Influence of empathy for advertising models on purchase intention: Comparison between celebrity models and unknown models

Abstract Primarily originating from characters (models) in a story, empathy is essential for enhancing the effectiveness of marketing communications with the condition that concept-based realism is practiced. However, previous literature primarily considered celebrities when investigating the relationship between models and products in marketing communications. Remarkably, extant literature does not include examples of unknown models that focus on empathy in product usage scenes, as depicted in marketing communications, or their effects. Therefore, this study considers consumers’ observations of product usage scenes, focusing on personal computers (PCs) and smartphones (SPs) in Japan. The research question is how empathy—or consumers’ belief that the model uses the products in the same way they do in their daily lives—positively affects purchase intentions. The Japanese have a strong social culture that emphasizes empathy; therefore, it is thought that the above effects would be easy to evaluate. The effects of empathy were examined in the context of both celebrity and unknown models and the results demonstrated that empathy positively affected purchase intention in both. However, the effect was higher for celebrity models. These findings suggest the importance of determining and using a model for marketing communications, whether celebrity or not, based on a clear product concept.


Introduction
Empathy is a cognitive experience formed through an emotional understanding based on observations (Decety & Jackson, 2004). It deepens the understanding of a story's content and enhances its persuasive effect (Green & Brock, 2000). Therefore, it is an effective way to convey a message to consumers and encourage them to act. This applies not only to novels but also to marketing communications, such as websites and advertisements. Presenting a product concept and gaining empathy through concept-based communication is more likely to induce purchasing interest (Li & Liu, 2020). This is because consumers remain loyal to the brand when they understand and empathize with its concept (Kato, 2022).
Empathy originates primarily from characters (models) in a story. When consumers empathize with a story, they share the model's emotions and experience from their perspective. The factors that create empathy are consistency and realism. First, inconsistencies in the story can confuse consumers and make them less likely to accept a product. Second, the narrative world must also be understood as similar to the real world, or less fictitious (Argo et al., 2008;Busselle & Bilandzic, 2008). This condition is considered necessary for both celebrity and unknown models.
However, previous literature primarily considered celebrities when investigating the relationship between models and products in marketing communications, particularly in advertising (Chung et al., 2013;Huang, 2017;Hussain et al., 2020;Paul & Bhakar, 2018;Till, 1998). Remarkably, extant literature does not include examples of unknown models that focus on empathy in product usage scenes, as depicted in marketing communications or their effects.
Hence, this study considers consumers' observations of product usage scenes, focusing on personal computers (PCs) and smartphones (SPs) in Japan. The study focuses on Japanese individuals, who traditionally value empathy and social harmony greatly (Karremans et al., 2011;Kelley, 2008; S. S. Kim et al., 2007;Nghiêm-Phú & Shibuya, 2021;Sugino, 2023;Watanabe et al., 2021). We can therefore assume that it will be easier to evaluate the effect of empathy on the model treated in this study, due to this national characteristic. PCs and SPs were selected because they are widely used in daily life, and their marketing applications include various depictions of their use. To remove brand bias, four brands were selected using both types of models. The research question is how empathy-or consumers' belief that the model uses products in the same way they do in their daily lives-positively affects purchase intentions. The effects of empathy were examined in the context of both the celebrity and unknown models. Consistency is essential for brand management (Keller & Swaminathan, 2019), and the results of this study warn of marketing communication that disregards the product concept and only temporarily attracts the attention of consumers.

Effects of empathy
Consumers focus more on scenarios and models in marketing communications than on product features (Escalas, 2007), and empathy is created among these consumers by generating emotions toward models that attract consumers' attention (Van Laer et al., 2014). This tendency is especially observed in novels, as readers empathize with the characters' emotions and have an imaginative experience based on the perspective of the characters in the story (Coplan, 2004). Compared to novels, it is more difficult for those observing marketing communications to immerse themselves, as these types of communications are less tangible and include the company's commercial intentions. Nevertheless, empathy can occur; therefore, consumers are not only aware of the communication's innate commercial intent but also observe their scenarios and models (Eagle, 2007).
