Digitizable Product Trade Development and Carbon Emission: Evidence from 94 Countries

Abstract

In the face of increasingly severe climate change and its destructive effects, how to effectively reduce carbon dioxide emissions has become a challenging task. Developing a digital economy provides opportunities for countries to reduce pollution and carbon emissions and reach a goal of carbon neutrality. As an emerging trade form, digitizable product trade is of great significance to promoting economic growth and carbon emission reduction. This paper selects panel data for 94 countries from 2001 to 2019 and adopts the STIRPAT model to analyze the impact effect and impact mechanism of digitizable product trade on carbon emissions. Research results show that developing digitizable product trade will help countries reduce carbon emissions. The conclusion is robust by replacing the explained variable and core explanatory variable. The carbon emission reduction effect has heterogeneity due to differentiated national income levels and product categories. Mechanism analysis shows that digitizable product trade reduces carbon emissions through the technology effect. Our analysis indicates that countries developing digital trade and digital technology and actively responding to environmental issues have a greater chance of reduced carbon emissions.

1. Introduction

In recent years, global carbon dioxide emissions continue to rise, which intensifies the trend of global warming and seriously affects the survival and development of humans [1,2]. From 2001 to 2019, total global CO₂ emissions have increased by 63.92% and per capita CO₂ emissions have increased by 12.82% (see Figure 1). The increase in extreme weather, melting glaciers, rising sea level, and reduction in food production caused by excessive greenhouse gas emissions will endanger human security and development. With a global agenda for making our planet greener and more sustainable [3], the European Union and national governments have set clear objectives to reduce carbon emissions [4]. Other major economies, including the USA and China, are working hard to reduce carbon emissions [5].

As for the determinants of carbon emissions, international trade has drawn particular attention. The role of international trade in carbon emissions is widely concerned by scholars [6,7,8,9]. However, the conclusions of the relationship between international trade and carbon emissions are not consistent. Some scholars report a positive impact of international trade on carbon emissions [10,11,12], they posit that the increase in resource consumption generally results in trade scale expansion and increased carbon emissions. While Shahbaz, Lean, and Shabbir and other scholars confirm the reduction role of international trade in carbon emissions [6,13,14,15,16]. Contrary to the above studies, Iwata, Okada, and Samreth believe that the impact of trade on carbon dioxide emissions is not significant [17].

These studies mainly focus on the role of traditional goods trade or services trade in carbon emissions. Few studies focus on the impact of new trade forms, such as digital trade, on carbon emissions. With the development of digital technology, increasingly more categories of traditional trade are evolving into digital trade; digitizable product trade is one of the typical. Digitizable products refer to products that are traded in both physical form and online form, such as music, electronic books, software, etc. [18]. The online trade and trade of digitizable products that are physically delivered but ordered via the internet are classified as digital trade [19]. With the spread of ICT, the more significant part of digitizable products will realize online delivery or online order; as a result, digitizable products trade has become or is transforming into digital trade. Owing to the difference in transaction mode with traditional trade, the impact of digitizable product trade on carbon emissions may exhibit new characteristics.

This paper contributes to the current literature in the following aspects: First, different from most literature that select traditional goods trade or service trade as a research focus, this paper focuses on the digitizable product trade and studies its effect on carbon emissions, which will be more helpful in revealing the impact of new trade patterns on the environment. Second, different from previous literature that focus on the overall impact effect of trade on carbon emissions, our study explores the heterogeneous effect of trade on carbon emissions by distinguishing different countries and different digitizable product groups. Additionally, this paper also discusses the impact mechanism of digitizable product trade on carbon emissions. Third, our empirical findings offer practical and up-to-date insights for policymakers and trade practitioners, and play a reference role in trade and environmental policy development in the digital and green economy era.

The remainder of the paper is organized as follows: Section 2 presents a literature review. Section 3 is the theoretical analysis that investigates the possible impact of digitizable product trade on carbon emissions and puts forward relevant hypotheses. Section 4 introduces the methodology and data source. Section 5 presents the empirical results, including the baseline regression results, endogeneity problem, and robustness check. Section 6 extends the analysis by carrying out heterogeneity analysis and mechanism analysis. The main findings are summarized in the discussion section. The conclusion part puts forward policy implications, discusses the limitations, and presents potential research directions.

2. Literature Review

A large amount of literature focuses on the impact of international trade on carbon emissions. The literature can be divided into three categories according to research findings: literature showing that trade promotes carbon emissions, literature positing that trade reduces carbon emissions, and literature indicating that trade has bidirectional or insignificant impact on carbon emissions.

