Interactive effects of disruptive environmental events and job characteristics on job engagement: Integrating event- and feature-oriented approaches

Characteristics and consequences of disruptive events

Major disruptive events, such as natural disasters, wars, or the local outbreak of diseases, are discrete, external incidents that entail the interaction of different entities (e.g. individuals, teams, environment). EST describes the nature of such events and explains their effects on different outcomes in work and organizational contexts (Morgeson et al., 2015). According to EST, the impact of an event depends on the interplay between three factors: event strength, space, and time. First, an event’s strength is defined as the extent to which an event causes change and requires attention, and is determined by its novelty, disruptiveness, and criticality. Event novelty refers to the extent to which an event is unexpected or new, that is, how much it differs from previous events. An event’s disruptiveness reflects how much it leads to changes and a discontinuity in people’s environment. Criticality reflects the meaningfulness and importance of an event for individuals and, thus, affects the degree to which they will attend to and react to the event (Morgeson et al., 2015; Morgeson and DeRue, 2006). Second, event space entails both the origin at a specific hierarchical level, such as the individual, team, organization, or broader environment, and the direction of an event’s effects. For example, events at the broader environmental level can have top-down effects on organizations or individuals (Tilcsik and Marquis, 2013). These effects include potential direct effects on outcomes at the organization, team, or individual levels, as well as moderating effects that affect the strength of the relationship between variables at one of these levels (Morgeson et al., 2015). Finally, event time includes the timing and duration of an event, as well as the dynamic change of event strength over time (Morgeson et al., 2015). Empirical studies largely support the propositions of EST regarding these event characteristics (e.g. Lin et al., 2021; Liu et al., 2021; McCarthy et al., 2021).

A key methodological approach to research transitions and individuals’ adaptation to transitions are discontinuous growth models (DGM; Bliese et al., 2017, 2020). To this end, Bliese et al. (2017) suggest a framework for researching dynamic changes in employees’ experiences and behaviors associated with the occurrence of disruptive events. In DGMs, time-related covariates are specified to reflect changes in the outcome during different phases of an event period. Specifically, DGMs include three parameters. First, the pre-transition slope reflects changes in the outcome in the period before the event. Second, the transition parameter is conceptualized as a discrete change from one timepoint to the next and contrasts observations before and after the transition into the event. Third, the recovery slope represents the subsequent, post-transition changes in the outcome (e.g. Bliese et al., 2017; Lang and Bliese, 2009). These parameters are useful for testing assumptions based on EST (see e.g. McFarland et al., 2020; Zacher et al., 2021), as they allow the investigation of change patterns caused by events, as well as individual differences in these changes (Bliese et al., 2017).

To test our hypotheses in the present study, we examine the effects of two unprecedented national lockdowns during the COVID-19 pandemic in Germany, which exemplify strong, disruptive events at the environmental level (e.g. Ménard et al., 2021; Sibley et al., 2020; Zacher et al., 2021). A national lockdown was first introduced in Germany in the early phase of the pandemic (i.e. March 2020; see Figure 2), thus representing a novel event, and was followed by a second lockdown that started in December 2020. Both lockdowns were highly critical and disruptive events that were characterized by severe restrictions of public and private life. These included bans of public gatherings, closures of public facilities and businesses, and personal contact restrictions, that were aimed at containing the spread of the coronavirus (Steinmetz et al., 2020). In this study, we examine the top-down influence of these lockdowns as novel, disruptive, and critical events at the environmental level on employees’ job engagement (i.e. an individual-level outcome). With a focus on the events’ novelty, we additionally compare employees’ reactivity to two consecutive lockdowns. Furthermore, we investigate if job characteristics measured prior to event onset, such as employees’ job autonomy and support from supervisors and coworkers, may play a role in employees’ reactivity and mitigate or strengthen the impact of the lockdowns on employees’ job engagement.

OPEN IN VIEWER

Figure 2.

Restrictions related to COVID-19 in Germany from March 2020 to September 2021.

Declines in job engagement in reaction to disruptive environmental events

Job engagement reflects employees’ affective-motivational state at work, including their attention to tasks and presence in the work role (Kahn, 1990; Rothbard and Patil, 2011). Based on Rich et al. (2010), who extended earlier work by Kahn (1990), we adopt a definition of job engagement as the simultaneous investment of physical, cognitive, and emotional energy in the work role. More specifically, physical engagement refers to employees being actively invested in their work and striving to complete their tasks. Second, cognitive engagement reflects the degree to which employees pay attention to and are absorbed by their work. Third, emotional engagement encompasses emotional states with positive valence and high activation that employees experience in their work role, such as feeling enthusiastic and energetic (Rich et al., 2010; see also Kahn, 1990). As an affective-motivational construct, job engagement has been shown to predict a number of important employee outcomes, including task performance and organizational citizenship behavior (e.g. Christian et al., 2011; Rich et al., 2010).

Drawing from EST (Morgeson et al., 2015), we expect that the transition into disruptive environmental events, such as national lockdowns, leads to declines in physical, cognitive, and emotional job engagement. Strong disruptive events that reside at a high hierarchical level may have top-down effects on employees in several ways. First, strong environmental events disrupt routines and established procedures. Particularly when facing highly novel events, such as the first national lockdown investigated in the present study, employees unlikely have behavioral scripts available for reacting to and dealing with these events (Morgeson et al., 2015; Morgeson and DeRue, 2006). Consequently, the transition into such an event requires the adaptation of behaviors or development of new behaviors and routines (Lang et al., 2021; Morgeson et al., 2015), which may detrimentally affect employees’ physical job engagement.

Second, the onset of disruptive environmental events requires high levels of additional information processing to comprehend the nature and causes of the event. This information processing distracts employees’ attention away from their job and core tasks (Morgeson et al., 2015) and can therefore be expected to detrimentally affect their cognitive job engagement. Third, disruptive environmental events, such as the national lockdowns, are likely perceived as a potential threat and may trigger negative emotions, such as fear and anxiety (e.g. Martinelli et al., 2020; Marzana et al., 2022). Employees therefore have to emotionally process and cope with such events (e.g. Rodas et al., 2022), which may negatively impact their emotional engagement at work. In summary, the occurrence of disruptive environmental events leads to a number of changes in individuals’ environment and requires them to master and spend energy on additional demands, such as directing their attention to the event, developing new behaviors and routines, and processing the event as a potential threat. As a result, we expect a decrease in employees’ physical, cognitive, and emotional job engagement due to their transitions into these events.

To date, few studies have investigated changes in job engagement in response to disruptive environmental events. A time-lagged field study by Liu et al. (2021) examined employees’ job engagement during the COVID-19 pandemic. They observed a negative effect of perceived COVID-19 crisis strength on job engagement, measured with a 3-week time lag, which provides initial support for the effect of major disruptive events on job engagement proposed by EST. Other studies have provided inconclusive results regarding this relationship (e.g. Kaltiainen and Hakanen, 2022; Oksa et al., 2023; Reinwald et al., 2021). Importantly, fully assessing and understanding changes in job engagement caused by the transition into a disruptive event requires repeated and regular measurements of job engagement as well as pre-event baseline assessments (Bliese et al., 2017). The present study uses such a design to test the hypothesized declines in employees’ physical, cognitive, and emotional job engagement caused by the transitions into disruptive environmental events.

Increases in job engagement in post-transition phases

After the transitions into the two national lockdowns, we expect gradual increases in employees’ job engagement during the post-transition phases (i.e. from April to December 2020 and from January to September 2021, respectively). In order to fully map these expected dynamic changes in job engagement associated with disruptive environmental events, we integrate EST with the transition-adaptation framework by Bliese et al. (2017). This framework outlines possible processes in the post-event period, the so-called “recovery trajectory,” which may include improvements due to recovery processes but also continued and stronger declines, for instance after a traumatic event (Bliese et al., 2017).

