This article was published on 6 June 2021 and was replaced on 28 June 2021

Aim of study: This study aims to analyse how psychological factors related to the Covid-19 lockdown affected the frequency of wine consumption among Iberian consumers. Area of study: To achieve this goal, we used data collected from an online survey in Europe comprising 4489 observations from Portuguese and Spanish samples. Material and methods: Using an ordered probit model, we analysed the wine consumption frequency as a function of a set of explanatory variables related to psychological factors and also sociodemographic variables, motivation-related variables and consumption characterisation. Main results: For Spanish respondents the fear of isolation was a decisive factor in increasing the probability of a higher frequency of wine consumption. Meanwhile, in Portugal, the fear of an economic crisis was the psychological factor leading to a higher consumption frequency. Research highlights: In both countries, psychological factors influenced the frequency of wine consumption during the lockdown due to Covid-19. However, the impact of the Covid-19 crisis has been felt differently in Spain and Portugal. Differences can be observed in both psychological and behavioural attitudes that have influenced the frequency of wine consumption and could also indicate significant cultural differences.


Introduction
Since the Second World War global wine consumption has increased remarkably.This trend has been driven by increasing globalisation, the emergence of new producing countries and regions, greater openness and expansion of international trade, and a growing standardisation of the typology, habits and attitudes of consumers (Anderson & Pinilla, 2018).However, in the traditional producing countries of Southern Europe, namely France, Italy, Portugal and Spain, consumption has decreased notably, while that of other alcoholic beverages, particularly beer, has increased.In general, in Western countries, there has been some convergence in the consumption patterns of alcoholic beverages.There is abundant literature explaining these trends and analysing the determinants of alcoholic beverage consumption (Fogarty, 2010;Holmes & Anderson 2017;Anderson et al., 2018).However, the behaviour of alcoholic beverage consumers can also be affected by unexpected shocks, such as economic or financial crises (Bor et al., 2013;Garbinti et al., 2020), wildfires (Thach & Eyler, 2017), earthquakes (Forbes & Wilson, 2018) and threats or terrorist attacks (Gergaud et al., 2018).
The Covid-19 pandemic and the subsequent lockdown, the resulting containment and associated public and individual control measures is an entirely new and unknown phenomenon in modern economic and social history, with disruptive psychological and behavioural effects.Specifically, Covid-19 has generated extreme fear and anxiety.It has disrupted routines and people have developed "disproportional cognitive, affective, or behavioural responses to the objects and situations that they associate with the Covid-19 pandemic and severe deteriorations may occur in the physiological and psychological functionalities" (Arpaci et al., 2020, p: 2).
At the microeconomic level, the behaviour of wine consumers has usually been analysed using variables related to both subjective and objective wine attributes, consumer motivations, together with sociodemographic characteristics (e.g.Charters & Pettigrew, 2008;Gonçalves et al., 2020;Ribeiro et al., 2020).Observations of the determinants of wine consumption frequency during the Covid-19 lockdown suggest that variables related to psychological behaviour should be added to the hedonic demand function as explanatory variables (taste, health or relaxation).
In this context, this study aims to analyse how psychological factors related to the Covid-19 lockdown crisis have affected the wine consumption frequency of Iberian consumers.Consumer behaviour is heterogeneous, not only between social classes but also between countries (Bor et al., 2013;Garbinti et al., 2020).Therefore, one research question in this study is whether the impact of the lockdown  has been homogeneous between countries.The impact of Covid-19 was probably felt differently in Spain and Portugal.Compared to Spain, Portugal had a lower mortality rate and a less restrictive lockdown, which was certainly manifested in the psychological and behavioural attitudes concerning the frequency of wine consumption.
For this purpose, based on samples collected for Spain and Portugal, we have estimated an ordered probit model using a stepwise approach, to determine consumer behaviour (consumption frequency).To do this we used significant psychological variables and control variables related to the sociodemographic characteristics of respondents as well as wine-related variables, such as the reasons for drinking wine and purchasing and drinking habits.

