In this paper we examine whether cultural characteristics lead to different emotion expressions whilst watching tourist destination ads via digital media, and thus, ads might be perceived differently depending on the country of origin of the viewer. To test this assumption, participants from two different countries, located on two different continents, Austria/Europe, and Colombia/South America, are exposed to a destination ad. Their faces are recorded and post-processing analysis on the recorded videos using the AFFDEX algorithm, that is capable of inferring emotions based on the facial action coding system (FACS), is conducted. Valence scores are compared among the viewers of the two countries over the time span of the whole commercial, and subpopulation differences of basic emotions (joy, surprise, anger, sadness, disgust, fear, and contempt) are explored using time series clustering along with optimizations for the dynamic time warping (DTW) distance. Screening sequences in this way reveals insight on the emotional reactions of different viewer groups. The findings instruct tourism marketers on how to fit the targeted emotions elicited by tourist destination advertising with various cultural settings.
1 Introduction
Destination image videos are strong instruments that influence people's perceptions of locations. Their purpose is to convey a certain image and to arouse the interest of both nationals and foreigners in a target destination and ultimately to persuade them to visit the destination. However, cultural differences have an impact on the perception and subsequently the effectiveness of advertising campaigns [1‐3].
As a result, investigating cultural differences in emotional responses to advertisements is supposed to shed light on the complex interplay between marketing communication and cultural contexts. Thus, this paper presents a methodology to explore how cultural factors contribute to the variation in emotional responses to advertisements. By examining the emotional dimensions that advertisements evoke in distinct cultural groups, we seek to uncover diverse affective responses, highlight key differences, and offer insights into the necessity of crafting culturally sensitive advertising campaigns. We answer the question if customers in various countries react differently to the same advertising appeal.
Anzeige
By considering affective responses as culturally constructed, this study contributes to the existing literature in three ways: First, this study responds to the demand of more emotion-based research [4] in tourism by measuring emotional reactions to tourism ads. However, in contrast to most previous studies in a tourism context, we do not rely on self-reported measures of emotions, which have inherent shortcomings [5]. Rather, we contribute to the short list of recent studies that employ psycho-physiological methods instead of, or in addition to, self-reporting measures by employing automatic facial expression analysis. Furthermore, our study goes beyond prior attempts, which have always assessed the effectiveness of the advertisement as a whole by examining responses on a frame-by-frame basis. Third, we employ time series clustering and dynamic time warping (DTW), which are new in this context and have not been applied before using such data. Thus, we analyse emotional responses in an objective, automated, and more detailed way than previous research has. The present study underpins and explains the results by employing a benchmark approach grounded on Hofstede’s national culture scores. The results finally allow us to offer valuable implications for practitioners seeking to engage diverse audiences effectively.
2 Literature Review
2.1 Cultural Aspects in Tourism Destination Promotion
According to Hofstede et al. [6], culture is the “collective programming” that distinguishes societies from one another. Minkov and Hofstede [7] demonstrate that at the national level, cultures can be evaluated and contrasted along four cultural dimensions, namely power distance, individualism-collectivism, masculinity-femininity, and uncertainty avoidance. Power distance describes the importance of individual hierarchy, the individualism-collectivism dimension indicates the level of interdependence among persons, masculinity-femininity the extent of stereotypical masculine or feminine characteristics, and uncertainty avoidance the degree of rule orientation [6].
These cultural dimensions reflect the value systems of a society, which influence lifestyle, employment, leisure, and consumer behaviour patterns [8]. These values are also crucial in explaining why attitudes toward advertising are not universal [3]. Thus, Hofstede’s typology was adopted by scholars to explore cross-cultural differences in advertising in general [2, 9] and in tourism destination promotion in particular [10‐12].
Moura, Gnoth, and Deans [11] adapted Hofstede’s cultural framework [6] for the evaluation of cultural values on tourism destination websites. Following this framework and applying it to destination commercials, individualism focuses on the uniqueness of the destination and therefore a commercial should contain images of solitude and self-fulfilment, whereas collectivism should be represented on the basis of the availability of activities related to interaction with the local community and depictions of families or teams. Power distance centred images contain prominent individuals in society and celebrities, as well as statements regarding the destination from people with societal power. To avoid uncertainty, facts about the destination and tourist services, contact information, maps, and so forth are recommended. Because tourism products are generally intangible, these visual aspects are especially important for destination advertising. Summarized, it is critical to pay particular attention to the visual appeals and to examine the emotional reactions to advertising in depth. For instance, Mele and Lobinger [10] argue that the representation of cultural values in tourist visuals is heavily connected to visual style parameters such as colours, lightning, scene composition, and the arrangement of actors.
