Trajectories of food insecurity for low wage workers from 2018 to 2022 Authors: Curtis J Antrum, Violeta Chacón, Maria F Gombi Vaca, Kim M Gans, Molly E Waring, Ran Xu, Caitlin E Caspi Published: 2026 ABSTRACT AbstractThe COVID-19 pandemic made many U.S. households susceptible to food insecurity and sparked a temporary expansion of federal food assistance. Recognizing that the effects of the pandemic on food insecurity were socially patterned and changed over time, this study aimed to identify different food insecurity trajectories from 2018 to 2022 and demographic/household factors associated with these trajectories in a low-income sample. We conducted a secondary analysis of data from a longitudinal annual survey (2018–2022) of 414 low-wage workers recruited in community settings in two U.S. cities: Raleigh, NC, and Minneapolis, MN. Annual survey measures included the 6-Item Food Security Module, self-reported demographics, and household economic factors including housing stability, employment, and receipt of Supplemental Nutrition Assistance Program (SNAP). Latent class analysis identified underlying classes (trajectories) of food security over time and assigned participants to a trajectory. Multinomial logistic regression models tested the association between demographic/household factors and trajectory membership. Food insecurity was high at baseline (72.7%). Latent class analysis yielded five trajectories from 2018 to 2022: (i) consistent high food security (23.9%), (ii) consistent moderate food security (28.3%), (iii) consistent very low food security (17.6%), (iv) improved food security (12.6%), and (v) temporarily improved food security (17.6%). Several demographic/household factors were associated with the likelihood of experiencing trajectories of less food security. Understanding trajectories of food insecurity during the COVID-19 pandemic can inform policy responses during future health or economic crises. Food security status is not a static condition; annual measurement is critical for promoting food security among low-income U.S. families.Keywords: Food insecurity, COVID-19 pandemic, Low-wage workers, Urban communities, Trajectory analysis Introduction Food insecurity is a household-level economic and social condition of limited or uncertain access to adequate food [1]. Prior to the COVID-19 pandemic, during a period of generally stable economic conditions in the U.S., rates of household food insecurity had steadily declined from 14.9% in 2011 to 10.5% in 2019 [2]. When the pandemic began in 2020, it posed extraordinary challenges to public health, the global economy, and social structures. As businesses closed and millions of Americans lost their jobs, these conditions threatened a sharp rise in food insecurity [3]. Yet, according to the annual USDA report using the U.S. Household Food Security Survey Module (HFSSM), the overall food insecurity rate in the U.S. remained unchanged from 2019 to 2020 and decreased slightly to 10.2% in 2021 [4, 5]. As early as March 2020, the U.S. government began to implement a suite of responsive measures, including offering financial support to individuals and families, expanding existing economic safety nets, and introducing new programs to address emerging food needs. For example, the 2020 Coronavirus Aid, Relief, and Economic Security (CARES) Act provided economic relief through direct stimulus payments, enhanced unemployment benefits, and business support [6]. The following year, the American Rescue Plan Act expanded the Child Tax Credit, helping reduce child poverty for millions of families [7, 8]. Additionally, the Families First Coronavirus Response Act (FFCRA) and Fiscal Year 2021 Omnibus Appropriations, made changes to the Supplemental Nutrition Assistance Program (SNAP) to ensure more Americans could access food assistance [9, 10]. SNAP is the largest U.S. federal food assistance program and addresses food insecurity by providing eligible low-income households with food benefits to supplement their grocery budget [11]. Pandemic-era program changes to SNAP included a temporary increase in the maximum SNAP benefit amount by 15% [12]. The USDA also provided states the opportunity to issue Emergency Allotments to SNAP households, increasing their benefit to the maximum amount for their household size or $95, whichever was greater. Thirty-two states issued monthly Emergency Allotments until the option expired in early 2023 [9]. There is growing evidence that the strong federal policy response mitigated an overall rise in food insecurity during this period [13–15]. Yet, despite the availability of these relief measures and the stability in national rates from 2019 to 2021, a deeper look suggests substantial social stratification in food insecurity trends for different groups. Food insecurity rates rose significantly in the early pandemic period among several key subgroups, including households with children and Black households, even while food insecurity rates decreased among other households [4]. These findings suggest that the initial economic effects of COVID-19 disproportionately impacted these groups, and that government