1 Introduction

Healthy oceans and marine ecosystem services underpin the ocean economy and play an important role in the existence and development of human society. The ocean is also a site for strategic capital investment and development. However, the sustainable development of the ocean cannot be separated from the conservation of marine ecosystems. The 19th National Congress of the Communist Party of China indicated that the country should adhere to ‘land and sea coordination’ and accelerate the pace of building China into a maritime power. Land–sea coordination and land–sea interactions are important considerations for the implementation of national spatial governance in coastal countries, and increasing maritime power has become an important policy goal of the Chinese government (Gao et al., 2022). General Secretary Xi Jinping emphasized that the construction of what is referred to as ‘a marine ecosystem civilization’ should be incorporated into the overall layout of marine development such that ‘China should continue to make great strides in the conservation of marine ecology, prevention and treatment of marine pollution, and protection of marine biodiversity in a bid to leave a clean marine environment to the world’s future generations. This concerns the common well-being of China and the world'.

Marine and coastal waters provide a variety of benefits to society that are generated through ecosystem services. Such services are, therefore, closely linked to the well-being of society. As a result, the valuation of marine ecosystem service benefits has become an important component for the implementation of a number of policy instruments, such as ecosystem-based management, ecosystem compensation programs, and payments for ecosystem service schemes (Börger et al., 2014; Xie et al., 2015; Ruslan et al., 2022). The study of the valuation of marine ecosystem services in China can be traced back to 1996 (Gao & Wang, 1996). Almost a decade later, in 2005, the State Oceanic Administration of China (SOA) initiated a 5-year program called ‘Service Evaluation of Marine Ecosystems in China Waters’. Based on the current situation of marine development and utilization in China, the SOA attempted to establish a set of valuation index systems and valuation models for marine ecosystem services that were in line with the marine ecosystem characteristics of China. Over nearly three decades, the valuation of marine ecosystem service benefits to China’s society has covered more than 10 types of ecosystems, such as coastal wetlands, coastal zones, bays, beaches, mangroves, and tidal flats (Shi et al., 2007; Chao & Liu, 2013). Valuation estimates include regulation, biological conservation, use, recreation, landscape, research and culture, and nonuse values (Wang et al., 2021).

The valuation methodologies employed gradually changed from market-based methods (represented by the replacement cost approach) to nonmarket valuation methods represented by stated (e.g., contingent valuation method) and revealed (e.g., travel cost method) approaches (Hanley et al., 2015; Tonin, 2018). Stated preference approaches are the only methodologies that can be used for simultaneously estimating the use and nonuse values associated with a change in marine ecosystems (Norton & Hynes, 2014). Value transfer methods beyond primary valuation approaches have also been developed; these methods use value estimates from previously conducted primary studies to estimate changes in ecosystem service values at other ‘policy’ sites for which little or no data are available in relation to benefit values (Rosenberger & Stanley, 2006; Hynes & O' Donoghue, 2020). Value transfer can be conducted by the simple transferal of the mean value from the original research, an income-adjusted mean value from the original research, or the original complete value function. Of particular interest to this work is the meta-analysis approach for value function transfer that uses regression methods to examine the influence of study- and site-specific characteristics from relevant primary studies on estimated values (Brouwer et al., 1999; Hynes et al., 2018).

Meta-analysis is an analytical method for the quantitative review and valuation of empirical research literature. A systematic quantitative analysis of empirical results is conducted by collecting values and associated study information from previous research on the same relevant ecosystem (Field & Gillett, 2010; Folkersen et al., 2018; Fan et al., 2022). This method was first introduced into the field of ecosystem economics by Walsh et al. (1989). Variables with a considerable effect on ecosystem service values can be established on the basis of a large number of independent empirical research results. Benefit transfer equations are then estimated by using regression analysis to deduce ecosystem service values (Liu et al., 2020; Wu & Zeng, 2021; Kang et al., 2022).

The key to meta-analysis is to develop a regression equation for the values of ecosystem services. Although numerous variables affect the values of ecosystem services, ecological parameters, such as the type, service category, and size of ecosystems (Guo et al., 2016), have received more attention than other variables. The economic values of various ecosystem types can differ substantially, according to Yang et al. (2017), and coastal wetlands and offshore ecosystems have high assessment values due to the high degree of exploitation of marine ecosystems, leading to high provisioning service values in particular. Within a single ecosystem, different associated ecosystem services can have varying values. Regulating services, such as climate regulation and water purification, are typically valued higher than other services (Barbier, 2013). A large increase in the assessed value of cultural services has been seen in recent years, and the willingness to pay to maintain the recreational value of marine ecosystems is growing (Sun et al., 2018; Kang et al., 2022). In addition to ecological parameters, valuation techniques affect the economic value of ecosystem services.

