Safety as a Sustainable Trust Mechanism: The Lingering Emotional Impact of the Pandemic and Digital Safety Communication in the Restaurant Industry

Abstract

This study investigates how pandemic-induced emotional disruption has reshaped sustainable consumer behavior in the digital age, with a focus on the continued influence of safety measures in the restaurant industry. As societies transition beyond COVID-19 restrictions, health-related anxieties persist, driving consumers to prioritize hygiene and risk reduction in their decision-making. Drawing on large-scale data from TripAdvisor and OpenTable, we analyze the effects of digitally communicated safety protocols on restaurant booking behavior across major U.S. cities. Our findings reveal that safety communication remains a salient factor in consumer choice, even after the acute phase of the pandemic. This effect is particularly pronounced in lower-tier restaurants, where visible digital safety signals help build trust and compensate for weaker brand equity. Conversely, in upscale establishments, where baseline hygiene standards are presumed, the marginal benefit of safety signaling is reduced. The study also identifies enduring patterns of emotional expression and anxiety in online reviews, indicating the long-term psychological imprint of the pandemic on consumer sentiment. By situating safety communication as both a psychological reassurance mechanism and a strategic digital marketing tool, this research contributes to the emerging discourse on sustainable marketing in post-crisis contexts. The results offer theoretical and managerial insights into how businesses can integrate health assurance into long-term brand strategies, reinforcing trust and resilience in digitally mediated, post-pandemic consumption environments.

1. Introduction

The COVID-19 pandemic caused significant economic and social disruptions beyond public health, with the restaurant industry among the most severely affected [1,2]. During the pandemic, social distancing, hygiene management, and government regulations presented new challenges for the restaurant industry, leading to dramatic shifts in consumer behavior [3]. Consumers, amid fear and uncertainty, reset their everyday consumption patterns, placing heightened importance on health and safety. This marked a pivotal transformation in consumption logic: health, hygiene, and perceived safety supplanted traditional criteria such as convenience and price [4,5,6]. As consumers sought reassurance amid fear and uncertainty, visible safety measures became instrumental in restoring trust and enabling restaurants to navigate the crisis and recovery period [7,8].

Although society has entered the “new normal” post-pandemic, the lingering psychological effects of the pandemic continue to influence consumer behavior, and consumers persist in prioritizing risk aversion and psychological reassurance [9]. As a result, the restaurant industry faces the critical question of whether it must continue to maintain enhanced safety measures to ensure consumer trust and economic performance in the digital age [10]. Restaurants need to transparently communicate safety measures through digital platforms to build consumer trust and brand loyalty. For example, restaurants on platforms like TripAdvisor clearly provide information about their safety measures, allowing consumers to make informed choices based on hygiene and safety [11,12]. This trust-building is not just a response to the pandemic but has evolved into a sustainable marketing strategy in the digital environment [13].

Despite the growing recognition of pandemic-related emotional distress as a key driver of post-pandemic consumer behavior, few studies have examined how heightened risk sensitivity stemming from these psychological impacts manifests in digital consumption contexts [6,13]. Specifically, there is a lack of empirical research linking psychological aftermath-related theories to digital safety communication and long-term brand engagement. Previous studies have acknowledged the role of safety protocols as crisis response mechanisms [1,10], but they tend to overlook how pandemic-related psychological effects—such as anxiety, hypervigilance, and distrust—continue to shape digital decision-making behaviors.

Moreover, although some literature frames safety practices as part of operational hygiene [14,15], their potential contribution to long-term consumer trust, emotional reassurance, and brand sustainability remains underexplored. This study addresses this critical research gap by proposing a psychologically informed framework of digital safety communication that integrates psychological risk management and consumer trust-building into sustainable marketing strategies.

Specifically, this study aims to answer the question “Should restaurants continue to maintain enhanced safety measures to ensure consumer trust and economic performance in the digital age?” The research focuses on analyzing the sustainability of safety measures in the digital environment, exploring their impact on consumer trust and revenue growth. The study also empirically investigates how digital transparency influences consumer engagement and brand loyalty, offering insights into how the restaurant industry can design sustainable marketing strategies in the “new normal” environment.

By drawing on trauma theory [16], perceived risk theory [17], and digital transparency frameworks [11,12], this study positions safety communication not merely as a health protocol but as a strategic response to consumer anxiety and uncertainty. It uses big data analysis and digital platform data to explore how safety signals continue to influence consumer engagement and purchase decisions after the acute phase of the pandemic. Additionally, it captures behavioral tendencies inherent in online interactions, offering new insights into sustainable marketing strategies that align with digital innovation and changing consumer psychology. This research contributes to bridging the gap in the literature connecting trauma, digital consumer behavior, and sustainable value creation, providing practical implications for businesses exploring the consumer environment shaped by both digitalization and the ongoing psychological impacts of global crises.

The structure of this study is as follows. Section 2 presents the theoretical background related to safety measures and research hypotheses. Section 3 describes the research methodology, and Section 4 analyzes the research results. Finally, Section 5 discusses the practical implications of the research findings and the limitations of the study.

2. Literature Review

2.1. Consumer Behavior Changes

The COVID-19 pandemic has had psychological and social impacts on consumer behavior [18]. Many individuals experienced sustained anxiety, uncertainty, and social isolation, which significantly disrupted consumption norms. These psychological responses, including hypervigilance and avoidance behaviors, have not dissipated but instead continue to influence consumer decisions in the post-pandemic landscape [12].

Drawing on psychological stress response theory, such behaviors reflect an adaptive strategy to maintain control in uncertain environments [11]. Such heightened sensitivity has led to durable preferences for environments perceived as safe and emotionally reassuring [16]. As a result, brands that offer visible cues of hygiene, predictability, and emotional reassurance are increasingly favored.