We propose that the onsets of disruptive environmental events are followed by a post-transition phase during which individuals can adapt to the changed environmental conditions caused by the events and gradually return to pre-event levels of functioning (e.g. Lang et al., 2021). First, the successful implementation of the new behaviors and routines (Morgeson et al., 2015) developed during the transition into the event should allow employees’ physical job engagement to recover. For example, after the transitions into the national lockdowns, employees may apply strategies to successfully work from home or have incorporated additional hygiene rules into their daily routines (e.g. Troll et al., 2022). Second, after the transition phase, employees should have collected the necessary information about the event and the resulting changes in the environment (Morgeson et al., 2015), allowing their cognitive job engagement to increase. Finally, once employees have understood and successfully adjusted to the event, it should be less threatening to them, allowing them to be more emotionally present in their work role. Overall, employees may gradually be less affected by the changes in their immediate environment that were caused by the lockdowns, which allows their physical, cognitive, and emotional job engagement levels to recover.

This reasoning is supported by previous empirical studies that observed recovery patterns in other employee outcomes following the transition into disruptive events. For example, recovery dynamics were studied by Zhu et al. (2016), who observed an increase in work adjustment over the first 9 months of new international assignments. In a more recent study, Zacher et al. (2021) observed an increase in employees’ self-reported work performance following a transition caused by the first national pandemic lockdown in Germany. In summary, due to successful adaptation to the changes caused by the disruptive environmental events, employees’ physical, cognitive, and emotional job engagement can be expected to gradually recover in the periods following their transitions into the events.

Recurring environmental events and the role of event novelty

In addition to employees’ reactivity to one-time events, repeated events and possible differences in their effects should be examined. Based on EST, a recurring event can be expected to be less novel and, consequently, have weaker effects on employees’ experiences and behaviors (Morgeson et al., 2015). First, the behavioral scripts and routines developed when facing a disruptive environmental event for the first time will be readily available to employees when the same type of event recurs (Morgeson et al., 2015). Thus, their physical job engagement should be less affected. For example, an event may lead to changes in employees’ working conditions, such as having to suddenly collaborate with colleagues virtually instead of face-to-face. Adapting to these changes for the first time (Klostermann et al., 2021; Liebermann et al., 2021) requires developing and establishing new skills and work routines (Ohly et al., 2017). By contrast, when facing the changes caused by an event for the second time (e.g. due to a second lockdown), these skills and routines are readily available to employees, and the cost of adapting to virtual work may be diminished.

Second, the effort required to search for and process information about an event is lower if it is already familiar (Morgeson et al., 2015). When employees are confronted with a disruptive environmental event for the second time, they should therefore be able to maintain higher levels of cognitive job engagement compared to the first event occurrence. Third, employees are likely to experience a familiar type of event as less threatening and may be able to build on personal resources (e.g. coping skills) that they developed when they were first confronted with the event (e.g. Zacher and Rudolph, 2021). Consequently, their emotional job engagement will be less impacted when the event recurs. Thus, whereas employees’ physical, cognitive, and emotional job engagement may suffer considerably in the face of a highly novel environmental event, this effect may be weaker when facing this event for the second time.

Some research supports this idea based on EST by showing that subjective ratings of event strength, including event novelty, were negatively associated with job engagement during the COVID-19 pandemic (Liu et al., 2021). However, the effects of individual differences in subjective perceptions of event novelty might not be identical and directly transferable to recurring environmental events, and research on other types of repeated events has been inconclusive so far (e.g. Luhmann and Eid, 2009; Seery et al., 2010). In the present study, we therefore examine employees’ reactivity to two consecutive disruptive events at the environmental level. Based on EST, we expect that the second, less novel event had a weaker impact on employees’ physical, cognitive, and emotional job engagement.

The moderating role of job characteristics

Disruptive environmental events require employees to adapt to changing circumstances, for example by developing new ways of working (Bliese et al., 2017; Morgeson et al., 2015). The degree to which adaptation to these events is necessary and feasible likely varies across jobs and job contexts. Accordingly, certain job characteristics may facilitate or impede employees’ coping with the transition.

EST describes two different types of moderation effects. First, events may moderate associations between behaviors, features, and other events (e.g. an organization’s bankruptcy may weaken the effect of social support on turnover intentions; Morgeson et al., 2015). Second, Morgeson et al. (2015) propose that event-oriented research should be further integrated with feature-oriented approaches, such as research on job or person characteristics, and that such features may moderate effects of events on outcomes at the individual level. However, additional integrative theory building is needed for describing the potential moderating effects of job characteristics in the context of events, and empirical research based on EST has not yet investigated them (Morgeson et al., 2015). We integrate EST and the job demands-resources model (Bakker and Demerouti, 2017; Demerouti et al., 2001) to develop and test specific hypotheses about the moderating effects of key job characteristics.

Based on the job demands-resources model, we expect that job resources support employees in adapting to the changes caused by disruptive environmental events and, thus, buffer the negative effects on employees’ job engagement. By contrast, high job demands are expected to diminish employees’ capacity and energy to effectively cope with these events, and therefore may amplify the declines in job engagement. In this study, we examine the potential moderating role of four important job demands and resources that have been studied extensively in the organizational literature, namely workload, job autonomy, and supervisor and coworker social support (De Lange et al., 2003). Research has studied these variables in relation to occupational wellbeing outcomes, but not as moderators in the context of disruptive environmental events (e.g. Crawford et al., 2010; Häusser et al., 2010).

We expect that a high (vs low) workload immediately before event onset amplifies the declines in job engagement due to the transitions into the disruptive environmental events. By contrast, we assume that high (vs low) pre-event levels of job autonomy and social support mitigate these declines. Figure 1 illustrates the expected trajectories in job engagement for high and low levels of workload, job autonomy, and social support. Importantly, the focus of this study is on pre-event levels of job characteristics rather than their possible changes during disruptive events. This approach allows examining whether job characteristics immediately before the onset of a disruptive environmental event influence employees’ reactivity and, thus, whether jobs can be designed to support successful adaptation to such events.

Transparency and open materials

The data and R code to reproduce the analyses, and complete results are available in our supplemental materials.

The data used in this paper were collected beginning in December 2019 as part of a larger longitudinal data collection effort that began approximately 3 months before the World Health Organization declared COVID-19 a pandemic in mid-March 2020. The second wave of survey data were collected at the beginning of March 2020, just before the first national lockdown in Germany, and the study was adapted in April 2020 to focus on employees’ monthly experiences and behaviors in the context of the pandemic. Several articles based on the same dataset, but with completely different research questions and completely different, non-overlapping substantive variables than the current study, have been published (see transparency matrix in Table S1 in the supplemental materials).

Context and design of the study

The specific disruptive environmental events examined in this study are the two national lockdowns in Germany in the context of the COVID-19 pandemic that were characterized by several nationwide public health measures to contain the spread of the coronavirus (SARS-CoV-2). They included severe restrictions of public life (e.g. banning in-person gatherings), the closure of public facilities and businesses (e.g. daycares, schools, hairdressers), and contact restrictions, such as limiting the number of personal contacts (Steinmetz et al., 2020).

A national lockdown was introduced in Germany for the first time in the early phase of the pandemic (i.e. March 2020; see Figure 2). Therefore, this event was highly novel, sudden, and unexpected. The first lockdown lasted about 2 months. Due to a high number of coronavirus infections, hospitalizations, and deaths, it was followed by a second national lockdown that began in December 2020, lasting about 5 months. The two national lockdowns caused a variety of significant changes in the environment and a disruption of people’s routines in several life domains and, thus, can be classified as disruptive. Finally, the lockdowns were highly critical, given the multitude of changes they caused and the variety of goals they made harder or impossible to attain (e.g. by increasing work-family conflict and job insecurity; e.g. Meyer et al., 2021). Overall, both national lockdowns can be classified as strong events that, although they differed in terms of novelty, can be expected to have had considerable top-down effects on employees.

