The previous section discussed how scholars frequently use the gravity model to analyze bilateral trade. Building on this foundation, the current study employs the trade gravity model but with specific modifications to suit our analysis. The basic formulation of the gravity model is as follows:

$Trad{e}_{ij}=\alpha \ast \left(GD{P}_i\ast GD{P}_j\right)/Di{s}_{ij}$ (1)

In Equation (1), $Trad{e}_{ij}$ represents the bilateral trade flow between country i and country j. $GD{P}_i$ and $GD{P}_j$ denote the gross domestic products of countries i and j respectively. The constant term is denoted by $\alpha$ and $Di{s}_{ij}$ represents the geographical distance between country i and country j.

Taking the logarithm of both sides of the equation is a classical approach to evaluate the basic gravity model, as it transforms the model into a linear form that is more amenable to regression analysis. Therefore, Equation (1) can be expressed in its logarithmic form as follows:

$\ln \left(Trad{e}_{ijt}\right)=\alpha +{\beta }_1\ln \left(GD{P}_{it}\right)+{\beta }_2\ln \left(GD{P}_{jt}\right)-{\beta }_3\ln \left(Di{s}_{ijt}\right)+{\epsilon }_{ij}$ (2)

In Equation (2), $\ln \left(Trad{e}_{ijt}\right)$ represents the natural logarithms of bilateral trade flow between countries i and j in year t. The indices i and j denote different trading partners, and t denotes the year. $\ln \left(GD{P}_{it}\right)$ and $\ln \left(GD{P}_{jt}\right)$ represent the natural logarithms of the gross domestic products of countries i and j in year t, respectively. $\ln \left(Di{s}_{ijt}\right)$ is the natural logarithm of the geographic distance between countries i and j. The constant term $\alpha$ represents the intercept, which captures effects in the model that do not vary with the explanatory variables. The coefficients ${\beta }_1$, ${\beta }_2$, and ${\beta }_3$, are the estimated parameters that measure the impact and direction of the respective explanatory variables (GDP and distance) on trade flow. ${\epsilon }_{ij}$ is the error term, representing the impact of other factors not observed in the model.

Given the constant geographic distance across the Taiwan Strait, this study excludes the impact of geographical distance on bilateral trade flows. It instead examines the effects of the economic sizes of the regions on either side of the Taiwan Strait, Chinese Mainland’s FDI in Taiwan region, and the exchanges in the exchange rate between the CNY and TWD on Chinese Mainland’s imports of agricultural products from Taiwan region. Consequently, Equation (2) is modified to focus on these variables as follows:

$\ln \left(Trad{e}_{ijt}\right)=\alpha +{\beta }_1\ln \left(GD{P}_{ijt}\right)+{\beta }_2\ln \left(FD{I}_t\right)+{\beta }_3\ln \left(Exc{h}_{ijt}\right)+{\epsilon }_{ij}$ (3)

Per capita real income is a crucial indicator for measuring the average economic level and purchasing power of a population within an economy. The disparity in per capita real income between two economies can significantly impact bilateral trade flows. In this study, per capita real income is represented by GDP per capita at constant US dollar values. The focus is on examining the impact of the GDP per capita gap between Chinese Mainland and Taiwan region on agricultural trade. The equation to model the GDP per capita gap’s effect on agricultural trade flows is formulated as follows:

$DGD{P}_{ijt}=\left|PercapitaGD{P}_i-PercapitaGD{P}_j\right|$ (4)

Consequently, incorporating the per capita GDP disparity as an explanatory variable, the trade gravity model for this study is specified as follows:

$\ln \left(T_{ijt}\right)=\alpha +{\beta }_1\ln \left(GD{P}_{ijt}\right)+{\beta }_2\ln \left(DGD{P}_{ijt}\right)+{\beta }_3\ln \left(FD{I}_t\right)+{\beta }_4\ln \left(Exc{h}_{ijt}\right)+{\epsilon }_{ij}$(5)

In Equation (5), $\ln \left(T_{ijt}\right)$ represents the natural logarithm of the agricultural trade flow from Chinese Mainland to Taiwan region in year t. Here, i denotes Chinese Mainland and j denotes Taiwan region. $\ln \left(GD{P}_{ijt}\right)$ denotes the natural logarithm of the product of GDPs of Chinese Mainland and Taiwan region in year t. $\ln \left(DGD{P}_{ijt}\right)$ represents the natural logarithm of the absolute value of the difference in per capita GDP between Chinese Mainland and Taiwan region in year t. $\ln \left(FD{I}_t\right)$ is the natural logarithm of the stock of direct investments from Chinese Mainland to Taiwan region in year t. $\ln \left(Exc{h}_{ijt}\right)$ denotes the natural logarithm of the exchange rate between the Chinese Yuan and the New Taiwan Dollar in year t. The constant term $\alpha$ captures the effects in the model that do not vary with the explanatory variables. The coefficients ${\beta }_1$, ${\beta }_2$, ${\beta }_3$, and ${\beta }_4$ are the estimated parameters that measure the impact and direction of the respective explanatory variables on trade flow. The error term ${\epsilon }_{ij}$ represents the influence of other factors not observed in the model.

