Peak-Season Price Adjustments in Shared Accommodation: The Role of Platform-Certified Signals and User-Generated Signals

Abstract

This study investigates the factors influencing landlords’ price adjustments during peak seasons on accommodation-sharing platforms, focusing on the role of platform-certified and user-generated signals. Utilizing a comprehensive dataset of 11,795 observations from a leading Chinese accommodation-sharing platform, we employ binary logit models to investigate how these signals affect landlords’ pricing strategies during “festival” and “weekend” peak times. Our analysis reveals that both platform-certified signals (such as “Preferred House” badges) and user-generated signals (such as customer satisfaction ratings) significantly increase the probability and magnitude of price adjustments during both festival and weekend peak seasons. Specifically, houses with a “Preferred” status are up to 28 times more likely to have price hikes during weekends compared to non-preferred ones. Further analysis reveals that higher levels of landlord professionalism, measured by the number of properties managed, amplifies the impact of user-generated signals on both the probability and magnitude of price adjustments. However, as the level of professionalism increases, this effect diminishes, indicating that highly professional landlords may have less flexibility to adjust prices due to already-high baseline rates. Interestingly, landlord professionalism did not significantly influence the impact of platform-certified signals on price adjustments, suggesting that the influence of such signals remains consistent across different levels of landlord professionalism. These results underscore the significant roles that both types of signals and landlord professionalism play in shaping pricing strategies, offering valuable insights for platform management and policy formulation aimed at enhancing consumer trust and competitive dynamics in the sharing economy.

1. Introduction

The sharing economy, particularly through platforms like Airbnb, has revolutionized the way accommodations are rented and managed, challenging traditional economic models by leveraging underutilized resources [1,2]. Shared accommodation, a significant sector of the sharing economy, has gained increasing attention from researchers, policymakers, and practitioners worldwide [3]. It involves the provision of short-term accommodation services to guests through accommodation-sharing platforms that allow property owners to rent out their homes or apartments to guests [4,5]. A Statista report that was recently published forecasts that the global revenue in the shared accommodation sector is projected to reach around USD 600 billion by 2027, from USD 113 billion in 2021, with a compound annual growth rate of around 32%.

Despite its rapid growth, the accommodation-sharing sector grapples with significant challenges, notably the uncertainties and informational asymmetry between landlords and renters [6,7]. These uncertainties can undermine trust and transaction smoothness, which are crucial for the sustainability of sharing platforms. While shared accommodation has risen as a compelling alternative to traditional hospitality avenues, it comes with its own set of unique challenges, primarily due to these uncertainties regarding property conditions and amenities, leading to potential misconceptions and unmet expectations [8]. To promote trust and facilitate transactions between renters and landlords, shared accommodation platforms offer various signals. For instance, on popular shared accommodation rental platforms, such as Airbnb and 58.com (a popular Chinese online classifieds marketplace known for its housing rental and real estate listings), they display badges such as “Superhost” and “Preferred House” as platform-certified signals, while also displaying user-generated signals such as “favorable rate”, “satisfaction rate”, and “satisfaction number” based on user-generated reviews. These signals provide additional information to customers, allowing them to assess the credibility of the shared accommodation before making a reservation. Such signals play a critical role in mitigating the trust issues that arise due to the uncertainties associated with shared accommodation transactions [9,10,11].

Shared accommodation is also an illustration of the sharing economy within the tourism sector [12]. It exhibits a clear festival effect on key performance indicators such as “occupancy rate” and “room rate” [13]. This sector is distinctly affected by peak times, such as major festivals and regular weekends, which we define in our study to examine their impacts on pricing strategies. Specifically, ‘festival peak season’ refers to periods during which significant national or cultural festivals occur, such as the Spring Festival (Chinese New Year), National Day, and the International Labor Day Holidays. These periods typically experience a surge in travel and accommodation demand. In contrast, the ‘weekend peak season’ includes regular weekends throughout the year (except the weekends during festivals), allowing us to examine the effects of platform-certified and user-generated signals during these consistently high-demand periods.

The increased competition in the shared accommodation industry has intensified the need for optimal pricing strategies for landlords to remain profitable [14]. Unlike traditional hotels, shared accommodations lack conventional evaluation indicators, like star ratings, complicating pricing strategies [15]. Recent research has predominantly focused on the impact of individual signal types—either platform-certified or user-generated—on pricing strategies and consumer trust [16,17]. However, there is a noticeable void in studies that explore the synergistic effects of both signal types within a single framework. This study aims to fill this gap by examining how the interaction between platform-certified signals (such as “Preferred House”) and user-generated signals (such as ratings and reviews) influences landlords’ pricing strategies during peak-demand seasons [18,19,20].

2. Literature Review

The pricing dynamics of shared accommodations continue to be a core focus of research in both academia and the industry, with studies exploring various influencing factors such as host reputation, accommodation characteristics, landlord characteristics, rental policies, etc. Contributing to this body of knowledge, Ikkala et al. and Wang et al. have provided insights into the complexities and determinants of pricing in networked hospitality exchanges, like Airbnb, defining conceptual boundaries and offering multifaceted views of pricing strategies [21,22]. Priporas et al. and Xie et al. further enriched the dialogue on service quality and host incentives in shared accommodations, emphasizing their roles in shaping consumer perceptions and preferences [23,24].

Research has also explored the impact of macro-level factors like race and culture on pricing. Benítez-Aurioles and Kakar et al. analyzed racial implications and the effects of various factors on Airbnb pricing, revealing disparities and impacts of location and ratings [25,26]. Mariani and Predvoditeleva studied the influence of cultural traits and perceived experiences on hotel online ratings, highlighting the diversity in consumer behaviors and evaluations in the online hospitality sector [27].

Landlords in the shared accommodation industry are active participants in the sharing economy and display high levels of responsiveness to information stimuli in the pricing of their listings. Studies by Gibbs et al. and Abrate et al. emphasized the uniqueness of pricing shared accommodation and the importance of effective price adjustments, revealing patterns in price variability and adjustment strategies among professional landlords during different circumstances, such as the COVID-19 pandemic [14,28].

The hedonic price model is a prevalent method for analyzing pricing in shared accommodations and hotels, suggesting prices are influenced by the utility and satisfaction derived from various product attributes [29]. Originally used in the automobile and real estate sectors, it is now extensively applied to study the correlation between product features and pricing in hospitality [30,31]. Research, such as the studies conducted by Deboosere et al. and Wang et al., utilizing this model has demonstrated the significant impacts of location, neighborhood characteristics, and other attributes on listing prices in shared accommodations, with factors like “Superhost” designation and reviews having a negative correlation with listing prices [32,33].

Table 1. An overview of existing research on shared accommodation.

