1. Introduction
Currently, we are facing a worldwide health problem, the COVID-19 pandemic. Pandemics are widely recognized as outbreaks that arise from large-scale infectious diseases and significantly increase mortality and morbidity in several regions of the globe. With pandemics, other problems emerge at the political, social, and economic levels (
Madhav et al. 2017;
Ruiz et al. 2021). On 11 March 2020, the World Health Organization (WHO) declared COVID-19 as a global pandemic, with proportions never seen before (
Kickbusch et al. 2020;
Phillipou et al. 2020).
After the appearance of COVID-19 pandemic, multiple countries around the world implemented distinct measures to mitigate, restrict, prevent, interrupt, or retard its spread. Such measures included recommendations to the population on individual disinfection, the use of mask, and physical and social distancing. The outcome of these procedures, resulted in a decrease in social interactions, consequently, decreasing employment, which affected the world economy (
Phillipou et al. 2020). Confinements/lockdowns never seen before were introduced, these types of impositions, forced individuals to “stay at home”. Stores, places of worship, churches, schools were closed, and traveling was paused or severely restricted. During confinement, individuals could only leave their homes to care for dependent people, to work (only for jobs considered essential, others should use teleworking wherever possible), go to urgent health treatments, and buy essential products such as food and medicine (
Ruiz et al. 2021).
Furthermore, the World Health Organization recommended limitations on the visualization time of generic news about the COVID-19 pandemic, searching for reliable information from official media, regular sleep, and healthy eating routines, physical activity, maintaining the family routine, and pursuit of a healthy lifestyle (
Antunes et al. 2020;
World Health Organization 2020). To achieve this, the practice of physical activity plays an essential role, particularly when taking into account that habits in daily routines can lead to an increase in sedentary behaviors (
Chen et al. 2020). Moreover, it is widely recognized to reduce anxiety levels, even if physical activities are practiced at home instead of outdoors (
Anderson and Shivakumar 2013;
Hammami et al. 2020;
Stubbs et al. 2017).
Although there was a state of emergency decreed on 18 March and 6 November 2020, the government allowed individual outdoor physical activity so there was no total restriction of movement imposed on the population. These measures allowed short-distance walking and pet hygiene walking. However, group physical activity was prohibited, and gyms and sports clubs were closed (
Antunes et al. 2020).
That said, the objective of the present research is to analyze the habits observed in the perception of the general physical health condition of Portuguese food consumers in the COVID-19 pandemic. Moreover, it is also intended to study consumers’ eating behaviors in this pandemic.
3. Methodology
The main objective of this research is to analyze the habits observed in the perception of the general physical health condition of Portuguese food consumers in the pandemic, through the assessment of the impact of weight, physical exercise, and the positive/negative eating behaviors in the pandemic on health. To accomplish this, a quantitative methodology was used. This type of methodology allows for the validation of theories and relationships between variables, generalizes results, and replicates with different samples (
Chrysochou 2017;
Malhotra et al. 2017).
The sample used is composed of 741 valid responses (representative of Portuguese consumers over 18 years, according to the Portuguese population data collected by PRODATA 2019), collected through a questionnaire made available online to Portuguese food consumers between November 2020 and February 2021. The applied questionnaire was adapted from the study “COVID-19 and Retail Management of Groceries: Insights from a Comprehensive Consumer Survey” (
Wang et al. 2020b), which analyzed changes in consumer retail supermarket shopping behavior in the pandemic, and was answered by 2500 adults in the USA. This questionnaire was pre-validated (it was applied to 75 respondents, an Alpha Cronbach of 0.92 was obtained, showing a good internal consistency) before its dissemination. The questionnaire consists of 32 questions divided into six groups: (1) Health in general measured by two questions; (2) Weight measured also by two questions; (3) Physical exercise with a question about the time spent (in minutes) the practice of daily physical exercise in the pandemic; (4) positive eating behaviors comprising twelve questions about the consumption of vegetables, fruits, whole grains, leguminous, foods low in saturated fats and rich in monounsaturated and polyunsaturated fats, use of natural sweeteners, water consumption, cooked foods and low-fat dairy products; (5) negative eating behaviors comprising nine questions about the consumption of white sugar and artificial sweeteners, snacks, fast food, sweets, sugary drinks, fried foods, frozen and pre-cooked meals, processed meats and additional consumption of salt and (6) sociodemographic characteristics of the respondents with six questions, namely, gender, age, education level, marital status, employment situation and annual household income in the last year.