In the present study, we investigate a pre-event period (i.e. December 2019–March 2020), the onset of the first and second national lockdowns in March 2020 and December 2020, as well as subsequent adaptation periods from April 2020–December 2020 and from January to September 2021, respectively (Steinmetz et al., 2020; see Figure 2). Job engagement was measured at all 20 measurement waves, beginning in December 2019 (Time [T]1). Starting in March 2020 (T2), data were collected at the beginning of each month (April 2020 = T3, May 2020 = T4, etc. until September 2021 = T20). Participants were asked to rate their job engagement during the previous 3 months at the first and second measurement waves (T1 and T2) and during the previous month at all subsequent waves (T3–T20). Accordingly, the time reference of the self-reports initially was 3 months and then changed to 1 month. Job characteristics (i.e. workload, job autonomy, supervisor support, and coworker support) were also measured at each of these waves, but only pre-event onset measurements collected at the beginnings of March 2020 and December 2020 (i.e. at T2 and T11) were included in the present analyses (for an overview of all measurement waves and the corresponding time coding, see Table S2 in the supplemental materials).

Participants

The data were collected online from a large employee sample in Germany by a professional, ISO 26362 certified panel company, which ensures the quality of the data. To be eligible to participate, respondents had to be at least 18 years old and be working full-time at each measurement wave. At the first measurement wave, 4,839 persons from the panel company’s database were invited to participate in the study. Of these persons, 1,848 initiated the first survey and provided all or at least partial responses (e.g. demographics, substantive variables; response rate of 38.19%). All of these respondents were invited to participate in each of the following waves.

In the present study, we only include participants who provided data on workload, job autonomy, and social support variables in March 2020 and December 2020 (i.e. the moderators included in our analyses). The resulting sample size was n = 652. Participants’ ages ranged from 21 to 69 years, with a mean age of 45.77 years (SD = 10.52), and 38.34% were female. Most participants (34.66%) held a higher educational degree and most worked in public service (13.04%) or manufacturing (11.81%). For further details, see Table S3 in the supplemental materials.

The multilevel models used to test our hypotheses were calculated based on a mean of 17.46 time points per person. They are well suited for handling missing data on endogenous outcomes (Singer and Willett, 2003). To address systematic patterns of attrition, incomplete responders (n = 404), who did not provide information on all demographic and substantive variables at all time points, were compared to complete responders (n = 248) on key demographic and substantive variables. There were no differences in these variables between complete and incomplete responders (p > 0.05), except for cognitive engagement, which was slightly higher in the group of complete responders (M = 5.26, SD = 1.04) compared to incomplete responders (M = 5.00, SD = 1.01; see Table S3).

Measures

We used the following self-report measures in this study. All measures were worded in past tense and referred to the past 3 months (T1 and T2) or the past month (T3–T20). Means, standard deviations, and internal scale consistency estimates (within- and between-person) are reported in Table 1.

OPEN IN VIEWER

Table 1.

Within- and Between-Person Descriptive Statistics and Correlations Between Variables.

Variable	M	SDlvl 1	SDlvl 2	1	2	3	4	5	6	7	8	9	10
1. Age	45.77	-	10.52	-	−0.04	−0.08	0.15	0.14	−0.05	−0.01	0.17	−0.07	−0.04
2. Sex (1 = Female)	0.39	-	0.49		-	−0.09	0.08	0.01	−0.03	0.06	−0.08	0.01	0.02
3. Waves participated	17.35	-	3.35			-	0.04	0.10	0.09	−0.01	0.02	0.01	0.01
4. Physical engagement	5.34	0.82	0.92				0.93/0.81	0.86	0.53	0.57	0.45	0.37	0.44
5. Cognitive engagement	5.10	0.84	1.01				0.65	0.97/0.87	0.66	0.52	0.49	0.41	0.44
6. Emotional engagement	4.51	0.83	1.17				0.55	0.59	0.97/0.87	0.17	0.51	0.52	0.50
7. Workload	4.27	0.81	1.18				0.32	0.31	0.19	0.95/0.83	0.08	0.13	0.24
8. Job autonomy	4.92	0.86	1.29				0.26	0.25	0.27	0.15	0.99/0.90	0.40	0.44
9. Supervisor support	4.14	0.86	1.22				0.22	0.23	0.25	0.13	0.28	0.94/0.86	0.78
10. Coworker support	4.47	0.77	1.07				0.28	0.27	0.28	0.18	0.29	0.52	0.93/0.85
ICC(1)							0.56	0.59	0.67	0.68	0.69	0.67	0.66
ICC(2)							0.96	0.97	0.98	0.98	0.98	0.98	0.98

Correlations ⩾ |0.08| are significant at p < 0.05. Lower triangular: Within-group (level 1). Upper triangular: Between-group (level 2). Boldface values: McDonald’s Omega, between/within-person. Number of people = 652. Number of observations = 11,313.

Analytical strategy

In the first step, we examined the reliability of the focal measures. We conducted multilevel confirmatory factor analyses (MCFA) to account for the nesting of the data and specified these models as homologous across levels of analysis (i.e. with the same factor structure at the within-person and between-person level of analysis; Chen et al., 2005). We examined MCFA models using the “lavaan” package (Rosseel, 2012) for R (R Core Team, 2019). The robust maximum likelihood estimator (MLR) was applied to account for non-normally distributed data. To determine whether job engagement should be analyzed using an overall, composite score or by differentiating the three facets of job engagement, we tested a 7-factor model against a 5-factor model and found that the 7-factor model (CFI = 0.932, TLI = 0.917, RMSEA = 0.049, SRMR _within = 0.031, and SRMR _between = 0.084, Χ² (418) = 7,620.77, p < 0.001) fit the data better than the alternative 5-factor model (Δχ² > 900, p < 0.05).

A common approach to evaluating model fit is to apply cutoff values for fit indices. For example, the study by Hu and Bentler (1999) suggests a cutoff value close to 0.95 (or higher) for CFI and TLI, close to 0.08 (or lower) for SRMR, and a cutoff value close to 0.06 (or lower) for RMSEA. However, the application of these thresholds can be problematic as they are not generalizable to other forms of misspecification than in the simulated model (e.g. Ropovik, 2015). In addition, more recent simulation studies have suggested that higher thresholds may be more appropriate (Schermelleh-Engel et al., 2003). These results, as well as the significant chi-square test, may indicate that the 7-factor model did not have an optimal fit. We therefore further examined our model with regard to factor loadings, item wordings, and standardized residuals (please see additional information on p. 8 of the supplemental materials).

In the second step, we ran DGMs applying a multilevel regression framework (Bliese et al., 2020; Bliese and Ployhart, 2002). We specified a set of time contrasts (i.e. time-related covariates) to model trajectories of job engagement, namely (1) a pre-event trajectory from December 2019 to March 2020 (2) a first event-transition slope between March and April 2020, (3) a linear first event recovery slope between April and December 2020, (4) a second event-transition slope between December 2020 and January 2021, and (5) a linear second event recovery slope between January 2021 and September 2021. We present the detailed time contrasts applied in Table S2 in the supplemental materials. We applied absolute rather than relative coding of time (Bliese and Lang, 2016). In other words, a specific time contrast ceased to “count up” once the next phase started (i.e. the next time variable started increasing). Applying absolute rather than relative time coding is advantageous for interpreting regression coefficients. For instance, the first transition slope reflects the difference between job engagement in December 2019–February 2020 (T2, measured in retrospect at the beginning of March) and March (T3, measured in retrospect at the beginning of April). The transition slope therefore represents the changes in job engagement that occurred at the onset of the first national lockdown on March 16, 2020. Based on the results of the MCFA and given that the multilevel regression-based approach allows studying one outcome at a time, we specified three DGMs: one each for physical, cognitive, and emotional engagement. The unconditional DGM excluding covariates provides a description of the trajectories of the focal variables across the period studied. We compared transition parameters of the two lockdowns using Wald Chi-Square Tests.

