In this nationwide cross-sectional study of a randomly selected sample of adults in Germany, it is observed that adults with a lower level of education are more physically active when they are working, less active in leisure time, and expend more energy in total than adults with a higher education level. These observations are in line with the findings of other studies [11, 13, 15–18]. Education shows the strongest independent associations with the physical-activity variables compared to occupation and income; this also confirms the findings of previous studies .
Studies suggest that individuals who engage in more vigorous activity during working hours may, as a response, have lower levels of physical activity during leisure time and vice versa [14–16]. We found that the association between education and leisure-time activity is mediated by vigorous work activity, which supports the hypothesis that low-SEP groups are less physically active in their leisure time because they are more physically active when working. It should be born in mind however, that some people are more active over all domains than others. 8% of adults with primary education, 7 with secondary education and 2.5 with tertiary education reported simultaneously that they fulfil vigorous working tasks and engage in sports activity at least 2 hours per week. Another mediator for the association between education and leisure-time activity is BMI. A high body mass index can act as a barrier to engaging in leisure-time activity, and the higher prevalence of overweight among people with a lower level of education might partly explain the education disparities in leisure-time activity.
Occupational physical activity dominates the 24-hour total energy-expenditure index because occupational activity usually corresponds to an 8-hour working day, and leisure-time physical activity is usually performed over shorter periods . Thus, high-SEP individuals who are mainly sedentary at work may be unable to fully offset the higher energy expended at work by individuals with a lower SEP. A representative study from the United States based on accelerometer data has observed that men and women with active jobs have more wear-time counts in total than men and women with sedentary jobs . The same database indicated that men with a lower education level had more wear-time counts on average than men with a higher level of education , which is in line with our results.
The described patterns change in higher age groups, where SEP-related differences in work-related activity tend to disappear. In this higher age group, the association between education and sports activity was stronger, the association between education and sitting time weekdays disappeared, and the negative association between education and high total energy expenditure turned into a positive association. It may be that individuals who used to do physically-active jobs during their working years do not increase their participation in sports activity when they retire, with the result that their level of total energy expenditure decreases. This may have something to do with the fact that they have not been socialized to participate in leisure-time physical activity to maintain a good level of physical fitness and may lack networks, opportunities and the physical health to start doing so at an older age when they stop working. By contrast, adults with higher education who may have compensated inactivity at work with leisure-time physical activity during their working years may use their increased free time in retirement to increase level of leisure-time physical activity.
Social cognitive theory distinguishes different types of behaviour according to the degree of volitional control an individual has in order to change behaviour. Non-volitional behaviour is defined as behaviour over which the individual has limited freedom of control, whereas volitional behaviour is a form of behaviour over which the individual does have the freedom of control . It may be reasonable to categorize occupational activity as non-volitional behaviour, since the individual has only limited freedom to change occupational activities that are determined by contract. Thus, work-related activity might be seen as a structural factor which influences leisure-time activity behaviour. Occupational activities are directly related to a person’s educational background, occupational position and level of income. Hence, the associations between SEP variables and occupational activities are particularly strong. The stronger the freedom to control behaviour, the weaker might be its socioeconomic determination. Perhaps this explains why the socioeconomic prediction of leisure-time activity was found to be lower than for occupational activity.
An important question may arise when considering our results: if physical activity is good for a person’s health, and adults with a higher level of education are less physically active in total, why are they healthier? In addition to a healthier diet and lower smoking prevalence, one hypothesis may be that occupational physical activity does not have the same health benefits as leisure-time physical activity. Most studies showing a reduction of mortality risk with increasing physical activity rely on leisure-time activity data . Findings on occupational physical activity and health are inconsistent, but in studies that assess both leisure-time and occupational activity, the dose–response relationship tends to be lower for occupational activity than for leisure-time activity [15, 31–36]. Two studies which investigated the relationship between physically demanding work and physical fitness showed that, although muscular strength was greater among physical workers than among sedentary workers, aerobic fitness was higher among sedentary workers in one study but showed no difference in the other [37, 38]. Lakka et al. has also shown that occupational energy expenditure is not positively related to cardiorespiratory fitness among Finnish men . Although physically demanding work tasks vary widely depending upon the occupation , it appears that work activity often correlates with muscle-strengthening activity (e.g. lifting and carrying heavy things), and leisure-time activity with aerobic physical activity (e.g. running, cycling, swimming). Muscle-strengthening activity and aerobic physical activity do not generate the same physiological adaptations and health benefits [40, 41], which may partially explain why low-SEP groups show lower cardiovascular health than high-SEP groups, even though they expend more energy in total.
Although the magnitudes of the SEP gradients on domain-specific activities may have changed since 1998 , we assume that the results of the study are still relevant, since the SEP gradient on physical-activity patterns has remained fairly consistent over the years [42, 43].
No causal inferences can be drawn when interpreting these results, since the study relies on cross-sectional data, and physical activity was assessed on the basis of self-reports. The GNHIES98 is a general health survey. It was not possible to assess physical activity in a comprehensive way. Total physical activity was only assessed using five intensity categories of physical activity with a distinction between weekdays and weekend days. The total activity questionnaire therefore produces rather rough estimates of total energy expenditure on a MET basis. As a result, it was decided to use the information obtained to rank individuals – rather than the continuous MET values – as the outcome for analysis. Studies show that physical activity questionnaires overestimate the duration and intensity of physical activity compared to objective measurements such as the accelerometer or the doubly labelled water method [44, 45], and there is evidence to suggest that the internal validity of questionnaires is lower than that of accelerometers when defining health risks based on biomarkers . Social desirability bias – as well as cognitive problems relating to recalling the durations of activity and categorizing the intensity of activities – compromise the internal validity of self-reported physical activity information [47, 48]. Reporting bias is particularly problematic if it differs systematically according to specific characteristics of the respondents, causing differential misclassification bias ; we cannot totally exclude this possibility in this study. The average MET score of total energy expenditure revealed here also seemed to be rather high. This may be partly attributable to the oversampling of formerly East German people in GNHIES98. Individuals who lived in the former East Germany were oversampled by 30% in order to increase the statistical power for East–west comparisons. In 1998, the economic structures of the former East and West Germany were different, since the ‘Fall of the Berlin Wall’ had taken place only ten years earlier in 1989. The economy in the former East Germany had large agricultural and industrial sectors involving high levels of physically demanding work, while the former West German economy had a large service industry sector with sedentary office jobs. In Germany as a whole, the service-industry sector has grown significantly since 1998, although the structural changes have been stronger in the former East than in western Germany. Furthermore, the prevalence of sports activities has also increased in Germany since 1998. A recent trend study demonstrated that the prevalence of performing ‘any’ sports activity had increased in Germany from 56.0% in 1998 to 61.4% in 2003 and 64.6% in 2009 . However, studies also demonstrate that the SEP gradient on physical-activity patterns has not essentially changed in Germany in the last fifteen years [42, 43].
The physical activity questionnaire used in this study differs from those used in other countries, which makes it difficult to compare results .