Author	Price-Influencing Factors	Sample	Results
Ikkala et al. [21]	Landlord reputation	11 Airbnb landlords with different levels of experience	Landlord reputation has a positive impact on listing prices
Wang et al. [22]	Landlord and accommodation characteristics, property amenities, rent rules, and reviews	180,533 listings across 33 cities on Airbnb	All five factors have significant impacts on listing prices
Benítez-Aurioles [25]	Flexible cancellation policy	497,509 listings across 44 cities globally	Flexible cancellation policies have a negative impact on listing prices
Kakar et al. [26]	Landlord ethnicity	2772 Airbnb listings in San Francisco	Landlord ethnicity has a significant impact on listing prices
Gibbs et al. [14]	Landlord and accommodation characteristics	15,716 Airbnb listings in Canada	Location and pictures have a positive impact on price while star ratings and review count have a negative impact
Abrate et al. [28]	Landlord professionalism	1.2 million observation of Airbnb data from Milan and Rome	The average intensity of price variability tends to increase with the degree of professionalization (number of listings)
Boto-García [29]	Landlord professionalism	24,000 Airbnb listings in Barcelona	Professional landlords show higher intertemporal price discrimination
Deboosere et al. [32]	Time variable, structural variable, host variable, and location and neighborhood variables	386,153 Airbnb listings in NYC	Location and seasonality have a significant impact on the price of Airbnb listings
Zhao et al. [34]	Functional attributes, locational attributes, reputational attributes, and host status attributes	1499 Airbnb listings in Beijing, China	Number of bedrooms, ratings, and transportation convenience have positive influences on listing prices, while reviews and the “Superhost” badge are negatively related to listing prices

provides a summary of the key findings from the existing literature on the pricing of shared accommodation. It is worth noting that there are instances of divergent and conflicting results among different studies.

Despite the growing popularity of shared accommodation, research on price adjustments during peak seasons remains limited. Existing studies mainly focused on differentiating between price adjustments on weekdays and non-weekdays, while largely overlooking the unique cyclical dimensions of the “festival” and “weekend” effects during peak seasons. Additionally, many previous studies have been limited by their utilization of single-source samples that do not fully account for the influence of geographical characteristics, emphasizing the need for comprehensive data when analyzing price adjustments during peak seasons.

More importantly, a notable research gap exists in the examination of the role of platform-certified signals and user-generated signals, both of which could significantly influence price adjustments during peak seasons. To the best of our knowledge, no earlier study has addressed this specific research gap by examining the interplay between platform-certified signals and user-generated signals, and how they jointly contribute to price adjustment in the shared accommodation market. By examining these two types of signals and their combined effects on pricing, this study contributes to a more comprehensive understanding of the factors driving price adjustments during peak seasons and the decision-making processes behind landlord price-adjusting behaviors.

3. Theoretical Framework, Conceptual Model, and Hypotheses

3.1. Signaling Theory

Information asymmetry is a common issue in business transactions, with the seller being the information superior and the buyer being the information inferior. The signaling theory has been proposed as a theoretical framework to address this issue, with the goal of reducing information asymmetry through information screening and signaling [35]. The information screening and signaling models describes processes where either the buyer or the seller acts first to transmit true information about the value or quality of a product through observable signals, reducing asymmetry [36,37]. In shared accommodation, guests often lack comprehensive knowledge about properties, relying on limited information provided by landlords or platforms, making the signaling theory highly applicable.

Several studies have applied the signaling theory to shared accommodation, exploring the impact of various signals, like “Superhost” status, identity verification, price, host reputation on consumer decisions, and perceived trust [38,39,40,41,42]. These studies reveal that such signals can significantly influence consumer expectations, purchase decisions, and landlords’ pricing decisions.

This study utilizes the signaling theory as a framework to operationalize the variables “favorable rate” (ratio of positive comments) and “Preferred Houses” (platform-preferred/recommended houses) as indicators of the “user-generated signal” and “platform-certified signal”, respectively. The choice of these signals is based on their prominence and empirical relevance. The “Preferred House” status, awarded to listings that consistently meet high guest satisfaction standards, is a significant indicator of platform endorsement and heavily influences consumer decisions. Similarly, the “favorable rate” reflects the percentage of positive reviews, serving as a crucial metric for prospective renters assessing listing quality. The literature on digital marketplaces frequently highlights the role of such certifications and ratings in influencing consumer trust and transactional decisions [16,17]. According to the signaling theory, effective signals are both observable and credible [37], which are criteria met by both “Preferred House” and “favorable rate” badges. These signals were selected for their demonstrated influence on accommodation pricing, particularly during peak-demand seasons. Additionally, this study examines the role of landlord professionalism, measured by the number of properties managed, as a moderating variable to assess how different levels of landlord experience might influence the impact of platform-certified and user-generated signals on pricing strategies. The inclusion of “landlord professionalism” is based on findings from prior research that indicated that more experienced and resource-rich landlords often adopt more sophisticated pricing strategies, particularly in how they leverage signals to optimize revenue [43]. It is hypothesized that landlord professionalism moderates the effects of these signals because more experienced landlords might use these indicators differently due to their greater experience and resources. Theoretical frameworks from both the signaling theory and resource-based views support this hypothesis, suggesting that more experienced service providers are better positioned to utilize reputation mechanisms effectively to distinguish themselves in competitive markets [37,44].

3.2. Conceptual Model

This study’s conceptual model, illustrated in Figure 1, distinguishes two peak seasons in shared accommodation: “festival” and “weekend”. This differentiation is crucial as it allows us to explore how different types of peak demands influence landlord behavior and pricing strategies in the shared accommodation sector. Employing the signaling theory, the model investigates the impact of both platform-certified and user-generated signals on landlords’ price adjustments during these seasons. Additionally, the model considers landlord professionalism, measured by the number of properties owned, as a moderating variable. Landlord professionalism is a critical factor due to the non-standardized services and variability in property conditions typical of shared accommodations. Landlord professionalism can significantly influence how uncertainties about properties are perceived and managed. Professional landlords can mitigate informational asymmetries through transparency, communication, and adherence to best practices, providing a sense of reliability and setting service quality benchmarks in an environment often marked by uncertainties and potential misconceptions about property states and amenities.

The relevance of platform-certified signals in our model is supported by the study conducted by Zervas et al., which demonstrates how Airbnb’s certification mechanisms, such as the “Superhost” status, influence pricing flexibility and consumer trust, particularly during significant events or peak seasons. Their analysis suggests that certifications can lead to optimized revenue strategies by landlords during high-demand periods, such as festivals, aligning with our focus on the festival peak season. Additionally, the impact of user-generated signals is underscored by research that found positive reviews and ratings to significantly affect landlords’ pricing decisions [3]. A study by Ert et al. specifically addresses the role of personal photos and trustworthiness perceptions in Airbnb, indicating that hosts perceived as more trustworthy can command higher prices and enjoy higher booking probabilities. This supports our discussion on the influence of user-generated content in shaping economic behaviors in shared accommodation platforms [16].

3.3. Research Hypothesis

The information asymmetry in shared accommodation platforms, where consumers often lack adequate information to assess property listings, poses significant challenges due to the private and unverified nature of properties [10,12]. This asymmetry can result in trust issues and deter bookings [16,39,45]. To mitigate this, platforms use certification mechanisms as signals of authenticity and quality, reducing consumer uncertainty and addressing information asymmetry issues [17,46].

Enhanced trust in listings, established through platform certifications, has been shown to increase consumers’ willingness to pay and grant landlords greater pricing power, especially during peak seasons [47,48]. This comparative advantage leads to increased bookings and pricing flexibility for endorsed landlords [46,49]. Essentially, the value of information conveyed through platform-certified signals, serving as indicators of trust and quality, impacts the pricing decisions in the shared accommodation market.

