Is technological innovation a push for trade friction?

Abstract

This study analyses the interrelation between technological innovation (TI) and trade friction (TF) by applying the bootstrap rolling-window full- and subsample Granger causality test from China in a sample from January 2002 to December 2021. Results show that the influence of TI on TF is twofold. On the one hand, TI is a push for TF. This finding is consistent with the ‘income effect’, which postulates that TI leads to more TF by affecting the income of other countries. On the other hand, TI has a negative influence on TF. This result confirms the ‘substitute effect’, implying that TI can benefit consumers by providing more high-quality and cheaper products. In turn, TF can hinder TI by reducing exporters’ profits. Based on these findings, governments should coordinate their efforts toward innovation and trade policies. At the same time, firms should master core technology and develop their high-performance products to avoid the risk of TF.

Keywords:

• Technological innovation
• trade friction
• rolling-window
• bootstrap

JEL CODES:

• C32
• F13
• O32

1. Introduction

This paper investigates the interrelationships between technological innovation (TI) and trade friction (TF). TI and TF are essential international trade issues that draw wide attention (Akcigit & Melitz, 2022). TF reflects conflicts of trading partners. A country may take trade remedies measures (e.g., anti-dumping, countervailing, and safeguard measures) to protect a particular industry when its economy is threatened by other countries’ trade activities, leading to trade conflicts (Tian et al., 2016). As a primary productive force, TI has played a vital role in economic and trade growth (Das & Chatterjee, 2021; Kogan et al., 2017; Su et al., 2022a), which is deeply connected with TF. TI enables a firm to improve the competitiveness of exports and expand overseas markets, which may harm other countries’ interests, and thus triggers TF (Houser, 2020). For example, according to the U.S. Department of Commerce, China’s high-tech trade surplus with the U.S. increased from US$21.089 billion to US$114.163 billion from 2003 to 2016, which prompted the Trump Administration to impose several rounds of tariffs on Chinese goods from 2018 (Lawrence, 2018). This event reveals that TI may be a push for TF. This relationship between TI and TF can also be observed in other countries. In the 1970s, confronted with the rapid growth in the Japanese semiconductor industry under the ‘ultra-large-scale integration’ plan formulated by the government, the U.S. took several protective measures such as anti-dumping, anti-investment and anti-merger against the semiconductor industry (Langdon, 1983). Another example is the TF between South Korea and Japan. In 2019, in order to restrict the development of the Korean semiconductor industry, the Japanese government limited the export of three core materials for manufacturing semiconductors to South Korea, and the two countries removed each other from the export whitelist of trade facilitation. Therefore, TI may positively impact TF, which puts firms in exporting countries at risk. However, as TI continues to improve, a country may have cheaper and irreplaceable products for export (Das & Chatterjee, 2021), which increases global welfare and causes TF to decrease. Conversely, TF has a mixed impact on TI by reducing overseas markets for exporters. On the one hand, it may hinder TI by reducing funds for research and development (R&D). On the other hand, it can stimulate TI by increasing firms’ incentive to escape from the competition (Liebman & Reynolds, 2013). In general, TF is an important issue that draws excellent attention worldwide; whether TI is a push for TF is controversial. Thus, this paper explores the time-varying correlation between TI and TF to solve this issue. This research has important implications for the government to formulate strategic trade policies to avoid the risk of TF and provides a reference for the government to achieve long-term prosperity through TI when confronting TF.

As China is implementing a national innovation-driven development strategy, it is already among the world’s leading players in some key TI indices (Song et al., 2017). According to the World Intellectual Property Organization (WIPO), China ranked first regarding patent applications in 2021. Besides, in the annual ranking of the Global Innovation Index (GII) released by WIPO in 2022, China climbed to the 11th position and ranked first among 36 upper-middle-income economies. TI's improvement helps exports gradually change from labour-intensive to technology-intensive industries, causing high-tech products to become the main target in trade disputes (Jabbour et al., 2019). For instance, the production of Chinese photovoltaic cells ranked first in the world for four consecutive years from 2007 to 2010, leading the European Union (EU) to launch a massive amount of anti-dumping investigation (Voituriez & Wang, 2015; Zhi et al., 2014). In addition, with the development of TI in aerospace and communication, the trade surplus in the above fields with the U.S. continues to increase, which prompted the U.S. administration to take protective measures in 2018 (Su et al., 2022b). Thus, we can infer that TI is a push for TF. However, this conclusion does not always hold. In 2021, the trends of TI and TF showed opposite directions. The number of authorised patents continued to soar under the government’s efforts to encourage TI, while the number of TF against China dropped significantly. In general, TI and TF have a close relationship, but whether TI is a push for TF is still ambiguous. According to China’s Ministry of Commerce (MOC), the country was the most prominent target of trade remedies investigations from 2000–2020, and India, the U.S., Ukraine, Argentina, South Korea, and the EU initiated 68.85% of the analyses. This makes TF the main risk for Chinese exporters (Jabbour et al., 2019; Miyagiwa et al., 2016). It is noteworthy that the country is upgrading its industrial structure by encouraging TI and is building a robust domestic market and a strong trading nation, which may change the relationship between TI and TF. Thus, the discussion about TI and TF has special significance for China.

The contributions of this paper are as follows. Firstly, previous literature rarely examines the causal relationship between TI and TF; they mainly focus on the impact of TI on trade and related policy (Miyagiwa & Ohno, 2007; Rossi et al., 2021) or the impact of trade policy on TI (Dorn et al., 2020; Guei, 2022; Liu et al., 2021). To our knowledge, this paper is the first to explore the interconnection between TI and TF by considering the number of TF cases that Chinese exporters counteracted worldwide. We find that TI and TF influence each other significantly. Secondly, existing studies mainly apply the full-sample causality test, neglecting the parameters’ structural changes. This paper contributes to the current investigations by using the bootstrap sub-sample rolling-window causality test, which enables us to examine the time-varying relationship between TI and TF. We find that TI is a push for TF, which is consistent with the ‘income effect’ that TI of one country leads to TF by threatening other countries’ income. This view, however, was not valid in 2021, during which period TI reduced TF. Encouraged by the national innovation-driven development strategy, firms were more capable of developing their high-tech products in 2021, benefiting their trading partners by producing high-quality and cheaper products. As a result, TF decreases. This result proves the ‘substitute effect’, implying that TI reduces TF by increasing the real purchasing power of consumers. Understanding the correlation between TI and TF is significant for China to survive in TF while achieving innovation-driven development.

The rest of this paper is organised as follows. Section 2 is the literature review. Section 3 analyses the interrelationship between TI and TF theoretically. The empirical methods and the data are introduced in Sections 4 and 5, respectively. Section 6 shows the empirical results. Section 7 further discusses the results and compares them with previous studies. Section 8 concludes the paper and presents limitations and recommendations for future studies.

