The Temporal Spillover Effect of Green Attribute Changes on Eco-Hotel Location Scores: The Moderating Role of Consumer Environmental Involvement

Abstract

This study focuses on a profound paradox in eco-hotel evaluations: why do consumer ratings for location, a static asset, exhibit dynamic fluctuations? To solve this puzzle, we construct a two-stage signal-processing theoretical framework that integrates Signaling Theory and the Elaboration Likelihood Model (ELM). This framework posits that the dynamic trajectory of a hotel’s green attributes (encompassing eco-facilities, sustainable practices, and ecological experiences) constitutes a high-fidelity market signal about its underlying quality. We utilized natural language processing techniques (Word2Vec and LSA) to conduct a longitudinal analysis of over 60,000 real consumer reviews from Booking.com between 2020 and 2023. This study confirms that continuous improvements in green attributes result in significant positive spillovers to location scores, while any degradation triggers strong negative spillovers. More critically, consumer environmental involvement (CEI) acts as an amplifier in this process, with high-involvement consumers reacting more intensely to both types of signals. The research further uncovers complex non-linear threshold characteristics in the spillover effect, subverting traditional linear management thinking. These findings not only provide a dynamic and psychologically deep theoretical explanation for sustainable consumption behavior but also offer forward-thinking practical implications for hoteliers on how to strategically manage dynamic signals to maximize brand value.

1. Introduction

Driven by the global wave of sustainable development, eco-hotels, as a core vehicle for the hotel industry’s transition to a green economy, are of increasing market value and research significance. They are not only commercial entities that meet the needs of environmentally conscious consumers but also a frontier for observing the dynamic interplay between modern consumption ethics, corporate social responsibility, and market performance. Within the complex evaluation system of eco-hotels, the location score has historically been regarded as a key evaluation metric. It is directly linked to a hotel’s pricing strategy, profitability, and brand reputation and is typically considered a core, relatively fixed asset. However, a profound paradox arises from this: why does a seemingly static attribute determined by geographical coordinates experience dynamic fluctuations in its consumer ratings over time? Unlike other variable service attributes, geographical location is a hotel’s most core and fixed physical asset. Therefore, when consumer ratings for this static asset exhibit dynamic fluctuations, it reveals a profound paradox: this is not a simple evaluation error, but suggests the consumer’s value judgment framework is being reshaped. Investigating which dynamic signals (such as the evolution of green attributes) can influence the evaluation of this core static asset is not only theoretically challenging but also points to the foundational logic of brand value creation. This is the fundamental reason why this study chooses the location score as its core dependent variable. This reveals that the value generation mechanism of location scores is far more complex than traditionally understood, concealing an underlying psychological and behavioral logic that has not been fully explained.

Currently, academic research on the spillover effect in hotel evaluations—the phenomenon where the evaluation of one attribute influences that of another—still has significant theoretical gaps. Previous studies have failed to effectively distinguish between static evaluation and dynamic signal processing. For instance, research by Han et al. [1] and Jones et al. [2] tends to analyze green attributes as a fixed label or static driver in cross-sectional studies, overlooking their unique value as a signal flow that evolves over time. Even as studies like that of Karaman et al. [3] begin to address evaluative spillover effects, they have not deeply explored how a signal’s ‘dynamic trajectory’ (i.e., continuous improvement or degradation) itself acts as a more potent form of information, driving consumers to re-evaluate a completely unrelated static asset like geographical location. In other words, the existing literature generally lacks a dynamic perspective to capture and explain the deep mechanism of this cross-category spillover effect. Therefore, investigating what factors, and through which mechanisms, lead to the dynamic changes in the core metric of location score over time constitutes a research gap that urgently needs to be filled.

To unravel the mystery of the dynamic changes in location scores, this study proposes examining a dynamic service attribute, green attribute changes (including eco-facilities, sustainable practices, and ecological experience), as the core independent variable. We argue that the key to understanding this issue is not the presence or absence of green attributes, but rather their improvement or degradation over time. A hotel’s continuous improvement or decline in green attributes is a highly diagnostic dynamic signal released to the market. This dynamic change transcends the attribute itself, becoming a critical clue for consumers to infer the hotel’s overall quality, management level, and value commitment. Therefore, examining the dynamic evolution of green attributes is a necessary path to explaining the phenomenon of fluctuating location scores.

However, when faced with this dynamic green information, consumers often find themselves in a decision-making dilemma: how to weigh a hotel with an outstanding environmental philosophy but a remote location? How to determine if a hotel’s environmental promises are sincere or mere greenwashing? To overcome this challenge, this study introduces consumer environmental involvement (CEI) as a moderating variable. CEI reveals individual heterogeneity in consumer information processing. It explains why different consumers, faced with the same green signal, produce evaluation spillovers of vastly different intensities, and sometimes even in opposite directions. By examining CEI, we can open the black box of consumer decision-making to understand how they use their motivations and knowledge reserves to process complex information, resolve internal conflicts, and ultimately form a judgment on the value of a hotel’s location.

Based on this, this study constructs an integrated “signal-processing” two-stage theoretical model to systematically explain the aforementioned influence mechanism. First, we apply Signaling Theory to explain why the spillover effect occurs: the dynamic changes in green attributes constitute market signals about the hotel’s intrinsic qualities (such as management level and social responsibility), which then spill over to the consumer’s value assessment of its geographical location by altering their overall perception and trust. Second, we use the Elaboration Likelihood Model (ELM) to clarify how the spillover effect works: the consumer’s environmental involvement determines whether they use a central or peripheral route to process these green signals, thereby moderating the final intensity of the spillover effect. This integrated model provides a complete causal chain to explain the dynamic formation mechanism of eco-hotel location scores.

To systematically address the aforementioned research gaps, this study aims to answer the following core research questions:

RQ1:

How do dynamic changes (improvements or degradations) in an eco-hotel’s green attributes (eco-facilities, sustainable practices, ecological experience) affect consumer ratings of its location, a static attribute, through a temporal spillover effect?

