Measuring City-Level Tourist Source-Market Linkage Capacity from Online Review Origin Data: Evidence from Liaoning, China

Abstract

Tourist source markets are often described through origin shares or flow volumes, but these measures cannot explain how destination cities build diversified, spatially extended, temporally stable, and network-embedded linkages with source markets. This study develops a system-oriented Market Accessibility and Interaction Index (MAI) for measuring city-level tourist source-market linkage capacity from online review origin data. Using Ctrip reviews collected during a unified data-collection window in October 2025 from 112 A-level attractions selected from the official Liaoning provincial catalog, we compiled a review-level dataset of 76,855 records and retained 29,327 reviews with valid origin, destination-city, timestamp, and distance information for the main analysis. The method proceeded in four steps: sample construction and origin standardization; city-to-origin distance measurement; seven-dimensional indicator construction; and normalization, aggregation, robustness testing, and benchmark comparison. The MAI integrates market scale, source diversity, interprovincial attraction, external effective radius, seasonal stability, input strength, and PageRank centrality. Results reveal pronounced inter-city differentiation. Dalian is the strongest source-market linkage hub, followed by Huludao, Dandong, and Jinzhou, while lower-ranked cities display narrower or less embedded source-market structures. Comparative benchmarking shows that MAI captures multidimensional linkage capacity more comprehensively than single indicators such as review volume, external share, distance reach, or PageRank alone. The study contributes a system-oriented diagnostic measure of city-level source-market linkage capacity and extends online review data from evaluative analysis to structural market measurement.

1. Introduction

1.1. Research Background and Problem Statement

Tourist source markets are a long-standing concern in tourism geography, destination management, and regional tourism planning because they shape not only where tourism demand originates but also how destinations position themselves, allocate marketing resources, and manage external dependence. Existing studies have commonly examined tourist source markets through visitor composition, travel flows, distance decay, gravity effects, and origin–destination interactions [1,2,3,4,5,6]. Although these approaches provide valuable insights into market structure and spatial constraints, they more often treat source markets as relatively static distributions than as structured systems of destination–origin linkages.

This limitation has become more important in platform-mediated tourism systems. A destination city is not connected to its source markets only through the number of visitors it receives; it is also connected through the diversity of represented origins, the degree of external openness, the effective spatial radius of demand, the continuity of demand over time, and the city’s position in a broader destination–origin network. A systems perspective is therefore appropriate because source-market linkage capacity emerges from interacting components rather than from a single-flow or market-share indicator.

1.2. Digital Footprints, UGC, and the Need for Structural Market Evidence

Online travel platforms record large volumes of user-generated content together with timestamps, ratings, and origin-related metadata, making it possible to observe tourism demand not only in terms of size, but also in terms of spatial breadth, temporal continuity, and relational structure. Early digital tourism studies showed that blogs, social media, and user-generated content increasingly shape tourism information search and travel planning [7,8,9,10,11,12]. Building on this foundation, online review research in tourism and hospitality has expanded to examine eWOM effects, booking outcomes, review helpfulness, destination image, topic extraction, service quality, and tourist satisfaction [13,14,15,16,17,18,19].

However, most UGC-based studies still treat platform records primarily as evaluative or perceptual materials rather than as structural evidence of destination–origin linkage. This creates a methodological gap: when review records contain origin-related metadata, they can be used not only to interpret what tourists say about destinations, but also to reconstruct how destination cities are connected to source regions. Comparative platform research further suggests that online review platforms differ systematically in how destinations are represented and that Ctrip is an analytically meaningful source in the Chinese tourism context [20].

1.3. Tourism Flows, Destination Networks, and the Systems Perspective

A second relevant stream of research uses digital footprints to reconstruct tourism flows, inter-city structures, and destination networks. Studies based on online reviews, travel blogs, mobile traces, and other platform-derived data show that digital traces can reveal destination hierarchies, tourism-flow differentiation, and network organization that are difficult to capture through conventional aggregate statistics [21,22,23,24,25,26,27,28,29,30]. At the same time, tourism-network research emphasizes that destinations should be understood as relational entities embedded in broader systems rather than as isolated points [31].

From this perspective, a tourist source market is not simply a collection of origin shares. It is a relational system through which destination cities receive, organize, and sustain demand from multiple source regions. Network centrality indicators are useful for describing the position of destinations in a network, but they cannot, on their own, capture market size, source diversity, external openness, spatial reach, temporal stability, and evaluative platform evidence simultaneously. A multidimensional diagnostic framework is therefore needed.

1.4. Research Gap, Objectives, and Contributions

To address this gap, the present study conceptualizes tourist source markets as a city-level linkage system. Using Ctrip reviews of A-level tourist attractions in Liaoning Province, China, it develops a Market Accessibility and Interaction Index (MAI) to measure city-level tourist source-market linkage capacity from seven dimensions: market scale, source diversity, interprovincial attraction, external effective radius, seasonal stability, input strength, and PageRank centrality. Rather than attempting to represent a full international tourism market system, the analysis focuses on domestic interprovincial source-market linkages observed in platform-based review behavior.

The proposed MAI differs from single-flow indicators, perception-oriented UGC analysis, and network centrality measures used on their own. Single-flow or review-volume indicators capture market size but ignore breadth, openness, temporal stability, and network position. UGC-evaluation studies explain satisfaction, sentiment, or image but do not directly measure source-market linkage capacity. Centrality indicators such as PageRank describe a node’s position in a network but do not simultaneously measure the multiple system properties that shape destination–origin linkage. MAI integrates these dimensions into a composite diagnostic measure.

This study makes three contributions. First, it reframes tourist source markets from a static description of origin composition into a measurable linkage system at the city level. Second, it extends the use of online review data from sentiment, satisfaction, and destination-image analysis to structural market measurement. Third, it provides a system-oriented framework that can support destination positioning, regional tourism coordination, and comparative analysis of source-market systems in platform-mediated tourism environments.

A related study using the same platform-based attraction system in Liaoning examined attraction-level tourist experience and sentiment structure through topic modeling and sentiment analysis [32]. The present study differs fundamentally in analytical objective, unit of analysis, and methodological design. Whereas the earlier study focused on attraction-level experiential themes and rating–sentiment dynamics, this study reconstructs city-level destination–origin linkages from IP-origin information and develops the MAI to measure source-market linkage capacity. The two studies therefore address different questions and produce non-overlapping analytical outputs.

2. Literature Review

2.1. UGC, Platform Data, and Destination Analysis

UGC and online review data have been widely used to examine destination image, eWOM effects, service quality, tourist satisfaction, and platform-mediated travel behavior. Earlier studies primarily treated reviews as textual or evaluative materials, while more recent studies have emphasized that platform data can also reveal spatial, relational, and structural characteristics of tourism systems. For example, TOURQUAL-based analysis demonstrates that UGC can be translated into structured service-quality evidence rather than simple sentiment signals [33]. Recent reviews and bibliometric mapping of tourist mobility, tourist tracking, and tourism footprints further show that digital traces have become important sources for measuring tourism dynamics at granular spatial and temporal scales [34,35,36].

2.2. Tourism Flows, Digital Footprints, and Network Methods

A second stream of research reconstructs tourist flows and destination networks from online travel diaries, mobile traces, geotagged data, and other digital footprints. These studies have advanced tourism geography by showing that destinations are embedded in relational networks rather than isolated administrative units. Recent work has further proposed multi-scale destination-network frameworks and digital-footprint-based analyses of tourist flow structures [37,38,39,40]. However, much of this research focuses on movement trajectories, destination co-visitation, or network topology. It does not directly measure how a destination city’s source-market linkage capacity is jointly shaped by market scale, diversity, openness, reach, temporal stability, and network embeddedness.

2.3. Composite Indicators and Source-Market Linkage Capacity

Traditional source-market studies and spatial-interaction models are useful for explaining origin shares, distance effects, and flow volumes [1,2,3,4,5,6], while network science provides indicators such as degree, betweenness, and PageRank to measure node position [31,37]. Composite-index approaches are also widely used in tourism and socio-economic systems when multidimensional performance cannot be captured by a single variable [41,42]. Nevertheless, a composite index must be theoretically justified, transparent in weighting, and interpreted as a diagnostic tool rather than as an absolute normative score. This is the rationale for constructing the MAI as a system-oriented measure of city-level source-market linkage capacity.

2.4. Comparison with Related Methods and Research Gap

Table 1. Comparison between related methods and the MAI framework.