In a final step, we examined the interplay of event occurrence and job characteristics immediately prior to these events. We added grand mean centered workload, job autonomy, supervisor support, and coworker support as between-person level covariates and specified cross-level interactions of transition by each job characteristic. We focused on job characteristics as reported in the survey wave preceding each of the transitions. For example, we referred to workload at T2 to predict variance in the first event-transition slope (changes in engagement experience from T2 to T3) and included workload at T11 to predict changes in job engagement from T11 to T12. Given the large sample size, we specified all interactions for both transitions in one model as a conservative test. Evidence of significant cross-level interactions would reflect that a specific job characteristic measured immediately before the transition predicts more or less pronounced changes in job engagement upon transitioning into these events.

Intercorrelations, measurement invariance, and confirmatory factor analyses

Descriptives, intercorrelations, and internal consistency estimates of our study variables can be found in Table 1.

Before running the focal analyses on changes in job engagement, we conducted MCFA. We found that a homologous 7-factor model (physical engagement, cognitive engagement, emotional engagement, workload, job autonomy, social support from supervisors, social support from colleagues) fit the data better than the alternative 5-factor model (Δχ² > 900, p < 0.05). We present fit indices and comparisons across models in Table S4, as well as additional information on model fit in the supplemental materials. Based on these results, we conducted separate analyses for each facet of job engagement and a supplemental analysis for the overall construct. Factor loadings ranged from 0.62 to 0.91 at the within-person level. Multilevel McDonald’s Omega ranged from ω_within = 0.82 (physical engagement) to 0.90 (job autonomy), suggesting that all scales were reliable.

Establishing measurement invariance over time is an important consideration for longitudinal data analysis (e.g. Putnick and Bornstein, 2016; Vandenberg and Lance, 2000). To establish measurement invariance, we specified a series of increasingly restrictive longitudinal confirmatory factor analysis (CFA) models with autocorrelated error terms for each of the job engagement facets and the job characteristics included in our analyses. A robust maximum likelihood estimator was applied to account for non-normally distributed data, and full information maximum likelihood (FIML) estimation was used to account for missing data. Table S6 in the supplemental materials shows the results of the configural (i.e. free factor loadings) and metric equivalence analyses (i.e. factor loadings invariant). With exception of workload, job autonomy, and coworker support, constraining the factor loadings to be equal across time in the metric equivalence analyses did not substantially affect model fit. Changes in CFI, RMSEA, and SRMR were lower than recommended cutoff values of ΔCFI greater than −0.010, ΔRMSEA smaller than 0.015, and ΔSRMR smaller than 0.030 (Chen, 2007; Cheung and Rensvold, 2002). These results suggest that the measures were equivalent across time points. For workload, job autonomy, and coworker support, we additionally conducted partial measurement invariance analyses. While freeing the factor loadings of one item at a time (as well as two items at a time for coworker support, measured with four items), a series of increasingly restrictive longitudinal CFA models was conducted. For each of these variables, partial measurement invariance was upheld when freeing loadings for certain items (i.e. items 1 or 2 for workload, item 1 for job autonomy, and items 2 or 3, items 1 and 2, items 1 and 3, items 2 and 3, items 2 and 4, and items 3 and 4 for coworker support; see Table S7). To further examine these constructs, we additionally plotted the factor loadings and 95% confidence intervals for all time points for the configural and metric invariance models (please see Figures S4–S6 in the supplemental materials). In these plots, no major deviations of the factor loadings in the configural from the metric invariance model were apparent. Thus, these observations do not point to substantial changes regarding the meaning of the items.

In Table 1, we present the intra-class correlation coefficients across the focal scales. The ICC(1) ranged from 0.56 for physical engagement to 0.69 for job autonomy. Hence, there was considerable variance both at the within-person level and at the between-person level of analysis, and our multilevel modeling approach is adequate to account for nesting of the data.

Hypothesis tests

DGMs predicting physical, cognitive, and emotional engagement can be found in Tables 2–4, respectively. The supplemental analysis for overall job engagement is reported in Table S8. In these models, the reported unstandardized coefficients correspond to changes in engagement for every month of the respective period on a 7-point scale. In each table, Model 1 is the unconditional DGM, Model 2 includes the job characteristics (workload, autonomy, and social support) before the onset of both lockdowns as predictors, and Model 3 contains the interaction of these variables with the transition effects of lockdowns 1 and 2.

OPEN IN VIEWER

Table 2.

Summary of discontinuous growth models predicting physical engagement.

Predictor	Model 1					Model 2					Model 3
	Β	SE	t	df	p	Β	SE	t	df	p	Β	SE	t	df	p
Intercept	5.590	0.042	131.838	10,726	<0.001	5.590	0.037	150.013	10,726	<0.001	5.590	0.037	149.668	10,718	<0.001
Level 1 (within-person)
Pre-Event Period	−0.014	0.014	−0.990	10,726	0.322	−0.014	0.014	−0.988	10,726	0.323	−0.014	0.014	−0.987	10,718	0.323
Lockdown1	−0.156	0.040	−3.867	10,726	<0.001	−0.156	0.040	−3.905	10,726	<0.001	−0.155	0.040	−3.919	10,718	<0.001
Recovery1	−0.015	0.004	−3.657	10,726	<0.001	−0.015	0.004	−3.644	10,726	<0.001	−0.015	0.004	−3.652	10,718	<0.001
Lockdown2	0.029	0.033	0.873	10,726	0.383	0.030	0.033	0.889	10,726	0.374	0.029	0.033	0.902	10,718	0.367
Recovery2	−0.009	0.004	−2.129	10,726	0.033	−0.009	0.004	−2.074	10,726	0.038	−0.009	0.004	−2.070	10,718	0.038
Level 2 (between-person)
Workload Pre Lockdown1						0.206	0.028	7.440	643	<0.001	0.237	0.031	7.711	643	<0.001
Workload Pre Lockdown2						0.114	0.023	4.978	643	<0.001	0.110	0.023	4.793	643	<0.001
Job Autonomy Pre Lockdown1						0.111	0.025	4.440	643	<0.001	0.110	0.028	3.972	643	<0.001
Job Autonomy Pre Lockdown2						0.083	0.024	3.479	643	0.001	0.080	0.024	3.364	643	0.001
Supervisor Support Pre Lockdown1						0.014	0.027	0.524	643	0.601	<0.001	0.030	0.015	643	0.988
Supervisor Support Pre Lockdown2						<0.001	0.028	0.015	643	0.988	0.002	0.028	0.055	643	0.956
Coworker Support Pre Lockdown1						0.056	0.033	1.693	643	0.091	0.050	0.038	1.326	643	0.185
Coworker Support Pre Lockdown2						0.053	0.029	1.828	643	0.068	0.048	0.029	1.635	643	0.103
Interactions
Workload Pre Lockdown1 × Lockdown1											−0.061	0.025	−2.463	10,718	0.014
Workload Pre Lockdown2 × Lockdown2											0.056	0.016	3.567	10,718	<0.001
Job Autonomy Pre Lockdown1 × Lockdown1											0.001	0.022	0.055	10,718	0.956
Job Autonomy Pre Lockdown2 × Lockdown2											0.035	0.016	2.151	10,718	0.032
Supervisor Support Pre Lockdown1 × Lockdown1											0.026	0.026	0.993	10,718	0.321
Supervisor Support Pre Lockdown2 × Lockdown2											−0.006	0.022	−0.273	10,718	0.785
Coworker Support Pre Lockdown1 × Lockdown1											0.010	0.034	0.294	10,718	0.769
Coworker Support Pre Lockdown2 × Lockdown2											0.061	0.024	2.600	10,718	0.009
	Model 1					Model 2					Model 3
Variance Components
Intercept	0.59					0.32					0.32
Lockdown1	0.23					0.21					0.20
Lockdown2	0.20					0.21					0.18
Residual	0.58					0.59					0.59
Deviance (df)	29,093.94 (13)					28,785.45 (18)					28,734.30 (26)
AIC	29,119.94					28,821.45					28,786.30
BIC	29,215.36					28,953.57					28,977.13
R 2 within	0.13					0.13					0.13