For instance, research by Zhao et al. underscores how certifications enhance consumer trust, which in turn boosts purchase intentions, illustrating the direct impact of platform certifications on consumer behavior and landlord pricing strategies [50]. Similarly, Fleischer and Ert highlight how trust indicators like “Superhost” status not only enhance Airbnb’s competitiveness against traditional hotels but also lead to increased profitability by enabling higher pricing flexibility [51]. Moreover, Ert and Fleischer have noted that while the “Superhost” program commands a price premium, this is not solely sufficient to compensate for an established reputation, indicating the complex effects of certifications on pricing [47]. Building on these empirical insights, we propose the following hypotheses:

Hypothesis 1a (H1a).

The platform-certified signal has a positive effect on the probability of landlords raising prices during the festival peak season.

Hypothesis 1b (H1b).

The platform-certified signal has a positive effect on the percentage of price adjustments during the festival peak season.

Hypothesis 1c (H1c).

The platform-certified signal has a positive effect on the probability of landlords raising prices during the weekend peak season.

Hypothesis 1d (H1d).

The platform-certified signal has a positive effect on the percentage of price adjustments during the weekend peak season.

Consumers on platforms like Airbnb.com (accessed on 2 March 2022) and 58.com (accessed on 2 March 2022) effectively mitigate information asymmetry by utilizing reviews, or electronic word-of-mouth, left by other users, which include both objective details and subjective sentiments about the listings [52]. Such user-generated signals, like the “favorable rate” on 58.com, have been shown to positively impact purchase intentions and willingness to pay, with consumers often willing to pay a premium for listings with positive assurance signals [17,53,54,55]. The impact of these signals is particularly pronounced during peak seasons, influencing the likelihood and extent of landlords’ price adjustments. Supporting this, a study by Daugherty et al. illustrated that consumer attitudes towards user-generated content can directly influence booking decisions, especially during peak travel times [56]. Similarly, Goh et al. found that user-generated content significantly impacts consumer purchasing behaviors, reinforcing the critical role of these signals in enhancing landlords’ pricing power during high-demand periods [57]. These insights corroborate the notion that user-generated signals not only affect consumer perceptions and willingness to pay but also directly impact the pricing strategies of accommodation providers during peak seasons. Therefore, based on these premises, we propose the following hypotheses:

Hypothesis 2a (H2a).

The user-generated signal has a positive effect on the probability of price increases by landlords during the festival peak season.

Hypothesis 2b (H2b).

The user-generated signal has a positive effect on the percentage of price adjustments made by landlords during the festival peak season.

Hypothesis 2c (H2c).

The user-generated signal has a positive effect on the probability of price increases by landlords during the weekend peak season.

Hypothesis 2d (H2d).

The user-generated signal has a positive effect on the percentage of price adjustments made by landlords during the weekend peak season.

Having a larger property portfolio is often linked to higher landlord professionalism, with such landlords having extensive rental experience and market insight, allowing them to optimize pricing during peak seasons [28,29]. Professional landlords are typically more responsive to booking surges during these times, adjusting prices more actively compared to their less-experienced counterparts [14,18]. This propensity to actively manage pricing is supported by recent studies, such as those by Xiao and Zhai, who found that professional landlords react positively to incentive structures on shared accommodation platforms, using these tools strategically to maximize revenue during peak booking periods [58]. Similarly, research by Jia and Wang demonstrates that professional hosts employ more sophisticated pricing strategies than nonprofessional hosts, effectively optimizing their responses to fluctuating market demands [59]. These findings suggest that landlord professionalism not only enhances the ability to adjust to market conditions but also positively moderates the impacts of both platform-certified and user-generated signals on the probability of pricing adjustments. Enhanced professionalism allows landlords to leverage these signals more effectively, thus improving their strategic positioning during high-demand seasons. Based on these premises, we propose the following hypotheses:

Hypothesis 3a (H3a).

The professionalism of the landlord positively moderates the effect of the platform-certified signal on the probability of price adjustment during the festival peak season.

Hypothesis 3b (H3b).

The professionalism of the landlord positively moderates the effect of the platform-certified signal on the probability of price adjustment during the weekend peak season.

Hypothesis 3c (H3c).

The professionalism of the landlord positively moderates the effect of the user-generated signal on the probability of price adjustment during the festival peak season.

Hypothesis 3d (H3d).

The professionalism of the landlord positively moderates the effect of the user-generated signal on the probability of price adjustment during the weekend peak season.

Previous research also suggests that professional landlords, who typically manage multiple properties, often set higher single-day (baseline) prices due to perceived value and reputation, which might limit their flexibility to further increase prices during peak seasons [28,29]. Tsai discusses the concept of price rigidity in housing markets, highlighting how landlords opt for higher baseline prices under fluctuating economic conditions, which subsequently limits their ability to increase prices during peak seasons [60]. Yang further discusses how the design of rental contracts can reflect the bargaining power and market strategy of professional landlords [61]. These landlords might impose higher initial rents to maximize returns, which also illustrates their limited flexibility in adjusting prices further. This approach not only maximizes returns but also indicates a strategic limitation in their ability to increase rent prices significantly during peak demand. Given these dynamics, professional landlords tend to experience a proportionally smaller percentage change in price adjustments during peak seasons compared to their less-experienced counterparts. This observation forms the basis for our next hypotheses, suggesting that while professional landlords are more likely to adjust prices in response to peak-season demands, the scope of their price adjustments may be limited by their higher baseline pricing strategies. Based on this understanding, we formulate the following hypotheses:

Hypothesis 4a (H4a).

The professionalism of the landlord negatively moderates the effect of the platform-certified signal on the percentage of price adjustments during the festival peak season.

Hypothesis 4b (H4b).

The professionalism of the landlord negatively moderates the effect of the platform-certified signal on the percentage of price adjustments during the weekend peak season.

Hypothesis 4c (H4c).

The professionalism of the landlord negatively moderates the effect of the user-generated signal on the percentage of price adjustment during the festival peak season.

Hypothesis 4d (H4d).

The professionalism of the landlord negatively moderates the effect of the user-generated signal on the percentage of price adjustments during the weekend peak season.

4. Data Collection and Research Methodology

4.1. Variables

The dependent variable in this study comprises the price adjustments of shared accommodation during different peak seasons, which are divided into four variables: (1) festival price adjustment option—the choice of landlords to adjust their prices or not during the festival peak season; (2) percentage of festival price adjustment—the percentage increase in the shared accommodation price during the festival peak season; (3) weekend price adjustment option—the choice of landlords to adjust their prices or not during the weekend peak season; and (4) weekend price adjustment percentage—the percentage increase in the shared accommodation price during the weekend peak season.

To investigate the impact of various signals on price adjustments during peak seasons, this study considers platform-certified signals and user-generated signals as the independent variables. The operational variables for these independent variables are “Preferred Houses” and “favorable rate” (with regard to landlords), respectively, and the number of properties owned by landlords is used as the operational variable for landlords’ professionalism.

To ensure the validity of the findings, 18 control variables are selected from 4 groups: accommodation characteristics, landlord characteristics, rent rules, and external characteristics. All variables are described in detail in

Table 2. Variable definitions and descriptions.

