Estimating the Value of Recreation and Ecotourism Using Meta-Regression Analysis

Abstract

Estimating the economic value of recreation and ecotourism is essential for sustainable ecosystem management and informed environmental policymaking. However, values derived from individual studies often vary because of subjective preferences and contextual variability, making it challenging to obtain generalizable estimates. To address this issue, this study employed a meta-regression analysis synthesizing 179 willingness-to-pay (WTP) observations obtained from 48 individual valuation studies conducted across various recreational and ecotourism sites in the Republic of Korea. Focusing specifically on national parks, which are prominent providers of cultural ecosystem services, we examined how site characteristics, study design factors, and valuation methodologies influenced estimated WTP values. Outliers were systematically identified and treated using statistical methods, with the random-effects model utilizing studentized residuals yielding the most robust results. Our findings revealed that national parks and studies employing the travel cost method (TCM) were associated with significantly higher WTP values. By applying the developed meta-regression model, we estimated that the total value of recreational and ecotourism services provided by national parks in the Republic of Korea was approximately USD 865.0 million in 2020. These results highlight the effectiveness of meta-regression analysis in synthesizing heterogeneous valuation studies, facilitating more accurate benefit transfers, and offering empirical insights to guide ecosystem service policy and management decisions.

1. Introduction

Cultural ecosystem services (CES) refer to the intangible benefits that humans derive from ecosystems or the interactions between humans and ecosystems [1,2]. Among these, recreation and ecotourism represent some of the most tangible cultural services derived from natural environments. Although CES include various subelements, this study focuses on recreation, such as nature-based leisure activities and ecotourism targeting natural areas. Recreation refers to activities that involve contact with nature in everyday life, such as walking, camping, and trekking, while ecotourism is defined as a responsible form of nature-based travel that aims to promote environmental conservation and support local communities [3].

As a significant category of cultural ecosystem services, recreational services contribute to improving individuals’ physical and mental health, while also offering an economic foundation for local communities and related industries [4]. Although both concepts are located within the broad framework of cultural services, they are distinguished from other services by their relatively high visibility, direct user participation, and measurable outcomes. These characteristics make recreation and ecotourism particularly suitable for empirical evaluation, especially monetary evaluation. Therefore, this study focuses on evaluating the value of recreation and ecotourism as representative subcategories of cultural services with relatively abundant data.

The sustainability of recreation and ecotourism is contingent upon the preservation of ecosystems and the wise use of environmental resources. Accordingly, it is essential to assess their value in a manner that reflects the multifaceted nature of cultural services. Cultural services are characterized by complex and layered structures in which components interact to generate diverse forms of value. Consequently, various methodological approaches—quantitative, qualitative, or a combination of both—can be employed depending on the characteristics of each service. However, services such as aesthetic enjoyment, inspiration, and cultural heritage, despite their high perceived value, often remain unmonetized because of the lack of available data and/or methodological tools [5].

Recreation and ecotourism are among the few cultural services that can be quantified in economic terms because of the relative availability of basic evaluation studies—especially when compared with abstract cultural values such as inspiration or spiritual meaning [6]. These characteristics have led to a relatively large number of valuation studies focused on recreation and ecotourism compared with other cultural services. However, most of these are case-specific studies, and thus it is difficult to generalize their findings to a broader context without a systematic synthesis. Moreover, as cultural services inherently involve subjective preferences, the derived values often vary among individuals, making it challenging to arrive at an objective valuation.

In conducting a meta-regression analysis, ensuring the reliability of the estimated relationships is critical—particularly when combining heterogeneous results across diverse study contexts. Outliers and overly influential observations can distort regression estimates [7]. Therefore, this study tries to identify and control outlier effects, thereby enhancing the reliability and objectivity of the meta-regression results. The application of such diagnostic procedures distinguishes this study from previous studies and contributes to improving the credibility of benefit transfer estimates based on meta-analytic models.

The purpose of this study is to comprehensively analyze domestic studies on the valuation of recreation and ecotourism to more accurately estimate the economic value of such services. To achieve this, we extracted willingness-to-pay (WTP) data from studies available in the Environmental Valuation Information System (EVIS), a comprehensive online database operated by the Korea Environment Institute, selecting those aligned with our research objectives. We then conducted a meta-regression analysis, taking into account the heterogeneity of study characteristics and methodologies.

While traditional meta-analysis aims to integrate effect sizes from multiple studies to estimate an overall average effect and evaluate consistency among results, meta-regression analysis has the characteristic of identifying heterogeneity [8]. In this study, we included research characteristics such as the type of location providing recreation and ecotourism (national parks), literature type, sample size, and valuation method as explanatory variables to analyze their influence.