Empathy reduces the likelihood that consumers will skip advertisements (Jeon, 2018), positively impacts brand attitudes (Chang, 2009), and motivates purchasing behavior (E. A. Kim & Muralidharan, 2020). Given this background, recent marketing communications have increasingly included examples of stories to convey content that can generate empathy through their models (Li & Liu, 2020). For example, Procter & Gamble conducted "Thank You, Mom" campaigns in multiple Olympic Games, and these included communication to promote gratitude and empathy for mothers worldwide (Mann & Ghuman, 2018). Many researchers have also documented the effects of empathy, such as through television advertisements for sunscreen in which empathy from observing the risks of skin cancer increased purchase intentions (Dunlop et al., 2010). Empathy's effectiveness in advertising has also been confirmed in many other industries, such as luxury fashion (J. E. Kim et al., 2016), automobiles (Brechman & Purvis, 2015), and the travel sector (Akgün et al., 2015). As positive experiences are dominated by emotional empathy, it is important to design marketing communication to create empathy among consumers (Umasuthan et al., 2017).

Empathy for the Japanese
Like all nations, Japanese consumers have certain characteristics; for example, they have a strict evaluation of food safety (Rupprecht et al., 2020), they have nonverbal cues that are not expressed in their verbal communication (Pandey & Charoensukmongkol, 2019), and they value the emotional interaction between humans and robots, rather than the functional aspects of robots, as they are familiar with robot animation (Choi et al., 2021). Empathy and harmony, which are the main contexts of this study, are deeply rooted in Japanese social values (Kelley, 2008). The Japanese have a strong interest in maintaining social harmony and empathy (Karremans et al., 2011;Watanabe et al., 2021). For example, the Japanese name of the new era "Reiwa," which was announced in 2019, means "harmony" (Osaki & Murakami, 2019). This characteristic is also reflected in corporate activities and decision-making. Japanese companies have been promoting the diversity of human resources in their organizations, but progress has been slow. This is because they see people who are fluent in Japanese, who understand Japanese culture, and who have acquired "Japaneseness" as highly important. In other words, they seek harmony and sympathy, even from foreign talent (Sugino, 2023). In terms of media, the Japanese tend to prefer Korean dramas, and one reason is because these dramas trigger their high level of empathy for the actors and actresses (S. S. Kim et al., 2007). In this way, the Japanese and empathy are inseparable. It is therefore possible that, compared to other countries, the effect of empathy regarding advertisements is higher in Japan.

Empathy factors
Content consistency and realism are fundamental for empathy, and inconsistencies in the story can confuse consumers, causing them to be less likely to accept it (Busselle & Bilandzic, 2008). This is consistent with the long-standing emphasis of marketing communication on brand management; the success or failure of establishing a strong brand depends on consistent embodiments of the concept (Keller & Swaminathan, 2019). Additionally, when content is recognized as highly fictitious, it becomes difficult for consumers to empathize with it; thus, the models' behaviors in this content must exhibit realism (Argo et al., 2008). This makes it easier for consumers to understand and imagine a situation; if they cannot imagine using the advertised product when they encounter marketing communications, it will be difficult for such communications to evoke a positive attitude (Chang, 2013). Therefore, it is important that the product usage scenesdescribed by models and adopted for marketing communications-match the concepts of the product brands and are realistic.
