The prevalence of standard, full-time employment in the United States (US) has been decreasing since the 1970s (Kalleberg, 2011). This trend was exacerbated by the 2008 Financial Crisis and COVID-19 Pandemic, both of which were associated with increases in part-time employment in the US (Cowan, 2020; Kalleberg and von Wachter, 2017). While some aspects of non-standard employment may be seen as beneficial (e.g., increased scheduling flexibility), decreases in standard employment have generally been associated with rising precarious employment (PE) in the US (Kalleberg, 2011). While there is no single, accepted definition of PE, one systematic review of definitions and operationalizations identified three core dimensions of the concept: employment insecurity, income inadequacy, and lack of workplace rights/protections (Kreshpaj et al., 2020). Moreover, recent literature has highlighted the importance of measuring PE’s components in a single, multi-dimensional operationalization to capture the full breadth of its effects on worker health (Hajat et al., 2023; Vives et al., 2020).

Mounting evidence suggests PE contributes to poor health (Hajat et al., 2023). In addition to increasing risk of disease, such as myocardial infarction (Matilla-Santander et al., 2022b), PE has a measurable effect on workers’ quality of life (Bhattacharya and Ray, 2021), potentially resulting in poor mental health (Vives et al., 2013). Results from European studies suggest a causal relationship between PE and poor mental health, including one cohort study (Canivet et al., 2016) and a cross-sectional study finding a dose-response relationship between the two (Vives et al., 2013). While fewer PE studies have been conducted in the US, published research indicates that workers in PE arrangements are more likely to report depression and poor mental health than those in standard employment relationships (Andrea et al., 2022; Peckham et al., 2019).

PE is theorized to lead to poor mental health through various pathways. For example, low control over one’s job and working schedule are associated with poor mental health (Dalgard et al., 2009; Lee and Kawachi, 2021). Other psychosocial occupation hazards experienced by workers in PE arrangements include workplace bullying and discrimination (Kvart et al., 2021). Outside of the workplace, the economic and social uncertainties inherent to precarious work can divert workers’ time and energy away from other meaningful social interactions (e.g., time spent with family or in one’s community), resulting in feelings of social isolation (Irvine and Rose, 2022).

Decades of worsening employment quality in the US have drawn attention to PE’s ramifications for worker health (Howard, 2017). However, few US studies have investigated associations between PE and mental health. For example, a recent review by Hajat et al. identified only four US studies examining PE’s relationship to mental health (Hajat et al., 2023). Of these, three used a typological approach towards measuring precarity (Andrea et al., 2022; Eisenberg-Guyot et al., 2020; Peckham et al., 2019), in which qualitatively distinct employment classes are identified within an analytic sample. Only one study (Bhattacharya and Ray, 2021) identified by Hajat et al. used a summative score approach, in which PE is measured as a continuum, allowing for the measurement of mental health outcomes at increasing levels of precarity. Other US studies have used a summative approach to measure associations with perceived stress during the COVID-19 pandemic (Oddo et al., 2023a) or have measured associations between individual PE measures, such as insecure income, with psychological stress (Thomas et al., 2022). To expand on these previous studies, we examined associations between PE, operationalized as a summative, multi-dimensional exposure, and self-rated mental health using 2008–2021 data from the US Medical Expenditure Panel Survey (MEPS), a nationally representative survey that collects data on health, employment, healthcare utilization, and health insurance coverage in the US.

This study used data from US MEPS (U.S. Department of Health & Human Services Agency for Healthcare Research and Quality, 2024). MEPS data are collected from a nationally representative sample of respondents in five rounds over a two-year period, or panel, during which information is collected on participants’ demographics, healthcare utilization, employment arrangements, and health insurance coverage. A new panel is selected each year. To increase sample size, we pooled data from MEPS panels 13–25, corresponding to years 2008–2021. This timeframe includes the 2008 financial crisis and its aftermath, during which the US experienced several labor market changes, such as increases in non-voluntary part-time work (Kalleberg and von Wachter, 2017). MEPS Panel 25, which includes data collected in 2020–2021, was the most recent panel for which five rounds of data were available at the time of analysis. To allow for analysis across five rounds of data collection, the longitudinal MEPS files for each panel were used. Information pertaining to how a participant was paid (i.e., salaried or not) and the year participants started their current main job was extracted from MEPS Jobs data files. This activity was reviewed by CDC, deemed research not involving human subjects, and was conducted consistent with applicable federal law and CDC policy.