2. Literature review

A growing body of literature has recognised TI’s important role in TF. Samuelson (2004) points out that when one country’s export increases through technological progress, other countries’ gain from trade will fall, resulting in a trade conflict. Besides, Aggarwal (2004) reveals that innovative exporters are a threat to foreign competitors, which triggers the use of an anti-dumping policy. Niels (2000) also suggest that anti-dumping policies are more used in technological than non-technological industries. Miyagiwa and Ohno (2007) reveal that while cost-saving TI facilitates exports, it also increases the likelihood of being arbitrated as a dumping case. In addition, Márquez-ramos et al. (2010) use a gravity trade model and prove a positive effect of TI on exports, but they do not further analyse the impact on TF. Azar and Ciabuschi (2017) also demonstrate that innovation facilitates exporting activities by strengthening exporters’ competitive advantage. Likewise, Rossi et al. (2021) suggest that innovative activities increase the likelihood of starting exporting and lower the probability of exiting foreign markets, which implies that TI increases the competition in international trade. Moreover, Tian et al. (2016) further point out that TF results from fierce competition.

Unlike most research that TI increases export and leads to anti-dumping, Damijan et al. (2010) reveal that product or process innovations have no significant promoting effect on export, implying that TI has no impact on TF. Besides, Miyagiwa et al. (2016) argue that improving R&D capability in developing countries can help avoid anti-dumping wars with developed countries. In addition, Zhang et al. (2018) point out that innovation is risky, which may impose firms’ unbearable financial burdens and discourage exports. In addition, Edeh et al. (2020) conclude that process and marketing innovation promote export performance, but product innovation hinders exports in Nigeria. Furthermore, Dai et al. (2020) suggest an inverted-U relationship between innovation intensity and exporters’ survival probability, implying that innovation has a mixed impact on TF.

Conversely, TF has specific impacts on TI, but the conclusion is mixed. Avsar and Sevinc (2019) find that Turkish anti-dumping trade barriers promote R&D expenditures by reducing competition. In addition, Buryi and Lahiri (2019) imply that domestic firms tend to increase R&D investments when there is a rise in import tariffs. Slavtchev (2020) also approves that a protectionist policy is conducive to German R&D inputs by reducing import competition from middle- and low-income countries. In addition, Melitz and Redding (2021) suggest that protectionist policy increases the incentive for innovation of domestic firms. Moreover, Dorn et al. (2020) reveal that trade liberalisation impedes innovation in U.S. manufacturing firms, suggesting that TF can promote TI. Likewise, Aghion et al. (2021) propose a negative influence of trade liberalisation on the innovative activities of some French firms, implying that TF facilitates TI.

In contrast, Atkeson and Burstein (2010) develop a general equilibrium model and suggest that trade liberalisation rather than protectionist policy increases exporter’s innovation rates. Besides, Lileeva and Trefler (2010) propose that Canadian exporters generate more innovations after Canada and the U.S. sign the free trade agreement. Likewise, Burstein and Melitz (2011) prove that a more extensive market facilitated by tariff cuts promotes innovation. In addition, Akcigit et al. (2018) conclude that reduced trade barriers bring an influx of foreign competitors into domestic markets, which boosts domestic innovation through intensified international competition. Coelli et al. (2022) draw a similar conclusion that the increasing import competition from the reduction in global tariffs contributes to more patents. Moreover, Aghion et al. (2005) suggest that the impact of import competition on innovation when trade is liberalised follows an inverted U shape. Extremely low or high levels of competition result in less innovation. Shu and Steinwender (2019) further imply that when there is no TF, the influx of foreign competitors into domestic markets spurs innovation in developing countries, while the impact is mixed in developed countries.

More recent studies have investigated the relationship between TI and China-related TF. Li and Li (2022) propose that innovation triggers anti-dumping investigations against China primarily via the ‘perceived threat’ channel. Wang (2022) suggests that technological progress in Huawei, a Chinese company, increases competition and results in the trade conflict. Besides, Wu et al. (2021) indicate that innovation promotes Chinese manufacturing exports, but they do not analyse the further impact on TF. Likewise, using firm-level data, Dong et al. (2022) reveal that innovation benefits China’s exports. Conversely, China-related international TF also influences TI. Xie et al. (2020) show that anti-dumping barriers hinder Chinese firms’ R&D investment and R&D intensity. Liu and Ma (2020) explore the impact of trade policy uncertainty (TPU) on innovation, finding that a rise in TPU decreases Chinese firms’ patent applications. Houser (2020) points out further that the Sino-U.S. TF will hinder the worldwide development of TI. However, Prud’homme and Cohen (2019) point out that Sino-U.S. TF has stimulated a quest for technological ‘self-reliance’ in China. In addition, Xu et al. (2022) propose that the Sino–U.S. TF is conducive to Chinese firms’ innovation. Moreover, Li et al. (2022) show that the U.S. tariff escalation increases Chinese firms’ R&D expenditure.

Most previous researchers examine the impact of innovation on specific trade policies (e.g., anti-dumping) (Miyagiwa & Ohno, 2007) while ignoring the direct impact of TI on TF. Although some studies discuss the reason and impact of Sino-U.S. TF (Li et al., 2022; Xu et al., 2022), there still needs to be an answer to whether TI is a push for TF. In addition, although some studies (e.g., Houser, 2020; Xu et al., 2022) show that the effects of TF on innovation vary by region, they still need to discuss whether these effects change over time. Furthermore, related research does not use time-varying parameters in the models, neglecting the structural changes in the full-sample time series, which may lead to inaccurate results. This study considers the influence of China’s TI policies when investigating the relationship between TI and TF. We use the bootstrap subsample rolling-window causality test (Balcilar et al., 2010; Su et al., 2019) to explore the influence of TI on TF at different periods. This study not only answers the question of whether TI is always a push for TF but also has implications for the government regarding promoting innovation in the environment of TF.

3. TI and TF interaction mechanism

We use a two-country theoretical framework (Miyagiwa et al., 2016; Samuelson, 2004) to explain the impact of TI on TF. Consider there are two countries, A and B, and there is intra-industry trade. Suppose TI contributes to the cost-reducing production in country A, which drives down the exporting price (Wu et al., 2021). This will increase the market share of country A while diminishing that of country B (Miyagiwa & Ohno, 2007). In this way, TI can usually increase the income of exports and promote the economic growth of country A (Dong et al., 2022; Palley, 2006). However, the income of country B falls because of the diminishing market share after country A achieves technological progress, which may cause TF between the two countries (Samuelson, 2004). Let y_B be the export income of country B, and TI_A be technological innovation in country A, then TF=f(y_B), y_B =g(${\mathit{\text{TI}}}_A),$ which means TF is a function of y_B, and y_B is a function of ${\mathit{\text{TI}}}_A.$ From another perspective, as TI reduces the price of the product, residents in country B can buy more products worldwide at a lower price, which means the real purchasing power of B increases. In this way, TI reduces TF. Let R_B be the real purchasing power of country B, then R_B=h(${\mathit{\text{TI}}}_A$), TF=l(R_B), indicating that TI of country A influences R_B, and TF is a function of R_B. As TI influences TF by affecting y_B and R_B, the total impact of TI on TF can be shown as Equation (1)(1) $$\frac{\mathit{\text{dTF}}}{d{\mathit{\text{TI}}}_A}=\frac{\partial \mathit{\text{TF}}}{\partial y_B}\times \frac{d{y}_B}{d{\mathit{\text{TI}}}_A}+\frac{\partial \mathit{\text{TF}}}{\partial R_B}\times \frac{d{R}_B}{d{\mathit{\text{TI}}}_A}$$(1) : (1) $$\frac{\mathit{\text{dTF}}}{d{\mathit{\text{TI}}}_A}=\frac{\partial \mathit{\text{TF}}}{\partial y_B}\times \frac{d{y}_B}{d{\mathit{\text{TI}}}_A}+\frac{\partial \mathit{\text{TF}}}{\partial R_B}\times \frac{d{R}_B}{d{\mathit{\text{TI}}}_A}$$(1)

