Gender-Specific Pathways Linking Unemployment and Depression: A Cross-National Panel Study, 2010-2023

Abstract

Objective

To examine whether population-level associations between national unemployment and depression burden differ for women and men across 18 major global economies from January 1, 2010 to December 31, 2023.

Patients and Methods

This cross-national longitudinal panel study analyzed gender-disaggregated, country-level annual data across predefined valid observation years (2010, 2013, and from January 1, 2015 to December 31, 2023) from 18 Group of twenty nations. Age-standardized depression disability-adjusted life year rates (Global Burden of Disease 2021) served as the primary outcome; international labor organization standardized 25-64-year gender-specific unemployment rates were the exposure. Linear mixed-effects models with random country and year intercepts were specified as the primary analytic framework, adjusted for log-transformed per capita gross domestic product, Gini coefficient, and annual health expenditure (% gross domestic product). Country-level Pearson correlations and multi-dimensional sensitivity analyses were conducted to assess cross-national heterogeneity and result robustness.

Results

The final analytic dataset contained 282 valid country-year-gender complete observations. Each 1 percentage-point national unemployment rise corresponded to a statistically significant 16.04 (95% CI, 8.89-23.19) disability-adjusted life year per 100,000 increases in female depression burden (P<.001). No meaningful association was detected among males (β=−2.52; 95% CI, −10.67 to 5.63; P=.54). The unemployment × male gender interaction coefficient was strongly significant (β=−18.56; 95% CI, −25.22 to −11.90; P<.001). Substantial cross-country heterogeneity existed, with within-nation unemployment-depression correlation coefficients ranging from −0.87 to 0.85; gendered effects were amplified in high-income countries. All sensitivity analyses confirmed the robustness of the primary findings.

Conclusion

Macroeconomic unemployment growth generates disproportionately larger population-level depression burden among women across major global economies. Public health surveillance, labor stabilization policies, and clinical mental health intervention frameworks must integrate gender-sensitive design to mitigate female mental health risks during economic downturns.

The global burden of depression continues to rise, with recent estimates indicating over 280 million people affected worldwide.¹ Depression represents not only a leading cause of disability but also a considerable economic burden through productivity losses and healthcare costs.² While multiple social determinants contribute to depression risk, economic factors—particularly unemployment—have emerged as critical predictors of population mental health.³

The relationship between unemployment and poor mental health is well-established in the literature.⁴ Unemployment often leads to financial strain, loss of social status, and reduced social networks, all of which can exacerbate depressive symptoms.⁵ Meta-analytic evidence suggests unemployment increases the risk of depression by approximately 30%-40%.⁶ However, this association may not be uniform across demographic groups, particularly between genders.

Gender represents a fundamental social determinant of health that shapes both exposure to economic stressors and vulnerability to their mental health consequences.⁷ Traditional gender roles in many societies position men as primary breadwinners, potentially making unemployment more threatening to masculine identity.⁸ Conversely, women often face double burden scenarios where unemployment combines with persistent domestic responsibilities.⁹ Theoretical frameworks suggest these gendered experiences may translate to differential depression pathways.¹⁰

Empirical evidence on gender differences in unemployment-depression associations remains mixed. Some studies report stronger effects for men, citing masculine identity threats,¹¹ while others find greater impacts on women, emphasizing caregiving burdens and economic dependencies.¹² Most existing research has relied on individual-level data from single countries or limited regional contexts,¹³ leaving cross-national patterns poorly understood.

Cross-national studies offer unique advantages for examining social determinants of health, allowing researchers to account for contextual factors like welfare state generosity, labor market policies, and cultural norms.¹⁴ Such designs are particularly valuable for studying gender differences, as they capture variations in gender equality and economic opportunity across societies.¹⁵ However, few studies have employed cross-national panel designs to examine gender-specific unemployment effects on depression at the population-level.

