Examining structural stability and time-varying causality between exports and economic growth in China amidst shocks

Can the Chinese economy still be driven by exports in the face of global shocks? This research question is explored in this brief study. Economists and public policy analysts widely asserted that China’s economic growth and development were built on a strong foundation of the export-oriented manufacturing sector (Johnston, 2024). Precisely, China embarked on the mission of export-led growth in the 1980s, marked by a series of economic reforms such as trade liberalisation (particularly joining the World Trade Organization [WTO] in 2001), the reindustrialisation process around 2004 and the processing of export goods (Yao, 2014). However, concerns have emerged about whether the Chinese economy can continue to rely on exports as its primary driver in the years ahead. Roubini (2024) argues that China’s export-led growth model has broken down. He provides a reason for this development. He notes that after three decades of robust economic expansion, during which the economy grew at an average annual rate of 10%, China’s growth has slowed significantly. Currently, the growth rate hovers around 5% and the International Monetary Fund projects that it could decline further to approximately 3.4% by 2028 ^[1]. Roubini attributes this economic outlook to structural, rather than cyclical problems. These include an ageing population, a collapsed real estate bubble, massive overhangs of private and public debt and a marked shift away from market-oriented reforms back to state capitalism.

While China faces internal challenges that can be structural, several external political and economic events have had a significant impact on its exports and economic growth. As shown in Figure A1 (see Appendix), four major global events have influenced both China’s economic growth and exports. The first event was the global financial crisis (GFC) of 2007–2008, triggered by the collapse of the financial giant Lehman Brothers (McKibbin and Stoeckel, 2010). As a result of the GFC, which had far-reaching global effects, China’s economic growth declined sharply from 14.23% in 2007 to 9.40% in 2009 before rebounding slightly to 10.64% in 2010. Similarly, exports as a percentage of GDP fell from 32.60% in 2007 to 27.19% in 2009. The second event is the trade tensions between China and the USA, during which both exports and economic growth declined in 2018 by 0.58 and 0.2%, respectively. The third event, the COVID-19 pandemic, did not significantly affect China’s exports but caused a sharp economic slowdown. Growth fell from 5.95% in 2019 to 2.24% in 2020, representing a total loss of 3.17% points in just one year. The fourth event is the ongoing Russia–Ukraine war, which has disrupted global value chains and driven a surge in global energy prices. According to Wang and Su (2023), the Russia–Ukraine war affected the Chinese economy through increases in the prices of energy, chemical and grain commodities. A key takeaway from this analysis is that China’s exports and economic growth have not returned to their pre-GFC levels, suggesting that these external shocks might have had some sort of impact on the country’s economy.

Global shocks usually lead to parameter instability in the relationship between macroeconomic variables, including exports and economic growth. Such instability can result in imprecise estimates and erroneous inferences, particularly when testing the export-led growth hypothesis (Ditzen et al., 2021; Raifu, 2023a, b). Moreover, conventional estimation techniques, such as vector autoregression–based methods, may not accurately capture the causal link between these variables (Raifu and Afolabi, 2023). The linear causality method usually assumes a stable, time-invariant relationship between variables over the study period. Under such an assumption, the linear causality method only captures the average long-run direction of causality, thereby masking potential structural changes in the series, caused by policy shifts and/or external shocks. In light of this, the present study uses novel methods or approaches to address two key objectives. First, we test the structural stability between exports and economic growth in China using Ditzen et al. (2021) method to determine whether the presence of instability, which could result in structural breaks in the relationship, affects the validity of the export-led growth hypothesis in the country. Second, we examine the time-varying causality between exports and economic growth, considering that their relationship has evolved due to various internal and external shocks. Precisely, we employed the Shi et al. (2020) time-varying causality method. Unlike the linear causality method, the time-varying causality method relaxes the assumption of structural stability by allowing the causal relationship to vary over time. While numerous studies have examined the relationship between exports and economic growth in countries worldwide, including China (see Ali and Li, 2018; Kalaitzi and Chamberlain, 2021; Kim et al., 2022), the current study contributes to the existing studies by simultaneously testing both structural stability and time-varying causality between exports and economic growth in China. These are unique contributions to the existing literature because the time-varying approach allows capturing of how causality between exports and economic growth evolves across different historical events or shocks. Thus, the hypothesis is that external and internal shocks do not fundamentally alter the structure of the relationship between exports and economic growth in China.