RQ2:

How does consumer environmental involvement moderate the intensity of this temporal spillover effect? Specifically, do high-CEI consumers exhibit stronger reactions to both positive and negative changes in green attributes?

RQ3:

What are the underlying characteristics of this spillover effect? Does it follow a simple linear pattern, or does it exhibit more complex, non-linear threshold characteristics?

In summary, this study is dedicated to theoretically integrating Signaling Theory and the ELM and applying them to the sustainable hotel discourse, aiming to test and innovate the explanatory power of these two theories in dynamic, digital consumption scenarios. Practically, this study aims to provide evidence-based strategic recommendations for hotel managers and online travel platforms, revealing how to strategically enhance comprehensive market performance, including location scores, by managing their green signals.

2. Literature Review

2.1. Eco-Hotel Evaluation and Green Attributes Research

With the global popularization of the sustainable development concept, the eco-hotel has become a hot topic in hospitality research. Scholars have pointed out that green attributes have evolved from a marginal differentiation strategy to one of the core factors influencing consumer choice [4]. Existing research generally confirms that implementing green management practices can significantly enhance market competitiveness [3]. These green attributes can typically be examined from multiple dimensions. This study focuses on three core aspects: eco-facilities, sustainable practices, and ecological experience.

To systematically capture the complete picture of green attributes, this study, based on a review of the existing literature, categorizes a hotel’s green efforts into three core and complementary dimensions: eco-facilities, sustainable practices, and ecological experience. This classification framework was not chosen arbitrarily; rather, it is because it comprehensively covers the entire process from tangible physical investments and intangible operational processes to the final perceived customer value, a multi-dimensional approach that is widely recognized in sustainable tourism research [2,4].

First, eco-facilities are the most direct and fundamental manifestation of a hotel’s environmental commitment, forming the physical basis of its green attributes. A significant body of literature focuses on tangible hardware and technical features, such as energy-saving lighting, water-saving appliances, the use of renewable energy, and environmentally friendly building materials [5]. These facilities are hard signals that consumers can directly observe and verify, and are thus widely regarded as a cornerstone for measuring a hotel’s environmental investment.

Second, sustainable practices reflect the depth and breadth to which a hotel integrates environmental concepts into its daily operations, representing its intangible managerial commitment. Research centers on non-physical processes in hotel operations and management, including waste sorting and recycling, green supply chain procurement, and environmental training for employees [5]. Compared to one-time facility investments, these ongoing practices better reflect a hotel’s authentic, long-term commitment to sustainability and are a key dimension for assessing its environmental sincerity (and avoiding greenwashing).

Finally, the ecological experience is the crucial measure of whether the aforementioned hardware and software investments can ultimately translate into a unique, perceivable value for the customer, constituting the ultimate outcome of green attributes. This dimension focuses on the extent to which a hotel’s environmental philosophy ultimately creates a unique stay experience for the guest, such as a sense of comfort, a tranquil atmosphere, and a feeling of connection with nature derived from its environmental ethos [6]. The formation of this experience also reflects the consumer’s role as a value co-creator in the service process [7] and ultimately influences their satisfaction and loyalty. Therefore, any assessment of green attributes would be incomplete without considering the experiential dimension.

At the consumer behavior level, numerous studies have shown that the consumer’s perception of a hotel’s green attributes significantly influences their intention to stay, satisfaction, and loyalty [8]. In particular, when a hotel’s green practices align with a consumer’s personal values (such as environmental awareness), it is more likely to evoke positive emotional responses and a higher willingness to pay [1], which is consistent with Perceived Value Theory [9]. However, the existing literature, when discussing green attributes, mostly adopts a static perspective, treating them as a fixed label (e.g., whether it has a certain eco-certification), and seldom focuses on how these attributes’ dynamic changes over time (such as service improvements or facility aging) affect consumers’ continuous evaluations.

2.2. Hotel Location Score and Spillover Effect Research

In the traditional evaluation system of the hotel industry, location has always been considered one of the most critical factors determining a hotel’s value. It directly affects the hotel’s customer base, pricing, and profitability and is generally regarded as an objective and relatively fixed physical asset. Therefore, research on location scores has mostly focused on its direct impact on the hotel’s financial performance. At the same time, the evaluation spillover effect, as an important phenomenon in consumer psychology, has also received attention in hotel evaluation research.

A notable recent trend is the use of sentiment analysis techniques to investigate spillover effects in online reviews. Online reviews, as a critical form of eWOM, have been shown to profoundly influence consumers’ hotel booking intentions [10,11]. Building on this, the study by Karaman et al. [3] clearly quantified this spillover effect through an empirical analysis of over 100,000 hotel reviews. For instance, their research found that a one-standard-deviation increase in the cleanliness rating significantly leads to a 0.15-standard-deviation increase in the service rating. This finding specifically confirms that a positive evaluation in one dimension (e.g., cleanliness) spills over to enhance consumer perceptions of other seemingly unrelated dimensions (e.g., service). However, while their work is foundational, it did not specifically address the unique case of cross-category spillover from dynamic green attributes to a static physical asset like location. Furthermore, although earlier service marketing research established the general halo effect, its specific mechanism in the context of sustainable consumption, especially the dynamic mechanism driven by online reviews, remains underexplored.

2.3. The Role of Consumer Environmental Involvement

To explain the heterogeneous responses of consumers to green information, the concept of consumer environmental involvement is crucial. Vermeir and Verbeke [12] defined CEI as the personal relevance and importance an individual attaches to environmental issues. Subsequent research, such as that by Yadav and Pathak [13], has consistently demonstrated that high-CEI consumers are more proactive in processing green information and show stronger green purchase intentions. The work of Kim and Choi [14] also found a significant positive relationship between consumers’ environmental concerns and their pro-environmental behaviors. However, these studies have mostly focused on purchase intentions or general behaviors and have not fully explored how CEI moderates the evaluation spillover effect in response to dynamic changes in green signals, which is a key focus of our study.

