Abstract
This study has two objectives: (1) to investigate the changes in the frequency, duration, and timing of solitary, family, and social meals in South Korea, and (2) to examine the effects of these meals on subjective well-being (i.e., life satisfaction). As for the latter objective, it was hypothesized that solitary and desynchronized meals would be associated with lower life satisfaction, whereas family, social, and synchronized meals would be linked with higher life satisfaction. Analyses of the time-use data between 1999 and 2014 revealed that family meals were being rapidly replaced by both solitary and social meals. Contrary to the thesis of temporal destructuration, however, the collective rhythms of eating became more pronounced in South Korea. Regarding the effects of the meals, eating together had positive effects on life satisfaction, whereas eating alone did not. The effects of solitary meals varied across individuals, based upon the level of voluntariness and scheduling. Moreover, eating together, especially with family members, protected individuals from the adverse impacts of a desynchronized eating rhythm.
What and how to eat is not just a physiological necessity but also a social construct associated with self-identity, values, world-view, and a sense of belonging (Fischler, 1980, 1988). For an entity to be considered edible, it must not be harmful and must be seen as morally desirable and acceptable to eat in the world that eaters have constructed (Durkheim, 1995 [1912]; Fischler, 1988). Moreover, the process of identifying and choosing food sources is collective and gives eaters feelings of belonging and distinction (Giacoman, 2016; Higgs, 2015). For example, enjoying a local flavor like kimchee promotes a sense of similitude and camaraderie, on the one hand, and distinction and otherness, on the other.
In addition to food choice, the way that food is consumed is highly social. Physical food itself is impossible to share with others (Simmel, 1910, in Symon, 1994). Thus, the appetite for food is a desire for something un-shareable. Paradoxically, however, no other desires brings humans together as much as hunger does. The consumption of food is experienced as a communion with others. When individuals get together to eat, according to Simmel (1910, in Symon, 1994), what they actually share is not the food itself but the most common needs that they all must eat and drink. As Symon (1994: 341) argues, ‘[T]he very universality of the basic needs inevitably gather people together more firmly and frequently than “higher or more spiritual motives.”’
Food expresses social relations, such as social hierarchy, inclusion and exclusion, boundaries and transactions across the boundaries (Douglas, 1972). For instance, ‘toilet meal,’ a practice once found in Japan, having a meal in the bathroom not to be seen as a social outcast, exemplifies the role that food plays in social relations. Whom to eat with signifies the position one holds within a group. Moreover, when and what to eat together can describe the quality of social relations. For example, having drinks together reveals social distance differently than having meals together does (Douglas, 1972). Likewise, having lunch together is different from having dinner together.
It is often argued that eating is not as socially significant as it once was and instead has become an increasingly individualized activity. Multiple factors have been pointed out for this change, some of which include changes in household composition, growing individualism, and diversification of temporal rhythm (Mestdag and Glorieux, 2009; Yates and Warde, 2017). Food has traditionally (or ideally) been provisioned and consumed by the family unit. However, due to the growing number of single-person households and the declining size of households, there are fewer persons to come together for a meal, and food provisioning at home becomes less economically practical. This encourages eating-out and the consumption of convenience foods, and blurs the distinction between home-cooked food and commercial food. Outsourced food reduces the incentives to gather together, as it is less expressive and more instrumental. Moreover, the shift toward a 24-hour society diversifies temporal rhythms and makes it more difficult for family members to coordinate their schedules for meals (Cheng et al., 2007; Southerton et al., 2012).
These changing eating patterns often become the basis to claim the crisis of the family, the most primary social group, a chief culture-transmitting agent, and even the basis of society. There is a growing concern that nothing may bring family members together, if food, a universal need, cannot. It is during family meals that members interact with one another, and younger ones are socialized into specific rules involving cooperation (Bossard, 1943; Fischler, 2011; Mestdag, 2005). Based on this, the decline of the family meal may indicate the decline of the family as a meaningful social group and even of society (Fischler, 1980). Empirical research, mostly conducted in European countries, appears to partially support the decline of the family meal (Cheng et al., 2007; Mestdag and Glorieux, 2009; Warde et al., 2007).
