The impact of digitalization on foreign direct investment inflows into cities in China

Abstract

This paper explores the impact of digitalization on Foreign Direct Investment (FDI) inflows in 270 Chinese cities from 2012 to 2019, focusing on regional disparities in income levels. Employing the System Generalized Moment Method (GMM), it aims to bridge the gap in understanding how digitalization influences FDI inflows across regions with different income levels. The findings indicate a positive correlation in low-income cities where digitalization significantly attracts FDI, but this effect is limited in medium and high-income cities. These results highlight that incentives and digital infrastructure development could be crucial for enhancing FDI in lower-income regions, making digitalization a potential strategic tool for economic growth.

Impact Statement

This paper explores the impact of digitalization on Foreign Direct Investment (FDI) inflows in 270 cities in China from 2012 to 2019, focusing on regional disparities in income levels by employing the System Generalized Moment Method (GMM). As previous research on digitalization’s impact on FDI inflows especially at the city level in China is scarce, this study makes two contributions: firstly, we apply the generalized method of moments (GMM) to analyze the digitalization-FDI nexus, providing empirical insights at the city level, which has often been overlooked in previous studies. Secondly, by categorizing cities based on income levels, our study reveals variations in how digitalization impacts FDI across different economic levels. The results of this study offer solutions for economic growth in low-income cities, as the research findings show a positive correlation between digitalization and FDI attraction in low-income cities, although this effect is limited in middle - and upper-income cities. It is emphasized that the development of digital infrastructure is crucial for boosting FDI in low-income regions, thereby making digitalization a potential strategic tool for enhancing economic growth in these areas.

Keywords:

• Digitalization
• foreign direct investment
• regional analysis
• GMM
• China

REVIEWING EDITOR:

• Goodness Aye, Academic Editor, University of Agriculture, Makurdi Benue State, Nigeria

SUBJECTS:

• Economics
• Political Economy
• International Economics

JEL:

• F23
• O33
• O53

Introduction

China’s Foreign Direct Investment (FDI) landscape has witnessed a significant transformation following the country’s accession to the World Trade Organisation (WTO) in 2001. Despite global economic uncertainties, including the COVID-19 pandemic, China has seen resilient growth in FDI inflows, which increased annually by over 6% from 2020 to 2022 (Ministry of Commerce of China, 2022). The importance of FDI to host economies has been well documented in the literature (see Bermejo Carbonell & Werner, 2018; Borensztein et al., 1998; Iamsiraroj, 2016), including the case of China (see Su & Liu, 2016; Tang et al., 2008; Zhang, 2001). However, the inflows of FDI have been unevenly distributed, as shown in Figure 1, with a pronounced concentration in coastal regions (Lee & Chang, 2009; Rodrik, 1999), and could be the answer to the findings of unfavorable effect of FDI on income inequality specifically in China as observed by Gries and Redlin (2009), Zheng et al. (2021), Chen and Wu (2005), and Miyamoto and Liu (2005) amid the positive outcome of FDI to host areas.¹

Figure 1. Spatial distribution of China’s FDI inflows performance in 2012 and 2019.

Source: China Urban Statistics Yearbook (2013–2020).

With the importance of FDI inflows as a source of capital, technology, and management skills to local industry for the development of the host economy or area,² lagged areas in China could get more advantageous by more inflows of FDI. Several strategies have been stated as the key to the success of China in luring a high volume of FDI and becoming among the world largest recipients of FDI could be large domestic market, labor quality (as well as cost), agglomeration, infrastructure and institutional quality (Boermans et al., 2011; Li & Park, 2006; Na & Lightfoot, 2006), leveraging on the rapid development of digitalization across the globe, we are interested on the role of digital transformation on FDI inflows into cities in China. While traditional theories like the Eclectic Paradigm (Dunning, 1977; Dunning, 2002) have laid a foundational understanding of location and internalization advantages, they largely overlook the increasingly pivotal role of digital infrastructure in the economic landscapes of cities. Digitalization can be defined by Brennen and Kreiss (2016) as the adoption or increased use of digital or computer technology across various sectors. From Figure 2, we can observe a rapid improvement in digital transformation across China between 2012 and 2019, but mainly occurs in the coastal and Eastern areas. Several studies examine the effects of digitalization, starting from the role of government initiatives to encourage more private innovation (Wang et al., 2023), till the application in various areas such as economic development (S. Luo et al., 2023; Wang, et al., 2022), regional disparities (Liu et al., 2024), manufacturing sector (Miao, 2022), small and medium enterprises (Zhou & Liao, 2024) and environmental issue (J. Wang, Dong, et al., 2022). However, research on digitalization’s impact on FDI inflows especially at the city level in China, to our limited reading, is scarce. A central question our research seeks to answer is whether increasing digitalization across all cities in China will help promote FDI inflows.

Figure 2. Spatial distribution of China’s digitalization in 2012 and 2019.

Source: China Urban Statistics Yearbook (2013–2020).

This study has two contributions. We first apply the generalized method of moments (GMM) to analyze the digitalization-FDI nexus, offering empirical insights at the city level—a scale often overlooked in previous studies (Noussan & Tagliapietra, 2020). Second, by categorizing cities based on income levels, our study reveals variations in how digitalization impacts FDI across different economic levels.

This paper analyses the determinants of FDI in the context of city-level digitalization in China from 2012 to 2019. The remainder of the paper is organized as follows. The related literature is briefly reviewed in Section 2. The model construction and methodology are outlined in Section 3. The empirical results and discussion are presented in Sections 4 and 5, and Section 6 concludes the study.

Literature review

This literature review discusses two critical streams of research relevant to our study, which include theoretical frameworks and the empirical review of the relationship between digitalization and FDI, to find the gaps in the existing research and propose the hypotheses of this paper.