This study addresses these gaps by analyzing data from 18 countries across 11 observation years (2010, 2013, and 2015-2023) to examine: (1) whether the association between unemployment and depression differs by gender; (2) the magnitude and direction of any gender differences; (3) the robustness of these patterns across different model specifications and country subgroups; and (4) cross-national heterogeneity in unemployment-depression associations. Drawing on stress process theory¹⁶ and gender role perspectives,¹⁷ we hypothesize that unemployment will show stronger associations with depression for women than men, reflecting gendered vulnerabilities in economic crises.

Patients and Methods

Study Design and Data Sources

This cross-national panel study analyzed aggregated population-level data from 18 countries across an 11 predefined valid observation window spanning January 1, 2010, to December 31, 2023. Valid single-year timepoints include 2010, 2013, and 2015-2023; consecutive years 2011-2012 and 2014 were fully excluded due to widespread covariate missingness across Group of twenty (G20) states. The core analytical objective was to estimate gender-specific associations between national unemployment rates and age-standardized depression burden. Countries were selected based on comprehensive data availability (Supplemental Figure 1, available online at http://www.mcpiqojournal.org).

Data Sources

Depression burden estimates were obtained from the Global Burden of Disease (GBD) Study 2021, which provides standardized, comparable estimates across countries and years using consistent methodology.¹⁸ Gender-specific unemployment rates were derived from the International Labour Organization (ILO) ILOSTAT modeled estimates database, adhering to internationally standardized 19th ICLS labor definitions.¹⁹ Country-level economic covariates, including gross domestic product per capita, Gini coefficient, and health expenditure as a percentage of GDP, were sourced from the World Bank World Development Indicators annual country-year dataset.²⁰ All data analyzed in this study are publicly available: GBD (https://ghdx.healthdata.org/gbd-results-tool), ILO (https://ilostat.ilo.org/data/), World Bank (https://databank.worldbank.org/source/world-development-indicators). Full reproducible analytical R code is provided in Supplemental Appendix 1, available online at http://www.mcpiqojournal.org.

Data Quality Assessment

All 3 primary data sources (GBD 2021, ILOSTAT, and World WDI) apply globally unified, peer-reviewed measurement frameworks to minimize cross-nation bias.

GBD Depression & DALY Data

GBD 2021 integrates nationally representative population surveys, psychiatric registry records, inpatient/outpatient clinical datasets, and peer-reviewed epidemiologic literature via Bayesian DisMod-MR 2.1 meta-regression modeling. Standardized DSM/ICD depression diagnostic criteria are applied uniformly across all jurisdictions. DALY metrics combine years lived with disability and years of life lost: global consensus disability weights for depressive episodes are derived from cross-cultural population preference surveys, consistently implemented in all country estimates. GBD reports country-specific uncertainty intervals to quantify measurement imprecision driven by local survey response/refusal rates (national survey response ranges 72%-94%). Cross-cultural depression screening instruments undergo standardized crosswalk calibration to improve inter-nation reliability, and all country burden estimates include measurement error terms in published uncertainty bounds. Individual survey-level refusal rates are not available in aggregated country-year public datasets, a limitation acknowledged in Section 4.4.

ILO Unemployment Data

All gender-stratified 25–64 unemployment metrics adhere to the 19th ICLS unified labor definition, harmonizing labor force classification across high- and middle-income nations.

World Bank Covariates

Annual country-year economic indicators are standardized to constant international currency units for cross-temporal and cross-national comparison.

Country Selection and Study Sample

Countries were systematically selected from the G20 nations based on complete gender-disaggregated annual data availability for all 11 valid observation years. The G20 framework was chosen as it represents approximately 85% of global GDP and two-thirds of the world's population, delivering an economically and geographically diverse sample.

Inclusion criteria applied uniformly to all original 20 G20 member states:

• 1)Complete gender-disaggregated, age-standardized depression DALY rates from GBD 2021 for every specified observation year (2010, 2013, 2015-2023).¹⁸
• 2)Complete ILO gender-stratified 25–64 unemployment annual series matching the identical 11-year window.¹⁹
• 3)Full annual country-year GDP per capita, Gini coefficient, and health expenditure records from World WDI for all study timepoints.²⁰

Two original G20 members failed all three data completeness rules and were excluded, yielding a final sample of 18 study countries: Argentina, Australia, Brazil, Canada, China, France, Germany, India, Indonesia, Italy, Japan, Mexico, Republic of Korea, Russian Federation, South Africa, Türkiye, United Kingdom, United States.