The remainder of the study is structured as follows: Section 2 presents the methodology and data sources. Section 3 discusses the findings, and Section 4 concludes the study with a discussion of policy implications.

Following Raifu (2023b), this study employs two estimation methods to explore the structural stability and time-varying causality between exports and economic growth in China. For the structural stability test, we use Ditzen et al. (2021). The method is used to detect whether structural instability has occurred between exports and economic growth during the period of investigation, which could potentially lead to structural breaks between the two variables and consequently weaken the validity of the exports-led growth hypothesis in China. We begin by specifying a neoclassical growth model that incorporates exports as the third variable apart from labour and capital, following Raifu (2024).

If we assume that A is constant and takes the natural logarithm of Eq. (1), we have

By incorporating the constant and error term, Eq. (2) is transformed as follows:

where $y$ is the output proxied by real gross domestic product, $k$ denotes capital represented by capital stock, $l$ is labour proxied by employed persons (10,000 persons) and $e$ denotes exports of goods and services. e is the error term assumed to be independently and identically distributed with zero mean and constant variance.

From Eqs. (1) to (3), a Cobb–Douglas production framework is used as a convenient analytical starting point to express the production process. However, the inclusion of exports does not convey the impression that exports are a direct input in output production. Rather, exports are introduced as an augmenting variable that influences total productivity through the channels such as technology transfer, economies of scale and efficiency gains associated with economic openness. In fact, Balassa (1978) argued that by concentrating on export expansion, a country can benefit from economies of scale, exposure to international competition and access to advanced technology through capital accumulation. Higher export earnings, in turn, can finance investment in productive capital and foster human capital development. The combined effect is the increase in productivity that spurs economic growth (Kumeka et al., 2024).

Given the above explanation, if we assume that Eq. (3) is characterised by structural breaks, then it can be re-specified as follows:

Following Ditzen et al. (2021), it is assumed that $t=Tj−1,...,Tj$ and $j=1,...,s+1$ with $T0=0$ and $Ts+1=T$⁠. From Eq. (4), we have $s$ breaks and $s+1$ regimes with regime $j$⁠, covering the observations $Tj−1,...,Tj$⁠.

The null hypothesis posits that the estimated coefficient in Eq. (4) remains unaffected by internal and external shocks, indicating no structural changes in the relationship between exports and economic growth in China. This hypothesis is tested against the alternative hypothesis. If structural breaks are detected, the ordinary least squares, fully modified ordinary least squares (FMOLS) and instrumental variables two-stage least squares (IV-2SLS) methods are employed to estimate the impact of exports on economic growth along the identified structural breaks. This is done in two ways: first, without considering other factors that can theoretically drive economic growth aside from exports and second, by accounting for other factors such as labour and capital that can drive economic growth in conjunction with exports (Raifu, 2017; Oladapo and Raifu, 2022; Raifu and Oshota, 2023a). The second approach allows us to model the exports-led growth hypothesis in China from a theoretical perspective, particularly the neoclassical growth framework (see Raifu et al., 2025). The second objective is to examine the time-varying causality between exports and economic growth. To do this, we employ a time-varying causality method developed by Shi et al. (2018).

We begin by specifying the time-varying causality framework based on a lag-augmented vector autoregressive (VAR) model suggested by Toda and Yamamoto (1995) as follows:

where k denotes the lag length, d is the maximum order of integration and t is the time trend.