2.1. Trade Promotes Carbon Emissions

Anderson [20] focuses on trade liberalization, environmental institution, legal and property rights, institutional risk, and exchange rate on carbon dioxide embodied in Chinese exports to developed countries from 1995 to 2008. Results indicate that trade positively affects carbon emissions. Sun et al. [21] investigate the role of service trade globalization on energy and carbon emission performance of 30 countries during the period 1980 to 2013. Results show that service trade openness has a positive and intensified effect on energy and carbon emission efficiency with time. Al-mulali and Sheau-Ting [12] find a positive linkage between trade and carbon emissions in six study regions. The positive impact of trade on carbon emissions also exists in most MENA countries. In 11 Eastern Asian countries from 1998 to 2011, Zhang [6] finds that exports of intermediate and final goods positively affect carbon emissions. Shahzad et al. [11] choose carbon emissions and trade openness of Pakistan as a research focus and conclude that one percent increase in trade results in 0.122 percent and 0.247 percent increase in carbon emissions in the short run and long run, respectively. The study of Liu [22] shows that an increase in trade openness or exports would enhance per capita carbon emissions. The research of Hossain [8]; Jayanthakumaran, Verma, and Liu [10]; Aghasafari [23]; and Al-mulali and Sheau-Ting [12] also confirm the promotional role of trade on carbon emissions.

2.2. Trade Reduces Carbon Emissions

Paramati et al. [24] investigate the role of financial deepening, green technology, and trade openness on carbon emissions using panel data from 25 OECD countries. They find that green technology, FDI, and trade are significant factors that help to reduce carbon emissions. Considering the dominant role in global carbon emissions, Hu et al. [25] select panel data from 25 major developing countries during 1996 to 2012 to explore the role of commercial service trade in generating carbon emissions. They find that expanding commercial service trade is helpful in reducing carbon emissions and promoting low-carbon economic growth. Zhang and Zhang (2018) [26] examine the impact of GDP, trade structure, exchange rate, and FDI inflows on China’s carbon emissions from 1982 to 2016. The research verifies the validity of EKC and the negative impact of service trade on carbon emissions. A study of South Africa also finds that trade reduces the growth of energy pollutants to improve environmental quality [14]. Leitao and Balogh [27] find that agriculture intra-industry trade has a diverse impact on carbon emissions for EU countries for the period 2000–2014, which is helpful for the EU to achieve the reduction target by 20% in 2020. Using provincial data from China, Ma et al. [28] find a negative impact of digital economy and exports on consumption-based carbon emissions. The negative effect of trade on consumption-based carbon emissions is also confirmed in the research [29] of 29 high-income countries during the period 1991 to 2008.

2.3. Trade Has a Bidirectional or Insignificant Impact on Carbon Emissions

Iwata, Okada, and Samreth [17] investigate the EKC for carbon emissions in 11 OECD countries by considering the role of trade. The results show that the impact of trade on carbon emissions is not significant. Adopting panel data from 49 high-emission countries in the Belt and Road regions, Sun et al. [30] show that trade has both positive and negative impacts on carbon emissions, but the impact varies in different country groups. Additionally, the research once again confirms the existence of EKC. Liddle [31] adopts a new consumption-based carbon emission database to test the importance of trade in national emissions. The research finds that trade has a significant role in consumption-based emissions but an insignificant role in territory-based consumptions. Further, exports decrease but imports increase consumption-based carbon emissions. In 55 middle-income countries, trade has a soothing effect on the environment in the short run but a harmful effect in the long run [32]. Shahbaza, Lean, and Shabbir [13] show that trade reduces carbon emissions in Pakistan in the long run, but in the short run, the impact of trade on carbon emissions is insignificant. For CIS countries, trade increases carbon emissions directly but indirectly decreases carbon emissions due to its negative effect on per capita income. The research of Pata [33] indicates that import has a positive impact on carbon emissions, while export decreases carbon emissions in the long run.

These studies show the impact effect of traditional trade on carbon emissions and reveal the positive, negative, bidirectional, or insignificant role of trade in environmental protection. However, few studies refer to the role of digitizable product trade in affecting carbon emissions. As a special trade form, digitizable product trade combines characteristics of both traditional trade and digital trade. In the future, with ICT development, digitizable product trade will gradually evolve into digital trade. In this process, transition from physical products to virtual products and realization of online delivery will produce a direct effect on carbon emissions. More digitizable product trade can also cause a change in traditional trade, hence affecting carbon emissions indirectly. In general, the impact of digitizable product trade on carbon emissions may present new features, and it is worthy to study the subject to bridge the relevant research gap.