As models used for marketing communications, celebrities have attracted attention because of their high effect on consumer behavior. Many studies on celebrities as models have been conducted because celebrities effectively convey messages (Till, 1998). For example, Tiger Woods' endorsement of Nike golf balls generated a price premium of approximately 2.5% (Chung et al., 2013). As previously mentioned, consistency and realism are important for enhancing the effectiveness of the celebrity model. For example, factors such as the similarities between celebrity images and products (Paul & Bhakar, 2018) and celebrities' product expertise (Pan & Meng, 2018) have been proven to be effective, although they have yet to address the empathy from everyday life that models exploit. Accordingly, the following hypothesis is proposed: H1: When observing usage scenes as depicted in marketing communications, empathetic consumers' beliefs-specifically that the celebrity model uses products in the same way as the consumers do in their daily lives-positively influences their purchase intentions.
However, appointing a celebrity carries the potential for risks that a company cannot manage. When a celebrity is appointed as an endorser for multiple products, the celebrity's credibility can decrease, and consumer attitudes toward the target product can become negative. A celebrity's image is also transmitted to the product, and consumers' trust in the celebrity is directly linked to their trust in the advertisement (Hussain et al., 2020); hence, if the celebrity is involved in a scandal, the advertising company will experience an economic loss (Huang, 2017). Accordingly, unknown models are often used for marketing communications, given their affordability and low risk. However, extant literature provides limited examples of models that also demonstrate factors that effectively evoke empathy and their effects. This is because consumers have no perceptions of unknown models, unlike celebrities. Although marketers and researchers have previously posited that it would be difficult to evaluate a model without initial existing images, consumers can exhibit measurable levels of empathy toward the model and the scenes described in marketing communications. However, from the perspective of empathy from everyday life that models exploit, it is considered that the same is true for an unknown model. Accordingly, the following hypothesis is proposed: H2: When observing usage scenes as depicted in marketing communications, empathetic consumers' beliefs-specifically that the unknown model uses products in the same way as the consumers do in their daily lives-positively influences their purchase intentions.
Finally, comparing the effects of empathy between the celebrity model and the unknown model, the former is assumed to be higher because, as demonstrated in previous studies, the celebrity model has the effect of attracting consumers (Chung et al., 2013;Till, 1998). Accordingly, the following hypothesis is proposed: H3: Compared to unknown models, celebrity models can predict purchase intentions based on empathy for advertising models.

Survey
An online survey was conducted from May 4 to 11 May 2021, targeting PC and SP owners in the Japanese market. All respondents provided informed consent, and the study design was approved by the appropriate ethics review board. The survey was e-mailed to a panel owned by a major Japanese research company, Cross Marketing, Inc., and the respondents were in their 20s to 30s. The PC and SP surveys were conducted using two different yet simultaneous surveys. The survey system was controlled to deter the same respondents from participating in both surveys. As illustrated in Table 1, each survey was conducted on 200 people, and 155 people who own PCs and 177 people who own SPs were the respondents of the study. Then, the respondents were randomly divided into two groups, and an image of marketing communication using celebrity models was used in one group and unknown models in the other group.
The brands selected for the evaluation were the Apple MacBook and Panasonic Let's Note for PCs and the Oppo Reno and Sony Xperia for SPs. Therefore, the sample size of the validation data was 332 people × 2 brands = 664 people or data points. These brands were chosen because, at the time of the survey, they used celebrity models and unknown models in their marketing communications. Additionally, to reduce bias in the results of this study, the small proportion of device owners of the target brands was also taken into account. Table 2 illustrates the brands owned by each respondent. In both surveys, the questions pertained to the following nine items: (1) gender, (2) age, (3) ownership status of PCs/SPs, (4) owned brand, (5) frequency of use, (6) purchasing emphasis point, (7) experience of contact with advertisements of the target brands, (8) purchase intention of the target brands, and (9) empathy for the model of the target brands. Question (3) was a screening question, and if the respondents did not own PCs/SPs, the survey ended without the respondents progressing to answer further questions. If the respondents had multiple PCs/SPs, they were instructed to respond primarily to the product they named in Question (4). Questions (5) through (6) asked the respondents about their product usage status and awareness. Questions (7)-(9) asked for two brands for each product category (PCs: Apple and Panasonic, SPs: Oppo and Sony). Question (7) was, "Have you recently seen advertisements for the following brands on TV, YouTube, street posters, etc.?" As the marketing communications targeted were developed through various channels, the content was set to ask comprehensive questions. The intent of Question (7) was to control the ease of empathy, primarily because it was presumed that the higher the frequency of contact with communication, the easier it is to understand and empathize with the content. Question (8) addressed this study's objective variable and was scored on a 7-point Likert scale, with responses ranging from 1 ("do not want to purchase at all") to 7 ("want to purchase very much").