Type	Variable	Definition
Dependent Variables	Price adjustments
	PriceFestAdj	=1 if prices are adjusted during festivals
	PctFestAdj	Percentage of festival price adjustment
	PriceWkndAdj	=1 if prices are adjusted during weekends
	PctWkndAdj	Percentage of weekend price adjustment
Independent Variables	Platform-certified signal
	Preferred	Platform preferred/recommended houses
	User-generated signal
	FavorRate	100 × the ratio of positive comments (“positive comments” are defined as those reviews where guests have given a listing a rating that meets or exceeds a certain threshold indicative of satisfaction. Specifically, our dataset categorizes reviews with ratings of 4 or 5 stars on a 5-star scale as “positive”.) received by landlords
Moderating Variable	Landlord professionalism
	LordProf	Number of properties listed (owned) by the landlord
Interaction Terms	Platform-certified signal interaction term
	Preferred × LordProf
	User-generated signal interaction term
	FavorRate × LordProf
Control Variables	Accommodation characteristics
	Area	House area (size)
	DailyPrice	Single-day price for the rental home
	Headline	Length for headline of the listing (in words)
	PicNum	Number of pictures provided by the landlord
	Describe	Descriptive text length (in words)
	Landlord characteristics
	SatisNum	Number of satisfactory ratings received by the landlord
	GoldLord	=1 if the landlord is certified as a gold landlord
	LordComment	Number of comments received by the landlord
	LordText	Comments text length (in words)
	Rent rules
	RentWay	=1 if whole-house rental; =0 if share with others
	Cancel	=1 if there is flexible cancellation
	DepositFree	=1 if there is no deposit
	Discount	=1 if there is a renewal discount
	StartNum	Minimum lease days
	External characteristics
	SalesAmount	Overall tourism revenue in the province
	Income	Total revenue of the shared accommodation industry in the province
	CPI	The consumer price index in the province
	HousePrice	Average sale price of pre-owned houses in the province

to facilitate the analysis and interpretation of this study’s results.

4.2. Data Collection

This study employed a self-developed web crawler to extract data from 58.com (accessed on 10 June 2022), a leading platform for shared accommodation in China, focusing on a diverse selection of 36 cities, which were chosen based on their representation of different economic regions as classified by the National Bureau of Statistics of China. The data collection spanned from 1 April to 31 May 2022, resulting in 13,982 listing data records, each containing daily price information collected over 61 days to ensure a comprehensive representation of the market conditions. These data were further enriched with publicly available information, such as the consumer price index (CPI) and the revenue of the shared accommodation industry in the respective provinces, sourced from the National Bureau of Statistics of China.

To maintain the integrity and reliability of this study, data points that were missing, abnormal, or lacked stable daily prices were removed. Stable daily prices were determined based on consistent room prices across specified regular working days, and considerations were also made for the pricing fluctuations during festivals and weekends in the peak season of accommodation. For instance, the average prices on 2 May and 3 May, corresponding to the International Labor Day holidays in China, were selected as the festival prices, and the average prices on six regular Saturdays were selected as the weekend prices.

Finally, after cleaning and supplementing, 11,795 data points were used for data analysis and model construction. The descriptive statistics of all variables used in the analysis are presented in

Table 3. Descriptive statistics of the variables of interest.

Variable	N	Mean	SD	Min	Max
PriceFestAdj	11,795	0.1	0.31	0	1
PriceWkndAdj	11,795	0.06	0.24	0	1
PctFestAdj	11,795	0.23	1.55	0	49.5
PctWkndAdj	11,795	0.02	0.11	0	3.98
Preferred	11,795	0.06	0.23	0	1
FavorRate	11,795	35	47	0	100
LordProf	11,795	5.85	9.63	0	63
Area	11,795	56.58	77.35	1	4917
DailyPrice	11,795	175.88	413.04	10	15,000
Headline	11,795	14.17	7.72	0	50
PicNum	11,795	11.63	7.99	1	72
Describe	11,795	174.41	195.08	1	2079
SatisNum	11,795	27.9	83.3	0	999
GoldLord	11,795	0.08	0.28	0	1
LordComment	11,795	3.91	17.77	0	464
LordText	11,795	195.83	618.77	0	5370
RentWay	11,795	1.54	0.66	0	2
Cancel	11,795	0.78	0.42	0	1
DepositFree	11,795	0.46	0.5	0	1
Discount	11,795	0.28	0.45	0	1
StartNum	11,795	4.83	9	1	31
SalesAmount	11,795	131.99	109	6.5	427.1
Income	11,795	70.78	57.22	3.6	235.9
CPI	11,795	102.46	0.45	101.5	103.6
HousePrice	11,795	21,348.57	17,119.47	2939	1.20 × 105

.

4.3. Research Model

The process of price adjustments made by landlords comprises two crucial decisions: firstly, determining whether to adjust prices during peak seasons, and secondly, deciding on the magnitude of the price adjustment. To precisely model the price adjustments made by landlords and evaluate the effect of explanatory variables, this study employed two distinct modeling techniques to analyze both the probability and magnitude of price adjustments. Specifically, we first constructed two separate Logit Binary Choice models for landlords’ choice of price adjustment during “non-weekend festivals” (i.e., festival peak seasons) and “non-festival weekends” (i.e., weekend peak seasons), respectively. The Logit model is a commonly used statistical method for binary classification tasks, such as predicting whether an event will or will not occur. In this study, the Logit model was used to examine the probability of price adjustment by landlords during festival and weekend peak seasons. By estimating the probability of price adjustments, this study gained a better understanding of the likelihood of landlords adjusting their prices in response to market conditions. Following the construction of the Logit model, two Ordinary Least Squares regression models were developed to estimate the percentage change in the landlords’ price adjustments during festival and weekend peak seasons, respectively. The OLS model was employed to determine the extent to which landlords adjusted their prices. By analyzing the relative magnitude of price adjustments, this study enhanced its understanding of the pricing strategies adopted by landlords during peak seasons.

In the Logit Binary Choice model, let $\mathrm{F}(\mathrm{x},\mathsf{\beta })$ denote the cumulative distribution function of the logical distribution, which is given by

$$\mathrm{P}\left(\mathrm{y}=1|\mathrm{x}\right)=\mathrm{F}\left(\mathrm{x},\mathsf{\beta }\right)=\frac{\mathrm{e}\mathrm{x}\mathrm{p}\left({\mathrm{x}}^{\prime }\mathsf{\beta }\right)}{1+\mathrm{e}\mathrm{x}\mathrm{p}\left({\mathrm{x}}^{\prime }\mathsf{\beta }\right)}$$

If the probability of the landlord adjusting the price during peak season is expressed as $\mathrm{p}\equiv \mathrm{P}\left(\mathrm{y}=1|\mathrm{x}\right)=\frac{\mathrm{e}\mathrm{x}\mathrm{p}\left({\mathrm{x}}^{\prime }\mathsf{\beta }\right)}{1+\mathrm{e}\mathrm{x}\mathrm{p}\left({\mathrm{x}}^{\prime }\mathsf{\beta }\right)}$, then the probability of the landlord not adjusting the price during peak season is $1-\mathrm{p}=\mathrm{P}\left(\mathrm{y}=0|\mathrm{x}\right)=\frac{1}{1+\mathrm{e}\mathrm{x}\mathrm{p}\left({\mathrm{x}}^{\prime }\mathsf{\beta }\right)}$. Therefore, the odds ratio of the landlord choosing to increase the price versus not increasing the price can be calculated as $\frac{p}{1-p}=exp\left(x^{\prime }\beta \right)$, where p represents the probability of the landlord choosing to adjust the price in the peak season. Next, take the logarithm of both sides, as follows:

$$\mathrm{L}\mathrm{n}\left(\frac{\mathrm{p}}{1-\mathrm{p}}\right)={\mathrm{X}}_{\mathrm{i}}{\mathsf{\beta }}_{\mathrm{i}}$$

When $x_i$ is a continuous variable, the coefficient ${\beta }_i$ represents the percentage change in the odds ratio $\left(\frac{\mathrm{p}}{1-\mathrm{p}}\right)$ for a one-unit increase in $x_i$. When $x_i$ is a discrete variable, ${exp(\beta }_i)$ represents the fold increase in the odds ratio $\left(\frac{\mathrm{p}}{1-\mathrm{p}}\right)$ for a one-unit increase in $x_i$.