The three most commonly used approaches—the contingent valuation method (CVM), choice experiments (CE), and travel cost method (TCM)—can produce widely varying results. Hynes et al. (2018) found that the CVM generates higher estimates of value than the TCM. The nonmarket valuation of ecosystem services using stated preference approaches depends on the respondents’ willingness to pay, which is influenced by factors such as income and ecological awareness. The study of Brander et al. (2012) on the value of mangrove ecosystem services in Southeast Asia revealed that a high GDP per capita is associated with a high willingness to pay for ecosystem services. The willingness to pay is also influenced by the respondents’ educational background and how important ecosystem services are to them (Qin et al., 2021). Meta-analysis, which has the advantages of transfer efficiency and the ability to control for underlying differences in the ecosystems being examined across original studies and policy sites, has become an important technical means for determining estimates of ecosystem service values (Groothuis, 2005; Navrud & Ready, 2007; Zhao & Wang, 2011a, b). However, the application of meta-analysis in the study of ecosystem service values in China is quite limited. Yang et al. (2017) conducted a meta-analysis of the values of wetland ecosystem services. However, their study was limited to a small number of coastal wetland ecosystems. In view of the increasingly extensive research on the valuation of marine ecosystem services in China, in this work, the meta-analysis method was applied to evaluate the drivers of the variation in the estimates of marine ecosystem service values in China comprehensively.

The research methods employed are fully described in the following sections. The database of the values of marine ecosystem services in China and the specific form of the meta-regression model are presented. The multilevel model (MLM) results are provided in Sect. 3. The findings are reviewed and discussed in Sect. 4, and some final conclusions are offered in Sect. 5.

2 Research methods

2.1 Data screening and coding

China National Knowledge Infrastructure; Elsevier, Google Scholar, and marine ecosystem service databases, such as The National Ocean Economics Program, Marine Ecosystem Services Partnership, The Economics of Ecosystems and Biodiversity, and Ecosystem Services Valuation Database, were used to gather relevant studies on the valuation of marine ecosystem services in China. Chinese and English literature published between 1996 and 2021 on the valuation of the economic value of marine ecosystem services in China was compiled and supplemented by manually searching gray literature. Search keywords included ‘ecosystem services’, ‘ecosystem value’, ‘resource value’, ‘ecosystem compensation’, and ‘ecosystem payment’. In addition to these words, ‘China’ and ‘Chinese’ were used for online searches.

Figure 1 summarizes the steps in literature screening. The inclusion criteria were as follows: (1) the study must cite literature on empirical studies on the valuation of marine ecosystem services; (2) the study must quantify the monetary value of specific marine ecosystem services; (3) the study must provide clear information on the methods, subjects, and timing of research; (4) the study must use the stated preference methodsFootnote 1 represented by the CVM and/or the CE approach; (5) the researcher should be able to obtain socioeconomic indicators for the specific research location and time of each study, such as GDP per capita and years of education per capita; and (6) the researcher must be able to convert the unit of value into ‘yuan/ hectare/year’.

Fig. 1
figure 1

Literature screening and results

Full size image

The quality control measures used included (1) screening the literature in accordance with the inclusion criteria, (2) excluding repeated reports and reports with little information, and (3) in putting the data from the selected literature into a suitably searchable database and double checking the data to ensure correct input.

According to Martínez-Estévez et al. (2013), each observation represents a single value estimate. When more than one value estimate is given by a study, they are recorded as separate observations, and the different methods and/or models used to produce the values are considered. By searching databases and considering the technical means and completeness of the literature, 57 studies (52 peer-reviewed journal articles and five gray literature papers) that used the stated preference method to evaluate marine ecosystem services were finally screened. A total of 72 valid sample observations were obtained, with more than one value estimate reported by a number of the studies. The specific literature information is shown in Appendix A.

2.2 Model setting

2.2.1 Meta-regression model

In this work, the value of marine ecosystem services in China was taken as the dependent variable in the model. The meta-regression model was set as

$$\mathrm{Y}=\mathrm{\alpha }+\mathrm{\beta X}+\varepsilon$$
(1)

where Y is the value of a marine ecosystem service in China (in natural log form), and X is the vector set of impact factors, including a variety of variables affecting the value of a marine ecosystem service.