The null hypothesis of no presence of Granger causality between y1t (economic growth) and y2t (exports) is specified as follows:

This is tested against the alternative hypothesis specified as

Given the above LA-VAR framework, Shi et al. (2018) developed three supremum Wald statistic tests used to test for the existence of time-varying causality between exports and economic growth. The three supremum Wald tests include recursive evolving, rolling window and forward recursive. The forward recursive Wald statistic [$f1,f2$] with a small sample side fraction $fw=f2−f1≥f0$ is given as $Wf2(f1)$ and the supremum Wald statistic version is given as

where $∧0={(f1,f1);0<f0+f1≤f2≤1,and0≤f1≤1−f0}$ for the small sample size $f0∈(0,1)$ in the regression. To test the time-varying causality, three dating rule procedures are provided as follows:

where $cv$ and $scv$ are critical values of $Wf$ and $SWf$ statistics, respectively. $fe∧$ and $ff∧$ are estimated chronologically, and their test statistics can exceed or fall below the critical values for the beginning and endpoints in the causal nexus.

The data used in this study include exports, real GDP, capital stock and the number of employed persons (10,000 persons) for the period of 1960Q1–2023Q4 ^[2]. We converted the annual data into quarterly data because the methods employed in this study require high-frequency data, particularly the time-varying causality approach developed by Shi et al. (2018, 2020). To ensure that our results are not biased, we followed a standard procedure commonly adopted in the literature. Specifically, the quadratic-match average method was used to disaggregate the annual series into quarterly series. This method has been widely applied in empirical studies (Türsoy and Faisal, 2018; Pata, 2025) because it adjusts for and addresses issues related to seasonal variation. Moreover, it preserves the originality of the annual data and the long-run trend of the series while producing a smooth quarterly pattern. The data on exports and economic growth are obtained from a new global macroeconomic database compiled by Müller et al. (2025). The capital stock data are curated from the Federal Reserve Bank of St. Louis ^[3]. The number of employed persons (10,000 persons) is taken from the China Compendium of Statistics (1949–2008) and the National Bureau of Statistics of China ^[4].

The descriptive statistics of the data are presented in Table 1, followed by the correlation analysis results in Table 2. As shown in Table 1, China’s exports and real GDP averaged CN¥4,850,000 million and CN¥24,000,000 million, respectively, during the period under consideration. Both exports and real GDP have grown over the years, ranging from a minimum of CN¥3,561.21 million to a maximum of CN¥25,345,392 million for exports and from CN¥674,000 million to CN¥110,000,000 million for GDP. The standard deviation values of exports and real GDP indicate a significant disparity between their mean and standard deviation, suggesting that both variables have experienced fluctuations. These fluctuations may be attributed to China’s reforms, external shocks, seasonality and changes in global demand, particularly for exports (Raifu et al., 2021a, b; Raifu, 2022; Raifu and Folarin, 2020; Raifu and Afolabi, 2024; Raifu et al., 2024a). Capital stock, used as a proxy for capital, has a mean value of $21,700,000 million (CN¥156.19 trillion), with values ranging from $623,680 million (CN¥4.48 trillion) to $100,500,000 million (CN¥722.68 trillion) ^[5]. This shows that capital stock has grown substantially over the years in China, reflecting its commitment to investment and growth. The average number of employed persons stood at 56,809.50 (10,000 persons), with a minimum of 25,569.06 (10,000 persons) and a maximum of 76,352.59 (10,000 persons). This indicates that the number of employed persons in China has been on the rise over the year, a reflection of reforms and population dynamics.

The summary statistics of growth rates of the variables support the level summary statistics. Exports grew from a minimum of −24.63% to a maximum of 123.75%, with an average growth rate of 15.66% over the period under review. A similar trend is observed for real GDP growth, which ranges from a minimum of −31.20% to a maximum of 21.26%, with an average of 7.79%. This reflects the significant progress China has made in terms of export expansion and economic growth, largely driven by the series of economic reforms implemented by the Chinese government over the years.

