Tweets to escape: Intercultural differences in consumer expectations and risk behavior during the COVID-19 lockdown in three European countries

Abstract

This study aims to understand the extent to which a time of emergency, (e.g. the COVID-19 pandemic), impacts consumer behaviour in terms of risk and expectations. The methodology involves the systematic content analysis of 15,000 tweets collected from three countries (UK, Italy and Spain) in April 2020. The results suggest that the top-of-mind expectation by consumers deals with escaping from home and enjoying freedom, either by having a good meal (UK), drinking alcoholic beverages (Spain), or travelling (Italy). They also suggest that the high levels of risk individuals were exposed to during the pandemic will not influence behavior in the long-term post-lockdown. Instead, they suggest consumers are willing to restore their consumption levels especially of activities that contribute to the sense of escapism. Finally, results provide evidence of the cultural differences emerging from consumers from different countries during the pandemic. Implications for international marketers and retailers are provided.

1. Introduction

Emergencies such as terrorist attacks, natural disasters, crisis situations, and diffusion of viruses such as the COVID-19 epidemic outbreak (starting in early 2020) are some of the main crises that the world has faced in the last 20 years. The magnitude and severity of each crisis was different, as well as their consequences to people, societies, and industries. During these crises, studies have explored consumers’ behavior (e.g. Forbes, 2017, Priporas et al., 2015, Rottier et al., 2003, Wen et al., 2005), as well as businesses’ responses amid the crises (e.g. Green et al., 2004, Hasanat et al., 2020).

The size and duration of this pandemic (COVID-19) seem to create a new reality for societies, economic sectors such as retailing, and economies at global level. Consumers’ concerns about their well-being has several implications for retailers, which adopted different responses (Pantano, Pizzi, Scarpi, & Dennis, 2020). McKinsey (2020) reports that most consumers expect to shop for groceries online more frequently and visit physical stores for other items, simultaneously shifting that spending online. Although many consumers have started shopping online for groceries during the pandemic, future intention for online grocery shopping is not universal. For example, consumers in countries like the UK and Italy allowed the opening of non-essential retailers only in a subsequent phase (even with limitations such as social distancing, hand sanitizers available for clients, the usage of masks by clients, etc.), while new lockdowns (late 2020 and early 2021) further imposed closures in the whole country or in limited areas (Regions/Counties).

Consumers’ buying intentions remain an important area for research especially when there are uncertainties about how consumers will react after states of emergency are lifted in terms of shopping (Goddard, 2020, Hall et al., 2020, Pantano et al., 2020, Roggeveen and Sethuraman, 2020), and consumers increasingly seek information about the virus spread online (Pantano, 2021). Consumers might further maintain habits they have adopted during the epidemic also when it is over. Retailers will need to identify those ‘new and lasting’ behaviors that are a genuine shift in consumer behaviour. For example, engaging with a whole new generation of shoppers in different ways, as older consumers learn to shop online and decide that it is much more convenient for them (Martucci, 2020), while other consumers have showed panic buying behaviours and physically and verbally assaulted retail employees . Therefore, there is a need to understand these changes and contribute to a more sustainable post-pandemic retailing sector. Furthermore, recent studies demonstrated that the previously developed algorithms to predict consumer behaviours (i.e., machine learning algorithms to predict consumers’ preferences) are not consistent anymore (Heaven, 2020).

In order to fill these gaps, the current study aims to understand the extent to which an emergency time, such as the one during the COVID-19 outbreak, impacts consumers new and lasting behaviors in terms of risk and expectations. This is explored through the analysis of spontaneous consumers’ online communications in the form of tweets in three different European Countries (UK, Italy and Spain). These three countries were selected due to fact that they have similar population but many cultural differences and different responses (national policies) to the outbreak, leading to high death tolls. Furthermore, the current study centred on the following key research questions:

RQ1: How does the emergency impact consumers new behaviours in terms of risk?

RQ2: How does the emergency impact consumers new behaviours in terms of expectations?

RQ3: How do the intercultural differences in European countries impact on the consumers new behaviors?

To address these questions, 15,000 tweets in total from consumers from the study’s countries referring to the activities to do after the pandemic were collected and analyzed. In other words, these tweets include the expected behaviours for when the pandemic is over. The findings of this study will contribute to ongoing research on post COVID-19 consumer behavior and to pro-actively planning for the post-pandemic retailing. By doing so, the present research contributes to the identification of new strategies that businesses can adopt in order to meet the new consumers’ expectations shaped by the emergency period. The remainder of the paper is organized as follows: the next section will review the literature on risk aversion behaviour, and intercultural differences in risk perceptions. The subsequent one will show the methodology of research and the main findings. Finally, the paper will discuss the results and provide implications for marketing theory and practice.

2. Theoretical background

2.1. Risk aversion behaviour

Experiencing a natural disaster changes consumers’ perception of background risks and motivates them to take fewer risks (Cameron & Shah, 2015). Indeed, consumer decision-making, or choice, is an important field in the consumer behavior domain. When consumers make choices, they face uncertainties caused by future consequences that are difficult to know with certainty (Taylor, 1974).

In 1960, Bauer introduced the concepts of risk and uncertainty to the marketing literature. According to Bauer all buying behaviors were subject to certain risks and that there was uncertainty regarding the effect of risk perception on shopping results. Specifically, Bauer (1960) pointed out that ‘‘consumer behavior involves risk in the sense that any action of a consumer will produce consequences which he cannot anticipate with anything approximating certainty, and some of which at least are likely to be unpleasant’’ (p. 390). Although in the literature risk and uncertainty are used interchangeably, they can be distinguished based on the probabilities of their outcomes. Risk exists in a decision when the probabilities of outcomes are known, while uncertainty exists when the probabilities of outcomes are not known (Quintal, Lee, & Soutar, 2010). For McConnell and Dillon (1997) uncertainty is always present while, risk might not be. Risk is only present when the uncertain outcomes of a decision are regarded by the decision maker as significant or worth worrying about, (i.e. the decision maker’s well-being).