2.2.2 MLM

The ordinary least squares method (OLS) is widely used in the estimation of regression models. However, the sample data should meet the assumption of being independent and identically distributed. Some of the variables used to predict the values of marine ecosystem services, such as the author, study, or location to which a given estimate pertains, may be constant across a number of estimates. In such cases, a generalized least squares regression technique called multilevel modeling may be appropriate and is often used in the meta-analysis of value transfer literature (Johnston et al., 2005; Brander et al., 2007; Ghermandi et al., 2008; Hynes et al., 2018). An MLM is mainly used to study data with a hierarchical or nested structure. By establishing a set of multilevel regression equations, the total error is divided into errors at each level, which can reflect the interaction effect between levels and indicates that the restrictive assumption of random error independence does not have to be adhered to (Bateman & Jones, 2003; Li et al., 2011).

In this study, a generalized linear two-level regression model explicitly incorporating regional-level effects was estimated. The specific settings are as follows:

Level 1:

$$Y={\upbeta }_{0}+{\upbeta }_{1}\mathrm{X}+\varepsilon$$
(2)

Level 2:

$${\upbeta }_{0}={\upgamma }_{00}+{\gamma }_{01}W+{u}_{0}$$
(3)
$${\upbeta }_{1}={\upgamma }_{10}+{\gamma }_{11}W+{u}_{1}$$
(4)

where X is the explanatory variable at the individual level, and W is the explanatory variable at the regional level. \(\mathrm{\rm E}\), \({\mathrm{u}}_{0}\), and \({\mathrm{u}}_{1}\) are errors.

2.2.3 Transfer error

The effectiveness of the meta-regression equation can be assessed on the basis of the transfer error. The transfer error is calculated as follows (Zhao & Wang, 2011a, b):

$$\mathrm{TE}=\left|\frac{{CS}_{RT}-{CS}_{ACT}}{{CS}_{ACT}}\right|\times 100\%$$
(5)

where \({CS}_{RT}\) is the predicted value from the regression model, and \({CS}_{ACT}\) is the actual estimate from the study.

2.3 Variable setting

2.3.1 Explained variables

The expression forms of the ecosystem service values from the literature included the annual average value per unit area of the ecosystem (unit: yuan/hectare/year), willingness to pay per capita (unit: yuan per capita), and total value (unit: yuan). The value of a marine ecosystem service was uniformly transformed into an average annual value per unit area (unit: yuan/hectare/year) to facilitate comparison and analysis.

The conversion equation for the willingness to pay per capita (unit: yuan per capita) is as follows:

$$\mathrm{Average\;annual\;value\;per\;unit\;area}=\left(\mathrm{willingness\;to\;pay\;per\;capita}\times \mathrm{region\;population}\right)/\mathrm{research\;region\;area}$$

The conversion equation for the total ecosystem value (unit: yuan) is as follows:

$$\mathrm{Average\;annual\;value\;per\;unit\;area}=\mathrm{total\;value\;of\;ecosystem\;services}/\mathrm{area\;of\;the\;study\;area}$$

In view of the large valuation timespan (1996–2019), the GDP deflator was used to adjust all values to the year 2018, and the base period was selected in reference to the study of Roldan et al. (2021).

2.3.2 Explanatory variables

On the basis of a previous meta-analysis of value estimates and the effective information provided by the database, 25 variables were first selected across three categories, namely, the ecosystem characteristics of the research subject, technical characteristics of the research method, and socioeconomic characteristics of the study region, as the independent variables in the meta-function. The relevant variables were coded, and the assigned values were recorded in accordance with the requirements of statistical and econometric data. The details are shown in Table 1.

Table 1 Variable selection and coding of the meta-regression model
Full size table

Stated preference methods rely on hypothetical scenarios to present the change in ecosystem services to respondents. Different scenario settings have different degrees of influence on the well-being of respondents, leading to differences in the willingness to pay (Tokunaga et al., 2020; Taye et al., 2021). In this study, the valuation scenario was included as an explanatory variable in the regression equation to test whether a ‘scenario bias’ existed in the selected study estimates. The scenarios were classified as ecosystem protection, compensation, improvement, or existence.

2.4 Data sources beyond valuation studies

In addition to the observations extracted from the literature, other socioeconomic data for the research areas under scrutiny in the original studies were obtained from The China Statistical Yearbook. In particular, the average years of education and average GDP of the study regions were recorded. The source data in the yearbook originated from the National Population Census bulletin.