To maintain consistency with previous literature, our PE operationalization included MEPS variables broadly representing three domains identified in a previous systematic review of PE definitions and operationalizations: employment security, income inadequacy, and lack of rights/protections (Kreshpaj et al., 2020). Variable selection was also informed by previous studies creating multi-dimensional measures of PE (Baek et al., 2023; Bhattacharya and Ray, 2021; Oddo et al., 2023a; Vives et al., 2010), as well as consideration of what information was available in our sample. For example, while some definitions of PE included measures related to disempowerment and education/training opportunities (Oddo et al., 2021), variables related to these concepts were not identified in MEPS. Ultimately, thirteen variables were selected for inclusion in our PE measure (Table 1). For each variable, participants were assigned a score of “1” if they reported the value of that variable associated with PE (e.g., an individual whose current main job is temporary would receive a score of “1” for that measure, while a non-temporary employee would receive “0”). While previous research using a summative, multi-dimensional PE measure to study associations with mental health have weighted individual employment variables to creating equal influence of each PE domains, these used formal scale validation approaches (Bhattacharya and Ray, 2021) or measured precarity using previous-validated precarity scales (Vives et al., 2013). As no formal validation of our PE measure was performed in this study, these variables were weighted equally and summed, resulting in a scale ranging from 0–13; a higher value was indicative of higher PE.

The outcome of this study was self-rated mental health. MEPS participants are asked to rate their mental health at each of five rounds using a Likert Scale with levels of “poor”, “fair”, “good”, “very good”, and “excellent”; we dichotomized this measure into a binary outcome with “poor/fair” and “good/very good/excellent” strata (McAlpine et al., 2018). Data were also collected on age, sex, race/ethnicity, nativity status, education, family size, marital status, region, and occupational group; each of these variables were also self-reported. As published research suggests that PE is differentially distributed among these factors (Hajat et al., 2023; Peckham et al., 2019; Thomas et al., 2022), these were considered potential confounders of the association between PE and poor mental health.

Our sample included 57,529 participants from MEPS panels 13–25, corresponding to 287,516 survey responses across five rounds of data collection. The weighted total for this sample was 106,990,169 US workers employed throughout the years 2008–2021. Of these, 60.5%, 26.1%, and 13.8% were in the low, medium, and high PE strata, respectively. Compared to the low PE strata, participants with high PE tended to be younger (mean = 38.7; SD = 16.6) and were more frequently female (56.3%), Hispanic (23.2%), foreign-born (22.4%), single (as opposed to married or divorced; 46.5%), non-college-educated (46.4%), and employed in service occupations (33.3%). Regarding PE-related variables, those with high PE frequently reported not belonging to a union (98.7%), not having leave to see a doctor (98.7%), not having paid sick leave (97.4%), not having a pension plan (96.2%), not having paid vacation (95.2%), and being paid though a non-salaried pay schedule (92.5%). Conversely, working temporary or seasonal jobs were the PE variables reported least-frequently by participants regardless of PE strata. Weighted descriptive statistics by self-reported mental health strata are reported in Supplemental Table 1.

The proportion of participants reporting poor/fair self-rated mental health was 3.2%, 4.2%, and 5.8% in the low, medium, and high PE strata, respectively (Table 3). Adjusted for potential confounders, those with medium and high PE values had 1.21 (95% CI = 1.11, 1.32) and 1.51 (95% CI = 1.36, 1.68) times the odds of reporting poor/fair mental health, respectively, compared to those with low PE values. Sensitivity analyses containing only time-varying covariates or using a complete case sample did not differ significantly from our main results (Supplementary Table 2).

Marginal effects analysis showed evidence of a dose response relationship between PE and poor self-rated mental health (Figure 1), with an overall positive relationship between the two. The covariate-adjusted probability of reporting poor/fair self-rated mental health ranged from a low of 2.80% (95% CI = 2.56, 3.05) in those with PE measure of zero or one, to 5.28% (95% CI = 4.21, 6.63) in those with a precarity value of 11–13.

