The role of job insecurity and work-family conflict on mental health evolution during COVID-19 lockdown

ABSTRACT

The aim of this intensive longitudinal study was (1) to explore the temporal evolution of two mental health indicators (anxiety and depressive symptoms, and insomnia) throughout COVID-19 lockdown in Spain, and (2) to examine its association with two work-related stressors (job insecurity and work-family conflict). A sample of 1519 participants responded to several questionnaires during the lockdown (between 16 March and 29 April 2020). Results of latent growth modelling showed a curvilinear increase of our two mental health indicators over time (a logarithmic growth for anxiety and depression, accentuated during the first part of the lockdown, and a quadratic growth for insomnia, accentuated during the second part). Regarding its association with work-related stressors, we found that higher levels of job insecurity and work-family conflict were related to higher levels of anxiety, depression, and insomnia. Additionally, we found a significant interaction between time and the two forms of work-family conflict (work-to-home and home-to-work), showing that people with more work-family conflict experienced stronger growth in all mental-health indicators. Overall, this study contributes to the description of the temporal dynamics of mental health during the COVID-19 outbreak in Spain, as well as its association with two key work-related stressors.

KEYWORDS:

• COVID-19 pandemic
• job insecurity
• work-family conflict
• mental health
• work stress

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1. Additionally, we tested all our models without control variables and the results did not differ substantially. However, according to the logic of Becker (2005), we maintained control variables in our final models.

2. We would like to thank the action editor and the two anonymous reviewers for their inspiring insights, that helped us to develop this additional section.

3. Additional information (fit indexes, model comparison, etc.) can be found in online supplemental material.

4. This classification has been removed from Rights & Sterba’s proposal (2020). As they reviewed, most of the proposed R² indicators for multilevel models present several shortcomings, mainly associated with the fact that those models have several sources of slope and intercept variation. Thus, for an accurate inspection of R², researchers should present the disaggregated measure, that enables to identify which source of variation is explaining whin portion of variability.

¹ We additionally run a test for the autoregressive structure and found that a model allowing autocorrelation improved the fit to the data (Phi = .16, p < .01), so this feature was included in the subsequent model (meaning that we controlled for autocorrelation). Additionally, we checked for heteroscedasticity, as the variance of anxiety and depression may vary over time and found that modelling the increase in variance did not significantly improve the fit to our data (Lratio = 3.21, p = .08).

² Additionally, we run a test for the autoregressive structure and found that a model allowing autocorrelation fits the data better (Phi = −.21, p < .01): subsequent models were therefore estimated including this feature. The increase in the variance was excluded because of convergence problems.