Research Stream/Representative Studies	Main Data or Method	Strength	Limitation for Linkage-Capacity Measurement	Relationship to This Study
Spatial-interaction and source-market models [1,2,3,4,5,6]	Origin shares, distance decay, gravity-type modeling	Explain flow intensity and spatial constraints	Often aggregate flows and do not diagnose multidimensional city capacity	Provides distance and source-market foundation
UGC/eWOM and review analysis [13,14,15,16,17,18,19,33]	Text mining, sentiment, quality or satisfaction evaluation	Captures perception, service quality, and destination image	Usually treats reviews as evaluative material rather than structural origin evidence	Extends UGC use from evaluation to structural measurement
Platform-comparison studies [20,43]	Cross-platform comparison and bias analysis	Clarify data heterogeneity and selection bias	Do not construct a city-level source-market index	Supports cautious interpretation of Ctrip-origin data
Digital-footprint tourism-flow studies [21,22,23,24,25,26,27,28,29,30,34,36,37,41,42]	Travel blogs, mobile traces, digital footprints	Reveal spatial-temporal mobility and tourist-flow networks	Often focus on movement patterns rather than source-market capacity	Provides methodological basis for relational measurement
Tourism network studies [31,35]	Social network analysis and centrality indicators	Reveal node position and network embeddedness	Single centrality indicators cannot measure scale, diversity, reach, and stability together	MAI incorporates PageRank as one dimension only
Composite and policy-oriented tourism indicators [39,40]	Multidimensional indicators and dashboard logic	Useful for systemic diagnosis and policy comparison	Can be sensitive to indicator selection and weighting	MAI adopts transparent equal weighting and benchmark comparison
The present MAI framework	Seven normalized dimensions from review-origin data	Integrates size, diversity, openness, reach, stability, input strength, and centrality	Dependent on platform-visible review behavior and IP-origin proxy quality	Provides a diagnostic measure of city-level source-market linkage capacity

Note: This table provides a systematic comparison of related research streams and clarifies how the MAI differs from single-flow, UGC-evaluation, and network-centrality approaches.

summarizes representative related approaches. Existing methods provide important foundations, but they either emphasize aggregate market shares, perceptual UGC evaluation, digital-footprint mobility reconstruction, or network centrality. The unresolved methodological gap is a city-level framework that integrates observable source-market size, heterogeneity, openness, spatial reach, temporal continuity, effective input strength, and systemic centrality into one transparent diagnostic measure. The MAI addresses this gap by converting online review origin data into a multidimensional index for comparing destination-city linkage capacity.

3. Materials and Methods

3.1. Analytical Framework

This study examines city-level tourist source-market linkage capacity rather than overall tourism competitiveness. The analytical focus is on how destination cities connect to multiple source regions across space and time within a platform-based behavioral dataset. In this framework, prefecture-level cities are treated as destination nodes, while the IP-origin field is used as a proxy for tourist source regions. The aim is to measure the structural capacity of destination cities to attract, sustain, and organize tourism demand from heterogeneous origins.

In this study, tourist source-market linkage capacity is defined as the ability of a destination city to establish, maintain, diversify, and stabilize effective connections with tourist source regions. It is observed through origin-based review interactions, market breadth, spatial reach, temporal continuity, and network embeddedness. The concept therefore refers to a structural capacity of the destination-city system, not to overall tourism competitiveness or total visitor volume.

To operationalize this concept, the study constructs a Market Accessibility and Interaction Index (MAI) from seven dimensions: market scale, source diversity, interprovincial attraction, external effective radius, seasonal stability, input strength, and PageRank centrality. The index is calculated at the prefecture-level city scale and uses city-to-origin distance measures derived from a review-level Ctrip dataset in which the IP-origin field serves as a proxy for tourist source regions.

The seven dimensions were selected to correspond to complementary properties of a tourism source-market system. Market scale captures observable demand size; source diversity captures breadth and balance; interprovincial attraction captures openness beyond the local province; external effective radius captures spatial reach; seasonal stability captures temporal continuity; input strength captures effective external linkage intensity; and PageRank centrality captures systemic embeddedness in the directed source-to-city network. Together, these dimensions represent the size, heterogeneity, openness, reach, stability, intensity, and relational position of city-level linkage capacity.

The analytical procedure consists of four steps. First, attraction-level review records were collected, cleaned, and standardized to identify valid IP-origin information, destination cities, timestamps, and rating information. Second, standardized source-region and destination-city coordinates were assigned to each retained record, and city-to-origin distance measures were generated. Third, city-level indicators were derived from the retained review matrix and integrated into the MAI. Fourth, the robustness of the resulting city-level structure was assessed using a stricter rating-complete subsample.

3.2. Data Source and Sample Construction

This study uses online review data collected from Ctrip, one of the largest online travel platforms in China. The sampling frame was established from the official catalog of A-level tourist attractions released by the Liaoning Provincial Department of Culture and Tourism. According to the official provincial catalog available during the study period, Liaoning Province had 584 A-level scenic spots, which served as the full attraction population frame for this study.

Online review data were extracted from Ctrip during a unified collection window from 1 to 31 October 2025. Because of the platform’s display settings, no more than 3000 publicly visible reviews could be accessed for a single scenic spot at the time of collection. The dataset therefore represents a platform-visible review snapshot rather than the complete historical review archive of each attraction. Among the 584 officially listed A-level scenic spots, 332 were found to have publicly visible reviews on Ctrip. To ensure sufficient review volume for city-level source-market analysis, attractions with at least 100 visible reviews were first retained, yielding 106 scenic spots. In addition, six highly reviewed sub-attractions were included because they were displayed separately on Ctrip but belonged to officially recognized A-level scenic systems in the provincial catalog. The final attraction-level analytical frame therefore comprised 112 scenic spots, covering all 14 prefecture-level cities of Liaoning Province.

Within these 112 attractions, a total of 76,855 review records were collected and organized into a review-level Ctrip dataset. This dataset served as the basis for constructing the city-level source-market linkage dataset. After standardizing the IP-origin field and retaining only records with identifiable IP-origin information, 29,361 reviews remained. Excluding records with invalid timestamps yielded 29,328 reviews, and excluding the single record without a valid distance value yielded a final main analytical sample of 29,327 reviews for MAI estimation. In addition, a stricter rating-complete subsample of 28,337 reviews was constructed for robustness testing by retaining only main-sample records with non-missing rating information.

Because this study focuses on city-level tourist source-market linkage capacity, the attraction-level review data were subsequently aggregated to the prefecture-level city. The IP-origin field was interpreted as a proxy for the tourist source region rather than an exact residential location. The resulting dataset therefore captures the structure of destination–origin linkages observed in platform-based review behavior. Since overseas-origin observations were extremely sparse, the empirical interpretation focuses primarily on domestic interprovincial source-market linkages.

The present study draws on the same platform-based attraction system as the earlier attraction-level review-text study but uses a different analytical dataset and workflow [32]. The earlier article analyzed cleaned review texts to identify experiential themes and sentiment structures at the attraction level, whereas this study retains review records with valid IP-origin information, destination-city information, timestamps, and distance measures to reconstruct city-level source-market linkages. The analytical unit, variables, and outputs are therefore different. Raw review data were collected using the Octopus web crawler software (version 8.7.7). Data cleaning, origin standardization, coordinate assignment, distance calculation, indicator construction, network analysis, robustness testing, and visualization were conducted in Python 3.9.12. The main Python packages used included pandas 2.1.3, NumPy 1.26.4, SciPy 1.12.0, NetworkX 2.7.1, and Matplotlib 3.5.1. The sample construction procedure and analytical design are summarized in Figure 1 and

Table 2. Sample construction and analytical design.

Stage	Unit	Count	Share/Note
Official A-level attraction catalog in Liaoning Province	Scenic spots	584	Full official population frame
Attractions with publicly visible Ctrip reviews	Scenic spots	332	56.8% of the official catalog
Threshold-qualified attractions (≥100 visible reviews)	Scenic spots	106	Base attraction sample
Supplementary high-review sub-attractions	Scenic spots	6	Displayed separately on Ctrip but belonged to officially recognized A-level scenic systems
Final attraction-level analytical frame	Scenic spots	112	Covers all 14 prefecture-level cities in Liaoning
Review records collected from Ctrip	Reviews	76,855	Review-level Ctrip dataset
Reviews with identifiable IP-origin information	Reviews	29,361	Valid-origin review pool
Valid origin + destination city + timestamp	Reviews	29,328	Valid-origin review pool with valid timestamps
Main analytical sample for MAI	Reviews	29,327	Valid origin + destination city + timestamp + distance information
Rating-complete subsample	Reviews	28,337	Main analytical sample with non-missing rating information

Note: The attraction-level analytical frame was constructed from the official A-level attraction catalog released by the Liaoning Provincial Department of Culture and Tourism. Online review data were extracted from Ctrip in October 2025. Due to platform display constraints, a maximum of 3000 publicly visible reviews could be accessed for each scenic spot at the time of collection. The final city-level MAI analysis is based on the retained review records aggregated to the prefecture-level city.