2.4. Research Gaps and Contributions of This Paper

A critical review of the literature reveals several significant gaps this study aims to address. First, existing research primarily assesses the impact of green attributes from a static perspective. Studies by Han et al. [1] and Jones et al. [2], for example, treat green attributes as a fixed label, thereby overlooking the shaping effect of their dynamic changes (improvement or degradation) on consumers’ long-term, continuous evaluations. As noted by Xiang et al. [15] in their review of social media analytics in tourism, there is a need for more dynamic, longitudinal research to capture complex consumer behaviors. Second, the psychological mechanisms and causal pathways behind the evaluation spillover effect, especially the cross-category spillover from dynamic service attributes (green practices) to static physical assets (geographical location), have not been clearly and systematically elucidated or empirically tested. Finally, the specific role of consumer environmental involvement in moderating this dynamic spillover process remains unclear, leading to an insufficient understanding of why different consumers exhibit systematic evaluation differences. This study aims to fill these gaps. By constructing a “signal-processing” two-stage model that integrates Signaling Theory and the Elaboration Likelihood Model and by introducing a temporal dimension to examine green attribute changes, we expect to more comprehensively and deeply reveal the dynamic formation mechanism of eco-hotel location scores, thereby providing valuable references for theoretical development and industry practice in the field of sustainable tourism.

3. Research Model and Hypotheses

Online travel platforms (OTPs) like Booking.com have fundamentally altered consumer behavior by providing centralized access to a vast array of accommodation options, including cost-performance hotels. These platforms seamlessly integrate social and e-commerce functionalities, allowing users to search, compare, and book travel products with unprecedented ease [15]. The inclusion of cost-performance information on these platforms empowers consumers with detailed information about sustainable practices, eco-friendly amenities, and authentic reviews from fellow travelers. This accessibility not only fosters trust and engagement but also enables consumers to make informed decisions based on a holistic assessment of both traditional and green attributes [16].

The theoretical model of this study is rooted in a deep integration of Signaling Theory and the Elaboration Likelihood Model. Signaling Theory explains why the spillover effect occurs: the dynamic changes in an eco-hotel’s green attributes (eco-facilities, sustainable practices, ecological experience) serve as signals to the market about its underlying high quality or managerial deficiencies. The ELM, in turn, explains how the spillover effect works: consumer environmental involvement, as a moderating variable, determines whether consumers process these signals through a central or peripheral route, thereby leading to evaluation spillovers of varying intensity.

3.1. Positive Signals from Green Attributes and Positive Spillover Effects

One of the core innovations of this study is to define the positive dynamic changes in green attributes as a high-fidelity market signal that is more diagnostically valuable than a static label. According to Signaling Theory, the effectiveness of a signal depends on its transmission cost and credibility [17]. A hotel’s continuous improvement in eco-facilities, sustainable practices, and ecological experience inevitably requires real, long-term financial and managerial costs. This makes such a dynamic positive trajectory a quality commitment that is harder to imitate and more persuasive than static environmental slogans or one-time certifications.

When consumers perceive this positive signal, they interpret this dynamic improvement as a diagnostic cue about the hotel’s underlying quality. Subsequently, consumers engage in cognitive attribution, likely inferring that a hotel willing to continuously invest in the green cause —a non-directly profitable area—is also highly likely to possess higher standards in less directly observable dimensions such as managerial rigor, service detail, integrity, and even forward-thinking strategy [18]. This attribution process effectively reduces the information asymmetry and perceived risk consumers face in decision-making and enhances their trust in the hotel brand [19]. Ultimately, the trust and overall positive impression (i.e., the halo effect) built upon this positive signal will spill over to the evaluation of other attributes through a psychological mechanism of evaluative transference. Consumers might even reinterpret the hotel’s location, viewing a slightly remote location as perfectly aligned with the tranquil, natural, and sustainable brand philosophy it pursues, thereby giving it a higher location score. Based on this, we propose the following hypotheses:

H1: 

Positive changes in eco-facilities scores, as a positive signal, have a positive spillover effect on the location score changes of eco-hotels.

H2: 

Positive changes in sustainable practices scores, as a positive signal, have a positive spillover effect on the location score changes of eco-hotels.

H3: 

Positive changes in ecological experience scores, as a positive signal, have a positive spillover effect on the location score changes of eco-hotels.

3.2. Negative Signals from Green Attributes and Negative Spillover Effects

Compared to positive signals, the study of negative dynamic changes as a negative signal is particularly insufficient, yet in reality, it may constitute a more impactful market signal. Based on the psychological principle of Negative Bias, negative information often carries more weight and influence than positive information when affecting individual judgments [20]. When a hotel’s green attributes deteriorate (e.g., decreased cleanliness, unrepaired eco-friendly facilities), it constitutes a blatant violation of its environmental promises, sending a highly destructive red flag to the market.

Consumers will first perceive this negative change as a warning signal. Its destructive power lies not just in being a service flaw, but in being potentially attributed to deeper, systemic problems within the hotel, such as managerial chaos, the hypocrisy of its commitments, or even the eventual exposure of greenwashing [21]. This deep-seated negative attribution will quickly lead to a collapse of consumer trust. Once trust is broken, consumers activate a defensive mindset of skepticism and scrutiny, leading to a Horns Effect, where they are inclined to evaluate all aspects of the hotel negatively. This distrust will directly spill over to the location evaluation. This collapse of trust and value reassessment triggered by negative signals will inevitably lead to a significant drop in the location score. Therefore, we propose the following hypotheses:

H4: 

Negative changes in eco-facilities scores, as a negative signal, have a negative spillover effect on the location score changes of eco-hotels.

H5: 

Negative changes in sustainable practices scores, as a negative signal, have a negative spillover effect on the location score changes of eco-hotels.

H6: 

Negative changes in ecological experience scores, as a negative signal, have a negative spillover effect on the location score changes of eco-hotels.

