A Survival Analysis of China’s Trade of Environmental Goods

Abstract

Based on data at the HS-6 digit level, this paper uses survival analysis to present characteristics concerning the survival of China’s exports and imports of environmental goods during the period 2002–2020, and compares them with those of other major exporting and importing countries. The major findings are summarized as follows: For exports, the average length of all trade spells was 8.36 years, and the length of only 18.72% of them was more than 18 years, which shows a disadvantage compared with other countries. The hazard rate was 0.31 in the first year of export, and the export survival rate of China’s environmental goods was always at the highest level compared with the others. By category, the export survival of environmental monitoring and evaluation goods was the best, and the export survival of renewable energy goods was the worst. For imports, the average length of all trade spells was 6.08 years, and the length of 50.37% of them was equal to 1 year, which are worse results than those of other countries. The hazard rate was up to 0.48 in the first year of imports, which is higher than those of other countries. By category, the environmental monitoring and evaluation goods had worse early import survival and better long-term survival, whereas the renewable energy goods experienced the opposite. Whether for exports or imports, the survival rate is higher when China trades with regions that signed a regional trade agreement with it.

1. Introduction

With industrialization, urbanization, and the rapid increase in the world population, the issues of environmental pollution and global warming are becoming more and more concerning. Countries are increasingly aware of the importance of environmental protection for sustainable development, and it has become a consensus of global development to continuously promote a low-carbon economy, green economy, and circular economy. In order to alleviate the deterioration of the global environment, environment-related industries in various countries have been developed greatly, and the importance of the trade of environmental goods in international trade has gradually emerged [1].

The trade of environmental goods has greatly reduced global environmental problems while promoting global economic development [2]. China has seen rapid growth of the trade of environmental goods since its accession to the World Trade Organization (WTO) in 2001 and has become the country with the highest trade value of environmental goods in the world. According to the BACI database, developed by CEPII [3], the export value of China’s environmental goods accounted for 18.77% of the world’s export value in 2020, and the import value accounted for 13.16%. China’s share was only 2.52% for exports in 2002 and 6.06% for imports. Maintaining and further promoting the trade of environmental goods will not only help China achieve its environmental goals, such as carbon peaking and carbon neutrality, but also contribute to high-quality trade development, and sustainable and high-level “double-cycle” development.

Most advances in the theoretical and empirical trade literature mainly concern the differences in the roles of the intensive and extensive margins in trade growth [4,5]. Besedeš and Prusa introduced survival analysis, which is widely used in biology and medicine, into the research of international trade [6,7]. Recent contributions have begun to emphasize the importance of enhanced survival of existing trade relationships to the long-term trade growth [8,9]. Due to the uncertainty of demand and costs, most trade relationships are difficult to maintain over time [10,11]. This leads to some significative research questions: What is the survival status of the trade relationships of China’s environmental goods for exports and imports? Is there any instability in trade relationships? In the context of the rapid growth of China’s trade of environmental goods, it is of great theoretical and practical significance to examine these issues for China to improve the trade stability of environmental goods and maintain the sustainable trade growth.

To better understand the survival of China’s exports and imports of environmental goods, this paper uses survival analysis to summarizes the survival characteristics over the period 2002–2020 based on the data at the HS-6 digit level and compares these results with those of major exporting and importing countries. Moreover, the survival analysis for different categories of environmental goods and for different groups by regions whether they signed a regional trade agreement (RTA) with China were also performed. As a result, the paper’s goals are to reveal the problems concerning the stability of China’s exports and imports of environmental goods and provide meaningful guidance for China to maintain and further expand its trade of environmental goods.

2. Literature Review

2.1. About Trade Survival

The heterogeneous firm models with complete information assume that once the trade relationships are established, they will survive for a long time, as they face no uncertainty, which is obviously contrary to the evidence that a large amount of zero trade flow exists in the international market [12,13,14,15,16]. Using TS-7 and HS-10 digit level data, Besedeš and Prusa found that the trade relationships of the imports of United States (USA) did not survive long during the period 1972–2001, and most had multiple trade spells [6]. By incorporating informational uncertainty, the extension of the Melitz’s model by Segura-Cayuela and Vilarrubia addressed the dynamic issues for trade survival [17]. Globally, literature regarding trade survival has been gradually aroused in the empirical trade fields, mainly focusing on two aspects. On the one hand, according to the import or export data at different levels, the trade survival characteristics of different countries or economies are explored. Using HS-8 digit-level data, Nitsch found that most German import trade relationships lasted only 1–3 years during the period 1995–2005 [18]. Using firm-level data from Spain, Esteve-Perez et al. found that nearly half of Spanish firms chose to stop exporting at the first year of export [19]. Wang et al. analyzed the export survival of seafood products from ASEAN countries during the period 1996–2014 using HS-6 digit-level data and found that the average length of trade spells was only 4.42 years [20]. Most trade relationships tend to be short-lived, having numerous repeated entries and exits [21,22,23,24,25]. On the other hand, the determinants of trade survival are examined, and most researchers believe that the economic scale [26,27], geographical distance [28,29], income level [30,31], and trade barriers [32,33] between trading parties are key factors affecting the trade survival.