As mentioned above, the TI in country A harms the income of country B, then $\frac{\partial \mathit{\text{TF}}}{\partial y_B}$<0. In addition, the reduction of income in country B brings more TF, then $\frac{\mathit{\text{dTF}}}{d{y}_B}<0.$ Therefore, $\frac{\partial \mathit{\text{TF}}}{\partial y_B}\times \frac{d{y}_B}{d{\mathit{\text{TI}}}_A}$>0, which is defined as ‘income effect’, suggesting that the TI of country A brings more TF by reducing the income of country B. Furthermore, as TI in country A raises the real purchasing power of country B, which causes less TF, we can infer that $\frac{d{R}_B}{d{\mathit{\text{TI}}}_A}$>0, and $\frac{\partial \mathit{\text{TF}}}{\partial R_B}$<0. Then $\frac{\partial \mathit{\text{TF}}}{\partial R_B}\times \frac{d{R}_B}{d{\mathit{\text{TI}}}_A}$<0, indicating that TI reduces TF by increasing the real purchasing power of country B. We define this effect as the ‘substitute effect’. In summary, TI has both an income effect and an substitute effect on TF. When the income effect outweighs the substitute effect, TI is a push for TF, and vice versa.

In turn, TF can also influence TI. On the one hand, TF impedes the TI of the exporter for the following two reasons. First, it is generally believed that a more significant market stimulates innovation because fixed costs can be spread over more units of products (Geng & Kali, 2021). When TF occurs, a country can negatively affect other countries’ market share by taking protective measures (Crowley et al., 2018), thus reducing firms’ inventiveness to innovate in other countries. For instance, when country B takes safeguard measures against subsidised imports or dumped imports from country A, the overseas markets of A shrink, suppressing the innovation motives of firms in country A. Second, TF may reduce the exporters’ profit, which negatively impacts the R&D expenditure that TI needs (Melitz & Redding, 2021). For example, the levying of anti-dumping and countervailing duty from country B increases the price of country A's products in country B, which weakens the competitiveness of country A's products and leads to a reduction in profit. Lack of financing is one of the biggest obstacles that innovative firms face (Krastanova, 2014; Wu et al., 2021), so the reduced profit impedes TI. Let TI=l(TF), which means TI is a function of TF. We can infer that $\frac{\mathit{\text{dTI}}}{\mathit{\text{dTF}}}<0,$ suggesting that TF hinders TI of the exporting country. On the other hand, TF promotes the TI of exporters by increasing the competitive pressure. Exporters may be incentivised to escape from the competition by developing products with higher performance through TI (Aghion et al., 2005; Galdon-Sanchez & Schmitz, 2002; Xu et al., 2022). In this situation, $\frac{\mathit{\text{dTI}}}{\mathit{\text{dTF}}}>0.$ In summary, TI and TF are interactive, but the exact direction of the impact is uncertain.

4. Methodology

4.1. Bootstrap full-sample causality test

The standard Granger causality test based on the vector autoregression (VAR) model usually assumes that statistics like Likelihood Ratio (LR) or Lagrange multiplier (LM) obey the standard asymptotic distribution in full samples (Sun et al., 2021). However, such an assumption may not hold because of structural changes in the time series (Toda and Phillips, 1993, 1994), which can lead to inaccurate estimates. Shukur and Mantalos (1997) suggest that the critical values of residual-based bootstrap (RB) estimation can improve estimation performance. Furthermore, Shukur and Mantalos (2000) prove that RB-based corrected LR- statistics exhibit relatively better power and size properties even in small samples, which can increase the robustness of the Granger test.

Therefore, RB-based modified-LR statistic is applied to explore the causal relationship between TI and TF. The VAR model is shown in Equation (2)(2) $${\mathbf{y}}_t={\mathbf{\Phi }}_0+{\mathbf{\Phi }}_1{\mathbf{y}}_{t-1}+\dots +{\mathbf{\Phi }}_p{\mathbf{y}}_{t-p}+{\mathbf{\epsilon }}_t,t=1,2\dots T$$(2) : (2) $${\mathbf{y}}_t={\mathbf{\Phi }}_0+{\mathbf{\Phi }}_1{\mathbf{y}}_{t-1}+\dots +{\mathbf{\Phi }}_p{\mathbf{y}}_{t-p}+{\mathbf{\epsilon }}_t,t=1,2\dots T$$(2) where ${\mathbf{y}}_t$ is a column vector of variables, ${\mathbf{\epsilon }}_t$ is the white-noise vector, T is the number of samples, ${\mathbf{\Phi }}_0,$ … ${\mathbf{\Phi }}_p$ are matrixes of coefficients to be estimated and p is the lag length. In this paper, we choose the lag length based on the Akaike information criterion (AIC), Schwarz information criterion (SIC) and Hannan-Quinn information criterion (HQ). In addition, as mentioned in the theoretical analysis that export is related to TI and TF (Dong et al., 2022), we use exports (EX) as the control variable in the VAR model (Jabbour et al., 2019; Xu et al., 2022). Thus, Equation (2)(2) $${\mathbf{y}}_t={\mathbf{\Phi }}_0+{\mathbf{\Phi }}_1{\mathbf{y}}_{t-1}+\dots +{\mathbf{\Phi }}_p{\mathbf{y}}_{t-p}+{\mathbf{\epsilon }}_t,t=1,2\dots T$$(2) can be expressed as follows: (3) $$\begin{array}{c}\left[\begin{array}{c}T{I}_t\\ T{F}_t\end{array}\right]=\left[\begin{array}{c}{\varphi }_{10}\\ {\varphi }_{20}\end{array}\right]+\left[\begin{array}{c}{\varphi }_{11}{}^{(1)}\\ {\varphi }_{21}{}^{(1)}\end{array}\begin{array}{c}\begin{array}{c}\begin{array}{cc}& \begin{array}{c}{\varphi }_{12}{}^{(1)}\\ {\varphi }_{22}{}^{(1)}\end{array}\begin{array}{cc}& \begin{array}{c}{\varphi }_{13}{}^{(1)}\\ {\varphi }_{23}{}^{(1)}\end{array}\end{array}\end{array}\end{array}\end{array}\right]\left[\begin{array}{c}T{I}_{t-1}\\ \begin{array}{c}T{F}_{t-1}\\ E{X}_{t-1}\end{array}\end{array}\right]+\dots +\left[\begin{array}{c}{\varphi }_{11}{}^{(2)}\\ {\varphi }_{21}{}^{(2)}\end{array}\begin{array}{c}\begin{array}{c}\begin{array}{cc}& \begin{array}{c}{\varphi }_{12}{}^{(2)}\\ {\varphi }_{22}{}^{(2)}\end{array}\begin{array}{cc}& \begin{array}{c}{\varphi }_{13}{}^{(2)}\\ {\varphi }_{23}{}^{(2)}\end{array}\end{array}\end{array}\end{array}\end{array}\right]\left[\begin{array}{c}T{I}_{t-2}\\ \begin{array}{c}T{F}_{t-2}\\ E{X}_{t-2}\end{array}\end{array}\right]+\\ \dots +\left[\begin{array}{c}{\varphi }_{11}{}^{(p)}\\ {\varphi }_{21}{}^{(p)}\end{array}\begin{array}{c}\begin{array}{c}\begin{array}{cc}& \begin{array}{c}{\varphi }_{12}{}^{(p)}\\ {\varphi }_{22}{}^{(p)}\end{array}\begin{array}{cc}& \begin{array}{c}{\varphi }_{13}{}^{(p)}\\ {\varphi }_{23}{}^{(p)}\end{array}\end{array}\end{array}\end{array}\end{array}\right]\left[\begin{array}{c}T{I}_{t-p}\\ \begin{array}{c}T{F}_{t-p}\\ E{X}_{t-p}\end{array}\end{array}\right]+\left[\begin{array}{c}{\epsilon }_{1t}\\ {\epsilon }_{2t}\end{array}\right].\end{array}$$(3)