All responses to the questions were measured using a 7-point Likert scale, except for the question related to the daily physical exercise, which was measured in minutes, and questions related to sociodemographic characteristics. The group of questions of Health in General, to describe the general health condition in the pandemic a scale was used where 1 is very bad, and 7 is very good. As for the questions regarding the alteration of physical health in the pandemic, the respondents used a scale where 1 = it worsened a lot to 7 = it improved a lot. In the group of questions related to weight, the scaling was: 1 = much below the ideal weight and 7 = much above ideal weight, and 1 = I lost a lot of weight, and 7 = I gained a lot of weight in the pandemic. In the groups of questions related to positive and negative eating behaviors, the scale of 1 = totally disagree to 7 = totally agree was used.
For a general characterization of the sample, a statistical analysis of the sociodemographic variables was performed to describe the characteristics of the sample participants and an analysis of the mean and standard deviation of the answers to each of the questions that comprise the G1 to G5.
The analysis of the sociodemographic characteristics of the sample is summarized in
Table 1. It should be noted that 67.2% of the Portuguese consumers surveyed are women; out of which 68.4% are single and 26.6% are married or live in a de facto union. The average age is 29.9 years, with 35.6% of respondents aged less than 20 years (and over 18 years), 30.8% aged between 21 and 30 years; 23.3% between 31 and 50 years old, and 10.3% are over 50 years old (the maximum age being 77 years old). A total of 42.5% have completed secondary level education (12th grade) and 40.8% have completed a degree. Regarding the employment status, 38.5% are dependent employees, 43% are students, and 6.1% are self-employed. In terms of the annual household income registered in the previous year, 38.5% of the respondents have an annual income below EUR 20,000 and 34.3% between EUR 20,000 and EUR 39,999.
Table 1 contains the statistics of the sociodemographic characteristics of the Portuguese food consumers surveyed.
SPSS v.25 was also used to calculate the average and standard deviation of the questions that compose G1 to G5, and an analysis of the intra-group questions.
Considering the main objective of this research, five latent variables were created corresponding to the groups of questions from Q1 to Q5: general health, weight, physical exercise, positive eating behavior, and negative eating behavior. The relationships between the latent variables were established taking into account the main objective and the hypotheses to be tested.
Regarding the group of questions of general health (G1), weight (G2), physical exercise (G3), positive eating behavior (G4), and negative eating behavior (G5), the mean and standard deviation of the answers are found described in
Table 2.
In average terms, the Portuguese respondents described their physical health as reasonably good (5.39) in the pandemic and that it neither improved nor worsened. In terms of weight, in average terms, respondents consider their current weight, neither very high nor slightly below the ideal weight (4.50) and without significant changes, in average terms (4.39). Additionally, in average terms, the Portuguese respondents practice 25.51 min of physical exercise daily.
Regarding positive eating behaviors, the highest average values were: fruits (4.30), vegetables (4.24), foods rich in monounsaturated and polyunsaturated fats (4.04), water (4.81), of cooked, steamed, grilled, or poached foods (4.26) and leaner meats (4.37). We also concluded that the Portuguese respondents did not significantly improve positive eating behaviors in the pandemic, with the highest average value being (4.81). Regarding negative eating behaviors, the consumption of sugary drinks, such as soft drinks, fruit juices, and sports drinks (2.77), more frozen and/or prepackaged meals (2.85), more processed meats as sausages, bacon, ham (2.93), the addition of more salt to the food (2.31) and higher consumption of fast food (2.59), were, on average, the most disagreeable matters for the respondents. This means, that in the pandemic, on average, Portuguese consumers did not increase their negative eating behaviors.
Once the statistical analysis of the sample was performed, the theoretical model was estimated using Partial Least Squares (PLS) in the Smart PLS 3.0 software (SmartPLS GmbH, Bönningstedt, Germany), the results of which are shown in the next section.
4. Results
4.1. Results Presentation
A descriptive model was created to explain the above measures (
Figure 1). This model was estimated by Partial Least Squares (PLS) in the Smart PLS 3.0 software (
Ringle et al. 2015). PLS is a covariance-based structural model used to estimate complex interrelationships between latent and observed variables and, in recent years, has been increasingly applied in empirical studies (
Ringle et al. 2019,
2020). The use of PLS has the main advantages of allowing to estimate complex models with several latent variables, indicators, and structural paths without imposing the assumption of data distribution; allow testing the complexity of theoretical relationships defined by the supporting literature and, finally, increases the probability of identifying significant relationships between variables when in fact these relationships exist in the sample (
Ketchen 2013;
Leguina 2015;
Ringle et al. 2019).