In digitally mediated environments, these shifts manifest in increased reliance on online information for assessing physical safety risks [19]. According to digital transparency theory, clear and accessible communication of safety protocols builds consumer trust and becomes a crucial differentiator [11,12]. Digital safety signals such as mask policies, cleanliness ratings, or safety icons serve as proxy indicators of environmental control, helping reduce ambiguity and reinforce consumer confidence.

This shift in consumer psychology also aligns with perceived risk theory, which posits that consumers act to minimize potential losses when uncertainty is high [17]. During and after the pandemic, safety information has emerged as a non-negotiable element of service evaluation, influencing both selection and loyalty decisions. Trust is no longer derived from brand image alone, but from clear, credible communication of risk mitigation measures [9].

Moreover, the convergence of psychological distress and digital technology has given rise to a new kind of consumer behavior—what we term digitally mediated safety-seeking. In this behavior mode, consumers rely on platform-embedded safety information, real-time reviews, and peer feedback to assess the trustworthiness of physical venues [12]. This form of online risk validation has fundamentally altered how consumers interact with brands in high-contact industries like hospitality and dining [1].

The pandemic also accelerated the adoption of contactless consumption models such as QR code menus, mobile payment systems, and delivery platforms [19]. Initially deployed for emergency purposes, these practices have become part of the default service architecture. Simultaneously, behaviors like takeout preference, deferred dining, and reduced social proximity now persist among trauma-sensitive consumers [9].

In sum, the pandemic catalyzed a paradigm shift in consumer priorities, away from convenience and price, and toward safety, transparency, and predictability [1,6]. These evolving expectations have transformed digital safety communication into a strategic trust mechanism, not merely an informational add-on. For today’s consumers, such signals fulfill an ongoing psychological need for emotional security, making them essential in the construction of sustainable consumer–brand relationships [7].

2.2. Safety Measures in the Restaurant Industry

The restaurant industry, characterized by unavoidable physical proximity and direct service interactions, has been particularly affected by the lasting psychological impact of COVID-19 [1,14,15,20]. Safety measures such as sanitization protocols, contactless service, and visible hygiene indicators have evolved beyond regulatory compliance into strategic trust mechanisms that address not only immediate health concerns but also residual psychological unease [13,21,22].

From the perspective of perceived risk theory [17], these measures reduce uncertainty and restore a sense of control, especially among trauma-sensitive consumers. This aligns with trauma theory’s assertion that individuals exposed to crises retain heightened vigilance toward future risks [16]. Thus, consumers continue to prioritize explicit safety signals in evaluating dining options, reinforcing the ongoing relevance of these measures in consumer decision-making. Safety communication, therefore, becomes a form of psychological infrastructure, reinforcing emotional stability and enabling re-engagement with high-contact environments.

In the digital era, the influence of safety signaling is further amplified by online platforms, where consumers access real-time information, peer reviews, and visual representations of safety practices [11,12]. Review platforms, social media, and booking websites act as both repositories and amplifiers of hygiene-related information, making safety an integral part of the digital evaluation process. In this environment, the digital visibility of a restaurant’s safety practices becomes part of its brand equity and trust profile [23,24,25].

However, the effectiveness and perceived necessity of these safety communication methods vary across restaurant types, reflecting differences in consumers’ baseline trust levels and expectations. In upscale restaurants, consumers often assume a high level of hygiene based on established reputations, brand prestige, visual ambiance, and price points [26]. As a result, additional safety communication may offer limited marginal benefit in shaping consumer trust and behavior. These restaurants often benefit from implicit trust, where cleanliness and professionalism are perceived as embedded qualities of the overall service experience.

In contrast, downscale restaurants—where baseline consumer trust is typically lower and hygiene expectations may be more variable—stand to benefit more significantly from overt safety signaling [26,27]. For these establishments, digital safety communication serves a compensatory function, offsetting the absence of strong brand cues and helping establish a foundational level of credibility. Importantly, this differentiation does not imply that trauma-induced vigilance is exclusive to one segment; rather, it suggests that consumers in different segments deploy their psychological sensitivity in different ways. Consumers with persistent psychological concerns stemming from the pandemic are likely to interpret safety signals more intensely in downscale contexts, where such information plays a larger role in shaping risk-related perceptions.

2.3. Hypothesis Development

Based on the theoretical foundations of trauma theory [16], perceived risk theory [17], and digital transparency [12], this study posits that the psychological aftereffects of the pandemic continue to shape consumer behavior, particularly in how individuals assess and respond to safety signals in restaurant settings. Specifically, persistent emotional disruptions resulting from pandemic-induced trauma have led to a reconfiguration of risk perception and risk management behaviors among consumers. These disruptions have created an enduring cognitive schema through which health-related threats are prioritized, even in the absence of acute danger.

In the post-pandemic context, this heightened vigilance leads consumers to interpret hygiene and safety cues not as optional enhancements but as essential indicators of service quality and personal well-being. The digital display of such cues provides consumers with psychological reassurance and a perceived sense of control. These cues also serve a symbolic function, communicating a restaurant’s commitment to health and responsibility, which strengthens the emotional bond between the brand and the consumer.

In this study, we focus on the relationship between the persistent psychological effects of the pandemic and the influence of safety measures on consumer behavior. As emotional vigilance continues to shape consumer preferences and evaluations, visible safety practices remain integral to consumption choices in high-contact environments like restaurants. The display of safety-related information in online platforms reduces perceived health risk, enhances consumer confidence, and encourages engagement, ultimately translating into increased sales. Thus, we propose the following hypothesis:

Hypothesis 1.