As discussed above, eating is a social activity that is associated with self-identity, a sense of belonging, social embeddedness, and subjective well-being. The current study examines the changes in the frequency, duration, and timing of eating in South Korea between 1999 and 2014, and the effects of various types of eating on life satisfaction. To do this, I categorized eating episodes reported in the Korean Time Use Survey (KTUS) by the presence of companions (solitary, family, and social meals) and by temporal conventionality (meals at conventional and unconventional times). Unlike most previous studies, the study analyzes not just the temporal duration but also the frequency of the various eating episodes. Moreover, it contributes to the existing body of research by examining the effects that various types of eating have on subjective well-being, which deserve more research attention than they have hitherto received.
Given the cultural influence on eating, patterns and changes in eating may be diverse across cultures. For example, where cultural collectivism is prevalent, solitary eating may result in greater shame, as seen by the incidence of toilet meals mentioned above. In South Korea, for instance, eating is not considered to be a private activity, rather it is seen as a means to achieve social solidarity. In fact, group members, such as family members, classmates, and co-workers, are expected to eat together. Unlike in more individualist cultures, individuals who eat alone in South Korea face social judgments, making them self-conscious about that behavior. This study in this sense expands the ongoing research by adding more diverse experiences, and helps find common and different features and trends across the world.
It has to be noted, however, that social norms are neither fixed, nor uniformly shared. Recent studies in Korea (Jo and Cheon, 2018; Lee et al., 2012; Park, 2017) show meaningful changes in eating norms, especially among youth, that challenge the stigma associated with solitary eating. The rapid growth in single-person households and a trend toward a 24-hour society are both associated with eating alone, and they are likely behind evolving eating norms. The discrepancy between the dominant and emerging norms associated with eating makes the Korean case particularly interesting, so studying the effects of de-socialized eating on subjective well-being in this cultural context is especially important and valuable.
Changes in eating behavior
Changes in eating habits may affect the frequency, duration, and timing of meals. Previous studies have reported a (slight) decrease in the duration of a meal in many European countries (Cheng et al., 2007; Mestdag and Glorieux, 2009; Warde et al., 2007). Considering the increase of solitary meals in those countries, this finding appears reasonable. Compared to family and social meals, solitary meals are more of an instrumental behavior that has a clear goal and sign of accomplishment (i.e., satisfying hunger). It then tends to be briefer, as there is little reason to continue the behavior once the goal is accomplished. A similar change in meal duration is expected in South Korea, given the large growth in single-person households, among others. According to national censuses, single-person households increased from 15.5% in 2000 to 27.2% in 2015 (http://kostat.go.kr).
It seems also reasonable to expect a decrease in meal frequency. For eating to be considered a meal it has to satisfy various qualifications prescribing time, place, sequence, and social synchronization (Douglas, 1972; Lhuissier et al., 2013). Eating that does not properly satisfy these is often considered a snack or a skipped meal. Solitary meals are less likely to satisfy these requirements than family and social meals, so their increasing prevalence might lead to a decline in meal frequency. Analyzing Flemish data in 1988 and 1999, Mestdag (2005) found a slight decrease in the average number of eating events. Lund and Gronow (2014), however, detected no such a trend in the four Nordic countries they analyzed. It has to be noted that the unit of analysis in both studies is an eating event, including both meals and snacks. In addition to the duration and frequency, growing solitary meals may also produce temporal destructuration, as they take place at less predictable times and are more likely desynchronized from collective routines (Southerton et al., 2012).
Eating behavior and subjective well-being
Eating triggers the endorphin system that boosts happiness and underpins primate and human social bonding (Dunbar, 2017). Moreover, such activities as laughter, reminiscences, jokes, and drinking alcohol that are often associated with commensality further release endorphins (Dunbar, 2017). This means that eating together brings about better subjective well-being than eating alone. In fact, Dunbar (2017) finds that people who eat socially are more likely to feel better about themselves in terms of life satisfaction, happiness, and worthwhileness of life than those who tend to eat alone. The former group also tends to have a wider social network capable of providing social and emotional support. Fischler (2011) reports a similar finding: People experience more positive emotions than negative emotions when they eat in company, compared to when they eat alone, in all the samples taken from the USA, France, and Denmark. Based on these arguments, the following hypothesis is derived:
Hypothesis 1: Family and social meals have a positive effect on life satisfaction, whereas solitary meals have a negative effect.