Models fitted using maximum likelihood estimates. Level-2 predictors were grand mean centered. Deviance = −2 × log-likelihood. R²_within after Raudenbush and Bryk (2010). Random slopes were orthogonalized (i.e. their intercorrelations fixed to 0). Number of people: 652. Number of observations: 11,383. Estimates in boldface indicate p < 0.05.

AIC: Akaike information criterion; BIC: Bayesian information criterion.

OPEN IN VIEWER

Table 3.

Summary of discontinuous growth models predicting cognitive engagement.

Predictor	Model 1					Model 2					Model 3
	Β	SE	t	df	p	Β	SE	t	df	p	Β	SE	t	df	p
Intercept	5.265	0.046	113.608	10,726	<0.001	5.265	0.040	131.375	10,726	<0.001	5.265	0.040	131.112	10,718	<0.001
Level 1 (within-person)
Pre-Event Period	0.009	0.014	0.614	10,726	0.539	0.009	0.014	0.612	10,726	0.540	0.009	0.014	0.612	10,718	0.541
Lockdown1	−0.185	0.042	−4.448	10,726	<0.001	−0.185	0.041	−4.506	10,726	<0.001	−0.185	0.041	−4.526	10,718	<0.001
Recovery1	−0.004	0.004	−0.968	10,726	0.333	−0.004	0.004	−0.958	10,726	0.338	−0.004	0.004	−0.966	10,718	0.334
Lockdown2	−0.034	0.034	−1.005	10,726	0.315	−0.034	0.034	−0.987	10,726	0.324	−0.034	0.034	−0.998	10,718	0.318
Recovery2	>−0.001	0.005	−0.005	10,726	0.996	<0.001	0.005	0.032	10,726	0.975	<0.001	0.005	0.050	10,718	0.960
Level 2 (between-person)
Workload Pre Lockdown1						0.217	0.031	6.931	643	<0.001	0.255	0.034	7.515	643	<0.001
Workload Pre Lockdown2						0.107	0.026	4.143	643	<0.001	0.104	0.026	4.022	643	<0.001
Job Autonomy Pre Lockdown1						0.133	0.028	4.712	643	<0.001	0.142	0.031	4.650	643	<0.001
Job Autonomy Pre Lockdown2						0.119	0.027	4.455	643	<0.001	0.119	0.027	4.419	643	<0.001
Supervisor Support Pre Lockdown1						0.026	0.030	0.863	643	0.388	0.018	0.034	0.539	643	0.590
Supervisor Support Pre Lockdown2						0.023	0.031	0.756	643	0.450	0.022	0.031	0.697	643	0.486
Coworker Support Pre Lockdown1						0.025	0.037	0.665	643	0.506	0.016	0.042	0.376	643	0.707
Coworker Support Pre Lockdown2						0.048	0.033	1.474	643	0.141	0.045	0.033	1.375	643	0.170
Interactions
Workload Pre Lockdown1 × Lockdown1											−0.078	0.025	−3.059	10,718	0.002
Workload Pre Lockdown2 × Lockdown2											0.044	0.016	2.713	10,718	0.007
Job Autonomy Pre Lockdown1 × Lockdown1											−0.020	0.023	−0.878	10,718	0.380
Job Autonomy Pre Lockdown2 × Lockdown2											0.021	0.017	1.221	10,718	0.222
Supervisor Support Pre Lockdown1 × Lockdown1											0.015	0.027	0.564	10,718	0.573
Supervisor Support Pre Lockdown2 × Lockdown2											0.027	0.022	1.206	10,718	0.228
Coworker Support Pre Lockdown1 × Lockdown1											0.017	0.036	0.467	10,718	0.641
Coworker Support Pre Lockdown2 × Lockdown2											0.038	0.024	1.557	10,718	0.119
	Model 1					Model 2					Model 3
Variance components
Intercept	0.79					0.43					0.43
Lockdown1	0.26					0.23					0.22
Lockdown2	0.21					0.22					0.20
Residual	0.61					0.61					0.61
Deviance (df)	29,754.13 (13)					29,452.31 (18)					29,411.49 (26)
AIC	29,780.13					29,488.31					29,463.49
BIC	29,875.55					29,620.43					29,654.32
R 2 within	0.13					0.12					0.12

Models fitted using maximum likelihood estimates. Level-2 predictors were grand mean centered. Deviance = −2 × log-likelihood. R²_within after Raudenbush and Bryk (2010). Random slopes were orthogonalized (i.e. their intercorrelations fixed to 0). Number of people: 652. Number of observations: 11,383. Estimates in boldface indicate p < 0.05.

AIC: Akaike information criterion; BIC: Bayesian information criterion.

OPEN IN VIEWER

Table 4.

Summary of discontinuous growth models predicting emotional engagement.

Predictor	Model 1					Model 2					Model 3
	Β	SE	t	df	p	Β	SE	t	df	p	Β	SE	t	df	p
Intercept	4.677	0.052	90.770	10,726	<0.001	4.677	0.045	103.285	10,726	<0.001	4.677	0.045	103.146	10,718	<0.001
Level 1 (within-person)
Pre-Event Period	>−0.001	0.014	−0.024	10,726	0.981	>−0.001	0.014	−0.024	10,726	0.981	>−0.001	0.014	−0.024	10,718	0.981
Lockdown1	−0.100	0.040	−2.487	10,726	0.013	−0.101	0.040	−2.504	10,726	0.012	−0.101	0.040	−2.512	10,718	0.012
Recovery1	−0.015	0.004	−3.805	10,726	<0.001	−0.015	0.004	−3.791	10,726	<0.001	−0.015	0.004	−3.791	10,718	<0.001
Lockdown2	−0.014	0.034	−0.420	10,726	0.674	−0.013	0.035	−0.379	10,726	0.705	−0.013	0.034	−0.388	10,718	0.698
Recovery2	0.001	0.004	0.287	10,726	0.774	0.002	0.004	0.345	10,726	0.730	0.002	0.004	0.344	10,718	0.731
Level 2 (between-person)
Workload Pre Lockdown1						0.069	0.039	1.802	643	0.072	0.094	0.041	2.301	643	0.022
Workload Pre Lockdown2						−0.023	0.032	−0.709	643	0.478	−0.025	0.032	−0.778	643	0.437
Job Autonomy Pre Lockdown1						0.184	0.035	5.295	643	<0.001	0.194	0.037	5.280	643	<0.001
Job Autonomy Pre Lockdown2						0.120	0.033	3.628	643	<0.001	0.122	0.033	3.693	643	<0.001
Supervisor Support Pre Lockdown1						0.125	0.037	3.347	643	0.001	0.148	0.040	3.693	643	<0.001
Supervisor Support Pre Lockdown2						0.097	0.038	2.524	643	0.012	0.097	0.038	2.538	643	0.011
Coworker Support Pre Lockdown1						−0.016	0.046	−0.347	643	0.729	−0.034	0.050	−0.679	643	0.497
Coworker Support Pre Lockdown2						0.077	0.040	1.908	643	0.057	0.070	0.040	1.733	643	0.084
Interactions
Workload Pre Lockdown1 × Lockdown1											−0.048	0.025	−1.904	10,718	0.057
Workload Pre Lockdown2 × Lockdown2											0.029	0.017	1.703	10,718	0.089
Job Autonomy Pre Lockdown1 × Lockdown1											−0.020	0.022	−0.918	10,718	0.359
Job Autonomy Pre Lockdown2 × Lockdown2											−0.012	0.018	−0.687	10,718	0.492
Supervisor Support Pre Lockdown1 × Lockdown1											−0.044	0.027	−1.645	10,718	0.100
Supervisor Support Pre Lockdown2 × Lockdown2											0.001	0.024	0.023	10,718	0.982
Coworker Support Pre Lockdown1 × Lockdown1											0.034	0.035	0.969	10,718	0.333
Coworker Support Pre Lockdown2 × Lockdown2											0.089	0.026	3.444	10,718	0.001
	Model 1					Model 2					Model 3
Variance components
Intercept	1.13					0.74					0.74
Lockdown1	0.22					0.22					0.21
Lockdown2	0.24					0.26					0.24
Residual	0.60					0.60					0.60
Deviance (df)	29,693.11 (13)					29,470.12 (18)					29,435.25 (26)
AIC	29,719.11					29,506.12					29,487.25
BIC	29,814.53					29,638.24					29,678.08
R 2 within	0.14					0.14					0.14