In Equation (3)(3) $$\begin{array}{c}\left[\begin{array}{c}T{I}_t\\ T{F}_t\end{array}\right]=\left[\begin{array}{c}{\varphi }_{10}\\ {\varphi }_{20}\end{array}\right]+\left[\begin{array}{c}{\varphi }_{11}{}^{(1)}\\ {\varphi }_{21}{}^{(1)}\end{array}\begin{array}{c}\begin{array}{c}\begin{array}{cc}& \begin{array}{c}{\varphi }_{12}{}^{(1)}\\ {\varphi }_{22}{}^{(1)}\end{array}\begin{array}{cc}& \begin{array}{c}{\varphi }_{13}{}^{(1)}\\ {\varphi }_{23}{}^{(1)}\end{array}\end{array}\end{array}\end{array}\end{array}\right]\left[\begin{array}{c}T{I}_{t-1}\\ \begin{array}{c}T{F}_{t-1}\\ E{X}_{t-1}\end{array}\end{array}\right]+\dots +\left[\begin{array}{c}{\varphi }_{11}{}^{(2)}\\ {\varphi }_{21}{}^{(2)}\end{array}\begin{array}{c}\begin{array}{c}\begin{array}{cc}& \begin{array}{c}{\varphi }_{12}{}^{(2)}\\ {\varphi }_{22}{}^{(2)}\end{array}\begin{array}{cc}& \begin{array}{c}{\varphi }_{13}{}^{(2)}\\ {\varphi }_{23}{}^{(2)}\end{array}\end{array}\end{array}\end{array}\end{array}\right]\left[\begin{array}{c}T{I}_{t-2}\\ \begin{array}{c}T{F}_{t-2}\\ E{X}_{t-2}\end{array}\end{array}\right]+\\ \dots +\left[\begin{array}{c}{\varphi }_{11}{}^{(p)}\\ {\varphi }_{21}{}^{(p)}\end{array}\begin{array}{c}\begin{array}{c}\begin{array}{cc}& \begin{array}{c}{\varphi }_{12}{}^{(p)}\\ {\varphi }_{22}{}^{(p)}\end{array}\begin{array}{cc}& \begin{array}{c}{\varphi }_{13}{}^{(p)}\\ {\varphi }_{23}{}^{(p)}\end{array}\end{array}\end{array}\end{array}\end{array}\right]\left[\begin{array}{c}T{I}_{t-p}\\ \begin{array}{c}T{F}_{t-p}\\ E{X}_{t-p}\end{array}\end{array}\right]+\left[\begin{array}{c}{\epsilon }_{1t}\\ {\epsilon }_{2t}\end{array}\right].\end{array}$$(3) , when ${\varphi }_{21}{}^{(q)}=0,$ (q = 1, 2, …, p), TI is not a Granger cause of TF. Likewise, when ${\varphi }_{12}{}^{(q)}=0,$ TF is not a Granger cause of TI can be tested. In this study. RB-based modified-LR-statistics and p-values are used to test full-sample causality. If the null hypothesis that ${\varphi }_{21}{}^{(q)}=0$ is rejected, then TI is the Granger cause of TF. Also, TF is the Granger cause of TI if the null hypothesis that ${\varphi }_{12}{}^{(q)}=0$ is rejected.

4.2. Parameter stability test

We assume the VAR model parameters to be constant over time in the full-sample causality test, which means only one causality can be obtained in every period (Su et al., 2020). However, the relationship between dependent and independent variables may undergo structural changes brought about by demand or supply shocks in the economy or may result from institutional shifts. In this situation, it is highly possible that the parameters will not be constant, which leads to an unreliable result of the full-sample Granger causality test (Balcilar and Ozdemir, 2013). Therefore, it is necessary to test the stability of parameters. This study uses the Sup-F, Mean-F, and Exp-F tests proposed by Andrews (1993) and Andrews and Ploberger (1994) to check the stability of the parameters. In addition, we apply the Lc test proposed by Nyblom (1989) and Hansen (1992) to test the long-term parameter stability. These tests can be used to check the stability of the parameters to determine whether structural changes exist at unknown time points.

4.3. Rolling-window subsample causality test

When structural mutations exist in the full-sample, although devices such as dividing the samples or using dummy variables can be employed to solve this problem, biases still exist, which affect the results of the Granger causality test. Thus, this study uses the bootstrap sub-sample rolling-window granger causality test (Balcilar et al., 2010), which not only allows the causality between variables to change over time but also enables us to observe the difference caused by structural changes in different subsamples and avoids biases (Qin et al., 2021; Su et al., 2021). This method divides the whole sample into fixed-size subsamples for causality testing. Suppose the full-sample length is T, and each subsample includes L observations; then the subsamples are τ − L + 1, τ − L + 2, …, τ, where τ = L, L + 1, … T. In this way, we can obtain T − L + 1 subsamples. When deciding the size of subsamples, L, there is no uniform standard (Balcilar et al., 2010). On the one hand, small subsamples can reduce the impact of potential heteroscedasticity, but the estimated variance will be more considerable, and therefore, the result is not effective. On the other hand, large subsamples can improve the estimation’s validity, but heteroscedasticity may lead to an unreliable result. It is usually believed that bias-minimizing window size should not be less than 20 observations (Pesaran & Timmermann, 2005).

We can then investigate the Granger causal relationship between TI and TF in each subsample by applying the RB-based modified LR causality test. The significance of the causality between TI and TF can be observed by calculating the p-value of the LR statistic. The impact of TI on TF can be obtained using the formula $N_b^{-1}{\sum }_{q=1}^p{\widehat{\varphi }}_{21}{}^{(q)},$ where N_b is the frequency of bootstrap iterations, and ${\widehat{\varphi }}_{21}{}^{(q)}$ is the bootstrap estimator in the VAR model. Similarly, $N_b^{-1}{\sum }_{q=1}^p{\widehat{\varphi }}_{12}{}^{(q)}$ shows the impact of TF on TI. The confidence interval is 90%, with the lower limit equal to the fifth quantiles of ${\widehat{\varphi }}_{12}{}^{{}^{(q)}}$ and the upper limit equal to the 95th quantiles of ${\widehat{\varphi }}_{21}{}^{{}^{(q)}}$ (Balcilar et al., 2010).