Theoretical approach and hypothesis
Covid-19 has caused a global health crisis unprecedented in our time.The WHO (World Health Organization) recognised it as a pandemic on 11 March 2020.From that moment, many countries around the world rushed to decree a general State of Emergency to confine the population and restrict numerous non-essential activities.The Spanish government declared the state of alarm on 14 March 2020, and, after several extensions, it remained in force until 21 June.Portugal meanwhile declared the state of alarm on 18 March and it lasted until 2 May.Since the end of the state of national emergency, local lockdowns have been imposed in some municipalities or districts because of the emergence of serious outbreaks of the infection in a second wave during the summer months.
This situation has forced millions of citizens to stay in their homes for weeks and has caused the closure, at least temporarily, of many workplaces.Confinement reduces social contacts and limits both work-related and personal activities.This has essential cognitive, affective and behavioural consequences (Arpaci et al., 2020).Isolation generates feelings of loneliness; the evolution of the crisis causes negative thoughts and behaviours that are inexorably modified.Fear of contagion and anxiety about the future, among other factors, have a negative psychological impact on many citizens (Atalan, 2020;Brodeur et al., 2020;Chaix et al., 2020;Sibley et al., 2020;Van Hoof, 2020).
Being a catastrophe, this lockdown constitutes an extraordinary economic and natural social experiment for many social sciences, since it allows us to study the changes that occur in the lives of the people affected.A particularly exciting aspect is their behaviour as consumers, especially of products that are controversial due to the consequences of their excessive consumption.The most representative case is that of beverages with an addictive alcoholic component (Bentzen et al., 1999), particularly dangerous in the case of young people (Gil & Molina, 2009).Additionally, the interest in studying these kinds of products is more significant due to the limitations imposed during confinement, and even afterwards, on the HORECA channel (on-trade and off-trade), whose establishments have been closed for weeks.This channel is highly important for virtually all alcoholic beverages since it represents an important part of their sales.The HORECA channel represents around 30% of the total volume of wine sales -31% in Portugal (2017) and 29% in Spain (2018)-due to the habit of eating out and tourism.
In terms of added value, they represent between 59% in Portugal (2017) and66% in Spain (2018).Therefore, the closure of catering establishments has had a significant impact on wine sales in this channel (ICEX, 2019;Interprofesional del Vino en España, 2020).For consumers, it represents a moment of social consumption.Within these drinks, wine deserves special attention since its consumption has a specific cultural dimension given that it is closely linked to gastronomy, although in recent decades its hedonic function has become increasingly important (Platania et al., 2016).Diet frameworks, such as the "Mediterranean alcohol drinking pattern" consider the wine consumption patterns identified as being healthy (Gunay & Baker, 2011;Bazal et al., 2019).However, this does not mean that situations of abuse do not occur (Goldsmith & d'Hauteville, 1998).
In general, consumption is conditioned by psychological and cultural factors (McGuire, 1976;Callwood, 2013).Its pattern responds to motivations, perceptions, learning, beliefs and attitudes of consumers as well as customs and social guidelines.Often, the two types of factors are linked (Silva et al., 2017).