This study also includes a meta-regression analysis of existing domestic studies that estimated the value of recreation and ecotourism. In doing so, we focus on national parks, which serve as prominent providers of such services and are distinguishable in terms of experiential value. Meta-regression analysis enables the synthesis of prior research findings and provides direction for future studies while avoiding the inefficiencies of redundant primary data collection in the valuation of nonmarket goods. Furthermore, it includes studies with conflicting results, reducing the likelihood of researcher bias. In this way, meta-regression can help address the inherent difficulty of objectifying cultural service values. Given that meta-regression results are recommended over single-study results for higher-level policy evaluation [9], this study seeks to provide policy implications based on its findings. In particular, we aim to quantify the factors that influence the value of services provided by national parks, which are representative environmental resources offering recreation and ecotourism and can be considered differentiated in terms of experiential aspects, and to provide empirical data that can be utilized for policy-making by estimating the total value of national parks using the derived meta-regression function.

2. Literature Review

Cultural ecosystem services (CES) include the intangible benefits that people obtain from ecosystems, such as recreation, aesthetic appreciation, spiritual enrichment, and cultural identity [1]. Recreation refers to leisure activities in natural environments, which contribute to mental recovery and physical health and provide an economic foundation for local communities [4]. Ecotourism is defined as a responsible form of travel that conserves natural areas while benefiting local communities [10]. The use of recreation and ecotourism in ecosystems is on the rise and constitutes an important component of the economies of many of the MA subglobal assessment study areas [1]. Recreation and ecotourism are most frequently analyzed because of their concrete and experience-based characteristics and the relative ease of monetary evaluation [6].

The valuation of CES values such as recreation and ecotourism has primarily relied on explicit preference methodologies, such as contingent valuation (CVM) and choice experiments (CE), to estimate willingness to pay (WTP). Numerous case studies have been conducted to estimate WTP for recreation and ecotourism in various ecological and regional contexts.

According to the systematic review by Kosanic and Petzold [11], recreation and tourism are among the most frequently analyzed subcategories, with study ecosystems including urban green spaces, coastal areas, marine environments, and forests. While these studies provide valuable insights into site-specific preferences and user behavior, differences in survey designs, sampling strategies, and site characteristics can influence WTP estimates.

When there is heterogeneity among studies, meta-analysis can be used as a methodology to statistically control and integrate individual study results to derive more generalized estimates. Meta-analysis is a statistical technique that synthesizes and analyzes results from prior studies to derive representative estimates. While it was initially utilized in fields such as medicine and education—where accumulated experimental results under controlled conditions are common—it began to be applied in economics with the pioneering work of Smith and Kaoru, and Stanley and Jarrell [12]. One of the primary advantages of meta-analysis is its ability to statistically control for heterogeneity among study subjects and methodological diversity during the synthesis process. Because of this, it has been widely adopted in the field of environmental valuation, particularly for estimating willingness to pay (WTP).

Meta-analysis is generally divided into two approaches: the pooled model, which estimates a combined value by aggregating similar studies, and meta-regression analysis, which conducts regression based on summary statistics and study/methodological characteristics from individual studies. If the studies under consideration exhibit high homogeneity, the pooled model is appropriate; otherwise, meta-regression analysis is recommended. In the meta-regression herein, WTP served as the dependent variable, while study population characteristics, site characteristics, and methodological variables were used as independent variables. This allowed us to quantitatively assess the influence of these independent variables on WTP [13].

Several studies have used meta-regression analysis to estimate the value of recreation and ecotourism. For example, Neher et al. [14] analyzed data from 58 distinct park-level surveys collected as part of the U.S. National Park Service (NPS) Visitor Services Project. The average WTP per visit across the entire NPS system in 2011 was USD 102, with park-level values ranging from USD 67 to 288. The total WTP for all NPS visitors in 2011 was estimated at USD 28.5 billion. The model they developed was then applied to other units within the NPS system to generate predicted average WTP by park, which were found to be consistent with prior studies.

Rosenberger et al. [15] conducted a meta-analysis estimating the value of individual recreational activities. Reviewing 2709 value estimates derived from 342 studies, they focused on 14 outdoor activities recognized by the USDA Forest Service’s NVUM program. The study found that the average recreational value in national forests was approximately USD 80 per day. The final WTP estimates varied depending on the type and base value of each activity, and the authors suggested that the total benefits of recreation could be estimated using these average values in combination with visit frequency and duration.

Ahn and Won [16] conducted a study estimating the economic value of urban parks through meta-regression analysis and benefit transfer. Similarly to the present study, they utilized data from the Environmental Valuation Information System (EVIS). Using 27 observations derived from 18 studies, they performed meta-regression and applied a transferred function to estimate the value of Busan Citizens Park. The predicted value was USD 11.86 per person per year, with a total value of approximately USD 40.3 million as of 2014.

