	93.905	0.088	6.006	54.373	41.042	4.584	10.098	0.113	89.789
8	93.905	0.088	6.006	54.373	41.042	4.584	10.098	0.113	89.789

present the variance decomposition results for the three time series across the five distinct phases, covering the first 8 forecast periods.

As shown in Table 5, during Phase I, the forecast error variance of Chinese cotton futures is entirely explained by its own innovations in the first period (100%). As the forecast horizon increases, the self-contribution rate gradually declines and stabilizes at 86.555% by period 4. Meanwhile, the contributions from Chinese cotton yarn futures and U.S. cotton futures increase over time, peaking at 0.469% and 12.976% in periods 4 and 5, respectively. For cotton yarn futures, only 28.887% of the forecast error variance in period 1 is explained by its own shocks, while a substantial 71.113% is attributed to Chinese cotton futures, indicating strong dependence. In contrast, U.S. cotton futures display a high level of self-explanatory power, with 89.941% of the variance in period 1 explained by its own shocks and 9.905% explained by Chinese cotton futures. The influence of cotton yarn futures on U.S. cotton futures is negligible. Overall, the variance decomposition results in Phase I suggest that the newly launched cotton yarn futures contract exerted limited influence within the broader cotton-related futures market, while it was already notably affected by U.S. cotton futures.

As shown in Table 6, during Phase II, the forecast error variance of Chinese cotton futures remained entirely explained by its own innovations in period 1, gradually declining to a stable contribution rate of 86.555% by period 4. The contribution from Chinese cotton yarn futures increased relative to Phase I, stabilizing at 1.181% by period 5, while the contribution from U.S. cotton futures decreased accordingly, stabilizing at 9.411% by period 4.

For Chinese cotton yarn futures, the proportion of variance explained by its own shocks was 33.369% in period 1, with Chinese cotton futures accounting for the remaining 66.631%. As the forecast horizon increased, the contributions from both sources declined, stabilizing at 60.523% (own) and 29.973% (from Chinese cotton futures), while the contribution from U.S. cotton futures rose from 0.000% in period 1 to 9.504%. In the case of U.S. cotton futures, the variance was largely explained by its own shocks—86.888% in period 1, stabilizing at 86.478%. The contributions from Chinese cotton futures and Chinese cotton yarn futures stabilized at 11.337% and 2.185%, respectively.

Taken together, these findings suggest that during Phase II, as the cotton yarn futures market matured, its influence within the cotton-related futures system in China increased. Following the escalation of China-U.S. trade tensions, China’s implementation of diversified cotton import strategies and other countermeasures effectively reduced the influence of U.S. cotton futures on the Chinese cotton futures market. Meanwhile, cotton yarn futures both exerted and received greater influence from U.S. cotton futures. This indicates that the introduction of cotton yarn futures in China has further strengthened the country’s ability to manage risks and enhance price discovery mechanisms within the cotton industry, thereby contributing to the internationalization and growing influence of China’s cotton-related futures markets.

As shown in Table 7, during Phase III, the forecast error variance of Chinese cotton futures in period 1 was still entirely explained by its own innovations (100%). However, as the forecast horizon extended, the contributions from Chinese cotton yarn futures and U.S. cotton futures stabilized at 2.360% and 15.091%, respectively. For Chinese cotton yarn futures, 41.898% of the forecast variance in period 1 was attributed to its own shocks, while 58.102% was explained by Chinese cotton futures. Over time, the contributions from the three markets stabilized at 52.709% (own), 38.350% (Chinese cotton futures), and 8.941% (U.S. cotton futures). Meanwhile, the contributions of Chinese cotton futures and cotton yarn futures to the variance of U.S. cotton futures stabilized at 8.162% and 2.037%, respectively, both showing a decline compared to Phase II.