.

3.3. Origin and Destination Standardization and Distance Measurement

3.3.1. Standardization of Source-Origin and Destination-City Fields

The IP-origin field was standardized into a unified source-region variable. Domestic origins were normalized to the provincial level, with directly administered municipalities retained as independent units, including Beijing, Shanghai, Tianjin, and Chongqing. Hong Kong, Macao, and Taiwan were treated as separate Chinese regional units. Foreign observations, where identifiable, were retained as country-level origins. Destination-city fields were standardized to the prefecture-level city names within Liaoning Province.

The IP-origin field was treated as an approximate source-region indicator rather than an exact residential address. Potential mismatches may arise from mobile roaming, VPN use, corporate IP pools, temporary travel locations, and platform-side location inference. To reduce the influence of these uncertainties, this study aggregated domestic origins to the provincial level, excluded records with unknown or invalid origin information, interpreted results as platform-observed source-market linkages rather than census-level tourist flows, and focused on domestic interprovincial patterns. This treatment is consistent with recent discussions of selection bias and representativeness in crowdsourced tourism big data [43].

3.3.2. Coordinate Assignment

Destination coordinates were assigned using city-centroid proxies because the analytical unit of this study is the destination city rather than the individual scenic spot. Source coordinates were assigned using a standardized coordinate lookup table. For domestic origins, provincial-capital coordinates were used as source proxies; for foreign origins, national-capital coordinates were used where available. Liaoning as a source region was represented by the provincial-capital proxy. This treatment provides a consistent geographic basis for city-level distance measurement.

3.3.3. Distance Measurement

Geographic distance was measured using the great-circle distance between the source-region centroid and the destination-city centroid. Let (${\mathrm{\varphi }}_{\mathrm{i}}$, ${\mathrm{\lambda }}_{\mathrm{i}}$) denote the latitude and longitude of source region $\mathrm{i}$, and let (${\mathrm{\varphi }}_{\mathrm{j}}$, ${\mathrm{\lambda }}_{\mathrm{j}}$) denote the latitude and longitude of destination city $\mathrm{j}$. The haversine distance is computed as:

$${\mathrm{d}}_{\mathrm{i}\mathrm{j}}=2R\arcsin \left(\sqrt{{\sin }^2\left({\displaystyle \frac{{\mathrm{\varphi }}_{\mathrm{i}}-{\mathrm{\varphi }}_{\mathrm{j}}}{2}}\right)+\cos ({\mathrm{\varphi }}_{\mathrm{j}})\cos ({\mathrm{\varphi }}_{\mathrm{i}}){\sin }^2\left({\displaystyle \frac{{\mathrm{\lambda }}_{\mathrm{i}}-{\mathrm{\lambda }}_{\mathrm{j}}}{2}}\right)}\right),$$

(1)

where $\mathrm{R}$ is the mean radius of the Earth, set to 6371 km.

For each retained review record, the city-to-origin distance measure was recorded as ${\mathrm{d}}_{\mathrm{i}\mathrm{j}}$. In addition, the transformed term $\ln (1+{\mathrm{d}}_{\mathrm{i}\mathrm{j}})$ was generated for subsequent use in the external effective radius dimension. This specification aligns the distance variable with the city-level analytical unit used in the MAI and supports the distinction between local dependence and extra-provincial spatial reach.

3.4. Construction of the MAI

Table 3. Indicator system of the Market Accessibility and Interaction Index (MAI).

Dimension	Indicator	Definition	Operationalization	Expected Implication
Market scale	Retained review volume	Observable size of the city’s review-based source market	Total number of retained reviews associated with city i; log-transformed before normalization	Higher values indicate a larger observable source-market base
Source diversity	Origin richness	Breadth of represented source regions	Number of distinct source origins in city i	Higher values indicate broader source coverage
Source diversity	Shannon entropy	Balance of source-market composition	$\mathrm{E}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{p}{\mathrm{y}}_{\mathrm{i}}=-{\sum }_{\mathrm{j}}{\mathrm{p}}_{\mathrm{i}\mathrm{j}}\ln ({\mathrm{p}}_{\mathrm{i}\mathrm{j}})$	Higher values indicate a more even source structure
Interprovincial attraction	External share	Degree of attraction beyond Liaoning Province	$\mathrm{A}\mathrm{t}\mathrm{t}\mathrm{r}\mathrm{a}\mathrm{c}\mathrm{t}\mathrm{i}\mathrm{o}{\mathrm{n}}_{\mathrm{i}}=1-\mathrm{L}\mathrm{o}\mathrm{c}\mathrm{a}\mathrm{l}\mathrm{S}\mathrm{h}\mathrm{a}\mathrm{r}{\mathrm{e}}_{\mathrm{i}}$	Higher values indicate stronger extra-provincial openness
External effective radius	External median distance	Central tendency of extra-provincial market reach	Median distance of non-Liaoning origins only	Higher values indicate broader external spatial reach
External effective radius	External mean log distance	Average distance-based extension of extra-provincial demand	Mean of $\mathrm{l}\mathrm{n}(1+{\mathrm{d}}_{\mathrm{i}\mathrm{j}})$ for non-Liaoning origins only	Higher values indicate farther-reaching external linkages
Seasonal stability	Inverse monthly coefficient of variation	Stability of monthly review fluctuations	$\mathrm{C}\mathrm{V}\mathrm{I}\mathrm{n}{\mathrm{v}}_{\mathrm{i}}={\displaystyle \frac{1}{1+\mathrm{C}{\mathrm{V}}_{\mathrm{i}}}}.$	Higher values indicate lower temporal volatility
Seasonal stability	Active-month share	Continuity of market activity across time	Share of months with non-zero reviews	Higher values indicate greater temporal continuity
Input strength	Effective interprovincial linkage strength	Total effective external linkage received by city i	Sum of source-region shares above the threshold τ = 0.01	Higher values indicate stronger combined external input
PageRank centrality	Network embeddedness	Systemic centrality of city i in the province-to-city network	Weighted PageRank score on the directed network	Higher values indicate stronger embeddedness in the source-market system
MAI	Composite index	Overall city-level source-market linkage capacity	Arithmetic mean of the seven normalized dimensions	Higher values indicate stronger city-level linkage capacity

Note: The MAI integrates seven dimensions: market scale, source diversity, interprovincial attraction, external effective radius, seasonal stability, input strength, and PageRank centrality. The external effective radius is calculated only from non-Liaoning origins so that local dependence and extra-provincial spatial reach are measured separately.

presents the indicator system of the Market Accessibility and Interaction Index (MAI). The index integrates seven dimensions: market scale, source diversity, interprovincial attraction, external effective radius, seasonal stability, input strength, and PageRank centrality. Together, these dimensions capture the observable size, breadth, openness, spatial reach, temporal continuity, and network embeddedness of city-level tourist source-market linkage capacity.

3.4.1. Market Scale

Market scale captures the size of the observable review-based source market of city $\mathrm{i}$. It is measured by the number of retained reviews associated with the city:

$${\mathrm{N}}_{\mathrm{i}}={\sum }_{\mathrm{j}}{\mathrm{n}}_{\mathrm{i}\mathrm{j}}$$

(2)

where ${\mathbf{n}}_{\mathbf{i}\mathbf{j}}$ is the number of retained reviews from source region $\mathbf{j}$ to city $\mathbf{i}$. Because review counts are right-skewed, the scale dimension is log-transformed before normalization:

$$\mathrm{S}\mathrm{c}\mathrm{a}\mathrm{l}{\mathrm{e}}_{\mathrm{i}}=\ln (1+{\mathrm{N}}_{\mathrm{i}})$$

(3)

3.4.2. Source Diversity

Source diversity reflects the breadth and balance of a city’s source-market structure. It is measured through two components: origin richness and Shannon entropy.

Origin richness is defined as:

$$\mathrm{R}\mathrm{i}\mathrm{c}\mathrm{h}\mathrm{n}\mathrm{e}\mathrm{s}{\mathrm{s}}_{\mathrm{i}}={\sum }_{\mathrm{j}}\mathrm{I}({\mathrm{n}}_{\mathrm{i}\mathrm{j}}>0)$$

(4)

where $\mathbf{I}(·)$ is an indicator function equal to 1 when the condition is satisfied.