Panel Structure and Missing Data Breakdown

A theoretical fully balanced panel covering 18 countries × 11 years × 2 gender strata yields a maximum of 396 potential country-year-gender observations. Two layers of data removal reduced the final analytical sample to 282 complete records:

• •Full year exclusion: 2011, 2012, 2014 removed from all countries (108 theoretical observations eliminated, 27% of the full 396 dataset).
• •Within-window missingness: Among the remaining target 11 years, 114 additional country-year-gender rows contained incomplete unemployment or covariate data and were dropped before analysis (∼29% of the 288 remaining theoretical balanced panel).

A stepwise sample selection flowchart is visualized in Supplemental Figure 1; Supplemental Table 1 (available online at http://www.mcpiqojournal.org) now reports exact observation counts per individual country, stratified by gender and year. No minimum year threshold was enforced for country inclusion during primary analysis; sensitivity analyses excluding nations with ≤3 total observations are reported in Supplemental Table 2 (available online at http://www.mcpiqojournal.org) to test bias from sparse data contributors. Intraclass correlation coefficient decomposition showed 62% of total depression DALY variance originates from cross-country between-group differences, while 38% reflects within-country temporal changes over the study period.

Ethics Exemption Statement

This analysis exclusively uses fully aggregated, de-identified national public population datasets without access to individual human subject records. The institutional review board confirmed this research meets criteria for ethics review exemption; no informed consent procedures were required.

The selection of the 11-year observation window was guided by dual criteria: full cross-country data accessibility, and economic-historical relevance capturing post-2008 recovery (2010), mid-cycle stabilization (2013), and the full COVID-19 pandemic economic shock period (2015-2023). Descriptive country-level correlation analyses utilize the identical full 18-country sample.

Measures and Variables

Primary Outcome

The primary outcome was age-standardized depression disability-adjusted life year (DALY) rates per 100,000 population, separated by gender for each country-year. DALYs integrate two standardized population health metrics: years of life lost due to premature mortality and years lived with disability. Global uniform depressive disorder disability weightings are applied to enable valid cross-national comparison; age standardization eliminates confounding from divergent national population age pyramids.²¹ All covariate values utilized in regression are country-year matched annual records, with full year-by-country raw covariate data provided in Supplemental Table 3, available online at http://www.mcpiqojournal.org.

Primary Exposure

The primary exposure variable is annual gender-specific unemployment rates (%) for the working-age population aged 25-64, defined per official 19th ICLS standards: individuals without current employment, immediately available for work, and actively seeking labor market opportunities.²² This study’s main exposure metric captures cross-sectional annual unemployment prevalence and cannot differentiate continuous, episodic, or long-term unemployment spells. A supplementary sensitivity analysis using available long-term unemployment subrates (subset of nations with data) is presented in Supplemental Table 2; limitations around unemployment duration measurement are expanded in Section 4.4. Complete variable operational definitions and data source citations are documented in Supplemental Appendix.

Covariates

Three time-varying, country-level annual economic covariates were incorporated to mitigate confounding and evaluate macroeconomic contextual moderation:

• (1)GDP per capita (constant 2017 international dollars, natural log-transformed) to capture national average economic development level.²³
• (2)Gini coefficient (0-100 scale) measuring national household income inequality.²⁴
• (3)Health expenditure as a percentage of national GDP to proxy health care infrastructure and social safety net scope.²⁵

Statistical Analyses

Primary Analyses

All primary regression analyses implement linear mixed-effects models with random intercept terms for individual countries and individual calendar years, consistent with all results presented in Table 1. Two-way country-gender and year fixed-effects panel regression is conducted solely as a supplementary robustness check (Supplemental Table 2); Hausman test results (χ²=9.24, P=.16) validated random-effects specification as the appropriate primary framework.