Theoretical framework of digitalization and FDI

Theoretically, many classical theories can explain FDI activities, in which Ownership, Location and Internalization (OLI) Framework or Eclectic Paradigm (Dunning, 1981, 1988; Dunning, 1977) explains why MNEs choose to invest in particular host countries, or why they choose to invest in particular locations within particular host countries. The location advantages of host countries are ignored by theories like Monopoly Advantage Theories. Meanwhile, Internalization Theories, which is taken into account by the Eclectic Paradigm, suggest the location advantages of host countries as an important influencing factor into the research framework.

The theory of location advantage uses policies, economic variables, and production costs to explain why different locations are more or less attractive for FDI. According to location advantage, location determinants like market size, labor cost, infrastructure development, government incentives, location, and openness are considered the main location determinants of FDI. Based on the Eclectic Paradigm and other relevant theories, this paper draws on previous studies for variable selection.

Digitalization can be regarded as technological progress, which greatly changes the way and speed of information transmission (Cenamor et al., 2017; Ciampi et al., 2022), and for multinational enterprises, digitalization helps to improve the return on investment (Luo, 2021). Investment income is the focus of multinational enterprises when making investment (Dunning & Lundan, 2010). Digitalization improves the attractiveness of FDI by reducing the information cost required by enterprises for cross-border investment. Specifically, if the local digitalization level is high, market phenomena have been formed and recorded in the institution, which means more data sources and makes it easier for investors to obtain and analyze, the collection of data on newly developed markets will also become more precise, which is also convenient for enterprises to make relevant investment plans (George & Schillebeeckx, 2022; Luo, 2021). Digitalized equiped region also enables foreign enterprises to quickly get familiar with the market and social situation of the host country (World Economic Forum, 2021). In addition, digitalization reduces the cost of talent search. Regions with higher levels of digital development often mean stronger technological and talent strength, which provides a large number of human resources for multinational enterprises (Grimpe et al., 2023). Moreover, digitalization can rely on information and communication technology to shorten the communication distance, effectively reduce the communication cost caused by geographical distance in business activities of enterprises, and enable enterprises to better integrate into the market environment of the host country.

Dunning’s Eclectic Paradigm is based on the old reality of the 1970s and 1980s. These realities and the assumptions behind them have changed considerably. Therefore, this paper believes that theoretically, digitalization can be introduced into the location advantage of OLI framework as a new motivation for transnational investment.

In addition to digitalization as a factor affecting FDI inflows, theories such as Neoclassical growth theory, Investment Development Path (IDP) theory and New trade theory also believe that other factors also have an important impact on FDI inflows. In Neoclassical Growth Theory, per capita GDP growth rate is considered to represent the overall improvement of the economy, indicating that a country has a good investment environment and potential large market demand, which is one of the factors considered by multinational enterprises to choose investment places. The Investment Development Path Theory highlights the importance of government spending, especially in infrastructure, in attracting FDI (Narula & Dunning, 2010). This aligns with the ‘location advantages’ in the Eclectic Paradigm. What is more, urbanization also plays a significant role in attracting international capital, as New trade theory indicated that agglomeration economies in urban areas are key factors influencing FDI (Fujita & Thisse, 1996; Guimarães et al., 2000; Tuan & Ng, 2004). Past studies have also shown that the exchange rate has an uncertain impact on FDI inflows. Cushman (1985) and Froot and Stein (1991) argue that the depreciation of the Renminbi promoted the inflow of foreign direct investment (FDI) based on the ‘Relative Production Cost Theory’ (Cushman, 1985) and the ‘Relative Wealth Hypothesis Theory’ (Froot & Stein, 1991). They believed that the depreciation of a country’s currency would lower the production costs of local goods, increase the return on FDI, and enhance the relative wealth of foreign investors.

Empirical reviews of digitalization and FDI

With China’s emergence as a digital leader (Woetzel et al., 2017), the role of digitalization in FDI has gained empirical attention. Studies indicate that digital factors like communication facilities and internet infrastructure attract FDI (Boermans et al., 2011; Mensah & Traore, 2022). According to Ha and Huyen (2022), digitalization is crucial in attracting FDI in the short- and long-term. They assert that digitalization can help overcome the challenges posed by the COVID-19 pandemic by leveraging data in 23 European countries. Some studies have shown that individual indicators of digitalization have a positive attractiveness of FDI, such as communication facilities and telecommunication level (Boermans et al., 2011; Mensah & Traore, 2022). Boermans et al. (2011) show that provinces with good communication facilities attract foreign investors. Mensah and Traore (2022) indicate that making high-speed internet an infrastructure quality indicator induces FDI in the banking and technology sector in the Belt and Road Initiative (BRI) countries. In another study, Sinha and Sengupta (2019) discerned that Information and communication technology significantly enhances productivity, efficiency, FDI inflows, and economic growth in developing countries.

Digitalization can be measured using various relevant indicators, some of which are discussed as follows. First is internet coverage, an indicator often used to measure digitalization. Choi (2003) selects 14 investing countries and 53 host countries to examine the relationship between the Internet and FDI. The study finds that for every 10% increase in internet-related indicators, FDI increases by at least 2%. Second, the proportion of computer services and software employees reflects the degree to which an economy leverages digital technologies and employs workers with digital skills to drive growth, innovation, and productivity (Kraus et al., 2021). The increasing use of digital technologies in businesses has led to a growing demand for workers with skills in computer services and software development (Möller et al., 2020). Third, telecommunications services, including advanced services such as broadband internet, mobile data, and other high-speed connectivity services, represent the extent to which people can access telecommunications services such as telephone, internet, and other communication technologies. Navas-Sabater et al. (2002) discuss how access to information and communications technologies has become crucial to sustainable economic development and poverty reduction agendas. Fourth, the postal service amount can reflect the efficiency and coverage of a city’s postal service. This also reflects digitalization through e-commerce. Efficient postal services can promote the flow of goods and documents, which may attract foreign investors seeking to establish a business in a city. Fifth is mobile phone penetration, an indicator used to measure digitalization (Corrocher & Ordanini, 2002). According to China Internet Network Information Center, in 2020, 99.7% of China’s Internet users (986 million) accessed the internet via their mobile phones, while 32.8% and 28.2% of them accessed the internet via desktop and laptop, respectively (Wang & Liu, 2021). Sixth is government science and technology (S&T) expenditure, an essential characteristic of digitalization. It refers to the allocation of fiscal revenue by the treasury to fulfill the goal of S&T transformation and innovation. Knowles et al. (2021) discuss that foundational research software infrastructure is critical for accelerating science but is often unsustainably funded. Solving this problem requires an appreciation of the importance of digital public goods and a commitment to invest in government.