This study employs balanced panel data spanning from 2012 to 2022, totaling 253 observations (11 × 23 = 253). According to the two-digit HS codes, the agricultural products imported by Chinese Mainland from Taiwan region are included in HS Chapter 01 to 24, excluding Chapter 02. This statistical classification is based on the research by Yang (2012) and the directory provided by the China Chamber of Commerce of Import and Export of Foodstuffs, Native Produce, and Animal By-products. Since no agricultural products under HS code Chapter 02 have ever been imported from Taiwan region by Chinese Mainland, these products are excluded from the data utilized in this research.

In this study, $T_{ijt}$ serves as the dependent variable, representing Chinese Mainland’s imports flows from Taiwan region agricultural products. The data is sourced from China Customs Statistics and is denominated in US dollars.

The primary explanatory variables of this study include:

1) $GD{P}_{ijt}$, representing the economic size of the Taiwan Cross-Strait region, calculated as the product of the Gross Domestic Products (GDPs) of Chinese Mainland and Taiwan region. We expect that a larger economic size corresponds to a higher import flows. The data is sourced from the Key Indicators Database of APEC, denoted in constant 2015 US dollars, and denominated in millions of US dollars.

2) $DGD{P}_{ijt}$, which is the absolute difference in per capita real income between Chinese Mainland and Taiwan region, represented by the absolute value of the difference in per capita GDP in constant US dollars. We hypothesize that a larger GDP per capita disparity between the two economies leads to a decrease in import flows. This data is also sourced from the Key Indicators Database of APEC and is denoted in constant 2015 US dollars.

3) $FD{I}_t$, the stock of direct investment from Chinese Mainland in Taiwan region. It is expected that there is a positive correlation between FDI and import flows. The data comes from the Ministry of Commerce of China, denominated in ten thousand US dollars.

4) $Exc{h}_{ijt}$, the exchange rate between the Chinese Yuan (CNY) and the New Taiwan Dollar (TWD). It is anticipated that a depreciation of the CNY against the TWD leads to a decrease in import flows. The exchange rate data, represented as the annual average, is sourced from the International Monetary Fund (IMF) Database ( Table 1 ).

Table 2 briefly summarizes the descriptive statistics of the dependent and independent variables before they are transformed into logarithmic form. All variables have 253 observations, as the data is balanced panel data. During the period from

Source: Author own work.

Source: Authors’ compilation from EViews 13.0.

2012 to 2022, the minimum import value of agricultural products from Taiwan region to Chinese Mainland was $¹614 (HS Chapter 14), and the maximum import value was $193.5 million (HS Chapter 03), reflecting the varying preferences of Mainland Chinese consumers for different categories of agricultural products in Taiwan region. The cross-strait per capita income disparity ranged from a minimum of $14,464.33 to a maximum of $17,345.67, with a standard deviation of $940.88, indicating a gradually widening income gap over the past 11 years. The stock of direct investment from Chinese Mainland in Taiwan region ranged from $135.32 million to $1.67 billion, with a mean value of $107 million and a standard deviation of $495.97 million, reflecting the continuous expansion of Chinese Mainland’s investment in Taiwan region over the past 11 years. The exchange rate between the CNY and the TWD had a minimum value of 0.1952 and a maximum value of 0.2333, with a standard deviation of 0.0117, indicating the stability of the exchange rate between the CNY and the TWD.

Pesaran (2004) proposed a general test for cross-sectional dependence in panel data, which involves using residuals to assess cross-sectional dependence. Before conducting pooled OLS, a Panel cross-section dependence test should be performed to determine whether the sample data exhibits cross-sectional dependence or is mutually independent. Failure to conduct this test may result in biased and inconsistent regression results based on the assumed gravity equation ( Irshad et al., 2018 ).

The results of the CD test for this study are shown in Table 3 . The P-values for the Breusch-Pagan LM, Pesaran scaled LM, and Pesaran CD tests are all below 0.05, indicating significant cross-sectional dependence in the residuals of the gravity model. Therefore, there is interdependence among the cross-sectional units in the panel data.

Source: Authors’ compilation from EViews 13.0.