3. Theoretical Analysis

As a unique trade form, digitizable product trade may have dual impacts on carbon emissions. On the one hand, the development of digitizable product trade relies on the support of well-developed digital infrastructure. The construction and operation of relevant infrastructure will produce a massive amount of carbon emissions. For example, operation of a 5G base station results in significant power consumption. Maintenance of a data center requires extensive cooling facilities in addition to consuming massive power; on the other hand, with further development of digitizable product trade, increasingly more tangible trade products such as photographic films, prints, sound and media, and software and video games, will acquire an intangible form to realize online delivery. In this process, digitizable product trade not only reduces carbon emissions in the production process, but also eliminates the dependence on logistics and transportation in the traditional delivery process, so as to promote carbon emission reduction. Given the possibility of these positive and negative impacts on carbon emissions caused by digitizable product trade, the following hypotheses are put forward:

Hypothesis 1.

Development of digitizable product trade will increase carbon emissions.

Hypothesis 2.

Development of digitizable product trade will reduce carbon emissions.

Grossman and Krueger [34] put forward that trade is one crucial factor that impacts environmental quality through the effects of scale, composition, and technique. With the expansion of digitizable product trade, more production factor inputs and energy resource consumption are required. The growth of economic activity will result in an increase in carbon dioxide emissions. As a result, digitizable product trade may affect carbon emissions through the scale effect; the composition effect of trade on the environment is ambiguous, depending on the countries’ comparative advantage. By introducing and absorbing foreign advanced digital services and production technologies, digitizable product trade will promote the modernization and intelligent and independent innovation of domestic enterprises and industries, but the promotion role is affected by the country’s industrial structure. If the service sector dominates, digitizable product trade is more likely to reduce carbon emissions through the composition effect. Otherwise, it may promote carbon emissions if the primary industry type in one country lies in the secondary industry. The technique effect refers to the effect that trade reduces carbon emissions by transferring modern and clean technologies to the local economy. Digitizable product trade represents a deep integration of information technology and traditional trade. Compared with traditional trade, digitizable product trade presents more robust technical attributes and can play the technical effect on carbon emission reduction by upgrading low-carbon technologies and reducing unit energy carbon emissions. Digitizable product trade may also produce a technology spillover effect and contribute to clean economy development. As a result, our work puts forward the following hypotheses:

Hypothesis 3.

Digitizable product trade will affect carbon emissions through scale effect.

Hypothesis 4.

Digitizable product trade will affect carbon emissions through composition effect.

Hypothesis 5.

Digitizable product trade will affect carbon emissions through technique effect.

Countries with different income levels differ in digitizable product trade development, carbon emission stage, internet development level, industrial structure, and technical level. Therefore, the impact of digitizable product trade on carbon emissions may have heterogeneity in different countries. High-income countries have entered a post-industrialization stage; their energy-saving technologies and clean and efficient production methods are in a leading position in the world. As a result, their carbon emission reduction potential may be lower than that for middle- and low-income countries. Middle- and low-income countries are committed to developing digital trade and taking measures to promote digital trade development. As a result, the transformation from traditional trade to digital trade is more likely to reduce carbon emissions in these countries.

Digitizable product trade may have different impacts on carbon emissions due to different product categories. Digital transformation, delivery, and presentation of some categories such as printed matter products may significantly reduce the consumption of paper and ink, and then significantly reduce the energy consumption in the trade process. Some categories, such as software products, are easier to digitize. As a result, trade in these products is more likely to reduce carbon emissions. Compared with the above trade categories, trade of other product categories may be limited by countries’ conservative attitude on intellectual property protection, national security, and other issues, and their carbon emissions effect may present different characteristics. Given the differentiated income level and product categories, our paper puts forward the following hypothesis:

Hypothesis 6.

The impact of digitizable product trade on carbon emissions is heterogeneous due to differentiated countries and product categories.