However, some studies believe that digitalization can have stochastic influences on FDI inflows. For example, Sangroya et al. (2010) indicate that adopting a cloud computing paradigm may have adverse effects on the data security of service consumers. Brougham and Haar (2018) show that employees’ awareness of technological advancements is negatively related to organizational commitment and job satisfaction. Thus, FDI may become less attractive in the eyes of domestic regions.

Despite the extensive literature, there is a gap in understanding the comprehensive impact of digitalization, particularly at the city level, on FDI in China. Existing studies often rely on single indicators of digitalization. Our study addresses this gap by employing a comprehensive digitalization index, considering various digital aspects. We hypothesize a positive relationship between the level of digitalization and FDI inflows that will be tested using a dynamic model across most cities in China.

Methodology

In developing our model to examine the relationship between digitalization and FDI inflows in Chinese cities, we have drawn upon extensive literature and empirical studies, which has guided our variable selection, ensuring that each variable is relevant and provides meaningful insights into the dynamics of FDI. The model aims to quantify the factors influencing FDI inflows into Chinese cities empirically. The literature has highlighted the importance of government spending, GDP growth, urbanization, and exchange rates in determining FDI inflows. To incorporate these insights, the FDI inflows model is specified as follows: (1) $${\mathit{\text{FDI}}}_{i,t}={\beta }_0+{\beta }_1{\mathit{\text{GOVS}}}_{i,t}+{\beta }_2{\mathit{\text{GDPC}}}_{i,t}+{\beta }_3{\mathit{\text{URBP}}}_{i,t}+{\beta }_4{\mathit{\text{EXRT}}}_{i,t}+{\mu }_{i,t}$$(1)

Where $\beta$s are parameters to be estimated. $\mu$ is the stochastic term, i and t refer to cities and years, respectively. FDI denotes FDI inflows to Chinese cities, the amount of foreign capital utilized annually. GOVS is government spending, GDPC represents the annual per capita real GDP growth rate, URBP stands for the proportion of permanent urban residents in the city’s population, including the city’s urban and rural areas, EXRT is the exchange rate, the amount of 1 US dollar in Chinese yuan. Aiming to explore the relationship between digitalization and FDI inflows, the model is expanded to include digitalization (DIG) as a key variable: (2) $${\mathit{\text{FDI}}}_{i,t}={\beta }_0+{\beta }_1{\mathit{\text{GOVS}}}_{i,t}+{\beta }_2{\mathit{\text{GDPC}}}_{i,t}+{\beta }_3{\mathit{\text{URBP}}}_{i,t}+{\beta }_4{\mathit{\text{EXRT}}}_{i,t}+{\beta }_5{\mathit{\text{DIG}}}_{i,t}+{\mu }_{i,t}$$(2)

Traditional panel estimators such as pooled OLS, random effect, and fixed effect, can be biased and inconsistent due to the correlation between the lagged dependent variable and the error term (Ibrahim & Law, 2014). Considering these econometric concerns, this study applies a panel data model using the Generalized Method of Moments (GMM) technique for estimation, addressing endogeneity and biases associated with non-exogenous explanatory variables. This method aligns with methodologies outlined in studies by Holtz-Eakin et al. (1988), Arellano and Bond (1991), Arellano and Bover (1995), and Blundell and Bond (1998). Hence, in GMM format, setting X to represent the vector of explanatory variables, the dynamic panel model can be simplified as follows: (3) $${\mathit{\text{FDI}}}_{i,t}={\beta }_0+\gamma {\mathit{\text{FDI}}}_{i,t-1}+\alpha {\mathit{\text{DIG}}}_{i,t}+\beta X_{i,t}+{\lambda }_t+{\delta }_i+{\mu }_{i,t}$$(3)

In this study, all variables enter in natural logarithmic as indicated by ln, where $\gamma$ is the coefficient of the lagged dependent variable, α is the coefficient of the core variable DIG, X is a set of other independent variables, and ${\lambda }_t$ is a period-specific effect common to all countries ${\delta }_i$ is the city-specific effect, ${\mu }_{i,t}$ is the random variable.

The GMM estimator uses lagged explanatory variables as instrumental variables, which are used to address the possible correlation between the lagged dependent variable and the error term as well as the endogeneity of the explanatory variables. This method, known as first-order difference GMM, transforms Equation (2)(2) $${\mathit{\text{FDI}}}_{i,t}={\beta }_0+{\beta }_1{\mathit{\text{GOVS}}}_{i,t}+{\beta }_2{\mathit{\text{GDPC}}}_{i,t}+{\beta }_3{\mathit{\text{URBP}}}_{i,t}+{\beta }_4{\mathit{\text{EXRT}}}_{i,t}+{\beta }_5{\mathit{\text{DIG}}}_{i,t}+{\mu }_{i,t}$$(2) into first-order differences and helps to eliminate city-specific effects. However, Arellano and Bover (1995), Blundell and Bond (1998) and other studies have shown that using lagged levels of variables as differential regression instruments may lead to weak instrumental variables and biased estimates. In addition, Bun and Kiviet (2006) show that this method would lead to significant deviations in the first-order difference GMM of highly persistent variables.

To address these issues, this study employs a two-step system GMM estimator, which is more efficient than the first-order difference GMM estimator. Newey and West (1987) propose the two-step GMM estimator using the optimal weighting matrix, which showed better efficiency compared with the one-step estimator. Windmeijer (2005) also agrees on the effectiveness of two-step system GMM estimator.