2.2. About China’s Trade of Environmental Goods

With the strengthening of environmental protection awareness in various countries [34], the studies on the trade of environmental goods have gradually emerged, and most of them concern the relation of trade liberalization and trade performance [35,36], and the relation of environmental policy and trade performance [37,38].

Early studies on China’s exports of environmental goods have mainly focused on trade competitiveness. Ma assessed the trade competitiveness of China’s environmental goods from the perspective of value, quantity, and price based on the list of WTO negotiations on environmental issues and argued that China’s advantage in trade value was not obvious, and the number of tax codes that were both superior in trade quantity and price was small [39]. You and Wen performed a comprehensive study and judgment on the export competitiveness of China’s environmental goods through the construction of a trade competitiveness index, and believed that on the whole, the exports of China’s environmental goods mainly had the advantage of quantity, and the level of export competitiveness was similar to that of China’s overall export competitiveness [40]. Further, the constant share market (CMS) model was also applied to the study of factors influencing the exports of China’s environmental goods, and the researchers believed that the improvement of the overall competitiveness of environmental goods was the main source of the export growth of China’s environmental goods, and the structural effect and cross-effect contributed less to the export growth [2]. In addition, the dynamic evolution of export technology structure of China’s environmental goods was also the focus of many studies [1,41]. With the development of trade liberalization, recent research has begun to focus on the impact of RTAs on China’s exports of environmental goods, and most argued that signing RTAs has significantly promoted the export growth of China’s environmental goods.

Relative to exports, there is less research on China’s imports of environmental goods. Liu et al. found that the imports of environmental goods that were mainly intended to address a certain environmental issue but can be used in some non-environmental end uses that did not inevitably benefit the environment, and an effect of the import of goods that only had environmental end uses on reducing pollution intensity existed [42]. Additionally, Mao et al. analyzed the changes in the scale, product structure, and trade network structure of the imports of China’s environmental goods during the period 1996–2019 [43].

Although the literature on trade survival and China’s trade of environmental goods has been growing rapidly in recent decades, few studies have presented the characteristics related to the export and import survival of China’s environmental goods. To sum up, the main marginal contribution of this paper is the application of survival analysis to the study of the trading of environmental goods by China, and it reveals the survival characteristics for China’s exports and imports of environmental goods.

3. Material and Methods

3.1. Data Collection and Concept Definition

According to the list of environmental goods proposed by Asia–Pacific Economic Cooperation (APEC) in 2012, environmental goods can be classified as 53 HS-6 goods by the 2012 version of HS code. We converted these codes to the codes of the 1992 version with a total of 54 HS-6 goods, which can be grouped into three categories—namely, environmental monitoring and evaluation goods, renewable energy goods, and environmental protection goods, as shown in Table S1. The annual export and import data of environmental goods at the HS-6 digit level during the period 2002–2020 were used for analysis. The data came from the BACI database, developed by CEPII [3], and the 1992 version of HS code was adopted. By collating, there were 140,731 observations for China’s exports of environmental goods during the period 2002–2020 and 49,223 observations for imports. An observation contains information about China’s exports or imports of a specific HS-6 environmental good to or from a specific region (in this paper, “regions” are for those areas or countries which have specific iso codes) in a specific year.

We define an export trade relationship as exports of a specific HS-6 environment good from a specific region to a specific destination region and define an import trade relationship as imports of a specific environment good of a specific region from a specific source region. A trade spell is defined as a period during which a trade relationship continuously existed, and the duration of the trade spell is a discrete time in years that the trade spell has experienced from the start time point to the observed time point without disruption. The survival time of the trade spell is the duration of this trade spell from the start year to the observed disrupted year (i.e., there is no export or import in the next year).

Table 1. China’s exports of environmental good “841290” to Afghanistan.

Export Trade Relationship			2002	2003	2004	2005	2006	2007	2008	2009	2010	2011	2012	2013	2014	2015	2016	2017	2018	2019	2020
Exporting Region	HS-6 Environment Good	Destination Region
China	”841290”	Afghanistan								#			#	#				#	#

Note: “#” denotes that there is an observation in the specific year for China’s exports of environmental good “841290” to Afghanistan.

shows an example of an export trade relationship of China involving an environmental good. As displayed, the trade relationship of China’s exports of environmental good “841290” to Afghanistan had observations in 2009, 2012, and 2013, and 2016–2019. The trade relationship had a total of 3 trade spells over the period 2002–2020. The first spell was represented only by 2009, and the survival time was 1 year. The second spell contained observations from 2012 to 2013, and the survival time was 2 years. The third spell contained observations from 2017 to 2018, and the survival time was also 2 years.