To use the PLS model, it is necessary to validate the sample size, which must be at least equal to one of the following conditions: (1) ten times greater than the number of indicators or (2) ten times greater than the number of directed structural paths to a latent variable in the structural model (
Ketchen 2013;
Leguina 2015). As can be concluded, the size of our sample is 741 observations, which is more than ten times the number of indicators (26 indicators) and, thus, the sample size fulfills the conditions to be applied to the PLS method.
According to
Figure 1, the Theoretical Path Model contains 26 indicators (represented in the rectangles), that is, the answers to the 26 questions in
Table 2 and the five latent variables created—General Health; Negative Eating Behavior; Physical Exercise; Positive Eating Behavior and Weight (represented in circles). The relationship between the latent variables according to the hypotheses to be tested are: the latent variables Negative Eating Behavior; Physical Exercise; Positive Eating Behavior and Weight have a direct influence on the General Health latent variable.
Reliability and validity measures must be used to validate the estimated model, that is, instruments to mediate the relationship between the latent and observed variables of the model, which implies an analysis of the reliability of each latent variable at the indicator level and the convergent validity and discriminant. The validation in the model of this study is described in
Table 3, which contains the outer loadings of each indicator (question) used, reliability, and average variance extracted (AVE) of the latent variables.
Latent variables (General Health; Negative Eating Behavior; Physical Exercise; Positive Eating Behavior and Weight) have high external loadings (greater than 0.647). The reliability coefficients of latent variables must be greater than 0.70 (
Ringle et al. 2019). In this model, the values obtained for reliability coefficients of latent variables are higher than the reference value (General Health > 0.810; Negative Eating Behavior > 0.933; Physical Exercise > 1000; Positive Eating Behavior > 0.936; Weight > 0.841) and thus, reliability coefficients are “satisfactory to good”. Therefore, all latent variables are above the acceptable values for the outer loadings, reliability, and validity of the estimated model.
The model has acceptable validity and convergence measured by Cronbach’s Alpha (all the results of this indicator are greater than 0.700—reference value, except for Cronbach’s Alpha of the latent variable Weight, which is 0.695, which is only 0.005 below) and Average Variance Extracted—AVE (0.50 is the reference value for the AVE). We conclude that, all latent variables have a stroke above the reference value, that is, General Health > 0.682; Negative Eating Behavior > 0.608; Physical Exercise > 1000; Positive Eating Behavior > 0.552; Weight > 0.730).
The Fornell–Larcker criterion was also used, as a measure of Discriminant Validity. This criterion analyzes the cross-loadings that are indicators of the discriminant validity of latent variables. As we can see in
Table 4, each AVE of the latent variables is superior to all the square correlations of the latent variables (elements outside the diagonal), thus establishing the discriminant validity of each of the five latent variables.
In summary, we conclude that the model shown in
Figure 1 complies with the measures of reliability and validity, and discriminant validity, and, therefore, it is a valid model.
After validating the model, the next step is to estimate the model by PLS. To do this, it is necessary to check if the stopping criteria of the PLS algorithm is reached before the maximum number of iterations (repetition programming), which must be lower than that defined in the settings for the parameter of the PLS–SEM algorithm (in this case 300 iterations). The estimated model is shown in
Figure 2 and the algorithm converged after the 7th iteration for the parameter of the PLS–SEM algorithm (out of 300 iterations).
The validation of the predictive precision of the model is performed through the values of R Square (R
2) of the endogenous (dependent) latent variable, that is, General Health. According to
Ritchey (
2000), in social sciences (as is the case in this study), the reference values for R
2 from 0.04 to 0.16 are considered moderately weak and from 0.20 to 0.49 are considered moderately strong. According to this criterion, the PLS algorithm calculated R
2 moderately weak (0.109).
Path coefficients establish significant relationships between latent variables. As shown in
Figure 2, a 10% variation in the Weight variable has a negative impact of 6.3% on General Health; a 10% variation in the Physical Exercise variable has a 16.8% positive impact on General Health; a 10% variation in the Positive Eating Behavior variable has a 20.6% positive impact on General Health and a 10% variation on the Negative Eating Behavior variable has a 20.2% negative impact on General Health.
Once the path coefficients were calculated, a bootstrap analysis was performed to assess their statistical significance (95% confidence interval).