The source-region share of city $\mathbf{i}$ from origin $\mathbf{j}$ is:

$${\mathrm{p}}_{\mathrm{i}\mathrm{j}}={\displaystyle \frac{{\mathrm{n}}_{\mathrm{i}\mathrm{j}}}{{\sum }_{\mathrm{j}}{\mathrm{n}}_{\mathrm{i}\mathrm{j}}}}$$

(5)

Shannon entropy is then defined as:

$$\mathrm{E}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{p}{\mathrm{y}}_{\mathrm{i}}=-{\sum }_{\mathrm{j}}{\mathrm{p}}_{\mathrm{i}\mathrm{j}}\ln ({\mathrm{p}}_{\mathrm{i}\mathrm{j}})$$

(6)

After normalization, the diversity dimension is calculated as:

$$\mathrm{D}\mathrm{i}\mathrm{v}\mathrm{e}\mathrm{r}\mathrm{s}\mathrm{i}\mathrm{t}{\mathrm{y}}_{\mathrm{i}}={\displaystyle \frac{1}{2}}(\mathrm{R}\mathrm{i}\mathrm{c}\mathrm{h}\mathrm{n}\mathrm{e}\mathrm{s}{\mathrm{s}}_{\mathrm{i}}^{*}+\mathrm{E}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{p}{\mathrm{y}}_{\mathrm{i}}^{*}),$$

(7)

where the superscript * denotes min–max normalized values.

3.4.3. Interprovincial Attraction

Interprovincial attraction measures the extent to which a city attracts reviews from outside Liaoning Province. Let ${\mathbf{L}}_{\mathbf{i}}$ denote the number of Liaoning-origin reviews associated with city $\mathbf{i}$, and let ${\mathbf{N}}_{\mathbf{i}}$ denote the total number of retained reviews of city $\mathbf{i}$. The local-share term is:

$$\mathrm{L}\mathrm{o}\mathrm{c}\mathrm{a}\mathrm{l}\mathrm{S}\mathrm{h}\mathrm{a}\mathrm{r}{\mathrm{e}}_{\mathrm{i}}={\displaystyle \frac{{\mathrm{L}}_{\mathrm{i}}}{{\mathrm{N}}_{\mathrm{i}}}}$$

(8)

Interprovincial attraction is then defined as:

$$\mathrm{A}\mathrm{t}\mathrm{t}\mathrm{r}\mathrm{a}\mathrm{c}\mathrm{t}\mathrm{i}\mathrm{o}{\mathrm{n}}_{\mathrm{i}}=1-\mathrm{L}\mathrm{o}\mathrm{c}\mathrm{a}\mathrm{l}\mathrm{S}\mathrm{h}\mathrm{a}\mathrm{r}{\mathrm{e}}_{\mathrm{i}}$$

(9)

3.4.4. External Effective Radius

The radius dimension is defined only from non-Liaoning origins. This specification separates local dependence from extra-provincial spatial reach and therefore measures the spatial extension of a city’s external source-market linkages more directly.

Let ${\mathsf{\Omega }}_{\mathbf{i}}$ denote the set of non-Liaoning reviews associated with city $\mathbf{i}$. The first component of the external-radius dimension is the median external distance:

$$\mathrm{M}\mathrm{e}\mathrm{d}\mathrm{D}\mathrm{i}\mathrm{s}{\mathrm{t}}_{\mathrm{i}}^{\mathrm{e}\mathrm{x}\mathrm{t}}=\mathrm{m}\mathrm{e}\mathrm{d}\mathrm{i}\mathrm{a}\mathrm{n}\left({\mathrm{d}}_{\mathrm{i}\mathrm{j}}\right),\mathrm{j}\in {\mathrm{\Omega }}_{\mathrm{i}}$$

(10)

The second component is the mean log-transformed external distance:

$$\mathrm{L}\mathrm{o}\mathrm{g}\mathrm{D}\mathrm{i}\mathrm{s}{\mathrm{t}}_{\mathrm{i}}^{\mathrm{e}\mathrm{x}\mathrm{t}}={\displaystyle \frac{1}{\left|{\mathrm{\Omega }}_{\mathrm{i}}\right|}}{\sum }_{\mathrm{j}\in {\mathrm{\Omega }}_{\mathrm{i}}}\ln (1+{\mathrm{d}}_{\mathrm{i}\mathrm{j}})$$

(11)

After normalization, the external-radius dimension is defined as:

$$\mathrm{R}\mathrm{a}\mathrm{d}\mathrm{i}\mathrm{u}{\mathrm{s}}_{\mathrm{i}}={\displaystyle \frac{1}{2}}\left[{\left(\mathrm{M}\mathrm{e}\mathrm{d}\mathrm{D}\mathrm{i}\mathrm{s}{\mathrm{t}}_{\mathrm{i}}^{\mathrm{e}\mathrm{x}\mathrm{t}}\right)}^{\ast }+{\left(\mathrm{L}\mathrm{o}\mathrm{g}\mathrm{D}\mathrm{i}{\mathrm{s}\mathrm{t}}_{\mathrm{i}}^{\mathrm{e}\mathrm{x}\mathrm{t}}\right)}^{\ast }\right]$$

(12)

This dimension captures the spatial reach of a city’s extra-provincial source-market connections rather than the overall distance profile of all reviews.

3.4.5. Seasonal Stability

Seasonal stability captures the extent to which a city’s review-based demand remains active and relatively stable across months. Let ${\mathbf{m}}_{\mathbf{i}\mathbf{t}}$ denote the number of reviews of city $\mathbf{i}$ in month $\mathbf{t}$, and let $\mathbf{T}$ denote the total number of months in the observed period. The monthly mean is:

$${\mathrm{m}}_{\mathrm{i}}={\displaystyle \frac{1}{\mathrm{T}}}{\sum }_{\mathrm{t}}{\mathrm{m}}_{\mathrm{i}\mathrm{t}}$$

(13)

The monthly standard deviation is:

$${\mathrm{\sigma }}_{\mathrm{i}}=\sqrt{{\displaystyle \frac{1}{\mathrm{T}}}{\sum }_{\mathrm{t}}({\mathrm{m}}_{\mathrm{i}\mathrm{t}}-{\mathrm{m}}_{\mathrm{i}}{)}^2}$$

(14)

The coefficient of variation is:

$$\mathrm{C}{\mathrm{V}}_{\mathrm{i}}={\displaystyle \frac{{\mathrm{\sigma }}_{\mathrm{i}}}{{\overline{\mathrm{m}}}_{\mathrm{i}}}}$$

(15)

To convert fluctuation into a positive stability measure, the inverse coefficient term is defined as:

$$\mathrm{C}\mathrm{V}\mathrm{I}\mathrm{n}{\mathrm{v}}_{\mathrm{i}}={\displaystyle \frac{1}{1+\mathrm{C}{\mathrm{V}}_{\mathrm{i}}}}.$$

(16)

The active-month share is:

$$\mathrm{A}\mathrm{c}\mathrm{t}\mathrm{i}\mathrm{v}{\mathrm{e}}_{\mathrm{i}}={\displaystyle \frac{1}{\mathrm{T}}}{\sum }_{\mathrm{t}}\mathrm{I}({\mathrm{m}}_{\mathrm{i}\mathrm{t}}>0)$$

(17)

The seasonal-stability dimension is then calculated as:

$$\mathrm{S}\mathrm{t}\mathrm{a}\mathrm{b}\mathrm{i}\mathrm{l}\mathrm{i}\mathrm{t}{\mathrm{y}}_{\mathrm{i}}={\displaystyle \frac{1}{2}}(\mathrm{C}\mathrm{V}\mathrm{I}\mathrm{n}{\mathrm{v}}_{\mathrm{i}}+\mathrm{A}\mathrm{c}\mathrm{t}\mathrm{i}\mathrm{v}{\mathrm{e}}_{\mathrm{i}}).$$

(18)

This formulation jointly captures both continuity and fluctuation.

3.4.6. Province-to-City Network Construction

To represent the structural relation between source regions and destination cities, a directed weighted network was constructed in which source regions are origin nodes and destination cities are destination nodes. The edge from source region $\mathbf{j}$ to city $\mathbf{i}$ is based on the source-region share ${\mathbf{p}}_{\mathbf{i}\mathbf{j}}$, but only when this share exceeds a minimum effective-link threshold.