Another important preliminary test conducted is the correlation test among our variables of interest. This test is performed to examine the degree of association between these variables. The correlation results are presented in Table 2. As evident from the table, there is a strong, positive and significant correlation between exports and economic growth, with a correlation coefficient of 0.992. This suggests that export performance has been a major driver of GDP growth in China, reinforcing the export-led growth hypothesis, which aligns with the Chinese economic growth model (Wu, 2000). Such a strong relationship implies that policies promoting trade, investment in export-oriented industries and infrastructure development have played a crucial role in sustaining economic growth (Raifu et al., 2021c; Raifu and Oshota, 2023b; Raifu et al., 2024a). Furthermore, both capital stock and employed persons are also positively correlated with economic growth, with correlation coefficients of 0.997 and 0.936, respectively. This indicates that capital stock and employed persons are essential contributors to economic growth (Adeboje et al., 2021).

To determine the stationarity properties of our variables of interest, we conduct unit root tests. Two methods are employed for this purpose. The first is the Phillips-Perron (1998) unit root test, which assumes that the variable contains a unit root. In other words, the variables are not stationary at their level, meaning they follow a random walk (Raifu and Abodunde, 2020; Raifu, 2022). Therefore, they need to be differenced to achieve stationarity. The second method used in this study is the Kwiatkowski et al. (1992) unit root test (KPSS). Unlike the Phillips–Perron test, the KPSS test assumes that the variables are already stationary at their level (Kwiatkowski et al., 1992). The results of the unit root tests are presented in Table 3. As shown in the table, and based on the Phillips–Perron test results, all variables, except for employed persons, which is stationary at the level, are stationary at the first difference. This suggests that these variables contain a unit root and must be differenced to become stationary. The KPSS test results further confirm the findings of the Phillips–Perron unit root test.

In this section, we present the main empirical findings from our analysis. We begin by discussing the results of structural breaks between exports and economic growth, which are displayed in Table 4. The annual data used in this study were converted to quarterly data (1960Q1–2023Q4) using the quadratic-match average method, a method widely employed in the literature (Raifu and Oshota, 2023a, b). This is important because the time-varying causality method requires high-frequency data, which are not consistently available for the period under consideration. As shown in Table 4, the null hypothesis of no structural breaks between exports and economic growth is rejected as we identified approximately five structural breakpoints. These breaks occurred in 1969Q2, 1978Q4, 1990Q2, 2002Q3 and 2014Q1. These break dates were endogenously determined or identified by the econometric method used (Ditzen et al., 2021), which allows for multiple structural changes over time without requiring the breakpoint to be known ex ante. Besides, the rejection of the null hypothesis of no structural breaks across the sample period further reinforces that the relationship between exports and economic growth in China is time-varying and suffers structural instability over time. This further supports the use of the time-varying causality approach to test the causality between exports and economic growth.

Some of the identified break dates above align with significant periods in China, Asia and the global economy. For instance, the structural break in 1978Q4 coincides with the economic reforms initiated by the Communist Party of China in the late 1970s. Precisely, 1978 marked the beginning of developmental reforms in China (Yao, 2014). According to Brandt and Rawski (2020), these reforms include the reduction of trade barriers, the introduction of market-driven pricing mechanisms and increased investment in heavy industries, all of which contributed to accelerated economic growth witnessed since 1970. The surge in economic growth during this period was accompanied by a significant expansion in exports throughout the 1980s.

Similarly, the decline in economic growth towards the end of the 1960s prompted the adoption of reforms that revitalised the Chinese economy. Throughout the 1980s, the Chinese economy experienced simultaneous growth in exports and economic output (Morrison, 2019). However, exports and economic growth have exhibited fluctuations over the years, likely influenced by external factors such as the Asian financial crisis of the 1990s, as well as internal economic dynamics. Despite these fluctuations, China’s overall economic performance remained robust due to the continuous pursuit of reforms.