Pictures of the marketing communications evaluated in this study were presented to the respondents before Question (9). This enables an evaluation without biasing purchase intention. Question (9) involved empathy, which was the subject of interest in this study, as follows: "This is a picture used on the website in [Product] of [Corporate Brand]. Can you empathize with the person in the picture as if they were using it in the same way in their daily life?" As with Question (8), a 7-point Likert scale was adopted. Measuring empathy with a scale is common in existing literature (Cartabuke et al., 2019;Ghaedi et al., 2020;J. Kim, 2018). At the time, each respondent was presented with the target, as illustrated in Figure 1. The pictures were selected from each product's website or YouTube account under the following conditions: (a) the unknown model has a low degree of recognition in Japan, and the celebrity model has a high degree of recognition; (b) the image depicted the product and the face of the person using it; and (c) the image was listed at the top of the website while satisfying the conditions of (a) and (b).
For Apple, the photographer Mika Ninagawa, who was adopted in the product's marketing communications, was selected as the celebrity model, and a picture was extracted from a YouTube video about Mac (Apple, 2021), and the unknown model was selected from videos on the 16-inch MacBook Pro website (Apple, n.d..). For Panasonic, the actor Manami Higa was selected as the celebrity model, and a picture was extracted from a YouTube video about Let's Note (Panasonic, 2020), and the unknown model was selected from a picture on the Let's Note FV Series website (Panasonic, n.d..). At Oppo, the pop idol Rino Sashihara was selected as the celebrity model, a picture was selected from the Reno 5A website (Oppo, 2023), and the unknown model was selected from a picture on the A73 website (Oppo, n.d.-a). At Sony, the film director Eiji Uchida was selected as the celebrity model, a picture was selected from the Xperia website (Sony, n.d.-a), and the unknown model was selected from a picture on the Xperia 1 III website (Sony, 2023).

Verification
Randomized controlled trials are desirable as a hypothesis verification method when randomization is possible. However, this study's subject of interest-empathy for the model-cannot be randomly assigned. Therefore, a propensity score was applied, a typical method for estimating effects when randomization is impossible (Rosenbaum & Rubin, 1983). Using this method, multiple covariates are aggregated into one variable, called the propensity score, which is used to correct the covariates in the treatment and control groups (Loke & Mattishent, 2020;Zhang, 2017). Respondents who empathized with the communications were the treatment group, and those who did not were the control group. As the true value of the propensity score for each subject is unknown, it is common to estimate this value using a logistic regression model. In estimating the effect, one method involves matching people with similar propensity scores between both groups; however, this reduces the sample size. Accordingly, the effects were estimated by inverse probability weighting, which produces a reciprocal weight of the propensity score. To verify H1 and H2 (Test 1), in the survey of each model, the difference in purchase intention was compared between those who empathized and those who did not. The propensity scores were estimated using logistic regression models; the objective variable was empathy (Item No. 2 in Table 3)-at this time, empathy was defined as "empathy" in the top three of the 7-point Likert scale and was made into a dummy variable-and the explanatory variables were the celebrity model dummy (No. 3), target brand dummy (Nos. 4-6), respondent attributes and values for the device (Nos. 7-15), and experience of contact with the advertisements (No. 16).