3.3. The Moderating Role of Consumer Environmental Involvement

Another theoretical innovation of this study is the application of the Elaboration Likelihood Model to explain the heterogeneity in how consumers process dynamic market signals in real consumption scenarios [22,23]. We argue that the level of consumer environmental involvement determines the degree of cognitive resources invested, leading consumers to choose different information processing routes when faced with green signals:

For high-CEI consumers, who possess the intrinsic motivation and knowledge to process environmental information, they tend to engage in central route processing. When faced with signals of green attribute changes, they will conduct systematic, elaborate cognitive processing. They will actively seek more information, carefully scrutinize the quality of arguments (such as specific details in reviews), and engage in deep logical attribution [13]. For instance, a high-involvement consumer, upon reading in reviews that the hotel has started implementing waste sorting and using local ingredients, might actively seek more information. They would likely perceive this as a sign of excellent management and social responsibility, thereby reinterpreting a location they previously considered ‘remote’ as ‘tranquil and close to nature’. This deep cognitive engagement results in judgments and attitudes that are more robust, lasting, and resistant to change. Therefore, both strong positive signals and destructive negative signals will trigger a more intense evaluation response in them, meaning the spillover effect is significantly amplified.

In contrast, for low-CEI consumers, who lack the motivation or ability for deep processing, they tend to engage in peripheral route processing. They will treat changes in green attributes merely as a heuristic, non-core peripheral cue. For example, “most reviews are good” or “the website has an eco-label” might be sufficient to form a vague positive impression. They will not invest significant cognitive resources to analyze the deeper meaning behind the signals [8]. This shallow, heuristic processing results in judgments that are relatively weak and transient. Consequently, the evaluation spillover triggered by green attribute changes signals, whether positive or negative, will be significantly weaker in intensity. Based on this, we propose the following hypotheses:

H7: 

Consumer environmental involvement positively moderates the spillover effect of positive changes in green attribute scores. That is, for consumers with high CEI, the positive spillover effect is stronger.

H8: 

Consumer environmental involvement positively moderates the spillover effect of negative changes in green attribute scores. That is, for consumers with high CEI, the negative spillover effect is also stronger.

3.4. Research Model

Based on the above hypotheses, this study constructs a research framework that integrates Signaling Theory and the ELM (see Figure 1). This model clearly shows the interrelationships between green attribute changes (signal source), location score changes (signal reception result), and CEI (signal processing moderator).

4. Data Acquisition and Empirical Analysis

4.1. Data Collection and Sample Construction

To accurately capture the dynamic evolution of eco-hotels’ green attributes and their temporal spillover effect on location scores, this study designed and implemented a rigorous, multi-stage quantitative longitudinal research plan. The cornerstone of this plan is its interdisciplinary integration, deeply merging cutting-edge natural language processing (NLP) techniques with mature econometric models. The goal was to transform a massive volume of unstructured consumer online review texts into structured panel data suitable for rigorous causal inference, thereby empirically testing the proposed “signal-processing” two-stage theoretical model.

The data corpus for this study was sourced from Booking.com, a leading global online travel platform. This platform was chosen for its vast user base, rich review content, and clear labeling system for eco-friendly accommodations, providing an ideal digital field for observing the target phenomenon. We employed automated web scraping techniques to systematically collect data from January 2020 to December 2023. We targeted approximately 9000 cost-performance eco-hotels across multiple countries and regions, each with over 50 reviews, ultimately constructing a large longitudinal dataset of over 60,000 valid consumer reviews. This data scale not only ensures the statistical validity and generalizability of the research conclusions, but its time span also allows us to capture the evolving trends in consumer attitudes towards sustainable tourism in the post-pandemic era.

To transform theoretical constructs into measurable empirical variables, this study employed a multi-step quantification process based on natural language processing. First, we conducted standardized data cleaning on all review texts, including converting to lowercase, removing punctuation, tokenization, removing stop words, and lemmatization, to ensure the consistency and comparability of the text data. We utilized a Word2Vec-based word embedding model to quantify the sentiment scores for each attribute. The process was as follows: First, we defined a set of core seed words for each attribute to be measured (see

Table 1. Hotel related properties and seed words.

Attribute Dimension	Subordinate Variables	Seed Words
Eco-facilities	Cleanliness	tiny, floor, cleanliness, cleaning clean, tidy, cleaned, dirty…
	Room facilities	room, facilities, beds, bathroom, courtyard, size…
Sustainable practices	Cost-performance	price, value, money, paid, pay, precious, charge, cost…
	Services and staff	staff, breakfast, service, check, careful, friendly…
Ecological experience	Comfort	uncomfortable, comfy, safely, home, comfort, smart…
Location score	Location	location, place, walk, centrale, walking, located, distance…

). Second, using a Word2Vec model pre-trained on a large corpus, we identified the most semantically similar words for each seed word by calculating the cosine similarity of their word vectors. This allowed us to construct more comprehensive and accurate semantic dictionaries for each attribute. Finally, we applied a sophisticated sentiment analysis algorithm to calculate the sentiment polarity (positive/negative) and intensity score for the words from each attribute dictionary that appeared in a review, assigning these scores to the corresponding attribute. In this manner, we transformed the vague descriptions of various dimensions in each review into precise, continuous sentiment scores.

This study adopts a dictionary-based method to quantify consumers’ psychological constructs, an approach with a solid theoretical foundation in psychology and consumer research. The foundational research by Pennebaker et al. [24] demonstrated that the words people use can effectively reflect their internal psychological states and traits, providing a theoretical basis for inferring psychological constructs from text. More specifically, Humphreys and Wang [25], in their review of consumer research, pointed out that dictionary-based automated text analysis is a valid deductive tool for measuring known theoretical constructs, such as the CEI in this study [26]. Therefore, this study’s approach of constructing an environmental topic dictionary to measure CEI is built upon a mature and validated methodology.