Table 1.

Characteristics of Study Countries by Income Level (2010, 2013, and 2015-2023)

Characteristic	Overall (N=18)	High-income (n=11)	Middle-income (n=7)
Depression DALY rate, mean ± (SD), per 100,000
Women	930.0 ± 267.5	947.4 ± 232.1	983.0 ± 310.2
Men	560.0 ± 163.2	575.3 ± 153.4	558.1 ± 173.8
Female-to-male ratio	1.66 ± 0.26	1.65 ± 0.21	1.76 ± 0.32
Unemployment rate, mean ± (SD), %
Women	7.1 ± 4.1	7.1 ± 2.9	8.0 ± 5.4
Men	6.6 ± 2.9	6.7 ± 2.4	7.0 ± 3.6
Economic indicators, mean ± (SD)
GDP per capita, US$	26,762 ± 21,145	41,252 ± 13,678	8275 ± 5642
Gini coefficient	37.5 ± 7.5	34.3 ± 5.2	42.5 ± 8.3
Health expenditure, % GDP	8.8 (4.2)	10.8 (3.8)	5.9 (2.1)

Countries grouped by World Bank income classification.

Fully specified primary mixed-effects regression equation:

$${\text{DALY}}_{\text{it}}={\beta }_0+{\beta }_1{\text{Unemp}}_{it}+{\beta }_2\left({\text{Unemp}}_{it}\times {\text{Male}}_i\right)+{\beta }_3{\text{GDP}}_{it}+{\beta }_4{\text{Gini}}_{it}+{\beta }_5{\text{HealthExp}}_{it}+{\alpha }_i+{\gamma }_t+{ϵ}_{it}$$

Where:

• •i=unique country-gender cluster identifier
• •t=calendar observation year
• •αi=random country-gender intercept
• •γt=random year intercept
• •Malei binary indicator (1=male population subgroup; 0=female reference)
• •β2 interaction coefficient quantifies differential unemployment-depression slopes between men and women, with negative values indicating weaker male associations. Temporal gender unemployment fluctuations across country strata are fully captured by the unemployment × male interaction term paired with time-varying annual unemployment predictors, allowing the model to estimate distinct within-nation temporal changes for each gender subgroup. Full step-by-step analytical workflows and reproducible R v4.3.1 code (plm and fixest packages) are attached as Supplemental Appendix to enable full replication.

Secondary and Exploratory Analyses

To visualize national heterogeneity and complement regression outputs, country-specific gender-stratified Pearson correlation coefficients were calculated between annual unemployment rates and matched depression DALY burdens. Standard linear regression diagnostic testing (homoscedasticity, residual normality, and linearity) was completed; all diagnostic plots are available in Supplemental Figure 2, available online at http://www.mcpiqojournal.org.

Sensitivity and Robustness Analyses

We conducted comprehensive sensitivity analyses to assess the robustness of our findings:

• (1)Model specification checks: fixed-effects panel models, first-difference transformation mixed models.
• (2)Stratified subgroup regression: World Bank high-income versus middle-income nation stratification (full outputs completed in Supplemental Table 3, available online at http://www.mcpiqojournal.org).
• (3)Temporal subset testing: excluding 2010 post-crisis baseline; excluding 2020-2022 COVID pandemic shock years.
• (4)Alternative covariate adjustment sets to test confounding sensitivity.
• (5)Outlier trimming: removing country-year records with unemployment >20% or DALY >1200 per 100,000.
• (6)Sparse data country exclusion: models dropping nations with ≤3 total observations.
• (7)Weighted regression: observations weighted by each country’s total number of valid data points.
• (8)Jackknife leave-one-country iterative analysis to assess single-nation leverage on pooled estimates.
• (9)Alternative exposure: long-term unemployment subrates (limited country sample).