3 Results

3.1 Descriptive statistics

As expected (Table 2), the degree of the data concentration of the values of marine ecosystem services, ecosystem area, and number of questionnaires have significantly increased after log-transformation. In terms of sample characteristics, the subjects of the evaluation of marine ecosystem service in China are mainly offshore and wetland ecosystems, and the value categories are mainly recreational and nonuse values. The dominant valuation approach is the CVM, and the dominant valuation scenario is ecosystem protection.

Table 2 Descriptive statistics of variables
Full size table

3.2 Meta-regression model estimation

OLS and MLM were estimated as part of the meta-analysis. The results are shown in Table 3 and graphically summarized in Fig. 2. Figures 2a and b report the results of the separate regression of OLS and MLM. The OLS regression results show that the F statistic is overall significant at the 1% confidence level and that R2 is 0.67 after adjustment. As shown in Table 4, the P of the White test result is 0.1985, which suggests that heteroscedasticity is not a significant problem. Moreover, the maximum value of the variance inflation factor does not exceed 10, indicating that no significant multicollinearity exists among the variables.

Table 3 Results of OLS and MLM regression
Full size table
Fig. 2
figure 2

Coefficients of each variable in the two regression models

Full size image
Table 4 Variance inflation factors and white test results
Full size table

The results of MLM regression demonstrate that the overall model is statistically significant. The MLM is statistically different from a one-level ordinary linear regression model in accordance with the results of a likelihood ratio test (chi2 = 9.06, P < 0.01), indicating that in Eq. (3), \({u}_{j}\ne 0\). The model has a significant hierarchical structure, showing that even after controlling for the variables at the individual level of the model, statistically significant variation remains in the estimates of the values of marine ecosystem services between regions. The variance of the regression equation random error \({\sigma }_{e}^{2}\) is 1.352, and the variance of the random error \({\sigma }_{u}^{2}\) at the second level is 1.092. If \({\sigma }_{u}^{2}\) is greater than 0, differences exist in the estimated values between regions. The intragroup correlation coefficient ICCFootnote 2 is approximately 0.45, indicating that 45% of the difference in marine ecosystem values is caused by regional differences. The MLM regression can produce a better statistical imitative effect than the OLS model, and the significance of parameter estimation is considerably improved.

Among the ecosystem characteristic variables, ecosystem area, ecosystem type, and value category significantly affect thevalues of marine ecosystem services. Every 1% increase in ecosystem area reduces the values of marine ecosystem services by 0.783%, indicating that the marine ecosystem area has a negatively inelastic effect on ecosystem service values. Coastal wetlands and beaches have higher ecosystem service values than offshore. Furthermore, the recreational value is higher than the nonuse value.

Among the technical characteristic variables, the valuation method, respondent category, and valuation scenario significantly affect ecosystem service values. The values obtained by the CVM are lower than those acquired through the choice experiment method. Among respondents, professional staff shows the highest willingness to pay for marine ecosystem services, whereas tourists have the lowest willingness to pay. The willingness to pay of indiscriminate respondents and local residents is in between that of professional staff and tourist groups. Among the four valuation scenarios, the willingness to pay for ecosystem compensation is the lowest, and that for ecosystem protection is the highest.

Among the socioeconomic characteristic variables, income level significantly affects the values of marine ecosystem services at the 5% significance level. All else being equal, if the GDP per capita increases by 1%, the values of marine ecosystem services increase by 0.559%.

3.3 Transfer error

In this work, the full-sampled literature (72 study observations) was used to calculate transfer error (Fig. 3). The mean transfer errors of the OLS and MLM models are 14.19% and 15.69%, respectively. According to Brouwer (2000) and Ready & Navrud  (2006), if the mean transfer error is less than 40%, a model can be considered effective. In this study, the mean transfer error is less than 20%, indicating that both models effectively predict the in-sample value.

Fig. 3
figure 3

Transfer error results

Full size image

4 Discussion

The variation in the estimates of the values of marine ecosystem services in studies from China is affected by the ecosystem characteristics of the study areas, technical characteristics of research methods, and regional socioeconomic characteristics. Consistent with the situation in other countries, the willingness to pay for marine ecosystem services in China has a diminishing marginal utility and wealth effect, indicating that the people’s willingness to pay for ecosystem services decreases as the study are increases (Woodward & Wui, 2001; Brander et al., 2006; Zhang et al., 2015). In line with other international studies, this work found that the willingness to pay increased with the increase in disposable income per capita (Trujillo et al., 2016; Hynes et al., 2018; Zambrano-Monserrate & Ruano, 2020; Kang et al., 2022). Consistent with the studies of Fan et al. (2017) and Hynes et al. (2018), this study discovered that the values of marine ecosystem services obtained by the choice experiment method were significantly higher than those obtained by the CVM.