However, each person may have various assumptions or perceive differently the risks depending on their knowledge, past experiences, personality traits, emotions, and acceptable risk levels (Aren and Hamamci, 2020, Dohmen et al., 2011, Mandrik and Bao, 2005, Nicholson et al., 2005, Nguyen and Noussair, 2014). Outreville (2014) rightfully points out that different people will respond to similar risky circumstances in quite different ways. Similalry, individuals might respond to risk different accordingly to their culture (Rieger, Wang, & Hens, 2015).

Consumers can be distinguished into different segments based on level of risk they are willing to take in each situation (Mandrik & Bao, 2005). This attitude toward risk is known as risk aversion (Matzler, Grabner, & Bidmon, 2008). In general, risk aversion is the level of risk that people do not want to accept. Risk aversion has been described as a cultural value, a personality trait, or a consumer decision making style (Bao, Zhou, & Su, 2003). Hofstede and Bond (1984) defined it as “the extent to which people feel threatened by ambiguous situations and have created beliefs and institutions that try to avoid these” (p. 419).

In general, it is presumed that people are risk averse, however this aversion may differ between people (Sun, 2014) as various biases and psychological traits affect peoples’ risk aversion behavior (Aren & Hamamci, 2020). Risk aversion could influence consumers’ decision-making especially under uncertain situations and it is closely linked to the concept of uncertainty avoidance (Meroño-Cerdán, López-Nicolás, & Molina-Castillo, 2017). In the same vein, the impact of risk aversion on consumer decision-making may vary in situations that are dominated by various types of risks, such as high social risk (Mandrik & Bao, 2005). High risk-averse individuals are inclined to feel threatened by risky and uncertain situations (Hofstede, 1991).

Although recent marketing campaigns during the pandemic would raise awareness and concern about the risk of contagion (Kirk and Rifkin, 2020, Roggeveen and Sethuraman, 2020), there are no studies on the possible consumer behaviors after the pandemic, and on the understanding of the extent to which risk aversion behavior might persist.

2.2. Intercultural differences in risk perceptions and individual behaviors

A relevant theoretical basis to understand how people from different cultures react to risk perceptions is provided by the Hofstede, 1980, Hofstede, 2001, Fukuyama, 1995 frameworks. Previous literature has demonstrated that the dimensions which define cultural differences according to Hofstede and Fukuyama enable to highlight the difference in risk perceptions, and how individuals cope with risky and/or uncertain situations. For instance, Anglo-American and Mediterranean cultures have been found to react differently to privacy threats as a function of their different approaches to risk (Dinev et al., 2006), or to the introduction of new products on the market (Griffith & Yalcinkaya, 2008). More generally, extant research found that culture shapes the extent to which individuals are willing to accept risks, including personal, economic and financial risk (Rieger et al., 2015, Sharma and Singh, 2018). According to Hofstede (2001), national cultures can be operationalized and quantified by relying on six major factors: individualism-collectivism, uncertainty avoidance, power distance, masculinity-femininity, long-term orientation, and indulgence.

The individualism dimension reflects the degree to which individuals are integrated into groups. On the one hand, in an individualistic society, personal achievement and individual values are stressed, which may lead to overconfidence and increased optimism (Cheon & Lee, 2018). As a consequence, risk-seeking attitudes might be cultivated and/or socially accepted. On the other hand, it has been found that countries with collectivistic cultural traditions tend to perceive less financial risk and appear to be less risk averse (Bontempo et al., 1997, Hsee and Weber, 1999). In this regard, cultural theory posits that individuals exhibit different levels of risk perceptions which tend to be consistent with their cultural way of conceiving the social organization (Kahan, 2012). Accordingly, individuals from individualistic cultures tend to take risks into account as long as they see their own freedom threatened by external events. Conversely, individuals from collectivistic cultures tend to be less concerned with their own achievements, and more risk averse for things that might threaten the well-being of the society as a whole. This is to say, risk aversion behaviors can be found both in individualistic and collectivistic societies, although dealing with different foci of risk (Xue, Hine, Loi, Thorsteinsson, & Phillips, 2014).

In general, albeit with significant differences between specific countries, the Anglo-American countries have a highly individualistic culture, whereas Mediterranean countries display higher levels of collectivism (Triandis, Bontempo, Villareal, Asai, & Lucca, 1988). The COVID-19 pandemic exerted a global effect on countries imposing several limitations on consumers worldwide (Pantano et al., 2020). In the present work, we argue that individuals from countries whose culture is more consistent with individualistic or collectivistic models do not differ on the extent to which they perceived to be at risk as a result of the pandemic emergency, but rather on the aspects of life that were perceived to be more threatened. Specifically, we advance that:

Proposition 1: After the restrictive measures, consumers from individualistic countries will be more concerned with restoring those activities which fulfil the self, whilst consumers from collectivistic countries will be more concerned with restoring their social activities.

Another dimension which helps explaining cultural differences in risk perception is given by uncertainty avoidance. This index captures the extent to which a society can tolerate an uncertain or ambiguous situation. Accordingly, one might infer that a high uncertainty avoidance corresponds to risk, although Hofstede (2001) warned that this is not necessarily the case since people in uncertainty-avoiding cultures may also take more risks to reduce ambiguity (Shah, 2012).

In this vein, literature has proposed that individuals tend to be more reluctant to changes and to adopt more rigid behavioral standards in cultures denoted by high levels of uncertainty avoidance (Steenkamp, Ter Hofstede, & Wedel, 1999). Therefore, individuals are less likely to diverge from their established patterns and accept changes which might alter their normal behaviors by adding more ambiguity (Astakhova, Beal, & Camp, 2017). In the specific domain of marketing research, these considerations have been translated by observing how, for instance, uncertainty avoidance affects individuals’ tolerance for errors in the delivery of products and services (Bolton & Agarwal, 2010). Similarly, uncertainty avoidance has been found to also shape consumers’ expectations about their future consumption standards (Guesalaga, Pierce, & Scaraboto, 2016) which, in turn, affect, consumers’ preference for variety in the assortment of options they can choose from (Herrmann & Heitmann, 2006), as well as the extent to which word of mouth communication is capable of affecting one’s choices (Chiu, Chen, Wang, & Hsu, 2019).

