There are a number of papers aimed at studying the consequences of the COVID-19 crisis for the working population and identifying key factors of this influence. First of all, let’s consider two articles which dataset provide the basis for this paper, thereby showing the novelty of our approach.

The first of these was an article by F. Knolle et al. (2021). In it, the researchers published the results of two waves of a three-part survey (May and October) of respondents from the UK and Germany (the first part – presence or absence of symptoms of coronavirus infection, the second – SCL-27 (symptom check list – survey to assess psychological problems), the third – schizotypal personality questionnaire).

The authors worked with the data from the first wave of the survey, paying much attention to the country effect in the differences in the impact of lockdowns: respondents from the UK noted a stronger direct effect of restrictions on health, financial condition and family relationships: they had a higher self-perception of psychological problems, that is, their mental state was worse according to the data in questionnaire SCL-27 (dedicated to mental state diagnosis). I addition, according to the data of the first part of the questionnaire, the sub-sample consisting of British people had more cases of infection, i.e. higher morbidity, which is an important result, since coronavirus also affects mental health.

The second work, based on the same dataset, is published by S.Daimer et al. (2021). In contrast to the first paper, data from the second wave of the survey are added to the analysis. Based on the data of GSI-27 (global symptom index), the authors note that the mental state of respondents in the first wave of the survey was worse than in the second one. Due to the design of the survey, the researchers were able to separate the current temporary effects from the offset of the pandemic from cumulative ones: it turns out that during the second period, respondents noted that their general mental health was better compared to the previous one, since they were able to adapt to the new reality, but the results on the schizotypal questionnaire were higher, which indicates a cumulative effect of the lockdowns: respondents are finding it increasingly difficult to cope with loneliness and financial difficulties over time. Therefore, the authors conclude that the longer the restrictions last, the more individuals will be susceptible to depression.

It is worth noting that in both articles the authors focused on the mental health of respondents in general and its changes because of the pandemic, paying great attention to the results obtained after filling out the part of the schizotypal personality questionnaire. In this paper, we study a narrower issue, focusing on the effect of WFH, which has not been considered before.

In addition to these studies, we will review papers aimed at studying the consequences of the COVID-19 crisis for the mental health of employees and identifying key factors of this influence. For example, in the article by A.Oksanen et al. (2021), the researchers assessed the possible stressful effects of social media communication² at work during the pandemic.

The results showed that social media communication at work increased during the pandemic, which, according to the authors, should have led to a deterioration in the mental health of employees. It turned out that there is a difference in the effect for different groups of workers: those who were used to using social networks at work before the pandemic coped better, and their stress and fatigue levels decreased. One of the conclusions of the researchers was the understanding that building a balance between personal life and work, as well as mental well-being, help to cope with the corona crisis. This paper indicates the heterogeneity of the impact of the pandemic on employees, and emphasizes the importance of a competent policy behalf the companies’ HR managers and measures taken to improve the well-being of employees.

The work-life balance of those employees who started working remotely in connection with the COVID-19 outbreak had a positive effect on mental well-being. N. Yüceol et al. came to this result (2021). The main objective of the paper was to determine the impact of remote working on mental well-being, due to the Covid-19 pandemic. In this context, surveys were applied to 397 generation Y³ academicians working at public and private universities. However, characteristics such as gender, age, and degree status did not significantly affect work-life balance or mental well-being.

According to C. Pieh et al. (2020), self-isolation and the COVID-19 pandemic led to an increase in depression and a decrease in quality of life. The authors assessed the mental health of respondents during the lockdown and wanted to find out how age, gender, income, work and physical activity affected mental well-being. It turned out that the effect is more pronounced in young people, women, the unemployed and people with low income.

Phadnis et al (2021) are the authors of a research aimed at studying the impact of remote work on mental health. According to the results of the research over half of the Indian organizations employees faced an increase in workload when working from home, some felt lonely and lost, as well as disconnected from the real world, which indicates the negative impact of switching to a remote format on the mental health of employees.

However, in foreign and domestic literature, the issue of the relationship between the format of work during the pandemic and the mental well-being of staff has not been sufficiently investigated. Therefore, in this study, we want to answer the question: “Has the transition to a remote format affected the mental health and well-being of employees, and if so, how?”.