relief measures may not have equally mitigated the economic effects for all groups. Other measures of food hardship beyond the HFSSM showed alarmingly high rates of food insecurity among Black and Hispanic households and widening disparities during the COVID-19 pandemic [16, 17]. Moreover, food insecurity is not a stable condition within a household [18]. The severity of household food insecurity is dynamic as households move in and out of eligibility for food assistance and experience other economic shocks, such as disruptions to employment status or housing stability [19, 20]. Research using longitudinal designs in higher-risk groups is needed to understand the variability in temporal patterns and levels of severity during different phases of a crisis like the COVID-19 pandemic. Recognizing that experiences of food insecurity are socially patterned and mutable over time, the current study aimed to use latent class analysis to identify trajectories of food insecurity among the households of urban, low-wage workers at a high risk of food insecurity, followed longitudinally from 2018 to 2022. Additionally, the study explored whether demographic factors and household economic factors, including housing and employment stability and participation in SNAP at different timepoints, were associated with food security trajectories. Methods Design We used data from the Wage$ study to conduct a secondary analysis [21, 22]. The Wage$ Study is a prospective cohort study of low-wage workers from two U.S. cities. This parent study collected data from 2018 to 2022, measuring health and economic outcomes among participants. The Wage$ study aimed to assess the impact of a minimum wage increase on health outcomes by comparing a city that passed an ordinance incrementally increasing minimum wage (Minneapolis, MN) and one that did not (Raleigh, North Carolina). Participants (n = 969) were recruited in community settings from February through October 2018 when they completed a survey and other study measures for which they were compensated $70. From 2019 to 2022 the same participants were followed annually and completed measures after each incremental wage increase in Minneapolis occurring on July 1 of each year. The annual survey asked questions about food insecurity, housing instability, employment, participation in food assistance programs, and demographics (race/ethnicity, age, gender, educational attainment, income, and household composition), and dietary behaviors. The timing of this survey provides an opportunity to assess longitudinal outcomes related to food insecurity during the COVID-19 pandemic. Sample The eligibility criteria for the primary study required individuals to be at least 18 years of age, earning less than or equal to $11.50 per hour, and working at least 10 h per week when they were recruited in 2018 in one of the two study cities [21]. Participants were also eligible if they were currently unemployed but had worked at $11.50 per hour or less in the previous 6 months and were looking for work in one of the two study cities. The current analysis includes the subset of participants who were followed for all years of the study (2018 to 2022) and responded to questions on food insecurity, SNAP participation, housing instability, employment, and demographics (race/ethnicity, age, gender, educational attainment, income, and household composition) at all time points. Measures Food insecurity status at each time point was measured using the Six-Item Short Form of the HFSSM [23]. This module provides scores ranging from 0 to 6. Scores from 0 to 1 are categorized as high/marginal food security, scores ranging from 2 to 4 are categorized as low food security, and scores ranging from 5 to 6 are categorized as very low food security (VLFS) [23]. Participants reported demographic characteristics. Race and ethnicity responses were condensed into the categories “Hispanic,” “Non-Hispanic White,” “Non-Hispanic Black,” and “Non-Hispanic Other.” Gender was reported as “Female,” “Male,” or “Non-binary.” Age in years was obtained as a continuous measure and for descriptive analysis was categorized as “<30”, “30–39”, “40–49”, “50–59”, and “60+”. Educational attainment categories include “Less than high school diploma,” “High school diploma,” “Some college,” “Associate/Technical degree,” and “Bachelor’s degree or higher”. Additionally, household composition categories were “No children,” “Single adult with children,” and “Multiple adults with children.” We categorized monthly household incomes ranging from “Less than $500” to “More than “$1,500” per month. At each time point, respondents provided their employment status as employed or unemployed and their receipt of SNAP in the last 30 days with “Yes”, “No”, or “Unsure” responses. Participants who responded that they were unsure (n ≤ 3 at each timepoint) were categorized as not receiving SNAP. We assessed housing stability using three questions concerning the ability to pay the mortgage, the number of places lived, and having a steady place to sleep. In each year, participants who had an instance of not being able to pay their mortgage or rent on