Models fitted using maximum likelihood estimates. Level-2 predictors were grand mean centered. Deviance = −2 × log-likelihood. R²_within after Raudenbush and Bryk (2010). Random slopes were orthogonalized (i.e. their intercorrelations fixed to 0). Number of people: 652. Number of observations: 11,383. Estimates in boldface indicate p < 0.05.

AIC: Akaike information criterion; BIC: Bayesian information criterion.

Hypothesis 1 suggests that, due to the transitions into the national lockdowns in Germany, there were declines in employees’ physical, cognitive, and emotional job engagement (from March to April and from December 2020 to January 2021, respectively). The parameters of the first transition slope (see Model 1 in Tables 2–4) suggested significant negative transitions in physical (B_{Lockdown 1} = −0.156, SE = 0.040, p < 0.001), cognitive (B_{Lockdown 1} = −0.185, SE = 0.042, p < 0.001), and emotional engagement (B_{Lockdown 1} = −0.100, SE = 0.040, p = 0.013) between March and April 2020. However, there were no significant transitions in physical, cognitive, or emotional engagement between December 2020 and January 2021. Thus, Hypothesis 1 was only supported for the first national lockdown by our data.

According to Hypothesis 2, there was an increase in employees’ physical, cognitive, and emotional job engagement in the post-transition phases of the national lockdowns (from April to December 2020 and from January to September 2021, respectively). Contrary to this prediction, the parameters of the first recovery slope (see Model 1 in Tables 2–4) suggested significant declines in physical (B_{Recovery 1} = −0.015, SE = 0.004, p < 0.001) and emotional engagement (B_{Recovery 1} = −0.015, SE = 0.004, p < 0.001), and no significant changes in cognitive engagement between April and December 2020. Also contrary to our hypothesis, the parameters of the second recovery slope indicated significant declines in physical engagement (B_{Recovery 2} = −0.009, SE = 0.004, p = 0.033) and no significant changes in cognitive and emotional engagement between January and September 2020. Thus, Hypothesis 2 was not supported by our results.

Hypothesis 3 states that the transition into the second national lockdown was associated with less pronounced changes in employees’ physical, cognitive, and emotional job engagement when compared to the changes in job engagement associated with the transition into the first national lockdown. In line with this hypothesis, Wald Chi-Square Tests revealed that the transition slope parameter of physical engagement at the first lockdown was stronger compared to the second lockdown (Δχ²_{[Δdf = 1]} = 14.870, p < 0.001). Also in line with our predictions, the transition in cognitive engagement at the first lockdown was stronger compared to the second lockdown (Δχ²_{[Δdf = 1]} = 9.221, p = 0.002). However, there was no significant difference between the transition parameters of emotional engagement at the first and second national lockdown (Δχ²_{[Δdf = 1]} = 3.275, p = 0.070). Thus, Hypothesis 3 was only supported for physical and cognitive engagement.

Hypothesis 4 suggests that pre-event levels of workload strengthened the decline in physical, cognitive, and emotional job engagement at the transitions into the national lockdowns. Considering Model 3 in Tables 2–4, the interaction terms suggest that, in line with our hypothesis, a high (vs low) pre-event level of workload strengthened the decline in physical engagement due to the first lockdown (B_{Lockdown 1 × Workload} = −0.061, SE = 0.025, p = 0.014). Also in line with the hypothesis, a high (vs low) pre-event level of workload strengthened the decline in cognitive engagement due to the first lockdown (B_{Lockdown 1 × Workload} = −0.078, SE = 0.025, p = 0.002; results of simple slopes analyses are reported in Table 5 and Figure S1). Contrary to our predictions, the second lockdown led to an increase in physical engagement when pre-event levels of workload were high and to a decrease in cognitive engagement when pre-event levels of workload were low (see results of simple slopes analyses reported in Table 5 and Figure S1). Also contrary to our hypothesis, workload was not associated with changes in emotional engagement due to the first or second national lockdown. Thus, Hypothesis 4 was only partially supported.

OPEN IN VIEWER

Table 5.

Summary of simple slopes analyses.

Moderator levels	Estimate	SE	df	t	p
	Physical engagement
Workload Pre Lockdown 1 + 1 SD	−0.229	0.050	10,718	−4.611	<0.001
Workload Pre Lockdown 1 Mean	−0.155	0.040	10,718	−3.919	<0.001
Workload Pre Lockdown 1 − 1 SD	−0.082	0.050	10,718	−1.653	0.098
Workload Pre Lockdown 2 + 1 SD	0.114	0.040	10,718	2.814	0.005
Workload Pre Lockdown 2 Mean	0.029	0.033	10,718	0.902	0.367
Workload Pre Lockdown 2 − 1 SD	−0.055	0.040	10,718	−1.357	0.175
Job Autonomy Pre Lockdown 2 + 1 SD	0.085	0.041	10,718	2.039	0.041
Job Autonomy Pre Lockdown 2 Mean	0.029	0.033	10,718	0.902	0.367
Job Autonomy Pre Lockdown 2 − 1 SD	−0.026	0.042	10,718	−0.614	0.539
Coworker Support Pre Lockdown 2 + 1 SD	0.115	0.046	10,718	2.482	0.013
Coworker Support Pre Lockdown 2 Mean	0.029	0.033	10,718	0.902	0.367
Coworker Support Pre Lockdown 2 − 1 SD	−0.056	0.046	10,718	−1.209	0.227
	Cognitive engagement
Workload Pre Lockdown 1 + 1 SD	−0.279	0.051	10,718	−5.454	<0.001
Workload Pre Lockdown 1 Mean	−0.185	0.041	10,718	−4.526	<0.001
Workload Pre Lockdown 1 − 1 SD	−0.090	0.051	10,718	−1.770	0.077
Workload Pre Lockdown 2 + 1 SD	0.033	0.042	10,718	0.783	0.433
Workload Pre Lockdown 2 Mean	−0.034	0.034	10,718	−0.998	0.318
Workload Pre Lockdown 2 − 1 SD	−0.100	0.042	10,718	−2.405	0.016
	Emotional engagement
Coworker Support Pre Lockdown 2 + 1 SD	0.110	0.050	10,718	2.224	0.026
Coworker Support Pre Lockdown 2 Mean	−0.013	0.034	10,718	−0.388	0.698
Coworker Support Pre Lockdown 2 − 1 SD	−0.137	0.050	10,718	−2.757	0.006

Moderators were grand mean centered. Plots of the interaction effects can be found in Figures S1–S3 in the supplemental materials.