These findings indicate that during this period, the pricing independence of China’s cotton yarn futures further increased. The outbreak of the COVID-19 pandemic contributed to a deeper fragmentation of the cotton industry chain between China and the United States, significantly weakening the pricing influence of Chinese cotton futures on U.S. cotton futures. Consequently, global cotton price discovery became increasingly centralized around the Intercontinental Exchange (ICE) in the United States.

As shown in Table 8, during Phase IV, the forecast error variance of Chinese cotton futures was still entirely explained by its own shocks in period 1. As the forecast horizon extended, its self-contribution stabilized at 88.109%, while the contributions from Chinese cotton yarn futures and U.S. cotton futures stabilized at 0.018% and 11.873%, respectively. Compared with the previous phase, the influence of U.S. cotton futures on Chinese cotton futures showed a slight decline. For Chinese cotton yarn futures, 42.777% of the variance in period 1 was explained by its own shocks and 56.808% by Chinese cotton futures. Over time, the contributions stabilized at 51.190% (own), 38.675% (Chinese cotton futures), and 10.135% (U.S. cotton futures). Regarding U.S. cotton futures, 92.850% of the variance in period 1 was explained by its own shocks, with 7.150% attributed to Chinese cotton futures and 0% to Chinese cotton yarn futures.

These results suggest that during Phase IV, the autonomy of both Chinese cotton and cotton yarn futures markets increased. Despite this, the primary source of variation in cotton yarn futures remained Chinese cotton futures. The influence of both Chinese markets on U.S. cotton futures continued to diminish, highlighting a further divergence in cross-market transmission effects.

As shown in Table 9, during Phase V, the forecast error variance of Chinese cotton futures in period 1 was still entirely explained by its own shocks. As the forecast horizon extended, its self-contribution stabilized at 93.905%, while contributions from Chinese cotton yarn futures and U.S. cotton futures stabilized at 0.088% and 6.006%, respectively. Compared to the previous phase, the influence of U.S. cotton futures on Chinese cotton futures further declined. For Chinese cotton yarn futures, 41.042% of the variance in period 1 was explained by its own shocks, while 57.022% was attributed to Chinese cotton futures. As the forecast period extended, contributions from the three markets stabilized at 54.373% (own), 41.042% (Chinese cotton futures), and 4.584% (U.S. cotton futures). Regarding U.S. cotton futures, the contribution from its own shocks was 90.025% in period 1, with 9.896% from Chinese cotton futures and only 0.079% from Chinese cotton yarn futures.

These findings suggest that in Phase V, the autonomy of both Chinese cotton and cotton yarn futures further improved. Although the influence of Chinese cotton futures on U.S. cotton futures increased slightly, it remained relatively limited overall, indicating that the global pricing power of China’s cotton-related futures markets had not yet undergone a fundamental shift.

Based on the variance decomposition results across the three futures markets over the five phases, the following conclusions can be drawn:

First, the contribution of U.S. cotton futures to Chinese cotton futures remained around 10% before and after the onset of the China-U.S. trade friction and the COVID-19 pandemic. Notably, following the escalation of trade tensions, the influence of U.S. cotton futures on Chinese cotton futures actually declined. This may be attributed to China’s proactive policy responses aimed at mitigating the impact of U.S. tariff barriers on key domestic industries. As a strategic commodity, cotton was prioritized in policy protections. By the time the U.S. imposed its second round of tariffs on cotton and related textile products, China had already implemented measures to cushion the impact—such as stabilizing and increasing domestic cotton production and diversifying imports from other major cotton-producing countries—to reduce exposure to price volatility in the global cotton market. With the outbreak of COVID-19, domestic cotton production in China experienced some disruption, which in turn negatively affected the downstream textile industry. Simultaneously, India—another major cotton producer—suffered from severe labor shortages due to the pandemic, reducing cotton cultivation. As a result, orders from Western textile companies shifted toward China, increasing China’s demand for raw cotton. Given the high level of mechanization in the U.S. cotton industry, its production was less affected by the pandemic, prompting China to continue importing large volumes of U.S. cotton to meet industrial needs. Consequently, the influence of U.S. cotton futures on Chinese cotton futures increased during this phase. However, as the pandemic receded and sanctions related to cotton and textiles eased, China’s domestic cotton output steadily recovered, market confidence rebounded, and the independence of Chinese cotton futures strengthened.