The effective edge weight is defined as:

$${\mathrm{w}}_{\mathrm{j}\mathrm{i}}=\left\{\begin{array}{c}{\mathrm{p}}_{\mathrm{i}\mathrm{j}},{\mathrm{p}}_{\mathrm{i}\mathrm{j}}\ge \mathrm{\tau }\\ 0,{\mathrm{p}}_{\mathrm{i}\mathrm{j}}<\mathrm{\tau }\end{array}\right.$$

(19)

where $\mathsf{\tau }$ = 0.01 is the threshold used in this study.

3.4.7. Input Strength

Input strength measures the total effective external linkage received by city $\mathbf{i}$. It is defined as:

$$\mathrm{I}\mathrm{n}\mathrm{S}\mathrm{t}\mathrm{r}\mathrm{e}\mathrm{n}\mathrm{g}\mathrm{t}{\mathrm{h}}_{\mathrm{i}}={\sum }_{\mathrm{j}}{\mathrm{w}}_{\mathrm{i}\mathrm{j}}$$

(20)

3.4.8. PageRank Centrality

To capture a city’s embeddedness in the wider province-to-city source-market system, PageRank centrality is calculated on the directed weighted network. The PageRank score of node $\mathbf{i}$ is given by:

$$\mathrm{P}{\mathrm{R}}_{\mathrm{i}}={\displaystyle \frac{1-\mathrm{d}}{\mathrm{N}}}+\mathrm{d}{\sum }_{\mathrm{j}\in {\mathrm{M}}_{\mathrm{i}}}{\displaystyle \frac{{\mathrm{w}}_{\mathrm{j}\mathrm{i}}}{{\sum }_{\mathrm{k}}{\mathrm{w}}_{\mathrm{j}\mathrm{k}}}}\mathrm{P}{\mathrm{R}}_{\mathrm{j}}$$

(21)

where $\mathbf{d}$ is the damping factor, set to 0.85; $\mathbf{N}$ is the total number of nodes in the network; and ${\mathbf{M}}_{\mathbf{i}}$ is the set of nodes linking to node $\mathbf{i}$. In this study, the city-level PageRank value is used as the final network-centrality component of the MAI.

3.5. Normalization and Aggregation of the MAI

Because the seven dimensions are measured in different units, all component indicators are standardized using min–max normalization:

$${\mathrm{x}}_{\mathrm{i}}^{\ast }={\displaystyle \frac{{\mathrm{x}}_{\mathrm{i}}-{\mathrm{x}}_{\mathrm{m}\mathrm{i}\mathrm{n}}}{{\mathrm{x}}_{\mathrm{m}\mathrm{a}\mathrm{x}}-{\mathrm{x}}_{\mathrm{m}\mathrm{i}\mathrm{n}}}}$$

(22)

where ${\mathbf{x}}_{\mathbf{i}}$ is the observed value of indicator $\mathbf{x}$ for city $\mathbf{i}$, and ${\mathbf{X}}_{\mathbf{m}\mathbf{i}\mathbf{n}}$ and ${\mathbf{X}}_{\mathbf{m}\mathbf{a}\mathbf{x}}$ are the minimum and maximum values of that indicator across all cities.

The MAI is then calculated as the arithmetic mean of the seven normalized dimensions:

$$\mathbf{M}\mathbf{A}{\mathbf{I}}_{\mathbf{i}}={\displaystyle \frac{1}{7}}\left(\mathbf{S}\mathbf{c}\mathbf{a}\mathbf{l}{\mathbf{e}}_{\mathbf{i}}^{\mathbf{\ast }}+\mathbf{D}\mathbf{i}\mathbf{v}\mathbf{e}\mathbf{r}\mathbf{s}\mathbf{i}\mathbf{t}{\mathbf{y}}_{\mathbf{i}}+\mathbf{A}\mathbf{t}\mathbf{t}\mathbf{r}\mathbf{a}\mathbf{c}\mathbf{t}\mathbf{i}\mathbf{o}{\mathbf{n}}_{\mathbf{i}}^{\mathbf{\ast }}+\mathbf{R}\mathbf{a}\mathbf{d}\mathbf{i}\mathbf{u}{\mathbf{s}}_{\mathbf{i}}+\mathbf{S}\mathbf{t}\mathbf{a}\mathbf{b}\mathbf{i}\mathbf{l}\mathbf{i}\mathbf{t}{\mathbf{y}}_{\mathbf{i}}^{\mathbf{\ast }}+\mathbf{I}\mathbf{n}\mathbf{S}\mathbf{t}\mathbf{r}\mathbf{e}\mathbf{n}\mathbf{g}\mathbf{t}{\mathbf{h}}_{\mathbf{i}}^{\mathbf{\ast }}+\mathbf{P}{\mathbf{R}}_{\mathbf{i}}^{\mathbf{\ast }}\right).$$

Equal weighting is adopted for three reasons. First, the MAI is designed as a transparent diagnostic framework rather than a normative destination-performance score; assigning equal weights avoids privileging one theoretical dimension without external policy consensus. Second, alternative data-driven methods such as principal component analysis or entropy weighting may over-emphasize dimensions with higher variance in a small 14-city sample and reduce interpretability. Third, expert-based approaches such as Delphi weighting are useful for policy prioritization but introduce another layer of subjective judgment. Equal weighting is therefore used as the baseline specification, while the benchmark comparison and robustness checks are reported to show how the composite index differs from single-indicator rankings. The implication is that MAI should be interpreted as a balanced structural diagnostic measure; future applications may test alternative weighting schemes when larger samples or expert panels are available.

Table 4. Descriptive statistics of the MAI dimensions.

Dimension	Mean	SD	Min	Max
Scale index	0.535	0.325	0.000	1.000
Diversity index	0.625	0.275	0.122	0.963
Interprovincial attraction index	0.619	0.288	0.000	1.000
External radius index	0.351	0.270	0.006	1.000
Seasonal stability index	0.718	0.276	0.000	1.000
Input strength index	0.627	0.257	0.000	1.000
PageRank index	0.368	0.287	0.000	1.000
MAI	0.549	0.143	0.271	0.770

Note: All indices are min–max normalized to the [0, 1] interval. The MAI is calculated as the arithmetic mean of the seven normalized dimensions.

reports the descriptive statistics of the normalized MAI dimensions.

3.6. Robustness Analysis

To test whether the observed city-level structure is sensitive to stricter record completeness, the MAI was also estimated for the rating-complete subsample of 28,337 reviews. Robustness was assessed in two ways.

First, score consistency between the main analytical sample and the rating-complete subsample was examined using the Pearson correlation coefficient:

$$\mathrm{r}={\displaystyle \frac{{\sum }_{\mathrm{i}}({\mathrm{X}}_{\mathrm{i}}-\overline{\mathrm{X}}\left)\right({\mathrm{Y}}_{\mathrm{i}}-\overline{\mathrm{Y}})}{\sqrt{{\sum }_{\mathrm{i}}({\mathrm{X}}_{\mathrm{i}}-\overline{\mathrm{X}}{)}^2}\sqrt{{\sum }_{\mathrm{i}}({\mathrm{Y}}_{\mathrm{i}}-\overline{\mathrm{Y}}{)}^2}}}$$

(23)

where ${\mathbf{X}}_{\mathbf{i}}$ and ${\mathbf{Y}}_{\mathbf{i}}$ denote the MAI scores of city $\mathbf{i}$ under the two sample specifications.

Second, rank consistency was assessed descriptively through a city-level scatter comparison between the two MAI estimates. Because the number of cities is small and the purpose is interpretive rather than inferential, the robustness analysis focuses on structural stability rather than formal hypothesis testing.

4. Results

4.1. Overall Differentiation in City-Level Source-Market Linkage Capacity

The MAI reveals substantial inter-city variation in tourist source-market linkage capacity across Liaoning. As shown in Figure 2 and

Table 5. Normalized city-level MAI dimensions and rankings.