In the 2000s, while exports continued to grow significantly, the economy experienced a slowdown, which might have contributed to the structural break observed in 2002Q3. Elizaveta (2018) attributes this slowdown to a negative correlation between consumption and investment, driven by excessive investment in heavy industries and the neglect of light industries. This imbalance led to economic growth outpacing income growth, resulting in subdued demand for goods and services domestically. Although the economy eventually rebounded, the recovery was gradual and was further disrupted by the GFC of 2008. The effects of the GFC on exports and economic growth are reflected in the structural break of 2014Q1. As illustrated in Figure A1, both exports and economic growth have yet to return to their pre-GFC levels. This suggests that the long-term impacts of the crisis, coupled with other internal and external factors, continue to influence China’s economic trajectory. Following Ditzen et al. (2021), it is essential to test the validity of the structural breaks identified in this study. To do so, we begin by assuming the null hypothesis that there are no structural breaks, against the alternative hypothesis that there are five breaks. The results of this test lead us to reject the null hypothesis, thereby validating the presence of the five structural breaks in the relationship between exports and economic growth (see Column 3 of Table 4). Based on this finding, we conclude that the five breaks are statistically significant and valid.

Next, we examine whether the export-led growth hypothesis holds along the identified structural breaks. Each structural break identified above corresponds to a distinct regime in which the coefficients of exports are permitted to vary across regimes, as determined by the five break dates. However, the coefficients for capital stock and the number of employed per persons remain constant in a similar version to Adu-Darko and You (2025). To estimate the coefficients along the regimes, we employ three estimation methods: OLS, FMOLS and IV-2SLS. While three methods are used for robustness, there is a key distinction between them. OLS is an estimation technique that minimises the sum of squared residuals; however, it does not account for serial correlation or endogeneity biases in long-run relationships between variables. In contrast, FMOLS, proposed by Phillips and Hansen (1990), modifies OLS by adjusting for serial correlation and endogeneity, making it more suitable for estimating long-term relationships in time series data (Kao and Chiang, 2001). In time series analysis, the issue of endogeneity usually arises because the main independent variable is believed to be endogenous. In our study, our main independent variable, exports, is believed to be endogenous, meaning it is correlated with the error term due to factors such as reverse causality, omitted variable bias or measurement error. Thus, IV-2SLS is deployed to address this issue of endogeneity by using an external instrument that is correlated with the endogenous variable but uncorrelated with the error term, thereby producing consistent and unbiased estimates. In this study, the lags of independent variables (exports and capital stocks) are used as instruments. We used the lag of independent variables for two reasons. First, it has been argued that if a variable is endogenous, its lag values may not be correlated with the error term; hence, it can be used as a valid instrument (Wooldridge, 2010). Second, it has been widely used in many empirical studies (Raifu, 2019; Raifu and Afolabi, 2024).

The results in the Table 5 are divided into two parts. The first part provides the baseline estimates, where we regress economic growth on exports alone (without structural breaks) and then on exports along with the structural breaks (exports with breaks 1 to 5). The second part focuses on robustness checks. These robustness checks are conducted for two main reasons. First, in line with Balassa’s (1978) export-led growth hypothesis, which posits that exports, along with capital and labour, drive economic growth. We use capital stock as a proxy for capital, while labour is proxied by the number of employed persons. Thus, we regress economic growth on exports, capital stock and the number of employed persons. This approach provides a theoretical foundation for our analysis. Failure to account for those two variables could lead to an overestimation of the effect of exports on economic growth.

For the baseline results based on OLS estimation, exports (without structural breaks), that is, pre-break, exhibit a positive but statistically insignificant effect on economic growth. However, when FMOLS is used for estimation, the effect of exports on economic growth becomes positive and statistically significant at the 1% level. Specifically, a 1% increase in exports leads to a 0.546% rise in economic growth. Similarly, the IV-2SLS results also show a positive and significant effect of exports on economic growth. In this case, a 1% increase in exports leads to an increase in economic growth by 0.575%. This finding supports the validity of the export-led growth hypothesis in China, indicating that exports play a crucial role in driving economic growth in the country. Our results align with numerous studies focusing on China and other nations. For instance, Azam et al. (2015) highlighted that China’s participation in global international trade has been instrumental in spurring its economic growth for decades.