The relationship between solitary meals and life satisfaction may not be as clear as Hypothesis 1 states. Research shows that alone time may enhance psychological health, as it can be used for coping, emotional release, and self-renewal (see Larson, 1997 for detail). In fact, Larson (1997) finds that spending an intermediate amount of time alone has a positive after-effect on early adolescents’ emotional states, the age at which children begin to voluntarily pursue solitude. It is notable in Dunbar (2017: Figure 6) that the variations in levels of trust, life satisfaction, happiness, and worthwhileness of life are much greater among solitary eaters than they are among social eaters, although the latter group has considerably higher averages. This implies that the effects of solitary meals are more heterogeneous across individuals.
Solitary eating itself is not new. Instead, novel is that more people voluntarily choose it. Research conducted in South Korea (Jo and Cheon, 2018; Lee et al., 2012; Park, 2017) shows heterogeneous motives behind solitary eating; approximately half of solitary eaters surveyed reported involuntary reasons, including ‘hard to find company,’ and ‘hard to coordinate schedules,’ whereas the other half gave more voluntary reasons, such as ‘to save time,’ ‘to enjoy a relaxed meal,’ and ‘to have freedom to choose the menu.’ According to Jo and Cheon (2018), voluntariness was the most important factor associated with the satisfaction of solitary meals, and it was most closely and negatively associated with age.
The KTUS does not measure whether a given activity is voluntary. However, the voluntariness of solitary meals may be inferred from the amount of time one spent alone during the survey day. Voluntariness presupposes the presence of alternative(s). For people to decide to eat alone over alternative options, they must have others around who are willing to eat together. If this holds, then they are expected to spend, on average, less time alone while awake. Given the social characteristic of eating, those eating alone against their will, on the other hand, are less likely to have people to associate with. In addition, solitary eating becomes increasingly involuntary along with age. To sum up, the effect of solitary meals on life satisfaction is expected to be contingent upon how much time one spends alone and how old one is. For these reasons, the second part of Hypothesis 1 is elaborated as:
Hypothesis 2: The negative effect of solitary meals becomes stronger, as one spends more time alone and gets older.
Temporal conventionality and subjective well-being
Institutionalization of activities becomes inevitable as society grows larger. Temporal institutionalization is the process of reducing spontaneity and individual freedom in scheduling one’s time (Zerubavel, 1981). That is, the way that various daily activities are allocated across time is highly patterned in terms of temporal location (when), duration (how long), frequency (how often), and sequence. Note that a small and shrinking part of temporal regularity is governed by natural imperative, and most of the aspects governing temporal regularity are socially based. Without rigidifying the temporal structure of life, or if everyone behaved spontaneously, it would be impossible to form social organizations. The temporal regularity introduces orderliness and structure, facilitates collective cooperation, and minimizes uncertainty (Zerubavel, 1981).
A similar observation is made from the perspective of practice theory, which treats practices, rather than individuals, as the fundamental unit of social analysis (Southerton, 2006). Everyday life consists of numerous activities ordered by spatio-temporal flow, and the activities are largely exclusive of one another in the sense that ‘engaging in one activity can rule out engagement in another’ (Southerton, 2006: 440). Southerton (2013) then argues that practices not only compete for time, but also for temporal locations in order to take on habitual and routine forms of action. Those practices which successfully hold a ‘fixed’ position shape the temporal rhythm of the day (Southerton, 2006). Other practices are allocated around the fixed ones, and their fulfillment is contingent on the emergence of opportune moments in the sequence of fixed practices.
Fixed practices have certain characteristics (see Southerton, 2006 for detail). Important for this study is that they tend to involve the co-participation of other individuals. For this reason, these practices usually require prior arrangement, obligation, and personal commitment. The practices conducted alone, on the other hand, are less likely to be fixed and thus fill empty time slots between fixed practices. Viewed from this perspective, the collective patterning of individual schedules is due largely to fixed practices that shape everyday life.
For its universality and social characteristic, eating tends to be a fixed practice, punctuating and giving temporal order to daily life (Cheng et al., 2007; Southerton et al., 2012; Yates and Warde, 2017). This is evidenced by the high concentration of eating in very limited time slots (Lhuissier et al., 2013; Lund and Gronow, 2014). Solitary eating, however, is somewhat different. It does not involve other people and thus requires little prior coordination and obligation. It is then likely to be a non-fixed practice, relatively free from fixed location and duration, and subject to spontaneity.