To avoid spurious regression results, this study follows the methodology of Cheng (2012) by employing the LLC test for common unit root and the PP-Fisher test for individual unit root. If both tests reject the null hypothesis of a unit root, the series is considered stationary; otherwise, it is non-stationary. The unit root results in Table 4 show that the LLC test and PP-Fisher test P-values for $\ln \left(T_{ij}\right)$, $\ln \left(GD{P}_{ij}\right)$, $\ln \left(FDI\right)$, and $\ln \left(Exc{h}_{ij}\right)$ are all below 0.05, indicating stationarity. However, the results for $\ln \left(DGD{P}_{ij}\right)$ exceed 0.05, indicating non-stationarity at the level. After first differencing, the LLC test and PP-Fisher test P-values for $\ln \left(DGD{P}_{ij}\right)$ are both below 0.05, indicating stationarity in the first-difference.

Source: Authors’ compilation from EViews 13.0.

The multicollinearity results show that the Variance Inflation Factor (VIF) for the explanatory variables $\ln \left(GD{P}_{ij}\right)$ and $\ln \left(Exc{h}_{ij}\right)$ exceeds 10. After removing the explanatory variable $\ln \left(GD{P}_{ij}\right)$, the VIFs for the remaining explanatory variables in the gravity model are all Table 5 .

Source: Authors’ compilation from EViews 13.0.

Therefore, the gravity model should exclude the explanatory variable $\ln \left(GD{P}_{ij}\right)$, and the explanatory variable $\ln \left(GD{P}_{ij}\right)$ should be adjusted to its first-difference before performing Ordinary Least Squares (OLS) regression.

The Ordinary Least Square (OLS) regression results, as shown in Table 6 , indicate

*Significant at the 10% level. **Significant at the 5% level. ***Significant at the 1% level. Source: Authors’ compilation from EViews 13.0.

that the coefficients of the independent variables $\ln FDI$, $\ln \left(Exc{h}_{ij}\right)$, and the first-difference of $\ln \left(DGD{P}_{ij}\right)$ are not significant, and both R-Squared and Adjusted R-Squared are below 5% (0.05). This implies that the explanatory variables in the gravity model hardly explain the variations in the dependent variable. The Durbin-Watson statistic result is 0.153679, suggesting a strong positive correlation between residuals, indicating that the gravity model may suffer from omitted variable bias and that OLS regression is no longer appropriate, necessitating an adjustment in the regression method.

We take the lagged term of the dependent variable $\ln \left(T_{ij}\right)$ as an explanatory variable and employ the Generalized Method of Moments (GMM) for instrumental variables regression. The regression results, as shown in Table 6 , indicate that all explanatory variables have significant coefficients at the 1% level. Chinese Mainland’s import flows from Taiwan region’s agricultural products to exhibits dynamic effects and positive inertia effects, with a 1% increase in imports from the previous year promoting approximately a 0.256% increase in current year imports. In the GMM estimation, the coefficient of the one-year lagged foreign direct investment from Chinese Mainland to Taiwan region $\ln \left(FD{I}_{t-1}\right)$ is significant and consistent with the estimation. A 1% increase in foreign direct investment from Chinese Mainland to Taiwan region leads to a decrease of about 1.49% in agricultural imports from Taiwan region. The coefficient of the exchange rate between the Chinese Yuan (CNY) and the New Taiwan Dollar (TWD) $\ln \left(Exc{h}_{ij}\right)$ is significant and aligns with expectations. A 1% depreciation of the CNY against the TWD results in approximately a 3.49% decrease in agricultural imports from Taiwan region.

The study initially hypothesized a negative significant relationship between the explanatory variable $\ln \left(DGD{P}_{ij}\right)$ and import flows, predicting that a widening per capita income gap between Chinese Mainland and Taiwan region would reduce agricultural imports from Taiwan region. However, in the GMM estimation, the coefficient of the two-year lagged first-difference of the Cross-Strait per capita real income disparity $\left(D\ln \left(DGD{P}_{ij},-2\right)\right)$ is significantly positive, indicating that a 1% increase in the growth rate of the per capita GDP disparity between Chinese Mainland and Taiwan region leads to approximately an 11.46% increase in agricultural imports from Taiwan region. This result suggests that although the income gap between Chinese Mainland and Taiwan region is widening, this change has not reduced the demand for Taiwan region agricultural products in Chinese Mainland. On the contrary, other economic factors might be contributing to an increase in import volume. Since the explanatory variable $\ln \left(GD{P}_{ij}\right)$ was excluded due to a VIF greater than 10, which originally was expected to have a significant positive relationship with import volume, we have reason to believe that the expansion of economic size contributes to the increase in agricultural imports from Taiwan region to Chinese Mainland.

