Abstract
Background:
The present study explored the influence of eating habits, body weight and television programme preference on television viewing time and domestic computer usage, after adjusting for sociodemographic characteristics and home media environment indicators. In addition, potential substitution or complementarity in screen time was investigated.
Methods:
Individual level data were collected via questionnaires that were administered to a random sample of 2,946 Germans. The econometric analysis employed a seemingly unrelated bivariate ordered probit model to conjointly estimate television viewing time and time engaged in domestic computer usage.
Results:
Television viewing and domestic computer usage represent two independent behaviours in both genders and across all age groups. Dietary habits have a significant impact on television watching with less healthy food choices associated with increasing television viewing time. Body weight is found to be positively correlated with television screen time in both men and women, and overweight individuals have a higher propensity for heavy television viewing. Similar results were obtained for age groups where an increasing body mass index (BMI) in adults over 24 years old is more likely to be positively associated with a higher duration of television watching. With respect to dietary habits of domestic computer users, participants aged over 24 years of both genders seem to adopt more healthy dietary patterns. A downward trend in the BMI of domestic computer users was observed in women and adults aged 25–60 years. On the contrary, young domestic computer users 18–24 years old have a higher body weight than non-users. Television programme preferences also affect television screen time with clear differences to be observed between genders and across different age groups.
Conclusions:
In order to reduce total screen time, health interventions should target different types of screen viewing audiences separately.
Keywords
Introduction
There has been a rapid growth of evidence demonstrating that sedentary screen-based behaviours are linked with major health impacts and risks,1,2 with television (TV) viewing to have mostly been explored as the dominant sedentary behaviour 3 and the most frequently measured leisure time activity.4,5 TV screen time appears to be related to severe health consequences, including weight gain and obesity, cardiovascular risk factors, type 2 diabetes and the metabolic syndrome.6 –11
In a comprehensive review, Boulos et al. analyse a host of factors that correlate watching TV with weight gain, including decreased physical activity, food marketing and food commercials, eating while watching TV, various forms of product placement in TV programmes and obesity stereotypes. 12 Engagement in TV viewing may displace the time allocated in physical activity resulting in a positive energy balance. 13 However, there is evidence indicating that the relationship between TV viewing time and body weight is independent of levels of physical activity. 8 Thus, the association between watching TV and being overweight may be attributed to the dietary patterns and the increased energy intake via the snacking habits and consumption of high-calorie foods in front of the TV set. 14 It is likely that watching TV may impede individual’s ability to react to internal hunger and satiety cues, and instead result in a strong dependence on external cues related to TV screen time (e.g. individuals complete their meal by the end of a TV show).12,15 Furthermore, the bulk of food advertising consists of commercials of energy-dense foods with low nutritional quality, 16 which direct individuals to evaluate these foods more favourably by increasing their desirability and acceptance. 17
In contrast to watching TV, computer usage encompasses a broader range of aspects including leisure time, occupational, educational and personal computer activities. Altenburg et al. investigated the association between the time engaged in computer usage and biomarkers of cardiometabolic risk and concluded that there is no significant effect of involvement with computers on any of the cardiometabolic biomarkers, including body weight. 18 On the contrary, Shuval et al. identified the time allocated to computer tasks as a ‘risk marker’ for weight gain and being overweight. 19 Other recent studies in adult populations have showed conflicting results regarding the relationship between recreational internet usage and body weight. Namely, a recent study by Van Dyck et al. revealed a non-significant association between leisure time internet usage and body mass index (BMI), 20 whereas Vandelanotte et al. underlined the strong link between high recreational internet time and being overweight, even in the case of a highly active lifestyle. 21
Given the well-documented health impacts of screen time, the investigation of the factors that explain screen-based behaviours is considered to be crucial for designing and implementing health interventions in order to reduce screen time. Besides, behaviour change can be accomplished after identifying and modifying the main determinants of the target behaviour. 22 Under this premise, this study seeks to explore (1) the influence of eating habits, body weight and TV programme preferences on TV viewing time; (2) the association of domestic computer usage with eating habits and body weight; and (3) the potential interdependency between watching TV and domestic computer usage after controlling for sociodemographic characteristics and home media environment indicators. Within the frame of the ‘behavioural economics’ approach, individuals may choose between behaviours that could be classified as either complementary or substitute behaviours. 23 In the first case, TV viewing time is followed by more/less time allocated in other sedentary behaviours, such as computer usage, reflecting a more generalised increase/decrease of screen viewing time. 24 In the case of substitution, TV viewing is negatively associated with computer usage, indicating ‘a compensation effect in operation’, with individuals switching between different screen viewing behaviours.25,26
Methods
Sampling
This study used data from the German General Social Survey (ALLBUS 2004), covering a wide range of topics, such as social inequality, attitudes, health, free time activities, digital divide and the welfare state. The dataset used for this study is the most recent and complete dataset providing information for media use that includes both television viewing and computer usage parameters. In addition, the data on domestic computer usage encompass all the aspects of domestic computer usage including the time spent on social and recreational activities as well as the time dedicated to accomplish work tasks and personal errands or improve learning skills.