The long-lasting psychological effects of the pandemic positively influence the impact of safety measures on sales in the restaurant industry.

While this consumer sensitivity to safety information is widespread, its behavioral expression varies depending on the environmental context—particularly the type of restaurant. In upscale restaurants, consumers tend to rely on ambient and reputational cues that implicitly signal cleanliness and professionalism [26,28]. As such, the psychological reassurance derived from explicit safety communication may be less critical. Conversely, in downscale restaurants, where such cues are less prevalent or less trusted, consumers are more likely to depend on digital safety communication as a primary means of assessing risk and forming trust [27]. Therefore, the same psychological mechanism—persistent trauma-induced vigilance—interacts differently with contextual factors, leading to heterogeneity in the impact of safety communication.

We argue that pandemic-induced psychological vigilance interacts with these contextual cues to affect the impact of digital safety communication on consumer behavior. In environments where consumers already assume a high standard of hygiene (e.g., upscale restaurants), the marginal utility of explicit safety communication is lower. In contrast, where such assumptions are absent or weaker (e.g., downscale restaurants), consumers place greater emphasis on these signals to reduce perceived risk and gain confidence.

Therefore, the following hypothesis is proposed:

Hypothesis 2.

The long-lasting psychological effects of the pandemic enhance the effectiveness of safety measures more strongly in downscale restaurants than in upscale restaurants.

Together, these hypotheses build a psychologically informed, digitally contextualized understanding of post-pandemic consumer behavior. They also contribute to the sustainability discourse by framing safety communication as a strategic mechanism for maintaining trust and resilience in consumer–brand relationships over time.

3. Methods

3.1. Data Collection

Based on the stylized facts outlined above, it is evident that COVID-19 continues to have an ongoing impact on our daily lives. In the subsequent section, we delve into an investigation to determine whether safety measures implemented by restaurants are still effective in boosting restaurant sales. To achieve our objective, we collected data from two primary sources: TripAdvisor.com and OpenTable.com. TripAdvisor.com provides users with information regarding whether the restaurants listed on its platform have adopted safety protocols in response to the ongoing pandemic, even as conditions return to a state of normalcy. On the other hand, OpenTable.com offers users daily booking statistics for the restaurants featured on its platform, which serves as a reliable proxy for assessing sales activity. In more specific terms, we aggregated restaurant-specific information from TripAdvisor.com to assess the continuous implementation of safety measures across 12 cities (Austin, Dallas, Houston, Jacksonville, Los Angeles, Miami, Orlando, San Antonio, San Diego, San Francisco, San Jose, and Tampa) within the U.S. Building upon this dataset, we consistently retrieved daily booking figures for these restaurants from OpenTable.com. This comprehensive data collection effort spanned one month, precisely from mid-April to mid-May (second quarter) in the year 2023. For the purpose of data collection, we employed Visual Basic.NET, a programming language developed by Microsoft that is built on the NET Framework.

3.2. Stylized Facts of Ongoing COVID-19

Before delving into the empirical section to examine the economic value of safety measures in the new normal, we initially assess whether there continues to be an enduring impact of COVID-19 on our daily lives. While many governments have lifted anti-COVID-19 measures since 2022 (Kantis et al., 2023 [29])), the lingering psychological effects of the pandemic may continue to impact people’s daily lives. The effects of such emotional disruptions may be discernible in online reviews where individuals express their opinions anonymously. It is believed that these reviews provide a convenient avenue to capture the ongoing influence of these psychological responses. Therefore, we gathered online reviews for restaurants in London from TripAdvisor.com. The sample includes online reviews from 2013 to the first quarter of 2023. This dataset was selected due to its longitudinal structure and high volume of user-generated text.

Considering the enduring impact of COVID-19 on online reviews related to restaurants, people may have become more sensitive to restaurant services than before the pandemic. Consequently, this heightened sensitivity could have had a negative influence on service evaluations. Additionally, increased uncertainty among individuals might have elevated anxiety levels [30].

We used Linguistic Inquiry and Word Count (LIWC) to extract two primary textual sentiment indicators: “emotional tone” and “anxiety.” These variables are widely used in marketing and hospitality research to identify patterns of affective expression in online consumer narratives [31]. Our analysis treats these measures as evidence of affective language shifts over time, reflecting how consumers express perceived risk, reassurance, or discomfort in restaurant contexts. Also, we acknowledge that anxiety-related language is often closely linked to perceived safety, hygiene, and environmental conditions. Thus, we interpret the LIWC anxiety scores in this study as contextually grounded reflections of service-related uncertainty.

Taking these factors into account, we observe the sustained impact from COVID-19 across three dimensions: review ratings, positive emotional tone, and anxiety. We used these three as the dependent variables. If the lasting influence of COVID-19 has negatively affected review ratings [32], it is plausible that it has also continued to exert a negative impact on the positive emotional tone [33]. To explore this sustained influence, we exploit a longitudinal time series regression model developed by Garthwaite [34]. Garthwaite [34] used this model and showed the impacts of celebrity endorsements on book sales by focusing on the impacts of it on book sales, reflected in discontinuity in the sales over time after the endorsement. For this, we use the following equation:

$$\begin{gathered} y_{rq}=\alpha +{\sum }_l{\beta }_l\times I\left\{l_{Quarter}^{th}Since COVID-19\right\}+{\tau }_y\times I\left\{{Year}_y\right\}+ \\ {\zeta }_m\times I\left\{Mont{h}_m\right\}+{\mu }_d\times {I\{Day of the Week}_d\}+{\delta }_r+{\epsilon }_{rq}, \end{gathered}$$

(1)

where q = quarter, r = a restaurant. Also, y_rq means the following dependent variables: review ratings, the level of positive emotional tone, and the level of anxiety. “I” is defined as an indicator variable (binary variable). It is 1 if the current quarter is lth-quarter since COVID-19; otherwise, it takes 0. These indicators depict the influences of COVID-19 on the dependent variables over the sample period since COVID-19. $I\left\{{Year}_y\right\}$ represents the year dummy variable, which plays the role of controlling year-specific fixed effects on comment generation. $I\left\{Mont{h}_m\right\}$ is the month dummy variable, which controls month-specific fixed influences on comment generation. ${\delta }_r$ reflects restaurant-level specific heterogeneity, and ε_rq shows the error term. In this analysis, we transform the information from individual reviews into a quarterly basis. For instance, we compute the quarterly average rating for restaurants by utilizing the ratings from individual reviews posted for those restaurants within a given quarter.