According to Hypothesis 5, high pre-event levels of job autonomy buffered the decline in job engagement at the transitions into the national lockdowns. Contrary to our hypothesis, our results suggested that the second lockdown led to an increase in physical engagement when pre-event levels of job autonomy were high, but had no effect on physical engagement when pre-event levels of job autonomy were low (see results of simple slopes analyses reported in Table 5 and Figure S2). Also contrary to our prediction, job autonomy was not associated with changes in physical engagement due to the first national lockdown or changes in cognitive or emotional engagement due to either of the national lockdowns. Thus, Hypothesis 5 was not supported by our data.

Hypothesis 6 states that high pre-event levels of (a) supervisor support and (b) coworker support buffered the decline in physical, cognitive, and emotional job engagement at the transitions into the national lockdowns. Contrary to Hypothesis 6a, pre-event levels of supervisor support were not associated with changes in job engagement at the transitions into the two national lockdowns. In line with Hypothesis 6b, the second lockdown led to a decrease in emotional engagement when pre-event levels of coworker support were low and led to an increase in emotional engagement when pre-event levels of coworker support were high (see results of simple slopes analyses in Table 5 and Figure S3). Contrary to our prediction, results suggested that the second lockdown led to an increase in physical engagement when pre-event levels of coworker support were high but had no significant effect on physical engagement when pre-event levels of coworker support were low (see results of simple slopes analyses in Table 5 and Figure S3). Also contrary to the hypothesis, coworker support was not associated with changes in physical engagement at the transition into the first lockdown, changes in cognitive engagement, or changes in emotional engagement at the transition into the first lockdown. Thus, Hypothesis 6b was only partially supported.

Summary and interpretation of findings

The present study investigated how disruptive environmental events, exemplified by two national lockdowns in Germany during the COVID-19 pandemic, impacted employees’ job engagement. In line with propositions of EST (Morgeson et al., 2015), we observed a decline in physical, cognitive, and emotional job engagement in reaction to the first disruptive environmental event. In contrast, there were no declines in job engagement when a similar event occurred several months later. The comparison of employees’ reactivity to the two consecutive events showed that the rather small changes in emotional engagement due to the transition into the first event did not differ from the changes due to the second event. However, the transitions in physical and cognitive engagement due to the second event were substantially weaker compared to the first event. These findings are consistent with the previously observed, negative relationship between subjective ratings of event strength, including event novelty, and job engagement (Liu et al., 2021). Overall, our results provide some support for the role of event novelty for employee experiences as outlined by EST. According to this reasoning, employees did not need to invest in additional information processing and behavioral adaptation at the transition into the second lockdown because they were already familiar with this type of disruptive environmental event. It may also be that employees were able to build up important resources (e.g. coping skills) during the first disruptive event that enabled them to better adapt to and cope with the second event. Interestingly, employees’ reactivity to the second event was not only weaker compared to the first event, but even negligible, suggesting that event novelty may have played a more important role than expected.

Contrary to our predictions and previous empirical findings on recovery processes following disruptive events (e.g. Zacher et al., 2021), we did not observe increases in job engagement in the months following the transitions into the lockdowns. Instead, there was a continued decrease in physical and emotional engagement after the first event, as well as a decrease in physical engagement after the second event. These findings are in line with previous studies observing sustained declines in psychological resources and outcomes after disruptive events (e.g. Lucas et al., 2004; Rauvola et al., 2022) and suggest that employee motivation may suffer longer-term due to these transitions.

We additionally examined whether key characteristics of employees’ jobs, measured immediately before the onsets of the disruptive environmental events, moderate changes in job engagement. Based on EST and work design research (Demerouti et al., 2001; Morgeson et al., 2015; Parker, 2014), we expected that workload would strengthen, whereas the job resources of job autonomy, supervisor support, and coworker support would buffer the declines in physical, cognitive, and emotional job engagement. Overall, our hypotheses were only partially supported. As expected, we observed that pre-event levels of workload strengthened the decline in physical and cognitive engagement due to the transition into the first disruptive event. Also in line with our hypotheses, the transition into the second event led to a decrease in emotional engagement when pre-event levels of coworker support were low and, unexpectedly, to an increase in emotional engagement when pre-event levels of coworker support were high. Surprisingly, our results also suggested that the second event led to an increase in physical engagement when pre-event levels of job autonomy and coworker support were high but had no effect on physical engagement when pre-event levels of job autonomy or coworker support were low. Also contrary to our predictions, the second lockdown led to an increase in physical engagement when pre-event levels of workload were high and to a decrease in cognitive engagement when pre-event levels of workload were low.

These findings suggest that only certain job characteristics may support employees in adapting to disruptive environmental events. Supervisor support had no moderating effect in our study, whereas the effects of workload, job autonomy, and coworker support showed different results for the two events and different facets of job engagement. Regarding the facets of job engagement, our results did not reveal systematic differences in the moderating effects between the three dimensions. The events’ influence on physical engagement was found to be conditional upon all three moderators. In addition, there was a moderating effect of workload on changes in cognitive engagement, and of coworker support on changes in emotional engagement. Overall, the distinction between the three facets of job engagement does not appear to provide additional insights into the moderating role of key job characteristics.

In contrast, the comparison of the moderating effects between the two consecutive events revealed a consistent pattern of results. For the first lockdown, the moderating effects observed were in line with our hypotheses (i.e. there was a decrease in job engagement that was stronger if workload was high). However, for the second event, the observed effects did not reveal the expected pattern. Specifically, our results showed an increase (rather than a weaker decrease) in job engagement for employees with high levels of job autonomy and coworker support, but also workload. These differential findings for the two consecutive events stress the importance of studying recurring events and the role of repeated event exposure for employees’ adaptation processes.

Theoretical and practical implications

This study has several implications for organizational research and practice. First, we extend theory and empirical research on job engagement by investigating employees’ reactivity to disruptive environmental events. We combine EST (Morgeson et al., 2015) with the transition-adaptation framework by Bliese et al. (2017) to fully map this dynamic process, including the pre-event trajectories, transitions into the events, as well as post-transition trajectories of job engagement. The relatively strong decline in job engagement observed in reaction to the first environmental event supports the propositions of EST regarding top-down effects on employees caused by disruptive events at the environmental level. In contrast, our results did not support predictions regarding potential recovery dynamics in the post-transition phases of disruptive environmental events (e.g. Bliese et al., 2017), as we observed sustained declines in job engagement after the transitions into the lockdowns. These unexpected findings suggest that employee motivation may suffer longer-term and stress the need to further investigate the whole dynamic process related to disruptive environmental events, including a pre-event baseline and post-transition trajectories. Possible ways to obtain data for such studies may be collecting representative panel data on a permanent, regular basis, or to use data generated during social media use (e.g. Min et al., 2021). Future theory development could focus on whether adaptation processes and recovery dynamics in the context of disruptive environmental events are limited to specific employee outcomes and whether certain event or person characteristics determine how employees adapt to such events.