Accordingly, individuals might display different mechanisms to cope with risk perceptions due to an unpredicted emergency situation and develop different expectations about their future consumption depending also on the level of uncertainty avoidance. Specifically, the present research posits that individuals belonging to cultures high in uncertainty avoidance will be more likely to express their desire to reduce risks and ambiguities by restoring their previous habits, whilst cultures scoring lower in uncertainty avoidance will exhibit higher tolerance of the new risky situation and the new habits it imposes. Specifically, we advance that:

Proposition 2: After the restrictive measures, consumers from cultures scoring high on uncertainty avoidance will be more likely to express their desire to restore their previous habits than consumers from cultures scoring lower on uncertainty avoidance who will be more concerned with the new patterns of behavior which need to be adopted to cope with risk.

3. Methodology

3.1. Case research selection

The present study investigated Italy, Spain and UK. These countries are chosen based on the different timing and severity of the containment measures adopted by their respective governments (lockdown), and comparable number of inhabitants (about 66 millions UK, 60 Italy and 47 Spain). Although the World Health Organization (WHO) declared the COVID-19 outbreak a public health emergency of international concern on the 30th of January 2020 (ECDC, 2020), Italian, Spanish and UK Governments’ speed of response was very different. Specifically, Italy was the first to adopt very restrictive containment measures immediately at the outbreak of the pandemic emergency in Europe, whilst the UK adopted milder containment measures after Italy and Spain. Indeed, Italy stopped all non-essential activities (i.e., closing non-grocery storesc.) on the 26th of March and Spain on the 29th, while UK did not stop (EU, 2020). Similarly, Spain closed the land borders on the 16th of March, while UK and Italy did not; and the non-essential movement was banned (lockdown) on the 10th of March in Italy, 16th of March for Spain and 24th of March for UK (EU, 2020). Furthermore, the three countries are characterized by different levels of individualism and uncertainty avoidance (Hofstede, 2001) which might determine both different risk perceptions and future expectations with regards to their behavioral intentions after the emergency situation imposed by the pandemic emergency. Specifically, the UK and Spain are the countries scoring respectively higher and lower on individualism; the situation is reversed with regards to Uncertainty Avoidance which is highest for Spain and lowest for the UK, as detailed in Fig. 1 .

Fig. 1.

Hofstede cultural indices for Italy, Spain and the United Kingdom.

3.2. Data collection and procedure

Consumers’ online communications on Twitter are largely considered a trustworthy source of data for consumer research based on consumers’ spontaneous and volunteer expressions of interest towards specific products and brands (Aleti et al., 2019, Arora et al., 2019, Athwal et al., 2019, Klostermann et al., 2018, Pantano and Stylos, 2020, Tellis et al., 2019, Villarroel Ordens et al., 2019). The analysis of tweets (through content analysis) enables determining the frequency with which certain concepts are mentioned (Berman, Melumad, Humphrey, & Meyer, 2019).

Among the different options for selecting tweets, the present research is based on selection by keyword. Specifically, researchers interrogated the Twitter database to provide all the tweets including the hashtag “#TheFirstThingIDoAfterTheLockdown”, (the hashtag has been further translated in Spanish and in Italian as “#PrimeroQueHagoDespuesdela Pandemia” and “#PrimaCosaCheFaccioDopolaPandemia” respectively). The research employed Wolfram Mathematica^TM software for data collection, as increasingly used in marketing studies to collect consumers’ online spontaneous communication analysis (Pantano & Stylos, 2020). We collected tweets between the 1st and the 15th of April 2020. We chose this period since almost any country in Europe had already experienced at least two weeks in lockdown, and they were already preparing to the phase 2 with the slow reopening of non-essential businesses.

The software can automatically download tweets through the ‘ServiceConnect’ function, by using an automatic connection via Twitter API and importing the parameters as required (i.e., inclusion of the specific hashtag, time and location). This procedure led to 5,013 for the UK, 5,002 Italy and 5,034 for Spain. Since these tweets are considered consumers’ spontaneous communications, thus not solicited by any specific request, we might assume that these are representative of all people in the three countries considering the specific hashtag (“#TheFirstThingIDoAfterTheLockdown”). However, the software automatically removed the tweets that only contained the hashtag without more text (i.e., the tweets only including a GIF/picture/video). This procedure resulted to 5,000 per each country, for a total of 15,000 tweets by unique Twitter users.

The tweets were entered into three independent databases (one for UK, one for Italy and one for Spain) with the full text of the collected tweets and other meta-data, such as user ID, to uniquely identify the author of each tweet, date and time of publication, language etc.

The research consisted of two different text classification techniques based on systematic analyses of the specific contents included in tweets, to support the extraction of relevant contents in terms of words and phrases extraction, through WordStat software. This software is able to process and manage alphanumeric data and provide varied text-analytical techniques through the integration of quantitative and qualitative features, while allowing the interpretation of textual data as the contents of tweets through the identification of significant concepts (words and word associations included in the tweets) (Davlembayeva, Papagiannidis, & Alamanos, 2019). In this way, it nurtures the objectivity, the replicability and the generalizability of the research methodology and findings (Davlembayeva et al., 2019).

3.3. Contents extraction

WordStat^TM software analyzes the words co-occurrences and phrases (as the word association with a meaning included in the “categorization dictionary”). Specifically, a multidimensional scaling map will be used to represent the co-occurrence of keywords or similarity of tweets (Fig. 2, Fig. 3, Fig. 4 ). For the phrases’ extraction, the software will further automatically extract the most important phrases from tweets in order to identify the thematic structures through computing a word frequency matrix. To this end, a factor analysis with Varimax rotation is performed in order to extract the most important factors (all the words with factor loadings higher than a specific criterion are considered as part of the extracted phrases).

Fig. 2.

Multidimensional scaling map representing the co-occurrence of words in at least 60 tweets (UK).

Fig. 3.

Multidimensional scaling map representing the co-occurrence of words in at least 60 tweets (Italy).

Fig. 4.

Multidimensional scaling map representing the co-occurrence of words in at least 60 tweets (Spain).