In the case of wine, a positive association has been found between its consumption and psychological factors such as sensory gratification, social bonding (Dunbar et al., 2017), pleasant experience (Peele & Grant, 1999), the creation of a positive or relaxed mood and the ability to cope with difficulties and to adapt (Sayette, 1993;Cooper, 1994;Stritzke et al., 1996;Foster et al., 2014).In the same way, Baum-Baicker (1985) identified five areas of psychological benefits derived from the moderate consumption of alcoholic beverages: (a) reduction of stress; (b) improved mood; (c) higher cognitive performance; (d) reduction of clinical symptoms and the incidence of depression (Gea et al., 2013); and (e) better functioning in the elderly.These results are explained by the fact that alcohol releases dopamine and serotonin, neurotransmitters associated with the experience of pleasure (Lovinger, 1997;Banerjee, 2014).
With this background, the first hypotheses that this study will test is the impact of psychological variables on the consumption frequency during the lockdown .In other words, whether the foreseeable increase in feelings of anguish, anxiety and worry during confinement caused changes in consumption patterns, in particular its frequency, since this can be an indicator of increased moments of discomfort.Frequency is, together with intensity, one of the two parameters that define the demand and consumption patterns of any good.In the case of alcoholic beverages, they cannot be treated as substitutes for one another, since each has different psychological and cultural implications (Heckley et al., 2017), and they all have significant but different consequences on health (Attard et al., 2019).Frequency is especially relevant for its ability to create habits that influence purchasing and consumption decisions in the medium and long term (Ji & Wood, 2007).Our hypothesis is: H1: Psychological variables influenced the consumption frequency during the lockdown Cultural factors have a significant national component, although interdependencies between countries are increasing.Thus, for example, research on the effects of the 2008 crisis on the consumption of some substances, including alcoholic beverages (Dom et al., 2016) revealed significant differences between countries of the European Union.However, Leifman (2001) and Smith & Mitry (2007) found, at least in Europe, a relative cultural convergence between European countries regarding wine consumption patterns.Cultural patterns are related to consumption occasions in which the same consumer adopts different behaviours.For example, Hall & Lockshin (2000) established three models of consumption (consumption at parties, at work meals, and private consumption); the first two were eliminated during the lockdown.
In the case of the Covid-19 crisis, all countries have been affected, although not with the same intensity.Similarly, the quality of the public management of the crisis has not been the same either and this also affects the psychological state of the citizens.A particularly interesting case is that of Spain and Portugal, neighbouring countries of Iberia and culturally very close, although with differences in certain social behaviours and the evolution and incidence of the Covid-19 pandemic.Compared to Spain, Portugal had a lower mortality rate and a less restrictive lockdown (OECD, 2020).Our second hypothesis to test will be: H2: The impact of Covid-19 on wine consumption was not homogeneous in Portugal and Spain with differences both in psychological and behavioural attitudes and in the frequency of consumption.