5. Data

This study uses monthly data from January 2002 to December 2021. China’s accession to the World Trade Organization (WTO) in December 2001 promoted its exports to proliferate. The larger overseas market gives firms a stronger motivation for TI (Geng & Kali, 2021). Besides, China has adopted a national strategy of innovation-driven development to improve firms’ long-term competitiveness and upgrade industrial structure. Hence, China’s granted patents have increased dramatically. In this paper, we use the number of granted patents each month to measure TI, which is widely considered a good indicator of innovation (Su et al., 2022a). In addition, with the growth in exports and technological advances, many countries impose remedial measures such as anti-dumping, countervailing and safeguards against China to protect their own industries (Jabbour et al., 2019). According to China’s MOC, 27% of global trade remedies investigations are against China, making it the world’s biggest target in trade remedies cases since 1995. Confronted with this challenging situation, Chinese exporters respond to the investigations actively to safeguard their interests. This paper uses the number of trade remedies cases that Chinese exporters counteract to measure TF (Tian et al., 2016), released by MOC.¹ We can infer that TI and TF may be correlated, with TI seemingly being a push for TF in most periods. Furthermore, the relationship between TI and TF is connected with changes in China’s exports (Dong et al., 2022). As China’s exports are enormous and continue to grow, its trading partners’ market share may decrease, thus leading to TF (Samuelson, 2004). This paper uses China’s export value to measure EX (Jabbour et al., 2019; Xu et al., 2022), which is drawn from the CEIC Data.²

Figure 1 shows TI and TF trends. The solid line indicates the changes in TF, while the dashed line describes the changes in TI. It can be observed that TI has an overall rising trend. It reached to peak in 2021. Besides, a high TI coincides with an increase in TF in most periods. In addition, as TI rises, the trade remedies investigations against China are also shifting from labour-intensive (e.g., textiles) to technology-intensive products. For example, when the growth of TI accelerated significantly in 2012, TF also increased rapidly. In 2012, the U.S. and EU implemented several rounds of anti-dumping investigations against Chinese photovoltaic cells, which substantially negatively impacted the photovoltaic industry (Voituriez & Wang, 2015; Zhi et al., 2014). Similar changes can be observed after the release of the Outline of the National Strategy of Innovation-Driven Development (ONSID) in May 2016. Encouraged by ONSID, Chinese firms allocate more resources to innovation, causing TI to rise. The figure shows that TF also ascends rapidly in the corresponding period. In particular, TF peaked in 2018 when Sino-U.S. trade frictions began. Therefore, we can infer that TI is a push for TF.

Figure 1. Trends of TI and TF.

Note: TI: technological innovation; TF: trade friction; CPC: Communist Party of China

Source: The authors.

However, the trends of TI and TF are sometimes different. Affected by the Sino-U.S. trade frictions, the Chinese government emphasises technological ‘self-reliance’ (Prud’homme & Cohen, 2019), which causes the relationship between TI and TF to change. According to the Outline of the 14th Five-Year Plan (2021–2025) for National Economic and Social Development, China will make more efforts to achieve innovation-driven development. Besides, the Chinese government is building a robust domestic market and a strong trading nation. These events increase firms’ innovation inventiveness, driving TI to soar in 2021. However, TF drops sharply during this period. Moreover, China’s exports affect the relationship between TI and TF. On the one hand, export is closely connected with TF. When China’s export is vast, and it keeps a rising trade surplus with its trading partners, the overseas market of other countries may shrink, leading to TF. On the other hand, rising export influences TI by increasing exporters’ profits. If firms attach importance to TI, they will invest more export profit in R&D activities, leading to more TI. However, if firms can make huge profits only by exporting large quantities of labour-intensive products, they may lack incentives for TI. As export is deeply connected with TI and TF, we choose export (EX) as the control variable (Jabbour et al., 2019; Xu et al., 2022). In summary, TI and TF have a time-varying relationship, which is also connected with exports.

Table 1 shows the descriptive statistics of the variables. The means of TI indicates that there are 19833.58 granted patents on average each month. The average value of TF suggests that Chinese exporters counteract 7.738 trade remedies cases on average monthly. The average weight of EX is US$144523.8 million. All three variables have positive skewness, which means they follow the right-skewed distribution. In addition, the kurtoses of the TI and TF are more significant than 3, demonstrating a leptokurtic distribution, whereas that of EX is less than 3, indicating a platykurtic distribution. Moreover, the Jarque–Bera test shows that TI and TF obey a nonnormal distribution at the significance of 1%. And EX follows a nonnormal distribution at the significance of 10%. Hence, the estimation of parameters is inaccurate when we use the traditional Granger causality test. The subsequent analysis takes all variables from the natural logarithms to avoid potential heteroscedasticity. Besides, EX is further transformed by taking the first difference to avoid non-stationary.

Table 1. Descriptive statistics of the sequence of TI, TF and EX.

Statistics	TI	TF	EX
Observations	240	240	240
Mean	19833.58	7.738	144523.8
Median	16001.5	7	154916.5
Maximum	68857	23	340498.8
Minimum	1386	0	19137
Std. Dev.	16619.15	3.954	71213.82
Skewness	0.918	0.681	0.030
Kurtosis	3.002	3.570	2.253
Jarque–Bera	33.744***	21.779***	5.616*

Notes: *** indicates that the statistics are significant at the 1% level. The unit of EX is US$million.

TI: technological innovation; TF: trade friction; EX: exports.

Source: The authors.

6. Empirical results

Before performing the full-sample Granger causality test, this study performs ADF (Dickey & Fuller, 1981) and PP (Phillips & Perron, 1988) unit root tests to ensure the stability of the sequences. Table 2 shows that all sequences are stationary.

Table 2. Unit root tests.

Series	ADF	PP
TI	−4.327 (3)***	−10.998[9]***
TF	−14.952 (0)***	−15.261[7]***
EX	−3.051(13)**	−32.072[12]***

Notes: Numbers in parentheses indicate the lag order, which is selected based on the AIC.

Numbers in the brackets refers to the bandwidth, which uses the Bartlett Kernel as suggested by the Newey–West test (1987).

*** and **denote significance at the 1% and 5% levels, respectively.

TI: technological innovation; TF: trade friction; EX: exports;.

ADF: Augmented Dickey-Fuller test; PP: Phillips and Perron test.

Source: The authors.