Since some words were rarer than others but equally more predictive, it was necessary to weight them more heavily. Thus, if considering tf as the total frequency, and idf as i word in the document d (part of D total documents) frequency, formula (1) adjusts the infrequent occurrence of words as (Humphreys & Wang, 2018):

$$\mathit{tf}\bullet idf=[1+\log \left(\mathrm{n}\mathrm{u}\mathrm{m}\mathrm{b}\mathrm{e}\mathrm{r}\mathrm{o}\mathrm{f}\mathrm{o}\mathrm{c}\mathrm{c}\mathrm{u}\mathrm{r}\mathrm{r}\mathrm{e}\mathrm{n}\mathrm{c}\mathrm{e}\mathrm{s}\mathrm{o}\mathrm{f}wordi\mathrm{n}d\right)\times \mathrm{l}\mathrm{o}\mathrm{g}\left(\frac{\mathrm{t}\mathrm{o}\mathrm{t}\mathrm{a}\mathrm{l}\mathrm{n}\mathrm{u}\mathrm{m}\mathrm{b}\mathrm{e}\mathrm{r}\mathrm{o}\mathrm{f}\mathrm{d}\mathrm{o}\mathrm{c}\mathrm{u}\mathrm{m}\mathrm{e}\mathrm{n}\mathrm{t}\mathrm{s}\mathrm{i}\mathrm{n}D}{\mathrm{n}\mathrm{u}\mathrm{m}\mathrm{b}\mathrm{e}\mathrm{r}\mathrm{o}\mathrm{f}\mathrm{d}\mathrm{o}\mathrm{c}\mathrm{u}\mathrm{m}\mathrm{e}\mathrm{n}\mathrm{t}\mathrm{s}\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{a}\mathrm{i}\mathrm{n}\mathrm{i}\mathrm{n}\mathrm{g}w}\right)$$ (1)

In particular, this formula assumes that the weight of each phrase considers that the more often a particular term occurs in a text, the higher its representativeness of its content, and the more text (tweets) in which the term occurs, the less discriminating it is. In other words, the algorithm embedded in the software scans the tweets and identifies the most frequent phrases (words association with a meaning included in the “categorization dictionary”).

To further reduce the number of phrases to the most meaningful ones and avoid redundancy, the software provides the function ‘distance’ (a machine learning algorithm able to evaluate the similarities of phrases). This algorithm allows the identification of the similarities among phrases through a number/weight as the Levenshtein distance between strings (the higher the number the higher the distance) (Levenshtein, 1965). Consequently, the system removes the phrases with the smallest distance, thus the resulting set of phrases do not include overlap. For this research, we also parameterized the system to consider only the phrases with minimum length of 3 (three words). For validity purposes, a stratified random sampling has been adopted by researchers to ensure that each phrases is reflected consistently across the tweets’ dataset, by considering 10–20% of the entire corpus of tweets (Humpreys, 2010).

Subsequently, each phrase was grouped in a topic. To this end, we used human coders, and the phrase’s grouping was circulated among three researchers who were unfamiliar with the study purposes. They were specifically instructed to evaluate the phrases included in the group per each defined group of phrases, per each case analysis. The phrases that were finally included in the groups were only the ones for which there was a match between the evaluations of at least two out of three researchers (Humpreys & Wang, 2018; Table 2, Table 4, Table 6).We further grouped the phrases according to their similarity to a main topic.

Table 2.

Phrases grouped according to the different behavior they describe (UK).

BEHAVIOURS	PHRASES
Eating and drinking	MILLION PEOPLE ARE FREE TO DINE OUT
	EAT AGAIN XXX
	HAVE A GOOD DRINK OUT
	DRINK WITH FRIENDS
Back to the previous life
	NIGHTOUT LIKE NOTHING HAPPENED
	FINALLY GETTING MARRIED
	GOING TO CHURCH AGAIN
	MEETING AGAIN WITH FRIENDS
Shopping
	REOPENING MY FAVOURITE SHOPS
	SELLING XXXX AFTER THE LOCKDOWN
Travel and tourism
	ESCAPE THE HOUSE
	SUMMER VACATION
Fear of new contagion
	BUSIEST ROADS SHOULD BE CLOSED
	FEAR OF GOING OUTSIDE
	KEEP SOCIAL DISTANCING
Other Business
	BREXIT HAPPENS
	MASTERCHEF AUDITION AFTER LOCKDOWN HARD TIME IS COMING FOR BARBERS

BEHAVIOURS	PHRASES
Eating and drinking	MILLION PEOPLE ARE FREE TO DINE OUT
	EAT AGAIN XXX1
	HAVE A GOOD DRINK OUT
	DRINK WITH FRIENDS
Back to the previous life
	NIGHTOUT LIKE NOTHING HAPPENED
	FINALLY GETTING MARRIED
	GOING TO CHURCH AGAIN
	MEETING AGAIN WITH FRIENDS
Shopping
	REOPENING MY FAVOURITE SHOPS
	SELLING XXXX2 AFTER THE LOCKDOWN
Travel and tourism
	ESCAPE THE HOUSE
	SUMMER VACATION
Fear of new contagion
	BUSIEST ROADS SHOULD BE CLOSED
	FEAR OF GOING OUTSIDE
	KEEP SOCIAL DISTANCING
Other Business
	BREXIT HAPPENS
	MASTERCHEF AUDITION AFTER LOCKDOWNHARD TIME IS COMING FOR BARBERS

¹Name of the fast food chain replaced.

²Name of the specific brand replaced.

Table 4.

Phrases grouped according to the different behavior they describe (Italy).

BEHAVIOURS	PHRASES
Travel and tourism
	I WANT ONLY TRAVEL
	GO TO THE SEA
	SWIMMING IN THE SEA
	I WANT TO GO SOMEWHERE
Back to the previous life
	FIRST DAY OUT WITH (FEMALE) FRIENDS
	THE FIRST REAL HUG
	I WANT TO GO OUT WITH MY (FEMALE) FRIENDS
	THE BEST NIGHT OUT EVER
	AS THE OUTBREAK NEVER HAPPENED
	I JUST WANT TO GO OUT
	GO HOME (FINALLY)
	I WANT TO MEET MY GIRLFRIEND
Eating and drinking
	JUST A BEER
Shopping	GO SHOPPING ANYWHERE I WANT
Other Business
	IMMEDITALY TO THE GYM
	GO TO BARBER FOR SHAVING
	AT LEAST ONE TATOO

Table 6.