It is necessary to conduct specification tests and diagnostic procedures to ensure the consistency of the GMM estimator. The Hansen overidentifying restriction test is included to assess the validity of the instrument and model specification. In addition, first-order serial correlation (AR(1)) and second-order serial correlation (AR(2)) need to be performed. When the model is valid, the null hypothesis of the absence of first-order serial correlation should be rejected, but the null hypothesis of the absence of second-order serial correlation should not be rejected.

This digitalization (DIG) index measurement involves three stages. The first stage involves identifying six indicators as the dimensions in digitalization. They are Internet penetration rate (INT), the proportion of computer services and software employees per 100 People (COM), total telecommunications service amount per 100 people (TEL), postal service amount per 100 people (POS), mobile phone user penetration rate (MPH), science and technology expenditure of government (SCE). The sub-indicators for the DIG index were selected with the support of relevant literature, including Boermans et al. (2011) and Mensah and Traore (2022), who emphasized the importance of digital infrastructure elements such as telecommunications facilities and mobile connectivity, and the following: Studies by Ha and Huyen (2022) and Choi (2003) highlight the importance of Internet penetration, e-commerce, and government technology investment as indicators of digitization in attracting investment.

In the second stage, each of them is to be normalized into a standard measurement by turning them into an index following the same formula by the United Nations in the construction of the Human Development Index (HDI) as follows:³ $$\begin{array}{c}\text{Dimension index =}\frac{\text{actual value ‐ minimum value}}{\text{maximum value ‐ minimum value}}\times 100\\ \end{array}$$

The third step, following the methodology used in constructing the HDI index, the third step involves computing the composite DIG index by averaging the six dimensions: $$\mathit{\text{DIG}}=\left(\frac{\mathit{\text{INT}}+\mathit{\text{COM}}+\mathit{\text{TEL}}+\mathit{\text{POS}}+\mathit{\text{MPH}}+\mathit{\text{SCE}}}{6}\right)$$

The rest of the variables are discussed in Table 1. The study uses panel data collected from 270 Chinese cities between 2012 and 2019. Table 1 displays the variables utilized in the study and their respective data sources.

Table 1. List of variables, descriptions, and sources.

Variables	Proxy	Descriptions	Sources
FDI	FDI inflows	Amount of foreign direct investment utilized per year	China Urban Statistics Yearbook (2013–2020)
GOVS	Government spending	Amount of government fiscal spending
GDPC	Economy development	Annual per capita real GDP growth rate
URBP	Urbanization level	Proportion of permanent urban residents in the city’s population
EXRT	Exchange rate	Period average price, Chinese Yuan (CNY) per 1 US Dollar (USD)
DIG	Digitalization Index	Digitalization Index is calculated by UNDP method

Most variables utilized in this study are sourced from the China Urban Statistics Yearbook for 2013–2020. Exchange rate data is obtained from China’s National Bureau of Statistics for the same period.

Results

Descriptive and correlation analyses are omitted to conserve space but are available upon request. This study presents full-sample regression analysis results using the two-step system GMM estimator, as shown in Table 2. To assess the contribution of each sub-indicator, separate regression analyses for sub-indicators were conducted in models 2a to 2f. Specification tests conducted before interpreting the regression results show that the lagged dependent variable is significant and positive across various estimations, validating the dynamic model. The null hypothesis of first-order autocorrelation (AR1) is not accepted, while the absence of second-order autocorrelation (AR2) cannot be rejected. The Hansen test supports the validity of the instruments used in the study.

Table 2. Regression Results of the Full sample [DV: FDI].

	Model 1: DIG = DIG	Model 2a: DIG = INT	Model 2b: DIG = COM	Model 2c: DIG = POS	Model 2d: DIG = MPH	Model 2e: DIG = SCE	Model 2f: DIG = TEL
FDIt-1	0.755***	0.743***	0.732***	0.693***	0.745***	0.753***	0.721***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
DIG	0.512*	0.081	0.144	0.082	0.059	0.136*	−0.010
	(0.068)	(0.360)	(0.194)	(0.150)	(0.633)	(0.059)	(0.849)
GOVS	0.337**	0.369**	0.408**	0.403***	0.346**	0.257	0.429***
	(0.049)	(0.012)	(0.020)	(0.005)	(0.015)	(0.133)	(0.005)
GDPC	−0.042	−0.039	−0.052*	−0.036	−0.030	−0.051*	−0.052*
	(0.187)	(0.171)	(0.086)	(0.226)	(0.352)	(0.065)	(0.084)
URBP	−0.051	0.356	0.412	0.363	0.327	0.054	0.686*
	(0.894)	(0.375)	(0.343)	(0.352)	(0.453)	(0.856)	(0.089)
EXRT	−2.250	−2.737*	−3.139	−2.723*	−2.259	−1.451	−3.802**
	(0.248)	(0.081)	(0.100)	(0.077)	(0.125)	(0.475)	(0.015)
	Model criteria
#Obs	1,512	1,512	1,512	1,512	1,512	1,512	1,512
#Cities	270	270	270	270	270	270	270
AR(1)	0.000	0.000	0.000	0.000	0.000	0.000	0.000
AR(2)	0.924	0.973	0.985	0.969	0.978	0.995	0.978
Hansen	0.030	0.078	0.112	0.111	0.081	0.308	0.128

Note. The p-values are reported in parentheses. *, **, and *** denote 10%, 5%, and 1% level of significance, respectively. The model is estimated with 2-step system GMM.AR and Hansen refer to p-value.

The analysis reveals that overall, DIG significantly positively impacts FDI, suggesting that an increase in the digitalization level corresponds with a rise in FDI. Within the regression results for sub-indicators, government Science and Technology Expenditure (SCE) notably positively influences FDI, indicating SCE as a potent factor within the context of digitalization’s impact on FDI attraction.