Although Hynes et al. [5] did not directly estimate recreational value, they built a global database and conducted a meta-regression analysis to identify factors influencing the value of recreational services in marine and coastal ecosystems. Based on 311 value estimates drawn from 96 studies, they found that sites associated with multiple ecosystems tend to have higher recreational value. Additionally, the explanatory power of their models improved as they sequentially incorporated study-specific, location-specific, and cultural variables.

In a recent study, Pisani et al. [17] conducted a meta-regression analysis of over 180 case studies on the valuation of recreation and tourism (including ecotourism) published worldwide from 1975 to 2021, identifying key factors influencing willingness to pay (WTP). The study focused particularly on protected areas and natural parks. The results revealed a significant trend toward higher WTP in areas with higher legal status in protected areas. These findings can serve as a basis for policy decisions such as expanding protected areas or increasing public funding.

Taken together, these prior studies show that meta-regression analysis has been actively used to estimate the value of recreation and ecotourism and to facilitate benefit transfer to sites where primary valuation studies have not been conducted. Depending on the research objective, studies may incorporate specific recreational activities or cultural considerations as explanatory variables but generally include study characteristics and site-specific variables. Building on this foundation, the present study applies meta-regression to comprehensively examine the factors influencing the value of recreation and ecotourism and uses the derived function to estimate the recreational and ecotourism value of national parks in the Republic of Korea.

3. Methodology

3.1. Data

This study utilizes data provided by the Environmental Valuation Information System (EVIS). EVIS is an online platform operated by the Korea Environment Institute (KEI), which has been a government-sponsored national research institute under the Office of the Prime Minister since 2011. It systematically compiles and organizes metainformation on the monetary valuation of environmental services into a structured database and builds toolkits based on case studies of integrated environmental and economic analysis. These resources are used to support both ex ante and ex post evaluation of environmental policies and projects.

For this study, we reviewed 130 database entries related to ecosystem services in the categories of “recreation, leisure, and aesthetic appreciation” (studies conducted up to 2020). From these, we selected studies that estimated the value of recreation and ecotourism associated with visits to specific sites, with the unit of measurement expressed as Korean won per person per visit (KRW/visit/person).

To ensure the suitability of data for meta-analysis, we excluded studies that lacked essential information required for pooled models (e.g., standard errors, confidence intervals, t-values) or for standardizing value estimates (e.g., sample size, survey year). Additionally, studies conducted before the year 2000 were excluded if data on the average household size at the time of analysis were not available.

After applying these selection criteria, we identified 179 willingness-to-pay (WTP) observations from a total of 48 relevant studies (See

Table A1. The list of original studies used in the meta-regression analysis.

Id	Site Type	Year	Valuation Method	Sample Size	Literature Type	No. of WTP Estimates
1	Prov. Park 1 (3), N/A 2 (6)	2000	VT 5	204	Acad. J. 9	9
2	N/A	2000	CVM 6	412	Acad. J.	1
3	N/A	2000	CVM	800	Acad. J.	2
4	Natl. Park 3	2000	CVM	400–550	Acad. J.	6
5	Natl. Park	2001	CVM	309,350	Acad. J.	2
6	For.Rec.Area 4	2002	CVM	590	Acad. J.	1
7	Natl. Park	2002	TCM 7	26,76,48	Acad. J.	3
8	N/A	2003	CVM	511	Acad. J.	5
9	N/A	2003	TCM	600	Acad. J.	3
10	Natl. Park	2004	TCM	400	Acad. J.	1
11	Natl. Park	2005	CVM	312	Acad. J.	2
12	N/A	2005	CVM	313	Acad. J.	3
13	Natl. Park	2005	CE 8	323	Acad. J.	4
14	N/A	2005	CVM	1065	Acad. J.	1
15	Prov. Park	2005	CVM	1270	Diss. 10	2
16	Prov. Park	2005	CVM	407–420	Acad. J.	6
17	N/A	2006	TCM	100	Diss.	1
18	N/A	2006	TCM	168	Acad. J.	4
19	N/A	2006	CVM	399	Acad. J.	1
20	N/A	2006	CVM	800	Acad. J.	2
21	Natl. Park	2006–2012	CVM	255–810	Report	19
22	Natl. Park	2006–2012	CVM	255–810	Acad. J.	19
23	N/A	2007	TCM	517	Acad. J.	1
24	N/A	2007	CVM	1000	Acad. J.	1
25	N/A	2008	TCM	287	Diss.	1
26	Prov. Park	2008	TCM	325	Acad. J.	1
27	N/A	2008	CVM	370	Acad. J.	2
28	N/A	2008	CVM	400	Acad. J.	4
29	N/A	2008	CVM	405	Acad. J.	1
30	N/A	2008	CVM	530	Acad. J.	2
31	Natl. Park	2009	CVM	178, 225	Diss.	4
32	N/A	2009	CVM	332	Diss.	3
33	N/A	2010	CE	400	Acad. J.	4
34	N/A	2010	CVM	540	Acad. J.	1
35	N/A	2011	CE	234	Acad. J.	6
36	Prov. Park	2012	TCM	81–435	Acad. J.	7
37	N/A	2013	TCM	300	Acad. J.	2
38	N/A	2013	TCM	300	Acad. J.	2
39	N/A	2013	CVM	400	Acad. J.	1
40	N/A	2014	TCM	284, 288, 572	Acad. J.	12
41	N/A	2015	CVM	467–1489	Acad. J.	4
42	N/A	2016	CVM	455	Acad. J.	1
43	N/A	2017	CVM	1000	Acad. J.	9
44	Natl. Park	2018	CVM	237	Acad. J.	3
45	N/A	2018	CVM	700	Acad. J.	3
46	N/A	2020	CVM	105	Acad. J.	1
47	For.Rec.Area	2020	CVM	500	Acad. J.	6