All statistical significance thresholds set at 2-tailed P<.05. Analysis adheres to STROBE observational cross-sectional panel reporting guidelines.²⁶

Statistical Software and Reporting

All analyses were conducted using R version 4.3.1 with the plm and fixest packages for panel data analysis.²⁷ Statistical significance was set at a 2-tailed P<.05. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for observational studies.²⁸

Results

Study Population and Descriptive Statistics

This cross-national panel study included 282 complete country-year-gender analytical records (141 women and 141 men subgroup observations) across 18 countries and 11 predefined valid years. Supplemental Table 1 lists exact yearly observation counts per country to visualize data imbalance (eg, India and South Africa with limited longitudinal records; China, France, Türkiye, and Russian Federation with near-full 10-year coverage). As summarized in Table 2, age-standardized depression DALY burden was persistently higher for women across all income tiers (overall female mean 930.0 vs male 560.0 per 100,000; female/male ratio=1.66). Average gender unemployment gaps were modest (women 7.1% and men 6.6%) with substantial cross-country variation.29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39 High-income nations averaged 10.8% GDP spent on health care, whereas middle-income nations carried higher female depression population burden. The full country sample selection flowchart is Supplemental Figure 1; complete annual covariate values by country are listed in Supplemental Table 3.

Table 2.

Mixed-Effects Model of Unemployment and Depression DALY Rates with Gender Interaction^a

Variable	β (95% CI)	SE	t Valuea	P
Unemployment rate, %	16.04 (8.89-23.19)	3.65	4.40	<.001
Male (ref: Female)	−238.80 (−287.66 to −189.94)	24.93	−9.58	<.001
Unemployment × Male	−18.56 (−25.22 to −11.90)	3.40	−5.46	<.001
Log (GDP per capita)	−7.02 (−96.08 to 82.04)	45.37	−0.15	.88
Gini coefficient	1.13 (−5.88 to 8.14)	3.58	0.32	.75
Health expenditure, % GDP	13.12 (−7.15 to 33.38)	10.34	1.27	.21

^aModel includes random intercepts for country and year (N=282). Marginal R²=0.42; conditional R²=0.85.

Primary Analysis: Gendered Association of Unemployment With Depression

Linear mixed-effects random intercept regression models (primary prespecified analysis) generated significant gender differentiation in unemployment-linked depression burden (Table 2). A one percentage-point national unemployment increase predicted a 16.04 (95% CI, 8.89-23.19) per 100,000 DALY rise among women (P<.001). Male populations showed a negligible, nonsignificant inverse association (β=−2.52; 95% CI, −10.67 to 5.63; P=.54). The cross-gender interaction term was highly statistically significant (β=−18.56; 95% CI, −25.22 to −11.90; P<.001), confirming substantially stronger female mental health risk linked to labor market contraction. Marginal gender effect slopes are visualized in Figure 1.

Figure 1.

Marginal effects of unemployment on depression by gender. Predicted depression DALY rates across unemployment levels (2%-20%), showing steeper slope for women (orange) versus men (blue). Based on linear mixed-effects model with 95% CIs.

Figure 1 visually depicts these marginal effects, illustrating the steeper slope of the unemployment-depression relationship for women across the observed range of unemployment rates (2%-20%).

Substantial Cross-National Heterogeneity

Primary gender-differential effects remained consistent across robustness testing, yet country-level stratified analyses revealed pronounced cross-national heterogeneity in unemployment-depression correlations, with substantial variation in both magnitude and direction across countries (Figure 2). Wide country-specific 95% CIs visible in Supplemental Figure 3 (available online at http://www.mcpiqojournal.org) predominantly stem from nations with limited annual longitudinal observations (India and South Africa), lower national data quality in middle-income jurisdictions, and divergent domestic social protection frameworks. Random-effects meta-analysis pooled cross-country correlation was nonsignificant (r=−0.13; 95% CI, −0.44 to 0.17; P=.39), reporting lack of uniform within-country trends. Income subgroup regression (Supplemental Table 4, fully populated now) reported considerably stronger gender interaction effects in high-income nations, with marginal nonsignificant interaction among middle-income countries. Supplemental Figure 4 (available online at http://www.mcpiqojournal.org) visualizes income-stratified effect differences; Supplemental Figure 5 (available online at http://www.mcpiqojournal.org) evaluates macroeconomic moderators; Supplemental Figure 6 (available online at http://www.mcpiqojournal.org) reports sensitivity results excluding extreme outlier country-year records.