Before proceeding directly to econometric calculations, we should notably mention the dataset design used in this paper.

The data were collected by researcher F. Knolle et al. to study a related topic (in the part devoted to the literature review, we explained in detail the scientific novelty of our paper in comparison with those published). Respondents mainly from Germany and the UK were asked to take an online survey on the EvaSys platform (https://www.evasys.de, Electric Paper Evaluationssysteme GmbH, Luneburg, Germany). Following the snowball sampling strategy⁴ respondents from other European countries were added, but the questionnaire was available only in German and English. The survey consisted of three parts: the presence of symptoms and/or the fact of contracting COVID-19 disease, a Checklist of symptoms (SCL-27) and a Schizotypal Personality Questionnaire, which reflected the impact of the pandemic on the financial situation and work, the consumption of psychoactive substances and media content, as well as questions of self-assessment of mental health.

4 waves of the survey were conducted: in the spring and autumn of 2020, in the winter and spring of 2021. The completion of the survey took approximately 35 min. The respondents took part voluntarily and did not receive monetary compensation.

We conducted a preliminary analysis of the variables, highlighting only those that directly relate to the research topic and are control ones. The interpretation of the variables is presented in Appendix 1, and the descriptive statistics are given below (see Table 1).

We conducted an initial analysis of the results and received the following:

• The average age of the respondents was 43.12 years;
• Most of the survey participants are women.
• The most of respondents in the sample have a higher education, which can rather be attributed to a bias in the data;
• The mental state of the respondents at the time of the survey averaged 3.31 units, which according to the encoding is close to normal;
• About a third of the respondents live with children;
• One in five respondents did not change the format of work in connection with the pandemic;
• Almost half of the respondents switched to a remote work format;
• 10% of employees faced overwork.

To assess the degree of closeness of the relationship between variables and to identify the direction of the relationship, we constructed a correlation matrix (see Figure 1. Correlation matrix of characteristics.

There is a negative correlation between the indicator of mental health and remote format of work. However, according to the survey design, the lower the indicator of mental health, the better it actually is⁵. Consequently, home office and mental health are positively correlated.

• Sports and mental well-being of respondents positively correlate.
• The younger the individual, the worse the mental state currently is (as of the survey).
• Over time, mental health deteriorated on average in the sample (according to the sign of the correlation coefficient between the survey wave number and the mental health indicator).
• On average, individuals living with children rated their mental health worse during the pandemic.
• Respondents with higher education were less likely to be dismissed, and they also assessed their mental state at a higher level.
• The presence of a confirmed diagnosis or symptoms of COVID-19 negatively correlates with the mental health of respondents.

We also plotted the distribution density of the mental health indicator to compare respondents who worked remotely and from the office during the pandemic. They show that the mental state of people who worked from home is better than that of those survey participants who did not change the format of work (see Figure 2 and Figure 3).

Having studied the entire econometric apparatus in the scientific literature on related topics, we came to the conclusion that it is worth starting the work with a linear regression constructed be means of the least squares method.

Taking the mental state of an individual as a dependent variable, we set all other variables as regressors (except for the individual participant number and the observation code). After checking them for multicollinearity, we found out that it is missing from the data, since all the obtained VIF coefficients are less than 5 (see Table 2).

It should also be noted that despite the low value of the VIF coefficient, it is necessary to more accurately disclose the definition of the variables “overtime” and “reduction of hours of work” and justify their simultaneous inclusion in the model. At first glance, if an individual had a decrease in working hours (Work_ReductionHours = 1), then it can be concluded that there is no overtime work (Work_Overtime = 0). Conversely, if an individual worked overtime (Work_Overtime = 1), then he did not have a decrease in working hours (Work_ReductionHours = 0). However, there are other combinations of these two variables: an individual can get a formal limitation of working hours (Work_ReductionHours = 1), but in fact overwork relative to the specified limitation (Work_Overtime = 1) and vice versa. It follows that there is no unambiguous relationship between the combinations of values of these variables. Accordingly, there is no linear relationship between the variables, and the both should be included in the model.