¹ Provincial park, ² not applicable, ³ national park, ⁴ forest recreation area, ⁵ value transfer, ⁶ contingent valuation method, ⁷ travel cost method, ⁸ choice experience, ⁹ academic journal, ¹⁰ dissertation. Source: Environmental Valuation Information System (EVIS).

). The variables extracted from each study are park type, literature type, sample size, valuation method, and year of study implementation. All amounts were converted to USD based on the average exchange rate in 2020 (1 USD = 1180 KRW). All statistical analyses were conducted using Stata 16.0.

The descriptive statistics of the dataset used in the analysis are presented in

Table 1. Descriptive statistics of variables used in the meta-regression analysis.

Category		Frequency Willingness to Pay (WTP)		(USD/Visit·Person)
		Freq.	Percent	Mean	Std. Dev.
Type of Site	National Park	63	70.79	17.73	22.56
	Provincial Park	19	21.35	51.79	69.53
	County Park	not applicable
	Forest Recreation Area	7	7.87	12.01	16.59
Type of Literature	Academic Journal	151	84.36	50.47	81.91
	Dissertation	9	5.03	16.69	15.03
	Report	19	10.61	11.69	3.05
Sample Size	≥1000	14	7.18	23.09	50.88
	<1000	165	92.18	46.48	78.11
Valuation Method	CVM	117	65.36	18.35	32.04
	CE	14	7.82	5.45	2.61
	TCM	39	21.79	140.49	109.09
	VT	9	5.03	32.30	31.12

Source: Appendix A Table A1.

. Among the 179 observations, 89 were related to natural parks (including national, provincial, and county parks) and forest recreation areas. Of these, 63 observations (70.79%) were based on studies targeting national parks, making it the most frequently studied category. In terms of publication type, academic journal articles accounted for the majority (84.36%, or 151 observations). Regarding sample size, the vast majority of observations (92.18%, or 165 observations) were based on samples with fewer than 1000 respondents. As for valuation method, the contingent valuation method (CVM) was the most commonly used, accounting for 65.36% (117 observations) of the total.

Outliers in the dependent variable—here, willingness to pay (WTP)—can lead to over- or underestimation of analytical results. To address this, the present study applied two outlier detection and removal methods. First, we excluded WTP values falling within the top and bottom 2.5%, as recommended by the OECD [18]. Second, we excluded observations of which the studentized residuals had absolute values greater than or equal to 2, in accordance with the approach suggested by Barnett and Lewis [19]. Studentized residuals are commonly used in regression diagnostics, and observations with absolute values of 2 or more are typically considered outliers.

Figure 1 illustrates the distribution of WTP values and the identified outliers based on this criterion.

Table 2. Sample composition after outlier removal.

Category	Full Sample	After Outlier Removal (Method 1) 1	After Outlier Removal (Method 2) 2
Number of Observations	179	170	169
Number of Studies 3	48	47	43

¹ WTP values in the top and bottom 2.5% were excluded [18]. ² WTP observations with studentized residuals having an absolute value of 2 or greater were excluded. ³ While the number of observations may remain the same, the number of studies can vary depending on the method used to define outliers. Source: authors.

presents the final sample composition after outlier treatment.

The variables constructed for the meta-regression analysis are summarized in

Table 3. Summary of variables used in the meta-regression analysis.