3.2. Data Processing

Two common issues in trade survival research need to be explained and processed. ① Multiple trade spells. During the sample period, a trade relationship may have multiple trade spells. As displayed in Table 1, China has exported the environmental good “841290” to Afghanistan in 2009 and over the periods 2012–2013 and 2017–2018. Hence, this export trade relationship has had three trade spells. By treating these trade spells as mutually independent, the results of the survival analysis are robust [7,44]. ② Censored data. There are two cases for censored data. One is right-censoring, i.e., if a trade spell remains in 2020, then the specific disrupted year cannot be recorded. Another is left-censoring, i.e., if a trade spell exists in 2002, then the specific start year cannot be recorded. Although the right-censoring does not result in estimation bias for the survival analysis, this is not the case for left-censoring [45,46]. In the data, there are 4552 trade spells with left-censoring for exports, which account for 24.45% of the total of export trade spells, and 1953 trade spells with left-censoring for imports, which account for 20.43% of the import total. Thus, these spells cannot simply be deleted. In order to alleviate the left-censoring, the method of extending the scope of data collection was adopted [32]. The data collection scope for the duration and survival time was expanded to the period 1995–2020. Therefore, if a trade spell existed for the period 1996–2016 only, the survival time of this trade spell was recorded as 21 years rather than 15 years. After processing, the numbers of trade spells that still had left-censoring (i.e., the trade spell existed in 1995) were 1228 for exports and 1064 for imports; and among them, the numbers of trade spells which had a recorded duration of 26 years were 1204 for exports (i.e., these trade spell have existed continuously over the period 1995–2020), and these were 1003 for imports. In view of these figures, we did not delete the trade spells that still had left-censoring after processing for the basic analysis but presented the analysis results after deleting them in the form of robustness tests.

3.3. Kaplan–Meier Method

In survival analysis, the hazard function or survival function is often used to describe the trend of the hazard rate or survival rate over time. Let T_i represent the survival time of the trade spell i. Define the probability that trade spell i happens to be disrupted at a given duration l (i.e., T_i is equal to the duration l) as a hazard function:

$$h_i(l)=P(T_i=l|T_i\ge l)$$

(1)

Further, the probability that T_i exceeds duration l is defined as the survival function:

$$S_i(l)=P(T_i>l)={\displaystyle \prod _{k=1}^l[1-h_i(k)}]$$

(2)

where k is a series of durations of no more than l. Obviously, the survival function is a cumulative distribution function (CDF), and the hazard function is a conditional probability density function (CPDF).

The Kaplan–Meier (K–M) method is a common non-parametric estimation of a survival function and hazard function. Let n_k denote the number of trade spells which had survival time of no less than duration k, and d_k denote the number of trade spells which had survival time exactly equal to duration k. Then, the non-parametric estimation of the survival function obtained by the K–M method at a given duration l is:

$$\hat{S}(l)={\displaystyle \prod _{k=1}^l\frac{n_k-d_k}{n_k}}$$

(3)

Further, the non-parametric estimation of the hazard function obtained by the K–M method at a given duration l is:

$$\hat{h}(l)=\frac{d_l}{n_l}$$

(4)

where n_l denotes the number of trade spells which had observed survival time of no less than duration l, and d_l denotes the number of trade spells which had observed survival time as exactly equal to duration l.

4. Results

4.1. For Exports

4.1.1. Distribution of Trade Spells

Before presenting the K–M analysis, the distribution of export trade spells of China’s environmental goods was displayed. Since the trade spells with right-censoring do not have a recorded survival time, the concept of length of trade spell was used for analysis. The trade spells with right-censoring had their longest duration as length, and the trade spells without right-censoring had their survival time as length.

Table 2. Distribution of the trade spells of China’s environmental good exports for the period 2002–2020.

Spells across Relationships			Length across Spells
Number of Spells	Total Number of Relationships	Frequency (%)	Length (Years)	Total Number of Spells	Frequency (%)
1	5770	55.36	1	6091	32.72
2	2375	22.79	2	2192	11.77
3	1366	13.11	3	1153	6.19
4	624	5.99	4	708	3.80
5	235	2.25	5	509	2.73
6	46	0.44	6	375	2.01
7	5	0.05	7	311	1.67
8	2	0.02	8	275	1.48
			9	265	1.42
			10	251	1.35
			11	269	1.44
			12	285	1.53
			13	329	1.77
			14	381	2.05
			15	463	2.49
			16	392	2.11
			17	442	2.37
			18	441	2.37
			19	430	2.31
			20	335	1.80
			21	561	3.01
			22	345	1.85
			23	248	1.33
			24	195	1.05
			25	166	0.89
			26	1204	6.47
Total	10,423	100.000		18,616	100.000

presents the distribution of all trade spells of China’s exports of environmental goods during the period 2002–2020. For exports, there were 10,423 trade relationships and 18,616 trade spells. As seen in the first three columns, 44.64% of trade relationships had no less than two trade spells, indicating that many export trade relationships have multiple spells. The maximum number of trade spells for a specific trade relationship was eight, and there were two export trade relationships owning it, namely, China’s export of the environmental good “841960” to Albania and China’s export of the environmental good “901390” to French Polynesia. As seen in the last three columns, 32.72% of the trade spells were only 1 year in length, 50.69% of the trade spells had a length of no more than 3 years, and only 18.71% of the trade spells had a length of more than 18 years.