The quantification of CEI is a core innovation of this study. As defined by Vermeir and Verbeke [12], CEI refers to the personal relevance and importance an individual attaches to environmental issues. Traditional text analysis methods, such as Term Frequency–Inverse Document Frequency (TF-IDF), are often used to measure the weight of a topic, but this approach only considers word frequency and cannot distinguish the consumer’s emotional attitude. To overcome this limitation, this study innovatively combines Latent Semantic Analysis (LSA) with sentiment analysis. We began by constructing a foundational environmental topic dictionary containing words such as sustainability, eco-friendly, and recycling, based on the classic literature in the field of environmental research like Kim and Choi (2005) [14]. We then applied LSA to the entire review corpus for topic modeling to identify latent semantic topics related to the environment. The crucial step was to assign sentiment weights to these environmental topics as they appeared in each consumer’s review, using the previously calculated sentiment scores. This weight reflects the consumer’s affective involvement in the discussion of environmental issues. Concurrently, we retained the frequency of environment-related words as a proxy for their cognitive involvement. By combining sentiment-weighted LSA topics with word frequency, we constructed a more comprehensive CEI measurement that captures both the cognitive and affective components of consumer involvement. This method is far superior to traditional word frequency counts, as it more profoundly and accurately reveals a consumer’s true level of involvement with environmental issues.

To ensure the validity of this innovative measurement method, we validated it in multiple ways. First, the method’s content validity is guaranteed by its theoretical foundation: our initial environmental dictionary is rooted in the established academic literature, ensuring the relevance of the measurement. Second, by examining its construct validity, we found that our calculated composite CEI score is significantly and positively correlated with simpler metrics like environmental word frequency (convergent validity), while having a low correlation with unrelated constructs such as service satisfaction (discriminant validity). Finally, and most importantly, the method’s predictive validity was confirmed by its performance in the subsequent models. Our results show that this CEI variable successfully and significantly moderated the spillover effect in a manner fully consistent with theoretical predictions, thereby demonstrating the method’s effectiveness and robustness.

Before proceeding to formal regression analysis, we subjected the raw dataset to a rigorous preprocessing routine. This included the prudent imputation of missing values, identification and handling of statistical outliers, and standardization of all variables to eliminate potential interference from differing scales on model estimation. The entire process, from theoretical construction to data collection and variable quantification, was interconnected, aiming to ensure a high degree of scientific rigor and logical consistency at every step, providing a reliable methodological guarantee for ultimately uncovering the dynamic formation mechanism of eco-hotel location scores. (A flowchart of this process is shown in Figure 2).

4.2. Variable Measurement and Operationalization

To transform theoretical constructs into measurable empirical variables, this study rigorously operationalized each core variable:

Dependent variable: location score changes. This variable was quantified by conducting sentiment analysis on the text content of each review related to the hotel’s geographical location. We first built a location-related keyword dictionary, then used a Word2Vec model to calculate the sentiment score for location in each review, and finally, through time-series differencing, obtained the dynamic change in the location score for each quarter.

Independent variables: green attribute score changes. We used the same word vector sentiment analysis method to quantify the three core green attributes separately. Eco-facilities were primarily measured by the sentiment scores of words related to cleanliness and room facilities; sustainable practices were reflected by words like cost-performance, service quality, and staff; and ecological experience was captured mainly through the sentiment scores of comfort-related words. Similarly, we calculated the quarterly change for each green attribute score.

Moderating variable: consumer environmental involvement. The quantification of CEI was a challenge and an innovation of this study. We utilized Latent Semantic Analysis for the deep topic mining of the vast review texts. By identifying and extracting latent semantic structures and topic clusters that reflect consumers’ environmental awareness, knowledge, and emotional investment, we were able to quantify each reviewer’s environmental involvement as a continuous value, thus achieving an objective measurement of this psychological construct.

5. Data Analysis and Results

5.1. Data Preprocessing

Prior to analysis, the dataset was rigorously cleaned to handle missing values and outliers. Missing data were addressed using appropriate imputation techniques (e.g., mean substitution or multiple imputation) to maintain data integrity without introducing significant bias. Outliers were identified using statistical methods (e.g., Z-scores, IQR) and were handled accordingly to prevent a distortion of the analysis. Furthermore, data normalization was performed to ensure a uniform distribution of variables, facilitating accurate and meaningful comparisons across different metrics.

Descriptive statistics provided a foundational understanding of the dataset’s characteristics. Measures such as mean, standard deviation, minimum, and maximum values were calculated for each of the green attribute scores and location score changes. For instance, the analysis revealed that the mean for cost-performance scores was negative, indicating a general decline in perceived value, whereas the mean for location scores was positive, reflecting overall satisfaction with hotel locations. These statistics offered valuable insights into the central tendencies and variability within the data, guiding the subsequent inferential analysis. The descriptive statistical results are shown in

Table 2. Descriptive statistics.

Fundamental Variables	Variables	Mean	Std. Error	Min	Max
	Location	0.00389	0.62554	–4.00000	5.00000
	Consumer environmental involvement	0.00001	0.00400	–0.03100	0.03500
Eco-Facilities	Cleanliness	0.00076	0.19069	–3.00000	3.00000
	Room facilities	0.00049	0.98713	–12.75000	10.0000
Sustainable Practices	Cost-performance	–0.00043	0.24302	–3.50000	4.00000
	Services and staff	–0.00301	1.01791	–10.40000	10.0000
Ecological Experience	Comfort	–0.00135	0.55300	–4.33333	5.00000

Note: The measurement of eco-facilities is based on variables cleanliness and room facilities, sustainable practices are determined by variables cost-performance and services and staff, and ecological experience is evaluated using variable comfort.

.

5.2. Regression Results Analysis

To improve the data distribution and reduce the influence of outliers, we applied a logarithmic transformation to the dependent variable location score and standardized all variables to eliminate unit differences. To test the moderating effect, we performed a median split on the comprehensive CEI score. This approach, which divides the sample into two distinct subgroups (a high-CEI group and a low-CEI group), is a robust and widely accepted method in consumer research for clearly examining the differential effects of a moderating variable. As argued by Iacobucci et al. [27], the median split is particularly effective for illustrating how a moderating variable alters the relationship between independent and dependent variables across groups, making it an ideal choice for testing our hypotheses H7 and H8. We then conducted OLS regression and quantile regression analyses.

