Income Shielded Minds Differently for Men and Women During COVID-19, Bayesian Study Finds

When the COVID-19 pandemic upended daily life in 2020, it did not weigh equally on every household. A new study published in PLOS Mental Health by Simon Busch-Moreno, Xiao Fu and Etienne B. Roesch of the University of Reading offers one of the most statistically careful portraits yet of how household income shaped self-reported symptoms of depression and generalised anxiety in the United Kingdom during the crisis. Using data from the COVID-19 Psychological Research Consortium study, known as C19PRC, the researchers applied Bayesian hierarchical ordered-logistic mediation models to trace how age and income influenced mental health across female and male respondents at two very different moments of the pandemic: the first wave in spring 2020, when lockdowns were at their most severe, and the sixth wave, when much of the acute disruption had subsided but its psychological aftershocks persisted.

The choice of statistical machinery matters here, and it is what sets the study apart from the many cross-sectional surveys that appeared during the pandemic. Self-reported mental health measures are typically collected on ordered scales, such as severity bands for depressive or anxious symptoms, and ordinary regression approaches can struggle with the ceiling and floor effects, skewness and confounding that such data exhibit. Ordered-logistic models are designed for exactly this kind of outcome, treating the observed categories as coarse readings of an underlying continuous liability. By wrapping that model in a Bayesian hierarchical framework, the authors could pool information across waves and demographic strata while allowing effects to vary, quantifying uncertainty in full posterior distributions rather than relying on single point estimates and thresholded p-values. The result is a more honest accounting of what the data can and cannot support.

Confounding is the central challenge in any observational study of income and mental health. Age and gender are entangled with both earnings and psychological wellbeing in ways that can easily masquerade as income effects. Older adults in the UK tended to report better mental health during the pandemic than younger adults, a finding that surprised many commentators early in the crisis but has since been replicated repeatedly. If age also correlates with accumulated wealth and stable housing, then a naive analysis might attribute to income what is really an age gradient, or vice versa. The mediation modelling approach adopted by Busch-Moreno and colleagues explicitly separates the direct associations of age and income from the pathways through which they may act, allowing the researchers to ask not merely whether income matters, but how its influence is structured across the life course and across genders.

The headline result is a striking asymmetry between men and women in how income related to mental health, and in how that relationship evolved over time. For male respondents, higher household income was consistently associated with fewer symptoms of depression and anxiety in both the first and sixth waves of the study. The protective association was present from the outset and remained stable as the pandemic wore on, suggesting that economic resources served as a reliable buffer for men throughout the entire period. For female respondents, the picture was more complicated. Early in the pandemic, higher income was only weakly linked to symptom levels, and the direction of the association offered little comfort: if anything, wealthier women reported marginally more symptoms during the initial wave, a pattern the authors interpret cautiously but which plausibly reflects the particular burdens that fell on women in the earliest weeks of lockdown.

By the sixth wave, however, the relationship for women had transformed. Higher income had become a substantially protective factor, associated with meaningfully lower levels of depression and anxiety. The researchers argue that this shift captures the delayed nature of the pandemic’s economic and social toll on women. In the first wave, the immediate shocks, school closures, cancelled childcare, disrupted work and the sudden collapse of social support networks, may have been distributed in ways that income could not easily offset, particularly for women who shouldered a disproportionate share of unpaid care and domestic labour. As the crisis lengthened, the cumulative advantages of financial security, the ability to maintain housing, absorb income losses, access private healthcare and reduce chronic financial stress, appear to have asserted themselves, turning income into a genuine shield for women’s mental health by the later stage of the study.

The gender comparison across income groups adds a further layer of nuance. Across most of the income distribution, male respondents reported better mental health than their female counterparts, consistent with the well-documented pattern that women experienced elevated rates of internalising symptoms during the pandemic. But at the lowest income group, that gap largely disappeared. Men and women in the most financially precarious households reported comparable levels of distress, and both fared markedly worse than their higher-income peers. This convergence at the bottom of the distribution is one of the study’s most policy-relevant findings. It suggests that severe economic hardship is a great leveller of gender differences in mental health, overwhelming the protective factors that otherwise sustain men’s relative advantage, and that the poorest households of both genders constituted a shared zone of high vulnerability throughout the pandemic.