Table 1. Empirical results of longitudinal time series model.

Variable		Coefficient (SE)
	Quarterly Average Rating (1)	LIWC: Emotional Tone (2)	LIWC: Anxiety Language (3)
The quarter during which COVID-19 emerged (1st quarter of 2020)	−0.010 (0.024)	0.032 (0.581)	0.017 (0.010)
1st quarter since the emergence of COVID-19	−0.001 (0.069)	−0.039 (1.503)	0.017 (0.028)
2nd quarter since the emergence of COVID-19	−0.065 (0.056)	−3.695 *** (1.251)	−0.000 (0.017)
3rd quarter since the emergence of COVID-19	−0.118 ** (0.059)	−1.877 (1.279)	−0.041 *** (0.012)
4th quarter since the emergence of COVID-19	−0.080 (0.062)	−4.447 *** (1.529)	0.104 *** (0.037)
5th quarter since the emergence of COVID-19	−0.135 *** (0.042)	−3.546 *** (0.988)	0.032 ** (0.017)
6th quarter since the emergence of COVID-19	−0.152 *** (0.038)	−3.132 *** (0.869)	0.012 (0.015)
7th quarter since the emergence of COVID-19	−0.198 *** (0.035)	−3.450 *** (0.795)	0.027 ** (0.013)
8th quarter since the emergence of COVID-19	−0.145 *** (0.036)	−3.079 *** (0.834)	0.063 ** (0.024)
9th quarter since the emergence of COVID-19	−0.098 *** (0.031)	−3.221 *** (0.736)	0.040 *** (0.012)
10th quarter since the emergence of COVID-19	−0.206 *** (0.034)	−3.164 *** (0.774)	0.030 ** (0.013)
11th quarter since the emergence of COVID-19	−0.271 *** (0.034)	−4.412 *** (0.737)	0.028 ** (0.012)
12th quarter since the emergence of COVID-19	−0.164 *** (0.037)	−3.412 *** (0.838)	0.021 (0.017)
Year dummy	Included	Included	Included
Month dummy	Included	Included	Included
Restaurant fixed effect	Included	Included	Included
Constant	Included	Included	Included
R2	0.798	0.815	0.211
# of observations	57,429	57,429	57,429

Note: where parentheses include the standard errors, ** p < 0.05; *** p < 0.01.

presents the empirical results derived from the aforementioned model. The initial column illustrates the effects of COVID-19 on review ratings. The overall estimated coefficients display a negative trend since the onset of COVID-19, with significance becoming apparent in the 5th quarter following the commencement of the pandemic (spanning from the 2nd quarter of 2021 to the 1st quarter of 2023). Given that positive emotional tone influences review ratings, we posit that the estimated coefficients for each quarter are also notably negative. Table 1 reveals that the majority of estimated coefficients were significantly negative, starting from the 3rd quarter since the initiation of COVID-19, excluding the 4th quarter of 2020. Regarding anxiety, the estimated coefficients turned significantly positive from the 1st quarter of 2021 and continued to remain mostly significantly positive until the end of 2022, though the estimated coefficient of the 1st quarter of 2023 is insignificantly positive. This suggests an increase in the level of anxiety led by the uncertainty reflected in textual content since the advent of COVID-19. These estimated coefficients confirm that COVID-19 still has an impact on our lives even in the new normal period. Consequently, we investigate whether safety measures implemented by restaurants could have a positive influence on restaurant sales.

3.3. Measures

3.3.1. Independent Variable

The independent variable employed in this study focuses on safety measures, particularly whether a restaurant was implementing precautions against COVID-19, even in 2023, when most countries had lifted restrictions related to COVID-19. Safety measures implemented by restaurants are recognized for their impact on consumers’ choices between dine-in and takeaway options during the COVID-19 pandemic [1]. In this research, we investigate the economic value of safety measures in the new normal. This variable is denoted by a binary representation: 1 signifies the implementation of safety measures, while 0 indicates their absence. Summary statistics for this independent variable are presented in

Table 2. Summary statistics of the independent variables.

Variable	# of Restaurants Taking Safety Measures	# of Restaurants Not Taking Safety Measures
Restaurants	192 (1.45%)	8229 (98.55%)

. Among the total sample of 8421 restaurants, it is worth highlighting that 192 establishments (approximately 1.5% of the total) have chosen to maintain these safety measures. We designate restaurants implementing safety measures as the treatment group, while those not implementing safety measures serve as the control group.

3.3.2. Dependent Variable

The dependent variable in this study is the number of daily restaurant bookings, sourced from OpenTable.com, which serves as a proxy for sales. Economic transactions (bookings) on OpenTable.com have been utilized as a proxy for sales in previous studies [35], so this approach provides a reliable indicator of consumer demand. As indicated in

Table 3. Summary statistics of the dependent variables.