Collinearity diagnostics showed no multicollinearity among the variables, with all VIF values below 10. Additionally, a Durbin–Watson test was conducted to check for autocorrelation in the linear regression model. The DW value ranges from 0 to 4, with values closer to 2 indicating less autocorrelation. In our study, the DW values were between 1.6 and 1.8, suggesting no significant autocorrelation among the variables.

Table 3. Regression results.

Index	OLS Coefficient	15%	35%	55%	75%	95%
Results under low CEI
constant	0.003 (0.382)	−0.953 ** (−56.311)	−0.224 ** (−18.503)	0.277 ** (27.234)	0.721 ** (89.181)	1.107 ** (205.759)
Positive changes in cost-performance	0.079 ** (10.874)	0.105 ** (5.233)	0.079 ** (5.614)	0.074 ** (7.164)	0.052 ** (6.076)	0.061 ** (10.933)
Negative changes in cost-performance	–0.062 ** (–7.337)	−0.078 ** (−3.521)	−0.068 ** (−5.735)	−0.070 ** (−7.658)	−0.052 ** (−7.209)	−0.030 ** (−7.800)
Positive changes in cleanliness	0.083 ** (11.462)	0.099 ** (5.287)	0.104 ** (8.092)	0.085 ** (7.831)	0.042 ** (5.411)	0.039 ** (6.160)
Negative changes in cleanliness	–0.061 ** (–5.866)	−0.063 ** (−2.978)	−0.062 ** (−5.163)	−0.082 ** (−8.560)	−0.040 ** (−6.318)	−0.058 ** (−10.377)
Positive changes in comfort	0.163 ** (18.912)	0.183 ** (9.323)	0.177 ** (13.912)	0.166 ** (17.354)	0.117 ** (15.013)	0.095 ** (19.533)
Negative changes in comfort	–0.179 ** (–17.686)	−0.202 ** (−9.139)	−0.197 ** (−14.271)	−0.182 ** (−17.037)	−0.116 ** (−12.359)	−0.103 ** (−16.955)
Positive changes in services and staff	0.158 ** (16.494)	0.168 ** (7.814)	0.161 ** (11.848)	0.151 ** (15.339)	0.122 ** (18.701)	0.071 ** (15.695)
Negative changes in services and staff	–0.203 ** (−19.364)	−0.219 ** (−10.069)	−0.215 ** (−15.502)	−0.216 ** (−21.162)	−0.174 ** (−25.815)	−0.102 ** (−18.955)
Positive changes in room facilities	0.175 ** (14.266)	0.180 ** (7.557)	0.171 ** (11.389)	0.187 ** (16.700)	0.141 ** (18.396)	0.110 ** (18.658)
Negative changes in room facilities	−0.172 ** (−17.784)	−0.163 ** (−7.466)	−0.192 ** (−14.810)	−0.173 ** (−18.295)	−0.155 ** (−24.244)	−0.096 ** (−18.453)
$R^2$	0.116
Results under high CEI
constant	−0.003 (−0.378)	−0.925 ** (−53.060)	−0.236 ** (−20.691)	0.246 ** (24.617)	0.716 ** (82.434)	1.104 ** (205.164)
Positive changes in cost-performance	0.065 ** (8.701)	0.071 ** (3.177)	0.057 ** (4.636)	0.055 ** (5.875)	0.057 ** (7.241)	0.035 ** (7.851)
Negative changes in cost-performance	−0.088 ** (−10.546)	−0.097 ** (−4.015)	−0.092 ** (−6.648)	−0.083 ** (−7.585)	−0.057 ** (−5.657)	−0.035 ** (−6.825)
Positive changes in cleanliness	0.062 ** (6.430)	0.076 ** (3.888)	0.074 ** (6.551)	0.066 ** (6.853)	0.045 ** (4.816)	0.023 ** (5.161)
Negative changes in cleanliness	−0.085 ** (−10.398)	−0.111 ** (−5.505)	−0.106 ** (−8.685)	−0.085 ** (−7.644)	−0.058 ** (−6.398)	−0.058 ** (−7.521)
Positive changes in comfort	0.190 ** (21.516)	0.246 ** (10.513)	0.206 ** (15.332)	0.185 ** (17.106)	0.127 ** (12.600)	0.100 ** (16.628)
Negative changes in comfort	−0.174 ** (−21.598)	−0.186 ** (−9.003)	−0.178 ** (−14.343)	−0.183 ** (−18.937)	−0.125 ** (−14.404)	−0.091 ** (−18.082)
Positive changes in services and staff	0.213 ** (21.871)	0.239 ** (10.168)	0.218 ** (16.243)	0.224 ** (21.664)	0.208 ** (27.972)	0.094 ** (16.609)
Negative changes in services and staff	−0.179 ** (−14.627)	−0.191 ** (−7.499)	−0.189 ** (−14.210)	−0.180 ** (−18.065)	−0.154 ** (−21.550)	−0.076 ** (−16.345)
Positive changes in room facilities	0.170 ** (13.806)	0.186 ** (8.242)	0.187 ** (15.516)	0.164 ** (17.942)	0.150 ** (22.388)	0.092 ** (18.316)
Negative changes in room facilities	−0.182 ** (−11.245)	−0.217 ** (−7.945)	−0.182 ** (−12.898)	−0.189 ** (−17.738)	−0.182 ** (−24.161)	−0.122 ** (−25.573)
$R^2$	0.122

Note: The measurement of eco-facilities score changes is based on variables cleanliness and room facilities, sustainable practices score changes are determined by variables cost-performance and services and staff, and ecological experience score changes are evaluated using variable comfort; ** p < 0.05.

shows the regression results. Under high CEI, the R² value for the OLS regression model was 0.122, indicating that the independent variables explained 12.2% of the variance in location score changes. Using robust standard errors, the model passed the F-test (F = 108.415, p < 0.05). Similarly, under low CEI, the R² value for the OLS regression model was 0.116, explaining 11.6% of the variance, and it also passed the F-test (F = 104.391, p < 0.05).

