On 11 March 2020, the World Health Organization (WHO) declared COVID-19 a pandemic1. Health authorities quickly realized that the pandemic posed a physical and mental health threat. On 18 March 2020, the WHO wrote, "this time of crisis is generating stress throughout the population" and called upon policymakers, health care professionals, and the general population to "support mental and psychosocial wellbeing in different target groups during the outbreak." Based on experience with previous pandemics, such as the Spanish flu (1918–1920), the Asiatic flu (1956–1957), the Severe Acute Respiratory Syndrome (SARS, 2002–2003), the "Swine" flu (2009), and Ebola (2013–2014), researchers proposed that the mental health consequences of the COVID-19 crisis were likely to be present for a long time and peak later than the actual pandemic2–4. They called upon the research community to study the mental health effects of COVID-19.
COVID-19 and mental health
Studies of the development of mental health and wellbeing during the pandemic have found mixed results. Some studies observed negative mental health outcomes5–10, whereas others reported positive aspects of the pandemic11,12 or found no evidence of changes in mental health outcomes during the pandemic11,13,14. Robinson et al. observed a high degree of unexplained heterogeneity in mental health responses to COVID-198. The most reported symptoms have been post-traumatic stress5–7, depression6–8, 10, and anxiety5,9–12,15−17. Other reported symptoms include insomnia5,18 and loneliness19.
Scholars have proposed three potential pathways by which the pandemic may affect mental health: the disease itself, the quarantine measures, and the economic consequences of the pandemic. As for the first pathway, the disease (threat) itself may directly affect mental health. People may fear that they or their significant others may be infected4,15. Those who caught the disease may suffer from post-infection consequences, such as fatigue and pain20 and may fear being a burden to their environment4. The second pathway acknowledges that measures to contain the disease, such as quarantine and social distancing, may affect mental health by reducing opportunities for physical and mental health activities, such as recreational activities and routines15,21−23. The third pathway assumes that mental health may suffer from the economic consequences of the pandemic15,22. In the current study, we focus on this economic pathway, particularly the potential role of financial stress in explaining mental health trajectories.
Research suggests that socio-economically disadvantaged groups are more vulnerable to decreasing mental health during the pandemic11,12,19–21,24−27, although the evidence is mixed5,28,29. Notably, Pijpker et al. found no differences in mental health between low and high socioeconomic status respondents in a sample of the Dutch population30. The literature suggests that the three pathways mentioned above play a role in the adverse mental health consequences for socio-economically disadvantaged groups. First, low socioeconomic status is associated with a higher chance of COVID-19 infection, resulting in higher mental distress31,32. Second, low-income jobs are less likely to be executed from home, so they are most affected by the lockdown and social distancing measures33. This may also result in an increased level of role conflicts having to combine work and family obligations33,34. Third, socio-economically disadvantaged and financially vulnerable groups are more likely to suffer from the economic consequences of the pandemic. They are more likely to work in sectors that suffered most from COVID-19, such as restaurants, travel, entertainment, and certain retail branches. Also, workers most likely to be affected by unemployment are less educated and have fewer financial resources. An empirical study among people across the European Union in the first six months of the pandemic showed high job insecurity among those with temporary contracts. Also, the unemployed had difficulty making ends meet, and people with low job insecurity had considerable mental health issues35. A cross-sectional study among 1,441 US citizens in the first two months of the pandemic showed that financial stressors and low assets were associated with higher odds of depression36. Financial stressors were defined as losing a job, a member of the household losing a job, having financial problems, and having difficulty paying rent. Assets included social assets (education and marital status), physical assets (homeownership), and financial assets (household income and household savings).
Entrepreneurs, particularly self-employed, are another group that suffered from the economic consequences of the pandemic. They experienced a higher loss of working hours than others during the pandemic37,38. Several studies indicate that self-employed are susceptible to mental health problems due to the economic consequences of the pandemic26,39,40.
Financial stress and health
Research on the relationship between the economic situation of households and mental and physical health has a long history. In the 1980s, Rose and Marmot followed more than 17,000 municipal officials in London. Their well-known Whitehall Studies showed that lower-paid civil servants were more likely to develop cardiovascular disease than their colleagues with higher positions41. Since then, studies have shown the relationship between poverty and many different types of physical and mental conditions, such as diabetes, cancer, chronic lung disease, schizophrenia, depression, substance use, and anxiety disorders42–44.
There is also evidence that stress plays a mediating role in the relationship between poverty and health45. Poor households often have fewer resources (for example, financial buffers and social support) to deal with life events. This lack of resources may result in stress and health problems46,47. Debt is also associated with stress and mental problems48. Income fluctuations cause uncertainty and, therefore, stress49,50. Having a financial buffer to deal with setbacks reduces stress and increases financial wellbeing51. Simonse et al. supported these notions by showing that low income, a sufficient financial buffer, and debts are positively related to financial stress52.