The survey, sample, and data collection
The survey was based on previous research (e.g., Bruwer & Buller, 2013;Martinez-Carrasco et al., 2006;Thach & Olsen, 2015) and on the methodology used by the European Social Survey (https://www.europeansocialsurvey.org).To assess attitudes, beliefs and behavioural patterns, the survey was designed by a group of researchers from the board of the European Association of Wine Economists, EuAWE (https://www.euawe.com/about-us/) -and from the INSEEC (School of Business and Economics of University of Bordeaux).Subsequently, the survey was pre-tested by conducting an online survey in European countries, namely France, Italy, Portugal and Spain between 17 April and 10 May 2020, through the SurveyMonkey platform and using the exponential discriminatory snowball sampling technique 1 .This method allows the urgency in the data collection process to be addressed.However, one possible limitation is that it can generate sample bias, failing to cover all socio-demographic layers of the target population, which tends to reduce as sample size increases (Taherdoost, 2016).The main goal of the study was taken into consideration and was inherent in the design of the survey, both in the content of questions as well as in the respective scales.Thus, the survey includes four groups of questions related to purchasing and consumption patterns, the consumption situation, Likert scales to assess feelings and emotions during the lockdown and sociodemographic information2 .The preliminary results from the European study, released on 30 April 2021, indicate that, in general, the frequency of wine consumption increased during the lockdown period for more than 35% of respondents.However, the expense of alcoholic beverages decreased, with the average price of wine being lower when compared to normal conditions.The frequency of wine consumption from personal cellars increased as did consumption in a non-social context and especially drinking alone.When analysing the determinants of the increased drinking frequency, the anxiety generated by the Covid-19 crisis played a relevant role, with respondents in all countries expressing a powerful fear about the economic consequences of the crisis, as various factors of precariousness may arise or may be intensified.
For a better understanding of consumer behaviour in this time of crisis, we need to conduct additional research on how psychological factors affect the consumption of alcoholic beverages, particularly wine.Therefore, the goal of this research is to study how psychological variables influenced consumption frequency during the lockdown period in the Iberian market, given a set of control variables related to consumption motivations, consumption behaviour, and socioeconomic characteristics.The sample comprises 1,940 respondents in Portugal and 2,549 in Spain, resulting in 4,489 valid responses (collected from 17 April to 10 May 2020).Table S1 [suppl] provides the socioeconomic characteristics of the respondents and the wine and Covid-19 related variables in both countries, as well as the results of the different means tests.
The majority of respondents are male (61.9% in Portugal and 59.6% in Spain), living in urban areas (50.1% and 62.3%, respectively), and employed in the services sector.In Portugal, unemployment is more prevalent, with 13.4% of respondents unemployed compared to 4.4% in Spain3 .In the Portuguese sample, 30.9% are in the 41-50 age group, followed by 22% in the 51-60 age group.In Spain, 28.7% and 28.6% of the respondents belong to the 41-50 and 51-60 age categories, respectively.More than half of the sample in Portugal claim to be "coping on present income" (54.3%), while in Spain, the majority of respondents claim to be "living comfortably" (64.6%).Concerning the wine-related variables, the Portuguese show a higher degree of consumption with online interactions, with 34.5% (22.5% in Spain) having a wine app on their smartphones, 20.5% (12.5% in Spain) gaining improved wine knowledge through online content during the lockdown and 54.0% (40.3% in Spain) receiving online offers from winemakers to buy their wine online.In both countries, and during the lockdown, personal cellars were the second most important place for wine purchase after supermarkets, followed by the online channel.Regarding the motivations for wine consumption, accompanying food and wine tasting are the most common reasons for drinking wine, followed by socialising with friends.Of the Portuguese respondents, 73.8% stated that during the lockdown they organised digital gatherings to have a drink with family or friends almost daily, 12.5% at least once a week, 12.9% rarely, and less than 1% did not organise a digital gathering.On the other hand, 48% of the Spanish respondents organised such meetings daily, 12.6% at least once a week, 35.6% rarely and 2.5% never.With respect to the psychological variables related to the Covid-19 crisis, in both countries, the respondents feared the virus and the economic crisis as a consequence of the pandemic.Nevertheless, they also agreed that this period provided an opportunity for positive initiatives.However, the fear of isolation is more prevalent in Spain, possibly due to the more widespread adverse effects of Covid-19 in that country 4 .
Figure 1 displays the observed values of wine consumption frequency during the lockdown by country.In both countries, almost half of the sample reports a lower frequency of consumption when compared to normal circumstances, before Covid-19, this being more prevalent in Portugal.Less than 20% claim that they consumed Wine consumption frequency during lockdown in the Iberian markets wine as usual and more than 30% said that they consumed wine more frequently than usual.