Regardless of the CEI level, positive changes in green attribute scores had a significant positive impact on location score changes (β > 0, p < 0.01), indicating a positive spillover effect and thus supporting H1–H3. This is consistent with previous findings on positive spillover effects in consumer evaluations. Conversely, negative changes in green attribute scores had a significant negative impact on location score changes (β < 0, p < 0.01), indicating a negative spillover effect and supporting H4–H6, which aligns with research on negative spillovers in consumer behavior.

In the high CEI scenario, positive changes in services and staff had the most significant positive spillover effect on location score changes (β = 0.213, p < 0.01). This effect is likely driven by the direct impact of enhanced sustainable practices—such as eco-initiatives, staff engagement, and the integration of green policies—on customer perception and overall service quality. When customers perceive a stronger commitment to sustainability, they are more likely to rate the hotel higher, especially in aspects related to location and reputation. Conversely, negative changes in room facilities had the most significant negative spillover effect (β = −0.182, p < 0.01). This is because with a heightened customer awareness of environmental issues, poor eco-facilities lead to dissatisfaction that spills over into lower scores for location and overall hotel performance.

In the low CEI scenario, positive changes in room facilities had the most significant positive spillover effect (β = 0.175, p < 0.01). This suggests that in a low CEI context, guests are more likely to value eco-friendly facilities and green certifications, as they have fewer opportunities to engage with sustainable practices in other areas. The availability of these eco-friendly facilities—such as energy-efficient appliances, water-saving systems, and sustainable building materials—directly influences guest satisfaction and the hotel’s location. On the other hand, negative changes in services and staff had the most significant negative spillover effect (β = −0.203, p < 0.01). In this context, a lack of visible sustainable practices, such as staff training on eco-friendly behaviors or the absence of sustainability policies, severely damages customer perception. This is particularly impactful in a low CEI environment, where customers may place more emphasis on visible sustainable practices in service and staff.

These differing magnitudes of spillover effects across CEI levels support H7 and H8, confirming the moderating role of CEI. This finding is consistent with the Elaboration Likelihood Model, which suggests that highly involved individuals are more influenced by attribute-specific information.

5.3. Further Findings

Linear regression only reveals the surface of the spillover effect; its “proportional input-return” perspective is misleading. To explore deeper mechanisms, we employed quantile regression and discovered non-linear, counter-intuitive threshold effects [28]. These findings are the most innovative contributions of this study, challenging traditional management logic.

One subversive discovery is that the positive spillover from improvements in sustainable practices (service and staff) does not increase linearly but presents an “inverted L-shaped” curve (see Figure 3 and Figure 4). This means its positive effect remains at a relatively stable high level across most low-to-mid quantiles (a plateau) but drops sharply once the location score reaches the highest quantiles. This contradicts the intuition that “better service leads to higher scores.” This phenomenon reveals a saturation cliff effect of high-quality signals: once a hotel establishes an excellent image through continuous effort, its positive signals become saturated, and new improvements can no longer “wow” consumers. For hotels whose location scores are already top-tier, consumers may shift their attention to other, more critical aspects, leading to a significant reduction in the spillover effect of green attributes. This proves that “good” cannot be infinitely cumulative; excellence has a cliff-like threshold for spillover effects.

In stark contrast, the negative spillover triggered by the deterioration of sustainable practices (service and staff) shows a “J-shaped” curve (see Figure 5 and Figure 6). This finding completely upends our initial trust cliff hypothesis. The negative regression coefficient (Y-axis) starts from a lower negative value and significantly rebounds towards 0 in the higher quantiles. This does not mean the negative impact is amplified but quite the opposite: the negative impact is significantly weakened. This counter-intuitive phenomenon reveals a “buffer effect” of excellent core assets: for hotels with already excellent location scores (high quantiles), their superior geographical location acts as a powerful cushion, absorbing and offsetting a portion of the negative impact from service-level deterioration. The consumer’s logic might be as follows: “Although the service at this hotel has worsened, the location is just too good, so I’m willing to tolerate it.”

Combining these two asymmetric findings, we arrive at a more profound conclusion: the trajectory of consumer evaluation is far more complex than imagined. Building goodwill is like rock climbing, where one encounters a saturation plateau where effects stagnate; the impact of a service decline, however, is moderated by the hotel’s strongest core advantage. This deepens Signaling Theory: the destructive power of a negative signal (like poor service) is not isolated; it is weakened by an overwhelming positive signal (like a prime location). This also explains why dynamic changes are so important and why the response curve for high-CEI consumers is steeper—because they are more sensitive to the interplay and value trade-offs between signals. These findings collectively paint a complex picture of consumer decision-making, emphasizing that businesses must shift from linear thinking to non-linear, dynamic threshold management and contextualized signal management.

5.4. Discussion

Our findings not only confirm the importance of green attributes but also specifically reveal the mechanism, magnitude, and key drivers of their influence on consumer perceptions. First, regarding the mechanism of influence, this study shows that dynamic changes in green attributes act as signals of a hotel’s underlying quality, reshaping consumers’ overall perceived value through a spillover effect, which in turn alters their evaluation of a static asset like location. Second, concerning the magnitude of importance, our model quantifies the effect size of this influence. For example, for highly environmentally involved consumers, improvements in sustainable practices (represented by service and staff) are the most predictive, with a regression coefficient for positive changes reaching 0.213, indicating that even a minor enhancement in green practices can significantly increase the hotel’s perceived value. Finally, this study clarifies that the most predictive green attribute is not static but depends on the consumer type: for high-CEI consumers, intangible sustainable practices are most influential, whereas for low-CEI consumers, tangible eco-facilities (e.g., room facilities, β = 0.175) become the more critical driver.

The confirmation of both positive and negative spillover effects underscores the need for hotel managers not only to enhance green attributes but also to prevent declines in these areas. The moderating role of CEI suggests that marketing strategies should be tailored to different consumer segments. For consumers with high environmental involvement, emphasizing sustainable practices may yield greater improvements in location scores. For those with lower involvement, highlighting eco-facilities and ecological experiences may be more effective.