4. Methodology and Data Source

4.1. Econometric Model Specification

The STIRPAT model [32,35], which is widely used to analyze the impact of trade on the environment, is adopted to investigate the impact of digitizable product trade on carbon emissions. Considering specific characteristics of digitizable product trade, the empirical model is constructed as follows:

$${\mathrm{lnCO}}_{2\mathrm{it}}={\mathsf{\delta }}_0+{\mathsf{\beta }}_0{\mathrm{export}}_{\mathrm{it}}+{\mathsf{\gamma }\mathrm{X}}^{\prime }{}_{\mathrm{it}}+{\mathsf{\omega }}_{\mathrm{i}}+{\mathsf{\omega }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{it}}$$

(1)

where $\mathrm{i}$ represents the country, $\mathrm{t}$ is the year, and ${\mathsf{\delta }}_0$ is a constant. The explained variable ${\mathrm{CO}}_{2\mathrm{it}}$ denotes carbon dioxide emission; ${\mathrm{export}}_{\mathrm{it}}$ is the core explanatory variable, representing development of digitizable product trade. ${\mathrm{X}}^{\prime }{}_{\mathrm{it}}$ is a control variable consisting of variables that represent economic development, industrial structure, technique level, financial development, foreign direct investment, and urbanization level. ${\mathrm{lnGDP}}_{\mathrm{it}}$ and ${\mathrm{lnGDP}}_{\mathrm{it}}^2$ are adapted to proxy economic development. $\mathrm{indu}$, $\mathrm{tech}$, $\mathrm{fin}$, $\mathrm{fdi},$ and $\mathrm{urban}$ represent industrial structure, technique level, financial development, foreign direct investment and urbanization level, respectively. ${\mathsf{\omega }}_{\mathrm{i}}$ and ${\mathsf{\omega }}_{\mathrm{t}}$ represent country-fixed effects and year-fixed effects, respectively, which are used to control possible impact of systematic risks and time-invariant national level factors on carbon emissions. ${\mathsf{\epsilon }}_{\mathrm{it}}$ is random error term.

According to theoretical expectation, digitizable product trade may produce positive or negative effects on carbon emissions. As a result, the coefficient of ${\mathsf{\beta }}_0$ is uncertain. If the environmental Kuznets curve exists, the impact of ${\mathrm{lnGDP}}_{\mathrm{it}}$ on carbon emissions is usually positive, but the impact of ${\mathrm{lnGDP}}_{\mathrm{it}}^2$ is expected to be negative. The impact of industrial structure on carbon emissions is uncertain. Technique development may reduce carbon emissions in the long run, but in the short run it may increase carbon emissions due to large-scale scientific research investment and scientific research facility construction. Financial development may induce greater carbon emissions due to its role in attracting FDI and promoting consumer credit. Foreign direct investment will stimulate the production and export of industrial products to increase carbon emissions. Urbanization generally promotes carbon emissions.

4.2. Variable and Data Source

The explained variable ${\mathrm{lnCO}}_2$ is measured by carbon dioxide emissions per capita. Relevant data are from the IEA database. Following Banga [18], a total of 49 products of HS six-digit code (see Appendix A) is selected as digitizable product and ratio of digitizable product export in commodity export is used to measure digitizable product trade development. Considering different economic development level, geographical locations, and data availability, 94 countries or regions are selected as sample countries (see Appendix B). These samples cover economies in different development stages and countries located on different continents, which can serve as representative samples. Relevant data are from WITS. Economic development is measured by per capita GDP and the square of per capita GDP. The data are from the World Bank. Our work adopts data for energy intensity level of primary energy to measure technology development, and the data are from the United Nations. Industrial structure, financial development, foreign direct investment, and urbanization are measured by industry value added as a percent of GDP, broad money as a percent of GDP, outward foreign direct investment as a percent of GDP, and urban population as a percent of total population, respectively. The data listed above all come from the World Bank. A detailed description of main variables is listed in

Table 1. Variable description and data sources.

Variable	Description	Source
${\mathrm{lnCO}}_2$	Logarithm of carbon dioxide emissions per capita	IEA
$\mathrm{export}$	Percentage of digitizable goods export in goods export	WITS
$\mathrm{lngdp}$	Logarithm of GDP per capita	World Bank
${\mathrm{lngdp}}^2$	Logarithm of squared GDP per capita	World Bank
$\mathrm{tech}$	Energy intensity measured in terms of primary energy and GDP	United Nations
$\mathrm{indu}$	Industry, value added (of GDP)	World Bank
$\mathrm{fin}$	Broad money as a percent of GDP	World Bank
$\mathrm{urban}$	Urban population (% of total population)	World Bank
$\mathrm{fdi}$	Outward foreign direct investment, net inflows (% of GDP)	World Bank

.

5. Empirical Results

5.1. Baseline Results

Using the panel random-effects regression method, this paper presents the empirical results in

Table 2. Baseline regression results.