The age findings reinforce the broader picture of a pandemic whose psychological burden was not distributed as many initially feared. Older age was linked to less depression and less anxiety in both genders across both waves. While older adults faced greater physical health risks from the virus itself, their mental health proved more resilient, possibly reflecting greater emotional regulation, more established life circumstances, accumulated coping experience and, in many cases, the financial stability that comes with later career stages or retirement. The mediation framework allowed the authors to examine whether age’s association with mental health operated partly through income, and the hierarchical structure of the model ensured that these pathways were estimated separately for men and women rather than assumed to be identical.

Technically, the Bayesian approach brings several advantages worth emphasising for readers unfamiliar with the methodology. Hierarchical models treat parameters at different levels of the data structure, individuals nested within waves and demographic groups, as drawn from shared distributions, which regularises estimates in sparse subgroups and prevents noisy cells from dominating conclusions. Ordered-logistic link functions respect the ordinal nature of the symptom scales rather than forcing them into a misleading linear metric. And the Bayesian posterior distributions give a full probabilistic description of each effect, making it possible to state, for example, the credible range within which the income effect for women in Wave 6 lies, rather than reducing the finding to a binary significance verdict. In a field where pandemic-era studies were often small, hastily analysed and prone to contradictory headlines, this level of inferential care is a meaningful contribution in its own right.

The authors are explicit about the limits of what observational survey data can establish. The C19PRC study measures self-reported symptoms rather than clinical diagnoses, and household income is a coarse proxy for the wider socioeconomic circumstances, debt, job security, housing quality and social capital, that shape mental health. Causal claims must therefore be made cautiously: the models illuminate patterns of association and plausible mediating structure, but they cannot fully rule out unmeasured confounders such as pre-existing mental health conditions that influence both earnings and symptom reporting. Longitudinal design, comparing Wave 1 and Wave 6 within the same study, strengthens the temporal interpretation considerably, yet the mechanisms behind the female income effect’s emergence remain a matter of informed interpretation rather than direct observation.

Even with those caveats, the study’s implications for policy are concrete. The authors argue that their evidence can inform targeted mental health interventions and economic support programmes in future public health emergencies. If income becomes protective for women only after the acute phase of a crisis passes, then early support cannot be designed around the assumption that financial assistance alone will immediately equalise psychological risk; it must be paired with measures addressing the care burdens and service disruptions that fall hardest on women in the first weeks of an emergency. Conversely, the persistent distress concentrated in the lowest income group, where gender differences vanish, points to a clear priority: households at the bottom of the income distribution of both genders warrant sustained, structural economic and psychological support, not only during the headline phase of a crisis but across its entire duration. As governments prepare for future pandemics and climate-related disruptions, the lesson from the UK’s pandemic experience is that the mental health consequences of a crisis are written, in large part, by the distribution of money, and that this distribution acts differently on different people at different times.

Subject of Research: Effects of household income and age on depression and anxiety symptoms by gender during the COVID-19 pandemic in the UK

Article Title: A Bayesian approach for exploring the effects of household income on self-reported mental health measures during the COVID-19 pandemic

Article References: Busch-Moreno, S., Fu, X., & Roesch, E. B. (2026). A Bayesian approach for exploring the effects of household income on self-reported mental health measures during the COVID-19 pandemic. PLOS Mental Health, 3(8), e0000691. https://doi.org/10.1371/journal.pmen.0000691

Image Credits: AI Generated

DOI: 10.1371/journal.pmen.0000691

Keywords: mental health, COVID-19, household income, depression, anxiety, Bayesian statistics, ordered-logistic models, gender differences, mediation analysis, C19PRC, United Kingdom, public health policy