Certain sub-indicators like INT, COM, and POS did not significantly impact FDI in the whole sample, potentially due to low digitalization levels in cities. Addressing this, squared (DIG²) and cubic (DIG³) terms of digitalization were introduced⁴. Results in Table 3 indicate that even with these adjustments, the impact of squaring the digitalization level on FDI attractiveness remained insignificant in model 3(a, c, e, g, and i). Adding the cubic term of DIG in model 3(b, d, f, h, and j) reveals that sub-indicators such as INT and COM negatively affect FDI attractiveness at higher DIG levels, while other sub-indicators remain statistically insignificant.

Table 3. Regression results of the full sample with square and cubic [DV: FDI].

	Model 3a	Model 3b	Model 3c	Model 3d	Model 3e	Model 3f	Model 3g	Model 3h	Model 3i	Model 3j
	DIG = INT		DIG = COM		DIG = POS		DIG = MPH		DIG = TEL
FDIt-1	0.753*** (0.000)	0.746*** (0.000)	0.801*** (0.000)	0.750*** (0.000)	0.758*** (0.000)	0.755*** (0.000)	0.743*** (0.000)	0.735*** (0.000)	0.691*** (0.000)	0.692*** (0.000)
DIG	0.511* (0.091)	0.343** (0.048)	0.175** (0.041)	0.245** (0.024)	−0.107 (0.638)	0.030 (0.805)	0.038 (0.760)	0.208 (0.177)	−0.329 (0.469)	−0.164 (0.519)
DIG^2	−0.069 (0.136)		−0.068 (0.278)		0.017 (0.431)		−0.005 (0.951)		0.023 (0.458)
DIG^3		−0.009** (0.049)		−0.050* (0.067)		0.000 (0.751)		−0.063 (0.134)		0.001 (0.482)
GOVS	0.330** (0.010)	0.327*** (0.010)	0.314** (0.019)	0.297** (0.023)	0.257** (0.044)	0.277** (0.034)	0.276*** (0.010)	0.257** (0.036)	0.358*** (0.008)	0.348*** (0.009)
GDPC	−0.022 (0.513)	−0.021 (0.527)	−0.043 (0.194)	−0.057* (0.075)	−0.025 (0.388)	−0.025 (0.393)	−0.037 (0.301)	−0.038 (0.356)	−0.047 (0.225)	−0.047 (0.211)
URBP	0.340 (0.327)	0.342 (0.331)	0.190 (0.544)	0.436 (0.192)	0.267 (0.467)	0.290 (0.425)	0.470 (0.188)	0.354 (0.365)	0.888** (0.023)	0.892** (0.022)
EXRT	−2.639* (0.063)	−2.476* (0.073)	−2.233 (0.139)	−2.066 (0.159)	−1.108 (0.465)	−1.533 (0.307)	−1.678 (0.100)	−1.182 (0.362)	−2.517 (0.103)	−2.611* (0.072)
	Model criteria
#Obs	1,512	1,512	1,512	1,512	1,512	1,512	1,512	1,512	1,512	1,512
#Cities	270	270	270	270	270	270	270	270	270	270
AR(1)	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000
AR(2)	0.976	0.969	0.952	0.989	0.972	0.976	0.957	0.966	0.963	0.966
Hansen	0.117	0.127	0.309	0.288	0.199	0.183	0.0561	0.0933	0.0267	0.0255

Note. The p-values are reported in parentheses. *, **, and *** denote 10%, 5%, and 1% level of significance, respectively. The model is estimated with 2-step system GMM. AR and Hansen refer to p-value.

Discrepancies in the entire sample results may be attributed to regional imbalances (Li & Park, 2016). To further explore this, the model was re-estimated by dividing the total sample based on the average per capita income from 2012 to 2019 into low- (Table 4), middle- (Table 5), and high-income (Table 6) city sub-samples.

Table 4. Regression results of the sub-sample of low-income cities [DV: FDI].

	Basic model							Squared model			Cubic model
	Model 4 DIG = DIG	Model 4a DIG = INT	Model 4b DIG = COM	Model 4c DIG = POS	Model 4d DIG = MPH	Model 4e DIG = SCE	Model 4f DIG = TEL	Model 4g DIG = COM	Model 4h DIG = MPH	Model 4i DIG = TEL	Model 4j DIG = COM	Model 4k DIG = MPH
FDIt-1	0.636***	0.666***	0.712***	0.620***	0.712***	0.656***	0.632***	0.717***	0.687***	0.589***	0.727***	0.688***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
DIG	0.976**	0.508**	−0.164	0.240*	0.205	0.249***	−0.051	−0.287**	−1.149*	−1.240**	−0.005	−0.865
	(0.034)	(0.017)	(0.405)	(0.056)	(0.672)	(0.009)	(0.588)	(0.030)	(0.096)	(0.016)	(0.977)	(0.148)
DIG^2								0.131	−1.526*	0.093**
								(0.347)	(0.067)	(0.020)
DIG^3											−0.043	1.472*
											(0.654)	(0.076)
GOVS	0.778*	0.876**	0.764**	0.993**	0.793*	0.604	0.834**	0.710**	0.625*	0.939**	0.938**	0.635**
	(0.059)	(0.028)	(0.032)	(0.023)	(0.062)	(0.122)	(0.020)	(0.033)	(0.050)	(0.020)	(0.015)	(0.030)
GDPC	0.032	0.060	−0.000	−0.007	0.017	0.001	0.019	−0.025	0.020	0.032	−0.013	0.016
	(0.682)	(0.470)	(0.996)	(0.934)	(0.820)	(0.995)	(0.801)	(0.772)	(0.799)	(0.642)	(0.891)	(0.825)
URBP	0.622	0.357	0.787	0.627	0.418	1.019	0.855	0.616	0.216	1.060	0.842	0.300
	(0.432)	(0.676)	(0.311)	(0.505)	(0.619)	(0.183)	(0.242)	(0.441)	(0.737)	(0.180)	(0.245)	(0.654)
EXRT	−8.788	−9.067	−8.231	−10.504*	−7.827	−7.484	−8.343*	−7.581	−5.082	−7.564	−10.889**	−5.383
	(0.148)	(0.125)	(0.107)	(0.086)	(0.199)	(0.183)	(0.093)	(0.153)	(0.263)	(0.150)	(0.049)	(0.205)
	Model criteria
#Obs	501	501	501	501	501	501	501	501	501	501	501	501
#Cities	92	92	92	92	92	92	92	92	92	92	92	92
AR(1)	0.004	0.004	0.003	0.004	0.004	0.003	0.004	0.003	0.003	0.004	0.003	0.003
AR(2)	0.950	0.797	0.857	0.809	0.832	0.841	0.671	0.902	0.748	0.769	0.865	0.734
Hansen	0.779	0.821	0.866	0.826	0.877	0.797	0.811	0.996	0.999	0.993	0.998	0.999