Variable	Description	Mean	Std. Dev.
ln_esti	Log of WTP for ecosystem services (KRW/person/visit, constant 2020 prices)	9.788	1.478
NationalPark	Dummy = 1 if the study site was a national park; 0 otherwise	0.352	0.479
Year	Year of study implementation (2000–2020); year dummy variables	-	-
Journal	Dummy = 1 if the publication was an academic journal article; 0 otherwise	0.844	0.364
Sample_1000	Dummy = 1 if sample size was 1000 or more; 0 otherwise	0.078	0.269
CVM	Dummy = 1 if valuation method was contingent valuation method (CVM); 0 otherwise	0.654	0.477
CE	Dummy = 1 if valuation method was choice experiment (CE); 0 otherwise	0.078	0.269
TCM	Dummy = 1 if valuation method was travel cost method (TCM); 0 otherwise	0.218	0.414
VT	Dummy = 1 if valuation method was value transfer (VT); 0 otherwise	0.05	0.219

Source: authors.

. Through a standardization process, the WTP values reported in each individual study were converted to constant 2020 Korean won, and this standardized value was used as the dependent variable.

Dummy variables were included as explanatory variables to capture differences across study and site characteristics: a dummy for national parks (to distinguish them from other types of sites), year dummies (to account for the time of study implementation), dummy variables for publication type and sample size (to identify heterogeneity across literature and study scale), and dummy variables for valuation methodology (to test for differences in WTP by method used).

The reference category for publication type was set as academic journal articles, and the threshold of 1000 respondents—commonly used in nationally representative CVM sampling designs in the Republic of Korea [20]—was used as the criterion for the sample size dummy. Valuation methods were categorized into four groups: contingent valuation method (CVM), choice experiment (CE), travel cost method (TCM), and value transfer (VT).

3.2. Analytical Model

Meta-regression analysis is a method that uses summary statistics from previous studies as the dependent variable and incorporates study-specific and data-related characteristics as explanatory variables. The standardized outcome derived from prior research is referred to as the effect size; in this study, the effect size corresponds to the standardized willingness to pay (WTP). This WTP serves as a comparable indicator of the economic value of recreational ecotourism services in the study.

Depending on the assumptions made about the population and parameter variation of the summary estimates, meta-regression models are generally categorized into two types: the fixed effect model and the random effect model.

The fixed effect model assumes that all studies share a common population and a single true effect size, meaning the observed variation in WTP estimates across studies arises solely from sampling error. The basic specification of the fixed effect model is as follows:

$$Y_i=\beta +e_i,i=\mathrm{1,2},\dots ,n$$

(1)

In the fixed effect model, it is assumed that all studies share the same underlying population and parameter for the effect size. Therefore, any variation in estimated values across individual studies arises solely from within-study variance (${\mathsf{\sigma }}_i^2$). A smaller variance indicates a more precise estimate of the true effect, and thus, studies with lower variance are assigned greater weight. The inverse of the within-study variance is used as the weight to calculate a weighted average.

The weighted mean and variance under the fixed effect model are given by the following equations:

$$\overline{\beta }={\displaystyle \frac{\sum w_i{Y}_i}{\sum w_i}},\hspace{1em}{\widehat{\sigma }}^2={\displaystyle \frac{1}{\sum w_i}}$$

(2)

In contrast, the random effect model assumes that the true effect size of each individual study is not fixed but rather drawn from a normal distribution with an unobserved mean and variance. In this framework, individual studies are considered to be randomly sampled from a population of studies distributed around a population mean effect size. The variation in effect sizes across studies is therefore attributed to both sampling error and between-study variance.

Accordingly, the total variance in the random effect model consists of the sum of the within-study variance (sampling error) and the between-study variance. The basic specification of the random effect model is as follows:

$$Y_i={\beta }_0+e_i+u_i,\hspace{1em}i=\mathrm{1,2},\dots ,n$$

(3)

Ultimately, the choice between the fixed effect model and the random effect model in meta-analysis depends on the nature of the error structure in the effect size—in this case, the willingness to pay (WTP). To assess the degree of heterogeneity across studies, Cochran’s Q test and Higgins’s I² statistic can be employed.

Cochran’s Q test indicates the presence of heterogeneity when the p-value is less than 0.1. Higgins’s I² statistic categorizes heterogeneity as low when below 25%, moderate at approximately 50%, and high when exceeding 75%. If the effect sizes are highly homogeneous, the fixed effect model is appropriate. When heterogeneity is present, the random effect model is preferred. In cases of substantial heterogeneity, a meta-regression analysis is recommended to explore the sources of variation. In this study, a meta-analysis was conducted using 179 effect sizes. The Cochran’s Q statistic was 9726.3 (df = 178, p < 0.001), and the Higgins’s I² statistic was 98.2%, indicating a very high level of heterogeneity among the included studies. Accordingly, a random-effects model was applied, and a meta-regression analysis was conducted to explore potential sources of heterogeneity. This section refers to Chapter 3 of Ahn et al. [21] for the meta-regression modeling framework. For more details, please refer to the reference.