A stepwise method was adopted to select the variables, as there were several. Assuming that the propensity score of individual i is ps i , the expected value of the treatment group, result Y 1 (No. 1), and the control group, result Y 0 (No. 1), can be obtained using Equations (1) and (2). The average treatment effect (ATE) was calculated using E Y 1 ð Þ À E Y 0 ð Þ. When the ATE was positive, and the 95% confidence interval did not include 0, it was judged to be significant at the 5% level. This hypothesis was supported at that point. Further, N represents the sample size, z = 1 represents the treatment group, and z = 0 represents the control group.
Similarly, to verify H3 (Test 2), the difference in purchase intention was compared between those who empathized with the celebrity model and those who empathized with the unknown model. In other words, the objective variable in the logistic regression model that estimates the propensity score is the celebrity model dummy (No. 3). The R statistical analysis software suite was used to perform the analyses.

Results
First, when observing the status of empathy used by each brand (Table 4), the Apple celebrity model and Panasonic unknown model were highly empathetic for more than 45% of people. Consumers empathize with realism, which is consistent with the product concept. For instance, Apple's MacBook appeals to the concept of "the ultimate work tool for professionals who innovate in the world (Apple, n.d..)," and it is thought that the image of photographers working on PCs was recognized as a match.
Next, the propensity score was estimated using a logistic regression model. To test the validity of the models, the c-statistics (the area under the curve [AUC]) should be 0.7 or higher (Austin et al., 2015;W. Y. Kim et al., 2018;Yao et al., 2016). As Table 5 indicates, in the present study, all models exceeded 0.7, confirming a certain degree of validity. When examining the variables that became significant at the 5% level, Apple demonstrated a positive effect in the celebrity model in Test 1. As explained earlier, this demonstrates that the Apple celebrity model has a high degree of empathy. Brand and usability as purchasing emphasis points have positive effects, and they are thought to represent the characteristics of people who empathize with Apple's communication.
Additionally, the experience of contacting advertisements also contributes to the ease of empathy. Similarly, in the unknown model, Panasonic, which has high empathy, has a positive effect, and Sony, which has low empathy, has a negative effect. Test 2 demonstrates the characteristics of people who empathize with the celebrity model, compared with those who empathize with the unknown model. Therefore, Apple, which is significant only for the celebrity model, has a positive effect, and frequency, which is significant only for the unknown model, has a negative effect.  Notes: SD = standard deviation.
In Test 1, as the effect of empathy included an ATE of 0.769 in the celebrity model and 0.562 in the unknown model, and the 95% confidence intervals did not include 0, these were significant at the 5% level (Table 6). Hence, hypotheses H1 and H2 are supported. Similarly, the ATE of the celebrity model in Test 2 was also significant, meaning that H3 was supported.

Implications and limitations
Currently, media websites and corporate official YouTube accounts are at the center of marketing communications. With visual appeals to obtain consumers' empathy for a product, empathy for the model is required. Moreover, although many of the models used in websites' visual marketing content are not well-known celebrities, previous research on the relationship between models and products has focused on celebrities (Chung et al., 2013;Huang, 2017;Hussain et al., 2020;Pan & Meng, 2018;Paul & Bhakar, 2018;Till, 1998;Tripp et al., 1994). Very little academic discussion on the effective factors for adopting unknown models has taken place. Additionally, research is yet to be conducted regarding the empathy factor toward models using products in daily life, as depicted in an advertisement scene. Hence, the results of the present study help to bridge this gap in the literature.
This study has some practical implications. First, depictions used to symbolically convey product concepts in everyday use scenes should be carefully considered. Apple's MacBook demonstrated the highest level of empathy. Consumers evaluated the picture as a clever depiction of the concept, which is a work tool used by professionals such as photographers, designers, artists, and engineers (Apple, n.d..). Second, a clear concept is essential for effective marketing communication. It has been reported that effective factors are matching a celebrity's images with products (Paul & Bhakar, 2018) and the impression that a celebrity's expertise is related to the target product (Pan & Meng, 2018). However, it is impossible to decide who should be appointed as a product endorser without a clear concept, even when a celebrity is selected. For example, in this study, Oppo Reno had the lowest empathy in the celebrity model. The true reason the company hired this celebrity is unknown, but one could be that the names are pronounced the same (brand name: Reno, celebrity: Rino Sashihara). The results demonstrate that it is difficult to understand the product concept from solely a factor such as pronunciation, and thus, it is difficult to achieve empathy.