When structural breaks (exports 1–5) are incorporated, all estimation methods, that is, OLS, FMOLS and IV-2SLS, confirm that exports positively influence economic growth. Let us consider the effect of exports2 on economic growth as an example. The break point (1978) coincides with the period of significant economic reforms (trade liberalisation [open door policy] and establishment of export zones) that paved the way for China’s economic expansion. Our results reveal that the impact of exports on economic growth remains positive for both OLS and FMOLS estimates, indicating that exports played a crucial role in driving China’s economic growth as the reforms took effect. Specifically, a 1% increase in exports leads to a 0.225% rise in economic growth based on the OLS estimation method and 0.543 and 0.574% increases based on the FMOLS and IV-2SLS estimation methods, respectively. In fact, the coefficient estimate based on OLS indicates that the impact of exports on economic growth was greater after the reforms took effect than in the pre-break period. Although numerous internal and external shocks may have influenced the trajectory of exports and economic growth via different channels, including the trade channel (Murach and Wagner, 2021), these shocks do not significantly affect the positive relationship between exports and growth. In other words, export-led growth hypothesis remains valid even in the presence of structural breaks caused by domestic and global factors. This demonstrates the resilience of China’s economy to both internal and external shocks (Zheng et al., 2024). Zheng et al. (2024) argue that this resilience is evident in key economic sectors and can be attributed to the initial economic reforms, as well as the continuous reform measures implemented by the Chinese government. These reforms have not only strengthened the economy but also sustained the positive relationship between exports and growth. Thus, even without accounting for other factors influencing economic growth, it is evident that despite structural breaks, exports continue to have a significant impact on China’s economic expansion. This further solidifies the validity of the export-led-growth hypothesis in the Chinese context.

When capital stock and the number of employed persons are controlled for, the magnitude of the effect of exports on economic growth decreases significantly. In the OLS results, the effect of exports on economic growth becomes negative in most cases and is not statistically significant at 5%. The results may be attributed to the weaknesses of OLS mentioned above. However, for FMOLS and IV-2SLS, although the effect of exports – both with and without structural breaks – remains positive and statistically significant, the magnitudes of the effects decline. The results show that, without structural breaks, exports exhibit a robust positive impact on economic growth, with estimated elasticities of 0.645% (FMOLS) and 0.201% (IV-2SLS), respectively. This implies that, holding other factors constant, a 1% increase in exports corresponds to approximately a 0.645 and 0.201% rise in economic growth. However, when structural breaks are incorporated into the analysis, the marginal effects show progressive attenuation across successive break periods (exports2 = 0.570% [0.192%], export3 = 0.530% [0.182%], export4 = 0.496% [0.175%]) ^[6].

This diminishing pattern in export elasticity likely reflects the impact of the internal and external economic shocks previously discussed. Aside from this, it also underscores the importance of incorporating capital and employment, as suggested by economic theory, to avoid overestimating the impact of exports on economic growth. Nevertheless, regardless of whether structural breaks or other variables are included, the export-led growth hypothesis remains valid in China. Exports are indispensable for sustaining economic growth in the country.

We now turn to the effect of control variables – capital stock and employed persons – on economic growth. From the three estimation methods used (OLS, FMOLS and IV-2SLS), it is evident that both capital stock and employed persons have significant effects on economic growth ^[7]. Precisely, the OLS results show that a 1% increase in capital stock and employed persons leads to economic growth by 0.857 and 0.645%, respectively. Similarly, based on FMOLS results, a 1% increase in capital stock and employed persons spurs economic growth by 1.282 and 0.811%, respectively. For the IV-2SLS estimation, a 1% increase in capital stock leads to a 0.689% increase in economic growth. Though the effect of employed persons is positive, it is not statistically significant. Generally, our findings clearly show that capital stock drives economic growth more significantly than labour input. This divergence stems from two key trends: (1) sustained growth in capital investment and (2) adverse population dynamics marked by declining fertility rates and rising life expectancy. Together, these factors have accelerated population ageing, reduced labour productivity and ultimately constrained economic growth (Meng and Wen, 2024). Our findings are also consistent with the existing studies (see Raifu et al., 2021b, 2025a, b).