The population from which our sample was drawn included German-speaking adults from both Western and Eastern Germany who resided in private households. Participation was voluntary and the study protocol was given ethical approval by GESIS – Leibniz Institute of the Social Sciences. The sampling procedure was completed in two phases and comprised probability sampling techniques. First, municipalities in Western Germany and in Eastern Germany were selected with a probability proportional to the number of adult residents. Therefore, the completion of the first phase ended with the collection of 111 sample points in 104 Western Germany municipalities and 51 sample points in 46 Eastern Germany municipalities. In the second phase, individuals were picked up at random from the residents’ municipal registers. A formal standardised questionnaire administered via personal interviews, as well as an additional self-completion questionnaire comprised the basic tools for the data selection. The response rate for the questionnaire completion was 44.9% and 47.6% in Western and Eastern Germany, respectively. Finally, 2,946 valid questionnaires were used. 27
Measures
Respondents were asked how long on average they watched TV per day and how much time they spent on the computer at home. Two indicators were constructed to encompass TV watching and computer usage respectively. First, the self-reported TV viewing time was defined by an indicator taking the value of one if the respondent limits TV viewing to less than 2 h a day (limited TV viewers). The value of two is for respondents who watch TV for 2–4 h a day (moderately high TV viewers) and the value of three corresponds to prolonged TV viewing that extends beyond 4 h a day (heavy TV viewers). 28 The cutoffs for the time spent on TV viewing were defined according to previous studies linking TV viewing with health risks.29,30
Furthermore, respondents were asked to report the time engaged in domestic computer usage (including the time spent using the internet) for several reasons, such as formal, educational, professional, leisure or personal activities. Then, domestic computer usage was divided into three categories and assessed through a 3-point ordinal scale, as follows: value 1 is for zero domestic computer usage, while the values of 2 and 3 correspond to 0.5–15 h per week (low to moderate computer usage) and more than 15 h per week (high computer usage), respectively. In this study, 41.7% of the total sample was defined as ‘domestic computer users’. The cutoff for the upper level of the computer usage indicator was specified according to the 75th percentile for the subsample of computer users. Since the 75th percentile was 15 h, high computer usage was set at more than 15 h per week.
In order to account for home media environment, participants were asked to record the number of the available TV sets and computers in their household. Since the median number for both TV sets and computers was equal to one, two binary indicators were created to stand for respondents having more than one TV set and more than one computer in their home-environment.
To elicit specific preferences for TV programmes, respondents were asked to score a 10-item variable regarding the level of their interest in various TV programme contents on a 5-point semantic differential scale from ‘very strong interest’ to ‘no interest at all’. A factor analysis through principal component analysis (PCA) with varimax rotation was employed, aiming to group the variables according to their relevance and create a smaller number of manageable size factors. The internal consistency of each factor was evaluated by calculating the Cronbach’s alpha coefficients. The factor analysis provided a four-component solution and the total variance explained was 64.438%. The Cronbach’s alpha coefficients were .668, .580 and .576 for the first, second and third components, respectively (the fourth factor consists of only one item). Table 1 analytically presents the main findings from the factor analysis application. To create measures of TV viewing preferences, the detached items of each factor were added and the sums were used to construct and define four dichotomous indicators reflecting the respondent’s interest in informative/educational TV programmes, traditional entertainment TV programmes, movies/films, and sports.