Phrases grouped according to the different behavior they describe (Spain).

BEHAVIOURS	PHRASES
Eating and drinking
	DRINKING ALCOHOL
	ALL WINE ABLE TO DRINK
	HAVING A STEAK AND WINE WITH MY FRIENDS
	WALKING DRINKING A BEER
Back to the previous life
	GOING TO THE BEAUTICIAN
	ENTERING AGAIN THE SQUARE
	KISSING
	HAVING THE FIRST PARTY WITH ALL FAMILY
	ORGANIZING A PARTY
	MEETING AGAIN FRIENDS
	VISITING FRIENDS AGAIN
	MEETING AGAIN MOM
	FINALLY ENDING THE SOCIAL DISTANCING
	MEETING AGAIN MY FAMILY
	GO TO CINEMA
	GO TO FOOTBALL STADIUM
	THE FIRST NIGHTOUT
Other Business
	DOING A TATOO WITH ALL FRIENDS TOGETHER

4. Findings

4.1. Study 1: UK

Fig. 2 represents the map by considering the co-occurrence value of 60 (meaning that each couple of words should appear at least in 60 tweets). Table 1 represents the obtained 19 phrases¹ from the original set of 631.

Table 1.

Phrases extracted (UK).

PHRASES	FREQUENCY	% CASES	TF • IDF
MILLION PEOPLE ARE FREE TO DINE OUT	97	1,94%	166,1
EAT AGAIN XXX2	65	1,30%	122,6
NIGHTOUT LIKE NOTHING HAPPENED	37	0,74%	78,8
SMOKE A FEW CIGARETTES EVERYDAY	35	0,70%	75,4
REOPENING MY FAVOURITE SHOPS AND PUBS	25	0,50%	57,5
BUSIEST ROADS SHOULD BE CLOSED	22	0,44%	51,8
BREXIT HAPPENS	20	0,40%	48
MEETING AGAIN WITH FRIENDS	13	0,26%	33,6
MASTERCHEF AUDITION AFTER LOCKDOWN	11	0,22%	29,2
FEAR OF GOING OUTSIDE	8	0,16%	22,4
HARD TIME IS COMING FOR BARBERS	8	0,16%	22,4
KEEP SOCIAL DISTANCING	6	0,12%	17,5
HAVE A GOOD DRINK OUT	5	0,10%	15
DRINK WITH FRIENDS	5	0,10%	15
SELLING CONVERSE AFTER THE LOCKDOWN	5	0,10%	15
FINALLY GETTING MARRIED	4	0,08%	12,4
GOING TO CHURCH AGAIN	3	0,06%	9,7
ESCAPE THE HOUSE	3	0,06%	9,7
SUMMER VACATION	3	0,06%	9,7

¹ Please note that the system automatically removes the emoticons and pictures associated with the tweets. Thus, they do not appear in our tables.

²“XXX” replaces the name of a specific food from an international fast food chain for anonymity purposes of consumers and brands.

These phrases can be grouped according to the different behavior they describe (Table 2 ).

4.2. Study 2: Italy

Fig. 3 represents the map by considering the co-occurrence value of 60 (meaning that each couple of words should appear at least in 60 tweets).

Table 3 represents the obtained 18 phrases from the original set of 1,398. The phrases were analyzed in Italian and translated in English afterwards for convenience purposes.

Table 3.

Phrases extracted (Italy).

	FREQUENCY	% CASES	TF • IDF
I WANT ONLY TRAVEL	142	2,88%	218,9
AT LEAST ONE TATOO	142	2,88%	218,9
FIRST DAY OUT WITH (FEMALE) FRIENDS	41	0,83%	85,3
THE FIRST REAL HUG	25	0,51%	57,4
I GO TO THE FIRST RESTAURANT I FIND	23	0,47%	53,6
I WANT TO GO SOMEWHERE	22	0,45%	51,7
I WANT TO GO OUT WITH MY (FEMALE) FRIENDS	18	0,36%	43,9
THE BEST NIGHT OUT EVER	10	0,20%	26,9
AS THE OUTBREAK NEVER HAPPENED	7	0,14%	19,9
I JUST WANT TO GO OUT	7	0,14%	19,9
GO SHOPPING ANYWHERE I WANT	6	0,12%	17,5
I WANT TO MEET MY GIRLFRIEND	6	0,12%	17,5
IMMEDITALY TO THE GYM	4	0,08%	12,4
GO TO THE SEA	3	0,06%	9,6
SWIMMING IN THE SEA	3	0,06%	9,6
GO TO BARBER FOR SHAVING	3	0,06%	9,6
GO HOME (FINALLY)	3	0,06%	9,6
JUST A BEER	3	0,06%	9,6

These phrases can be grouped according to the different behavior they describe (Table 4 ).

4.3. Study 3: Spain

Fig. 4 represents the map by considering the co-occurrence value of 60 (meaning that each couple of words should appear at least in 60 tweets).

Table 5 represents the obtained 18 phrases from the original set of 1,597. The phrases were analyzed in Spanish and translated in English afterwards for convenience purposes.

Table 5.

Phrases extracted (Spain).

	FREQUENCY	% CASES	TF • IDF
DRINKING ALCOHOL	107	1,81%	186,5
GOING TO THE BEAUTICIAN	34	0,57%	76,2
ENTERING AGAIN THE SQUARE	34	0,57%	76,2
KISSING	25	0,42%	59,4
HAVING THE FIRST PARTY WITH ALL FAMILY	22	0,37%	53,5
ORGANIZING A PARTY	21	0,35%	51,5
MEETING AGAIN FRIENDS	20	0,34%	49,4
VISITING FRIENDS AGAIN	18	0,30%	45,3
ALL WINE ABLE TO DRINK	11	0,19%	30
DOING A TATOO WITH ALL FRIENDS TOGETHER	8	0,13%	23
MEETING AGAIN MOM	7	0,12%	20,5
FINALLY ENDING THE SOCIAL DISTANCING	6	0,10%	18
MEETING AGAIN MY FAMILY	6	0,10%	18
HAVING A STEACK AND WINE WITH MY FRIENDS	4	0,07%	12,7
GO TO CINEMA	3	0,05%	9,9
GO TO FOOTBALL STADIUM	3	0,05%	9,9
WALKING DRINKING A BEER	3	0,05%	9,9
THE FIRST NIGHTOUT	3	0,05%	9,9

These phrases can be grouped according to the different behavior they describe (Table 6 ).