Then we improved the model based on the Akaike criterion, obtaining an optimal set of regressors and an MLS model. We also conducted a model comparison test and received a p-value of 0.8778, which means that at all levels of significance, the “short” model is more accurate (the corresponding screenshot of the console in R is given in Appendix 2. Statistical tests)

However, such a model may be incorrect, since there may be a temporary effect due to the presence of several waves of the pandemic, which is reflected in the Timepoint variable. Therefore, we introduce models with random and fixed time effects into the analysis. After conducting the necessary tests to compare all the models, we found out that the model with random effects is optimal from an econometric point of view (see Table 3).

We also conducted a series of tests for heteroscedasticity for all models, according to the p-value of which (3.315e-06 for MLS; 4.206e-07 for FE and 4.206e-07 for RE, the screenshots are also given in Appendix 2. Statistical tests) it is advisable to use robust standard errors to determine the significance of variables and test hypotheses. Then we formed a summary table with the evaluation results of all three models (see Table 4).

Let’s give a statistical interpretation to some variables in a model with random effects:

• There is a causal relationship between the indicator of mental state and the format of work: all other things being equal, the indicator of mental health is on average 0.1 better for respondents who worked remotely from home as compared to those who did not switch to the remote format.

For a more detailed study, we plotted the dynamics of the average value of the mental health indicator depending on the survey wave for those who worked remotely from home all this time, and those who worked from the office (see Figure 4).

The effect of time and the tightening of covid restrictions during the third wave of the survey is clearly traced here: there is a significant deterioration in the average value of the mental health⁶ index for both groups. The further dynamics of the indicator is also interesting: in the fourth wave of the survey, respondents who worked remotely noted an improvement in their mental state by 0.019. Most likely, this is due to the fact that people have adapted to the new format of work during the introduction of restrictions, and some employers have also begun to realize the problem of the mental health of employees and implement stimulating policies. So, to improve their workers well-being many companies launched pulse surveys of their staff mental state, which resulted in educational workshops and trainings conducted by company employees and invited speakers, more online meetings, including informal ones, and psychological services.

The situation among those who worked at the office is the exact opposite: during the strengthening of the lockdown, due to the lack of relevant experience, they could not adapt to the changed conditions, which made them feel more depressed in the future: the average for this group increased by 0.005 units during the transition from the third to the fourth wave. All these conclusions are confirmed by the dynamics of the group averages in the context of the survey waves (see Table 5).

Let us turn to the meaningful interpretation of the obtained coefficients. Here it is also worth noting that the conclusions concerning the interpretation of the remaining variables apply to both formats of work in the pandemic.

• There are no statistically significant differences in the indicator of mental health depending on gender.
• There is a causal relationship between the indicator of mental state and the age of employees: all other things being equal, with an increase in age per unit, the indicator of mental health improves by an average of 0.01. This conclusion coincides with the results of studies on related topics that were conducted earlier.
• There is also a connection between the presence of children in the household and the indicator of mental well-being: other things being equal, individuals with children rated their mental health on average 0.16 worse than individuals without children. It is probably more difficult to maintain a work-life balance with children.
• The indicators of physical activity and mental health of an individual are also interrelated: other things being equal, with an increase in the number of sports activities per week, one’s mental health improves by an average of 0.11. Indeed, sport is an effective way to deal with stress.
• Interestingly, all other things being equal, individuals with higher education assessed their mental state on average 0.26 better than individuals without higher education.
• In addition, other things being equal, individuals who had a confirmed diagnosis of COVID-19 or its symptoms rated their mental health on average 0.17 lower, which is due to the direct effect of the disease on the nervous system.

This approach may still be unreliable, since the number of values that the dependent variable can take is finite. This method does not guarantee that the estimate of the dependent variable will be strictly in the range from 1 to 5. Therefore, we turned to another method that is often used in this kind of research — ordered logistic regression (see Table 6). We also decided to build 3 models for each of the survey waves in order to study the effect of the remote format of work that interests us.

To interpret the coefficients, we used a graphical method to understand the impact of the work format on the probability of what level of mental health the respondent had in a particular wave of the survey.