Factor analysis results on TV programmes preferences
Respondents also reported their weight (kg) and height (m), while body weight indices were measured by calculating the BMI. 31 Furthermore, information on food and alcohol consumption patterns was included for describing participants’ eating habits. In particular, food consumption frequencies were depicted by eight indicators that measured respondents’ consumption of specific food groups, such as whole grain or multigrain bread or rolls, toast and white bread, fresh fruits, fresh and frozen vegetables, meat and meat products, deep fried foods and confectionery (sweets, cakes, biscuits, pastries). The list of these indicators mapped basic nutritional categories 32 and represented distinct food groups. The introductory question asked how often the participants consumed foods from each of the aforementioned food groups separately and a seven-point scale was rated, ranging from several times a day to never (several times a day, every day/almost every day, several times a week, about once a week, twice or three times a month, once a month or less often, never). This seven-point frequency scale was also used to assess wine-beer and spirits consumption frequency. In the context of the present study, a frequency of ‘at least daily’ was chosen as the cut-off point to designate frequent consumption of a food/beverage category. 33
The explanatory variables also included demographic and socioeconomic characteristics. Demographic variables pertained to dichotomous indicators depicting gender, age (young adults 18–24 years old, 25–35 years old, 36–46 years old, 47–59 years old, older than 60 years old), marital status (separated/divorced/widowed, married, single) and household size. Furthermore, dichotomous indicators standing for educational attainment (primary education, level I/high school education, university education, technical education, students) and disposable income were used to specify socioeconomic characteristics.
Econometric analysis
Given the ordered nature of the dependent variables (TV viewing time and computer usage) and to capture possible interdependency between them after adjusting for various covariates, the analysis procedure adopted a seemingly unrelated bivariate approach. TV viewing time and time engaged to domestic computer usage were conjointly estimated through the seemingly unrelated bivariate ordered probit model. 34 This approach also estimated the correlation between the error terms of both equations as an auxiliary variable and tested for statistical significance. The estimation was accomplished by means of the general Full Information Maximum Likelihood algorithm. The likelihood ratio test was also performed to assess the independence of equations under the null hypothesis ρ = 0.
For this study, the first and the second equation described TV viewing time and time allocated to domestic computer usage, respectively. TV viewing time was represented by an ordinal indicator to distinguish among limited, moderate to high and heavy TV viewers. Computer usage was also illustrated by an ordinal indicator corresponding to three different levels of time allocated to domestic computer usage: namely, no computer usage, low to moderate computer usage and high computer usage. The covariates included individual’s BMI, food and beverage consumption frequencies, TV and computer availability, as well as demographic and socioeconomic characteristics. In addition, the TV viewing equation included the four TV programme preference indicators produced by the factor analysis components. The analysis was performed separately for males, females and across different age categories since recent research supports that domestic screen viewing behaviour, especially TV watching, presents gender differences,35–37 and there is a noticeable age-related upward trend in TV viewing time.35,38 To better interpret the results obtained from the econometric analyses, respondents’ age was classified in three groups, namely, young adults 18–24 years old, adults 25–60 years old and adults over 60 years old.
Results
Table 2 provides an analytical description of the sample. In all, 58.3% of the respondents reported no domestic computer usage and 48.7% designated themselves as limited TV viewers (less than 2 h a day). The predictors presented a wide degree of variability between genders and across age groups. The application of chi-square tests revealed that both gender and age classification are strongly related with domestic screen viewing time and eating patterns. Furthermore, parametric test (t-test) demonstrated a statistically significant difference in the BMI of men and women, whereas the application of analysis of variance (ANOVA) showed that the mean BMI is not statistically equal across age groups.