The two Logit Binary Choice models for landlords’ decisions on price adjustment are as follows:

$$\mathrm{P}\mathrm{r}\mathrm{i}\mathrm{c}\mathrm{e}\mathrm{F}\mathrm{e}\mathrm{s}\mathrm{t}\mathrm{A}\mathrm{d}\mathrm{j}={\mathsf{\beta }}_0+{{\mathsf{\beta }}_1\mathrm{P}\mathrm{r}\mathrm{e}\mathrm{f}\mathrm{e}\mathrm{r}\mathrm{r}\mathrm{e}\mathrm{d}+{\mathsf{\beta }}_2\mathrm{F}\mathrm{a}\mathrm{v}\mathrm{o}\mathrm{r}\mathrm{R}\mathrm{a}\mathrm{t}\mathrm{e}+{\mathsf{\beta }}_3\mathrm{P}\mathrm{r}\mathrm{e}\mathrm{f}\mathrm{e}\mathrm{r}\mathrm{r}\mathrm{e}\mathrm{d}\ast \mathrm{L}\mathrm{o}\mathrm{r}\mathrm{d}\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{f}+{\mathsf{\beta }}_4\mathrm{F}\mathrm{a}\mathrm{v}\mathrm{o}\mathrm{r}\mathrm{R}\mathrm{a}\mathrm{t}\mathrm{e}\ast \mathrm{L}\mathrm{o}\mathrm{r}\mathrm{d}\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{f}+\mathsf{\delta }\mathrm{Z}}_{\mathrm{i}}+{\mathsf{\epsilon }}_{\mathrm{i}}$$

(1)

$$\mathrm{P}\mathrm{r}\mathrm{i}\mathrm{c}\mathrm{e}\mathrm{W}\mathrm{k}\mathrm{n}\mathrm{d}\mathrm{A}\mathrm{d}\mathrm{j}={\mathsf{\beta }}_0+{{\mathsf{\beta }}_1\mathrm{P}\mathrm{r}\mathrm{e}\mathrm{f}\mathrm{e}\mathrm{r}\mathrm{r}\mathrm{e}\mathrm{d}+{\mathsf{\beta }}_2\mathrm{F}\mathrm{a}\mathrm{v}\mathrm{o}\mathrm{r}\mathrm{R}\mathrm{a}\mathrm{t}\mathrm{e}+{\mathsf{\beta }}_3\mathrm{P}\mathrm{r}\mathrm{e}\mathrm{f}\mathrm{e}\mathrm{r}\mathrm{r}\mathrm{e}\mathrm{d}\ast \mathrm{L}\mathrm{o}\mathrm{r}\mathrm{d}\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{f}+{\mathsf{\beta }}_4\mathrm{F}\mathrm{a}\mathrm{v}\mathrm{o}\mathrm{r}\mathrm{R}\mathrm{a}\mathrm{t}\mathrm{e}\ast \mathrm{L}\mathrm{o}\mathrm{r}\mathrm{d}\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{f}+\mathsf{\delta }\mathrm{Z}}_{\mathrm{i}}+{\mathsf{\epsilon }}_{\mathrm{i}}$$

(2)

Here, ${\mathrm{Z}}_{\mathrm{i}}$ is a set of control variables, including LordProf, Area, DailyPrice, Headline, PicNum, Describe, SatisNum, GoldLord, LordComment, LordText, RentWay, Cancel, DepositFree, Discount, SatisNum, SalesAmount, Income, CPI, and HousePrice.

A set of Ordinary Least Squares (OLS) linear regression models are then constructed to estimate the percentage change in price adjustment:

$$\mathrm{P}\mathrm{c}\mathrm{t}\mathrm{F}\mathrm{e}\mathrm{s}\mathrm{t}\mathrm{A}\mathrm{d}\mathrm{j}={\mathsf{\beta }}_0+{{\mathsf{\beta }}_1\mathrm{P}\mathrm{r}\mathrm{e}\mathrm{f}\mathrm{e}\mathrm{r}\mathrm{r}\mathrm{e}\mathrm{d}+{\mathsf{\beta }}_2\mathrm{F}\mathrm{a}\mathrm{v}\mathrm{o}\mathrm{r}\mathrm{R}\mathrm{a}\mathrm{t}\mathrm{e}+{\mathsf{\beta }}_3\mathrm{P}\mathrm{r}\mathrm{e}\mathrm{f}\mathrm{e}\mathrm{r}\mathrm{r}\mathrm{e}\mathrm{d}\ast \mathrm{L}\mathrm{o}\mathrm{r}\mathrm{d}\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{f}+{\mathsf{\beta }}_4\mathrm{F}\mathrm{a}\mathrm{v}\mathrm{o}\mathrm{r}\mathrm{R}\mathrm{a}\mathrm{t}\mathrm{e}\ast \mathrm{L}\mathrm{o}\mathrm{r}\mathrm{d}\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{f}+\mathsf{\delta }\mathrm{Z}}_{\mathrm{i}}+{\mathsf{\epsilon }}_{\mathrm{i}}$$

(3)

$$\mathrm{P}\mathrm{c}\mathrm{t}\mathrm{W}\mathrm{k}\mathrm{n}\mathrm{d}\mathrm{A}\mathrm{d}\mathrm{j}={\mathsf{\beta }}_0+{{\mathsf{\beta }}_1\mathrm{P}\mathrm{r}\mathrm{e}\mathrm{f}\mathrm{e}\mathrm{r}\mathrm{r}\mathrm{e}\mathrm{d}+{\mathsf{\beta }}_2\mathrm{F}\mathrm{a}\mathrm{v}\mathrm{o}\mathrm{r}\mathrm{R}\mathrm{a}\mathrm{t}\mathrm{e}+{\mathsf{\beta }}_3\mathrm{P}\mathrm{r}\mathrm{e}\mathrm{f}\mathrm{e}\mathrm{r}\mathrm{r}\mathrm{e}\mathrm{d}\ast \mathrm{L}\mathrm{o}\mathrm{r}\mathrm{d}\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{f}+{\mathsf{\beta }}_4\mathrm{F}\mathrm{a}\mathrm{v}\mathrm{o}\mathrm{r}\mathrm{R}\mathrm{a}\mathrm{t}\mathrm{e}\ast \mathrm{L}\mathrm{o}\mathrm{r}\mathrm{d}\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{f}+\mathsf{\delta }\mathrm{Z}}_{\mathrm{i}}+{\mathsf{\epsilon }}_{\mathrm{i}}$$

(4)

By analyzing the coefficients of the independent variables, we can quantify the impact of each variable on the percentage change in price adjustment. The inclusion of a control variable (${\mathrm{Z}}_{\mathrm{i}}$) in the models helps to account for other factors that may influence price adjustment, ensuring a more accurate estimation.

