Longitudinal Research in Occupational Health Psychology

Toon Taris, Toon W. Taris

• Publisher: Routledge
• Year: 2017
• Print ISBN: 9781138933460
• eBook ISBN: 9781317391555
• Edition: 1
• Topic: Psychology
• Subtopic: Applied Psychology
• Index: Psychology

Occupational health psychology (OHP) involves the application of psychology to improving the quality of work life and to promoting and protecting the safety, health and well-being of employees. Achieving these aims requires researchers and practitioners to possess in-depth knowledge of the processes that are presumed to bring about the desired outcomes. To date, most studies in OHP have relied on cross-sectional designs in examining these processes. In such designs all variables of interest are measured simultaneously. Although this has generated useful insights in how particular phenomena are associated, such designs cannot be trusted when it comes to drawing causal inferences: association is not causation.

This book therefore focuses on longitudinal research designs in OHP, whereby the concepts of interest are measured several times, offering much stronger evidence for causal relationships. The authors focus on design issues in longitudinal research (such as the number of measurements chosen, and the length of the time lags between these measurements), and illustrate these issues in the context of applied research on topics such as the work-family interface, conflict at work, and employee well-being. By doing so this volume provides a state-of-the-art overview of current research in OHP, both in terms of its findings and methodologies.

This book is based on a special issue of the journal Work & Stress.

Cause and effect: Optimizing the designs of longitudinal studies in occupational health psychology

According to the National Institute for Occupational Safety and Health, occupational health psychology (OHP) involves “the application of psychology to improving the quality of work life, and to protecting and promoting the safety, health and well-being of workers” (NIOSH, 2013). Although not everyone may agree with this definition (is OHP-research only about the application of currently available psychological knowledge to working life, or does it also generate new and even fundamental insights?), to make a difference in workers’ lives is certainly a key concern in our discipline. Achieving this aim requires in-depth knowledge of the causal processes that affect these desiderata: healthy work and healthy workers. In order to obtain such knowledge, we need longitudinal studies in which the same variables are measured at least twice across time for the same set of participants (e.g. Hassett & Paavilainen-Mäntymäki, 2013; Menard, 2007; Ployhart & Vandenberg, 2010). Although cross-sectional designs can tell us whether particular variables are associated in ways that are proposed by theories, longitudinal research designs can also provide us with information about the temporal order of the events underlying these associations, show how the presumed “outcomes” have changed across time and whether this change can be ascribed to (changes in) the alleged “independent” variables. Accordingly, over the last two decades the number of OHP studies examining causal processes through longitudinal research designs has increased steadily (see, for instance, Austin, Scherbaum, & Mahlmann, 2002; Stone-Romero, 2011). A number of such studies are included in this special edition of Work & Stress, which is devoted to longitudinal research.

An intriguing question is whether our understanding of the causal underpinnings of occupational safety, health and well-being has increased at a similar pace. To some extent it has: longitudinal studies have often confirmed and clarified relations that were previously mainly obtained in cross-sectional research. For example, by now we know that high job demands, low support and low control adversely affect well-being longitudinally and, by implication, causally (Häusser, Mojzisch, Niesel, & Schulz-Hardt, 2010). However, longitudinal studies have also produced unexpected and confusing findings, and sometimes have not replicated associations that had been firmly established cross-sectionally. We believe that this is at least partly due to poor a priori consideration of what exactly should be the right longitudinal design for studying the specific causal relations at hand. That is, researchers should carefully consider their choice of research design in the light of theories on the specific relations under study, previous empirical studies on these relations, and practical considerations, paying due attention to each of these three aspects. In this editorial we focus on two pivotal sets of issues that warrant more attention, namely (1) the length of the time intervals employed in longitudinal designs, and (2) the issue of reciprocal (i.e. bi-directional) effects.