Table 5 shows the results of this significance test. We conclude that the latent variables negative eating behavior, physical exercise, and positive eating behavior are very significant (
p < 0.000), and the latent variable Weight is not significant to explain the general health of the Portuguese surveyed.
4.2. Results Discussion
We started the present research with a first assumption that fits H1, that the weight above the ideal has a negative impact on the assessment of the general physical health of the respondents in the COVID-19 pandemic. It was possible to ascertain that, although the relationship between the latent variables is negative as formulated in hypothesis 1, this relationship is not significant, so the first hypothesis is rejected. It should be noted that, according to
Table 2, the Portuguese respondents reported that, on average, they did not see significant habit variation in terms of weight in the pandemic and consider that the current weight is neither above nor below the ideal weight, which may justify the fact that, in general, weight does not have a significant impact on the assessment of physical health of the Portuguese respondents.
These data indicate that, although the respondents consider that the strategies pointed out in several studies (
Abbas and Kamel 2020;
Arora and Grey 2020;
Muscogiuri et al. 2020), as ways of overcoming contexts of stress, anxiety, or other problems associated with isolation, such as snacking behaviors, comfort eating and drinking, have brought with it a consequent increase in weight, this reality does not influence the perception of worse general physical health.
The second hypothesis under study focuses on the influence of daily physical exercise for a positive impact on the assessment of physical health in general, in the COVID-19 pandemic, this hypothesis is confirmed in the present research. In this regard, several studies point out that the increase in leisure-time physical activity is associated with better physical health, that the increase in sedentary time is associated with less physical health, and that physical activity decreased in the COVID-19 pandemic. The present research allows us to perceive that there is a positive relationship between the respondents, who associate better physical health in general with the practice of physical exercise. Thus, these considerations are in line with the findings of
Cheval et al. (
2021),
Chen et al. (
2020),
Dwyer et al. (
2020),
Matias et al. (
2020) and
Raiol (
2020), which demonstrate that physically active behaviors and a reduction in physical inactivity in the confinement caused by COVID-19 improve the perception of physical health. It should be noted that, on average, the Portuguese respondents practice 25.51 min of physical exercise daily, demonstrating their concern for their physical health in general.
The third hypothesis of the research focuses on determining a relationship between positive eating behaviors and the positive impacts that result from this in the perception of physical health in general, in the COVID-19 pandemic, by the respondents. This positive eating behavior such as consumption of vegetables, fruits, whole grains, legumes, foods low in saturated fats and rich in monounsaturated and polyunsaturated fats, use of natural sweeteners, water consumption, cooked foods, and low-fat dairy products have a positive impact on respondents’ assessment of physical health in general in the pandemic. Respondents who adopt these food consumption behaviors tend to demonstrate a more positive perception of their physical health.
These findings are in line with the contributions of
Arora and Grey (
2020),
Amekran and El Hangouche (
2021) and
Chen et al. (
2020), since a significant part of eating behaviors in a period of greater isolation may correspond to the incorporation of healthy habits described above. By identifying this relationship which reveals that respondents positively associate these eating behaviors with the perception of better physical health, we can confirm H3.
The fourth hypothesis under study focuses on determining a relationship between the adoption of negative eating behaviors and a more negative assessment of physical health in general, in the COVID-19 pandemic, amongst the respondents. These negative eating behaviors are measured by the consumption of white sugar and artificial sweeteners, snacks, fast food, sweets, sugary drinks, fried foods, frozen and pre-cooked meals, processed meats, and additional consumption of salt.
The results show that respondents who adopt these food consumption behaviors tend to show a more negative perception of their physical health. These findings are in line with the contributions of
Ammar et al. (
2020),
Durães et al. (
2021) and
Ribeiro-Silva et al. (
2020), so we can say that H4 is confirmed in the present study.
A summary of the hypotheses under study is aggregated in
Table 6, which describes which of the hypotheses are supported or not.
Although the analysis of the results obtained corroborates much of the relevant literature, the present study focuses only on physical health. Nonetheless, the importance of the balance between physical and mental health is important in times of normality. Given the current context, it would be important to determine the perception that these individuals have concerning this other component of their health. We consider this a gap to be addressed in future studies.
It is also recommended to carry out longitudinal studies, which allow us to perceive the evolution overtime on the perception of the physical health condition. Such should be done through the application of other indicators and widening the scope for physical health. It is also suggested to carry out comparative studies, to allow comparing perceptions of individuals from different cultural, geographic, and demographic backgrounds, under the effect of the same extreme global phenomenon as the pandemic COVID-19, with identical social and economic consequences.