To deeply understand the survival of China’s exports of environmental goods internationally, countries with a total export value of more than USD200 billion for environmental goods during the period 2002–2020 were selected for comparative analysis, including China, Germany, the USA, Japan, Republic of Korea, Italy, United Kingdom, France, and the Netherlands.

Table 3. Distribution of export trade spells across major exporting countries for the period 2002–2020.

Num	Country	Total Number of Relationships	Total Number of Spells	Average Length (Years)	Frequency (%) Length = 1 Year	Frequency (%) Length < 4 Years	Frequency (%) Length > 18 Years
1	Germany	10,506	18,863	9.4997	34.13	52.04	27.46
2	USA	10,659	19,481	9.3795	33.68	51.91	26.66
3	China	10,423	18,616	8.3568	32.72	50.69	18.72
4	Italy	9499	18,385	8.3265	37.42	56.35	23.20
5	United Kingdom	10,242	20,709	7.9220	37.35	57.57	20.77
6	France	9712	19,577	7.8026	37.79	57.81	20.46
7	Japan	9065	18,726	6.7640	41.87	62.62	16.46
8	Netherlands	9726	20,737	6.3690	41.85	63.04	14.32
9	Republic of Korea	8178	16,770	5.9927	43.66	64.75	12.70

presents the distribution of the trade spells in each country summarily. As displayed, there were four countries which had more than 10,000 trade relationships for exports of environmental goods, including the USA, Germany, China, and the United Kingdom. During the sample period, the average length of trade spells of China was 8.36 years, ranking the third among the major exporting countries. Although the frequency for China’s spells whose length was equal to 1 year or less than 4 years was not so high, the frequency was 18.72% for trade spells whose length was more than 18 years—only ranking sixth. These data indicate that China’s exports of environmental goods have a higher frequency for mid-term trade relationships, and there is a gap with other major exporting countries in having long-term trade relationships.

Further, the distribution of China’s export trade spells across different categories of environmental goods over the period 2002–2020 was also analyzed comparatively.

Table 4. Distribution of China’s export trade spells across different categories of environmental goods for the period 2002–2020.

Num	Category	Total Number of Relationships	Total Number of Spells	Average Length (Years)	Frequency (%) Length = 1 Year	Frequency (%) Length < 4 Years	Frequency (%) Length > 18 Years
1	Environmental monitoring and evaluation goods	3602	5927	9.2158	30.62	47.04	21.87
2	Environmental protection goods	4157	7792	8.4565	33.60	52.37	17.57
3	Renewable energy goods	2664	4897	7.8777	33.86	52.42	16.72

presents the distribution of the spells across different categories. By category, the export trade spells of environmental monitoring and evaluation goods had the longest average length, reaching up to 9.22 years. Moreover, the spells of environmental monitoring and evaluation goods had the lowest frequency for length equal to 1 year or less than 4 years and had the highest frequency for lengths of more than 18 years. These indicate that China’s exports of environmental monitoring and evaluation goods may be relatively stable compared with its exportation of renewable energy goods and environmental protection goods. Although China’s exports of renewable energy goods had the least number of trade relationships, the export trade spells of renewable energy goods had the shortest average length, at only 7.88 years. Moreover, the spells of renewable energy goods had the highest frequency for length equal to 1 or less than 4 years and had the lowest frequency for length of more than 18 years across all categories. These data indicate that China’s exportation of renewable energy goods may be relatively unstable compared with its exports of other environmental goods.

4.1.2. Results of K–M Estimation

In this section, the results of K–M estimation are displayed.

Table 5. Survival rate and hazard rate of China’s exports of environmental goods by K–M estimation.

Duration (years)	With Left-Censoring		Without Left-Censoring
	Survival Rate	Hazard Rate	Survival Rate	Hazard Rate
1	0.6894	0.3106	0.6669	0.3331
2	0.5825	0.1552	0.5519	0.1725
3	0.5300	0.0901	0.4953	0.1026
4	0.4998	0.0569	0.4626	0.0659
5	0.4806	0.0384	0.4418	0.0451
6	0.4686	0.0249	0.4287	0.0296
7	0.4599	0.0186	0.4191	0.0223
8	0.4528	0.0155	0.4115	0.0182
9	0.4495	0.0072	0.4080	0.0086
10	0.4459	0.0081	0.4039	0.0100
11	0.4435	0.0054	0.4012	0.0068
12	0.4409	0.0057	0.3985	0.0067
13	0.4395	0.0034	0.3968	0.0041
14	0.4378	0.0038	0.3948	0.0050
15	0.4364	0.0031	0.3932	0.0042
16	0.4356	0.0018	0.3922	0.0025
17	0.4344	0.0028	0.3915	0.0018
18	0.4338	0.0014	0.3906	0.0022
19	0.4335	0.0007	0.3902	0.0011
20	0.4332	0.0007	0.3897	0.0013
21	0.4326	0.0014	0.3884	0.0032
22	0.4321	0.0011	0.3871	0.0033
23	0.4307	0.0032	0.3871	0.0000
24	0.4295	0.0029	0.3871	0.0000
25	0.4287	0.0017