Gaps in the literature and the current study
Although the evidence is mixed, most studies have found that mental health declined during the COVID-19 pandemic. Research also indicated a high degree of unexplained heterogeneity in mental health trajectories. Many studies on COVID-19 and mental health cannot adequately examine these trajectories because these studies have cross-sectional designs. When studies used longitudinal designs, data collections (understandably) started only after the pandemic outbreak. The current study examined mental health trajectories by including data collected before and after the pandemic outbreak; this was possible by connecting long-running data on mental health to an ongoing data collection on financial stress53. The current study specifically focused on how financial stress and vulnerability might explain these mental health trajectories.
Moreover, we examined how households' financial situation before COVID-19 and their income development during COVID-19 explained financial stress. Having a financial buffer may protect against financial stress because a financial buffer can absorb income loss or unexpected expenditures. Especially in economically uncertain times, lacking a financial buffer may result in feelings of not being in control of one's financial situation and in worries about not being able to meet financial obligations. Thus, the absence of a buffer may result in increased financial stress. Similarly, having debts in economically uncertain times may trigger worries about being unable to repay them because of the anticipation of future income drops. Also, having debts may increase feelings of dependency on others54. Thus, having debts in economically uncertain times such as COVID-19 may increase financial stress. Income is likely to be negatively associated with financial stress. Low-income households are more vulnerable to becoming unemployed. Also, low-income households may have fewer opportunities to cut spending. Finally, it stands to reason that income and financial stress are dynamically related: income drops are likely associated with increasing financial stress. We tested three hypotheses:
-
Changes in financial stress during COVID-19 negatively relate to mental health trajectories.
-
Adverse income development during COVID-19 and low incomes, insufficient buffers, and debts before COVID-19 positively relate to financial stress trajectories during COVID-19.
-
Financial stress trajectories mediate the association between financial vulnerability (in terms of lacking sufficient buffers, having debts, low incomes, and adverse income changes) and mental health trajectories.
Research design
Data and variables
We used longitudinal data from the Long-running Internet Studies on Social Sciences (LISS) panel (initial N = 1,114). The LISS panel consists of a representative sample of approximately 5,000 households, drawn by the Central Bureau of Statistics of the Netherlands55. Respondents fill in monthly questions on various topics, such as health, family, work, personality, and economic situation. To ensure that vulnerable households can participate, they are supplied with a laptop and internet connection if necessary. The rich dataset enabled us to examine the relationship between developments in households' economic situation, financial stress, and mental health. We used three measurements to compare the situations before and during COVID-19: April – November 2018 (t = 0), December 2019 – March 2020 (t = 1), and December 2020 – March 2021 (t = 2).
Mental health
To assess mental health, we used the Mental Health Index (MHI-5), a brief and reliable measure of mental health, and a subset of the validated SF-36 Health Survey56 (Cronbach's α = .87). MHI-5 asks respondents how often they felt nervous, down, calm, depressed, and happy in recent weeks (see Appendix for the complete questionnaire). Respondents' scores on each item ranged from 1 (never) to 6 (continuously). We recoded the items so that a higher score reflected better mental health. LISS' health questionnaire measures MHI-5 every year. We used the measurements administered in November / December 2018, 2019, and 2020.
Explanatory variables
We used the Psychological Inventory of Financial Scarcity (PIFS) (Cronbach's α = .93) to measure financial stress53. The PIFS assesses the subjective experience of financial scarcity and captures appraisals of insufficient financial resources and lack of control over one's financial situation, responses in terms of financial rumination and worry, and a short-term focus (see Appendix). Respondents' scores on each item range from 1 (totally disagree) to 7 (totally agree). Higher scores indicate more financial stress. The PIFS was administered in April 2018, February 2020, and August 2020.
We included three aspects of a household's economic situation in the analyses because previous research has shown that these are strongly associated with financial stress: income, financial buffers, and the number of debts52. We used monthly income data for 2018, 2019, and 2020. For financial buffers and debts, we used the last available measurement before the outbreak of COVID-19. This measurement was held in June/July 2019, and it concerned households' financial situation at the end of 2018.
Income. The LISS panel measures net monthly household income in euros. We summed the net monthly household incomes for 2018, 2019, and 2020 to obtain yearly net household incomes. Since the needs of a household grow with each additional member, we corrected for household size. To consider economies of scale, we adjusted household income by dividing it by the square root of household size, in line with OECD guidance57. We included income at the first measurement and income changes between the three measurements as independent variables in our model.
Buffer. Savings may serve as buffers against unexpected expenditures and income shocks. Ruberton et al. stressed "the importance of holding minimal financial savings, but also the relative unimportance of having wealth above sufficiency levels" 51. We, therefore, defined buffer as a dichotomous variable that equaled one if a household's liquid assets exceeded a threshold depending on income and household size and zero otherwise. We argue that higher-income households need a higher buffer because they have more fixed expenditures and own more property. Based on the Buffer Calculator for the Netherlands, provided by the National Institute for Family Finance Information58, we used the following formula to define the threshold for having sufficient buffer: required buffer = € 600 + monthly income + € 400 * household size. Respondents were asked: "What was the total balance of your banking account, savings accounts, term deposit accounts, savings bonds or savings certificates, and bank savings schemes on 31 December 2018?". If they responded "I don't know," the questionnaire asked, "To what category did the total balance (total value) belong on 31 December 2018 (positive or negative)?" and given 15 categories (less than € 50 to € 25.000 or more). We used the category midpoints to calculate whether the household's liquid assets exceeded the buffer threshold.