Variable	Mean	Standard Deviation	Min	Max
# of booking	27.13	32.73	1	394

, the mean number of daily bookings is approximately 27. In the empirical model specification, we applied a log transformation to the number of daily bookings to enhance the normality of the dependent variable.

Table 4. The results of the t-test.

Groups	Observations	Mean of # of Booking	t-Value	p-Value
Control	69,734	26.62	26.96	0.00
Treatment	1683	48.28

illustrates the model-free sales comparison between the control and treatment groups. The treatment group comprises restaurants that have implemented safety measures. The t-test conducted reveals a noteworthy disparity, with the treatment group exhibiting significantly higher numbers of daily bookings in comparison to the control group. This observation suggests a positive impact of these safety measures on sales, even in the context of returning to a new normal. In the empirical results section, it is crucial to ensure a thorough analysis by introducing a diverse set of control variables that might have the potential to influence the daily booking figures.

3.3.3. Control Variables

To assess the impact, as previously mentioned, we have included a set of control variables to mitigate potential confounding effects. Specifically, we have incorporated restaurant star ratings obtained from OpenTable.com. These star ratings are derived from assessments made in online reviews (online review ratings) and serve as indicators of restaurant quality [36]. These star ratings are rounded up or down based on a 0.5 benchmark. In other words, if the average rating of online reviews is 2.26, it will be displayed as 2.5, and if it is 2.24, it will be set as a 2.0-star rating. By including star ratings as a control variable, our aim is to account for the influence of customer-evaluated quality. The distribution of star ratings is presented in

Table 5. Distribution of star ratings of restaurants.

Star Ratings	Frequency	Percent
1.0	35	0.42
1.5	11	0.13
2.0	46	0.55
2.5	75	0.89
3.0	308	3.66
3.5	795	9.44
4.0	2749	32.64
4.5	2359	28.01
5.0	2043	24.26
Total	8421	100.00

. Among the total of 8421 restaurants under study, approximately 25% of these establishments have received the highest possible star rating, which is 5.0. According to Mayzlin et al. [37], when online reviews receive a rating of 5.0, those reviews are considered extremely positive. Therefore, in the empirical section, we classify restaurants with a 5-star rating as high-quality establishments.

In conjunction with star ratings, our analysis encompasses the incorporation of day of the week, city, and restaurant fixed effects. The inclusion of these factors is vital as they could potentially impact sales trends. Firstly, the day of the week might play a role in influencing sales patterns. Secondly, local economic conditions and dynamics could be intertwined with sales fluctuations. Hence, we account for these variations by introducing city fixed effects. Lastly, we acknowledge that each restaurant possesses its unique set of unobservable characteristics that may elude the researchers’ control. To address this potential source of bias, we introduce restaurant-level fixed effects. This comprehensive approach, incorporating a spectrum of control variables and fixed effects, aims to provide a robust analysis that captures the nuanced interplay of various factors on sales trends.

3.4. Model Specification

Drawing upon the aforementioned variables, we present the subsequent model specification as the foundational framework:

$$\begin{array}{cc}\mathrm{l}\mathrm{n}(\# \mathrm{o}\mathrm{f} \mathrm{b}\mathrm{o}\mathrm{o}\mathrm{k}\mathrm{i}\mathrm{n}\mathrm{g}{)}_{it}& \hfill \\ & ={\beta }_0+{\beta }_1·{Safety measures}_i+{\beta }_2·{Star ratings}_i\\ & +{\sum }_T{\tau }_t·D_t+{\sum }_L{\sigma }_l·C_l+{\sum }_I{\rho }_i·R_i+{\epsilon }_{it}\hfill \end{array}$$

(2)

where i denotes a restaurant, and t signifies a day. ${\beta }_1$ measures the impact of safety measures, ${\sum }_T{\tau }_t·D_t$ stands for the vector of day-of-week dummies, ${\sum }_L{\sigma }_l·C_l$ corresponds to the vector of city dummies, and ${\sum }_I{\rho }_i·R_i$ signifies the vector of restaurant dummies. ε_it accounts for the error term.

4. Empirical Results

Table 6. Empirical results of testing H1.

Variables	DV: ln(# of Booking)
	Model (1)	Model (2)	Model (3)
Treatment (safety measures)	0.463 *** (0.034)	0.525 *** (0.033)	1.314 *** (0.052)
Star ratings	−0.034 *** (0.008)	−0.043 *** (0.008)	0.703 *** (0.037)
Day of the week effects	Not Included	Included	Included
City fixed effects	Not Included	Included	Included
Restaurant fixed effects	Not Included	Not Included	Included
Intercept	Included	Included	Included
R-squared	0.33%	4.89%	86.74%
# of observations	71,417	71,417	71,417

Note: where parentheses include the standard errors, *** p < 0.01.

presents the empirical findings derived from Equation (2). In Model (1), the analysis exclusively incorporates star ratings as the control variable. In Model (2), day of the week and city fixed effects are introduced in conjunction with star ratings as control variables. Lastly, Model (3) extends the analysis by encompassing restaurant-level fixed effects alongside the previously mentioned control variables. We estimated the three models using the Least Squares Dummy Variable (LSDV) approach. Model (3) specifically includes fixed effects by incorporating dummy variables for each individual restaurant. This controls for unobserved, time-invariant heterogeneity across restaurants, isolating the effect of the predictors on the outcome.

Across all three models, consistent evidence emerges indicating that safety measures exert a positive influence on sales. Notably, the estimated coefficients associated with safety measures exhibit statistical significance, reinforcing the notion of their beneficial impact on sales outcomes (β_{Safety Measures} = 0.463, p-value < 0.01 in Model (1), β_{Safety Measures} = 0.525, p-value < 0.01 in Model (2), β_{Safety Measures} = 1.314, p-value < 0.01 in Model (3), respectively).