Then, we conduct the Granger full-sample causality test by constructing VAR models shown in Equation (3)(3) $$\begin{array}{c}\left[\begin{array}{c}T{I}_t\\ T{F}_t\end{array}\right]=\left[\begin{array}{c}{\varphi }_{10}\\ {\varphi }_{20}\end{array}\right]+\left[\begin{array}{c}{\varphi }_{11}{}^{(1)}\\ {\varphi }_{21}{}^{(1)}\end{array}\begin{array}{c}\begin{array}{c}\begin{array}{cc}& \begin{array}{c}{\varphi }_{12}{}^{(1)}\\ {\varphi }_{22}{}^{(1)}\end{array}\begin{array}{cc}& \begin{array}{c}{\varphi }_{13}{}^{(1)}\\ {\varphi }_{23}{}^{(1)}\end{array}\end{array}\end{array}\end{array}\end{array}\right]\left[\begin{array}{c}T{I}_{t-1}\\ \begin{array}{c}T{F}_{t-1}\\ E{X}_{t-1}\end{array}\end{array}\right]+\dots +\left[\begin{array}{c}{\varphi }_{11}{}^{(2)}\\ {\varphi }_{21}{}^{(2)}\end{array}\begin{array}{c}\begin{array}{c}\begin{array}{cc}& \begin{array}{c}{\varphi }_{12}{}^{(2)}\\ {\varphi }_{22}{}^{(2)}\end{array}\begin{array}{cc}& \begin{array}{c}{\varphi }_{13}{}^{(2)}\\ {\varphi }_{23}{}^{(2)}\end{array}\end{array}\end{array}\end{array}\end{array}\right]\left[\begin{array}{c}T{I}_{t-2}\\ \begin{array}{c}T{F}_{t-2}\\ E{X}_{t-2}\end{array}\end{array}\right]+\\ \dots +\left[\begin{array}{c}{\varphi }_{11}{}^{(p)}\\ {\varphi }_{21}{}^{(p)}\end{array}\begin{array}{c}\begin{array}{c}\begin{array}{cc}& \begin{array}{c}{\varphi }_{12}{}^{(p)}\\ {\varphi }_{22}{}^{(p)}\end{array}\begin{array}{cc}& \begin{array}{c}{\varphi }_{13}{}^{(p)}\\ {\varphi }_{23}{}^{(p)}\end{array}\end{array}\end{array}\end{array}\end{array}\right]\left[\begin{array}{c}T{I}_{t-p}\\ \begin{array}{c}T{F}_{t-p}\\ E{X}_{t-p}\end{array}\end{array}\right]+\left[\begin{array}{c}{\epsilon }_{1t}\\ {\epsilon }_{2t}\end{array}\right].\end{array}$$(3) . According to AIC, SIC and HQ, the optimal lag length is 4. The full-sample causality results are presented in Table 3. The null hypothesis that TI is not a Granger cause of TF and the null hypothesis that TF is not a Granger cause of TI cannot be rejected at the significance level of 10%. Results show that TI and TF have no relationship with each other, which is inconsistent with the theoretical analysis.

Table 3. Full-sample Granger causality tests.

	H0: TI is not a Granger cause of TF		H0: TF is not a Granger cause of TI
Tests	Statistics	p-values	Statistics	p-values
Bootstrap LR test	2.325	0.660	1.342	0.844

Notes: The null hypothesis is that no causal relationship exists between the variables. p-values are calculated using 10,000 bootstrap repetitions.

TI: technological innovation; TF: trade friction; LR: likelihood ratio.

Source: The authors.

The traditional full-sample Granger causality test requires that all parameters are constant, and we can get only a single Granger causal relationship within a fixed time interval. However, when structural changes exist in the parameters, the causality of TI and TF may change over time. In this situation, the results of the traditional full-sample Granger causality test may deviate from the actual situation (Zeileis et al., 2005). Hence, the stability test is performed to determine the presence of structural mutations. As mentioned above, this study uses Sup-F, Mean-F, Exp-F, and Lc tests to test the stability of parameters in the VAR models. The results are shown in Table 4.

Table 4. Parameter stability tests.

	TI equation		TF equation		VAR (4) system
	Statistics	Bootstrap p-value	Statistics	Bootstrap p-value	Statistics	Bootstrap p-value
Sup-F	37.669***	0.000	41.211***	0.000	36.266	0.109
Mean-F	17.251***	0.009	9.373	0.390	20.476	0.243
Exp-F	14.102***	0.001	15.488***	0.000	13.796	0.160
Lc					3.420	0.220

Notes: We calculate p-values using 10,000 bootstrap repetitions. ∗∗∗indicate significance at the 1% levels. Lc shows the results of the Hansen–Nyblom parameter stability test for all parameters in the VAR jointly.

TI: technological innovation; TF: trade friction; VAR: vector autoregression.

Source: The authors.

The results of the Sup-F tests indicate a sudden shift in the TI equation, the TF equation is at the 1% significance level. Mean-F and Exp-F tests are used to test the null hypothesis that parameters follow a martingale process. The results show that the null hypothesis is rejected, indicating that the TI equation and TF equation may evolve gradually with time. In summary, the results above show that parameters are unstable, and there are structural changes in the whole sample. Thus, the full sample Granger causality test results need to be more accurate. To improve the accuracy of the results, we adopt the bootstrap rolling-window Granger causality test to investigate the time-varying causal link between TI and TF in different subsamples. As the bias-minimizing window size should not be less than 20 observations (Pesaran & Timmermann, 2005), the rolling subsample data includes 24³ months of observations to ensure the reliability of the test.

Figure 2 shows the rolling bootstrap of the p-values of the LR-statistics using TF as the dependent variable. Results show that the p-values are less than 0.1 during 2012:M6–2012:M8, 2018:M2-2018:M3, 2021:M7-2021:M12, indicating that the null hypothesis—TI is not the Granger cause of TF—is rejected significantly at the 10% level. Figure 3 reports the sum of the rolling-window coefficients for the impact of TI on TF. Combining Figures 2 and 3, we can see that TI influences TF positively in 2012:M6–2012:M8 and 2018:M2-2018:M3, suggesting that China’s TI is a push during this period for TF. However, this influence becomes negative in 2021:M7-2021:M12, meaning that TI reduces TF.

Figure 2. Bootstrap p-value of the statistics (the null hypothesis is that TI is not a Granger cause of TF).

Source: The authors.

Figure 3. Sum of rolling-window coefficients of TI’s influence on TF.

Source: The authors.

China has implemented a series of programmatic policies to promote TI. For instance, in 2006, the Chinese government promulgated the ‘Outline of the National Medium- and Long-Term Science and Technology Development Plan (2006–2020)’, aiming to rank top five in the world by 2020 in terms of the number of national granted patents. In 2008, the ‘Outline of National Intellectual Property Strategy’ was issued, which lists the better protection of intellectual property as one of the strategic goals. Encouraged by such policies of China, the TI improves rapidly, promoting the growth of exports of the country. In 2012, several countries conducted trade remedies investigations against China (Qin et al., 2022; Su et al., 2022c), which involved not only labour-intensive (e.g., textiles) and capital-intensive products (e.g., steel and tires) but also technology-intensive products (e.g., photovoltaic). Moreover, 70% of the trade remedies investigations are initiated by developing countries or emerging economies, such as Brazil, India, Argentina and Thailand. The main reason that TI increases TF during this period is twofold. On the one hand, TI facilitates Chinese exports, which reduces its competitors’ market share and income. The conflicting interests between the two countries lead to TF. On the other hand, many countries have not recovered from the subprime mortgage crisis and the European sovereign debt crisis (Feng et al., 2021; Su et al., 2022d). According to WTO, the global commodity trade grew by only 0.2% in 2012. Such an economic downturn encourages many countries to safeguard against Chinese exports (Baldwin & Evenett, 2009). Therefore, TI has a positive influence on TF from 2012:M6–2012:M8.