Note. The p-values are reported in parentheses. *, **, and *** denote 10%, 5%, and 1% level of significance, respectively. The model is estimated with 2-step system GMM. AR and Hansen refer to p-value.

Table 5. Regression results of the sub-sample of middle-income cities [DV: FDI].

	Basic model							Squared model
	Model 5 DIG = DIG	Model 5a DIG = INT	Model 5b DIG = COM	Model 5c DIG = POS	Model 5d DIG = MPH	Model 5e DIG = SCE	Model 5f DIG = TEL	Model 5g DIG = DIG	Model 5h DIG = INT	Model 5i DIG = COM	Model 5j DIG = POS	Model 5k DIG = MPH	Model 5l DIG = SCE	Model 5m DIG = TEL
FDIt-1	0.697***	0.632***	0.722***	0.654***	0.674***	0.616***	0.652***	0.644***	0.717***	0.781***	0.597***	0.692***	0.621***	0.725***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
DIG	−0.100	0.121	−0.120	0.039	−0.271	0.065	−0.083	−4.450*	−0.307	−0.147	−0.304	0.054	−2.145*	0.437
	(0.728)	(0.453)	(0.471)	(0.555)	(0.146)	(0.589)	(0.435)	(0.052)	(0.716)	(0.330)	(0.567)	(0.784)	(0.059)	(0.648)
DIG^2								0.865*	0.057	0.046	0.037	0.269*	0.113**	−0.034
								(0.057)	(0.685)	(0.584)	(0.493)	(0.059)	(0.049)	(0.646)
GOVS	0.464***	0.411**	0.509***	0.463**	0.376**	0.416*	0.337**	0.433***	0.269*	0.486**	0.493**	0.278**	0.373**	0.287**
	(0.007)	(0.023)	(0.009)	(0.012)	(0.015)	(0.052)	(0.029)	(0.009)	(0.091)	(0.017)	(0.014)	(0.043)	(0.026)	(0.036)
GDPC	−0.018	−0.020	−0.019	−0.014	−0.027	−0.025	−0.020	−0.033	−0.015	−0.014	−0.022	−0.018	−0.043	−0.014
	(0.653)	(0.738)	(0.751)	(0.832)	(0.644)	(0.672)	(0.659)	(0.432)	(0.788)	(0.784)	(0.652)	(0.770)	(0.500)	(0.710)
URBP	−0.201	0.220	−0.335	−0.101	−0.208	−0.509	0.196	−0.148	0.342	−0.366	−0.460	−0.085	−0.432	0.634
	(0.627)	(0.631)	(0.469)	(0.803)	(0.623)	(0.371)	(0.741)	(0.726)	(0.354)	(0.392)	(0.369)	(0.850)	(0.422)	(0.330)
EXRT	0.297***	−1.867	0.256**	0.397***	0.019	0.377**	0.213**	–	–	–	–	–	–	–
	(0.003)	(0.421)	(0.032)	(0.008)	(0.833)	(0.034)	(0.047)	–	–	–	–	–	–	–
	Model criteria
#Obs	491	491	491	491	491	491	491	491	491	491	491	491	491	491
#Cities	89	89	89	89	89	89	89	89	89	89	89	89	89	89
AR(1)	0.014	0.016	0.016	0.015	0.014	0.016	0.014	0.015	0.016	0.015	0.016	0.014	0.017	0.014
AR(2)	0.334	0.362	0.322	0.342	0.323	0.349	0.358	0.351	0.338	0.299	0.345	0.337	0.352	0.336
Hansen	0.946	0.987	0.920	0.975	0.866	0.926	0.942	0.998	0.999	1.000	1.000	1.000	1.000	0.999

Note. The p-values are reported in parentheses. *, **, and *** denote 10%, 5%, and 1% level of significance, respectively. The model is estimated with 2-step system GMM. AR and Hansen refer to p-value.

Table 6. Regression results of the sub-sample of high-income cities [DV: FDI].