Second, our study adds to research on the role of event novelty as defined by EST, as it examines employees’ reactivity to repeated environmental events (Lang et al., 2021). We found some support for the propositions of EST regarding the role of event novelty by showing differential changes in physical and cognitive job engagement in reaction to two national lockdowns. These results are in line with previous findings based on subjective ratings of event strength (e.g. Liu et al., 2021). We extend this research by examining changes in employees’ job engagement in reaction to recurring disruptive events. In more general terms, our findings may contribute to research on the meta-concept of resilience, or people’s successful adaptation in the context of adversity (e.g. Fisher et al., 2019; Hartmann et al., 2020). In particular, investigating employees’ job engagement trajectories in reaction to repeated adverse events in their broader environment may add to research that conceptualizes resilience as the outcome of a dynamic process (rather than a general capacity; Arnold et al., 2023; Fisher et al., 2019). In this conceptualization, resilience is characterized by individuals’ return to (“bouncing back”) or development beyond (“bouncing beyond”) their levels of functioning before facing adversity, as opposed to remaining at a lower level of functioning (e.g. Hoegl and Hartmann, 2021). An important contribution to this line of research is the investigation of individuals’ reactivity to adversity from an event-oriented perspective, as well as potential differences in reactivity to recurring events (e.g. Backmann et al., 2021; Raetze et al., 2022). Our findings suggest that although employees’ job engagement did not “bounce back” immediately after the transitions into the disruptive environmental events, their reactivity to the second event was substantially weaker, indicating that they may have been able to build on their experience of the first event occurrence in coping with the second event. Future theory development and research should explore the specific mechanisms underlying adaptation processes in the context of repeated events. For example, future theorizing could elaborate under which conditions and to what extent employees’ reactivity is weaker due to their improved knowledge, behavioral scripts, and coping skills, such as emotion regulation.

Third, this study extends theorizing on events by integrating event- and feature-oriented approaches to understanding employee experiences. We combine EST and the job-demands resources model (Bakker and Demerouti, 2017; Demerouti et al., 2001) to derive and test hypotheses on the potential moderating role of job characteristics in the context of disruptive environmental events. Our study thereby advances knowledge on how contextual factors at the job level may influence the effects of disruptive environmental events on employees. Considering the role of contextual factors for individuals’ adaptation to adversity is an approach that has also been called for in resilience research (e.g. Fisher et al., 2019; Raetze et al., 2022). Our finding that job characteristics, measured before event onset, moderated the events’ effects on employees’ job engagement underscores the significance of this approach. Jobs with high levels of job autonomy and coworker support may mitigate such events’ harmful consequences. Our results showed consistent patterns of moderating effects for each of the consecutive events, but these patterns differed between events. Specifically, in reaction to the recurring event, employees showed an increase (rather than a weaker decrease) in job engagement if they had high levels of job autonomy and coworker support, but also workload. These findings suggest that employees’ adaptation processes, including “bouncing beyond” previous levels of functioning (e.g. Hoegl and Hartmann, 2021), and the way in which these can be supported by job characteristics are different for repeated events than for first-time, unknown events. Thus, studying repeated events and associated adaptation processes is also important when investigating the moderating role of employees’ work environment. Further theory development is needed on the specific mechanisms through which contextual factors may support or hinder employees’ adaptation to disruptive environmental events, which might also explain the differences in moderating effects between repeated events. Contextual characteristics may facilitate or impede employees’ coping with the overall magnitude of changes caused by the event, but these mechanisms might also be specific to the type of changes employees face. For example, job autonomy might specifically facilitate employees’ coping with an event if it requires them to make additional decisions or change the way they work. In addition, future research could further investigate these effects by testing potential moderators at different hierarchical levels (e.g. the individual or team level).

Regarding practical implications, our findings demonstrate how severe and long-lasting the top-down effects of disruptive environmental events on employees’ experiences can be. Human resource managers should be aware of the potential impact on employees caused by events that originate outside the workplace, such as lockdowns during a pandemic crisis (see Rudolph et al., 2021), and consider how to best support employees in times of crisis, for instance through human resource practices (Zacher and Rudolph, 2022). In this regard, our findings provide initial guidance on how to design jobs that support employees in dealing with disruptive, unexpected events. In particular, creating jobs with a high degree of autonomy and a team climate that encourages coworker support may help employees retain high levels of job engagement or even increase their job engagement levels when facing such events.

Limitations and future research

Notwithstanding its strengths (e.g. a large sample and analyses including baseline measures occurring in pre-event periods), our study has a number of limitations that could be addressed in future research. First, both job engagement and job characteristics were assessed using self-report measures. Given that we tested hypotheses on changes in, rather than levels of, job engagement and interactions of job characteristics with these transition slopes, common method bias (Podsakoff et al., 2012; Spector, 2006) may not account for specific patterns of focal results (Siemsen et al., 2010). Regarding the measures used, we did not observe universal evidence for measurement invariance over time (e.g. Widaman et al., 2010). However, we found partial measurement invariance for the variables in question and, importantly, did not use latent variable models for our main analyses. Nevertheless, our findings might still be subject to method biases. Thus, future research should also include other sources of data, such as objective measures of job characteristics, supervisor and coworker ratings, and behavioral observation. Related to this issue, we investigated the effects of repeated environmental events based on the concept of event novelty, but we did not ask employees directly about the extent to which the event was novel for them. To address this issue, future studies on repeated environmental events should also assess subjective ratings of event strength characteristics.

Second, the national lockdown periods began in the middle of March and the middle of December 2020, respectively. Because our measurement waves were conducted at the beginning of each month, there was a time lag of approximately 2 weeks relative to the onsets of the events. Our measurements were sufficiently sensitive to detect substantial shifts in job engagement at the onset of the first lockdown. However, our study might underestimate employees’ immediate reactivity to these events. Future studies could therefore examine similar events with measurement points that are more frequent or even closer in time to the onsets of the events.

Third, although our study provides a precise description of the changes in job engagement due to the disruptive environmental events, we did not assess the specific mechanisms that may explain why these events impacted employees’ job engagement. These potential processes could include additional information processing requirements, as well as the necessary changes of behaviors and routines (Morgeson et al., 2015). Future studies could investigate these mechanisms to examine the exact processes by which disruptive events lead to a decline in job engagement. In addition, we modeled the lockdowns as discrete events, which does not reflect the specific dynamics and the occurrence of additional events during the lockdowns. For example, employees may have first switched to working from home and then had additional childcare responsibilities due to school closures or they may have faced additional events, such as unemployment or COVID-19 infections of friends and family members (see e.g. Fleuren et al., 2023). Furthermore, the COVID-19 pandemic as such was an exceptional context in which the lockdowns took place and some of the measures to contain the spread of the coronavirus were also in place outside the lockdown periods, such as social distancing and restricted on-site work (see Figure 1 and Steinmetz et al., 2020). Future research could examine these more complex dynamics and the pathways through which environmental events trigger subsequent events. In addition, future studies could investigate the relative strength of different events and accumulation of their effects over time (see e.g. Seery et al., 2010).

Finally, our study was limited to full-time employees at each measurement wave. Thus, our findings do not generalize to employees working only for a limited number of hours per week or those in self-employment. Although the results of attrition analyses did not raise great concerns, our results might still be affected by selection effects (e.g. Nunan et al., 2018). Specifically, attrition analyses revealed that participants not included in the sample had a lower level of cognitive engagement compared to those who provided sufficient information. Thus, participants who were particularly affected by the first lockdown and could not adapt successfully might have dropped out of the study. This may include employees who had to switch to short-term work due to economic problems or part-time employment because of additional childcare responsibilities (Kniffin et al., 2021; Rudolph et al., 2021). Such potential selection effects also may have hindered the detection of effects caused by the second lockdown (i.e. Hypothesis 3). Moreover, the extent to which employees were affected by the lockdowns may have been influenced by person characteristics or additional job characteristics that were not examined in this study. For example, teleworkers may have had to change their work behavior less, or not at all, during the lockdowns and therefore been less affected by this particular type of event (e.g. Günther et al., 2022; Kaltiainen and Hakanen, 2023). Accordingly, future studies could investigate event-specific job and person characteristics, such as teleworking, to capture the extent to which employees were affected by the disruptive environmental event.