Interestingly, the estimated coefficients related to star ratings yield a significantly negative trend, suggesting a counterintuitive relationship between review ratings and sales. This intriguing observation implies that higher review ratings do not inherently translate to elevated sales figures. It underscores the complex interplay of various factors beyond mere customer evaluations in influencing sales outcomes.

In Model (3), the coefficient of determination (R-squared) is approximately 87%. This substantial value indicates that a significant proportion of the variance in the number of bookings can be elucidated by the independent and control variables incorporated within the model’s framework.

With these outcomes in perspective, a compelling conclusion can be drawn. The results robustly substantiate that safety measures wield positive effects on sales, even in the context of the “new normal” period, supporting the first hypothesis (H1).

As mentioned in the previous section, we classify restaurants with a 5-star rating as high-quality establishments, which account for approximately 25% of the total sample. In our analysis, we incorporate the interaction between high-quality restaurants and safety measures. By introducing this interaction term, we aim to examine how safety measures differentially impact sales. If the estimated coefficients of these interaction terms are significantly negative, it suggests that the positive effects of safety measures on restaurant sales diminish when the restaurants are of high quality. As indicated in

Table 7. Empirical results of testing H2.

Variables	DV: ln(# of Booking)
	Model (1)	Model (2)	Model (3)
Treatment (safety measures)	0.522 *** (0.035)	0.580 *** (0.034)	1.314 *** (0.052)
High * treatment (safety measures)	−0.722 *** (0.116)	−0.669 *** (0.122)	−0.330 *** (0.052)
Star ratings	−0.031 *** (0.008)	−0.041 *** (0.008)	0.703 *** (0.037)
Day of the week effects	Not Included	Included	Included
City fixed effects	Not Included	Included	Included
Restaurant fixed effects	Not Included	Not Included	Included
Intercept	Included	Included	Included
R-squared	0.39%	4.94%	86.74%
# of observations	71,417	71,417	71,417

Note: where parentheses include the standard errors, *** p < 0.01.

, the estimated coefficients of the interaction terms are indeed significantly negative (β_{High*Safety measures} = −0.722 p-value < 0.01 in Model (1), β_{High*Safety measures} = −0.669 p-value < 0.01 in Model (2), β_{High*Safety measures} = − 0.330 p-value < 0.01 in Model (3), respectively). This confirmation aligns with the second hypothesis (H2), providing empirical support for it. Based on Model (3), we observe that the positive impact of safety measures is 1.314, but this impact is negatively adjusted when the restaurants are of high quality (0.984 = 1.314 − 0.330). Furthermore, the R-squared value of Model (3) is 86.74%, indicating that the independent variables account for approximately 87% of the variation in restaurant sales.

5. Robustness Check

To investigate the economic value of safety measures in the restaurant industry, we used secondary data scraped from OpenTable.com and conducted an observational analysis. While this approach provides valuable empirical insights, it may be subject to endogeneity concerns due to potential confounding factors influencing both safety measure adoption and restaurant performance. To address this issue and strengthen causal inference, we applied a quasi-experimental approach commonly used in prior research [36]—Mahalanobis distance matching—to create a balanced sample of restaurants with and without safety measures.

We began with a raw dataset containing detailed information about restaurants, including restaurant ratings, review volume, price levels, cuisine types, and city locations. To ensure accurate numerical analysis, all relevant string-type variables—such as price (Price), restaurant ratings (Rating), and review count (Number of Reviews)—were converted into numeric form. Specifically, the price level was categorized into three ordinal levels, Low, Medium, and High, which were labeled accordingly. We excluded observations with missing values in any of the key matching variables (price, cuisine, or city). The cleaned dataset was saved for subsequent matching

To reduce selection bias and ensure covariate balance between treated and control groups (i.e., restaurants that did or did not take safety measures), we employed Mahalanobis distance matching. Matching was conducted on three covariates: price level, primary cuisine type, and city. The treatment variable was whether the restaurant took any safety measures. To further ensure comparability in consumer visibility and popularity, we retained only 1:1 nearest-neighbor matches based on Euclidean distance, calculated using restaurant ratings and review volume—two key indicators of consumer engagement and restaurant prominence.

Table 8. Quasi-experimental validation of H1 via Mahalanobis matching.

Variables	DV: ln(# of Booking)
	Model (1)	Model (2)	Model (3)
Treatment (safety measures)	0.100 ** (0.050)	0.099 ** (0.049)	0.851 *** (0.084)
Star ratings	0.273 *** (096)	0.391 *** (0.099)	1.446 *** (0.224)
Day of the week effects	Not Included	Included	Included
City fixed effects	Not Included	Included	Included
Restaurant fixed effects	Not Included	Not Included	Included
Intercept	Included	Included	Included
R-squared	0.54%	15.57%	90.29%
# of observations	2573	2573	2573

Note: where parentheses include the standard errors, ** p < 0.05; *** p < 0.01.

reports the results of robustness checks assessing the economic impact of safety measures on restaurant bookings. Across all three models, the treatment variable indicating whether a restaurant implemented safety protocols remains consistently positive and statistically significant. In Model (1), which includes only the treatment indicator and star ratings, the coefficient for safety measures is 0.100 (p-value < 0.05), suggesting that restaurants with visible safety protocols received approximately 10% more bookings than those without. Model (2) introduces additional controls for day of the week and city-level fixed effects, and the treatment coefficient remains stable at 0.099 (p-value < 0.05), reinforcing the robustness of the finding across different temporal and geographic conditions. Model (3) incorporates restaurant-level fixed effects to account for unobserved heterogeneity across establishments. In this fully specified model, the coefficient increases substantially to 0.851 (p-value < 0.01), indicating that within the same restaurant, the adoption of safety measures is associated with a nearly 85% increase in bookings. This result highlights the significant within-restaurant behavioral shift triggered by the presence of visible safety protocols.