shows the survival rate and hazard rate of China’s exports of environmental goods by K–M estimation (There is no survival rate or hazard rate for the twenty-sixth year of export due to the right-censoring). As the export duration increased, the survival rate gradually decreased, but the decline tended to shrink. In the first year of exports, the survival rate dropped sharply to 0.69, decreasing by 0.31. In the fourth year, the survival rate was 0.50, decreasing by 0.19 from the first year. In the twenty-fifth year, the survival rate was 0.43, only decreasing by 0.07 from the fourth year. Correspondingly, the hazard rate was negatively duration-dependent, i.e., as the export duration increased, the hazard rate tended to decrease. In the first year of exports, the hazard rate was up to 0.31. From the second year to the twelfth year, the average annual hazard rate was 0.04. From the thirteenth year to the twenty-fifth year, the average annual hazard rate was approximately 0.00. Further, we conducted the K–M estimation again by using the observations without left-censored trade spells, and the changes in the survival rate and hazard rate were almost the same as before, which verified the robustness.

Further, the export survival rates of environmental goods across major exporting countries are also reported for comparative analysis. As shown in

Table 6. Export survival rates of environmental goods across major exporting countries.

Num	Country	First Year	Fourth Year	Twelfth Year	Eighteenth Year	Twenty-Fifth Year
1	China	0.6894	0.4998	0.4409	0.4338	0.4287
2	Germany	0.6765	0.4783	0.3898	0.3761	0.3654
3	USA	0.6824	0.4759	0.3808	0.3607	0.3475
4	Italy	0.6422	0.4300	0.3398	0.3240	0.3147
5	France	0.6400	0.4174	0.3164	0.2977	0.2858
6	United Kingdom	0.6465	0.4166	0.3156	0.2911	0.2771
7	Netherlands	0.6013	0.3676	0.2741	0.2595	0.2519
8	Republic of Korea	0.5849	0.3514	0.2598	0.2460	0.2406
9	Japan	0.5983	0.3617	0.2553	0.2323	0.2157

, the survival rate of China’s exports of environmental goods has been always at the highest level in all durations across major exporting countries. In the twenty-fifth year of export, only China’s survival rate of exports of environmental goods was still more than 0.40. Moreover, the result of the log-rank (LR) test shows that the chi-square statistic is 2638.25 with a p-value of 0.00, so we reject the null hypothesis that there is no difference in survival functions across major exporting countries. These results indicate that China’s exports of environmental goods are relatively stable compared with other countries, which partly explains the rapid growth of China’s exports of environmental goods after access to the WTO.

According to the distribution of the export trade spells of China’s environmental goods above, in contrast to other major exporting countries, China has few long-term trade relationships. The reason why China’s exports of environmental goods have a relatively lower frequency of long-term trade relationships is not because the trade relationships are unstable, but because the trade relationships have not been established long enough. In fact, relatively stable trade relationships give China’s export growth of environmental goods great potential, although China’s exports of environmental goods have the highest share in the world currently.

We next compared the export survival rates across the different categories of environmental goods. As reported in

Table 7. Export survival rates across different categories of environmental goods.

Num	Country	First Year	Fourth Year	Twelfth Year	Eighteenth Year	Twenty-Fifth Year
1	Environmental monitoring and evaluation goods	0.7093	0.5398	0.4900	0.4844	0.4809
2	Environmental protection goods	0.6826	0.4835	0.4240	0.4163	0.4078
3	Renewable energy goods	0.6762	0.4771	0.4078	0.3995	0.3977

, the export of environmental monitoring and evaluation goods had the highest survival rate in all durations, and the export of renewable energy goods had the lowest survival rate in all durations. Moreover, the result of the LR test shows that the chi-square statistic is 76.14 with a p-value of 0.00, so we reject the null hypothesis that there is no difference in survival functions across different categories. These results verify the previous speculation based on the distribution of export trade spells across different categories: that China’s exports of environmental monitoring and evaluation goods are relatively stable compared with its exports of other environmental goods, and its exports of renewable energy goods are relatively unstable.

The regions importing environmental goods from China were divided into two groups based on whether they signed an RTA with China or not during the period 2002–2020. One contained the regions that signed RTAs with China, defined as the treatment group; the other contained the regions that did not sign RTAs with China, defined as the non-treatment group. The data of RTA signing came from the Gravity database, developed by CEPII [47]. Thus, a group comparison of export survival rate was conducted. As shown in

Table 8. Group comparison of export survival rate by whether each region signed an RTA with China.