Second, since its launch, China’s cotton yarn futures market has been influenced by both domestic cotton futures and U.S. cotton futures. Initially, limited market maturity, investor hesitation, and insufficient hedging participation by relevant enterprises led to poor liquidity and a weak correlation with cotton futures prices. During this early stage, price fluctuations were primarily self-driven. As the market matured, with greater investor engagement and improved functionality in price discovery and risk hedging, the linkage between cotton yarn and domestic cotton futures strengthened significantly. Moreover, as the United States is China’s largest importer of textile products, U.S. cotton futures increasingly influence Chinese cotton yarn futures indirectly through their impact on Chinese cotton futures. The heightened market uncertainty brought on by the trade war and the pandemic further incentivized stakeholders to use cotton yarn futures for risk management. As a key downstream product in the cotton value chain, cotton yarn futures gradually exhibited more dependency on upstream cotton prices. However, with the dissipation of adverse market conditions and a cooling of speculative enthusiasm, trading volumes in the cotton yarn futures market declined sharply, contributing to a partial recovery of its pricing independence.

Third, U.S. cotton futures have consistently exhibited a high degree of autonomy. Particularly in the wake of the U.S.-China trade conflict and the COVID-19 pandemic, the interdependence between the Chinese and U.S. cotton markets weakened. The influence of Chinese cotton-related futures on U.S. cotton futures further diminished. Although some level of interconnection was reestablished after market stabilization, China’s cotton futures markets remained in a relatively passive position in the transmission of global cotton price shocks, continuing to lack significant pricing power.

3.6. Volatility Spillovers Between Chinese and U.S. Cotton and Cotton Yarn Futures Markets

Before establishing the GARCH model, it is essential to ensure the model’s appropriateness by testing for the presence of ARCH effects in the residuals. ARCH effects are used to evaluate volatility clustering in time series data—that is, the tendency for large changes in asset prices to be followed by large changes, and small changes to be followed by small changes, indicating temporal dependence in volatility.

Given that the autocorrelation function (ACF) and partial autocorrelation function (PACF) of the three futures return series exhibit trailing characteristics, ARMA(p, q) models are employed to fit the return series of the Chinese cotton futures, Chinese cotton yarn futures, and U.S. cotton futures markets. The optimal lag orders (p and q) for each ARMA model are determined based on autocorrelation patterns and information criteria. Subsequently, the residuals from each fitted model are subjected to the Lagrange Multiplier (LM) test for serial correlation and the ARCH test for heteroskedasticity. As shown in

Table 10. LM and ARCH test results.

	CF	CY	CT
LM Test
F-statistic	0.574	0.861	0.678
p-value	0.546	0.874	0.827
Presence of Autocorrelation in Residuals	No	No	No
ARCH Test
F-statistic	4.698	10.879	20.564
p-value	0.011	0.000	0.000
Presence of ARCH Effects	Yes	Yes	Yes

, the LM test results indicate that none of the three return series exhibit significant autocorrelation (p-values > 0.05), whereas all three series demonstrate significant ARCH effects (p-values < 0.05), confirming the presence of conditional heteroskedasticity. Therefore, it is appropriate to proceed with further analysis using the BEKK-GARCH model.

To further investigate the bilateral volatility spillover effects among the three futures markets, this study employs the BEKK-GARCH(1,1) model, estimated using the WinRATS 7.0 time-series regression software. The BFGS algorithm is selected for optimization, and Student’s t-distribution is applied to better capture the empirical distributional characteristics of the data. The estimation results of the BEKK-GARCH parameters are presented in

Table 11. Results of the BEKK-GARCH analysis.