Taken together, the consistent and significant effect of safety measures across all model specifications provides strong empirical support for Hypothesis 1, which posits that the long-term psychological legacy of the pandemic continues to influence consumer behavior, making safety communication a key driver of economic performance. Although not reported in detail due to space limitations, Hypothesis 2 was also tested using the same matched dataset, and the results aligned with theoretical expectations, showing a stronger treatment effect in lower-tier restaurants.

Table 9. Robustness check using random-effects panel estimation.

Variables	DV: ln(# of Booking)
	Model (1)	Model (2)
Treatment (safety measures)	0.264 *** (0.114)	0.318 *** (0.112)
Star ratings	−0.017 (0.022)	−0.030 (0.022)
Day of the week effects	Not Included	Included
City fixed effects	Not Included	Included
Restaurant fixed effects	Not Included	Not Included
Intercept	Included	Included
Overall R-squared	0.33%	4.07%
# of observations	71,417	71,417

Note: where parentheses include the standard errors, *** p < 0.01.

presents the results of a robustness check using random-effects panel regression as an alternative estimation strategy. While our main analyses rely on the fixed-effects specification to control for unobserved, time-invariant restaurant characteristics, this random-effects approach provides an additional robustness check for the consistency of our findings. Across both models, the treatment variable (safety measures) shows a statistically significant and positive association with the number of bookings. Specifically, restaurants that maintained safety precautions in the post-pandemic period experienced higher log booking counts, with coefficients of 0.264 (p-value < 0.01) in Model (1) and 0.318 (p-value < 0.01) in Model (2). These results remain robust even after controlling for day of the week effects and city-level fixed effects in Model (2). Although the overall explanatory power of the model is relatively low (R² = 0.33% in Model (1) and 4.07% in Model (2)), the consistent significance and positive direction of the treatment effect provide empirical support for Hypothesis 1. This finding suggests that digital safety signaling continues to have an economic impact in the new normal era. While not reported due to space constraints, results for Hypothesis 2, which examines heterogeneous effects across restaurant tiers, were also supported within this matched panel data context.

6. Discussion

This study investigated whether digitally communicated safety measures remain economically and psychologically relevant in the post-pandemic restaurant industry. Our findings confirm that safety communication exerts a sustained, positive influence on consumer behavior, particularly through increased booking activity. Even as formal COVID-19 restrictions have receded, consumers continue to respond favorably to health-related transparency—signaling that pandemic-driven sensitivities have become embedded in consumer decision frameworks [6,13].

These results extend prior research by showing that digital safety cues have evolved from temporary risk-mitigation tactics into strategic trust signals. For emotionally attuned consumers, such cues not only reduce uncertainty but also serve as stabilizing forces, enabling re-engagement with high-contact service environments. In this sense, safety protocols function as “psychological infrastructure” that supports sustainable consumption in times of lingering anxiety.

Additionally, this study offers a novel contribution to trauma theory by positioning digital safety communication as a response not only to physical risk, but to the psychological imprint of collective trauma. Consumers impacted by pandemic-related stress exhibit patterns of heightened vigilance and selective trust, particularly in uncertain environments. Safety measures—when clearly communicated—help restore a sense of control, reinforcing emotional stability and brand attachment. This reframes trauma-informed behavior as not only a clinical concern, but a driver of digital-era consumption dynamics.

Importantly, our findings also clarify the moderating role of the market segment. In lower-tier restaurants, where brand equity and ambient cues are limited, safety communication plays a compensatory role by reducing perceived hygiene risk and enhancing credibility. In contrast, upscale restaurants already benefit from reputational trust, limiting the marginal value of overt safety signaling [17]. This asymmetric effect reinforces the need for tailored communication strategies based on a restaurant’s brand positioning and customer profile.

In sum, this study advances the discourse on post-pandemic marketing strategy by integrating risk reduction, emotional resilience, and digital transparency into a unified framework. By highlighting how consumers navigate a world shaped by past trauma, the research provides a roadmap for firms seeking to build sustainable, trust-centered brand ecosystems.

6.1. Theoretical Implications

This research contributes to the literature in three significant ways. First, this research differs from the existing literature, which has considered safety measures primarily as recovery tools during the pandemic era. Instead, it explores the economic role of safety measures as a sustainable marketing strategy grounded in trauma within both physical and digital spaces [38,39]. We explain this as a durable marketing signal that addresses consumer needs resulting from heightened health-related sensitivities, demonstrating that safety measures function as strategic elements that extend beyond short-term crisis response. Furthermore, the study expands social science and hospitality research by showing how pandemic-induced psychological responses influence not only consumer behavior but also platform-based decision-making. These findings provide a crucial foundation for understanding consumer psychological responses in future crisis situations and designing marketing strategies tailored to those responses.

Second, building on literature related to pandemic-induced consumer anxiety, this study demonstrates that perceived risk and anxiety remain deeply embedded in consumer logic even after the pandemic. The results indicate that, even after the public health crisis subsides, consumers continue to seek reassurance through visible and digitally accessible safety cues, suggesting that digital transparency plays an ongoing and critical role in alleviating anxiety within the consumer decision architecture [17,19]. This finding expands the perceived risk literature, contributing to the development of consumer choice theory centered on risk management and psychological stability in digital environments. It also suggests that consumer tendencies to prioritize safety and trust continue to influence behaviors in the post-pandemic era.