Model
The measure of wine consumption frequency is a categorical, ordinal variable.Therefore, the appropriate econometric model to represent the causal relationship, as opposed to a relationship that simply captures statistical associations, is the ordered probit model.This model identifies statistically significant relationships between explanatory variables and a dependent variable, accounting for unequal differences between the categories in this variable.
The ordered probit model, proposed by McKelvey & Zavoina (1975), is built around a latent regression, specified as follows: where i is the observation, y* is the unobserved N×1 dependent variable, β' is the vector of the K×1 estimated parameters or unknown marginal utilities, x is the of N×K covarites that are assumed to be independent of ε, and ε (N×1) is the error term including unobservable factors and assumed to be normally distributed across observations (with zero mean and unit variance).The probabilities underlying this model are: 5 The probability of spirits and beer was, respectively, estimated through an ordered probit model (with the dependent variable being the lockdown period consumption frequency and the explanatory variable the frequency of consumption in the pre-Covid period, i.e. consumption in normal situations, during the last year, 2019, before Covid).Only the statistically significant variables remain in the analysis.

𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃[𝑦𝑦
where Φ(•) denotes the cumulative distribution function, and μ j > μ j-1 .Also, including a constant term will require μ 0 = 0.The likelihood function for the estimation of the model parameters is based on the implied probabilities.The parameters estimated have no behavioural meaning, but the partial effects are particularly meaningful and given by: where φ(•) is the density function.
The marginal effect for a dummy variable (D) would be obtained by a difference of probabilities, rather than a derivative, as follows: where γ is the coefficient of D.
Given the empirical specification of the model used in this study and to solve the research questions, we adopted the following process.First, in this type of model endogeneity is not a common issue since decisions regarding the consumption frequency of wine, beer and spirits can be simultaneous and interdependent.This can give rise to endogeneity with at least one of the explanatory variables (frequency of consumption of beer and spirits) which is determined simultaneously with the explained variables and correlated with the error term (Woldridge, 2002).Therefore, we have addressed the issue by including as instrumental variables the probability of consumption of spirits and beer in the lockdown period, and the frequency of consumption in the pre-Covid period 5 .After assessing potential endogeneity in the explanatory variables, econometric practices recommend a stepwise regression approach when using a considerable number of explanatory variables.This process has the advantage of selecting only the statistically significant variables, eliminating non-significant outcomes, thus improving the statistical quality and the reading of the model.We performed a backward selection for the complete ordered probit regressions, with the decision level being significant at 10%.To assess potential structural differences between countries, we estimated the ordered probit regression model for both countries (jointly) and the regression re-run by adding the set of explanatory variables multiplied by a country factor (dummy variable).We also performed a likelihood ratio test with the nested and global model.This allowed us to identify structural (partial regression estimated parameters) differences between the country data.Therefore, we proceeded to estimate the ordered probit model, following a stepwise approach for Portugal and Spain separately.In the sequence of this procedure, the model estimated is described in Eq. ( 8), where a set of explanatory variables, including related psychological factors and consumer characteristics affect the probability of changes in consumption frequency. (8)

Variables
Table 1 includes the list of statistically significant variables used in the final econometric models.The dependent or explained variable is the frequency of wine consumption during the lockdown.Therefore, the "Lockwine" variable is ordinal, with three categories, evaluating potential changes, when compared to the pre-lockdown period (i.e., normal consumption): i) less frequent; ii) as usual; iii) more frequent.
We expect a series of factors to affect wine consumption frequency during the lockdown period.First, we are interested in the effects of psychological factors on

Frequency of wine consumption during the lockdown
Ordinal variable (1, less frequent; 2, as usual; 3, more frequent)

Isolation
The measure of the sense of isolation QCAL scale (5-point Likert scale) unified by a factor analysis which grouped 3 items into one factor.

Motiv_taste
Enjoying the taste as a factor of wine consumption Dummy variable (0, not relevant; 1, relevant)

Motiv_health
Health issues as a driver of wine consumption Dummy (0, not relevant; 1, relevant)

Motiv_relax
Wine consumption as a factor of relaxation and stress relief Dummy variable (0, not relevant; 1, relevant)

Lockonline Use of online channel during the lockdown to buy wine
Dummy variable (0, not used; 1, used)

Normwine
The measure of consumption frequency in the period pre-Covid Likert scale (1, never; 2, less than once a month; 3, at least once a month; 4, at least once a week; 5, daily)