Num	Group	First Year	Fourth Year	Twelfth Year	Eighteenth Year	Twenty-Fifth Year
1	Treatment group	0.8451	0.7318	0.7036	0.7036	0.7036
2	Non-treatment group	0.6726	0.4743	0.4113	0.4028	0.3955

, the export survival rate of the treatment group was always higher than that of the non-treatment group in all durations. Moreover, the survival rate of the treatment group was even as high as 0.70 in the twenty-fifth year of export, which was 0.30 higher than that of the non-treatment group. Further, the result of the LR test shows that the chi-square statistic is 495.21 with a p-value of 0.00, so we reject the null hypothesis that there is no difference in survival functions for the two groups. These results indicate that China’s exports of environmental goods survive better in the regions that signed an RTA with China.

4.2. For Imports

4.2.1. Distribution of Trade Spells

Table 9. Distribution of the trade spells of China’s environmental good imports for the period 2002–2020.

Spells across Relationships			Length across Spells
Number of Spells	Total Number of Relationships	Frequency (%)	Length (years)	Total Number of Spells	Frequency (%)
1	3105	58.34	1	4815	50.37
2	1018	19.13	2	1209	12.65
3	627	11.78	3	565	5.91
4	367	6.90	4	299	3.13
5	164	3.08	5	199	2.08
6	38	0.71	6	142	1.49
7	3	0.06	7	108	1.13
			8	96	1.00
			9	69	0.72
			10	85	0.89
			11	72	0.75
			12	75	0.78
			13	60	0.63
			14	60	0.63
			15	62	0.65
			16	75	0.78
			17	57	0.60
			18	66	0.69
			19	78	0.82
			20	65	0.68
			21	117	1.22
			22	89	0.93
			23	32	0.33
			24	30	0.31
			25	31	0.32
			26	1003	10.49
Total	5322	100.000		9559	100.000

presents the distribution of all trade spells of China’s imports of environmental goods during the period 2002–2020. For imports, there were 5322 trade relationships and 9559 trade spells. As seen in the first three columns, 41.66% of trade relationships had no less than two trade spells, indicating that many import trade relationships have multiple spells. The maximum number of trade spells for a specific import trade relationship was seven, and there were three import trade relationships with seven—namely, China’s import of the environmental good “840410” from Austria, China’s import of the environmental good “901380” from Morocco, and China’s import of the environmental good “840420” from Spain. As seen in the last three columns, 50.37% of the trade spells were only 1 year in length, 68.93% of the trade spells had a length of no more than 3 years, and 15.12% of the trade spells had a length of more than 18 years.

To deeply understand the survival of China’s imports of environmental goods internationally, countries with total import values of more than USD200 billion for environmental goods during the period 2002–2020 were selected for comparative analysis, including China, the USA, Germany, Republic of Korea, Japan, Mexico, France, and the United Kingdom.

Table 10. Distribution of import trade spells across major importing countries for the period 2002–2020.

Num	Country	Total Number of Relationships	Total Number of Spells	Average Length (Years)	Frequency (%) Length = 1 Year	Frequency (%) Length < 4 Years	Frequency (%) Length > 18 Years
1	Japan	4160	7676	6.2694	50.30	67.85	16.28
2	Germany	6756	13,524	6.1310	48.03	68.41	15.32
3	China	5322	9559	6.0750	50.37	68.93	15.12
4	Mexico	4558	8448	5.8185	48.80	69.19	13.45
5	USA	7254	15,024	5.6666	49.25	69.60	13.44
6	United Kingdom	6631	13,768	5.3911	50.05	71.06	12.18
7	Republic of Korea	5005	9671	5.2754	53.30	72.51	12.02
8	France	6564	13,636	5.1360	51.21	72.24	11.34

presents the distribution of the trade spells in each country summarily. As displayed, there were only two countries which had less than 5000 trade relationships for imports of environmental goods—Japan and Mexico. During the sample period, the average length of trade spells of China was 6.08 years, ranking third among the major importing countries. Although the frequency of China’s trade spells whose length was less than 4 years was not high, the frequency of 50.72% for spells whose length was equal to 1 year was only lower than the frequencies of the Republic of Korea and France. These results indicate that China’s imports of environmental goods may have the undesirable phenomenon of disrupting trade relationships immediately after they are established.

Further, the distribution of China’s import trade spells across different categories of environmental goods over the period 2002–2020 was also analyzed comparatively.

Table 11. Distribution of China’s import trade spells across different categories of environmental goods for the period 2002–2020.