time, lived in 3 or more locations within a year, or did not have a steady place to sleep were categorized as having unstable housing. Analysis We conducted statistical analyses using SAS version 9.4 software (SAS Institute Inc., Cary, NC, USA). A latent class analysis was performed for the reported category of food insecurity status at each time point using PROC LCA [24]. The Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) were used to determine the best model fit. These tests defined the appropriate number of latent classes in the study sample. We then used bivariate analysis to identify participant characteristics associated with latent classes. We used demographic characteristics of race/ethnicity, age, gender, city (Minneapolis vs. Raleigh), educational attainment, income, and household composition. We used chi-square tests with a significance criterion of p < 0.05 to identify statistically significant differences between classes (i.e., food security trajectories). Finally, we used multinomial logistic regression to test the association between household economic conditions (SNAP benefit receipt, employment status, and housing stability), and food security trajectory (class). For SNAP, we created three variables representing different durations of benefit receipt: consistent SNAP (responding yes to SNAP participation at all time points), ever had SNAP (responding yes at any time point), SNAP throughout pandemic representing exposure during the years in which COVID-19 expansions were provided (responding yes in all years from 2020 to 2022). For employment status, we created the variable pandemic job stability to demonstrate stable employment during the peak pandemic years (i.e., participants reported being employed in both 2020 and 2021). Similarly, we created the variable pandemic housing stability to demonstrate housing stability peak pandemic years (i.e., participants reported having housing stability in both 2020 and 2021). Adjusted logistic regression models controlled for age (continuous), education (high-school education or less vs. greater than high school), race/ethnicity (Non-Hispanic Black vs. other race/ethnicity), gender (female vs. other gender identity), and household composition (living in a household with children vs. no children in household). As a sensitivity analysis, we ran logistic regression models with participants who responded that they were unsure whether they had received SNAP categorized as having received SNAP. All logistic regression models met the criteria for convergence. Results Of the 482 participants completed the survey in all years from 2018 to 2022, we excluded participants who were missing data on race or ethnicity (n = 5), household composition (n = 9), education (n = 4), income (n = 9), gender (n = 6), receiving SNAP at any timepoint (n = 47), and/or food insecurity at any timepoint (n = 30), resulting in an analytic sample of 414 participants. Seventy-seven percent (77%) of the participants were Non-Hispanic Black, and 63% were female (Table 1). At baseline in 2018, the majority of participants had either low or very low food security status (72.7%). To identify different trajectories of food insecurity over time, we estimated models with two to seven classes to determine the appropriate number of classes. The five-class model indicated the best fit with an AIC of 281.82 and a BIC of 498.55. An entropy of 0.77 based on each individual’s probability of belonging to one class (0–1) indicated that the classes had reasonably good separation. Table 1Baseline demographic characteristics low-wage workers from Minneapolis, MN and Raleigh, NC (N = 414), 2018 N PercentRace/ethnicity Hispanic2< 1% Non-Hispanic Black31777% Non-Hispanic White7919% Non-Hispanic Other164%Age (years) 18–2910325% 30–399924% 40–496215% 50–5910425% 60+4611%Gender Female27263% Male13836% Non-Binary41%City Minneapolis, MN18244% Raleigh, NC23256%Education Less than a high school diploma5814% High school diploma15237% Some college4812% Associate/technical degree10726% Bachelor’s degree or higher4912%Monthly Income (USD) Less than $5008320% $501 to $100016941% $1001 to $150010826% More than $15005413%Household type Multiple adults with children10626% Single adult with children6616% No children24258% Baseline demographic characteristics low-wage workers from Minneapolis, MN and Raleigh, NC (N = 414), 2018 The five classes of food insecurity were descriptively labeled according to the patterns they exhibited from 2018 to 2022: (i) consistent high food security, (ii) consistent moderate food security, (iii) consistent VLFS, (iv) improved food security, and (v) temporarily improved food security (Figs. 1 and 2). In this latent class model, 23.9% of participants (n = 99) were in the class with consistent high food security, 28.3% (n = 117) were in the class with consistent moderate food security, 17.6% (n = 73) belonged to the class with consistent VLFS, 12.6% (n = 52) were in the class with improved food security, and 17.6% (n = 73) were in the class with temporarily improved food security. Five latent class trajectories (2018–2022) for (a) likelihood of high/marginal food