	(1)	(2)
	$\mathbf{l}\mathbf{n}\mathbf{C}{\mathbf{O}}_2$	$\mathbf{l}\mathbf{n}\mathbf{C}{\mathbf{O}}_2$
$\mathrm{export}$	−1.139 ***	−0.399 **
	(−4.62)	(−1.97)
$\mathrm{lngdp}$		1.223 ***
		(16.29)
${\mathrm{lngdp}}^2$		−0.057 ***
		(−12.36)
$\mathrm{tech}$		0.341 ***
		(9.06)
$\mathrm{indu}$		0.003 ***
		(2.97)
$\mathrm{fin}$		0.003 ***
		(8.14)
$\mathrm{fdi}$		0.003 ***
		(3.8)
$\mathrm{urban}$		0.015 ***
		(7.28)
Year	Yes	Yes
Country	Yes	Yes
N	1786	1320
R2	0.0569	0.4064

t value in parentheses, ** p < 0.05, *** p < 0.01.

. Column (1) shows the impact of digitizable product trade on carbon emissions. The results indicate that digitizable product trade negatively and significantly affects carbon emissions, suggesting that digitizable product trade is helpful in reducing carbon emissions. Column (2) shows regression results by adding all control variables; the conclusion that digitizable product export negatively and significantly affects carbon emissions still exists, showing the carbon emission reduction role of digital trade development. As a result, Hypothesis 1 is rejected and Hypothesis 2 is confirmed. Coefficients of GDP and squared GDP confirm the existence of the environmental Kuznets curve. The larger share of industry value added to GDP induces higher carbon emissions. Technology development promotes carbon emissions. Finance deepening and urbanization also increase carbon emissions.

5.2. Endogeneity Problem

This paper adopts the IV method to address the possible endogeneity problem. An ideal instrument variable is correlated with the endogenous explanatory variable but uncorrelated with the explained variable. Considering the ICT-enabled characteristic of digital trade and following the practice of Wang et al. [36], the logarithmic form of fixed telephone subscriptions from 1982 to 2000 is adopted as an instrument variable. On one hand, digitizable product trade highly depends on ICT development, which originates to a large extent from fixed telephone dial-up internet access. Therefore, countries with a large number of fixed telephone subscriptions in the early stage tend to have relatively developed ICT technologies, thus providing good conditions for developing digital trade. On the other hand, the early-stage fixed telephone subscriptions do not affect carbon emissions in the current period. As a result, fixed telephone subscriptions can serve as an ideal instrument variable. Apart from that, we also select one period lag of export as an instrument variable. The two instrument variables described above are placed into the equation simultaneously, and the 2SLS regression results are shown in

Table 3. Regression results with instrument variables.

	$\mathbf{l}\mathbf{n}\mathbf{C}{\mathbf{O}}_2$
$\mathrm{export}$	−2.253 ***
	(−3.91)
First-stage regression results
$\mathrm{lntelephone}$	0.002 ***
	(3.1)
$\mathrm{l}.\mathrm{export}$	0.809 ***
	(7.87)
Ctrls	Yes
Year	Yes
Country	Yes
Kleibergen–Paap	36.69
N	1245
R2	0.6791

Note: l.export is one-period lag of export; t value in parentheses, *** p < 0.01.

.

The first-stage regression results indicate that $\ln \mathrm{telephone}$ and $\mathrm{l}.\mathrm{export}$ significantly and positively affect digitizable product export. In the second-stage regression, the coefficient of $\mathrm{export}$ is still negative and significant, showing that digitizable product trade development reduces carbon emissions. The Kleibergen–Paap value is larger than 10% maximal IV size value in the weak identification test, indicating the appropriate use of instrumental variables. In general, the regression results show that digitizable product export negatively and significantly affects carbon emissions, which is consistent with the main conclusion of our study.

5.3. Robustness Check

To test the robustness of the model, this paper replaces the core explanatory variable with the proportion of digitizable product export in imports and exports and the proportion of digitizable product export in GDP and denote the two variables as $\mathrm{export}2$ and $\mathrm{export}3$, respectively. Regression results are shown in columns (1) and (2) of

Table 4. Robustness test results.