Time lags in longitudinal designs

The appropriate length of the time lags between study waves is a crucial issue in longitudinal research methodology. The length of this interval should correspond well with the underlying “true” causal lag (e.g. De Lange, Taris, Kompier, Houtman, & Bongers, 2004; Dormann & Zapf, 2002; Ployhart & Vandenberg, 2010; Zapf, Dormann, & Frese, 1996). If it is much shorter than the true causal lag, chances are that the antecedent has not yet had sufficient time to affect the outcome variable. Conversely, if this interval is too long, the effect of being exposed to the antecedent variable may already have disappeared. To complicate things further, in the intervening period between study waves all kinds of other events may take place that compete with the Time 1 exposure in affecting the outcome variables. Such competing factors may affect the internal validity of the study in that the strength of the effect between predictor and effect will be biased downwards. Further, employees are not just passive recipients of situational stimuli, but may try to change their work situation. Between the study waves, they may, for example, try to change their job content, working conditions or working time arrangements (Wrzesniewski & Dutton, 2001). When time intervals are too short or too long, the chances of detecting the effects of the antecedent variable on the outcome will decrease, as compared to when the study interval corresponds with the true underlying interval. All this implies that (1) the magnitude of the longitudinal effects may vary strongly with the length of the interval used, and (2) that to advance longitudinal research, scholars should carefully consider the possible underlying causal time lags before they conduct their study. When this true time lag is unknown or cannot be reasonably surmised, researchers should preferably employ multiphase designs in which measurements are taken from the same set of participants at several points in time (Ployhart & Vandenberg, 2010; Taris & Kompier, 2003), the lengths of the intervals between measurements being appropriate to the variables under study. We believe that these recommendations are appropriate because as yet the number of study waves and the length of the interval between these waves is often chosen on pragmatic grounds.

Despite the increase in longitudinal studies, to date the two-wave longitudinal design has continued to dominate the research scene, with the possible exception of the diary study (Bolger, Davis, & Rafaeli, 2003). Diary studies tend to focus on relatively volatile processes in which the phenomena of interest (e.g. mood or fatigue) change quickly over time. Typically, a relatively small number of participants completes short questionnaires repeatedly during the day, on a number of consecutive days. Due to this repeated measures design, within-participant changes in the phenomena of interest can be related to the “little experiences of everyday life” that precede these changes (Wheeler & Reis, 1991, p. 340). Similarly, spill-over effects from one day to another can be examined, for example, from working days to a non-work day. Unfortunately, the potential benefits of the diary design in examining causal processes are not always realized. Establishing temporal order requires that the outcome variable (e.g. well-being) is related not only to a particular predictor (e.g. recovery experiences) as measured earlier that day or on the previous day, but also to well-being as measured earlier (e.g. on the previous day). If this requirement is not met, diary designs are not more informative regarding the development of causal processes than a purely cross-sectional approach (Kelloway & Francis, 2013).

At present, when the research focus is not on explaining changes in day-to-day experiences, the two-wave longitudinal design remains the most frequently used approach. Thus, the choice of a particular interval between the study waves becomes paramount. Dormann and Zapf (2002) and De Lange et al. (2004) conducted studies using multi-wave longitudinal designs in order to examine how the strength of the effects in longitudinal research depended on the length of the interval between the study waves. Dormann and Zapf (2002), comparing findings over one-year, two-year and four-year intervals, found that it took at least two years for the longitudinal effects of stressors at work on mental health (measured as depressive symptoms mediated by irritation) to be demonstrated and that they were weakest at one year. Similarly, De Lange et al. (2004), using time lags of one, two and three years, reported that the longitudinal effects between job stressors and health (depression, emotional exhaustion and job satisfaction) were strongest for a one-year interval. By the end of 2013 these two studies had jointly received nearly 500 citations, but in the light of the above discussion their value as a general foundation for the choice for a particular time interval is limited. Not only do their findings regarding the optimal interval between the study waves differ, but it is also likely that their findings do not generalize beyond the type of stressors and outcomes studied. This reasoning follows from a seminal study by Frese and Zapf (1994) who distinguished among five types of developmental trajectories for the processes typically studied in OHP. For example, workers may get used to some stressors quickly, meaning that the effects of exposure to these stressors may be short-lived and that a short interval between the study waves is required to detect these effects. Other work factors may affect health only in the long run, and are referred to by Ford et al. (2014) in the first paper in this special edition of Work & Stress as lagged effects. Clearly, simple general rules of thumb regarding the appropriate length of the time interval for a study do not exist. This implies that any sensible choice for a particular time interval between the study waves must necessarily take into account the type of causes and consequences being investigated, as well as consider the development and context of the process that is being examined.