4. Results

The dataset used for meta-regression exhibited a structure similar to panel data, as the number of WTP observations collected from each study varied. This resulted in a likely unbalanced panel, which in turn increased the probability of violating the assumption of homoscedasticity in the error terms. Therefore, prior to estimation, a Hausman test was conducted to determine whether a fixed effect or random effect model would be more appropriate.

The fixed effect model assumes that unobserved heterogeneity across studies is structurally unrelated to the explanatory variables, whereas the random effect model treats this heterogeneity as a random variable that may be correlated with the explanatory variables. The null hypothesis of the Hausman test is $\mathrm{cov}\left({\mathsf{\sigma }}_i^2,\hspace{0.17em}{u}_i\right)=0$, indicating no correlation. If the null hypothesis is rejected, the fixed effect model is preferred; otherwise, the random effect model is considered more suitable.

The results of the Hausman test indicate that the null hypothesis could not be rejected at the 10% significance level for either the full sample or the outlier-adjusted sample, suggesting that the random effect model was appropriate for this analysis (see

Table 4. Results of the Hausman Test.

Specification	Full Sample	Outlier Removal–Method 1 2	Outlier Removal–Method 2 3
χ2 Statistic (p-value)	5.08 (0.6499) 1	3.87 (0.7945) 1	2.07 (0.9560) 1

¹ Values represent χ² statistics; numbers in parentheses indicate p-values (Prob > χ²). ² Outliers were removed by excluding the top and bottom 2.5% of WTP values. ³ Outliers were removed based on studentized residuals with absolute values ≥ 2. Source: Authors.

).

The meta-regression estimates using the random effect model are presented in

Table 5. Results of meta-regression analysis.

Category	Full Sample		Outlier Removal Method 1 1		Outlier Removal Method 2 2
	Coef.	Std. Err.	Coef.	Std. Err.	Coef.	Std. Err.
National Park	0.498	0.552	0.401	0.483	0.658 **	0.312
2000	0.037	1.235	0.075	1.079	0.027	0.698
2001	0.784	1.722	0.882	1.504	0.645	0.973
2002	1.423	1.405	1.593	1.229	1.301	0.805
2003	2.225	1.343	2.337 **	1.172	2.248 ***	0.749
2004	−1.422	1.858	−1.101	1.628	−1.572	1.086
2005	−0.124	1.134	0.103	0.996	−0.699	0.653
2006	−0.004	1.064	0.493	0.944	0.172	0.628
2007	0.394	1.075	0.742	0.947	0.418	0.635
2008	0.383	1.054	0.663	0.925	0.245	0.616
2009	0.652	1.069	1.017	0.943	0.687	0.63
2010	0.627	1.1	0.957	0.969	0.635	0.655
2011	0.665	1.083	1.028	0.955	0.649	0.644
2012	0.481	1.053	0.849	0.928	0.52	0.613
2013	0.663	1.282	0.816	1.121	0.645	0.732
2014	2.215	1.673	2.229	1.467	2.226 **	0.922
2015	1.812	1.607	1.842	1.402	1.959	0.893
2016	0.266	1.689	0.267	1.478	0.287	0.986
2017	−0.269	1.685	0.228	1.474	0.638	0.968
2018	−0.161	1.349	−0.111	1.177	−0.22	0.755
2020	reference
Journal	0.364	0.727	0.616	0.641	0.44	0.425
Sample 1000	0.283	0.575	0.169	0.513	−0.221	0.425
CVM	reference
CE	−0.663	0.849	−0.937	0.746	−0.506	0.484
TCM	1.616 ***	0.548	1.393 ***	0.485	1.626 ***	0.323
VT	1.154	1.5	1.117	1.307	1.504	0.822
cons	8.425	1.192	8.172	1.045	8.327	0.681
R2	0.67		0.64		0.81
N (group)	179 (48)		170 (47)		173 (45)

*** p < 0.01, ** p < 0.05. ¹ Outliers removed based on top and bottom 2.5% of WTP values. ² Outliers removed based on studentized residuals with absolute values ≥ 2. Source: authors.

, and the results can be interpreted as follows.

First, the national park dummy variable showed a statistically significant positive coefficient in the sample processed using Outlier Removal Method 2, indicating that WTP values for national parks are higher than those for other site types. As shown in the descriptive statistics in Table 1, 63 out of 89 nature-related observations (70.8%) were from national park studies, suggesting both a higher research interest in national parks and that sites with more research tend to show higher WTP estimates. However, it is important to note that WTP does not necessarily reflect the intrinsic ecological value of a site, but rather the value assigned through human experience.