Num	Category	Total Number of Relationships	Total Number of Spells	Average Length (Years)	Frequency (%) Length = 1 Year	Frequency (%) Length < 4 Years	Frequency (%) Length > 18 Years
1	Environmental monitoring and evaluation goods	2091	3610	6.1726	51.50	68.89	15.46
2	Environmental protection goods	1897	3508	6.1166	49.97	68.90	15.51
3	Renewable energy goods	1334	2441	5.8710	49.28	69.03	14.05

presents the distribution of the spells across categories. By category, the import trade spells of environmental monitoring and evaluation goods had the longest average length, reaching 6.17 years, and had the lowest frequency for lengths of less than 4 years. The spells of renewable energy goods had the lowest frequency for length equal to 1 year, and the spells of environmental protection goods had the highest frequency for a length of more than 18 years. These results indicate that there may be no overall difference in survival performance among different categories for China’s imports of environmental goods.

4.2.2. Results of K–M Estimation

Table 12. Survival rate and hazard rate of China’s imports of environmental goods by K–M estimation.

Duration (Years)	With Left-Censoring		Without Left-Censoring
	Survival Rate	Hazard Rate	Survival Rate	Hazard Rate
1	0.5172	0.4828	0.4538	0.5462
2	0.3982	0.2300	0.3179	0.2994
3	0.3461	0.1308	0.2575	0.1900
4	0.3196	0.0767	0.2263	0.1214
5	0.3037	0.0496	0.2073	0.0839
6	0.2916	0.0400	0.1925	0.0712
7	0.2843	0.0250	0.1835	0.0469
8	0.2789	0.0189	0.1777	0.0317
9	0.2761	0.0101	0.1742	0.0196
10	0.2722	0.0140	0.1701	0.0233
11	0.2685	0.0135	0.1662	0.0229
12	0.2636	0.0183	0.1626	0.0221
13	0.2617	0.0073	0.1602	0.0147
14	0.2600	0.0064	0.1588	0.0087
15	0.2583	0.0067	0.1570	0.0111
16	0.2565	0.0070	0.1545	0.0160
17	0.2551	0.0053	0.1527	0.0119
18	0.2543	0.0034	0.1516	0.0069
19	0.2533	0.0036	0.1504	0.0083
20	0.2528	0.0023	0.1493	0.0068
21	0.2525	0.0008	0.1493	0.0000
22	0.2519	0.0027	0.1493	0.0000
23	0.2516	0.0009	0.1493	0.0000
24	0.2511	0.0019	0.1447	0.0313
25	0.2511	0.0000

shows the survival rate and hazard rate of China’s imports of environmental goods by K–M estimation. As the import duration increased, the survival rate gradually decreased, but the decline tended to shrink. In the first year of imports, the survival rate dropped sharply to 0.52, decreasing by 0.48. In the fourth year, the survival rate was 0.32, decreasing by 0.20 from the first year. In the twenty-fifth year, the survival rate was 0.25, only decreasing by 0.07 from the fourth year. Correspondingly, the hazard rate was negatively duration-dependent. In the first year of import, the hazard rate was up to 0.48. From the second year to the twelfth year, the average annual hazard rate was 0.06. From the thirteenth year to the twenty-fifth year, the average annual hazard rate was approximately 0.00. Further, we conducted the K–M estimation again by using the observations without left-censored trade spells, and the changes in the survival rate and hazard rate were almost the same as before, which verified the robustness.

Further, the import survival rates of environmental goods across major importing countries were also reported for comparative analysis. As shown in

Table 13. Import survival rates of environmental goods across major importing countries.

Num	Country	First Year	Fourth Year	Twelfth Year	Eighteenth Year	Twenty-Fifth Year
1	China	0.5172	0.3196	0.2636	0.2543	0.2511
2	Japan	0.5159	0.3280	0.2546	0.2437	0.2395
3	Germany	0.5400	0.3164	0.2354	0.2252	0.2212
4	Republic of Korea	0.4881	0.2769	0.2174	0.2110	0.2083
5	USA	0.5259	0.2982	0.2197	0.2081	0.2027
6	Mexico	0.5272	0.3013	0.2233	0.2105	0.2023
7	France	0.5102	0.2754	0.1972	0.1899	0.1868
8	United Kingdom	0.5183	0.2826	0.1993	0.1834	0.1762

, China’s imports of environmental goods have an advantage in facilitating long-term trade relationships compared with other major importing countries. In the eighteenth year and the twenty-fifth year of import, the survival rate of China’s imports of environmental goods was still more than 0.25. However, China’s imports of environmental goods had a relatively lower survival rate in the first year of import compared with others, indicating that China’s imports of environmental goods have a phenomenon of high “infant death”; i.e., a lot of trade relationships are disrupted immediately after they are established, which is consistent with the expectation based on the distribution of import trade spells across major importing countries. Moreover, the result of the LR test shows that the chi-square statistic is 170.49 with a p-value of 0.00, so we reject the null hypothesis that there is no difference in survival functions across major importing countries. Once the trade relationships of China’s imports of environmental goods have successfully crossed the first year of import, they can more easily become long-term.

We next compare the import survival rates across the different categories of environmental goods. As reported in

Table 14. Import survival rates across different categories of environmental goods.

