**3. Data and Methods**

The analyses are based on the BIBB/BAuA Employment Survey 2018, a representative cross-sectional survey covering approximately 20,000 employed individuals in Germany who work at least 10 h per week and are at least 15 years old [50] (see Table A1 for an overview). The BIBB/BAuA Employment Survey is representative for the German labor force, including employed as well as self-employed individuals, and covers various occupational groups. For the analyses, we exclude individuals above the age of 65, as working individuals above that age (age 65 ≈ statutory retirement age in Germany) are a selective and highly heterogeneous group (e.g., self-employed, as well as individuals depending on additional income to increase their pension). Further, the analyses are restricted to individuals with valid data for the included variables. The analysis sample amounts

to 18,554 individuals, although the number of observations varies slightly according to the analyses performed. The sample consists of 54.5% men and 45.5% women, with the majority (51.2%) being aged between 35–54 years (see Table A2 in Appendix A). Before the BIBB/BAuA Employment Survey 2018 has been carried out, it has been inspected and received a positive vote of the ethics commission.

#### *3.1. Variables*

The BIBB/BAuA Employment Survey 2018 includes three different variables indicating cognitive demands: facing new tasks, improving work, and doing unlearned things. We consider all of the three items in order to cover different dimensions of cognitive demands. Although the wording of the questions on cognitive demands is not identical, the indicators are still comparable to those used in previous studies [13,28]. The respondents were asked how often they face these demands during their work and the response scale was frequently, sometimes, rarely, never (see Table 1). In order to keep the analyses simple and reduce complexity, we recode the variables into indicator variables. For new tasks and improving work, these dichotomous variables equal 1 if someone reports to experience a cognitive demand frequently and 0 otherwise. As comparatively few individuals report doing unlearned things frequently, we collapse sometimes and frequently into one category for this variable.



Source: Own calculations based on the BIBB/BAuA Employment Survey 2018, weighted results.

A special feature of the BIBB/BAuA Employment Survey is that individuals who report frequently facing a specific job demand are subsequently asked whether or not they perceive this job demand as stressful. This allows us to take the individual assessment whether the specific cognitive demand is perceived as a resource or rather as a demand into account. For the variable "improve work", the question whether it is stressful or not was not asked, as this question was not defined as a stressor due to its rather positive meaning.

Employee well-being is operationalized by three different outcomes. Work-related fatigue (yes = 1, no = 0) is based on the question of whether the respondent has frequently suffered from emotional exhaustion in the last 12 months during work, on working days, or both. A comparable dichotomous measure of fatigue is collected within the European Working Conditions Survey (EWCS) and has been previously used to study work-related differences in well-being [51]. Moreover, respondents were asked about their general state of health, which was recoded into an indicator variable for self-rated good health (excellent/very good/good = 1, not so good/poor = 0). Self-rated health has been a widely used measure and it was found that it likely considers chronic and acute illnesses [52,53]. Finally, we consider overall job satisfaction (by asking the question: "And now, all things considered, how satisfied are you with your work on the whole?"), recoded into an indicator variable (very satisfied/satisfied = 1, less satisfied/not satisfied = 0), as a measure for overall well-being. Single-item measures of global job satisfaction have been found to be as reliable as multiple-item measures [54].

As additional working conditions, we consider a measure for work intensity (by asking the question: "How frequently does it happen during your work that you have to work under great time pressure or pressure to perform?"), as well as an indicator for the respondent's autonomy at work (by asking the question: "How frequently does it occur that you can plan and arrange your own work

yourself?"). These variables equal 1 if the individuals report frequently facing the working conditions, and 0 otherwise (sometimes/rarely/never).

As control variables, we include gender, three age groups (15–34, 35–54, 55–65), three dummies for schooling (low, intermediate, high), four dummies for the occupational status, as well as five sector dummies (see Table A2 in Appendix A).