5. Discussion and conclusion

The discussion is organized into three sections which logically and chronologically follow the research questions stated earlier. We begin by exploring the content of consumers’ intentions about their life after the lockdown in order to address the extent to which consumers display high risk-aversion tendency (RQ1 and RQ2). Then, we focus on the comparison between the three countries to analyze if relevant differences in the countermeasures taken by the respective governments and the cultural differences between the three countries led consumers to develop different expectations about their consumption habits after the lockdown (RQ2).

The current research collected and analyzed a large volume of online communication on Twitter sharing the hashtag “#TheFirstThingIDoAfterTheLockdown” in order to listen into the thoughts in consumers’ minds during the restrictive measures adopted by governments to cope with the pandemic. By doing so, the present research allows to detect an unbiased set of expectations (Divakaran, Palmer, Søndergaard, & Matkovskyy, 2017) the consumers developed about their future behaviors after the lockdown.

A pattern of results emerged showing that consumers are expressing their willingness to feed their personal and social needs after the lockdown, suggesting that the emergency period has not drastically lowered consumers’ intention to consume as well as the lack of resources they have available to fulfill their desires.

When looking at the specific categories of expectations that emerge from the Tweets observed in the present research, results across the three countries involved in the study suggest that the top-of-mind expectation by consumers deals with escaping from home and enjoying freedom, either by having a good meal (Table 2) or drinking alcoholic beverages (Table 6), or even by travelling or having a new tattoo (Table 4). This finding suggests that the high levels of risk individuals were exposed to during the pandemic (Roggeveen & Sethuraman, 2020) are not expected to carry over the subsequent periods after the lockdown; instead, findings suggest that consumers are willing to restore their levels of consumption especially on those activities which positively contribute to the sense of escapism connected to consumption.

Consistently with this finding, another category of expectations which clearly emerged from our results concerns individuals’ intention to go back to their previous life, for instance by catching up with friends (Table 4), getting married (Table 2), or simply participating in a party (Table 6). This is to say, consumers are developing expectations which are mainly focused on social activities and which revolve around experiences rather than material possession. This finding aligns with Forbes (2017) who reported a higher tendency towards utilitarian consumption in the immediate aftermath of a natural disaster (i.e. an earthquake), followed by a steep increase in hedonic and experiential consumption.

Despite the various limitations to grocery shopping imposed during the lockdown (Pantano et al., 2020), it is worth noticing that no tweet in all our dataset explicitly refers to the queues out of the stores, or to the unavailability of multiple items such as toilet paper or hand sanitizers. Rather, tweets dealing with shopping expectations emphasize the possibility to visit again one’s favorite stores (Table 2) as well as the freedom to go shopping anywhere one wishes to (Table 6). The hedonic side of shopping thereby results the main driver of consumers’ expectations about their shopping behavior after the pandemic emergency, while during the pandemic the utilitarian value prevailed. Previous research has clearly related escapism and the hedonic side of shopping (e.g. Scarpi, 2006), while the present research builds on this finding by suggesting that the hedonic side of shopping contributes to smoothing risk aversion tendencies which follow an emergency situation. In other words, the depravation, or huge limit, of hedonic aspects of the shopping experience (i.e., the social aspect traditionally involved) makes consumers feel more rewarded by a hedonic experience. Summarizing, these results describe how consumers compare the expectations about shopping experience after the pandemic to the perception of the same experience during the pandemic outbreak.

Results from the present research provide converging evidence in favour of the cultural differences in the three different countries. In particular, results portray the cultural differences during the time of emergency, and consumption behavioral intentions in terms of expectations and risks after the emergency (i.e., after lockdown). Indeed, 14 out of the top 15 activities mentioned by Spanish individuals – scoring lowest on individualism - refer to social activities (Table 6), whilst only 2 social activities emerge in the top 15 mentioned by UK individuals – scoring highest on individualism (Table 2), with Italy lying in between with 5 social activities mentioned in the top 15 (Table 4). With specific regards to consumption, it might be worth noticing that consumers from the most individualistic countries mentioned consumption of products and/or services which pertain mostly to the individual sphere such as, for instance, “smoking cigarettes”, “visiting my favorite shops” (UK, Table 2), or “going shopping anywhere I want”, “going to the gym”, “barber” (Italy, Table 4). Conversely, in the least individualistic country (i.e. Spain), only one consumption occasion refers to individual consumption (“going to the beautician”, Table 6). This finding provides support to Proposition 1 by suggesting that consumers belonging to more collectivistic cultures are more likely to develop behavioral intentions pertaining to social activities after an emergency period in comparison with consumers from individualistic cultures who are more likely to focus on activities which fulfil the individual sphere. Accordingly, retailers and service providers from these different countries are encouraged to carefully think about the different extent to which consumers are willing to engage in social versus individual activities to reward themselves after the lockdown period.