Parameters	Phase I		Phase II		Phase III		Phase IV		Phase V
	Coeff	Signif	Coeff	Signif	Coeff	Signif	Coeff	Signif	Coeff	Signif
C(1,1)	0.188	0.001 ***	0.162	0.067 *	1.069	0.000 ***	0.626	0.129	−0.439	0.017 **
C(2,1)	0.119	0.009 ***	−0.179	0.009 ***	0.952	0.000 ***	0.530	0.067 *	−0.473	0.000 ***
C(2,2)	0.000	0.999	0.028	0.897	−0.007	0.944	0.062	0.382	−0.007	0.814
C(3,1)	−1.245	0.000 ***	−0.576	0.000 ***	−0.331	0.152	1.521	0.147	0.183	0.438
C(3,2)	0.000	0.999	0.328	0.013 **	−0.007	0.975	1.822	0.026 **	−0.617	0.048 **
C(3,3)	0.000	0.999	0.257	0.312	−0.001	0.995	0.108	0.947	1.082	0.000 ***
A(1,1)	0.311	0.000 ***	0.254	0.001 ***	0.527	0.009 ***	0.161	0.051 *	0.370	0.021 **
A(1,2)	−0.231	0.000 ***	−0.177	0.003 ***	−0.156	0.426	−0.193	0.014 **	−0.086	0.555
A(1,3)	−0.080	0.596	−0.045	0.731	−0.032	0.799	0.217	0.185	0.130	0.614
A(2,1)	−0.210	0.003 ***	−0.074	0.443	−0.485	0.011 **	0.054	0.630	0.102	0.448
A(2,2)	0.543	0.000 ***	0.538	0.000 ***	0.398	0.049 **	0.344	0.001 **	0.523	0.000 ***
A(2,3)	0.039	0.759	0.123	0.457	0.118	0.381	−0.333	0.086 *	−0.226	0.253
A(3,1)	−0.127	0.007 ***	0.093	0.052 *	−0.181	0.009 ***	0.252	0.000 ***	−0.251	0.000 ***
A(3,2)	0.013	0.648	0.073	0.074 *	−0.119	0.089 *	0.208	0.000 ***	−0.178	0.014 **
A(3,3)	0.374	0.000 ***	−0.081	0.217	0.229	0.000 ***	0.195	0.050 *	−0.023	0.857
B(1,1)	0.680	0.000 ***	0.917	0.000 ***	−0.192	0.428	0.838	0.000 ***	0.569	0.002 **
B(1,2)	−0.011	0.755	0.242	0.000 ***	−0.807	0.001 ***	−0.018	0.778	−0.246	0.153
B(1,3)	0.620	0.000 ***	−0.238	0.025 **	0.078	0.727	−0.347	0.102	0.332	0.291
B(2,1)	0.155	0.000 ***	0.017	0.821	0.375	0.166	−0.133	0.013 **	0.065	0.530
B(2,2)	0.887	0.000 ***	0.664	0.000 ***	0.712	0.005 ***	0.833	0.000 ***	0.892	0.000 ***
B(2,3)	−0.095	0.403	0.331	0.009 ***	−0.096	0.331	0.090	0.679	0.349	0.102
B(3,1)	0.276	0.000 ***	0.079	0.000 ***	0.404	0.000 ***	0.078	0.652	−0.389	0.000 ***
B(3,2)	0.096	0.003 ***	−0.037	0.236	0.394	0.000 ***	0.020	0.876	−0.208	0.016 **
B(3,3)	0.129	0.137	0.890	0.000 ***	0.942	0.000 ***	0.268	0.366	−0.414	0.005 **

*, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

. It is important to note that the magnitude of individual coefficients in the BEKK-GARCH model is not directly interpretable; thus, the Wald test is utilized to examine the statistical significance and directional implications of volatility spillovers among the futures markets. The Wald test results are summarized in

Table 12. Results of the Wald test.