Our second objective is to test whether the causality between exports and economic growth in China is time-varying. By “time-varying”, we mean that the causal relationship between exports and economic growth is not constant over time but changes significantly at different points. In other words, the direction, strength or even the existence of causality may vary depending on the specific context or period under analysis (Raifu and Afolabi, 2023). The rationale for examining the time-varying relationship between exports and economic growth has been discussed in the introduction.

The results of the time-varying causality analysis are presented in Table 6 (see Figure 1 for time-varying causality graphs). However, before discussing these results, we first present the findings of linear causality between exports and economic growth using the Toda and Yamamoto (1995) approach. This method assumes a stable, time-invariant relationship between variables over the study period. The linear causality test reveals the existence of a partial unidirectional causality running from economic growth to exports. This suggests that in China, economic growth drives exports by enhancing firms’ production capabilities, technological efficiency and economies of scale. These gains enable them to generate output that exceeds domestic demand and serves international markets. This finding is consistent with those of Emery (1967) and Kalaitzi and Chamberlain (2020), which support the validity of the growth-led export hypothesis in China.

For the time-varying causality analysis, three approaches are employed: the forward expanding causality test, the rolling window causality test and the recursive expanding causality test. Although these three methods offer complementary perspectives on the dynamics of the export–growth relationship, they differ in how they capture temporal variations in causality. The forward expanding test evaluates causality by progressively enlarging the sample window from the beginning of the series. This method is particularly sensitive to the detection of emerging causal relationships in the early part of the sample and is well suited for identifying the onset of structural changes or breaks (Thoma, 1994). In contrast, the rolling window approach applies a fixed-length moving window across time, allowing for the identification of local fluctuations in the strength and direction of causality. This method is especially useful for examining temporary or cyclical changes in relationships between variables such as exports and economic growth (Arora and Shi, 2016). Finally, the recursive expanding approach operates by continuously extending the estimation sample while updating the parameter estimates. This technique is effective in identifying persistent and evolving causal patterns within macroeconomic dynamics (Phillips et al., 2015).

While the linear causality test supports only a unidirectional relationship, the time-varying causality results reveal a bidirectional relationship between exports and economic growth across all three methods employed – forward, rolling and recursive Wald causality tests (Shi et al., 2020). This conclusion is based on the fact that the computed Wald test statistics exceed their corresponding critical values during several periods under analysis. This indicates that the linear method fails to fully capture the dynamic nature of the causal relationship over time. The evidence of bidirectional causality implies that exports can stimulate economic growth, while economic growth can also subsequently drive export expansion. The detailed results are reported in Table 6 and graphically illustrated in Figure 1, which shows how the causality between exports and economic growth changes over time. However, the specific timing of causality differs across the three methods. For example, the forward expanding Wald causality test shows that causality occurs early, particularly in the 1980s and into the 1990s, when exports Granger-cause economic growth. This period aligns with major economic reforms in China, such as the Open Door Policy initiated in 1978 and the establishment of Special Economic Zones in 1979 and 1980, which substantially boosted both exports and growth through rapid industrial expansion. Wei (1995) confirmed that China’s openness reforms triggered rapid export growth, industrial production and overall economic expansion during the 1980s and 1990s, with a strong positive export–growth relationship (see also Bohnet et al., 1993). Similarly, the recursive causality test identifies causality around the same reform-driven period. By contrast, the rolling causality test captures additional episodes of causality, including the early 1980s, the 1990s and notably the post-GFC period of the 2010s, for causality running both from exports to growth and from growth to exports. Thus, our findings align with several studies that have employed similar methods to examine time-varying causality between macroeconomic variables. For instance, Raifu and Afolabi (2023) identified bidirectional time-varying causality between military spending and certain macroeconomic variables in the USA. Similarly, Raifu (2023a) established bidirectional time-varying causality between oil prices and stock returns in Norway (see also Raifu, 2023b; Raifu and Lasisi, 2025; Raifu and Obaniyi, 2025).