	(1)	(2)	(3)
	$\mathbf{l}\mathbf{n}\mathbf{C}{\mathbf{O}}_2$	$\mathbf{l}\mathbf{n}\mathbf{C}{\mathbf{O}}_2$	$\mathbf{l}\mathbf{n}\mathbf{t}\mathbf{o}\mathbf{t}\mathbf{a}\mathbf{l}\mathbf{C}{\mathbf{O}}_2$
$\mathrm{export}2$	−1.033 **
	(−2.15)
$\mathrm{export}3$		−2.921 **
		(−2.33)
$\mathrm{export}$			−0.623 ***
			(−2.76)
Ctrls	Yes	Yes	Yes
Year	Yes	Yes	Yes
Country	Yes	Yes	Yes
N	1320	1320	1320
R2	0.4067	0.4071	0.5642

t value in parentheses, ** p < 0.05, *** p < 0.01.

. Our work also replaces the explained variable with total amount of CO₂ emissions (kt) and denotes it as ${\mathrm{totalCO}}_2$. The logarithmic form of this variable is adopted and the regression results are reported in column (3) of Table 4. Columns (1) and (2) show that $\mathrm{export}2$ and $\mathrm{export}3$ affect carbon emissions negatively and significantly. Column (3) still supports the conclusion that digitizable product trade development reduces the total carbon emissions. The main conclusion that digitizable product trade helps reduce carbon emissions is still valid after replacing the core explanatory variable and the explained variable, indicating that the results are robust.

6. Extended Analysis

6.1. Heterogeneity Analysis

The impact of digitizable product trade on carbon emissions may be heterogeneous due to different national income levels; as a result, our work divides the sample countries into different income groups. Following the thresholds of the World Bank, this paper classifies countries with per capita GNI more than USD 12,695 as high-income countries (see Appendix B) and other countries as middle- and low-income countries. The regression results are shown in

Table 5. Regression results for high-, middle-, and low-income countries.

	(1)	(2)
	$\mathbf{l}\mathbf{n}\mathbf{C}{\mathbf{O}}_2$	$\mathbf{l}\mathbf{n}\mathbf{C}{\mathbf{O}}_2$
High-income countries	9.510 ***
	(4.37)
Middle- and low-income countries		−0.397 *
		(−1.9)
Ctrls	Yes	Yes
Year	Yes	Yes
Country	Yes	Yes
N	396	924
R2	0.4927	0.4640

t value in parentheses, * p < 0.1, *** p < 0.01.

. Column (1) indicates that in high-income countries, the development of digitizable product trade increases carbon emissions and the promotion role is significant. Opposite to high-income countries, middle- and low-income countries achieve carbon emission reduction targets by developing digitizable product trade. A possible reason is that digitizable product trade of high-income countries has been highly digitized, thus weakening the carbon emission reduction effect brought by the transformation from tangible trade to digital trade. In the sample period, carbon emissions mainly come from the expansion of trade scale. For middle- and low-income countries, digitizable product trade gradually evolves into digital trade and online delivery, replacing physical delivery. The change in trade mode reduces carbon emissions.

Apart from that, the impact of digitizable product trade on carbon emissions may be differentiated due to different product categories. Our work distinguishes digitizable products into photographic plates and film, printed matter, sound and media, and software and video games, and reports the regression results in

Table 6. Regression results for differentiated product categories.

	(1)	(2)	(3)	(4)	(5)
	$\mathbf{l}\mathbf{n}\mathbf{C}{\mathbf{O}}_2$	$\mathbf{l}\mathbf{n}\mathbf{C}{\mathbf{O}}_2$	$\mathbf{l}\mathbf{n}\mathbf{C}{\mathbf{O}}_2$	$\mathbf{l}\mathbf{n}\mathbf{C}{\mathbf{O}}_2$	$\mathbf{l}\mathbf{n}\mathbf{C}{\mathbf{O}}_2$
Photographic plates and film	1.814 ***
	(3.38)
Printed matter		−0.761 **
		(−3.09)
Sound and media			−7.707 **
			(−2.04)
Software				−9.196 *
				(−1.9)
Video Games					3.449
					(1.42)
Ctrls	Yes	Yes	Yes	Yes	Yes
Year	Yes	Yes	Yes	Yes	Yes
Country	Yes	Yes	Yes	Yes	Yes
N	1000	1263	1154	1070	1247
R2	0.4165	0.4131	0.3977	0.4398	0.3973

t value in parentheses, * p < 0.1, ** p < 0.05, *** p < 0.01.

. Column (1) shows that photographic plates and film product trade have positive and significant effect on carbon emissions. Before digital film completely replaces it, the production and use of photographic plates and film will still promote carbon emissions. In column (5), the coefficient shows that video games product trade has positive impact on carbon emissions, but the effect is not significant. Coefficients in column (2) to column (4) demonstrate the negative and significant impact of printed matter, sound and media, and software product trade on carbon emissions. These product categories are easier to digitize. As a result, trade of these products is more helpful in reducing carbon emissions. This analysis confirms the existence of Hypothesis 6.