Null Hypothesis	Phase I		Phase II		Phase III		Phase IV		Phase V
	Wald	Signif	Wald	Signif	Wald	Signif	Wald	Signif	Wald	Signif
A(1,2) = B(1,2) = 0	20.456	0.000	22.292	0.000	10.388	0.006	9.724	0.008	4.185	0.123
A(2,1) = B(2,1) = 0	11.510	0.003	0.853	0.653	7.399	0.025	7.677	0.022	1.578	0.454
A(1,2) = B(1,2) = A(2,1) = B(2,1) = 0	36.945	0.000	31.053	0.000	52.332	0.000	20.514	0.000	4.484	0.344
A(1,3) = B(1,3) = 0	14.677	0.001	8.767	0.012	0.150	0.928	3.808	0.149	1.577	0.455
A(3,1) = B(3,1) = 0	23.224	0.000	15.680	0.000	25.318	0.000	27.901	0.000	59.979	0.000
A(1,3) = B(1,3) = A(3,1) = B(3,1) = 0	26.969	0.000	20.809	0.000	29.056	0.000	30.614	0.000	67.305	0.000
A(2,3) = B(2,3) = 0	0.726	0.696	17.239	0.000	1.118	0.572	3.128	0.209	3.586	0.166
A(3,2) = B(3,2) = 0	5.890	0.053	6.515	0.038	11.130	0.004	26.826	0.000	15.649	0.000
A(2,3) = B(2,3) = A(3,2) = B(3,2) = 0	8.358	0.079	22.767	0.000	11.676	0.020	29.238	0.000	30.259	0.000

.

(1) In Phase I, at the 1% significance level, coefficients A(1,1), A(2,2), A(3,3), B(1,1), B(2,2), and B(3,3) are statistically significant, indicating that price volatilities in China’s cotton futures, cotton yarn futures, and U.S. cotton futures all exhibit pronounced ARCH and GARCH effects. In other words, both short-term shocks and long-term volatility persistence from previous periods significantly influence current price fluctuations in these markets.

Moreover, coefficients A(1,2), A(2,1), and B(2,1) are also significant at the 1% level, coefficient B(1,2) is statistically insignificant at the 10% significance level, implying that China’s cotton futures exhibit a negative ARCH-type volatility spillover to cotton yarn futures, whereas cotton yarn futures exert both a negative ARCH-type and a positive GARCH-type spillover to cotton futures. This suggests that new information shocks in either market reduce volatility in the other, while the historical volatility of cotton yarn futures tends to amplify fluctuations in cotton futures. Based on the Wald test, hypotheses 1, 2, and 3 are rejected at the 1% significance level, confirming the existence of bidirectional volatility spillovers between China’s cotton and cotton yarn futures. Furthermore, the magnitude of spillovers from cotton yarn to cotton is greater—likely due to the high volatility and risk associated with the newly launched cotton yarn futures contract during this early phase of market development.

In addition, coefficients A(3,1), B(1,3), and B(3,1) are statistically significant at the 1% level, coefficient A(1,3) is statistically insignificant at the 10% significance level, and Wald test hypotheses 4, 5, and 6 are rejected, indicating bidirectional volatility spillovers between China’s cotton futures and U.S. cotton futures. Notably, China’s cotton futures exert a positive GARCH-type spillover effect on U.S. cotton futures, and the spillover from China to the U.S. is stronger in magnitude.

In contrast, only B(3,2) is significant among the coefficients involving China’s cotton yarn futures and U.S. cotton futures, while A(2,3), A(3,2), and B(2,3) are not statistically significant. Additionally, Wald test hypotheses 7, 8, and 9 cannot be rejected, suggesting that there is no clear bidirectional volatility spillover between these two markets. A unidirectional GARCH-type spillover from U.S. cotton futures to China’s cotton yarn futures is observed, likely because the latter market was still underdeveloped and had not yet established a strong linkage with international cotton price dynamics during this phase.