5. Main Result

5.1. Impact of Signals on the Probability of Price Adjustments during Peak Seasons

We first present results on the impact of platform-certified signals and user-generated signals on the probability of price adjustments during festival and weekend peak seasons. Hypothesis testing is conducted hierarchically, employing four models, as outlined in

Table 4. Logit model results for the probability of price adjustments.

Model Dependent Variable	Logit1 PriceFestAdj (OddsRatio)	Logit2 PriceFestAdj (OddsRatio)	Logit3 PriceWkndAdj (OddsRatio)	Logit4 PriceWkndAdj (OddsRatio)
Preferred	13.749 ***	6.989 ***	28.721 ***	26.794 ***
FavorRate	2.397 ***	3.088 ***	2.737 ***	4.358 ***
LordProf	1.022 ***	1.054 **	0.983 ***	1.088 ***
Area	1.002 *	1.002 *	1	1
DailyPrice	0.999 ***	0.999 ***	0.999 **	0.999 **
Headline	1.052 ***	1.054 ***	1.014 *	1.012
PicNum	1.032 ***	1.032 ***	1.015 *	1.014 *
Describe	1.001 ***	1.001 ***	1.001 *	1.000 *
SatisNum	1.006 ***	1.006 ***	0.991 *	0.989 **
GoldLord	0.543 **	0.583 *	0.719	0.76
LordComment	0.816 ***	0.832 ***	0.952	0.967
LordText	1.002 ***	1.002 ***	1.001	1.001
RentWay	2.590 ***	2.573 ***	3.488 ***	3.505 ***
Cancel	0.552 ***	0.554 ***	0.812	0.817
DepositFree	1.283 **	1.288 **	1.174	1.151
Discount	1.519 ***	1.519 ***	1.684 ***	1.688 ***
StartNum	0.957 ***	0.957 ***	0.962 ***	0.961 ***
SalesAmount	1.011 ***	1.010 **	1.029 ***	1.029 ***
Income	0.980 **	0.981 **	0.947 ***	0.946 ***
CPI	0.656 ***	0.672 ***	1.263	1.279
HousePrice	1	1	1	1
Preferred × LordProf		1.056 ***		1.033
FavorRate × LordProf		0.944 **		0.873 ***
N	11,795	11,795	11,795	11,795
Pseudo R2	0.4742	0.4762	0.4349	0.4384

* p < 0.05, ** p < 0.01, *** p < 0.001.

. Model Logit1 estimates the main effects of key independent variables (Preferred, FavorRate) on the dependent variable PriceFestAdj (H1a and H2a). Model Logit2 adds the interaction terms Preferred × LordProf and FavorRate × LordProf to the previous model to test the moderating effects of landlord professionalism on the effects of platform-certified signals and user-generated signals (H3a and H3c). Similarly, Model Logit3 estimates the main effects of key independent variables (Preferred, FavorRate) on the dependent variable PriceWkndAdj (H1c and H2c). To test the moderating effects of landlord professionalism, interaction terms are added to construct Model Logit4 (H4b and H4d). Utilizing a hierarchical model construction approach makes it easier to interpret the main effects of independent variables and the moderating effects of landlord professionalism. This approach enhances our understanding of how these factors individually contribute to the probability of price adjustments during festival and weekend peak seasons. Odds ratios were calculated and reported for each model, and the estimated results of all four models are presented in Table 4.

Table 4 shows that, during the festival peak season, landlords are 13.749 times more likely to adjust prices on Preferred Houses (Preferred) compared to Non-preferred Houses (p < 0.001). Additionally, for each percentage increase in the ratio of positive comments received by landlords (FavorRate), the probability of landlords increasing the price during the festival peak season, compared to not doing so, increases by a factor of 2.397. These results provide support for the validity of Hypotheses H1a and H2a.

Similar to the festival peak-season results, during the weekend peak season, landlords are 28.721 times more likely to adjust prices on Preferred Houses (Preferred) compared to Non-preferred Houses (p < 0.001). For each percentage increase in the ratio of positive comments received by landlords (FavorRate), the probability of landlords increasing the price during the weekend peak season, compared to not doing so, increases by a factor of 2.737. These findings provide support for the validity of Hypotheses H1c and H2c.

Interestingly, the logit binary choice model results reveal that platform-certified signals and user-generated signals have a stronger impact on the probability of price adjustments during the weekend peak season (odds ratio = 28.721 and odds ratio= 2.737) compared to the festival peak season (odds ratio = 13.749 and odds ratio = 2.397). During the festival peak season, demand spikes are highly predictable due to scheduled events and holidays, leading landlords to set prices at a premium from the start. The primary strategy here revolves around maximizing occupancy at these higher rates rather than increasing prices further as the event approaches. In contrast to festivals, weekends do not typically coincide with large-scale events but still attract a steady flow of travelers. The demand on weekends is consistently present but is influenced more by variable factors, like local events, weather conditions, and spontaneous travel decisions. This variability introduces a different dynamic, where landlords may start with more moderate pricing but remain highly responsive to short-term fluctuations in demand. This dynamic allows for greater price variability, enabling landlords to adjust rates more frequently and substantially based on real-time market conditions. The increased impact of signals during weekends indicates that landlords utilize these indicators more actively to navigate the less predictable and more fluctuating demands. Unlike festivals, where the demand pattern is set and pricing is largely predetermined, weekends require a more dynamic approach to pricing. This allows landlords to leverage platform-certified and user-generated signals effectively to make informed pricing decisions that respond to immediate market dynamics.

Regarding the moderating effect of landlord professionalism, results show that the effect of user-generated signals is strengthened as landlord professionalism increases (measured by the number of properties owned), during both the festival and weekend peak seasons. Therefore, H3c and H3d are supported. The moderating effects of landlord professionalism on the impact of platform-certified signals, however, are shown to be different for the two peak seasons. Specifically, the effect of platform-certified signals during the festival peak season is strengthened for landlords with a higher level of professionalism, supporting H3a. However, the moderating effect of landlord professionalism during the weekend peak season is positive but not statistically significant, thereby not supporting H3b. One possible explanation for the observed difference is probably related to the nature of demand patterns. During festival peak seasons, where demand is highly predictable and concentrated, professional landlords can utilize their experience and reputation to maximize the effectiveness of signals. This predictability allows for more strategic pricing decisions, where professionalism significantly enhances the impact of platform-certified signals. While the impact of signals is stronger overall during weekends, the role of landlord professionalism may not enhance this effect. This could be because during weekends, even though demand is variable and less predictable, it is regularly occurring, thus allowing less-professional landlords to gain experience and respond effectively. The weekends may also attract a broader mix of guests whose decisions might be less influenced by professional status but more by immediate value offerings and reviews. Therefore, while signals are critical, the added value of professional status does not significantly change the already-high responsiveness to these signals.

Figure 2 illustrates how the average marginal effects of platform-certified and user-generated signals vary with landlord professionalism. Specifically, Figure 2a illustrates how the effects of the platform-certified signal change with varying levels of landlord professionalism. Meanwhile, Figure 2b,c demonstrate similar dynamics for user-generated signals.