Regarding the year variables, the year 2003 showed statistically significant coefficients in the sample with Outlier Removal Method 1, while both 2003 and 2014 were significant in the Method 2 sample.

In terms of publication type, academic journal articles were positively correlated with WTP compared with other literature types, although this relationship was not statistically significant.

For the sample size variable, studies with a sample size of 1000 or more tended to show higher WTP in the full sample and Method 1 sample, whereas the opposite trend was observed in the Method 2 sample. However, none of these coefficients were statistically significant.

As for valuation methods, both the travel cost method (TCM) and value transfer (VT) showed higher WTP values relative to the reference category, contingent valuation method (CVM). Between them, only TCM produced statistically significant coefficients across all samples.

Regarding model explanatory power, the adjusted R² was highest in the sample processed by Method 2 (0.81), followed by the full sample (0.67) and the Method 1 sample (0.64). The statistical significance of the results also improved noticeably after applying Outlier Removal Method 2. Although it is not possible to generalize which outlier treatment method is universally superior, the findings suggest that excluding outliers based on studentized residuals was relatively more suitable for estimating the recreational and ecotourism value in this study.

In addition, the meta-regression function estimated in this study was used to calculate the aggregate willingness to pay (WTP) for all national parks in the Republic of Korea. As of 2020, there were 22 national parks officially designated in the country, and visitor statistics for these parks are publicly available. Based on the WTP estimates derived from the meta-regression model, the total value of recreational and ecotourism services provided by national parks was estimated at USD 865.0 million for the year 2020 (see

Table 6. Comparison of estimated recreational and ecotourism values of national parks in the Republic of Korea (2018–2022).

Application of Meta-Regression Results
	2018	2019	2020	2021	2022
WTP 1 (unit: USD)	24.5
Number of Visitors (unit: 10,000 persons)	4382	4318	3528	3590	3627
Total Value 2 (unit: USD million)	1074.6	1058.9	865.0	880.3	889.4
Value Transfer Using Previous Study by Sim et al. [22]
	2018	2019	2020	2021	2022
WTP 3 (unit: USD)	9.1–11.9
Number of Visitors (unit: 10,000 persons)	4382	4318	3528	3590	3627
Total Value 2 (unit: USD million)	502.0	494.5	404.8	410.6	413.3

¹ WTP derived using the meta-regression function, adjusted to 2020 constant prices. ² Total value = WTP × number of visitors. ³ WTP values transferred from Sim et al. [22], adjusted to 2020 prices. Values were categorized by park type (mountain, marine, and urban) and calculated using the average WTP for each category. Source: authors.

).

For comparison, a previous study by Sim et al. [22], which assessed the recreational value of national parks, reported per-visit WTP estimates ranging from USD 4.5 to USD 17.0. When these values were applied using a benefit transfer approach, the total recreational and ecotourism value of the Republic of Korea’s national parks in 2020 was estimated to be USD 404.8 million.

Although the estimates from the single-study approach and the meta-regression model differ, these results can be interpreted together to suggest a reasonable valuation range. Specifically, the total recreational and ecotourism value of national parks in 2020 is estimated to fall between USD 404.8 million and USD 865.0 million.

Beyond these results, it is also possible to estimate a broader range of values by incorporating alternative unit values such as minimum costs associated with experience (e.g., entrance or parking fees), average expenditures derived from travel cost methods, or WTP values obtained through other valuation approaches.

5. Discussion

This study involved a meta-regression analysis based on prior research that estimated the value of recreational and ecotourism services in the Republic of Korea. The derived model was then applied to national parks, which are representative destinations for such experiences. A total of 179 willingness-to-pay (WTP) estimates extracted from 48 relevant studies—collected via the Environmental Valuation Information System (EVIS)—were utilized in the analysis. To minimize estimation errors, outliers in the dependent variable (WTP) were addressed prior to analysis. Specifically, outliers were removed using two approaches: (1) excluding the top and bottom 2.5% of WTP values and (2) excluding observations with studentized residuals greater than 2 in absolute value. At this point, the estimated WTP was USD 29.5 in the model without any outlier treatment, USD 26.7 in the model with the top and bottom 2.5% of values trimmed, and USD 24.5 in the model with outliers removed based on studentized residuals. Notably, the national park variable—a key explanatory factor—was statistically significant only in the model using studentized residual diagnostics. These findings are consistent with Viechtbauer and Cheung [7], who noted that outliers can distort WTP estimates, that diagnostics such as studentized residuals are useful for identifying problematic observations, and that removing outliers enhances the explanatory power and interpretability of meta-analytic models.