(2) In Phase II, coefficients A(1,1), A(2,2), B(1,1), B(2,2), and B(3,3) are statistically significant at the 1% level, coefficient A(3,3) is statistically insignificant at the 10% significance level, indicating that China’s cotton and cotton yarn futures continue to exhibit strong ARCH and GARCH effects, while the volatility of U.S. cotton futures is driven primarily by GARCH effects.

Significance in A(1,2) and B(1,2), along with the rejection of Wald Hypothesis 1, and coefficient A(2,1), B(2,1) is statistically insignificant at the 10% significance level, suggesting that China’s cotton futures exert unidirectional ARCH- and GARCH-type volatility spillovers on cotton yarn futures. This indicates that, as cotton yarn futures matured, they became increasingly influenced by price volatility originating from upstream cotton markets.

The spillover relationship between Chinese and U.S. cotton futures remains directionally consistent with that of Phase I. However, in contrast to the previous phase, the volatility clustering effect from U.S. cotton futures to China’s cotton yarn futures becomes more evident, while its long-term persistence fades—reflecting a shift in the nature of cross-market influences as market conditions evolved.

(3) In Phase III, the coefficients A(1,1), A(3,3), B(2,2), and B(3,3) are statistically significant at the 1% level, while A(2,2) is significant at the 5% level. During this phase, the volatility persistence of China’s cotton futures is found to be insignificant, which may be attributed to the disruption caused by the COVID-19 pandemic. The volatility characteristics of China’s cotton futures became disordered, weakening the persistence of the impact from past information on current price movements.

Moreover, the significance of A(2,1) at the 5% level and B(1,2) at the 1% level, along with the rejection of hypotheses 1, 2, and 3 in the Wald test, suggest the presence of bidirectional volatility spillover effects between China’s cotton futures and cotton yarn futures during this phase.

The significance of A(3,1) and B(3,1) at the 1% level, whereas A(1,3) and B(1,3) show no statistical significance at the 10% level, coupled with the rejection of Hypothesis 2, indicates that the U.S. cotton futures market served as the primary transmitter of volatility to China’s cotton market in this period. As the pandemic spread globally, China achieved effective domestic containment while the U.S. cotton futures market experienced increasing instability, reinforcing the role of the U.S. as the main source of external risk spillovers.

(4) In Phase IV, A(1,1) and A(3,3) are significant at the 10% level, A(2,2) at the 5% level, and B(1,1) and B(2,2) at the 1% level. The coefficients A(1,2) and B(2,1) are also significant at the 5% level, and hypotheses 1, 2, and 3 are rejected based on the Wald test, whereas A(2,1) and B(1,2) show no statistical significance at the 10% level. These results indicate the presence of ARCH-type volatility spillovers from cotton futures to cotton yarn futures in China, implying that past volatility in cotton futures significantly influences the volatility of cotton yarn futures. Furthermore, a unidirectional GARCH-type volatility spillover from cotton yarn futures to cotton futures is observed, suggesting that new information in the cotton yarn market dampens short-term fluctuations in cotton futures.

This pattern may be explained by the post-pandemic recovery in China, during which overseas orders resumed and yarn mills rushed to secure raw material costs. The short-term volatility in cotton futures, driven by procurement sentiment, had a direct impact on cotton yarn futures, whose prices also directly reflect downstream textile demand. As the pandemic gradually subsided, clothing consumption began a slow and steady recovery, while the relatively low liquidity in the cotton yarn futures market contributed to more persistent volatility

Additionally, A(3,1) is significant at the 1% level, whereas A(1,3), B(1,3), B(3,1) show no statistical significance at the 10% level, and Hypothesis 2 is rejected in the Wald test, indicating that the U.S. cotton futures market exerts a unidirectional ARCH-type volatility spillover on China’s cotton futures. This may stem from China’s rigid dependence on cotton imports and herd behavior in the futures market, rendering China’s cotton futures a “passive recipient” of short-term volatility from the U.S. market. Moreover, China’s policy tools, due to implementation lags, can only mitigate medium- to long-term volatility (GARCH effects) and are less effective in buffering short-term ARCH shocks

The significance of A(2,3) and A(3,2) further reveals bidirectional ARCH-type volatility spillovers between U.S. cotton futures and China’s cotton yarn futures, suggesting that historical volatility in either market significantly influences the contemporaneous volatility in the other.