The extent to which one assimilates one’s schedule to the collective schedule can be an indicator of social integration. Although constraining individual freedom, according to Durkheim (1995 [1912]), commonly accepted and obligatory beliefs and practices attach individuals to one another, support collective life, provide a sense of belonging, and constitute society. Thus, insufficient common practices are in fact demonstrative of the insufficient presence of society in individuals, under which they are left free and independent. Given his conception of human nature (homo duplex), a physical and a social being that together constitutes the individual, a lack of society corresponds to denying one of the two necessary conditions for existence. Individuals are then subject to depression and suicidal tendencies.
In fact, it will be hard for a large number of people to share the same sentiments on the basis of spontaneity alone (Zerubavel, 1981). The following excerpt from Zerubavel (1981: 65) reveals the relationship between activity synchronization and social integration that he found in the Benedictine monasteries:
Temporal symmetry undoubtedly contributed to a sense of ‘togetherness’ in monastic life. It added to the element of similitude among the different monks and, thus, helped to form a strong basis for what Durkheim called ‘mechanical solidarity.’ . . . [N]ote also that a most common method of punishing monks was to segregate their activities temporally from those of the rest of their community, as by having them eat their meals three hours after them, rather than together with them.
Biological studies also show the positive effects of behavioral synchrony on mental health. Cohen et al. (2010), for example, found that synchronized training in a group condition significantly elevated the pain threshold of college rowers, compared with a similar training regime carried out alone. The authors relate this to a heightened endorphin surge, which is experienced as a light analgesic. This explains why other synchronized activities, such as religious rituals and music, also elevate mood and enhance a sense of bonding, cooperativeness, and generosity (Cohen et al., 2010).
Temporal conventionality is considered important in medical science. Circadian rhythms, referring to the ‘patterns of various physiological processes that vary systematically across the 24-hour period’ (Haynes et al., 2016: 75), are maintained by regular changes in core body temperature, cortisol, and melatonin. In order to adjust to the external environment, the human body develops the internal circadian clock that synchronizes physiological processes with a 24-hour cycle. The clock positioned in the suprachiasmatic nucleus (SCN) of the hypothalamus responds to regular external time cues (zeitgebers) and produces an output signal that regulates biological rhythms (Grandin et al., 2006; Haynes et al., 2016). The most exemplary zeitgeber is photic input (light), although non-photic and social zeitgebers are also suggested, including awakening in the morning, eating meals, and interacting with others.
Disruptions to circadian rhythms can cause mood disruptions, such as bipolar disorder, depression, and anxiety disorders. Studies show that individuals with affective disorders tend to have less routinized activities that are inconsistent with natural light–dark rhythms (Grandin et al., 2006; Haynes et al., 2016). For this reason, social rhythm therapies aim to promote the stability of daily behaviors and restore circadian rhythms. It is not quite clear yet whether social zeitgebers are genuine zeitgebers, influencing circadian rhythms in themselves. Even if they are not, desynchronized daily routines may still modify exposure to light and thereby disrupt the timings of circadian rhythms (Haynes et al., 2016).
To summarize the above arguments, individuals whose daily behaviors deviate from collective routines are more likely to dissociate themselves from collective beliefs and practices, to have limited opportunities to interact with others, and eventually to become socially isolated. This kind of lifestyle reduces hormonal releases that enhance feelings of well-being and social bonding, disrupts circadian rhythms, and may even cause symptoms of mood disorders. Given that eating is a fixed practice that shapes the collective rhythm of daily life, the timing of meals may reflect the level of activity synchronization and influence feelings of depression and anxiety. Based on this, the following hypothesis is formulated:
Hypothesis 3: Individuals with unconventional meal schedules have a lower level of life satisfaction.
Note that temporal conventionality is not a priori but a social construct and thus is empirically defined. It is defined in this study as 60 minutes before and after the points (10-minute slots) in which eating reaches its peaks during the morning, afternoon, and evening of the interview day. This results in three intervals of conventionality each of which is 130 minutes long.
While the impact of desynchronization applies to all the types of meals, it is expected to be more relevant to solitary meals for their greater association with desynchronization. This means that the negative effects of solitary meals (Hypothesis 1) are contingent upon their level of voluntariness (Hypothesis 2) and temporal conventionality (Hypothesis 3). Those who choose solitary meals over more social alternatives and who follow collective routines may not experience such negative consequences.