City	Scale	Diversity	Interprov. Attraction	External Radius	Seasonal Stability	Input Strength	PageRank	MAI	Rank
Dalian	1.00	0.96	0.83	0.36	0.94	0.82	0.47	0.770	1
Huludao	0.58	0.75	0.97	0.01	0.85	0.96	0.62	0.672	2
Dandong	0.86	0.86	0.66	0.59	0.83	0.61	0.20	0.659	3
Jinzhou	0.83	0.88	0.91	0.07	0.83	0.85	0.24	0.657	4
Tieling	0.07	0.60	1.00	0.45	0.79	1.00	1.00	0.655	5
Panjin	0.61	0.81	0.80	0.32	0.76	0.78	0.26	0.618	6
Anshan	0.70	0.73	0.58	0.40	1.00	0.69	0.25	0.583	7
Benxi	0.89	0.81	0.55	0.18	0.84	0.54	0.15	0.566	8
Fushun	0.61	0.68	0.53	0.35	0.93	0.51	0.26	0.537	9
Chaoyang	0.39	0.66	0.79	0.02	0.90	0.76	0.20	0.531	10
Shenyang	0.46	0.42	0.44	0.67	0.66	0.50	0.15	0.442	11
Liaoyang	0.02	0.15	0.41	1.00	0.00	0.50	0.48	0.366	12
Fuxin	0.00	0.12	0.26	0.26	0.64	0.45	0.87	0.357	13
Yingkou	0.48	0.37	0.00	0.14	0.90	0.00	0.00	0.271	14

Note: Higher MAI values indicate stronger city-level source-market linkage capacity. The seven component dimensions are min-max normalized to the [0, 1] interval. The table is included to improve transparency and allow readers to verify how the overall MAI ranking is derived from the seven normalized dimensions. Rankings are based on the MAI calculated from the main analytical sample of 29,327 retained review records.

, Dalian records the highest MAI score and emerges as the strongest source-market linkage hub in the province. It is followed by Huludao, Dandong, and Jinzhou, which together form the strongest follow-up group in the current city-level system. By contrast, Liaoyang, Fuxin, and Yingkou remain at the lower end of the ranking and display comparatively weak overall linkage capacity.

These ranking patterns indicate that city-level source-market linkage capacity is not reducible to market size alone. While Dalian combines large scale with strong multidimensional performance, several other cities achieve relatively strong positions through more specific combinations of openness, external reach, or network strength. Conversely, some cities with greater administrative importance do not necessarily rank highly in the MAI. This confirms that the index captures a structural property of destination–origin linkage rather than a general measure of urban prominence.

The ranking pattern also suggests that the provincial source-market system is highly uneven. Rather than displaying a smooth hierarchical structure, it contains one dominant hub, several strong but differentiated secondary cities, a broader middle group, and a small set of lower-linkage cities. This provides initial evidence that tourist source-market linkage capacity is a multidimensional and unevenly distributed system characteristic.

Although the empirical case is Liaoning, this tiered pattern has broader implications for other regional tourism systems. It suggests that provincial or regional destinations should not be evaluated only by flagship-city performance. Instead, regional planners need to identify which cities function as linkage hubs, which cities have specialized external-market strengths, and which cities remain structurally constrained by narrow source bases or weak network positions.

4.2. Multidimensional Profiles of High- and Middle-Linkage Cities

The city ranking alone does not explain why different cities occupy similar or different positions in the MAI. For this reason, the seven normalized dimensions are compared in Figure 3, which shows the multidimensional structure of source-market linkage capacity across the 14 cities.

Figure 3 shows that Dalian occupies a clear leading position because it performs strongly across most dimensions simultaneously. It combines the largest retained review base, high source diversity, strong interprovincial attraction, relatively strong seasonal stability, and a favorable network position. Its leading rank is therefore not driven by a single dominant component, but by a broad multidimensional advantage.

Dalian’s dominance can be interpreted as the combined outcome of economic, geographic, and infrastructure advantages. Economically, Dalian has a stronger urban service base and a more mature destination brand than many other cities in the province. Geographically, its coastal tourism resources, leisure image, and port-city identity provide distinctive external appeal. Infrastructurally, stronger air, rail, port, and urban service connections improve its accessibility for interprovincial visitors. Its leading MAI score therefore reflects an integrated source-market linkage system rather than a single-dimensional flow advantage.

The second group of cities—especially Huludao, Dandong, and Jinzhou—exhibits a more differentiated structure. Huludao benefits substantially from high interprovincial attraction and relatively strong input strength, despite a much smaller review base than Dalian. Dandong combines a comparatively large retained review volume with a broad external radius, which supports its high ranking even though its network position is not as strong as that of Dalian. Jinzhou performs strongly through a balanced combination of diversity, interprovincial attraction, and effective network linkage. These results indicate that relatively strong source-market systems may arise from different structural pathways rather than from a single dominant mechanism.

A broader middle group—including Panjin, Anshan, Benxi, Fushun, and Chaoyang—shows mixed performance across the seven dimensions. These cities do not stand out as clearly as the higher-ranked group, but they also do not exhibit uniformly weak profiles. Instead, each displays selective strengths alongside clear limitations. Panjin, for example, retains a relatively balanced multidimensional profile, while Benxi benefits more from review volume and source coverage than from external openness. This pattern confirms that medium-ranked cities should not be interpreted as homogeneous cases; rather, they represent structurally diverse forms of partial linkage capacity.

One special case in the upper half of the ranking is Tieling, which scores highly under the retained indicator structure. However, this result should be interpreted as sample-sensitive because its retained review base is relatively small (131 reviews). Tieling’s high MAI appears to be driven more by favorable normalized openness and network-related indicators than by a large observable market scale. Since min-max normalization can amplify small-sample cases when several proportional indicators are high, Tieling is retained for complete provincial comparison but should be treated as a special case rather than as a fully established high-capacity hub comparable to Dalian.

4.3. Lower-Linkage Cities and the Case of Shenyang

The lower part of the ranking includes Shenyang, Liaoyang, Fuxin, and Yingkou. These cities differ in detail, but all display weaker overall source-market linkage capacity than the higher-ranked cities in the MAI system.

Among them, Shenyang is the most noteworthy case. As the provincial capital, it might be expected to rank more strongly in a conventional urban hierarchy. However, the MAI places Shenyang in the lower group. This result does not imply that Shenyang is a weak city in general tourism terms. Rather, it indicates that, within the retained attraction-centered review sample, Shenyang does not perform strongly as a diversified, externally open, and strongly embedded source-market linkage system.

The structure of Shenyang’s retained profile helps explain this result. Its external market radius is not especially weak once only non-Liaoning origins are considered, which means that its lower ranking is not primarily due to distance compression caused by local-origin observations. Instead, Shenyang is mainly constrained by a relatively small retained review base, limited source diversity, modest interprovincial attraction, and comparatively weak network embeddedness. In other words, what is weak is not necessarily Shenyang’s overall urban tourism significance, but its representation in the retained review-based attraction system used in this study.

At the lower end, Liaoyang, Fuxin, and Yingkou display more clearly constrained structures. These cities combine limited retained review volume with relatively weak external linkage capacity or narrower source-market structures. Yingkou is particularly notable for its very low interprovincial attraction and weak overall network strength, which together help explain its last-place ranking in the MAI.

These results reinforce an important point: the MAI measures source-market linkage capacity in attraction-centered review behavior, not general urban status. The lower position of Shenyang therefore should be interpreted as a structurally meaningful empirical result rather than as a computational anomaly.

4.4. Source-Market Structure Across Cities and Major Origin Regions

To further examine how the city-level system is formed, Figure 4 visualizes the source composition of destination cities across the major retained source regions. This figure makes it possible to move beyond the aggregate index and observe how different cities depend on or extend beyond major origins.

Figure 4 shows that the city-level source-market structure is highly uneven. Several cities remain strongly dependent on intra-provincial demand, while others show broader engagement with major extra-provincial origins. This structural variation helps explain why cities with comparable overall review volume may still obtain very different MAI scores.

A key pattern in Figure 4 is that high-linkage cities tend to show a relatively more diversified and outwardly oriented source structure. Dalian, Huludao, Dandong, and Jinzhou all display stronger engagement with major extra-provincial origins than the lower-ranked cities. By contrast, lower-linkage cities tend to rely more heavily on a narrower subset of origins and exhibit weaker penetration into broader external markets.

This difference in source composition is particularly important because the MAI is designed to distinguish between cities that merely accumulate demand and cities that build wider and more open market linkages. Figure 4 therefore provides structural support for the ranking results in Figure 2 and the multidimensional profiles in Figure 3.

4.5. Interprovincial Source-Market Contribution and Network Structure

The source-market system is not shaped equally by all origins. To identify the strongest external contributors, Figure 5 presents the output strength of major extra-provincial source regions, and Figure 6 visualizes the province-to-city linkage network of major nodes.

Figure 5 shows that the external component of the source-market system is driven primarily by a limited set of major source regions rather than by a uniformly distributed national market. This indicates that city-level differentiation in Liaoning is structured through a selective interprovincial system in which a relatively small number of strong origins provide most of the effective external linkage.