Our results also point out the different levels of risk-aversion displayed by individuals from the three different countries. Interestingly, only tweets from UK consumers reveal a considerable level of risk perception after the lockdown, acknowledging that “busiest roads should be closed”, the “fear of going outside” and the importance to “keep social distancing” (Table 2). This finding might sound to somehow contradict previous studies suggesting that high levels of uncertainty avoidance are typically associated with low levels of risk taking (Shah, 2012). Given that the UK rates as the country displaying the lowest uncertainty avoidance index, one might expect tweets from UK consumers to reflect a relatively lower proportion of concern toward the potential risks connected to the activities performed after the emergency period. Instead, results show that tweets from consumers from countries scoring higher on uncertainty avoidance display a lower concern about social distancing and the risk connected to a new spread of the pandemic: tweets from Italian consumers emphasize the desire to return to normal life “as nothing happened” (Table 4), and tweets from Spanish consumers reveal intentions to engage again in activities which might even involve high risks due to scarce distancing such as “going to the cinema” or to the “stadium” (Table 6). This apparent contradiction can be explained in the light of the basic clarification provided by Hofstede (2001, p. 148) according to which “uncertainty avoidance does not equal to risk avoidance”. Rather, individuals in uncertainty-avoiding cultures might paradoxically display higher risk-taking tendencies as a deliberate strategy to reduce the ambiguity introduced by external factors (Rieger et al., 2015). In this vein, the behavioral intentions revealed by tweets from Spanish and Italian consumers at the end of the emergency situation support the notion that consumers from uncertainty-avoiding cultures are more prone to take risks in order to cope with the ambiguity carried by new situations. Specifically, consistently with Proposition 2, consumers from Spain and Italy displayed a higher tendency to take risks by engaging in activities which might reveal to be dangerous for scarce social distancing in order to re-establish the situation as it used to be before the emergency period. In opposite, consumers from the UK appeared to be more at ease with the new situation introducing ambiguity in their lifestyles by incorporating it in their thoughts about the future

Accordingly, retailers and service providers from countries differing on uncertainty avoidance might carefully modulate their offering bearing in mind that consumers in uncertainty-avoiding cultures are more likely to accept and be satisfied with offerings which keep them rooted in the situation prior to the emergency.

5.1. Theoretical contributions

This paper contributes to the literature around three key areas. These areas are contributions to literature on consumer reactions to emergency situations, risk aversion after an emergency and the intercultural differences that can take place. Firstly, the present work addresses consumer reactions to emergency situations. Scholarly literature has investigated prior cases of consumer reactions to states of emergency and catastrophes by showing, for instance, the relative shift in consumers’ perceptions of shopping convenience and shopping behavior when natural disasters occur (Cameron & Shah, 2015). In this vein, the majority of extant studies has focused on panic buying behaviors (e.g. Hall et al., 2020) which is a likely consumption pattern observed during, or immediately before, an emergency. Consumers have been found to cope with the expected product scarcity by altering their search and purchase patterns (Hamilton et al., 2019). Noticeably, consumers’ intentions and behaviors after an emergency period have not received the same amount of scholarly attention as for the phases during the emergency. In this light, the present research contributes to this stream of literature by identifying the nature and intercultural differences of consumption goals following an emergency period and proposes a simple and scalable solution for businesses to anticipate them.

Secondly, this paper contributes to the literature on risk-aversion, specifically during a case of emergency (Aren and Hamamci, 2020, Bao et al., 2003, Mandrik and Bao, 2005, Matzler et al., 2008, Meroño-Cerdán et al., 2017, Sun, 2014). This was explored through the use of tweets that help explore the degree to which consumers were risk-averse. The findings suggested that consumers have not drastically lowered their intention to consume once lockdown is lifted. This is especially so with activities that aid their sense of escapism. This aligns with views that the pandemic will not substantially change people’s behaviors and intentions in the aftermath of the lockdown. This contributes in particular to the literature emerging on COVID-19 (Kirk and Rifkin, 2020, Pantano et al., 2020, Roggeveen and Sethuraman, 2020), but also on that of escapism in consumption. This will be of interest to marketing scholars who are not only exploring COVID-19 research but also those who study consumption within emergencies and its impact on risk. This is of particular interest now as due to the current pandemic this is very timely. There were no studies currently exploring the level of risk aversion after the pandemic. This study helps acknowledge this gap.

Thirdly, this study helps address how both expectations and risk aversion compare across three different countries: UK, Italy and Spain. This gives a unique perspective to compare how the different countries differ in their perceptions in the current pandemic. In this vein, the COVID-19 emergency represents an unprecedented emergency situation because of its global reach and duration, thereby providing scholars with the possibility to effectively compare intercultural differences in consumers’ reactions which would have been hardly observable for most of the emergency situations examined by prior studies which focused mainly on local disasters such as earthquakes and hurricanes. Our findings are particularly focused on the notion of individualism and collectivism when viewing the responses. As a consequence, implications for scholars in both consumer behavior and international marketing emerge. Indeed, our results can further contribute to the literature on standardization and adaptation for marketing messages to be aimed at consumers in these countries.

5.2. Practical implications

This research has multiple practical implications for practitioners. These practical implications are of note on two levels. Firstly, for retailers coming out of lockdown now, and secondly for retailers to prepare for potential future lockdowns in the event of future waves of infections or new emergencies nationally and internationally. Summarizing, retailers should: (i) should continue to advertise and promote during the emergency, (ii) promote according to expectations, and (iii) focus on creating quality experience for returning consumers.

(i) Retailers should continue to advertise and promote during the emergency

The first implication for practitioners based on this is that if consumer expectation and risk aversion hasn’t been lowered considerably, then this highlights even more so that brands should continue to advertise and promote where possibly during this time. This is in line with the long-term effects shown by the work of Binet and Field (2013). As such, an implication from the current study is that as consumers expect to resume their behavior then retailers should prepare for this by continuing to carry out advertising and other forms of promotion where possible. This may have different advice for larger organizations and SMEs. However, the advice would be do what they can with the resources they have.

(ii) Retailers should promote according to expectations

Our findings could offer inspiration for how marketers can promote their organisations to consumers. Firstly, this could influence practitioners on which areas to focus in their promotional activity i.e. experiential aspects of their offer or utilitarian ones. Based on this research, it may be of interest to highlight experiences of promotions (i.e. highlighting the experiential nature of their offer) in order to begin trying to meet the expectations of what consumers want to do after lockdown ends. This could be taking into account the needs of the different countries and promote accordingly.