Figure 2.

Country-level correlations between unemployment and depression DALY rates. Bar chart showing Pearson correlation coefficients for 18 countries. Orange indicates positive correlations and blue indicates negative correlations. Marker shapes denote statistical significance (P<.05).

Sensitivity and Robustness Analyses

All sensitivity specifications reproduced the core finding of significantly stronger female unemployment-depression associations (Supplemental Table 2). Excluding the 2010 baseline year retained significant gender interaction (β=−23.52; P<.001); removing 2020-2022 pandemic years also preserved the interaction signal (β=−14.66; P<.001). Aggregate pooled gender-stratified bivariate correlations mirrored regression outputs: women r=0.27 (95% CI, 0.15-0.38), men r=−0.03 (95% CI, −0.15 to 0.10), with statistically significant cross-group difference (z=2.50; P=.012). Pandemic co-trends of unemployment and depression burden are plotted in Supplemental Figure 7, available online at http://www.mcpiqojournal.org.

Subgroup stratification by national income tier showed a significant gender interaction in high-income economies (P<.001), whereas the interaction was weaker and did not reach statistical significance in middle-income states (P=.085). Figure 3 visualizes seven representative national scatter plots; the 7 nations were intentionally selected to span high/middle-income, positive/negative correlation directions, and sparse/dense observation coverage to concisely illustrate heterogeneous patterns. Full 18-country-gender-stratified scatter plots are provided in Supplemental Figure 8 (available online at http://www.mcpiqojournal.org) for complete transparency.

Figure 3.

Country-specific relationships for representative countries. Scatter plot matrix showing associations for 7 representative countries. Blue dots represent annual observations and orange lines show linear regression fits with 95% CIs.

Discussion

Principal Findings

This cross-national longitudinal panel analysis identifies a robust, consistent gender disparity in population-level depression burden associated with rising national unemployment. Each percentage-point labor market contraction correlates with a 16.04 per 100,000 DALY elevation in female depressive disease burden (95% CI 0.08-0.24; P<.001), with no detectable parallel population-level association among men. The unemployment × male gender interaction coefficient remains highly significant across all model and sample sensitivity tests. While consistent average gender disparity exists globally, strong cross-national contextual heterogeneity modifies effect magnitude and directional trends.

Interpretation in Context

Our core finding of greater female mental health vulnerability aligns with stress process theory and gender role structural frameworks.^9,16 Three key national contextual mechanisms jointly moderate the gendered unemployment-depression link, though this ecological cross-national dataset cannot isolate independent causal effects of each factor (a limitation noted below):

Welfare State Generosity

Generous universal childcare subsidies, unconditional unemployment cash transfers, and universal health care buffer female dual care-labor burdens during job loss. Nations with limited social safety nets amplify women’s combined financial and domestic care strain post-unemployment, widening the gender depression gap.^35,36 Health expenditure per GDP partially proxies social protection capacity within our covariate set but cannot fully capture targeted family welfare policies.

Labor Market Regulatory Frameworks

Mandatory parental leave, legal anti-gender employment discrimination protections, and accessible part-time re-entry pathways reduce women's post-job-loss economic precarity.^31,39 Countries with rigid full-time-only labor norms create disproportionate re-employment barriers for women with caregiving responsibilities, strengthening the unemployment-depression linkage among females.