Data and methods
How people allocate time reflects in behavioral terms their values and involvement in the various domains of life (Warde et al., 2007; Zerubavel, 1981). To take advantage of this, I analyzed the data from the Korean Time Use Survey, conducted by the Korea National Statistical Office every five years since 1999 (see Choi, 2009 for more information). It is a two-day time-diary survey, in which respondents are asked to record their activities on a time diary consisting of 144 ten-minute time slots. The nationally representative sample varies in size across the four surveys (1999, 2004, 2009, and 2014), ranging from about 20,000 to 32,000 respondents, aged 10 or over. Since it is a two-day survey, the number of time diaries is twice as many as the respondents. I selected the respondents aged between 20 and 65 years of age. Those under 20 years of age spend most of the day at school without sufficient autonomy over eating, especially its scheduling and duration. The elderly are excluded as their patterns of eating and the consequences of the patterns may be different from younger adults.
Time-use data are often contrasted to time-estimate data, in which respondents are asked to estimate the time devoted to a specific activity in a typical week/month. This requires respondents not only to recall the activity but also to average out the time (Kan and Pudney, 2008). These tasks often invite normative and perceptual biases, because what people remember they have done is often subject to social norms and different from what they actually did (Mestdag, 2005). On the other hand, time-use surveys, although still recall-based, refer to a much shorter period that is also the most recent reference period (24 hours) and solicit a list of the entire activities respondents performed throughout the day. These characteristics contribute to a more accurate and behavior-oriented measurement.
For the proper analyses of the hypotheses above, I needed to identify when, how long, how often, and with whom each respondent had a meal on the survey days. This was done by combining the information on activity duration with the information on activity start and stop times and the parties accompanying the activity. The person(s) accompanying the activity is defined in the KTUS as the one(s) who actively participated in the activity with the respondent, and classified in 2014 as; (1) alone, (2) spouse, (3) child aged under 10, (4) child aged 10 or more, (5) parent, and (6) others. I recoded these categories into three: alone (1), family members (2–5), and other (6). Note that respondents were allowed to select multiple categories, if necessary, for each activity (e.g., 2 and 3). Social meals are defined in this study as an event in which no family member is involved. For example, when a respondent eats a meal with their spouse and a friend (2 and 6), it is considered a family meal.
One of the focal variables in this study regards the overall level of satisfaction with life, measured on a five-point scale, ranging from ‘Very Satisfied’ (1) to ‘Very Dissatisfied’ (5). This item was one of the topical questions in 2014, and thus not available in the other years. There are two activity codes that regard eating in the KTUS: having a meal and having a snack. This study focuses on the activity reported as meals only. Control variables include age, gender, marital status, educational attainment, personal income, and the time spent alone while awake. Personal income was measured with 12 evenly-spaced ordinal categories.
The data from the weekend (Saturday and Sunday) will be excluded from analysis. There are two reasons for this. One is obviously that time-use patterns are considerably different between the weekend and the week. The other is rather unique to South Korea: Saturday has changed from a half working day to a complete weekend day since the KTUS began. This makes it somewhat difficult to compare the time-use patterns on Saturday before and after the institutionalization of the five-day workweek.
To test the three hypotheses, in which life satisfaction (LS hereafter) is the dependent variable, I use ordinal logistic regression via PROC LOGISTIC in SAS. Since LS has five ordinal categories (j = 5), there are four logit functions: logit(Pr(LS ⩽ j), where j = 1,. . ., 4. Note that each function contrasts the lower (more satisfied) categories of LS with the higher (less satisfied) categories. For example, the first function compares LS = 1 to LS = 2 through 5. Each function has its own intercept and estimates for predictor variables, which means that each predictor has four estimates. If the assumption of proportional odds is held, that is, the four estimates are assumed to have an equal slope, the results become much simpler to present and interpret.
Results
Table 1 shows the frequency and duration of meals. According to the KTUS activity coding lexicon in 1999, meals are divided into binary categories (either family or non-family). Since no data on the accompanying people were collected in that year, it is impossible to distinguish which of the non-family meals are solitary or social. For this reason, a complete comparison is possible after 2004. As shown in the table, an average South Korean spent 83.8 minutes eating 2.69 meals on a typical weekday in 2014. The most noticeable change regards family meals, which rapidly declined in both frequency (from 54.6% in 1999 to 39.8% in 2014) and total duration (from 53.6% to 38.4%). Solitary and social meals, on the other hand, show substantial increases between 2004 and 2014; the former is conspicuous in its frequency, while the latter is so in length.