Third, this study investigates the impact of safety measures on dining experiences through big data analysis of online platform data. The analysis shows that displaying the implementation of safety measures on online platforms is a key element in consumer decision-making processes. This suggests that restaurants can leverage digital transparency as a critical strategy for building consumer trust and increasing customer engagement in the digital environment [11,12]. This method contributes to expanding platform-based research on safety measures in psychology and customer behavior, offering valuable insights into how digital communication influences consumer decisions.

In conclusion, this research provides new insights into building consumer trust and sustainable marketing strategies in the digital age. It offers direction for how safety measures, as both psychological and economic elements, can be integrated across various industries in the future.

6.2. Practical Implications

The findings from this research provide practical insights for restaurant operators and digital marketers looking to navigate the post-pandemic consumer environment through sustainable and digitally integrated strategies. In today’s digital age, consumers are increasingly relying on online cues for pre-visit evaluations. As a result, visually communicating health and safety practices on platforms like TripAdvisor and OpenTable has become a critical tool for building trust [13]. Digital transparency reassures health conscious consumers and enhances brand credibility in the post-pandemic context [13].

However, the impact of safety messaging appears to differ depending on the market segment. Restaurants with lower brand equity or less-established reputations—typically categorized as lower-tier restaurants—experience greater benefits by actively showcasing safety measures. These signals compensate for limited baseline trust and can serve as key differentiators in the highly competitive digital marketplace [1]. On the other hand, upscale restaurants, where hygiene is already assumed to be a standard, derive relatively less marginal value from overt safety communication. In these establishments, safety assurances are more effectively integrated into a broader experiential narrative that highlights service quality, exclusivity, and ambiance, subtly reinforcing hygiene without overpowering the premium experience [17].

Additionally, the strategic display of safety information can enhance algorithmic visibility on digital booking platforms. As recommendation systems increasingly use structured metadata to improve relevance and ranking, incorporating health-related information can increase discoverability and positively influence consumer conversion rates. In this context, safety cues function not only as a psychological reassurance mechanism but also as digital signals that contribute to platform-based marketing efficiency.

These observations collectively suggest that health assurance has become a crucial element of sustainable brand positioning in the digital era. By integrating safety into both their operational practices and digital communications, restaurants are better positioned to meet evolving consumer expectations, strengthen trust, and secure long-term competitiveness in hybrid consumption environments.

6.3. Limitations and Future Research

While this study provides timely insights into how digital safety communication affects consumer sentiment and restaurant performance in the post-pandemic context, several limitations must be acknowledged. These limitations not only qualify our findings but also suggest promising avenues for future research.

First, the study is based on observational data, limiting our ability to make strong causal claims. Although we applied restaurant-level fixed effects, city and day-of-week controls, and robustness checks, potential endogeneity remains a concern. For instance, restaurants that chose to implement safety measures may have had superior management practices, stronger customer loyalty, or greater marketing resources—unobserved variables that could bias results. To partially mitigate this concern, we employed a quasi-experimental approach using Mahalanobis distance matching as a robustness check. This method allows us to compare treated and control groups with similar pre-treatment characteristics, thereby improving covariate balance and lending further credibility to our main findings. However, we acknowledge that such techniques do not fully eliminate endogeneity concerns. Future studies would benefit from integrating controlled experiments or natural experiments to more rigorously establish causality and further address potential selection bias.

Second, the small size of the treatment group (only 1.5% of restaurants retained visible safety measures during the study period) raises concerns about statistical power and external validity. Although this reflects real-world behavior in the post-pandemic recovery stage, it may limit the generalizability of our findings. Subsequent research could expand the scope geographically or temporally, or focus on settings where safety communication remains more prevalent (e.g., healthcare, aviation, or public transport) to evaluate whether similar effects hold in higher-safety-sensitivity contexts.

Third, the use of LIWC sentiment indicators provides a scalable way to analyze emotional expressions in consumer reviews, but it also introduces interpretive constraints. While our analysis of “anxiety” and “emotional tone” aligns with standard practice in hospitality and marketing research, these constructs are not clinically validated psychological diagnoses. We have revised the manuscript to reflect that these indicators are linguistic approximations of affective states, not evidence of trauma or PTSD. Future research may incorporate clinical survey instruments, biometric data, or interviews to triangulate psychological interpretations of consumer sentiment with greater precision.

Fourth, our binary treatment variable does not reflect differences in how safety measures were implemented across restaurants. Some establishments may have maintained robust hygiene and distancing policies, while others adopted symbolic or minimal measures. This limitation partly reflects data availability—at the time of collection, TripAdvisor only indicated whether a restaurant had implemented any COVID-19 safety measures, without further breakdown. Future work should pursue more granular coding of safety measures—for example, distinguishing between sanitation efforts, employee health policies, and contactless technologies—to assess which specific practices most influence consumer behavior and trust.

Fifth, while our model includes multiple control variables, it does not fully account for other external drivers of consumer behavior, such as macroeconomic trends (e.g., inflation, unemployment), changes in consumer income, or shifts in dining preferences during the recovery period. These unmeasured factors may influence both booking behavior and review sentiment. Researchers could integrate consumer financial data, regional economic indicators, or survey-based attitudinal metrics to develop a more comprehensive understanding of how contextual forces interact with digital safety communication.

Finally, the emotional tone analysis is based on London restaurant reviews, while behavioral data on bookings comes from U.S. cities. While this mixed-source strategy enriches the study by drawing on different forms of consumer expression, it introduces potential cross-cultural inconsistencies. Cultural norms, public health regulations, and communication styles differ across regions, which may influence both sentiment expression and safety expectations. Future studies should strive for regionally aligned datasets—or conduct comparative research across countries—to assess how cultural context moderates the psychological and behavioral effects of safety communication.