	Basic model							Squared model
	Model 6 DIG = DIG	Model 6a DIG = INT	Model 6b DIG = COM	Model 6c DIG = POS	Model 6d DIG = MPH	Model 6e DIG = SCE	Model 6f DIG = TEL	Model 6g DIG = INT	Model 6h DIG = COM	Model 6i DIG = MPH	Model 6j DIG = TEL
FDIt-1	0.762***	0.719***	0.694***	0.708***	0.729***	0.687***	0.761***	0.001	0.707***	0.698***	0.777***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.112)	(0.000)	(0.000)	(0.000)
DIG	0.681**	0.103	0.153	0.213**	0.109	0.244*	0.147	0.001	0.091	0.035	0.329
	(0.027)	(0.389)	(0.220)	(0.015)	(0.468)	(0.092)	(0.138)	(0.761)	(0.531)	(0.875)	(0.506)
DIG^2								−0.053	−0.027	0.009	−0.012
								(0.985)	(0.752)	(0.933)	(0.695)
GOVS	0.319*	0.677***	0.525*	0.388**	0.530**	0.392**	0.485***	0.947	0.539*	0.560**	0.310**
	(0.084)	(0.009)	(0.080)	(0.019)	(0.036)	(0.024)	(0.009)	(0.784)	(0.057)	(0.044)	(0.020)
GDPC	−0.185	−0.160	−0.124	−0.142	−0.163*	−0.169*	−0.153	0.135	−0.111	−0.154	−0.147
	(0.103)	(0.108)	(0.167)	(0.120)	(0.095)	(0.061)	(0.117)	(0.990)	(0.166)	(0.136)	(0.109)
URBP	−0.603	−0.352	−0.241	−0.611	−0.162	−0.477	−0.255	0.001	−0.260	−0.149	−0.091
	(0.248)	(0.445)	(0.555)	(0.255)	(0.693)	(0.367)	(0.598)	(0.210.)	(0.552)	(0.683)	(0.834)
EXRT	−0.998	−1.174	−1.365	−1.111	−0.707	−0.795	−0.661	0.001	−1.075	−2.178	−0.529
	(0.484)	(0.412)	(0.344)	(0.379)	(0.564)	(0.545)	(0.590)	(0.314.)	(0.455)	(0.149)	(0.680)
	Model criteria
#Obs	520	520	520	520	520	520	520	520	520	520	520
#Cities	89	89	89	89	89	89	89	89	89	89	89
AR(1)	0.126	0.131	0.133	0.802	0.127	0.136	0.129	0.920	0.141	0.134	0.127
AR(2)	0.204	0.216	0.223	0.914	0.218	0.203	0.210	0.972	0.210	0.214	0.199
Hansen	0.903	0.884	0.777	0.912	0.823	0.841	0.859	0	0.999	0.997	0.997

Note. The p-values are reported in parentheses. *, **, and *** denote 10%, 5%, and 1% level of significance, respectively. The model is estimated with 2-step system GMM. AR and Hansen refer to p-value.

In the sub-sample analysis, digitalization’s impact on FDI is notably more significant in low-income cities compared to middle- and high-income ones, as demonstrated in Table 4. For low-income cities, the proportion of internet users (INT), postal service amount (POS), and government science and technology expenditure (SCE) positively affect FDI.

In low-income cities, the proportion of computer services and software employees (COM) does not significantly impact FDI. Meanwhile, higher-level total telecommunications services (TEL² in model 4i) significantly correlate with FDI.

Furthermore, the relationship between the proportion of mobile phone users (MPH) and FDI inflows in low-income cities is initially negative at a squared level (model 4h). However, it turns positive at a cubic level (model 4k).

The regression analysis presented in Tables 5 and 6 suggests that the influence of digitalization on FDI attraction in cities with middle- and high-income levels is limited. In middle-income cities, squared-level digitalization (DIG), only mobile phone users proportion (MPH), and government science and technology expenditure (SCE) show significant positive impacts on FDI. In high-income cities, only overall digitalization (DIG), postal service amount (POS), and SCE are significantly positive factors.

Across the total sample and various income sub-samples, SCE consistently positively influences FDI inflows, indicating that government support in science and technology is a critical driver for attracting FDI. Model 2e shows a coefficient of 0.136 for the total sample, significant at the 10% level. In low-income cities (model 4e), SCE is significant at the 1% level with a coefficient of 0.249. In contrast, in middle-income cities, the squared SCE (model 5 l) is significant at the 5% level with a coefficient of 0.113. Model 6e shows SCE in high-income cities with a positive coefficient of 0.244, which is significant at the 10% level.

Discussion

The results in the previous section show that digitalization has a significant effect on attracting FDI, which is in line with the results of Ha and Huyen (2022), indicating that the increase of digitalization degree corresponds to the increase of FDI inflow. But the details of the results tell us that the effects are not always consistent across groups of cities with different income levels, with factors such as government investment in technology and regional digital infrastructure playing an important role.

In the total sample, the positive effect of government science and technology expenditure (SCE) on FDI is significant. This finding is consistent with Zhu et al. (2022), Bai et al. (2019) and Knowles et al. (2021), suggesting that government investment in science and technology contributes significantly to creating an FDI-friendly environment. Which means that the government’s initiatives in digital infrastructure development are more important among the many digital indicators to enhance the city’s attractiveness to foreign investors.

In the case of sub-sampling, the significant impact of digitalization on attracting foreign direct investment in low-income cities is more obvious than that in other samples, as shown in Table 4. The proportion of Internet users (INT), the volume of postal services (POS) and government expenditure on science and technology (SCE) show a positive impact on the attractiveness of investment, which confirms the emphasis on the role of infrastructure and government support in attracting FDI, consistent with previous literature (Asongu & Odhiambo, 2020; Choi, 2003; Knowles et al., 2021; Ko, 2007; Zhu et al., 2022).

However, the insignificant impact of the proportion of computer services and software employees (COM) suggests that in low-income cities, foreign investors might prioritize other aspects, such as infrastructure, policy incentives, or market size, over the availability of a digitally skilled workforce. This observation points to the multifaceted nature of FDI determinants, where factors like industry specifics and regional characteristics can influence investment decisions.

Furthermore, the significant positive correlation of higher-level telecommunications services (TEL²) with FDI in Table 4 underscores the importance of advanced digital services in creating a business-friendly environment. Services such as broadband internet and mobile data are critical for efficient business operations and global market access, making regions with high levels of TEL more attractive to foreign investors.

The relationship between the proportion of mobile phone users (MPH) and FDI inflows, transitioning from an inverted U-shape at lower levels (model 4h) to a U-shape at higher levels (model 4k), suggests a threshold effect. This implies that initial increases in mobile phone usage may not translate into immediate productivity gains or business opportunities but become beneficial beyond a certain penetration level. This situation is consistent with Kurniawati (2022), highlighting the importance of maintaining good ICT infrastructure for economic growth and sustainable development in Asia, and also highlighting the importance of increasing mobile subscriptions for sustainable economic growth.