(5) In Phase V, A(1,1), B(1,1), and B(3,3) are significant at the 5% level, while A(2,2) and B(2,2) are significant at the 1% level. Based on the BEKK-GARCH model estimates and the Wald test results, there is no significant volatility spillover between China’s cotton futures and cotton yarn futures. With the waning impact of COVID-19 and the easing of China-U.S. trade tensions, China’s domestic cotton production became relatively stable. The policy of reserve cotton rotation increased market supply, and the import quota mechanism likely helped regulate the timing of cotton imports. On the demand side, yarn production capacity remained excessive, especially during periods of weak demand, and spinning mills operated at relatively low utilization rates.

The absence of strong shocks on either the supply or demand side reduced the sources and transmission channels of price volatility. However, A(3,1) and B(3,1) are significant at the 1% level, and A(3,2) and B(3,2) are significant at the 5% level, whereas A(1,3), B(1,3), A(2,3), B(2,3) show no statistical significance at the 10% level, while hypotheses 5 and 8 are rejected. These findings indicate that U.S. cotton futures exert unidirectional volatility spillovers on both China’s cotton and cotton yarn futures. This phenomenon may be explained by the global supply chain restructuring and weak domestic demand in China following the “dual downturn” of 2023. As a result, China’s cotton-related futures markets exhibited limited global price-setting power and became more susceptible to unidirectional risk spillovers from the U.S. cotton futures market.

4. Conclusions and Implications

This study investigates the impact of two major exogenous shocks—namely, the U.S.-China trade conflict and the COVID-19 pandemic—on the inter-market dynamics of cotton-related futures prices. Using daily settlement prices of the most actively traded contracts for Chinese cotton futures, Chinese cotton yarn futures, and U.S. cotton futures from 1 September 2017 to 31 March 2025, we employ Granger causality tests, a vector autoregressive (VAR) model, and the BEKK-GARCH model to empirically analyze changes in both mean and volatility spillover effects across five distinct stages before and after the onset of these events.

The empirical results reveal that the return spillovers among Chinese and U.S. cotton futures as well as cotton yarn futures exhibit temporal heterogeneity in response to external shocks. Key findings reveal that:

Initially, US cotton futures exerted a dominant influence over China’s cotton futures with strong internal linkages within China’s cotton-related derivatives prior to trade disputes. Subsequently, both trade friction and the pandemic significantly weakened price co-movements between Chinese and US markets while amplifying cross-market volatility risk transmission. Despite China’s marginally enhanced pricing power in cotton-related futures, US cotton futures maintained persistent dominance in price discovery throughout all phases. Notably, external shocks substantially disrupted the interdependence between Chinese cotton and yarn futures, with yarn futures ultimately demonstrating enhanced independence and influence in later stages.