Figure 6 further illustrates that destination cities are embedded in a differentiated province-to-city linkage network rather than in isolated bilateral flows. The network confirms the dominant role of Dalian as the strongest hub, while also showing that other higher-ranked cities occupy distinct secondary positions in the external market structure. Importantly, the network is not a simple reflection of city size. Instead, it captures the system-level pattern of how major origins connect to cities with different combinations of openness, diversity, and effective attraction.

Taken together, Figure 5 and Figure 6 demonstrate that the source-market system identified in this study is fundamentally relational. Cities differ not only because they have different numbers of reviews but because they occupy different positions in a broader structure of origin–destination connections.

4.6. Robustness of the MAI Structure

To evaluate whether the city-level structure of the MAI depends strongly on stricter record completeness, the index was also estimated for the rating-complete subsample of 28,337 reviews. The comparison between the main analytical sample and the stricter subsample is shown in Figure 7, while the city-level comparison can be summarized in

Table 6. Robustness comparison between the main analytical sample and the rating-complete subsample.

City	Main-Sample MAI	Rating-Complete MAI	Main Rank	Rating-Complete Rank	Rank Change
Dalian	0.770	0.771	1	1	0
Huludao	0.672	0.671	2	2	0
Dandong	0.659	0.547	3	8	5
Jinzhou	0.657	0.657	4	3	−1
Tieling	0.655	0.655	5	4	−1
Panjin	0.618	0.619	6	5	−1
Anshan	0.583	0.583	7	6	−1
Benxi	0.566	0.566	8	7	−1
Fushun	0.537	0.538	9	9	0
Chaoyang	0.531	0.531	10	10	0
Shenyang	0.442	0.442	11	11	0
Liaoyang	0.366	0.366	12	12	0
Fuxin	0.357	0.356	13	13	0
Yingkou	0.271	0.271	14	14	0

Note: The table compares city-level MAI scores and ranks between the main analytical sample (29,327 reviews) and the stricter rating-complete subsample (28,337 reviews). Positive rank change values indicate a lower position in the stricter subsample, while negative values indicate a higher position. Overall structural stability is high, although Dandong remains notably more sensitive than the other cities.

.

Figure 7 shows that the MAI is highly stable overall. The city-level scores derived from the stricter subsample remain closely aligned with those from the main analytical sample, indicating that the broad structure of inter-city differentiation is not strongly altered by the stricter completeness requirement. This supports the stability of the measurement framework.

At the same time, the robustness comparison suggests that not all city positions are equally stable. The top-ranked city remains clearly differentiated, and the lower group also remains broadly consistent. However, several cities in the middle-to-upper part of the ranking exhibit somewhat greater sensitivity. This implies that the empirical interpretation should focus more on linkage tiers than on the exact ranking position of each city.

Dandong is the most sensitive case in the robustness comparison. Its decline from third place in the main sample to eighth place in the rating-complete subsample indicates that the records removed by the stricter rating-completeness rule were not neutral for all MAI components. Because Dandong’s high position depends partly on a broad external-radius profile and a relatively large retained review base, removing non-rating records changes its relative component balance more visibly than in cities whose ranks are driven by more stable single-dimension advantages. This result does not overturn the overall structure, but it indicates that Dandong should be interpreted within the strong secondary tier rather than as a fixed third-ranked city.

In substantive terms, the robustness result strengthens the central claim of this study: the MAI captures a meaningful and stable system property of city-level source-market linkage capacity. Although some cases remain more sensitive than others, the overall pattern of one dominant hub, several strong secondary cities, a diverse middle group, and a constrained lower group remains unchanged.

4.7. Comparative Benchmark Against Single-Indicator Rankings

To further verify the value of the MAI, a comparative benchmark was conducted against several commonly used single indicators: retained review volume, origin richness, external share, external median distance, and PageRank centrality. These indicators correspond to scale, breadth, openness, distance reach, and network embeddedness, respectively. The benchmark compares each single-indicator ranking with the MAI ranking using Spearman rank correlation and top-five overlap. The purpose is not to reject single indicators, but to show that no single dimension can fully substitute for the multidimensional linkage-capacity measure.

Table 7. Comparative benchmark between MAI and single-indicator rankings.

Benchmark Indicator	Represents	Spearman Rho with MAI	Top-Five Overlap with MAI	Main Limitation
Retained review volume	Market scale	0.640	3/5	Cannot capture openness, reach, stability, or network position
Origin richness	Source breadth	0.639	3/5	Does not measure balance, intensity, or external distance
External share	Interprovincial openness	0.851	4/5	High openness may coexist with small sample size
External median distance	Spatial reach	−0.152	2/5	Long distance alone does not indicate strong linkage capacity
PageRank centrality	Network embeddedness	0.273	3/5	Centrality does not capture scale, diversity, or temporal stability

Note: Spearman rho values were calculated from the 14-city rankings in the main analytical sample. The comparison shows that MAI is related to several component indicators but cannot be reduced to any single indicator. This supports the methodological advantage of the composite framework.

presents the comparative benchmark between the MAI and selected single-indicator rankings.

The benchmark results confirm the rationale for a composite index. External share has the highest association with MAI, but it may overstate small-sample cases such as Tieling. Review volume and origin richness are moderately associated with MAI, but they do not identify cities that build wider or more embedded external linkages. PageRank captures network position, yet its low correlation with MAI shows that centrality alone is insufficient for diagnosing source-market linkage capacity. Therefore, the MAI adds value by integrating multiple system properties into a single transparent diagnostic measure.

5. Discussion

5.1. Tourist Source Markets as a System Property

The results of this study suggest that tourist source markets should be understood not merely as distributions of visitor origins, but as a system property of destination cities. Traditional source-market analysis usually identifies where visitors come from and how strongly specific origins contribute to destination demand. Although such approaches remain useful, they often describe source markets as static aggregates and do not fully capture how destination cities differ in their ability to maintain diversified, externally connected, and temporally stable market relations. This interpretation is consistent with the broader shift in tourism research from isolated destination analysis toward relational and networked understandings of tourism systems [31].

This interpretation is consistent with systems thinking in which outcomes emerge from interactions among components rather than from isolated attributes. In the present case, city-level linkage capacity emerges from the interaction between source-region diversity, spatial accessibility, temporal continuity, and network position. Network science further helps explain why cities with similar review volumes may occupy different structural roles: a city can be large but weakly embedded, small but externally open, or centrally positioned despite limited scale. The MAI is therefore designed to capture both component performance and relational embeddedness.

The MAI developed in this study addresses this limitation by integrating market scale, source diversity, interprovincial attraction, external effective radius, seasonal stability, and network embeddedness into a unified city-level framework. From this perspective, a city’s source market is not evaluated solely by the amount of observable demand but also by the structure and quality of its destination–origin linkages across space and time. The empirical results show that these dimensions do not always move together. Cities with relatively modest retained review volume may still achieve high linkage capacity if they are more open to extra-provincial demand and better embedded in the wider source-market network. Conversely, cities with larger local demand bases may remain structurally narrow in terms of market breadth and external linkage.

This shift from market size to market linkage capacity extends recent work on digital footprints, tourist market differentiation, and destination networks by offering a multidimensional destination-level systems measure rather than a descriptive account of flows alone [23,24,25,26,27,28,29,30,31]. In this sense, the MAI is intended not merely as a composite tourism index, but as a system-level measure of how destination cities maintain differentiated, externally connected, and temporally structured source-market relations.

5.2. Structural Pathways of Strong Linkage Cities

The empirical results indicate that a strong source-market linkage system is not defined by any single dimension. Instead, it emerges from the interaction of several reinforcing attributes. Dalian’s dominant position illustrates this clearly. Its advantage lies not only in scale, but also in the combination of high source diversity, substantial interprovincial attraction, broad external reach, relatively stable temporal activity, and strong network embeddedness. In other words, Dalian performs strongly because it is both large and well connected.

The next group of cities—especially Huludao, Dandong, and Jinzhou—shows that relatively strong source-market linkage capacity can arise from different structural pathways. Huludao benefits more from external openness and linkage intensity than from scale. Dandong combines a comparatively large retained review base with broad external market reach. Jinzhou performs strongly through a more balanced multidimensional profile, including diversity, interprovincial attraction, and effective linkage strength. These differences indicate that cities may achieve similar overall MAI scores through distinct combinations of system attributes.