Lockspirits
The measure of consumption frequency of spirits during the lockdown Estimated probability of consumption Wine consumption frequency during lockdown in the Iberian markets wine consumption, which are included to analyse how consumers perceive the emerging crisis, the insecurity, and changes to daily life.Crises affect consumer behaviours.
Crises spread poverty (Dobrovolskij & Stukas, 2013), increase stress (Mucci et al., 2016) and alter the well-being of populations (Van Hal, 2015).All crises change eating habits (Di Renzo et al., 2020).There is evidence that in the immediate aftermath of a crisis, consumers purchase higher levels of utilitarian products, but there is no evidence of increased consumption of hedonic products (Forbes, 2017).
In the case of Covid-19, Bracale & Vaccaro (2020) found an increase in Italians´ consumption of some types of food linked with their symbolic value and a tendency to carry on at home certain external socialization habits.Therefore, different responses can be expected depending on the value consumers give to a product, and wine is a good example.Thus, Likert scales measure the "fear of crisis", i.e., the expectations of future economic consequences of the Covid-19 crisis.Additionally, the lockdown fostered a sense of social isolation, which is expected to influence consumer behaviour.To account for this potential effect, the variable "isolation" was included.
Second, a group of control variables, including sociodemographic variables, consumption motivation variables, and consumption characterisation was used.The first subgroup includes the age and professional situation of the respondent.In order to better understand the results, a comparison between the survey sample results and the population of the countries should be made, as similarities and differences occur due to the sampling technique employed and the results should be read according to the sample.For example, it is worth mentioning that in both Portugal and Spain, the most populated age group is the 40-49 years segment (18.8% and 19.0% of the adult population) which is also the most represented age group in this sample (29.6% of the respondents).Additionally, it is noteworthy that in our sample, the unemployment rate is 8.3% (4.35% of Spanish respondents and 13.4% of Portuguese respondents), while the national unemployment rate in Spain is 14.1% and in Portugal, it is 6.5%.Moreover, the occupation of the respondents also diverges from that of the national populations.In Portugal and Spain, agricultural workers account for 5.7% and 4.0% of the workforce, while in our sample this rate is higher (19.3% and 15.6% of the respondents).The second subgroup, the reason-related variables, comprises a set of variables to assess the three main possible motivations for consuming wine: taste, as wine consumers may favour the taste of wine when compared to other substitute beverages; relaxation and stress relief, to assess the influence on the consumption of wine; and health, which has been a major concern during Covid-19 pandemic (Rhum & Black, 2002;Agrawal et al., 2008;Barrena & Sánchez, 2009;Helble & Sato, 2011;Moran & Saliba, 2012;Thach & Chang, 2015, 2016;Chang et al., 2016;Castriota, 2020).The final subgroup of variables included in this study is that of consumption characterisation.The habit of consuming alcoholic beverages (wine, beer or spirits) is included to assess how pre-lockdown consumption habits influenced changes in the consumption frequency during the lockdown period.A dummy variable for "online" buying is also included, to assess whether the lockdown has changed the use of the purchase channel.With specialised wine stores, wineries, cellars and other traditional channels closed, the online channel may have facilitated product availability, thus influencing wine consumption behaviour.

Results
Table 2 includes the final estimation results for Spain and Portugal.The results of the estimations suggest that there are similarities and disparities between Spain and Portugal.It should be noted that the variables that are statistically significant for both countries (Age, Motiv_taste, Motiv_relax and Lockonline) have the same direction effect, which allows us to conclude that the changes in consumption patterns are determined by similar factors.This appears to be in line with the analysis previously presented.However, there are some disparities in the key factors influencing changes in consumption behaviour.
Given the nature of the estimated model, the ordered probit coefficient results do not provide meaningful interpretations, apart from the significance and direction (sign) of the effects.Thus, we estimated the ordered probit marginal effects to show how wine consumption frequency probabilities change with a variation in the explanatory variables.Table 3 reports the results for the countries analysed, highlighting both differences and similarities between these two markets.
The analysis of the marginal effects allows us to corroborate the argument that the impacts of the determinants of wine consumption changes have similarities for both countries.It is noticeable that the direction of the marginal effects remains the same for both countries across all variables, for all outcomes.
The psychological factors included in this study, "feeling of isolation" and the "fear of a crisis" display different results in terms of significance.These results are different for each country, although both variables show the significance of the sample as a whole.The feeling of isolation, which is significant for Spain and the global sample, show that the higher this feeling, the higher the probability of increased frequency of wine consumption.A consumer with a greater sense of isolation is more likely to have increased the frequency of wine consumption during the lockdown.The different lengths of the lockdown in Spain (100 days) and Portugal (45 days) can explain this feeling.