Debt. Previous studies have indicated that the number of debts is positively related to financial stress and often more strongly than the total amount of debts52,59. We used the number of different types of loans as a proxy for the number of debts. We excluded mortgages and study loans and focused on consumer credit. The survey asked respondents to indicate whether they had (a) one or more personal loans, revolving credit arrangement(s), or financing credit(s) based on a hire-purchase or installment plan, (b) a loan or credit arrangement based on a pledge, (c) overdue payments on one or more credit cards (d) money loaned from family, friends, or acquaintances, and (e) any other credits, loans or debts. We observed that the number of respondents with more than one type of debt was negligible. We, therefore, defined debt as a dichotomous variable that equaled one if a household had at least one type of debt and zero otherwise.
Control variables
We used age, education level, household composition, and personality traits as control variables in our analyses. Age and education level may be confounders of the association between income and financial stress. Furthermore, previous research has shown that mental health during COVID-19 may differ between households with compositions12,19,20,20,29. We distinguished four household types: (1) no partner, no children, (2) children, no partner, (3) partner, no children, and (4) partner with children. Several studies have indicated that personality traits influence saving behavior, impulse buying, and debt. The findings for extraversion, conscientiousness, and emotional stability are the most consistent. For example, conscientiousness is positively associated with savings and negatively associated with debt60. Extraversion negatively predicts debt61. Of the Big-Five personality traits, emotional stability and conscientiousness show the strongest (negative) association with experienced financial scarcity62. For the current study, we used items from the 50-item Goldberg's Big-Five index concerning the trait variables emotional stability, conscientiousness, and extraversion (α = .77, .89, and .87, respectively)63. We parsed out the variance between six controls (age, education level, household composition, emotional stability, conscientiousness, and extraversion) and the independent variables. This allows us to examine the unique relationship between economic variables, financial stress, and mental health.
Model
A linear mixed model enables analyzing the dynamic relationship between variables of interest within and across subjects. We were interested in how financial stress and mental health trajectories were associated. Moreover, we wanted to establish indirect relations between income development during COVID-19, income, financial buffers, and debts before COVID-19 on the one hand, and mental health trajectories on the other. In addition, we wished to allow for individual heterogeneity in mental health. We, therefore, chose a random intercepts model, meaning that the average mental health and financial stress over the three observations may differ between individuals. At the same time, the slopes are homogeneous for the sample. We included time as an independent variable to test whether mental health and financial stress changed between measurements. Also, to test whether the relationship between mental health and financial stress differed between the three measurements, we added time as a moderator to our model.
Furthermore, we did not impose any restrictions in advance on the covariance between observations at different measurement moments (unstructured covariances). To ease the interpretation of the parameter estimates, we standardized the numeric variables. We estimated a mediation model to test our hypotheses, where mental health was the dependent variable, financial stress was the mediator, and income, buffer, and debt were the independent variables. The following equations describe the model mathematically:
\({\text{y}}_{\text{t}}={\alpha }+\beta {\text{x}}_{\text{t}}+\gamma {\text{m}}_{\text{t}}+\delta \text{z}+{\epsilon }_{1}\text{t}+ {\epsilon }_{2}\text{t}{\text{m}}_{\text{t}}+ {{\eta }}_{\text{t}}\) | (1) |
\({\text{m}}_{\text{t}}=\alpha \text{'}+{\beta }^{\text{'}}{\text{x}}_{\text{t}}+ {\delta }^{\text{'}}\text{z} +{{\alpha }}^{{\prime }}+ {\epsilon \text{'}}_{1}\text{t}+{{\eta }{\prime }}_{\text{t}}\) | (2) |
In these equations, \(t\) represents the time of the measurement (t = 1, 2, 3), \({\text{y}}_{\text{t}}\) is a vector with length N = 1,114 with the dependent variable mental health at measurement t for each of the respondents, \({\text{m}}_{\text{t}}\)is a vector with the mediator financial stress at measurement t. \({\text{x}}_{\text{t}}\)Is a vector with the time-dependent variable income at time t. \(z\) Is a matrix with constant variables over time: the independent variables (buffer and debt), control variables (age, education level, gender, household composition, and personality traits). \({\alpha }\) and \({{\alpha }}^{{\prime }}\)Are vectors with random intercepts. \(\beta , \gamma , \delta ,{\epsilon }_{1},{\epsilon }_{2}, {\beta }^{\text{'}}, {\gamma }^{\text{'}}, {\delta }^{\text{'}},{\epsilon \text{'}}_{1}\) are the regression coefficients, and \({{\eta }}_{\text{t}}\) and \({{\eta }{\prime }}_{\text{t}}\)are the prediction errors.