The ecosystem type and ecosystem service value categories were identified as important factors affecting the willingness to pay for marine ecosystem services. However, their effects were closely related to regional ecosystem governance policies and the public’s ecosystem awareness. Hynes et al. (2018) found that people’s willingness to pay for beach recreation, followed by coastal wetland recreation, was the highest, whereas the willingness to pay for the recreational value of mangroves was the lowest. This study in China showed that although people are willing to pay for ecosystem services that can provide recreational value, they are more willing to pay for wetlands than for beaches, likely because coastal wetlands are culturally more important in Asia than in many other parts of the world (Xie & Guo, 2018; Wang et al., 2022). An area for future research would be a separate meta-analysis on revealed preference travel costs for coastal recreational sites in China. Such a study would provide a clear picture of the coastal and marine ecosystem preferences of Chinese recreationalists.

The momentum of sustainable and rapid economic and social development in China’s coastal areas has not diminished in the twenty-first century. The shortage of land resources and the limitations on land use have become the key factors restricting economic development (Shan & Li, 2020). China launched the fourth round of land reclamation against this background. Previous land reclamation rounds resulted in a reduction of 1.3612 million hectares of coastal wetlands in China from 2003 to 2013, representing a reduction rate of 22.91%Footnote 3. With the promotion of the concept of the construction of a marine ecosystem civilization, China has taken a number of active measures to strengthen the protection of coastal wetlands. A total of 42 national marine parks were established from 2011 to 2019. Coastal populations also have a better understanding of the ecosystem services delivered by wetlands than of those delivered by other marine ecosystem types. By contrast, people have a limited understanding of the ecosystem function of beaches and only evaluate beaches from the perspective of recreation. In addition, in terms of public education, the publicity and education related to the importance of healthy marine ecosystems and marine biodiversity protection continue to lag, and the public lacks a sufficient understanding of the broad range of marine ecosystem functions and services. This situation may explain why ecosystem value categories other than recreation have no significant effect on the willingness to pay.

The CVM and CE are the two most common stated preference valuation methods. This work further confirmed that different valuation methods did have an influence on the valuation of ecosystem services. Although a large amount of literature exists on the comparison of the two methods, a consistent conclusion on which method is betterdoes not exist (Colombo et al., 2005; Jin et al., 2006; Heet al., 2017; Bostan et al., 2020). Most researchers agree that the effectiveness of the stated preference method is affected by questionnaire design and respondent attributes (Wu et al., 2017; Liuet al., 2019). During data collection, most of the studies were found to have not elaborated on the questionnaire design in detail, and only 45% of the studies included a questionnaire as an Appendix. Therefore, the design elements of the questionnaire, such as question form (open or closed), payment form, and whether the background information related to the valuation site is explained to the respondents, could not be readily verified across the majority of studies. This situation was also the reason why the above variables were excluded from the analysis.

Although more than 40% of the studies did not distinguish between respondent categories, the results showed that respondent identity significantly affected the valuation of ecosystem services. Qin et al. (2021) suggested that the respondents of stated preference surveys should be the main beneficiaries of the evaluated ecosystem services. This suggestion implies that tourists might be the respondents of interest for destination scenic areas, whereas local residents may be the respondents of interest for local amenities. Some studies (accounting for 6.9% of the sample) selected workers with professional backgrounds as the respondents. Given that professionals tend to be better educated than other respondents, they may have a deeper understanding of ecosystem functions and services and, therefore, tend to give a higher valuation (Marzetti et al., 2016).

The stated preference method obtains respondents’ willingness to pay through the description of a hypothetical scenario. Thus, the design of scenarios is a critical feature of a questionnaire, and poorly defined scenarios elicit meaningless answers (Lindemann-Matthies & Brieger, 2016; Lee et al., 2021). This work introduced hypothetical scenarios into the meta-analysis and found that designing different categories of scenarios can affect people’s willingness to pay. In the research sample, more than a quarter of the studies did not provide any information about the scenario design. The studies were classified into corresponding categories in accordance with the title of the papers and the focus of the study. Willingness to pay was found to be the lowest in the scenario of ecosystem compensation and the highest in the scenario of management for ecosystem protection. Willingness to pay for the latter was 1.4 times higher than that for the former. Most of the literature that did not provide scenario design information was classified into the category of continued ecosystem existence.