A R
Additionally, for Portugal and the whole sample, we can observe that the fear of a crisis displays the same effect.i.e., higher concerns about a crisis result in the higher probability to have increased consumption frequency.The results of the psychological factors show that insecurity, uncertainty and the threat to regular daily life had a positive impact on the probability of a higher frequency of wine consumption during the lockdown, which coincides with the interpretations of the frequency of consumption during the lockdown.Consumers have responded emotionally to the threat of Covid-19, which is linked to changes in consumption patterns.The high level of threat and uncertainty is positively linked to a higher frequency of wine consumption.
The sociodemographic variables considered in this study are age and the consumer's professional situation.The age of consumers is significant in both countries, and the marginal effects allow us to conclude that older consumers tended to be less likely to increase their frequency of wine consumption during the lockdown.Meanwhile, the probability that younger consumers maintained or increased their frequency of consumption during this period was higher.
The marginal effects of employment show that, on average, employed people are more likely than the unemployed to say that their wine frequency stayed constant during the lockdown or increased with the lockdown (in Portugal) and less likely to say their wine consumption decreased.For Portugal, it is possible to conclude that being employed improved the probability of increasing the frequency of consumption.
The control variables included in the estimation, i.e., the motivations and the consumption characteristics, display exciting results.The motivations for consuming wine are related to the consumers' loyalty to a product.We can observe that consumers driven by "taste" are more likely to increase the frequency of consumption which highlights the differentiation of wine when compared to other substitute beverages."Health" concerns, however, tended to reduce the frequency of consumption in Portugal, i.e., health concerns decrease the probability of a consumer to change to a higher frequency.This is especially interesting in the Covid-19 context, where we would expect health concerns to change consumer behaviour and patterns.In Portugal, this result is consequential in the same way as the variable Fear Covid Crisis.
The "relaxation" variable as a motivation has the expected result.For both countries, it shows that the higher a consumer values relaxation as a reason to drink wine, the more likely they are to increase the frequency of consumption.This result may have two interpretations.Wine is a good that generates a feeling of relaxation among consumers and the lockdown experience with social distancing, isolation and feelings of uncertainty drove consumers to increase the search for relaxing factors, such as drinking wine.
For the consumption characterisation variables, it is noteworthy that the "online channels" were indeed crucial for the demand for wine.Consumers who opted to buy wine online are more likely to have increased their frequency of consumption (those who do not opt to buy Wine consumption frequency during lockdown in the Iberian markets online have a higher probability to have reduced the frequency of consumption).The "previous consumption" behaviour effects are as expected, even though the significance is only found for Spain and the global sample.Thus, the higher frequency of spirit consumption (Spain and the whole sample) is linked to lower probabilities of positive changes in the frequency of wine consumption during the lockdown.Wine consumption habits in the pre-Covid period also display the expected outcomes (for Spain and the whole sample).A consumer who has a daily consumption habit is less likely to have experienced an increase in consumption frequency during lockdown but, inversely, consumers with a lower frequency of consumption (less than once a month) are more likely to have increased their frequency of wine consumption.This behaviour indicates that wine is a product that has attracted non-consumers and lower frequency consumers during the lockdown.This result is in line with the previous analysis of the frequency of consumption which found that the frequency of consumption increased for 34.5% of the respondents, whereas the majority of consumers had not increased their frequency, since they were already regular/daily consumers.