Anzeige
However, in addition to cultural differences, novelty-seeking plays a crucial role in the selection of a destination [13]. Mitas and Bastiaansen [14] explore effects of novelty on positive emotions and find that the influence of tourism experience on happy emotions is mediated in part by novelty. As tourism commercials are designed to inspire and motivate people to explore new destinations and experiences, novelty-seeking also plays an important part when people watch tourism advertising. Thus, this aspect must be taken into account when analysing tourism ads.
Therefore, considering the novelty-seeking aspect, this study shows how to explore whether cultural differences are reflected in the emotional response to a tourism ad.
3 Data Collection and Pre-processing
A comparison of Hofstede’s national culture scores for Colombia and Austria shows that the two countries differ most significantly in terms of power distance (Austria/Colombia: 11/67) and individualism (55/13) [15]. To allow for comparison between the two countries, located on different continents with innate cultural idiosyncrasies due to their geographical locations, histories, traditions, and societal norms, 15 participants from Austria/Europe, and 34 from Colombia/South America were asked to watch an Austrian tourist destination commercial online. The study uses a convenience sample composed of personal contacts of the authors as well as students. Participants’ faces were recorded while watching the commercial using face detection.
The AFFDEX algorithm was used to post-process 18 facial expressions based on 34 facial landmarks using the facial action coding system (FACS) developed by Ekman & Friesen [16]. In the subsequent emotion expression modelling task, these facial expressions were used as input to estimate the likelihood for an emotion to occur. The resulting emotion predictors either have a positive or a negative effect on the likelihood of an emotion (e.g., nose wrinkle and upper lip raise increases the likelihood of disgust, lip suck and smile decreases the likelihood of disgust). Seven emotions (anger, contempt, disgust, fear, joy, sadness, and surprise) and valence (overall positive or negative emotion) were traced [17]. Interpolation was used to rescale varying time intervals to 40 ms.
4 Methodology
4.1 Smoothing
To stress long-lasting emotional reactions that occurred due to the influence of the promotional video (e.g., smiling), and flatten short lasting facial expressions other than emotional ones (e.g., sneezing), the raw time series scores were smoothed using rolling means with a time window of \(n=50\). . This data preparation step, so-called simple moving average (SMA), shortens each time series by two seconds.
Due to different facial expression intensities across the 49 study participants, the smoothed emotion scores were normalized onto a scale ranging from 0 to 1.
Mean aggregation of the smoothed normalized emotion scores was conducted at the country level. The resulting group means are displayed together with their 95%-confidence intervals (CI).
In the literature, it has been shown that only about 50% of study participants express emotions with observable facial features [18]. As a result, comparisons across the entire samples are not well suited to reveal cultural differences in response to the commercial as not all sample members disclose visually observable reactions. Therefore, time series clustering is used to group respondents according to facial expression similarities. This fulfils the need of disclosing subpopulations with time-variant but synchronous facial expressions to discrete scenes. Time series clustering is split into three categories, whole time series clustering, subsequence time series clustering, and time point clustering [19]. The first group of this taxonomy is used here to classify individuals based on time series of a selected emotion and video sequence.
4.6 Dynamic Time Warping (DTW)
When it comes to the selection of a dissimilarity measure for the clustering procedure, ‘structure-based’ concepts focus on underlying correlational structures, whereby ‘shape-based’ concepts stress proximities [20]. As ‘shape-based’ concepts can fail if there is a lot of noise or anomalous records, which is the case for short-lasting facial expressions not based on emotional grounds (e.g., itchiness), prior normalization is required (see Sect. 4.2). In combination with time series clustering along with optimizations for the DTW distance, this is the method of choice due to its insensitiveness to local compression and stretches [21].
For DTW, the elements of two time series are aligned with the purpose of determining an optimal path through their cross-distance matrix. The warping functions, \({\varnothing }_{x}\) and \({\varnothing }_{y}\), optimize the average accumulated distortion, \(d\), of two time series, \(X={x}_{1},\dots ,{x}_{N}\) and \(Y={y}_{1},\dots ,{y}_{M}\), by locally remapping their time point indices to the warping curve elements \(\varnothing (k)\).
Instead of skipping time series elements while aligning \(X\) and \(Y\) to each other, duplication allows a single time point of \(X\) to match multiple consecutive time points of \(Y\). The symmetric multiplicative weight of such transitions, \({m}_{\mathrm{\varnothing }}(k)\), is 1, whereby direct transitions take on the value 2. These weights can be set differently. No limit was specified for the number of time expansions or compressions, i.e., a time point of \(X\) might be matched with an unlimited number of \(Y\) elements. The optional boundary constraint ensures that the warping path starts and terminates at the starting and ending points of \(X\) and \(Y\). To preserve time order, monotonicity is imposed, and the continuity condition limits transitions to adjacent points in time. Figure 1 exemplifies two study participants. The solid line represents the reference time series, the dashed line the query time series, and the dotted lines the match guidelines. Irrelevant whether respondents are smiling longer or shorter, or start smiling earlier or later, shape similarities are identified.