The willingness to pay for the scenario of continued ecosystem existence was 1.1 times higher than that for ecosystem compensation. However, the actual scenario design information of this category could not be verified. Studies have shown that the reliability of the survey results increases as the understanding of the respondents about the hypothetical situation deepens (Tao et al., 2012; Maldonado & Cuervo Sánchez, 2016). With increasing environmental awareness, people have become familiar with scenarios such as ecosystem protection and ecosystem improvement. However, ecosystem compensation is a concept that has emerged in China only in recent years, and the public has very limited understanding of this concept (Wan et al., 2021; Cui, 2022). Scenario rejection or scenario adjustment is likely to occur with insufficient information. This situation will reduce the effectiveness of ecosystem service valuation.

In addition, the lack of hypothetical scenario design information indirectly reflects the lack of policy application of the research on the valuation of marine ecosystem services in China. Most of the reviewed research could be argued to be aimed at academics instead of at solving practical problems or meeting policy needs. Guo & Kildow (2015) pointed out that most of the results of the research on the valuation of marine ecosystem services have not been put into practice, resulting in a serious gap between scientific research and policy application (Zhao, 2015; Torres & Hanley, 2017; Li & Wang, 2022). Studies that are oriented to meeting the needs of ecosystem management policies or solving actual ecosystem problems are likely to come up with reasonable and credible scenarios from which reliable valuation estimates can be obtained.

5 Conclusions

On the basis of empirical studies that estimated the stated preference values of marine ecosystem services in China from 1996 to 2021, this work constructed a database for use for value transfer. The key variables driving the variation of the value estimates of marine ecosystem services in China were explored by using meta-regression analysis. Value estimates in the literature were found to be affected by the ecosystem characteristics of the site under study, characteristics of research methods and technologies, and regional socioeconomic characteristics. In terms of ecosystem characteristics, the demand for marine ecosystem services in China displayed diminishing marginal utility, and high values were associated with coastal wetland ecosystems and recreational services. In terms of the technical characteristics of the research methods, the valuation results of the choice experiment method were higher than those of the CVM, and the identity of the respondents and the design of the hypothetical scenarios were factors that influenced the resulting valuation estimates. In terms of regional socioeconomic characteristics, income level had a positive effect on people’s willingness to pay.

In this study, the MLM was used for meta-regression analysis. A two-level model was constructed from the individual and regional levels, and individual samples were nested within regions. The statistical fit obtained by controlling for regional-level effects using the MLM was better than that acquired with the standard OLS model, which assumed that the data lacked a hierarchical structure. The variance associated with the regional-level residuals was of similar magnitude as the estimated variance for the overall error term in the MLM, indicating that factors associated with the regions wherein the studies were conducted are important for explaining the residual variation in the value of marine ecosystem service estimates. While the sample size used to estimate the models was relatively small, it still represented the majority of the stated preference study estimates currently available for such an analysis in China. It also reflected that the valuation of marine ecosystem services had received less attention than that of terrestrial ecosystem services in the country.

Some deficiencies may exist in the current approach to the evaluation of marine ecosystem services in China. Existing studies mainly focused on coastal wetlands, offshore bays, and islands and rarely involved other critical marine ecosystems indigenous to China, such as mangroves and seagrass beds. Furthermore, the absence of an established standardized system for ecosystem service valuation for feeding into policymaking affects valuation quality to a certain extent and the comparison of value estimates. The current interest in generating ecosystem accounting frameworks will also require a consistent valuation approach if estimates are to be used in such accounting systems.

In recent years, some major coastal provinces and cities in China have begun to explore the construction of a standardized valuation system for marine ecosystem services that would improve valuation quality. Echoing the call of Hynes et al. (2018), such systems should also require studies to report the results from standard ‘workhorse’ models from different types of valuation approaches (e.g., a conditional logit model for CE and a basic negative binomial model for a single-site travel cost model) because this approach should provide additional statistical control over the variation in values in future value transfer exercises. The quality and effectiveness of studies on the valuation of marine ecosystem services in China can be improved only by integrating valuation study results into marine ecosystem development and management decisions and by encouraging scholars to perform additional studies with practical applications and a basis on realistic policy needs.