To the authors’ knowledge, this is the largest study of PE and mental health in the US. Our analysis represented more than 106 million US workers employed throughout the years 2008–2021, a timeframe including the 2008 Financial Crisis and emergence of COVID-19 pandemic, events which likely increased the prevalence of PE in the U.S. (Cowan, 2020; Kalleberg and von Wachter, 2017; Oddo et al., 2023a). After controlling for potential confounders using multivariable GEEs, we detected significant associations between PE, operationalized as a summative multi-dimensional exposure, and self-rated mental health (Table 3). Furthermore, marginal effects analyses were suggestive of a positive dose-response relationship between our PE measure and the covariate-adjusted probability of reporting poor/fair mental health (Figure 1).

Few recent US studies of employment quality and mental health have measured PE using a summative approach, which operationalizes precarity as a numeric exposure and can measure health outcomes at increasing levels of PE. One example from Oddo et al. (2023) surveyed more than 600 US adults and found significant positive associations between PE and self-perceived stress (Oddo et al., 2023a). Similarly, a study using the Quality of Work Life module of the General Social Survey (GSS) found that participants with high levels of precarity had increased odds of job-related stress (Bhattacharya and Ray, 2021). Outside of the US, evidence of a dose-response relationship between PE and poor mental health was observed in a sample of Spanish workers, a trend also observed in the study’s sex-stratified analysis (Vives et al., 2013). Congruence between our results and these studies, each of which used slightly different summative operationalizations of precarity, supports a causal relationship between PE and poor mental health.

In contrast to studies using a summative PE measure, others have operationalized precarity using a typological approach, in which employed persons are grouped using statistical techniques such as structural equation modeling. While typological PE operationalizations are not designed to detect detection dose-response relationships – a strength of summative approaches – they benefit in their ability to identify taxonomically distinct clusters of employment relationships (Hajat et al., 2023). One such study from Peckham et al. used latent class analysis to group GSS participants into distinct employment clusters, including a “precarious” group of workers which had a high probability of part-time work or nonpermanent contracts, low wages, and low probability of union representation (Peckham et al., 2019). This precarious group was found to have higher probability of reporting poor general and mental health when compared to a “Standard Employment Relationship” group. Another study from Andrea et al. used data from the Health and Retirement Study identified distinct employment quality trajectories (Andrea et al., 2022). One trajectory, “Poor EQ to Delayed and Poor Retirement”, was similar to Peckham’s “precarious” cluster and had the highest prevalence of depression of any group attached to (i.e., participating in) the labor force. While these studies used a typological as opposed to summative approach, their results support our finding that workers with high levels of PE have high rates of poor mental health.

Regarding causal mechanisms, research suggests that the uncertainty, unpredictability, and variability intrinsic to PE contributes to increased job-related stress and, therefore, poor mental health (Bhattacharya and Ray, 2021; Irvine and Rose, 2022). This hypothesis is congruent with findings that workers with low control over their working conditions report higher levels of psychological distress, especially when paired with high demands within the workplace (Dalgard et al., 2009). Similarly, workers with high job insecurity have been found to have increased levels of stress-related hormones, findings which add biological plausibility to this association (Herr et al., 2017).

PE may also affects workers’ mental health by interfering with their life outside of work. A 2023 advisory from the US Surgeon General, titled “Our Epidemic of Loneliness and Isolation”, noted that one’s employment situation may play a role in generating feelings of loneliness (Office of the U.S. Surgeon General, 2023). For example, multiple job holding due to financial insecurity may result in less time spent with one’s family; preoccupation with work may distract an individual from engaging with their community and forming meaningful social connections; and so forth. In fact, research outside of the US has noted associations between PE and “social precarity”, or constraints in working life, social activities, or financial difficulties (Matilla-Santander et al., 2022a). In this context, it is possible that our results are partially mediated by the social isolation produced by precarious work.

The results of this study are suggestive of a dose-response relationship between PE and poor/fair mental health in the US. These findings are concerning given evidence that employment quality and institutional protections for US workers have been decreasing for decades (Kalleberg, 2011). Despite growing evidence of PE’s negative ramifications for workers’ well-being, few studies have assessed interventions aiming to reduce precarity’s impact on workers (Gunn et al., 2022). Future research is needed to understand how best to ameliorate the negative effects of precarity on workers’ mental health, including workplace interventions and broader policy changes related to employment quality.