Fig. 1.
DTW alignment example between two study participants.
×
The resulting minimum cumulative Euclidean distances are divided by \({M}_{\mathrm{\varnothing }}\), the sum of time points \(N+M\), to allow for comparison between different time series pairs. Finally, the global dissimilarity matrix, containing distances between all respondents’ time series, is handed over to an agglomerative hierarchical clustering algorithm using average linkage, whereby Ward.D2 and centroid linkage led to comparable results.
5 Results
The results section is split into two parts. First, valence scores are presented to explore the likelihood to perceive overall positive or negative emotions as well as country variations. Second, a selected basic emotion is examined for a particular sequence to exemplify differences between the study participants of the two countries.
5.1 Valence – Whole Commercial
Figure 2 contrasts the absolute smoothed normalized mean-aggregated valence scores along the time span of the Austrian commercial, 1 min 41.2 s, between Austrian and Colombian participants. Emotion scores run synchronously for the two countries after the introductory section of the commercial. Colombians show higher overall valence intensity levels compared with Austrians. The wider confidence interval of the Austrian sample is partly due to the smaller sample size of 15 participants, compared with 34 Colombians. But there are no striking differences, in the sense of non-overlapping confidence intervals, detected at the coarse level of overall valence.
Fig. 2.
Valence score means and 95%-CIs.
×
Other possibilities to stress these differences are presented in Fig. 3. Absolute (thick line) and relative (thin line) valence differences between the Austrian and Colombian sample, based on their smoothed normalized likelihood to experience overall positive or negative emotions, underline the higher intensity levels of Colombian participants. Negative differences and ratios less than one are black coloured if Colombians showed higher valence levels, in the opposite case coloured differently. The valence likelihood is higher for Colombian participants through a major part of the Austrian commercial.
Fig. 3.
Austrians and Colombians valence differences (thick line) and ratios (thin line).
×
As Colombian study participants watched a commercial that presents a country located at a different continent with possibly lots of new and inspiring content, compared with Austrians who watched a commercial about their home country in which not too much new impressions might be presented, likewise the cumulative overall valence score reaches a higher level throughout the whole commercial for Colombians (Fig. 4).
Fig. 4.
Valence for Austrian and Colombian participants.
×
In general, as exemplified in Fig. 4, the cumulative valence scores point into the expected direction through the time span of the Austrian commercial. On closer inspection, at the level of basic emotions, negative emotions like anger, contempt, disgust, or sadness sum up to lower levels for Colombians watching the Austrian commercial. On the contrary, surprise closes at higher levels. Fascinating content of promotional campaigns seem to affect emotions in a positive way.
Unexpectedly, Austrian study participants reached higher cumulative likelihood levels for the emotion joy. But this might be related to the findings for the basic emotion fear, since Colombian participants perceived higher levels of fear. Reasons might lie in a lot of new, unfamiliar, and unexpected content raising emotional discomfort. As a result, Austrian participants watching the Austrian commercial might experience their usual safety in their known environment, and only that opens gates to perceive higher levels of joy. Consequently, the dominance of the comfort zone seems to prevent Colombians from experiencing higher levels of joy. Nevertheless, in line with the cultural dimensions and the novelty arguments, all other single basic emotions pointed into the hypothesized direction.
5.2 Basic Emotions – Selected Sequence
Figure 2, 3 and 4 reveal differences between the participants of the two countries with respect to overall positive and negative emotions, i.e., valence. The following chapter presents an example at the differentiated level of single basic emotions. A remarkable difference between the two groups along the timeline of the whole commercial was identified between the 65th and 73rd second for the basic emotion surprise, depicted with dashed vertical lines (Fig. 5). The level of surprise is lower for the Austrian study participants. Whether all within group members behave in the same way, the home country group, and the foreign group respectively, a fine-grained analysis is needed.
Fig. 5.
Surprise score means and 95%-CIs.
×
As stated in the methodology section, only about 50% of the participants show feelings with facial expressions [18]. Therefore, time-series clustering along with optimizations for the DTW distance was used to identify clusters with similar patterns. As DTW is computationally burdensome for large datasets [20], instead of the whole standardized time series only the highlighted scene (Fig. 5) is used to identify four clusters with comparable behaviour patterns (Fig. 6).