Average frequency and duration by type of meal on a weekday in South Korea.
The proportions of solitary and family meals in 2014 (29.7% and 39.8% in frequency) are lower, whereas that of social meals (30.5%) is much higher, than the corresponding proportions (39%, 46%, and 15%, respectively) that Yates and Warde (2017) found in the UK for 2012. This may reflect the differences in the meanings attached to eating meals. South Koreans are still more reluctant than Britons to eat alone and more likely to see eating as a means to form and maintain social ties. Britons are more likely to consider eating a private task, such that they mind eating alone less than South Koreans and set the bar higher on eating companions.
When the duration of meals is divided by the frequency, the mean duration per meal is obtained. As expected, meals eaten alone are the briefest (e.g., 15.98/0.70 = 22.8 minutes in 2004), whereas social meals are the most leisurely (31.4 minutes). Interestingly, the meal duration has been increasing since 2004 for all meal types and especially for social meals. As a result, South Koreans spent about 10 minutes more on meals in 2014 than they did in 2004, although the frequency declined slightly from 2.75 to 2.69 meals per day. This is in contrast to prior studies which found either little change (Mestdag, 2005) or a decrease in duration (Cheng et al., 2007; Warde et al., 2007).
Figure 1 presents the rates of people eating meals on the 144 time slots during the survey day in 2004 and 2014. Similar to other countries (Lhuissier et al., 2013; Mestdag, 2005), a three-meal pattern is quite clear in South Korea, and the lunchtime peak is especially distinctive. The three peak points in 2014 were at 7:30AM, 12:20PM, and 7:10PM, corresponding to the traditional breakfast, lunch, and dinner schedule. These peaks are identical to what Mestdag (2005) found in his Flemish data, except for dinner (7:30AM, 12:20PM, and 6:20PM). The later dinner in South Korea reflects longer working hours. The meal timings in 2004 were quite similar to those in 2014. Differences include that the midday and evening peaks in 2004 were reached at 12:10PM and 7:20PM, respectively.
The rates of individuals eating by time of weekday.
Solitary meals were expected to show unconventional timing more than the other meal types. To examine this, the minutes on meals are divided based on how close they are from the peaks to determine whether they are conventional or unconventional. As discussed earlier, eating within a 130-minute range (a peak slot ± 60 minutes) is considered conventional. The conventional hours are from 6:30AM to 8:40AM for breakfast, from 11:20AM to 1:30PM for lunch, and from 6:10PM to 8:20PM for dinner. Table 2 shows that about three-fourths of the time spent eating meals occurred during the three intervals (72.0% in 2004 and 76.1% in 2014). As expected, solitary meals are least likely to take place within the conventional intervals (61.1% in 2004 and 67.3% in 2014), whereas social meals are most likely to do so (75.5% in 2004 and 81.0% in 2014). While all these percentages are contingent upon how wide the interval becomes, the pattern is consistent, no matter how they are defined.
Percentages of meal duration within conventional hours.
Interestingly, the collective rhythms of eating became more pronounced (from 72.0% to 76.1%), despite the growth of solitary meals that follow a less structured schedule. A part of the growing temporal synchrony has to do with the changes in the curves and the peaks between the years. The curves are less skewed to the right in 2014 than they were in 2004, which increases the area under the curve falling within the conventionality interval. Despite this, it is still clear that more meals, especially solitary ones, were taken during the conventional hours in 2014 than they were in 2004. This indicates that (1) there is little support for temporal destructuration of meals in South Korea, and (2) solitary meals in 2014 were more structured than those in 2004 with regard to timing.
Table 3 shows the distribution of LS and meal duration by LS levels. The most common response to LS is ‘Moderate’ (46.7%), whereas only a small fraction (3.9%) reported being ‘Very Dissatisfied.’ Slightly more respondents are (very) satisfied (26.3%) than they are (very) dissatisfied (20.6%). One can also find in the table that LS has an inverse relationship with the duration of solitary meals, whereas it has positive relationships with the duration of family and social meals. Table 4 presents the frequency distributions of the meals, the focal explanatory variables. As shown, a vast majority (about 80%) of the respondents either never ate solitary or social meals, or they ate them only once during the survey day. Family meals, on the other hand, demonstrate a more diverse distribution.
Distribution of life satisfaction and meal duration in minutes, 2014.
Distribution of respondents by meal frequency (%).