Num	Country	First Year	Fourth Year	Twelfth Year	Eighteenth Year	Twenty-Fifth Year
1	Environmental monitoring and evaluation goods	0.5045	0.3167	0.2754	0.2699	0.2694
2	Environmental protection goods	0.5229	0.3198	0.2595	0.2475	0.2429
3	Renewable energy goods	0.5277	0.3233	0.2516	0.2407	0.2361

, although the imports of environmental monitoring and evaluation goods have the phenomenon of high “infant death”, they have an advantage in long-term trade relationships compared with environmental protection goods and renewable energy goods. On the contrary, although the imports of renewable energy goods had the highest survival rate in the first year and the fourth year of imports, they had a disadvantage in maintaining long-term trade relationships compared with other environmental goods. Moreover, the result of the LR test shows that the chi-square statistic is 1.31 with a p-value of 0.52, so we cannot reject the null hypothesis that there is no difference in survival function across the categories. These results verify the previous speculation based on the distribution of import trade spells across different categories: that there are no overall differences in survival performance across different categories for China’s imports of environmental goods.

Similarly, a group comparison of import survival rate by whether each region signed an RTA with China was also conducted. The treatment was similar to the previous study on exports. As shown in

Table 15. Group comparison of import survival rate by whether each region signed an RTA with China.

Num	Group	First Year	Fourth Year	Twelfth Year	Eighteenth Year	Twenty-Fifth Year
1	Treatment group	0.5987	0.3890	0.3047	0.2940	0.2899
2	Non-treatment group	0.4945	0.3003	0.2525	0.2435	0.2407

, the import survival rate of the treatment group was higher than that of the non-treatment group for all durations. Further, the results of the LR test show that the chi-square statistic is 41.51 with a p-value of 0.00, so we reject the null hypothesis that there is no difference in survival functions for the two groups. These results indicate that as with exports, China’s imports of environmental goods survive better in the regions that sign RTAs with them.

5. Discussion and Conclusions

Little of the previous literature related to the trade of environmental goods focuses on trade survival. Trade survival is closely related to the intensive margin and the extensive margin of trade [7,48]. Based on the data at the HS-6 digit level, we used survival analysis to present the survival characteristics of China’s exports and imports of environmental goods over the period 2002–2020 and compared them with those of other major exporting and importing countries. As a result, our study contributes to advances in understanding the trade survival of environmental goods and has theoretical and practical relevance for helping China to improve trade stability for environmental goods and maintain sustainable trade growth. The main conclusions drawn in the research are as follows.

For exports, the average length of 18,616 trade spells of was 8.36 years, ranking third among the major exporting countries. Although the length of 50.69% of these trade spells was no more than 3 years, the length of only 18.72% of them was more than 18 years, which was not an advantageous figure compared to other countries. The analysis based on the K–M estimation showed that the hazard rate was 0.31 in the first year of export, which is lower than that of China’s export of all goods [49]. The hazard rate became lower and lower as the export duration increased, which is consistent with recent studies about China’s export survival [50,51]. Compared with other countries, the survival rate of China’s exports of environmental goods has been always at the highest level for all durations. By category, the export survival rate of environmental monitoring and evaluation goods was always higher compared with the others; and the export survival rate of renewable energy goods was always lowest. Moreover, the result of the group comparison indicates that the export survival rate of the RTA treatment group is significantly higher than that of the non-treatment group, which is also supported by Cyrus [30].

For imports, the average length of 9559 trade spells was 6.08 years, ranking third among the major importing countries. The frequency of 50.37% for trade spells of 1 year is worse than those of other countries. The analysis based on the K–M estimation showed that the hazard rate was up to 0.48 in the first year of import, which is higher than that of China’s import of all goods [52]. Moreover, compared with other countries, the survival rate of China’s imports of environmental goods was lower in the first year of import, and the survival rate was the highest level for the eighteenth year and the twenty-fifth year. There is a phenomenon of high “infant death” for China’s importation of environmental goods, which is also supported by Cadot [23]. By category, the import survival rate of environmental monitoring and evaluation goods was the highest among the relevant nations in the twenty-fifth year of import, but the lowest in the first year. The opposite was the case for renewable energy goods. On the whole, there are no overall differences in survival performance among categories for China’s imports of environmental goods. Moreover, the results of the group comparison indicate that the import survival rate of the RTA treatment group is also significantly higher than that of the non-treatment group.

According to the above conclusions, we give the following recommendations for the sustainable growth of China’s trade of environmental goods. China should continue to hold the current export status and develop more export trade relationships. Meanwhile, the export trade relationships for renewable energy goods should be further improved. For imports of environmental goods, China should pay as much attention as possible to the first year of import to prevent the undesirable phenomenon of trade relationships being disrupted immediately after they are established to make more relationships become long-term. Moreover, China should sign RTAs with more trading partners to reduce trade costs, thereby improving the survival of trade relationships. However, the main focus of our study was to explore the stability of export and import trade relationships of China’s environmental goods by using complete survival analysis, and the specific determinants of trade survival need to be further examined in the future.