Sample characteristics (N = 2,946)
ANOVA: analysis of variance; BMI: body mass index.
Mean, standard deviation in parentheses.
Besides investigating the factors influencing domestic screen time, the seemingly unrelated bivariate ordered probit model also examined the correlation between the two equations of the bivariate procedure for evidence of a potential conjoint process in decision-making. All the likelihood ratio tests performed to assess the independence of equations accepted the null hypothesis ρ = 0. More specifically, in the procedure adopted for men, it turned out that the two equations were not strongly correlated since the estimated correlation coefficient ρ took the value of −0.01 and it was statistically insignificant (χ2(1) = 0.10, p = .751). In the application to women, the correlation coefficient was equal to −0.05 and it was also statistically insignificant (χ2(1) = 1.32, p = .250), indicating that the two behaviours are independent. Similar results were obtained for age classification. Therefore, the bivariate models employed across age groups resulted in statistically insignificant correlation coefficients for young adults 18–24 years old (ρ = −0.115, χ2(1) = 2.16, p = .141), adults 25–60 years old (ρ = −0.032, χ2(1) = 0.87, p = .350) and adults over 60 years old (ρ = 0.060, χ2(1) = 0.58, p = .448). Thus, the time allocated to domestic computer usage has no effect upon TV viewing time and these two behaviours cannot be considered to be either substitute or complementary in both genders and across all age groups of our sample.
Tables 3 and 4 display the results derived from the seemingly unrelated bivariate ordered probit model analyses for both genders and across age groups, respectively. Statistically significant differences observed between males and females and across different age categories for several regressors illustrate that both gender and age conceivably influence differences in domestic screen viewing behaviours.
Seemingly unrelated bivariate ordered probit model estimates for male and female participants
BMI: body mass index.
Age: 36–46 years old (omitted variable); Marital status: married (omitted variable).
Educational attainment: technical education (omitted variable).
Seemingly unrelated bivariate ordered probit model estimates across age categories
BMI: body mass index.
*Marital status: married (omitted variable), Educational attainment: student (omitted variable).
n = 310.
n = 1,821.
n = 812.
With respect to individual’s body weight, it seems that BMI has a more intense influence upon domestic screen viewing behaviour in females and adults aged 25–60 years, since it was found to affect both computer usage and TV viewing time. Contrary to our anticipation, BMI was inversely related to domestic computer usage in women (β = −0.017, p = .036) and among individuals aged 25–60 years (β = −0.020, p = .002), whereas a statistically significant positive association between body weight and computer usage was observed in young adults 18–24 years old (β = 0.036, p = .052), suggesting that associations may be confounded in different ways among groups with different characteristics. As it was expected, our findings confirm a positive correlation between BMI and the level of TV watching in both males (β = 0.016, p = .053) and females (β = 0.034, p < .01). In the same way, increasing TV viewing was linked with a higher body weight among individuals over 24 years old (25–60 years old:β = 0.030, p < .01, over 60 years old: β = 0.021, p = .033).
Eating patterns were also found to influence both TV viewing time and domestic computer usage. A daily consumption of white-toast bread was positively related to TV viewing time in both genders (females:β = 0.262, p < .01, males:β = 0.174, p < .01) and adults over 24 years old (25–60 years old:β = 0.239, p < .01, over 60 years:β = 0.217, p = .017). A positive correlation between TV viewing time and wine/beer consumption was observed in young adults (β = 0.681, p = .028). Furthermore, participants aged 25–60 years who reported a daily consumption of meat products were more likely to spend more time watching TV (β = 0.188, p < .01). A frequent meat consumption also augmented the probability for more time watching TV in both males (β = 0.127, p = .058) and females (β = 0.165, p = .012). On the contrary, frequent whole grain bread consumption was inversely associated with TV viewing time in both genders (males: β = −0.187, p = .005, females: β = −0.133, p = .047) and adults over 24 years old (25–60 years old: β = −0.157, p = .009, over 60 years: β = −0.180, p = .044). A daily vegetable consumption was also negatively related to TV screen time in female participants (β = −0.161, p = .015) and young adults (β = −0.314, p = .076).