For multivariate analyses examining the relationship between the meal types and LS, ordinal logistic regression is used. The problem is that the assumption of proportional odds is not held. In fact, the statistical test of the assumption is known to be overly conservative and almost always rejects the assumption, especially when the sample size is large, or when there are many predictor variables (Allison, 2012) that include continuous ones. An obvious option when this happens is to run a series of binary logistic regressions for the four logit functions. However, the results are complicated to present, and hypothesis tests become less powerful, which explains why the actual choice does not depend solely upon the statistical test (Allison, 2012). Since the results of binary and ordinal logistic regressions are highly consistent, the ordinal logistic regression results are presented in Table 5. Note that the explanatory variables in Table 5 predict the probability of being more satisfied with life.
Ordinal logistic regression of life satisfaction on types of meals.
Log-transformed, if measured in minute.
*p < 0.05, **p < 0.01, ***p < 0.001.
Both Models 1 and 2 examine Hypothesis 1. Meals in Model 1 are measured in duration (minutes), whereas they are measured in frequency in Model 2. The results are almost identical and show that the time spent eating with others and the frequency of doing so increase the odds of being at a higher level of LS. For example, an additional 1% increase in the duration of family meals raises the odds of being more satisfied by 5.8% (exp.(0.056) = 1.058). Solitary meals, on the other hand, are not significant (p = 0.306 in Model 1 and 0.132 in Model 2). Binary logistic regressions and the partial proportional odds model (results not reported), which allows for some explanatory variables to have different slopes across the levels of LS, indicate that the length of social meals does not affect the odds of being ‘Very Satisfied,’ whereas it has significant and positive impacts on the lower levels of LS. To sum up, the results in Models 1 and 2 generally support Hypothesis 1, except for the part that claimed solitary meals have a negative influence.
Models 3 and 4 in Table 5 examine Hypothesis 2, stating that the effects of solitary meals on LS are contingent upon the meal’s voluntariness. The level of voluntariness is incorporated into those models via the two interaction terms. The interaction term between solitary time and solitary meal shows an expected sign (–) but is only significant at the 0.10 level in Model 3 (p = 0.090). The same term, however, is significant in Model 4 (p = 0.005). The other interaction term between solitary meal and age is significant in both models. To sum up, the results generally support Hypothesis 2. That is, the effects of solitary meals become negative, as one spends more time alone and grows older, both of which are considered correlates to the voluntariness of solitary meals.
Table 6 presents the results from the ordinal logistic regression to examine Hypothesis 3. The minutes on meals are divided into two component parts: those taking place within the conventional hours and those occurring outside those hours. Model 1 includes the unconventional minutes only, whereas Model 2 adds the conventional minutes to Model 1. As shown in the table, the unconventional minutes lower LS, if it is solitary (p = 0.003 in Model 2), but are inconsequential for family and social meals. That is, the more minutes one spends eating alone outside the conventional hours, the less satisfied one becomes with life. However, the same is not true for social and family meals: Eating with others does not lower LS, even if it takes place outside the hours. This is somewhat unexpected, as those meals, if accompanied, are desynchronized from collective routines.
Ordinal logistic regression of life satisfaction on timing of meals.
Log-transformed.
*p < .05, **p < 0.01, ***p < 0.001.
Model 2 clearly shows that all types of meals, if taking place within the conventional hours, increase the odds of being more satisfied with life. This result includes solitary meals, although the coefficient (0.024) is only half as much for family (0.049) and social (0.047) meals. To sum up, the results in Models 1 and 2 of Table 6 support Hypothesis 3.
As mentioned above, the conventionality interval (130 minutes) is rather arbitrary, so that the results in Models 1 and 2 may change as the interval gets wider or narrower. To examine this, the interval was operationalized as 90 minutes long (a peak slot ± 40 minutes) and as 170 minutes long (a peak slot ± 80 minutes). The results, presented in Models 3 and 4 of Table 6, demonstrate a clear pattern. As the interval lengthens, the effects of off-hour eating become increasingly negative. A good example is provided by social meals. Its coefficient (+0.012) is positive, although not significant, when the short interval (90 minutes) is applied. However, it becomes negative (–0.029) and significant (p = 0.015) for the long interval (170 minutes). This makes sense because the longer the interval gets, the more desynchronized the unconventional meal becomes. Also note that the coefficient of the conventional solitary meal decreases (from 0.030 to 0.024 and to 0.018), as the interval lengthens.