From 2012–2018, China’s TI accelerates under the national strategy of innovation-driven development (Song et al., 2017; Zhang, 2020). In 2018, for example, Huawei was the company that filed for most patents with the World Intellectual Property Organization (Houser, 2020). With the progress of TI, China has encountered multiple trade remedies investigations initiated by India, Mexico, Brazil, Costa Rica, Turkey, the U.S., the EU, and so on from 2018:M2 to 2018:M3. Among them, the investigation initiated by the U.S. has a relatively more considerable influence. The U.S. administration decided to impose a 25% tariff on more than 1,300 kinds of Chinese imports involving aviation, aerospace, information, and communication technology in 2018. Confronted with such a situation, China also imposes tariffs on U.S. exports (Lawrence, 2018; Li et al., 2020). It is believed that China’s rise to technological prominence affects U.S. interests, which leads to Sino-U.S. TF (Li & Li, 2022). Therefore, TI is a push for TF during 2018:M2-2018:M3. This result is consistent with the ‘income effect’, implying that the TI of China brings more TF by threatening the export income of other countries.

However, there was a negative impact of TI on TF in 2021. Guided by the outline of the 14th Five-Year Plan (2021–2025), China has been improving the system for TI, encouraging R&D investment, and accelerating the development of TI. Besides, the Sino-U.S. TF has raised Chinese companies’ awareness of developing high-tech products such as semiconductors. Furthermore, the Chinese government makes more efforts to build a robust domestic market and expand overseas markets under the Belt and Road Initiative, which enhances the incentives for Chinese firms’ TI. It can be observed that while TI increased, TF dropped sharply in 2021. The reasons are as follows. First, developing Chinese firms’ high-tech products enables them to circumvent technical trade barriers, thus decreasing TF. Second, TI boosts China’s economy and benefits its trading partners by producing high-quality and cheaper products (Dong et al., 2022). Hence, TI reduces TF during 2021:M7-2021:M12. This finding proves the ‘substitute effect’, suggesting that China’s TI reduces TF by increasing the real purchasing power of other countries.

Figure 4 presents the rolling bootstrap p-values of the LR statistic using TI as the dependent variable. The p-values are less than 0.1 during the periods of 2005:M2–2005:M7, showing that TF is the Granger cause of TI during this period. Figure 5 depicts the sum of the rolling-window coefficients of TF’s influence on TI. Specifically, TF exerts a negative effect on TI, which means that more TF hinders TI in China during that period.

Figure 4. Bootstrap p-value of the statistics (the null hypothesis is that TF is not a Granger cause of TI).

Source: The authors.

Figure 5. Sum of rolling-window coefficients of TF’s influence on TI.

Source: The authors.

In 2005, China’s exports continued to increase, reaching approximately $1.4 trillion. At the same time, China is the country that receives most trade remedies investigations globally (Jabbour et al., 2019). According to China’s MOC, its exporters respond to 43 trade remedies cases from 2005:M2 to 2005:7, involving various labour-intensive products such as textiles, daily necessities, and building materials. Among them, textiles are the most significant target for anti-dumping investigations. It is worth noting that all import quotas for textiles and clothing among WTO members should be abolished from January 1, 2005. As China has a comparative advantage in the textile industry, this event further benefits textiles export (Zhuo and Park, 2019). As the export of textiles threatens other countries’ interests, several economies, including the EU, the U.S. and India, decided to take various protectionist measures to restrict imports from China. In response to the protectionist measures and to protect the textile industry, the Chinese government makes many efforts to negotiate with its trading partners. Such TF hinders the development of TI. First, TF reduces the resources for TI. China was an export-oriented economy in 2005, whose income depended mainly on exporting labour-intensive products. As TF reduces exporters’ profits, it limits the R&D expenditure further and ultimately impedes TI. Second, the overseas export markets shrink when there is TF, which weakens exporters’ incentive for TI. Hence, TF reduces TI in 2005:M2–2005:M7.

However, under the national innovation-driven development, the industrial structural upgrade gradually. Chinese firms have realised the important role of TI in enterprise development and invested more resources in R&D (Dong et al., 2022; Yi et al., 2013). An increasing number of Chinese enterprises have developed their high-tech products, whose performance has reached or surpassed that of imported products. For example, although the Sino-U.S. trade friction has caused a shrinking overseas market and a plummeted profit for Huawei, the company put more emphasises on TI. (Xu et al., 2022). In addition, the Chinese government has been building a robust domestic market and a strong trading nation, which gives firms a greater incentive to innovate (Wu et al., 2021). Since TF has less impact on the resources and incentives of TI in Chinese firms, there is no significant influence of TF on TI from 2005:M7 to 2021:M12.

7. Discussion

In the empirical analysis, we apply the bootstrap sub-sample rolling-window Granger causality test to investigate the relationship between TI and TF in different subsamples and draw three conclusions. Firstly, we find that the influence of TI on TF is positive in 2012:M6–2012:M8 and 2018:M2-2018:M3, indicating that during these periods, TI is a push for TF. In 2012, TI facilitated Chinese exports, which led to conflicting interests between China and its trading partners. Besides, many countries have not recovered from the subprime mortgage crisis and European sovereign debt Crisis (Feng et al., 2021), leading to a positive influence from TI to TF in 2012:M6–2012:M8. In 2018, with the continuous progress of TI, China encountered multiple trade remedies investigations initiated by India, Mexico, Brazil, Costa Rica, Turkey, the U.S., the EU, and so on. Among them, the research undertaken by the U.S. has a relatively more considerable influence. It is believed that China’s rise to technological prominence affects U.S. interests, which leads to Sino-U.S. TF (Li & Li, 2022). Therefore, TI is a push for TF during 2018:M2-2018:M3. This result is consistent with the ‘income effect’, indicating that the TI of China brings more TF by threatening the export income of other countries. This finding supports other studies conducted in this area (Aggarwal, 2004; Li & Li, 2022; Miyagiwa & Ohno, 2007), but they mainly focus on the impact of TI on anti-dumping. For example, Miyagiwa and Ohno (2007) reveal that while cost-saving TI facilitates exports, it also increases the likelihood of being arbitrated as a dumping case. Li and Li (2022) propose that innovation triggers anti-dumping investigations against China primarily via the ‘perceived threat’ channel. Unlike these studies, this paper uses the number of trade remedies cases (anti-dumping, countervailing and safeguards against China) that Chinese exporters counteract to measure TF, drawing a more comprehensive conclusion that China’s TI is a push for TF.

Secondly, TI reduces TF in 2021:M7-2021:M12. During this period, developing Chinese firms’ high-tech products enabled them to circumvent technical trade barriers. Besides, TI boosts China’s economy and benefits its trading partners by producing high-quality and cheaper products. Hence, TI reduces TF during 2021:M7-2021:M12. This finding proves the ‘substitute effect’, suggesting that China’s TI reduces TF by increasing the real purchasing power of other countries. This result differs from most studies that China’s TI brings more TF (Samuelson, 2004; Wang, 2022). For example, Wang (2022) suggests that technological progress in Huawei, a Chinese company, increases competition and results in the trade conflict. In contrast, this paper proves that TI can reduce TF by promoting producing high-quality and cheaper products and achieving mutual benefit between trading partners. This finding has important implications for the government and firms to engage in TI.