Discussion
Covid-19 is a very recent phenomenon and we need to analyse its effects in different fields.We know very little about its impacts on alcohol consumption, specifically wine, indicating the need for research on the topic.This study aims to contribute to providing a better understanding of the behaviour of wine consumers during the lockdown, with a focus on the effects of psychological and cultural factors on wine consumption frequency.
The psychological effects of the lockdown on people's consumption behaviour is set to become a growing topic of research.The comparison with the consequences of other natural disasters is useful, but we must also bear in mind that the lockdown has very different characteristics, such as the absence of supply problems and a different perception of its impact on the risk of human lives, since it is not an instantaneous phenomenon but rather one that is prolonged over a long period of time and without the violence characterising other shocks.
In any case, the severity of the impact on lifestyle of disasters such as September 11, 2001, tsunamis or earthquakes is expected to be greater than a period of lockdown.The former imply an extreme risk to life, while in the case of Covid, the effects are related more to feelings or isolation during the lockdown.
Using data from Portugal and Spain, we have estimated an ordered probit model to investigate how wine consumption frequency was influenced by psychological factors and whether there are differences between these two countries.The econometric results and the marginal effects show that the psychological factors have affected the frequency of wine consumption, thus verifying our first hypothesis.However, these psychological factors have had a distinct impact in both countries.The fear of isolation increased wine consumption frequency for the Spanish sample and, in Portugal, the same effect was found as a result of the fear of a crisis.The feeling for consumers is similar.The different lockdown lengths of the two countries and the situation of fear related to the number of people infected reflect the same final result; anxiety and disruption in people's routines increase their wine consumption frequency.Thus, our second hypothesis has also been verified.
The impact of Covid-19 was probably felt differently in Portugal and Spain (lower mortality rates and different restrictive lockdown measures) which is certainly reflected in the different results for wine consumption frequency.
Age is a determining variable that should be included in future research as it requires further explanation.Two paths can explain the negative relationship between age and the increase of wine frequency consumption.First, older people did not increase their frequency of wine consumption as it was already high.But it is necessary to include the absolute variable of the quantity of consumed wine.And second, it constituted an opportunity to identify new possibilities to increase the wine frequency consumption of younger people.
Employment is decisive for the Portuguese people and the whole sample.This result must be related to consumers' incomes and the probability to increase the consumption frequency.
The variables that motivate wine consumption are related to hedonic demand functions such as taste or relaxation.The Iberian lifestyle forms part of the Mediterranean alcohol drinking pattern, and the relevance of health concerns must also be related to lifestyle or the Covid-related fear.This question remains open for future research.
However, one of the most interesting results of this study is related to the use of online media to buy wine.The growth of this channel is here to stay and wineries are developing different apps to sell wine by different consumer profiles.One remaining uncertainty is whether the amount sold in this channel exceeds other sale channels or not and whether the average sales price increases or not between the different vendors.
This result highlights the fact that the impact of the lockdown on consumer behaviour is not homogeneous between countries, so it is not possible to generalise the results of one market to other markets.As an example, it should be noted that 52.9% of Portuguese consumers have reduced the frequency of their wine consumption.This indicator in Spain is 44.7%.However, if we observe the impact of consumers who have increased the frequency of wine consumption, the situation is reversed.In Spain, 36% of consumers have increased the frequency of wine consumption, while in Portugal this percentage is 32.6%.Despite this, many similarities have been found between the two countries.
Our results can be compared with those of other studies on the impact on wine consumption of other natural disasters or of the same lockdown.
The impacts produced on consumption can vary depending on the kind of disaster.Thus, Foster (2017) reflects on consumption patterns after the 2011 Christchurch earthquake disaster.In this study, the immediate decrease in wine consumption (both red and white) represents around 27% in volume and 22% in value terms.The analysis of the extent of recovery of consumption over a longer period of time (4 weeks), reveals an increase of 10% in volume and and increase of between 6.6% and 10.8% in terms of value.These ratios indicate that short-term recoveries do not offset the post-disaster decline.However, it is Wine consumption frequency during lockdown in the Iberian markets necessary to point out that the situation derived from other natural disasters (earthquakes, tsunami, 9/S and others), implied a instantaneous greater vital risk as well as supply problems that have not occurred in the case of Covid-19.Braccale & Vacaro (2020) provided an analysis about food consumption in Italy during the Covid-19 emergency.
The situation is different depending on the type of drink considered.Thus, it is observed how the consumption of alcoholic beverages (beer and wine) has increased.The greatest increase in wine consumption occurred in Italian wine from controlled designation of origin (6.6%), beer (4.1%) and Italian wine from protected geographical indications (2.3%).
Our study has, however, some limitations.The sampling method, the snowball sampling technique, can produce biased results, since the survey sample may be different from the population of interest, namely wine consumers.This may lead to differences between the sociodemographic characteristics of the sample and the population of the countries (and therefore also the population of interest), which can be reflected in the results and their interpretation.As a consequence, our results should be taken with caution as they may not accurately reflect the behaviour of the population of interest.
Future research could complete these results as follows.First, by making a comparison with cross-country differences considered in this article.This could be further expanded with more data.In particular, a comparison with consumption behaviour after the lockdown period would be interesting.Second, future research could go further in terms of the explanatory variables, considering the differences in the price of wine before and during the lockdown.Third, we could exploit different sampling techniques.And, finally, the debate on whether wineries should formulate new strategies in the face of the increase in online shopping should be studied.
Another interesting question is the study of the recovery of wine consumption after lockdown and if there have been changes in wine consumers' profiles or in the purchasing channels, especially as a result of increased online shopping.

Figure 1 .
Figure 1.Wine consumption frequency during the lockdown, in %

Table 1 .
List of variables used in the econometric models

Table 2 .
List of variables used in the econometric models