Conclusion and discussion
This study examined the changes in eating behavior and its effects on LS in South Korea. The shared family meal has declined fast in both frequency and duration. As a consequence, a growing number of South Koreans either reluctantly choose to eat alone or actively pursue the comfort and freedom it provides. The latter reason is evidenced by the growing temporal synchronization of solitary meals (see Table 2), indicating that it is increasingly performed within a narrow time frame along with a large aggregate of others. In addition, social meals also increased simultaneously and to an equal extent. This might be a necessary reaction to the increasing number of solitary meals. To counteract the sense of isolation arising from solitary meals, people simultaneously searched for intimacy and the sense of belonging derived from eating socially (Park, 2017).
These trends are not unique to South Korea, as previous research has reported similar findings in other societies. For example, Mestdag and Glorieux (2009) found the displacement of family meals by solitary and social meals in their Flemish data. While an exact comparison is not possible, the changes in South Korea appear more dramatic than those in the Flemish study. Likewise, Lund and Gronow (2014) showed that the temporal rhythm of eating was more stable than was often claimed, and desynchronized eating was not prevalent. The findings of the current study were beyond common expectations and observations: Eating became slightly more synchronized in South Korea, despite the growing trend for solitary meals. While the rapid increase in social meals mirrors the collectivist perception of eating in Korean society, solitary meals being more synchronized may represent a divergence from this cultural norm.
The increase in social meals explains why South Koreans spend more time eating, despite the declining frequency of meals (see Table 1). As mentioned earlier, no such trend has been reported elsewhere. One explanation is that social gatherings in South Korea have increasingly involved eating and that the respondents reported the eating part of such socializing as time spent eating. This conjecture is supported by the fact that the reported time spent on face-to-face social interactions decreased during the same period (from 37.1 minutes on a typical weekday in 2004 to 26.4 minutes in 2014). This decline would be puzzling, if one does not properly understand the changes in eating patterns.
Eating with others, be they family members, friends, or colleagues, had positive effects on LS. Solitary meals, on the other hand, demonstrated neither a positive nor negative influence. The effects of solitary meals, in fact, were heterogeneous across individuals, depending upon the level of voluntariness and timing of the meal. That is, as long as it was a voluntary choice and took place within conventional hours, solitary meals improved LS.
As expected, the negative effects of eating at unconventional times increased in tandem with the width of the conventionality interval (see Table 6). For instance, if a person eats a meal at 8:30PM, the meal is considered unconventional in both the narrow (6:30PM–8:00PM) and medium intervals (6:10PM–8:20PM), while it is conventional in the wide interval (5:50PM–8:40PM). Thus, an activity that is considered unconventional when the wide interval is used deviates more from collective routine than does an activity that is considered unconventional when the narrow interval is used. Following, it makes sense that the negative effects of unconventional meal timing strengthen as the interval widens.
Despite this general trend, the extent to which this applies varies by meal type. This trend was most prevalent in solitary meals and least so in family meals. No matter the length of the intervals, eating at unconventional times had a negative effect on LS, when it was performed alone. However, it was less true for social meals and not true at all for family meals. These findings indicate that: (1) eating in company protects individuals from the adverse impacts of a desynchronized eating rhythm; (2) individuals are better insulated from social desynchronization when they are with family members than when they are with friends and colleagues.
The current study has some limitations. First, analyzing cross-sectional data, it argued that the way one eats influences one’s LS. However, the causality between the two variables may be reciprocal or even reversed. While prior research on this issue is very limited, one exception is Dunbar (2017). His path analysis supports a one-way causal path flowing from eating social dinner to higher LS. Secondly, the current study could not take into account the voluntariness of eating together. Eating together may not always be a pleasant experience (Giacoman, 2016; Park, 2017). In South Korea, where social hierarchy is emphasized, eating together may be less preferred unless it is with peers. Therefore, the precise effects of eating together may be contingent upon context. In the case of eating alone, the degree of voluntariness was indirectly inferred from the amount of time spent alone and age. These indicators may not precisely reflect the degree. Lastly, the boundary between family and social meals was somewhat ambiguous. For example, eating with spouse and friend(s) is considered a family meal in this study. Although not frequent, such cases blur the boundary.
Footnotes
Funding
The author received no financial support for the research, authorship, and/or publication of this article.