Fig. 6.
Surprise time series cluster 65th to 73rd second (individuals = solid; centroids = dashed).
×
Study participants with strong emotional face expressions in the middle of the scene are separated from the rest of the sample, cluster 2 (12 pax) and cluster 4 (5 pax), see bold dashed centroids. This peak seems to be responsible for the slight increase in surprise for Colombians. The two clusters together represent 35% of all study participants, being close to the theoretically established and empirically validated 50% [18]. As Colombian participants are confronted with new content (novelty) and behave differently for the two Hofstede’s cultural dimensions (power distance and individualism/collectivism), they should also show a peak of surprise. In line with these arguments, 14 out of 17 pax (82%) who belong to these two groups are Colombian participants, but only 3 are Austrian (Note: Individuals are indicated with their country initials and running number).
The resulting cluster solution further assorted 12 persons (cluster 1) presenting rather flat patters, and 20 persons (cluster 3) with mixed and only slightly varying reactions. The latter has the lowest likelihood levels for surprise towards the end of the selected sequence. The six Austrians of cluster 3, out of all 12 Austrian study participants (50%), are responsible for the low average scores for surprise for this sequence.
6 Discussion
The results of this study clearly show that there are cultural differences in the emotional response to the same advertising appeal. In the introduction, we assume that these differences are due to different cultural values on the one hand, and to novelty-seeking aspects on the other. Although cultural effects cannot be empirically separated from novelty effects here, we discuss the country-specific differences based on Hofstede's dimensions [6], as this is of particular importance to destination management organizations (DMOs) who have the difficult challenge of promoting their destinations in expanding markets.
One important finding of the study is, that Colombians show higher overall valence levels compared with Austrians while watching an Austrian tourist destination commercial. From a cultural perspective, this can be explained by the generally lower individualism and power distance values of Colombia (13 and 11, compared to 55 and 67 for Austria). Colombians are generally recognized for their passionate personalities who frequently value interpersonal interactions and emotional connections. Thus, it can be derived that they are more likely to respond positively to campaigns that evoke emotions or elicit empathy, which is the case for the commercial under study. It displays people enjoying their holiday including scenes such as a group of people swimming and having fun by splashing each other with water, a woman running through the forest and smiling, two boys running around with a kite and having fun, or people at different ages enjoying traditional food at an alpine pasture. In contrast, Austrians tend to prefer straightforward factual communication. Thus, emotional appeals need to be subtle and aligned with cultural norms to resonate with the Austrian audience without seeming overly dramatic or insincere. This could be realized by providing more factual information, such as entitling the areas/locations courted in the video, which was not the case for the commercial at hand.
In conclusion, the findings of this exploratory research indicate that there are cross-cultural differences in advertising perception which can be explained by Hofstede’s cultural dimensions. Thus, successful advertising campaigns need a thorough awareness of these distinctions as well as a specialized approach to resonate with each group. Therefore, destination marketers are highly recommended to use A/B testing that features cultural differences, such as collectivist cues (vs. individualist cues). Alternatively, marketers should take care to make advertising as neutral as possible in terms of cultural cues to outweigh any potential cultural differences. In any case, they must be culturally aware, use suitable imagery, and emotional signals. Advertisers may build compelling ads that effectively engage both Colombians and Austrians by considering these distinctions.
This study has also important implications for academia and future research. In terms of methodology, the problem of delays in the emotional reactions among multiple sequences must be mentioned. In this regard, future research should weaken the boundary constraint of the DTW optimization and free the ending parameter to allow for unconstrained endpoints, i.e., partial mapping. But the exclusion of overlaps in the identification of emotions belonging to one single sequence requires its separate presentation, and associated therewith, a modified face recording attempt, that does not allow for conclusions at the cumulative level as drawn for the present study.
An important limitation is the selection of an Austrian destination commercial. Thus, in addition to cultural differences, the analysis is based on different tourist segments (inbound vs outbound). This entails a different familiarity with the destination that could influence emotional reactions to the ad due to, e.g., novelty aspects. Although we controlled for destination familiarity, this could threaten the validity of the study.
Secondly, there should be a qualitative analysis of the scenes. For example, Mele and Lobinger [10] found that the representation of cultural values in visuals is connected to parameters such as shot composition (number of people in an image). A prototypical example of individualism consists of a picture showing a single subject positioned far from the viewer (long shot), as it reinforces the idea of independence, whereas increased proximity (e.g., close-up shot) with multiple persons (e.g., a family) can emphasize closeness and togetherness – which relate to collectivism. Detailed sequential analyses are needed to theoretically support the detected differences.
Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.
The images or other third party material in this chapter are included in the chapter's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the chapter's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.