Regarding the control variables, the type of rental indicator (RentWay) has a significant impact on the probability of landlords’ making price adjustments. This can be attributed to guests’ preference for renting entire homes during festivals and weekends. The increased demand for whole-home rentals in these periods leads to an increase in listing prices, resulting in a greater probability for landlords to adjust prices during these peak seasons. Other variables, like LordProf, DailyPrice, Headline, PicNum, Describe, SatisNum, RentWay, Discount, StartNum, SalesAmount, and CPI, show significant effects on the dependent variable as well, with varying magnitudes.

5.2. Impact of Signals on the Percentage of Price Adjustments during Peak Seasons

Next, we present our results on the impact of platform-certified signals and user-generated signals on the percentage of price adjustments during the festival and weekend peak seasons. Following a similar approach, we hierarchically tested our hypothesis by estimating four models as outlined in

Table 5. OLS regression results for the percentage of price adjustments.

Model Dependent_Variable	Reg1 PctFestAdj (Coef.)	Reg2 PctFestAdj (Coef.)	Reg3 PctWkndAdj (Coef.)	Reg4 PctWkndAdj (Coef.)
Preferred	3.003 ***	3.121 ***	0.088 ***	0.102 ***
FavorRate	0.043 *	0.061 ***	0.014 ***	0.016 ***
LordProf	−0.001	0.007 **	−0.000 *	0.001 *
Area	0	0	−0.000 *	0
DailyPrice	−0.000 ***	−0.000 ***	0	0
Headline	−0.01	−0.01	0	0
PicNum	0.011 **	0.011 **	0	0
Describe	0	0	0	0
SatisNum	0.000 ***	0.000 ***	−0.000 ***	−0.000 ***
GoldLord	0.006	−0.003	−0.005 *	−0.006 **
LordComment	−0.001 *	−0.001 **	0.000 *	0.000 **
LordText	0	−0.000 *	0	0
RentWay	0.045 ***	0.045 ***	0.006 ***	0.006 ***
Cancel	−0.027 *	−0.025 *	−0.003	−0.003
DepositFree	0.040 **	0.038 **	0.003	0.003
Discount	0.053 ***	0.054 ***	0.002	0.002
StartNum	0.007 **	0.007 **	−0.000 *	−0.000 *
SalesAmount	−0.004 **	−0.004 ***	0.000 ***	0.000 ***
Income	0.009 ***	0.009 ***	−0.001 ***	−0.001 ***
CPI	−0.130 ***	−0.130 ***	0.003	0.003
HousePrice	−0.000 ***	−0.000 ***	−0.000 *	−0.000 *
Preferred × LordProf		−0.004		−0.001
FavorRate × LordProf		−0.007 **		−0.001 **
_cons	13.325 ***	13.266 ***	−0.262	−0.267
N	11,795	11,795	11,795	11,795
adj. R2	0.2039	0.2055	0.0578	0.0584

* p < 0.05, ** p < 0.01, *** p < 0.001.

. Model Reg1 estimates the main effects of key independent variables (Preferred, FavorRate) on the following dependent variable: percentage of festival price adjustment—PctFestAdj (H1b and H2b). Model Reg2 adds the interaction terms Preferred × LordProf and FavorRate × LordProf to the previous model to test the moderating effects of landlord professionalism on the effects of platform-certified signals and user-generated signals (H4a and H4c). Model Reg3 estimates the main effects of key independent variables (Preferred, FavorRate) on the following dependent variable: percentage of weekend price adjustment—PctWkndAdj (H1d and H2d). To test the moderating effects of landlord professionalism, interaction terms are added to construct Model Reg4 (H4b and H4d). The regression results are presented in Table 5.

The results of the multiple linear regression models (Reg1 and Reg3 in Table 5) demonstrate that “Preferred Houses” (Preferred) have positive and significant coefficients (3.003, p < 0.001 and 0.088, p < 0.001), indicating that landlords tend to adjust prices higher for Preferred Houses compared to Non-preferred Houses during both festival and weekend peak seasons. The ratio of positive comments received by landlords (FavorRate) also has positive and significant coefficients (0.043, p < 0.05 and 0.014, p < 0.001), suggesting that an increase in the percentage of positive comments leads to higher price adjustment percentages during both festival and weekend peak seasons. Therefore, Hypotheses H1b, H1d, H2b, and H2d are supported.

Also shown in Table 5, the interaction term (FavorRate × LordProf) between user-generated signals (FavorRate) and the professionalism of landlords (LordProf) is significant (p < 0.01), confirming the presence of a moderating effect attributed to landlord professionalism. Figure 3 provides additional insights into the impact of this moderating effect. Specifically, it demonstrates that an increase in landlord professionalism weakens the impact of user-generated signals (FavorRate) on the percentage of price adjustments during festival and weekend peak seasons. Therefore, Hypotheses H4c and H4d are supported. The negative impact of the interaction term (FavorRate × LordProf) can be attributed to the explanation made in the hypothesis development. It is reasonable to consider that landlords with a higher level of professionalism, indicated by a greater number of properties, tend to set higher daily prices, which in turn leaves limited room for additional price increases during peak seasons. Consequently, as landlord professionalism (LordProf) increases, the influence of user-generated signals (FavorRate) on the percentage of price adjustments decreases.

In contrast, the interaction term (Preferred × LordProf) between the platform-certified signal “Preferred Houses” (Preferred) and the professionalism of landlords (LordProf) is not statistically significant, indicating that the landlord professionalism measured by the number of properties owned does not have a moderating effect on the relationship between the platform-certified signals (Preferred) and the percentage of price adjustments during festival and weekend peak seasons. Therefore, Hypotheses H4a and H4b are not supported. One possible explanation for the insignificant interaction term (Preferred × LordProf) is that the influence of Preferred House status on price adjustment percentages may be independent of the number of properties a landlord owns or manages. This could be due to the perceived higher value and demand for Preferred Houses, which prompts landlords to adjust prices based on a house’s status alone rather than considering the number of properties they own or manage. Consequently, the impact of platform-certified signals (Preferred) on the percentage of price adjustments during festival and weekend peak seasons does not appear to be influenced by the level of landlord professionalism, regardless of the number of properties owned by landlords.

Table 6. Summary of hypotheses testing results.

Hypothesis	Peak Season	Type of Signal	Probability of Price Adjustment	Percentage of Price Adjustment
H1a	Festival	Platform-certified	Positive (supported)
H1b	Festival	Platform-certified		Positive (supported)
H1c	Weekend	Platform-certified	Positive (supported)
H1d	Weekend	Platform-certified		Positive (supported)
H2a	Festival	User-generated	Positive (supported)
H2b	Festival	User-generated		Positive (supported)
H2c	Weekend	User-generated	Positive (supported)
H2d	Weekend	User-generated		Positive (supported)
Moderating Effect of Landlord Professionalism
H3a	Festival	Platform-certified	Positive (supported)
H3b	Weekend	Platform-certified	Positive (not supported)
H3c	Festival	User-generated	Positive (supported)
H3d	Weekend	User-generated	Positive (supported)
H4a	Festival	Platform-certified		Negative (not supported)
H4b	Weekend	Platform-certified		Negative (not supported)
H4c	Festival	User-generated		Negative (supported)
H4d	Weekend	User-generated		Negative (supported)

below provides a summary of the generalized results of hypotheses testing conducted in this study. Each hypothesis tested is listed along with its corresponding impact on the probability and price adjustment percentage. Additionally, the following table indicates whether each hypothesis is supported by the analysis:

5.3. Theoretical and Managerial Implications

Theoretically, our findings on the influence of platform-certified signals, such as “Preferred House” designation, on rental pricing strategies during peak seasons further enrich the existing literature on digital platform dynamics. Lawani et al. provide insights into how quality assessments through user reviews impact pricing, aligning with our observations that platform certifications also play a crucial role in shaping pricing decisions [62]. Additionally, the impact of these certifications can be compared to findings by Zhao and Peng, who explored the broader effects of online reviews on user behavior, suggesting that platform-certified signals similarly guide consumer choices by providing trusted indicators of quality and reliability [63].