China’s economic growth has largely been driven by export-led growth via industrialisation. However, recent internal and external developments pose significant risks to this growth model. These economic shocks could destabilise the export-led growth strategy, leading to structural breaks in the model. Given this concern, the study aimed to achieve two key objectives. The first was to assess the structural stability of China’s growth model and determine whether the export-led growth strategy remains viable despite structural breaks. The second objective was to examine the direction of causality between exports and economic growth using a time-varying causality approach. To achieve the first objective, the study employed the Ditzen et al. (2021) structural break test, which improves upon the Bai and Perron (1998) multiple structural breaks method. In addition, it used a time-varying causality framework developed by Shi et al. (2018, 2020) to analyse the direction of a causal relationship between exports and economic growth. The study utilised annual data, converted into quarterly data from 1960 to 2023, to ensure a detailed and dynamic analysis.

Our results reveal that structural instability exists between exports and economic growth, primarily driven by internal shocks rather than external ones. This instability causes structural breaks in the relationship between exports and economic growth. However, despite these breaks, the export-led growth hypothesis remains valid as exports continue to have a positive and significant impact on economic growth. Furthermore, our findings suggest that a time-varying approach is the most effective way to model the causal relationship between exports and economic growth. While the linear causality approach, based on the Toda–Yamamoto method, indicates a unidirectional causality running from economic growth to exports, the time-varying approach reveals a bidirectional causality. This means that exports can Granger-cause economic growth and economic growth can also Granger-cause exports, creating a feedback loop where both variables influence each other. In other words, as exports drive economic growth, economic growth, in turn, drives exports.

Our findings have both theoretical and policy implications. From the theoretical perspective, the evidence of time-varying bidirectional causality and structural breaks in the relationship between exports and economic growth supports dynamic theories of exports and growth models such as the export-led growth hypothesis or endogenous growth theory, which posits that exports and economic growth reinforce each other rather than following a one-way relationship. Thus, the results challenge the static and linear causality assumptions suggested by the traditional growth models, showing instead that the export–economic growth relationship evolves over time or at best regime or time-dependent and influenced by reforms or policy shifts and sometimes external shocks. Therefore, our findings confirm that trade and growth are mutually dynamic; hence, the theoretical model should account for the feedback effect where export and growth drive each other.

In terms of policy implications, our findings indicate that despite the presence of external shocks, export promotion remains vital to sustaining China’s economic growth. This confirms the validity of the export-led growth hypothesis, and vice versa, as a solid development strategy for China. However, to remain competitive in a rapidly evolving global environment, China must prioritise industrialisation and innovation. This includes promoting export expansion through deeper trade liberalisation while simultaneously investing heavily in infrastructure, technology and human capital development. Such pro-growth policies are crucial as our analysis clearly shows that stronger economic growth further enhances export capacity. By strategically balancing export promotion with domestic growth-enhancing policies, China can reinforce its economic resilience and maintain its growth trajectory. Moreover, institutional and structural reforms, such as the continued establishment and modernisation of special economic zones across different provinces, should be prioritised and periodically modernised, given their historical effectiveness in the 1980s and 1990s. Also, the structure of the Chinese economy is currently undergoing a significant transition from an export-driven model to a consumption-driven growth model. This shift is largely driven by the contraction of major export markets, rising trade protectionism, excess industrial capacity and the increasing reliance on pushed rather than demand-driven exports. In light of these developments, the Chinese government needs to implement policies and programmes such as household income support, social welfare expansion or targeted consumption incentives that will stimulate domestic demand.

Our study is not without limitations, particularly given the evolving economic dynamics across Asian countries, including China. Many of these economies are shifting away from an export-led growth model toward consumption-driven growth, in which economic expansion is primarily driven by household spending rather than exports, government expenditure or investment. In such a model, rising consumer demand stimulates production, generates employment, increases firm profits and further encourages investment, creating a self-reinforcing cycle of domestic demand-led growth. Consequently, countries such as India, South Korea, Singapore and China are increasingly rebalancing their growth strategies towards internal demand. Specifically, the World Bank’s 2025 China Economic Update emphasises the need for China to unlock and strengthen domestic consumption (World Bank, 2025). Therefore, future studies on China should focus on the examination of consumption-driven growth, potentially employing methodologies similar to those adopted in this study.

The supplementary material for this article can be found online.