So, if we consider the second wave of the survey, we can say that for those who worked remotely from home in the autumn of 2020 the probability of having a better mental state score turned out to be higher on average, other things being equal (see Figure 5). However, it should be noted that during this period, all employees, regardless of the format of work, on average assessed their mental well-being as “normal”.

As for the third wave of the survey (see Figure 6), in the winter of 2021, the proportion of people assessing their mental health as “good” and even more so “very good” was small and decreased compared to the second wave of the survey, and the proportion of those assessing it as “bad” is quite large and increased significantly compared to the previous period. Moreover, such a pattern can be traced for both groups of workers. It seems to us that this can be explained by the fact that the third wave of the survey is the period of the most severe covid restrictions in the countries studied, when the format of remote work became a forced measure taken by companies. Employees had to find a new model of work and life in the current situation and readjust.

In May 2021, people working remotely from home adapted to the new format. Most employees found a new, most comfortable model of work and interaction within the company. Of course, as we suppose, organizational and technical solutions of personnel management specialists played a major role. As a result, looking at the graph, we can say that those who worked remotely from home in May 2021 are more likely to have a better mental state indicator, on average higher, other things being equal (see Figure 7).

• Is the issue of physical health of employees important for your company? Do you have corporate insurance?
• Does your company measure the physical health of employees?

If the respondent answered affirmatively, the following questions were asked:

• How is this indicator measured?
• Do you conduct employee surveys?
• Is there a record of the number of sick days and calls for voluntary medical insurance?

• Have you noticed any changes in the physical health of your employees during the pandemic? Did they differ during the introduction of the first lockdown and subsequent ones?

If the respondent answered affirmatively, the following questions were asked:

• How did you notice that? Have there been any changes in health, behaviour, productivity? Has the number of sick days and voluntary medical insurance appointments increased?
• Did these changes differ depending on the gender and age of the employee?
• Were the changes positive or negative? What factors influenced this?

If the respondent answered that the changes were negative, then the following questions were asked:

• How did you counteract them? What events were held and how effective were they?

• Have you given any recommendations on equipping a home office?
• Have you provided office equipment for the employee’s home?
• Have you covered the expenses of employees incurred in connection with the transition to remote work?
• What is included in VMI? Has its content changed after the start of the pandemic?
• Have you carried out measures to encourage vaccination?

• Is the issue of physical health of employees important for your company?
• Does your company measure mental health of employees?

• How is this indicator measured?

• Is a psychologist included in VMI? Do you conduct surveys about the mental state of employees?
• Have you noticed any changes in the physical health of your employees during the pandemic? Did they differ during the introduction of the first lockdown and subsequent ones?

• How did you notice that? Have there been any changes in health, behaviour, productivity? Did they differ among employees of different genders and ages?
• Were these changes positive or negative? What factors influenced this?

If the respondent answered that the changes were negative, then the following questions were asked:

• How did you counteract them?
• What events were held and how effective were they? How often do you communicate with your employees?
• Did you give instructions on switching to a remote format during the first lockdown?
• How do you cope with the loss of a sense of belonging among employees?
• Have you conducted events/trainings to recreate the psychological field and external stimuli of employee motivation at home (by field we mean the external environment that activates the desired mental state)?
• Is there any information about employees who are at risk due to the housing and family factor (presence of children, elderly relatives)? Have they received special support?

If the respondent answered that the changes were positive, then the following questions were asked:

• How did you maintain the health of employees? Have you held online meetings, team building trainings?

• Does your company measure the physical health of employees?

• How is this indicator measured?

• Has the productivity of employees changed after switching to remote work and in which direction? Did the results of the employees’ work in the first lockdown and subsequent ones differ?

• What do you associate the increase/decrease in employee productivity with?
• Have employees started working after hours: early in the morning and late at night, at weekends?

• Has the employee-subordinate interaction model changed?
• Has the process of monitoring the fulfillment of tasks, the effectiveness of the employee and his work schedule changed?
• Have you provided a structured work plan for the day?
• What time is considered working, when and how does an employee report on the work done, through what channels does he interact with the employer?
• Has the remuneration of labour (from fixed to piecework) changed due to the transition to a remote format?

• What would you prefer for your employees in the future in the absence of a pandemic: remote/hybrid /full-time format of work?