6.2. Mechanism Analysis

In this part, the interaction term is added to the baseline regression model to analyze the impact mechanism of digitizable product trade on carbon emissions. The following equations are constructed:

$${\mathrm{lnCO}}_{2\mathrm{it}}={\mathsf{\delta }}_0+{\mathsf{\beta }}_0{\mathrm{export}}_{\mathrm{it}}+{\mathsf{\gamma }\mathrm{X}}^{\prime }{}_{\mathrm{it}}+{\mathsf{\beta }}_1{\mathrm{export}}_{\mathrm{it}}{\ast \mathrm{lnGDP}}_{\mathrm{it}}+{\mathsf{\omega }}_{\mathrm{i}}+{\mathsf{\omega }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{it}}$$

(2)

$${\mathrm{lnCO}}_{2\mathrm{it}}={\mathsf{\delta }}_0+{\mathsf{\beta }}_0{\mathrm{export}}_{\mathrm{it}}+{\mathsf{\gamma }\mathrm{X}}^{\prime }{}_{\mathrm{it}}+{\mathsf{\beta }}_1{\mathrm{export}}_{\mathrm{it}}{\ast \mathrm{indu}}_{\mathrm{it}}+{\mathsf{\omega }}_{\mathrm{i}}+{\mathsf{\omega }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{it}}$$

(3)

$${\mathrm{lnCO}}_{2\mathrm{it}}={\mathsf{\delta }}_0+{\mathsf{\beta }}_0{\mathrm{export}}_{\mathrm{it}}+{\mathsf{\gamma }\mathrm{X}}^{\prime }{}_{\mathrm{it}}+{\mathsf{\beta }}_1{\mathrm{export}}_{\mathrm{it}}{\ast \mathrm{tech}}_{\mathrm{it}}+{\mathsf{\omega }}_{\mathrm{i}}+{\mathsf{\omega }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{it}}$$

(4)

In the three equations above, ${\mathrm{export}}_{\mathrm{it}}{\ast \mathrm{lnGDP}}_{\mathrm{it}}$, ${\mathrm{export}}_{\mathrm{it}}{\ast \mathrm{indu}}_{\mathrm{it}}$, and ${\mathrm{export}}_{\mathrm{it}}{\ast \mathrm{tech}}_{\mathrm{it}}$ measure the interaction term of digitizable product export and economic scale; digitizable product export and industrial structure; and digitizable product export and technology development, respectively. The interaction term captures the impact mechanism that trade affects carbon emissions through scale, composition, and technology effects. The regression results are listed in

Table 7. The impact mechanism of digitizable product trade on carbon emissions.

	(1)	(2)	(3)
	$\mathbf{l}\mathbf{n}\mathbf{C}{\mathbf{O}}_2$	$\mathbf{l}\mathbf{n}\mathbf{C}{\mathbf{O}}_2$	$\mathbf{l}\mathbf{n}\mathbf{C}{\mathbf{O}}_2$
$\mathrm{export}$	−6.254 ***	−1.327	−0.340 *
	(−4.19)	(−1.25)	(−1.68)
${\mathrm{export}}_{\mathrm{it}}{\ast \mathrm{lnGDP}}_{\mathrm{it}}$	0.810 ***
	(3.96)
${\mathrm{export}}_{\mathrm{it}}{\ast \mathrm{indu}}_{\mathrm{it}}$		0.037
		(0.89)
${\mathrm{export}}_{\mathrm{it}}{\ast \mathrm{tech}}_{\mathrm{it}}$			−2.710 *
			(−3.10)
Ctrls	Yes	Yes	Yes
Year	Yes	Yes	Yes
Country	Yes	Yes	Yes
N	1300	1320	1320
R2	0.4092	0.4067	0.4110

t value in parentheses, * p < 0.1, *** p < 0.01.

.