The identification of non-linear spillover effects, such as inverted U-shaped and J-shaped relationships, adds complexity to our understanding of consumer behavior. These threshold effects imply that there are optimal levels of investment in green attributes, beyond which additional efforts may yield diminishing returns or even negative consequences.

6. Conclusions and Implications

6.1. Main Conclusions

Through the deep mining of large-scale, longitudinal real consumer data, this study systematically reveals the mechanism by which dynamic changes in eco-hotel green attributes affect their location scores. The core findings include the following: First, positive changes in green attributes (covering eco-facilities, sustainable practices, and ecological experience) act as a high-quality market signal, producing a significant positive spillover effect on location scores; conversely, their negative changes constitute a dangerous signal, leading to significant negative spillovers. Second, consumer environmental involvement plays a key moderating role in this process, with high-involvement consumers reacting more strongly to both positive and negative spillover effects. Third, the spillover effect is not a simple linear relationship but exhibits complex non-linear and threshold characteristics. These findings not only provide new empirical evidence for understanding complex consumer decision-making behavior in the context of sustainable tourism but also have profound and enlightening theoretical and practical implications.

6.2. Theoretical Contributions

By constructing an integrated theoretical framework and validating it with large-scale real consumer data, this study makes several contributions to consumer behavior theory in the field of sustainable tourism. First, by introducing a dynamic perspective, this study significantly enriches the application of Signaling Theory in service marketing. It innovatively points out that the dynamic trajectory of a signal (i.e., continuous improvement or degradation) constitutes a higher-fidelity meta-signal than a static label. It deepens our understanding of evaluation spillover effects, especially the spillover across attribute categories (from dynamic service attributes to static physical assets). This study confirms that dynamic, subjective green signals can systematically alter a consumer’s value perception of an objective, static core asset like location, demonstrating a powerful penetration of the spillover effect, sufficient to reframe a consumer’s cognitive framework regarding a brand’s most fundamental assets. This provides a more refined micro-foundation for Signaling Theory, explaining why continuous changes in non-core service attributes can reshape consumer value judgments of a hotel’s core physical assets, such as geographical location.

Second, this study expands the Elaboration Likelihood Model in terms of methodology and application scenarios. By using natural language processing techniques to quantify consumer environmental involvement from massive real reviews, this study moves ELM from a laboratory theory to the real world, validating its strong vitality in explaining how large-scale, naturally occurring consumer evaluations of specific attributes like hotel location become heterogeneous. A more central contribution is the construction and validation of a “signal-processing” two-stage integrated framework, which provides a more complete causal chain for explaining evaluation spillover effects: Signaling Theory defines how green signals indicate a hotel’s potential quality, while ELM explains how these signals are decoded by different consumers. This integration provides a logically tighter and more explanatory theoretical weapon for understanding complex consumer decision-making processes and pushes sustainable consumption research from static assessment to dynamic observation.

6.3. Managerial Implications

The findings of this study offer a series of forward-thinking strategic implications for eco-hotel managers and online travel platforms. The core idea is to transform sustainability from a static compliance task into a dynamic, strategic value-creation tool capable of actively managing consumer perception. This requires firms to cultivate a dynamic capability that integrates market signals with environmental practices [29]. First, hotels must drive a revolution in sustainability communication from a static asset display to a dynamic signal narrative. The research reveals that consumers react far more strongly to the dynamic process of continuous improvement than to static labels. Therefore, hotels should proactively showcase their green journey of continuous progress. This not only builds a more credible high-quality signal but also enhances their location score by endowing the hotel with a proactive brand personality, even if the location is not geographically advantageous.

Second, customer relationship management urgently needs to innovate from one-size-fits-all broadcasting to stratified signal outreach. Given that consumers with different levels of environmental involvement process green signals through different routes, managers should abandon a uniform marketing approach. Instead, they should provide detailed green evidence to high-involvement customers. For instance, when a hotel chain like Marriott begins to promote its efforts in reducing food waste and using local ingredients, it can precisely target high-involvement customer segments with this information through detailed reports or social media stories. These consumers are more likely to deeply understand the value of such initiatives and translate that into overall brand trust, potentially even boosting their scores of the locations of certain hotel branches. Meanwhile, for the broader low-involvement audience, the focus should be on concise, intuitive peripheral cues (such as sustainable travel labels) to achieve precise communication and maximize spillover effects.

Finally, brand reputation defense must shift from passive crisis public relations to proactive signal risk control. This is not merely a defensive act, but a strategy for value creation. Therefore, through the strategic management of dynamic signals like green attributes, managers can gain a powerful lever to break their most fundamental physical constraint: geographical location. This means that even hotels in less advantageous geographical spots can, through sustained green practices, transform ‘remote’ into positive associations like ‘tranquil,’ reshaping their ‘psychological location’ value in the minds of consumers and offering a novel space for brand strategy and asset management.

6.4. Limitations and Future Research Directions

Despite the comprehensive methodology, this study acknowledges several limitations that warrant future investigation. First, the data were sourced exclusively from Booking.com, which may limit the generalizability of the findings to other online travel platforms. Future research should incorporate multiple platforms to enhance external validity. Second, the reliance on automated sentiment and semantic analysis, while innovative, may not fully capture the nuanced complexities of consumer sentiment; incorporating qualitative methods could provide deeper insights. Additionally, this study focused on a specific timeframe (2020–2023), potentially overlooking long-term trends and external factors, such as global environmental policies or economic shifts that could influence consumer behavior. Future research could extend the timeframe to include a broader temporal context. Moreover, while CEI was a key moderating variable, other individual differences, such as cultural background, demographic factors, and prior environmental experiences, could also influence spillover effects and merit exploration. Finally, this study primarily examined direct spillover effects, overlooking potential indirect effects through mediating variables like brand reputation or consumer trust. Future research should examine these additional pathways to provide a more comprehensive understanding of the mechanisms behind consumer evaluations in sustainable tourism. Addressing these limitations will further clarify the complex dynamics of green attribute evaluations and their impact on location scores, thereby advancing both theory and practice in the field.