The results revealed that both the national park variable and the valuation method variable had statistically significant effects on WTP. Studies targeting national parks reported higher WTP values compared with those targeting other sites, and those employing the travel cost method (TCM) also yielded higher WTP estimates relative to those using the contingent valuation method (CVM), which served as the reference category. These findings suggest that differences in park type and methodological approach contribute to the variation in estimated values.

Among the samples analyzed, the model using studentized residuals for outlier removal demonstrated the highest explanatory power. Moreover, applying the estimated meta-regression function to all national parks in the Republic of Korea yielded an annual recreational and ecotourism value of USD 865.0 million as of 2020.

The results of this study are consistent with existing meta-analyses in the field of environmental value assessment. For example, Hynes et al. [5] stated that evaluation methods, ecosystem types, and site specificity influence WTP estimates in marine and coastal recreational value assessments. Similarly, Neher et al. [14] and Rosenberger et al. [15] emphasize that heterogeneity in WTP is influenced by site characteristics and survey methodology.

However, it is important to note that this does not imply that national parks are inherently more valuable than other natural areas, or that TCM is a superior valuation method; rather, it reflects differences in perceived or stated values based on method and site context.

In the case of the national park variable, this finding aligns with Neher et al. [14], which also reported higher WTP for designated national parks. The higher WTP observed for national parks in this study may be due to the fact that these areas are designated and managed by the government for the preservation of natural and cultural heritage in the Republic of Korea and therefore may offer higher levels of satisfaction in terms of recreational and ecotourism experiences.

For the valuation method, this may be attributed to methodological differences: revealed preference methods such as TCM reflect actual expenditures and the full recreational experience of visitors, whereas stated preference methods like CVM and CE tend to isolate values for individual attributes [23]. While it is difficult to determine which WTP value more accurately reflects true economic value, the comparison of values derived from these two approaches can help establish a reasonable range for estimating the recreational and ecotourism value of national parks.

This study highlights the practical utility of meta-regression analysis as a tool for integrating distributed valuation studies and deriving robust and generalizable estimates for nonmarket environmental goods. By systematically identifying the determinants of WTP for recreation and ecotourism and applying them to the context of Korean national parks, this study provides an empirical basis for policy evaluation and resource allocation. The ability to integrate heterogeneous research results and reduce researcher subjectivity is particularly valuable in the field of cultural ecosystem services, where subjective preferences and contextual variability pose significant challenges to valuation. Therefore, this study not only contributes to academic discourse but provides concrete, data-driven insights that can support high-level policy planning and enhance the efficiency of future valuation efforts.

6. Conclusions

As interest in recreation and ecotourism—both closely linked to human well-being—continues to grow, research estimating the value of these activities provides an important foundation for the development of informed management policies. This study emphasizes the economic importance of recreation and ecotourism as core components of cultural ecosystem services (CES) through meta-regression analysis. By identifying the determinants of willingness to pay (WTP) and applying them to national parks, this study provides methodological insights and practical tools for ecosystem service valuation in Korea.

This study demonstrates the feasibility of conducting nuanced value assessments that account for detailed factors such as survey methods and regional characteristics, moving beyond policy approaches based solely on simple mean estimates. The influence of the variables identified in this study can be used as reference material for standardized survey design and survey method selection in future ecosystem service value assessment studies. In addition, when commissioning related studies or constructing policy evaluation data in the public sector, it can provide practical criteria for reflecting structural factors that allow for meta-analysis in advance.

This study also contributes to the field by synthesizing existing valuation research through meta-regression and by estimating the value range of national parks’ recreational and ecotourism services. These results demonstrate the utility of meta-regression analysis as a decision-support tool for natural park policy and planning. This study suggests the possibility of a methodological approach using meta-regression techniques in future research on ecosystem service values and lays the foundation for integrated comparison and interpretation of existing individual studies.

The study focused exclusively on national parks as representative ecosystems offering recreational and ecotourism services, which may limit the generalizability of the results to other types of sites. However, as shown in Table 1, national parks account for the majority of related studies in the Republic of Korea, and they are the only park type with official visitor statistics available for all sites. For meta-regression analysis to be both valid and reliable, an adequate quantity and quality of prior studies is necessary [24]. In this respect, the findings of this study may serve as a useful reference for future research. For future expansion of the analysis to other areas, additional studies on various types of natural parks must be conducted, and visitor statistics for those sites need to be systematically collected through national data systems. Once a sufficiently comprehensive database of nonmarket valuation studies on recreation and ecotourism services has been compiled, an updated meta-regression analysis should be undertaken to improve the reliability of the present findings.