Reciprocal effects

Normal and reversed effects constitute another issue in longitudinal research. In a previous editorial on longitudinal designs (Taris & Kompier, 2003) we recommended researchers to more often employ full panel designs: designs in which both the presumed “outcomes” and “explanatory” variables are assessed at all study waves. We believe that, over the last decade, one of the strong developments in OHP has been the increase in the number of studies employing these designs. Full panel designs allow for the examination of both normal and reversed effects. Normal effects usually refer to the lagged effects of job characteristics on safety, health, well-being and performance-related variables. Reversed effects relate to effects of the latter categories of variables on job characteristics. Inspired by the seminal work of Zapf et al. (1996), De Lange et al. (2004) discussed a number of mechanisms that could account for reversed effects. They argued that there are multiple reasons why employees who report high levels of ill health at Time 1 could report higher levels of job demands at Time 2. For example, such employees (assuming that their coping capacity is limited) may only perceive their jobs as having become more demanding (i.e. the demands themselves did not change). It is also possible that ill health causes a “drift” towards an objectively more demanding job (see, for instance, Kompier & Taris, 2011, for a discussion). When a study supports both normal and reversed effects, researchers speak of reciprocal effects.

The increase in the number of longitudinal studies in OHP during the last decade went hand in hand with a more frequent examination of possible reversed and reciprocal effects (among others, De Jonge et al., 2001; De Lange et al., 2004; Rodriguez-Munoz, Sanz-Vergel, Demerouti, & Bakker, 2012). In itself this is a positive development. A strong example is a study by Finne, Knardahl, and Lau (2011) on workplace bullying and mental distress. Whereas self-reported workplace bullying predicted mental distress two years later (signifying normal causation), mental distress also predicted bullying across time (indicating reverse causation). Apparently, the associations between workplace bullying and mental distress constitute a vicious circle, with bullying leading to mental problems and the latter leading to higher levels of bullying, and so forth. However, such exemplary studies notwithstanding, as yet we do lack a more systematic integration of such effects into OHP theory, i.e. into the body of knowledge on the processes relating work characteristics to worker health, well-being, safety and performance. Currently a fairly impressionistic picture of these reversed/reciprocal effects emerges: sometimes such effects are absent, sometimes present (and in that case researchers are happy to report and interpret them), but we are still in need of a more integrated theory that describes (1) when and (2) what sort of effects can be expected (3) for whom in (4) which circumstances and (5) what the specific processes responsible for such effects could be. We might add that even the terms “normal” versus “reversed” effects themselves are problematic: What is normal about “normal” causality? This label derives from the fact that researchers “normally” examine the effects of work stressors on strains, rather than that it reflects a characteristic of these effects themselves. Consequently, it might be more accurate to speak of “stressor-to-strain effects” instead of normal causality, and of “strain-to-stressor effects” in the instance of reversed effects.

Another notable exception to the above observation that longitudinal studies have served to clarify the causal processes underlying occupational safety, health and well-being concerns the body of research on gain and loss spirals in OHP. Building on Fredrickson’s (2001) broaden-and-build theory of positive emotions, through longitudinal research it has been found that the presence of job resources (such as high levels of autonomy and support) tends to promote well-being (such as high levels of work engagement) over time, and that high levels of engagement in turn lead to high levels of job resources (see Salanova, Schaufeli, Xanthopoulou, & Bakker, 2010, for an overview). This research suggests that engaged workers tend to collect more resources in their job, leading to even higher levels of engagement (the so-called gain spiral), whereas low-engagement workers tend to...

Table of contents

• Cover
• Half Title
• Title
• Copyright
• Contents
• Citation Information
• 1. Cause and effect: Optimizing the designs of longitudinal studies in occupational health psychology
• 2. How do occupational stressor-strain effects vary with time? A review and meta-analysis of the relevance of time lags in longitudinal studies
• 3. The effect of conflict at work on well-being: Depressive symptoms as a vulnerability factor
• 4. Enjoyment and absorption: An electronic diary study on daily flow patterns
• 5. Effects of vacation from work on health and well-being: Lots of fun, quickly gone
• 6. Work to non-work enrichment: The mediating roles of positive affect and positive work reflection
• 7. Do you want me to be perfect? Two longitudinal studies on socially prescribed perfectionism, stress and burnout in the workplace
• 8. A participative intervention to improve employee well-being in knowledge work jobs: A mixed-methods evaluation study
• 9. Should I stay or should I go? Examining longitudinal relations among job resources and work engagement for stayers versus movers
• 10. Are job and personal resources associated with work ability 10 years later? The mediating role of work engagement
• Index