The regression analyses in Tables 5 and 6 suggest that the effect of digitalization on FDI, while positive, is more limited in high-middle income cities compared to lower-income cities. This can be attributed to the already advanced infrastructure and technology in these developed cities, which limits the potential for further digitalization to have a significant impact on FDI inflows. While digitalization is one factor that attracts FDI, there are other factors besides it, especially in more developed urban environments. Middle and high income urban policymakers need to not only rely on digitalization strategies to attract FDI, but should also consider more comprehensive ways to attract foreign investment. This situation is consistent with the view put forward by the revised OLI paradigm (Dunning & Lundan, 2010), which emphasizes that the investment decisions of companies are affected by various institutional forces, including government regulations, policies and incentives (Francis et al., 2009).

A combined observation of all sample results reveals that the continued positive impact of government science and technology spending on all income levels highlights its important role as a digitalized sub-indicator in attracting FDI inflows. This shows that regardless of the income level of a city, government support for science and technology is always important in creating an environment conducive to foreign investment. This finding is consistent with Institutional Theory, suggesting that institutional forces significantly influence investment decisions, especially government regulations, policies, and incentives, and in line with the OLI Framework, government funding for digital infrastructure and related innovation development also establishes locational advantages that attract investment from Mnes, a finding also supported by Zhu et al. (2022), Bai et al. (2019) and Knowles et al. (2021).

Conclusion and policy implications

This study examines the impact of digitalization on FDI inflows in 270 cities in China from 2012 to 2019, providing new evidence on the important role of digitalization in attracting FDI. The results found that government spending on science and technology is a strong driver of attracting FDI, underscoring the importance of government support for digitalization.

In particular, the contribution of digitalization to FDI inflows varies in cities with the level of income. Although the impact of digitalization is limited in middle- or high-income cities, it has become a powerful tool to attract FDI in low-income cities. This regional differential effect opens up new prospects for less developed cities, and the strategic push for digital development can be a long-term growth engine for lower-income cities. From a policy perspective, these findings argue for focused government investment in digital infrastructure, especially in less developed cities. Priority areas include broadband networks, data centers, and smart city technologies, creating a conducive business environment for attracting foreign investors. Alongside infrastructure development, it is critical to develop a digitally skilled workforce through education and training. This strategy would enhance business services in less developed cities, improving their attractiveness to foreign investments.

For low-income cities, a strategic emphasis on improving digitization levels is recommended as a key approach to attracting foreign investment. Governments should actively promote telecommunications services and widespread mobile phone adoption, supported by enabling policies and financial incentives. Such efforts are expected to yield a favorable investment environment that can help attract FDI inflows.

Although some meaningful results about digitalization in modeling Chinese cities at the digital level were obtained in this study, future research should be conducted to create a more robust analysis from different perspectives. If various data sources are used, more detailed information will become available. Moreover, expanding the study to include a broader geographic scope beyond China could help ascertain the generalizability of our findings and understand regional variations in the impact of digitalization on FDI.

Authors contributions

This research paper greatly benefited from the contributions of all three authors. Dr. Tajul Ariffin Masron played a key role in conceptualization, research question formulation, and experimental design. Dansha Zhang’s expertise enhanced data collection, analysis, and interpretation, she also critically revises the paper for intellectual content. Xiutuan Lu contributed significantly to the theoretical framework, literature review, and critical insights. All authors accept accountabilities for all aspects of this work.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The datasets used and analyzed during the current study are available from the corresponding author upon reasonable request.

Additional information

Notes on contributors

Dansha Zhang

Dansha Zhang is currently a PhD student at School of Management, Universiti Sains Malaysia, Malaysia.

Tajul Ariffin Masron

Tajul Ariffin Masron is an Associate Professor at School of Management, Universiti Sains Malaysia, Malaysia.

Xiutuan Lu

Xiutuan Lu is a lecturer at Naning College for Vocational Technology, China.

Notes

1 In other words, less FDI-receiving areas tend to develop less than high FDI-receiving areas, leading to bigger income gap.

2 Other potential benefits of FDI are creating a more competitive domestic market (Barrios et al., 2005; Ishikawa et al., 2010), bringing a greener technology (Amendolagine et al., 2021; Song et al., 2015), spurring local innovation (Wang & Wu, 2016), improvement in host-country institutions (Kwok & Tadesse, 2006).

3 For example, the internet penetration index it follows: $\mathit{\text{INT}}=\left(\frac{{\mathit{\text{INT}}}_{\mathit{\text{city}}}-{\mathit{\text{INT}}}_{\mathit{\text{lowest}}}}{{\mathit{\text{INT}}}_{\mathit{\text{highest}}}-{\mathit{\text{INT}}}_{\mathit{\text{lowest}}}}\right)\times 100$.

4 The following equations represent the models introduced squared and cubic terms of the digitalization variable. The interpretation of the model symbols is consistent with Equation (3)(3) $${\mathit{\text{FDI}}}_{i,t}={\beta }_0+\gamma {\mathit{\text{FDI}}}_{i,t-1}+\alpha {\mathit{\text{DIG}}}_{i,t}+\beta X_{i,t}+{\lambda }_t+{\delta }_i+{\mu }_{i,t}$$(3) :

${\mathit{\text{FDI}}}_{i,t}={\beta }_0+\gamma {\mathit{\text{FDI}}}_{i,t-1}+{\alpha }_1{\mathit{\text{DIG}}}_{i,t}+{\alpha }_2{\mathit{\text{DIG}}}^2{}_{i,t}+\beta X_{i,t}+{\lambda }_t+{\delta }_i+{\mu }_{i,t}$ and ${\mathit{\text{FDI}}}_{i,t}={\beta }_0+\gamma {\mathit{\text{FDI}}}_{i,t-1}+{\alpha }_1{\mathit{\text{DIG}}}_{i,t}+{\alpha }_2{\mathit{\text{DIG}}}^3{}_{i,t}+\beta X_{i,t}+{\lambda }_t+{\delta }_i+{\mu }_{i,t}$