Thirdly, it shows that TF influences TI negatively during the periods of 2005:M2–2005:M7, indicating that TF may hinder a country’s TI when its economic growth mainly depends on exports. From 2005:M2 to 2005:7, Chinese exporters respond to 43 trade remedies cases; among them, textiles are the most significant target for anti-dumping investigations. China was an export-oriented economy in 2005, whose income depended mainly on exporting labour-intensive products (e.g., textiles). As TF reduces exporters’ profits, it limits the R&D expenditure further and ultimately impedes TI (Melitz & Redding, 2021). In addition, the overseas export markets shrink when there is TF (Crowley et al., 2018), which weakens exporters’ incentive for TI. Hence, TF reduces TI in 2005:M2–2005:M7. This result agrees with related studies (Houser, 2020; Liu & Ma, 2020; Xie et al., 2020), but they fail to investigate the impact of all of the TFs that China encounters. For example, Xie et al. (2020) show that anti-dumping barriers hinder Chinese firms’ R&D investment and R&D intensity. However, they fail to study the influence of other TFs, such as countervailing and safeguards against China. Liu and Ma (2020) focus on the impact of trade policy uncertainty (TPU) on innovation, finding that a rise in TPU decreases Chinese firms’ patent applications, which does not discuss TF directly. Houser (2020) points out that the Sino-U.S. TF will hinder the worldwide development of TI, which does not aim to discuss the TF between China and other countries. Furthermore, we find that TF influences TI negatively during the periods of 2005:M2–2005:M7, while the influence is insignificant in different periods. This result differs from some studies (Li et al., 2022; Prud’homme & Cohen, 2019; Xu et al., 2022) that the Sino–U.S. TF is conducive to Chinese firms’ innovation. This finding implies that in order to promote TI, the government and firms should take measures to avoid TF. In addition, the Chinese government should build a robust domestic market to reduce the negative impact of TF on TI in Chinese firms.

8. Conclusion, implication and limitation

8.1. Conclusion

TF is an important issue that draws great attention worldwide. In particular, TF has become the main risk for Chinese exporters, deeply connected with TI. In this context, this paper aims to analyse whether TI is always a push for TF. Firstly, we conduct the full-sample Granger causality test. Results show that TI and TF have no relationship with each other, which is inconsistent with existing literature (Avsar and Sevinc, 2019; Miyagiwa et al., 2016; Samuelson, 2004). Considering that the full-sample Granger causality test assumes that there is only a single Granger causality in the whole sample, it is inaccurate in estimating the relationship between TI and TF. Then, we apply the parameter stability test, results show that the parameters are unstable and structural changes exist. Therefore, we apply the bootstrap sub-sample rolling-window Granger causality test to examine the time-varying causal relationship between TI and TF in different subsamples, drawing three main conclusions.

First, in 2005 and 2018, China’s TI led to more TF primarily by threatening trading partners’ overseas market share and exporting income. This result supports the ‘income effect’ in the theoretical analysis, implying that China’s TI can lead to more TF by reducing the income of the other country. This finding also supports other related literature (Aggarwal, 2004; Li & Li, 2022; Miyagiwa & Ohno, 2007), which reveals that TI leads to more anti-dumping investigations.

Second, TI had a negative effect on TF in 2021, during which time the development of Chinese firms’ high-tech products enabled them to circumvent technical trade barriers. Besides, TI promotes producing high-quality and cheaper products, which helps achieve the mutual benefit between China and its trading partners during this period. This result is consistent with the ‘substitute effect’ in the theoretical analysis, which suggests that TI can reduce TF by providing cheaper products, thereby increasing the real purchasing power of consumers. This result differs from most studies that China’s TI brings more TF (Samuelson, 2004; Wang, 2022).

Third, a negative influence from TF on TI is observed, suggesting that TF may hinder TI by reducing firms’ profit for innovation. In addition, TF causes the overseas export markets to shrink, which weakens exporters’ incentive for TI. This result agrees with related studies (Houser, 2020; Liu & Ma, 2020; Xie et al., 2020).

8.2. Suggestion and implication

Understanding the relationship between TI and TF is significant for China to survive in TF while achieving innovation-driven development. It provides the following insights for policymakers. First, as China’s TI may lead to TF and thus risks trade, policymakers should take some measures to reduce TF risks when applying the national innovation-driven development strategy. For example, policymakers should take actions such as bilateral consultation or the WTO dispute settlement mechanism to build a win-win relationship with its trading partner. Second, since TI can also help China to survive in TF by increasing the welfare of worldwide consumers, policymakers should continue to encourage TI. Third, as TF can impede technology progress when economic growth relies mainly on exports, the government needs to build a robust domestic market and avoid over-dependence on overseas markets to reduce the risk of TF.

Furthermore, the relationship between TI and TF has important implications for firms. Firstly, firms engaged in TI should take measures to reduce TF risk. When trade remedy cases against Chinese firms occur, they should respond actively to minimise the loss of profits. Secondly, firms should realise that low-value-added enterprises mainly engaged in processing, assembly, and component manufacturing will suffer more in TF. However, when TI brings high-quality, low-cost, and irreplaceable products, TF may decrease. Therefore, firms should try to master core technology and develop high-performance products to deal with TF.

8.3. Research limitations and recommendations

This paper has some limitations that can be considered recommendations for future studies. The first limitation is that this paper focuses on the relationship between TI and TF in China while not considering other countries. TF is a common phenomenon in international trade, and the impact of TI on TF differs among different countries. Therefore, future studies can examine whether TI is always a push for TF in other countries and compare the results with ours. However, considering that China has the most patent applications and is the most prominent target of trade remedies investigations, it has special significance to study this issue using China as a sample.

The following limitation is the relatively limited period. Using the sample from January 2002 to December 2021, we find that TI reduced TF in 2021. This is because Chinese firms pay more attention to developing their high-tech products, enabling them to circumvent technical trade barriers. They also benefit their trading partners by producing high-quality and cheaper products. However, we have yet to determine whether this phenomenon will be a long-term trend. Hence, future studies could extend the sample period to see whether this conclusion still holds. However, the current period is enough to draw a meaningful conclusion that TI has both positive and negative effects on TF.

Finally, this paper uses the number of trade remedies cases that Chinese exporters counteract to measure TF, which does not include the trade remedies investigations that Chinese firms do not respond to. In the first few years after China entered the WTO, Chinese exporters rarely responded to trade remedies investigations such as anti-dumping. As they become familiar with international trade rules, more exporters actively respond to trade remedies cases to safeguard their interests. Research can be continued by considering all of the trade remedies investigations to examine their relationship with TI. However, considering this paper aims to investigate TF rather than unilateral trade remedies investigations, the number of trade remedies cases that Chinese exporters counteract is a better indicator of the conflict between China and its trading partners.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

The work was supported by National Social Science Foundation of China (No.20&ZD101) and Fujian Social Science Planning Major Project(No. FJ2020MJDZ049).

Notes

1 http://cacs.mofcom.gov.cn/cacscms/view/notice/ckys#

2 https://www.ceicdata.com

3 This paper also uses the rolling-window widths of 20-, 28- and 32- months to explore the causality, and the results do not change significantly, which proves the robustness of the results.