With regard to domestic computer usage, the time engaged in involvement with computers was positively influenced by the frequent consumption of beer/wine (β = 0.309, p = .074), whole grain bread (β = 0.142, p = .065) and vegetables (β = 0.126, p = .097) in women. In the male sub-sample, involvement with computers was positively correlated with the frequent consumption of whole grain bread (β = 0.128, p = .065), vegetables (β = 0.163, p = .031) and confectionery (β = 0.268, p < .01). Conversely, male participants who daily consumed wine/beer, white-toast bread, fruits and fried foods were less likely to engage in tasks performed on a computer (Table 3). Concerning age classification, frequent fruit consumption was negatively related to involvement with computers in young participants 18–24 years old (β = −0.294, p = .050) and adults aged 25–60 years (β = −0.107, p = .094). In addition, respondents aged 25–60 years who reported a daily white-toast bread consumption spent less time allocated in domestic computer usage (β = −0.119, p = .068). On the contrary, an upward trend in the time spent on a computer was observed in 25- to 60-year-old individuals with a daily consumption of whole grain bread and confectionery, and in older adults over 60 years with a daily vegetable consumption (Table 4).
TV programme preference was also found to constitute a significant determinant of the time spent watching TV for both males and females. More specifically, watching movies and films was related with higher TV viewing time in both genders. Furthermore, the propensity for higher levels of TV viewing increased in women with a preference for traditional entertainment TV programmes and in men with a stronger interest in sports TV programmes (Table 3). Regarding TV programme preferences across age groups, our findings indicated a positive association between preferences in movies/films and watching TV in participants at all ages. In addition, TV viewing time increased for adults over 24 years with a stronger preference in traditional entertainment and sports TV programmes, whereas participants aged 25–60 years who favoured informative and educational TV programmes seemed to reduce TV screen time (Table 4).
Discussion
This study using data from a random sample of German adults provides evidence that watching TV and domestic computer usage constitute totally independent behaviours for both genders and across different age categories and there is neither a substitution nor complementary effect on screen time between them. A possible explanation of this independency might be the different nature of each behaviour. Watching TV is a predominantly leisure time activity, whereas domestic computer usage includes both work/study and leisure computer time. Thus, there might be differences in the amount of time allocated to multiple computer tasks that could not be taken into consideration. Recent research also underlined the role of age in the internet’s effect on TV viewing time with older age groups being less likely to reduce watching TV or increase internet usage. 39 Given that the majority of the participants were over 50 years old (49.4%), it seems that domestic computer usage will have a weak relationship with TV viewing time.
In line with previous studies, our findings also demonstrated that body weight is positively correlated with watching TV in both genders and adults over 24 years old, and overweight individuals have a higher propensity for heavy TV viewing.7,9,11,40 According to Levine et al., overweight individuals are more likely to engage in more sedentary activities because of their body weight and/or genetic predisposition. 41 One possible explanation of this positive correlation is that watching TV may partially replace physical activity, resulting in overall lower energy expenditure. 42 However, there is evidence to suggest that TV viewing time and body weight are associated regardless of the levels of physical activity, 8 implying that factors such as eating patterns may play a critical role in explaining the relationship between TV viewing time and BMI. 43
Less healthy dietary patterns were found to contribute to a higher TV viewing time in both genders and across age. Our findings add to the evidence in the literature that lower diet quality is associated with exposure to TV, with elements of less healthy dietary habits to be associated with TV screen time in both genders and all age groups.37,43,44 TV programme preferences also seem to substantially affect watching TV with clear differences to be observed between males and females and across age groups regarding the programme content. In particular, male participants tend to spend more time watching sport events, whereas in the female sub-sample, time spent on TV viewing increases when traditional entertainment programmes are on screen. In TV viewers aged over 24 years, a stronger interest in traditional entertainment and sports TV programmes contributed to a substantially higher TV viewing time, whereas participants 25–60 years old who preferred educational and informative TV programmes reduced TV screen time. For both genders and across all age groups, TV screen time also increased in the case of movies and films. Recent research indicated that movies and films are most likely to include food/beverage brand placement. 45 In addition, the great majority of TV food commercials present energy-dense foods with low nutritional quality. 16 Indirect effects of TV viewing on unhealthy eating patterns are also of major importance, especially for specific segments, such as women, since there are particular types of TV programmes mostly engaged in traditional entertainment categories (e.g. cooking shows), which promote food as a type of entertainment. 12 Further research should explore food advertising and brand placement tactics in TV programmes with different contents in order to investigate potential differences in the level of responsiveness towards food marketing environment among different profiles of TV viewers. In addition, these findings indicate the need for food policy interventions designed to inform and orientate people to enjoyable alternatives for relaxation during leisure time in order to decrease watching TV.