Based on the above findings, this paper proposes the following recommendations:

First, it is essential to further leverage the fundamental functions of the futures market to enhance the risk resilience of the cotton industry chain. Stakeholders across the cotton supply chain should strengthen cooperation, improve risk management mechanisms, and establish a coordinated early warning system to collectively mitigate the impact of market price fluctuations. From cotton cultivation at the upstream end, to processing in the midstream, and down to the sales of cotton textiles at the downstream, including the supply of chemical products needed in the spinning industry, all enterprises along the chain should actively utilize financial instruments such as futures and insurance to manage risks at both the input and output ends. By engaging in hedging operations through the futures market, firms can lock in production costs and sales margins, thereby strengthening the entire industry’s capacity to withstand risks. Meanwhile, it is necessary to improve the operational efficiency of cotton-related futures markets and to enhance their role in price discovery. This would enable enterprises within the supply chain to better anticipate price trends and optimize production and marketing decisions, leading to more efficient resource utilization, improved operational performance, and greater market stability. In addition, enterprises should closely monitor price movements in both domestic and international spot and futures markets for cotton, maintain dynamic risk surveillance, and respond swiftly to extreme market fluctuations triggered by unexpected events. A comprehensive risk management mechanism across the cotton industry chain should be established.

Second, market participants’ risk awareness and management capabilities must be improved through targeted guidance. Futures exchanges, brokerage firms, and industry associations should provide professional support to cotton enterprises, helping them make informed use of internal resources for risk mitigation. Enterprises should make dynamic adjustments in response to changing market risks under different backgrounds and at different stages. For instance, the effectiveness of the futures market’s price discovery function, as well as the direction and intensity of international risk spillovers, may differ between the early and later stages of a shock. Cotton enterprises should flexibly employ hedging strategies tailored to specific market conditions in order to mitigate the impact of severe spot price volatility on their production and sales activities. At the same time, speculative behavior in the market should be properly regulated. Rational investor participation should be promoted to ensure market liquidity, while excessive speculation should be curbed to prevent undue volatility. Strict prohibitions against market manipulation and other forms of misconduct are also necessary to optimize the trading environment in cotton-related futures markets.

Third, it is recommended that a responsive monitoring and emergency mechanism be established to address the stress caused by futures price shocks during major unforeseen events. This entails timely response capabilities, as well as the development of comprehensive mechanisms for the prevention, detection, and control of risks in the futures market. Contingency plans for non-conventional scenarios and emergency measures for dealing with market disruptions should be pre-arranged. Guidance on using the futures market’s hedging function should be strengthened to reduce speculative behavior, which is particularly significant in preventing risk during major disruptions. Timely feedback and communication mechanisms should also be in place to guide relevant stakeholders in dynamically adjusting the scale and direction of their hedging strategies. This ensures that sufficient preparedness is in place, allowing stakeholders to respond promptly in the early stages of an unexpected event, thereby reducing the impact of futures price volatility and ensuring the overall stability of the industry.

Finally, to effectively mitigate transnational price volatility in cotton futures markets, we advocate for the establishment of synchronized Sino-U.S. institutional mechanisms. This necessitates implementing a bilateral joint cotton reserve release protocol during supply shocks, harmonizing margin requirements between Zhengzhou and ICEs to suppress speculative feedback loops, and developing a real-time shipping corridor monitoring system through BRI (Belt and Road Initiative)-USA partnership frameworks.

This study systematically investigates the interconnectedness of cotton and cotton yarn futures prices between China and the United States. It analyzes how the price transmission mechanisms have evolved before and after the COVID-19 pandemic and offers empirical insights that may serve as a reference in the event of similar disruptions. These findings contribute to understanding how market volatility and associated risks are transmitted internationally.

However, it should be noted that, in addition to China and the United States, countries such as India, Brazil, and Australia are also major cotton producers with significant influence on global cotton trade and supply chains. These countries have also established cotton futures markets. Future research could therefore expand the analysis to include the futures markets of these countries. Moreover, considering that other futures products, such as PTA and short-staple fiber, play critical roles in the cotton industrial chain, future studies may incorporate them into the analysis of price transmission and risk spillover mechanisms within the broader cotton-related futures complex. Finally, with the resurgence of China-U.S. trade tensions, further research is warranted to investigate how the price transmission dynamics between the Chinese and U.S. cotton markets evolve under this new wave of trade frictions, and how these dynamics compare to those observed during the previous phase of the trade conflict.