A key methodological implication of this pattern is the value of the external-radius specification adopted in this study. By calculating the radius dimension only from non-Liaoning origins, the MAI separates two analytically distinct questions: whether a city attracts extra-provincial demand at all, and how far that extra-provincial demand extends once it exists. This prevents cities with strong local-market dependence from being penalized twice and improves the interpretability of the index. More broadly, the specification remains compatible with distance-based tourism research, which continues to show that spatial separation and effective reach remain important in shaping tourism demand [1,2,3,4,5]. Together, these results confirm that source-market linkage capacity should not be equated with tourism scale or administrative hierarchy alone, and that structurally strong cities may emerge through different combinations of openness, reach, and embeddedness [6,27,28,29,30].

5.3. The Case of Shenyang and the Interpretation of City Differences

Shenyang deserves special discussion because it is the provincial capital and might be expected to occupy a stronger position in a conventional tourism hierarchy. However, the MAI places it in the lower group. This result should not be interpreted as evidence that Shenyang is weak in overall tourism terms. Instead, it highlights the distinction between general urban prominence and attraction-centered source-market linkage capacity.

The distance-based results indicate that Shenyang’s lower ranking is not primarily a consequence of local-distance compression. Once only non-Liaoning origins are considered, Shenyang exhibits a relatively broad external market radius. Its lower MAI is instead associated with a comparatively small retained review base, limited source diversity, modest extra-provincial attraction, and relatively weak network embeddedness within the retained attraction-centered review system. This pattern suggests that platform-visible reviews of A-level attractions do not necessarily capture the full range of urban tourism functions. A provincial capital may possess strong business, leisure, transportation, or urban-consumption significance without being equally prominent in an attraction-centered source-market system derived from online scenic-spot reviews.

The robustness analysis further supports a cautious but confident interpretation of city differences. The overall city-level structure of the MAI is highly stable, and most cities either retain the same rank or shift by only one position under the stricter rating-complete subsample. Dandong remains the main sensitive case, indicating that some middle-to-upper positions are more contingent than others. This means that the strongest claim of the study concerns the persistence of the overall differentiation structure—one dominant hub, several strong secondary cities, a heterogeneous middle group, and a constrained lower group—rather than the fixity of every individual rank position. More broadly, this is consistent with recent work showing that platform-based tourism data are analytically powerful but remain subject to representational bias and sample sensitivity [20,43].

5.4. Theoretical Implications, Practical Applications, and Limitations

5.4.1. Theoretical Implications for Tourism, Geography, and Data Science

The results have theoretical implications for three areas. For tourism studies, the MAI reframes source markets from static market composition to dynamic linkage capacity, thereby connecting market analysis with tourism systems thinking. For tourism geography, the findings show that city-level tourism advantage is produced by the joint effects of spatial reach, market openness, and relational position rather than by administrative status alone. For data science, the study demonstrates how platform metadata can be transformed from descriptive digital traces into system-level indicators, while still requiring explicit attention to selection bias, proxy validity, and robustness.

5.4.2. Practical Implications and Limitations

The findings have practical implications for destination governance and regional tourism planning. Destination cities should not evaluate source-market performance solely in terms of review volume or local demand dominance. Recent studies of tourist market structure and tourism-flow networks in China show that destination strength may depend as much on market breadth and relational position as on aggregate volume [6,27,28,29,30]. A city with a large local visitor base may still face structural limitations if its market remains narrow, highly concentrated, or weakly embedded in extra-provincial linkages. Conversely, a city with more modest scale may show stronger long-term market potential if it maintains broader external connections.

The system perspective also suggests that regional tourism coordination should pay attention not only to flagship cities, but also to how cities occupy complementary positions in a broader source-market network. From this perspective, the provincial tourism system is not simply a collection of isolated destinations, but a differentiated network of cities with distinct linkage capacities and structural roles. This implication is consistent with destination-governance research emphasizing that regional tourism performance depends on relational coordination and governance capacity rather than on the isolated performance of individual places [44].

Several limitations should also be acknowledged. First, the study relies on platform-visible online reviews, which represent behavioral traces of reviewers rather than the full population of visitors. The observed source-market linkage structure should therefore be interpreted as a review-based market system rather than a complete census of actual tourist flows. Second, the IP-origin field is a proxy for source region rather than an exact measure of residence. Third, the attraction-level analytical frame is shaped by the review visibility and display mechanisms of Ctrip, including the maximum visible-review cap and the threshold-based scenic-spot selection. Fourth, some cities—most notably Tieling—achieve relatively strong scores under the retained indicator structure despite a limited retained review base and should therefore be interpreted cautiously.

Future research can extend this framework by integrating multi-platform review data, official mobility statistics, or booking data for external validation. It would also be useful to compare attraction-centered linkage capacity with broader urban tourism functions in cities such as Shenyang and to test the transferability of the framework in other regional destination systems [20,27,28,29,30,43,44].

6. Conclusions

6.1. Research Summary

This study developed a system-oriented framework for measuring city-level tourist source-market linkage capacity using online review origin data and applied it to the prefecture-level cities of Liaoning Province, China. Based on a review-level Ctrip dataset aggregated to the prefecture-level city, the study built a Market Accessibility and Interaction Index (MAI) from seven dimensions: market scale, source diversity, interprovincial attraction, external effective radius, seasonal stability, input strength, and PageRank centrality.

The empirical results reveal substantial inter-city differentiation in source-market linkage capacity. Dalian emerges as the strongest linkage hub in the provincial system, while Huludao, Dandong, and Jinzhou form the strongest follow-up group. A broader middle tier includes cities such as Panjin, Anshan, Benxi, Fushun, and Chaoyang, whereas Shenyang, Liaoyang, Fuxin, and Yingkou remain in the lower part of the ranking. These findings show that city-level tourist source-market systems are unevenly distributed and cannot be inferred from administrative status or tourism scale alone.

The results also demonstrate that relatively strong source-market linkage capacity may arise through different structural pathways. Some cities perform strongly because they combine large review volume with broad external connectivity, while others benefit more from external openness, spatial reach, or network embeddedness. This confirms that city-level source-market linkage is a multidimensional system property rather than a single-flow characteristic.

6.2. Innovations, Advantages, and Limitations

This study makes three main contributions. First, it reframes tourist source markets from a static description of origin composition into a measurable linkage system. Instead of asking only where tourists come from, it measures how destination cities differ in their capacity to sustain diversified, externally connected, and temporally stable source-market relations. This shift is consistent with research that treats destinations as relational entities embedded in broader tourism systems rather than as isolated units [31].

Second, it contributes a city-level measurement framework. In particular, the external effective radius calculated only from non-Liaoning origins improves the conceptual coherence of the index by separating local dependence from extra-provincial spatial reach. Third, it extends the use of online review data in tourism research. Rather than using review data only for sentiment, satisfaction, or destination-image analysis, it demonstrates that IP-origin information can also support structural market measurement and destination–origin system analysis. In this respect, the study complements recent work on digital footprints, tourism-flow structures, destination networks, and tourist market differentiation [6,20,27,28,29,30,31,32,35,36,37,38,39,40,41,42,43].

6.3. Theoretical Value and Potential Applications

The findings also have practical implications. Cities should not evaluate source-market performance solely through total demand volume or local visitor dominance. A city with a relatively modest review base may still possess strong development potential if it maintains broader and more effective external linkages. Conversely, a city with larger local demand may still face structural constraints if its source-market system remains narrow or weakly embedded in the broader interprovincial network. At the regional level, the results imply that tourism governance should pay greater attention to the complementary positions of cities within a wider source-market system and to the role of relational coordination in destination development [44].

6.4. Future Work

The main limitations of this study should be interpreted together with the innovations. The MAI provides a transparent and replicable diagnostic framework, but it is based on platform-visible online reviews rather than a complete census of tourist flows. The IP-origin field is useful for large-scale source-market observation, but it remains a proxy for source region rather than an exact residence measure. The attraction-level analytical frame is also shaped by platform display constraints, including the visible-review cap on Ctrip. Finally, cities with small retained review bases, especially Tieling, require cautious interpretation even when their normalized component scores are high.

Future research can extend the MAI framework in four directions. First, multi-platform validation can integrate data from Ctrip, Trip.com, Mafengwo, Dianping, Google Reviews, Booking.com, or other platforms to reduce platform-specific bias. Second, multi-source validation can compare review-origin results with official mobility statistics, booking data, mobile positioning records, or ticketing data. Third, cross-regional, national, and multi-country applications can test whether the MAI is transferable beyond Liaoning. Fourth, dynamic MAI models can be developed to examine how source-market linkage capacity changes before and after major events, infrastructure improvements, destination branding campaigns, or policy interventions.