Cultural Gender Role Norms

Societies assigning primary unpaid domestic and childcare labor to women create a double burden activated upon job loss.^29,32,33 Additionally, cross-cultural variation in mental health help-seeking and depression diagnostic labeling alters recorded DALY burden: male psychological distress often manifests as substance use or externalizing behaviors (not captured in depression DALY metrics), while female affective symptoms are more frequently clinically diagnosed and counted in population burden estimates.³⁴

Our results diverge from single-country individual-level studies reporting stronger male unemployment mental health harm,¹¹ which can be attributed to methodological differences: individual microdata captures short-term identity distress among laid-off men, although this population-level ecological analysis quantifies long-run aggregate national depression disease burden across full gender cohorts.¹³ Cross-nation heterogeneity (with correlations ranging from strongly negative to strongly positive) confirms national social, labor, and cultural contexts act as powerful effect modifiers.¹⁴ Stronger gender interaction in high-income nations may stem from amplified normative workforce participation expectations for women in advanced economies, rendering involuntary unemployment comparatively more psychologically distressing for women in these settings.^30,31

Policy Implications

Gender-Sensitive Macroeconomic Stabilization Policy

Labor crisis relief packages must integrate universal affordable childcare access to mitigate women's dual care-work burden during unemployment shocks.^35,36

Integrated Clinical and Occupational Mental Health Screening

Primary care, workplace health, and public employment service programs should implement routine depression risk screening targeting unemployed women during recessionary periods.^37,38

Equitable Social Protection Expansion

Labor market anti-discrimination rules, flexible work legislation, and gender-balanced unemployment benefit eligibility reduce the female mental health penalty of job loss.³⁹

Strengths and Limitations

Strengths

Multi-country longitudinal panel design drawing standardized ILO/GBD/World Bank global harmonized population data.18, 19, 20 Consistent gender-stratified exposure and outcome metrics with extensive multi-dimensional sensitivity testing across model forms, time subsets, income groups, and outlier removal. Random-effects regression accounts for unobserved permanent cross-country differences and universal yearly macroeconomic shocks.²⁶ Full Supplemental Appendix including country-level observation counts, annual covariate tables, complete diagnostic plots, and reproducible analytical code, maximize methodological transparency.

Limitations

Ecological study design

Population-level aggregate country-year data cannot establish individual-level causal inference; observed national gender associations do not guarantee identical individual-level relationships.¹³

Unmeasured cross-national confounding variables

Divergent national mental health screening protocols, variable clinical depression diagnostic thresholds, cultural stigma around mental illness, uneven health care access/utilization, and divergent gender role cultural norms are not adjusted in the model and may bias cross-country comparisons.^7,14

Data Measurement Constraints

• 1.GBD survey response/refusal rates are not available at aggregated country-year level, introducing unquantified sampling bias in some middle-income nations.¹⁸
• 2.The annual cross-sectional unemployment metric cannot distinguish short episodic job loss versus chronic long-term unemployment; limited cross-nation long-term unemployment data prevented full stratified testing of unemployment duration effects.⁶
• 3.Estimates from countries with only 1-3 annual observations carry elevated statistical uncertainty, visible in wide country-specific CIs.

Panel imbalance

Although the study restricts analysis to identical predefined calendar years, unequal per country longitudinal data volume creates unbalanced country-year-gender strata, mitigated through weighted and leave-one-country jackknife sensitivity tests.

Inability to disentangle 3 macro moderators (welfare generosity, labor policy, and cultural norms)

The study’s aggregate national covariates lack granular policy variables to independently quantify each contextual factor's unique moderating impact on gendered unemployment-depression pathways.

Conclusion

Across the 18 major global economies analyzed, rising national unemployment is on average a gendered population mental health risk factor, producing considerably larger depression burden increases for women than men. This consistent average gender disparity persists even amid substantial cross-national heterogeneity in effect magnitude and directional correlation patterns. Economic crisis response frameworks, population mental health surveillance systems, and clinical depression screening protocols must incorporate gender-informed design to mitigate disproportionate female mental health harms during labor market downturns.

Potential Comepting Interests

The authors report no competing interests.

Ethics Statement

This study used fully aggregated, de-identified national-level public population datasets without individual human subject records. Institutional review board exemption was obtained, and informed consent was not required.