Based on the regression results, we derive the following gravity model for this study:

$\begin{array}{c}\ln \left(T_{ijt}\right)=0.255649\ast \ln \left(T_{ijt-1}\right)+11.46061\ast \text{D}\left(\ln DGD{P}_{ijt}\right)\\ -1.487925\ast \ln \left(FD{I}_t\right)-3.487474\ln \left(Exc{h}_{ijt}\right)+{\epsilon }_{ij}\end{array}$ (6)

Based on the above research findings, we propose the following policy recommendations for consideration by the authorities on both sides of the Taiwan Strait:

1) Chinese Mainland and Taiwan region signed the Cross-Strait Currency Clearing Cooperation Memorandum in 2010, establishing a currency clearing mechanism between the Chinese Yuan (CNY) and the New Taiwan Dollar (TWD). However, to date, cross-border settlement in local currencies across the Taiwan Strait has not been realized. While a clearing mechanism exists, a settlement mechanism has not yet been established, and electronic wallet payments do not facilitate transactions in local currencies. Our study indicates that exchange rate stability facilitates Chinese Mainland’s agricultural imports from Taiwan region. The financial authorities of the two economies should sign a local currency settlement agreement and actively assist financial institutions in establishing settlement accounts and electronic wallet payment channels for CNY and TWD to facilitate transactions for importers, exporters, and individuals.

2) The study shows a significant negative relationship between Chinese Mainland’s direct investment in Taiwan region and the import volume of agricultural products from Taiwan region. This could be due to trade policy restrictions imposed by the authorities in Taiwan region, as well as production shifts and import substitution effects. To increase the export of agricultural products in Taiwan region to Chinese Mainland, agricultural authorities across the Taiwan Strait should negotiate and sign cooperation agreements in the field of agricultural investment, supporting agricultural technology and sustainable development projects. Additionally, authorities in Taiwan region need to optimize policies regarding Chinese Mainland’s FDI in Taiwan region, such as relaxing market entry restrictions, particularly in the agricultural and food processing sectors, to promote Cross-Strait agricultural trade growth.

3) We also recommend establishing a Cross-Strait agricultural trade information platform to share real-time market demand, price fluctuations, and seasonal changes. This will help farmers and agricultural exporters in Taiwan region better plan production and exports, reducing risks associated with market uncertainties in Chinese Mainland. Although this study does not include political stability factors such as the “1992 Consensus” as a variable, it is important to acknowledge that the “One-China principle” plays a crucial role in Cross-Strait institutional consultations. Therefore, when the authorities across the Taiwan Strait negotiate and sign cooperation agreements, adherence to the “One-China principle” is essential.

In our study, we analyzed the determinants of Chinese Mainland’s agricultural imports from Taiwan region using data collected from 2012 to 2022 with a gravity model. The regression results were obtained using both the OLS estimator and the GMM estimator. Based on the GMM estimation, we found that Chinese Mainland’s agricultural imports from Taiwan region exhibit a positive inertial relationship with the previous year’s imports. It indicates that the previous year’s imports from Taiwan region help increase the following year’s imports. Additionally, the widening gap in per capita real income between the two economies is significantly positively related to import volume, indicating that an increase in the growth rate of the income gap between Chinese Mainland and Taiwan region promotes higher agricultural imports from Taiwan region. This intriguing phenomenon aligns with the H-O theory, which posits that differences in factor endowments drive greater international trade flows. The expansion of the income gap reflects a growing divergence in factor endowments between the two regions of the Taiwan Cross-Strait, which in turn fosters increased agricultural trade across the Strait.

The paper analysis reveals a significant negative relationship between Foreign Direct Investment (FDI) from Chinese Mainland into Taiwan region and agricultural import volumes from Taiwan region, indicating that increased direct investment by Chinese Mainland in Taiwan region may lead to production shifts or substitution effects, thereby reducing the demand for agricultural imports from Taiwan region. This finding is consistent with Mundell’s theory that FDI can substitute for some aspects of international trade flows.

Furthermore, the depreciation of the Chinese Yuan (CNY) against the New Taiwan Dollar (TWD) is associated with a decrease in agricultural imports from Taiwan region, highlighting the suppressive effect of exchange rate volatility on trade. In this context, adopting a local currency settlement mechanism between the two economies could mitigate exchange rate risks, stabilize trade conditions, and ultimately promote agricultural trade. Therefore, it is advisable for the Cross-Strait authorities to pursue the establishment of a CNY-TWD Cross-border settlement framework to enhance trade flows.

This study has certain limitations, as it does not address the export of agricultural products from Chinese Mainland to Taiwan region. Additionally, the potential impacts of the Economic Cooperation Framework Agreement (ECFA) and cross-strait political stability on agricultural trade were not considered. Future research should explore these aspects in greater depth to provide a more comprehensive understanding of the dynamics of agricultural trade between the two economies of the Taiwan Strait.