Our results also showed that body weight is inversely correlated with the time engaged in computer activities in female participants and adults aged 25–60 years. TV viewing time and domestic computer usage seem to differ from each other in their association with body weight in these groups. Energy expenditure may be lower when watching TV rather than when using the computer, since the latter may require higher muscle activity. 18 In addition, a difference may be expected in the levels of food consumption and the subsequent energy intake between TV viewers and computer users since TV viewers’ hands can perpetually be used for eating, unlike involvement with computers. 46 On the other hand, our findings showed a positive correlation between body weight and domestic computer usage in young adults 18–24 years old. One of the potential mechanisms for BMI increase in young computer users may be the sedentary nature of the activity in connection to the kind of tasks accomplished with computer. Vandelanotte et al. underlined the relationship between the time engaged in computer activities and the time spent on other sedentary activities, 21 resulting in lower energy expenditure. Furthermore, young adults are more likely to allocate more time to computer activities, such as gaming, that are more linked to weight gain compared to other leisure time computer activities. 47 Health practitioners should also consider the potential consequences of computer gaming in the body weight increase of older adults since young gamers will apparently keep similar computer usage patterns across the life span. 48
With respect to dietary habits, domestic computer users seem to be oriented towards more healthy eating patterns compared to TV viewers. In particular, the frequent consumption of whole grain bread and fresh or frozen vegetables augments the probability for increasing time allocated in domestic computer usage in both genders and adults over 24 years old. At this point, it should be noted that this study included information about domestic computer usage without distinguishing between leisure time computer usage and computer usage for occupational or educational purposes. Future research should fill this gap, by examining food consumption patterns and their impact on recreational and non-recreational computer usage separately.
Limitations
The limitations of this study refer to the self-reported data on screen media time, food consumption frequencies and individual’s weight and height. The use of self-report may result in over- or under-reporting due to limited recall, social desirability or other biases.49,50 Although several researches in adolescents’ health have noted high correlations between self-reported and objectively measured weight and height, future studies should seek to include body weight measures assessed more precisely. 51 In addition, there were no available data for distinguishing domestic computer usage for watching TV. The cross-sectional design also excludes some statements concerning the causalities of the observed associations.
Conclusion
This study expanded upon previous research and investigated potential interdependency on screen time between TV viewing and domestic computer usage using a representative sample of German-speaking adults. It showed that watching TV and domestic computer usage comprise independent behaviours in both genders and across all age groups, and may encompass other lifestyle factors rather than screen viewing sedentary time per se.
This study also confirmed the complexity of TV viewing time and time spent on a computer that seem to be closely related to eating patterns and body weight. Watching TV remains the dominant screen-based activity, while the amount of time spent on computers at home involves both recreational and educational/occupational purposes. Future research examining the potential interdependency between TV viewing time and different domestic computer usages could help to improve the effectiveness of health interventions, by directing them towards the reduction of total screen time.
Footnotes
Conflict of Interest
The authors declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.
Funding
The authors received no financial support for the research, authorship and/or publication of this article.