This study also has several limitations. First, the targets were limited to PCs and SPs in Japan; thus, the generalizability of the results is limited. As mentioned earlier, the Japanese tend to attach great importance to empathy (Karremans et al., 2011;Kelley, 2008;S. S. Kim et al., 2007;Nghiêm-Phú & Shibuya, 2021;Sugino, 2023;Watanabe & Yabu, 2021). It is therefore possible that the effect of empathy regarding advertisements is higher in Japan, compared with other countries. Ideally, this study should control for all such national characteristics to assess the effects of empathy; however, as national characteristics are diverse, it is difficult to specify what variables should be controlled for. In fact, there are many studies that did not control for national characteristics in examining the effects of empathy. These include studies on the effect of empathy on acceptability in American charity advertising (Bae, 2021), the effect of service recovery empathy on consumer attitudes in China (Wei et al., 2020), the effect of employee empathy on service loyalty in Chinese banks (Bahadur et al., 2019), the effect of fair trade coffee empathy on happiness in South Korea (Hwang & Kim, 2018), and the effect of customer-employee interaction empathy on loyalty in Germany (Wieseke et al., 2012). Considering the differences in empathy effects due to national characteristics, it is necessary to compare the same effects in multiple countries.
Second, the present study did not examine differences in the effects of celebrities' characteristics. The effect may therefore vary depending on the celebrity's occupation, popularity, and favorability. This study is the first to introduce empathetic consumers' beliefs into advertising effectiveness-specifically, that the unknown/celebrity model uses products in the same way as the consumers do in their daily lives. This study does not establish systematic knowledge about the above; it is just one model case. As knowledge regarding this content is overwhelmingly lacking, it is desirable to build comprehensive knowledge through continuous research.
Third, the influence of the country of origin of the model used for marketing communications is unclear. It is easier for consumers to induce empathy if they possess characteristics similar to those of the model (Komeda et al., 2013). Therefore, even with the same image, racial differences between the consumer and model may affect the results, and no conclusions about racial or characteristic similarities can be drawn from this study. However, given the current push toward globalization, it is necessary to unify global marketing communication. Hence, from a practical perspective, it is more important to create content that can generate high empathy, regardless of race, rather than focusing on the differences in effects that depend on race.
Fourth, the reliability of measurements by the Likert scale remains controversial. The Likert scale is the most widely used psychometric scale in questionnaire-based research (Anjaria, 2022). The advantage of Likert scales is that they can be applied to a wide range of fields, from social sciences to engineering (Pescaroli et al., 2020). However, as research evolves, staying on top of best practices can be difficult (Jebb et al., 2021). In particular, when there is novelty in the measurement target, such as empathy dealt with in this study, there are of course few precedents in the existing literature. In addition, under certain conditions, such as when the element being measured is unique or narrow in scope, the use of a single item may be appropriate and should not necessarily be considered unsound (Allen et al., 2022;Bergkvist & Rossiter, 2007;Fuchs & Diamantopoulos, 2009;Jordan & Turner, 2008;Matthews et al., 2022;Wanous et al., 1997). Some studies even argue that multi-item scales are redundant and that respondents' perceptions are better represented by single-item scales (O'Brien et al., 2022;Trail et al., 2023). Unfortunately, methods to improve the reliability of the Likert scales themselves and single-item measurements are beyond the scope of this study. The new hypotheses of this study should be verified by various methods in the future. Addressing these limitations will be a direction for future research.