Column (1) shows the impact of digitizable product export on carbon emissions through the scale effect. The coefficient of $\mathrm{export}$ is negative and significant, showing that digitizable product trade decreases carbon emissions. The coefficient of ${\mathrm{export}}_{\mathrm{it}}{\ast \mathrm{lnGDP}}_{\mathrm{it}}$ is positive and significant, indicating that digitizable products promote carbon emissions through the scale effect. Hypothesis 3 is confirmed. In column (2), the impact of $\mathrm{export}$ and ${\mathrm{export}}_{\mathrm{it}}{\ast \mathrm{indu}}_{\mathrm{it}}$ on carbon emissions is negative but insignificant, indicating that the path of digitizable product trade influencing carbon emissions through the composition effect does not exist. As a result, Hypothesis 4 does not exist. The coefficients of $\mathrm{export}$ and ${\mathrm{export}}_{\mathrm{it}}{\ast \mathrm{tech}}_{\mathrm{it}}$ are negative and significant, which shows that digitizable product trade reduces carbon emissions through the technique effect channel. The results above confirm the existence of Hypothesis 5.

7. Discussion

In this paper, we focus on a new trade pattern and explore the impact effect and impact mechanism of digitizable product trade on carbon emissions. The regression results show that digitizable product trade has a significant inhibitory effect on carbon emissions. The conclusion is still robust after replacing the core explanatory variable and the explained variable. The impact of digitizable product trade on carbon emissions has heterogeneity due to differentiated national income levels and product categories. Promoting digitizable product trade in middle- and low-income countries is more conducive to carbon emission reduction. Exporting more printed matter products, sound and media products, and software products make the carbon emission reduction effect more pronounced. Mechanism analysis shows that digitizable product export helps reduce carbon emissions through the technique effect. The main findings are summarized in

Table 8. A brief description of main findings.

Empirical Analysis	Variable Description	Impact of Digitizable Product Trade on Carbon Emissions
Baseline regression	Digitizable product trade as the sole explanatory variable	Negative and significant
	Core explanatory variable and control variables	Negative and significant
Endogeneity problem	Fixed-telephone subscription and one-period lag of the core explanatory variable as instrument variables	Negative and significant
Robustness check	Proportion of digitizable product export in imports and exports as core explanatory variable	Negative and significant
	Proportion of digitizable product export in GDP as core explanatory variable	Negative and significant
	Total amount of CO2 emissions as the explained variable	Negative and significant
Heterogeneity analysis	High-income countries	Positive and significant
	Middle- and low-income countries	Negative and significant
	Photographic plates and film products	Positive and significant
	Printed matter products	Negative and significant
	Sound and media products	Negative and significant
	Software products	Negative and significant
	Video games products	Positive but insignificant
Empirical analysis	Variable description	Impact of interactive term on carbon emissions
Mechanism analysis	Scale effect	Positive and significant
	Composition effect	Positive but insignificant
	Technology effect	Negative and significant

. Our finding supports the existing research viewpoint that trade helps reduce carbon emissions (Paramati et al. [24]; Hu et al. [25]; Zhang and Zhang [26]; Leitao and Balogh [27]; Ma et al. [28]). However, our work also enriches the current literature by contributing more in-depth research on the impact that heterogeneity and mechanism of trade has on carbon emissions.

8. Conclusions

The empirical findings provide beneficial enlightenment for trade development and environmental protection. First, the negative and significant impact of digitizable product trade on carbon emissions indicates countries that develop digital trade and actively respond to environmental issues have a greater chance for reduced carbon emissions. Relevant countries may take the development of digital trade as a way to reduce carbon emissions. Second, heterogeneity analysis shows that the carbon emission reduction role of digitizable product trade is more obvious for middle- and low-income countries. As a result, middle- and low-income countries committed to carbon reduction may seize the opportunity for digital economy development and vigorously develop digital trade. Heterogeneity analysis also indicates that trade of printed matter products, sound and media products, and software products is more helpful in reducing carbon emissions. It is suggested that countries pay appropriate attention to the product trade development described above, and prioritize the development of relevant sectors to reduce carbon emissions. Third, considering that digitizable product trade reduces carbon emissions through the technique effect, firms are encouraged to adopt low-carbon technologies in production and countries are suggested to increase technical competence, to reduce the overall carbon emissions.

Our research still has limitations, which may also be future directions for follow-up research. First, adopting carbon dioxide emissions per capita or total amount of carbon emissions is unable to comprehensively measure carbon emissions. Constructing a composite carbon emission index is more appropriate. Second, the country-level analysis reported in this paper can be expanded and deepened to the industry level. Finally, due to data availability, our research just covers trade data for 49 products in 94 countries. With the progress of modern information and technology, increasingly more products may realize online delivery or online order; therefore, the scope of digitizable products will be further expanded. Additionally, a larger number of countries will be able to participate in digital trade. As a result, future research may follow up with data updates and reach more comprehensive and in-depth conclusions on the role of digital trade in carbon emissions.