However, our study introduces new insights into the dynamics of these effects during different types of peak seasons. Unlike the consistent increase in price responsiveness to user-generated signals found by Zhu et al., our study reveals a stronger impact during weekend peak seasons compared to festival periods [64]. This difference may be attributed to the more predictable nature of demand during festivals, which allows landlords to set optimal prices in advance, as suggested by our data. This observation is supported by the work of Rao, who noted that the predictability of event-driven demands could reduce the marginal utility of real-time data from reviews [65]. Moreover, our analysis of the moderating effect of landlord professionalism provides a different view that contributes to the existing literature. While previous research, such as that by Deboosere et al., has suggested that more-professional landlords are better at leveraging platform tools to optimize pricing [32], our findings indicate that this may not uniformly enhance the influence of user-generated signals. Specifically, we find that increased professionalism weakens the impact of positive comments on pricing adjustments, possibly because professional landlords, who manage multiple properties, may prioritize maintaining consistent pricing strategies over quick adjustments to fluctuating user-generated signals. This finding contrasts the conventional expectations and highlights the complex interplay between landlord experience and the operational use of digital platform signals.

Managerially, the variability in price adjustments based on platform and user-generated signals during different peak seasons suggests the need for more sophisticated dynamic pricing tools. Rental platforms and property managers should consider incorporating advanced algorithms that adjust prices in real time, reflecting changes in user feedback and market conditions. This approach could be particularly effective for properties that already enjoy high visibility and positive ratings, enabling owners to capitalize on peak demand periods by optimizing price adjustments more aggressively. Additionally, the role of landlord professionalism in leveraging platform features suggests that rental platforms could benefit from offering specialized training programs. These programs would help landlords understand and effectively utilize platform tools and data analytics. Furthermore, the distinction in demand dynamics between festival and weekend peak periods offers a strategic opportunity for targeted marketing. Platforms could use these insights to promote properties with high user ratings more aggressively during less predictable demand periods, thereby enhancing occupancy rates and maximizing revenue.

6. Endogeneity Issue Discussion and Robustness Checks

6.1. Discussion of Endogeneity Issues

In our study, the use of cross-sectional data raises concerns regarding endogeneity. First, in the context of shared accommodation, a feedback loop between consumer feedback and price adjustments may create a two-way causal relationship that warrants attention to potential endogeneity issues. To mitigate this problem, we implemented a methodological approach where the collection of our dependent variable data (i.e., price adjustments) was strategically delayed compared to the explanatory variable data (user-generated signals). Specifically, price adjustment data were collected from 1 April to 31 May 2022, while the data on user-generated signals were collected before 1 April. This temporal separation aligns with the best practices in causal inference, emphasizing the importance of timing to prevent reverse causality. Miao et al. reinforce this approach in their study on personalized pricing with instrumental variables, demonstrating that separating the timing of early indicators from outcomes can effectively mitigate endogeneity issues [66]. Their findings support our method, suggesting that such a timing strategy is critical in ensuring the directionality of causality from user-generated signals to later price adjustments.

Second, landlords may strategically manipulate their prices to acquire the platform’s certification of a “Preferred House”. Consequently, a two-way causality issue may exist. To address this concern, we adopted the instrumental variable approach. Specifically, we selected two instrumental variables: “New House”, which refers to a recently listed property with a lack of ratings and comments, and “RealPic”, which refers to house listings with photos that have been verified by the platform. While the lack of ratings and comments may affect the property becoming a “Preferred House”, it should not directly affect a landlord’s decision to adjust prices during peak seasons. Similarly, “Real Picture” verification may affect the property becoming a “Preferred House”, but it should not directly impact the pricing and price adjustments during peak seasons. We selected “New House” and “RealPic” as instruments based on their relevance and exogeneity, which are principles underlined in the study by Burgess et al., which provides a comprehensive review of instrumental variable estimators for continuous treatments, like ours [67].

Regression analysis was conducted using the two-stage least squares method (2SLS), and the over-identification test was performed to test the exogeneity of the two instrumental variables, yielding a p-value of 0.2272, indicating that the variables were exogenous. The 2SLS estimation was then performed again using the common standard error, and the first-stage regression’s F statistic was 71.5092, suggesting that there was no weak instrumental variable. Finally, the heteroscedasticity-robust Durbin–Wu–Hausman (DWH) test was used to investigate whether the “Preferred House” variable was endogenous. F statistics and chi-squared statistics were found to be p = 0.1807 and p = 0.1802, respectively. As such, it can be concluded that the “Preferred House” variable is not endogenous, and there is no two-way causal relationship between the “Preferred Houses” and the price adjustments made by the landlords.

6.2. Discussion on the Robustness of Results

In the setting of the model in this study, the two logit binary choice models estimating the probability of price adjustments during the festival and weekend peak seasons differ only in terms of their dependent variables, thus forming a condition for robustness testing of each other. Similarly, the two multiple linear regression models estimating the percentage of price adjustments during the festival and weekend peak seasons only differ in the explanatory variables and can also be robustly tested against each other. The comparison of key variables’ directions and significance between the four models is consistent, indicating a robust model setting that reinforces the credibility of the research results.

7. Conclusions and Future Research

This study, through empirically analyzing data from 58.com, explores the dynamics of price adjustments in shared accommodation during peak seasons, linking findings to the signaling theory. It fills theoretical gaps by empirically verifying the role of the signaling theory in shared accommodation pricing, focusing on platform-certified and user-generated signals during “festival” and “weekend” peak seasons. Four models are constructed to estimate the probability and percentage of price adjustments during these seasons, revealing that both types of signals positively impact price adjustments, with varying effects in different peak seasons.

Landlord professionalism, indicated by the number of properties owned, is found to positively moderate the impact of user-generated signals on the probability of price adjustments, suggesting that professional landlords, with their market knowledge and experience, are more responsive to such signals to align with market expectations and maximize profits. However, increased landlord professionalism weakens the influence of user-generated signals on the percentage of price adjustments, possibly due to already-higher base prices limiting their flexibility to increase prices further.

This research offers practical insights for platform operators, landlords, consumers, and researchers, emphasizing the importance of reliable platform-certified and user-generated signals in pricing decisions and market trust. It suggests that a balanced approach to leveraging platform certifications and user feedback can optimize pricing strategies and enhance market efficiency and fairness.

While this study provides valuable insights, it is not without its limitations. For instance, this study’s scope is limited to the shared accommodation market in China, which may not be generalizable to other countries or regions with different market characteristics. Another limitation is that this study only considers the impact of two types of signals (platform certification and user-generated reviews), while there may be other signals that influence price adjustments in the shared accommodation market. Possible future research could explore the impact of other types of signals on the pricing behavior of landlords, such as location, amenities, and property size. Furthermore, it could be valuable to conduct cross-country comparisons to examine whether the findings of this study are applicable to different market contexts.