(iii) Retailers should focus on creating quality experience for returning consumers

Related to the above point retailers should focus even more so on creating/preparing experiences for consumers. So, whilst the previous implication dealt with promoting the experience, this implication is to create an enjoyable experience for customers. This could be highlighted as a standard need for retailers, despite there being a pandemic or not. However, the results show post-lockdown there could be an increased need to create enjoyable experiences for consumers. This is especially as the expectation of such experience may be higher, and thus judged more discerningly. Retailers could focus on the experiential aspects of product/service offerings, but also in light of provisions put in place to make sure customers feel safe during the experience. As whilst the risk aversion may not be extreme, it is still important to ensure customers feel safe, as this could have impacts on the experience.

5.3. Limitations and future research

Despite the contribution to literature and implications for practitioners, some limitations should be taken into account. The first limitation is that this study just made use of one social media for data collection. As such it represents the views of only those who are users of Twitter. This understandably only represents one of the many social media user groups that exist. As such, future research could take into account other social media such as Instagram, Facebook etc. It is worth noting though, this does depend on the level of access that can be granted to researchers through these platforms. However, getting the experience of non-Twitter users could help build up a richer perspective of consumer perceptions. Future research could then compare amongst social media platform users, and also collect data from those who don’t use social media. This could be carried out in order to explore differences between those who use social media and those that don’t. Similarly, the data collection referred to a very specific hashtag (“#TheFirstThingIDoAfterTheLockdown”) as representative of a certain behavioural intention. However, it is worth noting that other tweets not included in the collection and analysis might have contributed to the conversation without specifying this particular hashtag. Thus, future studies adopting our methodology, might consider developing a machine learning algorithm able to also collect tweets containing “similar” hashtag, for instance by implementing the algorithm to automatically identify the similarities through Levenshtein distance (Levenshtein, 1965). This process would result in bigger databases even when related to a very limited period of time.

Building on from the previous point, another limitation is the use of just one form of data collection (social media posts). Whilst this is useful for building up an unbiased perspective of the consumers it may miss out on potentially rich descriptions offered by consumers through data collection methods such an interviews and focus groups. Moreover, studies adopting social media post analysis lie on the assumption of within-country homogeneity in Hofstede dimensions. That is to say, for example, it assumes that all people in Spain who are tweeting are more collectivist than people tweeting from the UK. However, literature has shown a considerable within-country variation to Hofstede scores which could be better addressed this by triangulating different methods. Again, this could also help introduce the opinions of those who don’t use social media.

The third limitation of is that the study collected data over one-time frame (between the 1st and the 15th of April 2020.) As such it is one snapshot in time of the consumer perceptions, while other lockdowns have been imposed for different timing in late 2020 and early 2021 (i.e., UK encountered a second national lockdown in November 2020 and from January 2021-to at least mid-Feburary 2021). An area for future research that would address this situation would be to carry out the study again after the COVID-19 outbreak ends. In this respect researchers could compare perceptions during lockdown to those post-COVID-19 outbreak. Collecting data at multiple points would also allow researchers to track how expectations and risk aversion changed over time. Similarly, due to the approaching mass vaccinations, future research could carry out the study during vaccine dissemination and compare accordingly. For example, exploring how consumers experience in one lockdown affects the expectation of the “new normal” lead due to the specific vaccine.

A fourth limitation is that only three countries were explored (two with a more collectivist culture and one with a more individualist culture). As the findings suggested that there was a difference between the countries that had collectivist or individualist cultures, this opens up the possibility of exploring this area further. As such two approaches could be adopted. One future area could simply include a greater range of countries to see how they compared in their responses. This would open up the findings for greater comparison. Another approach would be to focus on collecting tweets from both more collectivist and individualist countries. This could help explore the points raised in this study even further. Understanding how these cultures differed could not only affect how organisations could communicate with countries, but it could also inform how more national promotions by the respective governments were carried out.

Finally, the present methodology assumes that individuals sending tweets in a certain language and living in a certain country show a similar culture, representative of the country’s culture (representative of Hofstede’s dimensions). Moreover, the same country might further show diverse behaviour among individuals (i.e., when considering Northern and Southern Italy and the different historical colonizations impacting differently on citizens’ culture). Since Twitter does not provide indication on the nationality of the individuals for privacy issues, new studies involving also qualitative data from in-depth interviews, and quantitative surveys conducted in different areas of the same country would further corroborate the results, and achieve more a generalizable understanding of the inhabitants of the different countries.

Biographies

Eleonora Pantano, PhD, FHEA, MEng, is a Senior Lecturer (Associate Professor) of Marketing at University of Bristol. Her research activities mainly relate to the development of new customer solutions to improve retail analytics, strategies and management, with emphasis on the role of artificial intelligence, emotional analytics, and machine learning algorithms. Her papers have also consistently pushed the frontiers in the area of marketing and have become highly cited by academics in the field, appearing in Tourism Management, Journal of Business Research, Computers in Human Behavior and Psychology and Marketing among the others.

Constantinos-Vasilios Priporas, PhD, MCIM, FEMAB, is a Senior Lecturer in Marketing at Middlesex University Business School, UK. His research interests include consumer behavior and strategic marketing with main emphasis on tourism, retailing and food. He has published in several international academic journals and conferences, including Tourism Management, Journal of Travel Research, International Marketing Review, Journal of Business Research, International Journal of Contemporary Hospitality Management, Computers in Human Behavior. In addition, he co-authored a textbook on Technology and Innovation for Marketing and co-edited a book on Market Sensing. He is a member of several professional bodies and he is editorial board member of the Journal of Customer Behaviour and has acted as a guest editor, reviewer, and track chair in academic journals and conferences.

Luke Devereaux, PhD is a Lecturer of Marketing at Middlesex University London. His research interests include corporate identity, dynamic capabilities and social enterprise. His research has been published in International Studies of Management and Organisation and Journal of Business Research. Before joining academia, he worked as a journalist, editor and market researcher.

Gabriele Pizzi is an Associate Professor of Marketing at the Department of Management of the University of Bologna, Italy. His research interests deal with assortment management, the impact of innovative technologies on the retailing activity and the longitudinal analysis of customer satisfaction. His work has been published in international journals such as the Journal of Retailing, Journal of Business Research, Journal of Service Research and Journal of Retailing and Consumer Services, Journal of Interactive Marketing and others.