security (b) likelihood of low food security and (c) likelihood of very low food security Percentage of respondents with high/marginal, low, or very low food security at each time point (2018–2022) by latent class trajectory (i-v) The class labeled consistently high food security (i) was characterized by a high probability of high or marginal food security (> 60%) at all five time points. The class labeled consistently moderate food security (ii) was about equally likely to demonstrate high/marginal food security or low food security at all timepoints, but was unlikely (< 10%) to have very low food security at any time. Those in the consistent VLFS class (iii) had a high probability of very low food security (> 70%) at each time point. The improved food security class (iv) began with a high probability of very low food security (99%), but over the five years, increased their likelihood of high/marginal food security (77%). The class with temporary improvements in food security (v) was very likely (> 80%) to experience food insecurity (either low or very low food security) in most years; however, in 2021 37% reported food security before it diminished again in 2022. Bivariate descriptive analyses indicated statistically significant differences between classes by age, city, and household composition (p < 0.05, Table 2). For example, the class with improved food security had the highest percentage of households composed of a single adult with children (29% compared to 15% in the consistent high food security class). Among those in the improved food security class, 81% were from Raleigh and only 19% were in Minneapolis. Table 2Participant demographics by food security (FS) trajectories from 2018 to 2022 among low wage workersClass i: Consistent high FS(N = 99)Class ii: Moderate FS(N = 117)Class iii: Consistent very low FS (N = 73)Class iv: Improved FS(N = 52)Class v: Temporary improvement in FS (N = 73) p N (%)N (%)N (%)N (%)N (%)Race/ethnicity Hispanic2 (2%)0 (0%)0 (0%)0 (0%)0 (0%)0.1455 Non-Hispanic Black68 (69%)85 (73%)60 (82%)42 (81%)62 (85%) Non-Hispanic White24 (24%)26 (22%)13 (18%)8 (15%)8 (11%) Non-Hispanic Other5 (5%)6 (5%)0 (0%)2 (4%)3 (4%)Age categorical (Years) >3030 (30%)24 (21%)11 (15%)17 (33%)21 (29%)0.0445 30–3926 (26%)31 (26%)16 (22%)11 (21%)15 (21%) 40–499 (9%)21 (18%)16 (22%)5 (10%)11 (15%) 50–5923 (23%)26 (22%)25 (34%)17 (33%)13 (18%) 60+11 (11%)15 (13%)5 (7%)2 (4%)13 (18%)Gender Female65 (63%)69 (63%)51 (70%)37 (71%)50 (68%)0.0693 Male33 (36%)48 (36%)22 (30%)15 (29%)20 (27%) Non-binary1 (1%)0 (0%)0 (0%)0 (0%)3 (4%)City Minneapolis, MN35 (35%)68 (58%)35 (48%)10 (19%)34 (47%)< 0.0001 Raleigh, NC64 (65%)49 (42%)38 (52%)42 (81%)39 (53%)Education Less Than High School9 (9%)22 (19%)10 (14%)8 (15%)9 (12%)0.0742 High School/GED34 (34%)42 (36%)23 (32%)19 (37%)34 (47%) Some College14 (14%)13 (11%)5 (7%)6 (12%)10 (14%) Associates21 (21%)31 (26%)25 (34%)13 (25%)17 (23%) Bachelors21 (21%)9 (8%)10 (14%)6 (12%)3 (4%)Monthly income (USD) Less than 50018 (18%)26 (22%)18 (25%)9 (17%)12 (16%)0.3191 501 to 100033 (33%)51 (44%)26 (36%)26 (50%)33 (45%) 1001 to 150027 (27%)27 (23%)20 (27%)12 (23%)22 (30%) More than 150021 (21%)13 (11%)9 (12%)5 (10%)6 (8%)Household type Multi w/children19 (19%)36 (31%)23 (32%)7 (13%)21 (29%)0.0290 Single w/children15 (15%)17 (15%)6 (8%)15 (29%)13 (18%) No children65 (66%)64 (55%)44 (60%)30 (58%)39 (53%) Participant demographics by food security (FS) trajectories from 2018 to 2022 among low wage workers Using the consistent high food security class (i) as a reference group, crude logistic regression demonstrated participants who received SNAP in all five years were more likely to belong to the consistent moderate food security trajectory (OR = 2.19, 95% CI: 1.16–4.10), consistent VLFS class (OR = 2.33, 95% CI: 1.17–1.73), or the temporarily improved food security class (OR = 2.19, 95% CI: 1.09–4.40), but not the class with improved food security (OR = 1.87, 95% CI: 0.86–4.05). Additionally, those who experienced job stability were less likely to be in the consistent VLFS class than in the consistent high food security class (OR = 0.48, 95% CI: 0.25–0.89). Those who had ever received SNAP or received it throughout the pandemic years were more likely to be one of the less food secure classes (ii-iv) compared with the consistently high food security class (i) (Table 3). Table 3Household economic factors by food security (FS) trajectories from 2018 to 2022 among low wage workers(i) Consistent high FS(ii) Consistent moderate FS(iv) Consistent very low FS(iii) Improved FS(v) Temporary improvement in FSConsistent SNAP (2018–2022) Yes19 (15%)40 (32%)25 (20%)16 (13%)25 (20%) No80 (28%)77 (27%)48 (17%)36 (12%)48 (17%) Crude OR (95% CI)REF2.19 (1.16–4.10)*2.33 (1.17–1.73)*1.87 (0.86–4.05)2.19 (1.09–4.40)* Adjusteda OR 95% CI)REF1.66 (0.84–3.26)1.84 (0.88–4.80)2.06 (0.88–4.80)1.61 (0.76–3.40)Ever had SNAP (2018–2022) Yes56 (17%)95 (29%)60 (18%)49 (15%)67 (20%) No43 (49%)22 (25%)13 (15%)3 (3%)6 (7%) Crude OR (95% CI)REF3.32 (1.80–6.11)*3.54 (1.73–7.28)*12.53 (3.66–42.94)*8.57 (3.40–21.62)* Adjusted OR 95% CI)REF2.69 (1.36–5.32)*3.20 (1.46–7.03)*18.16 (4.95–66.63)*7.68 (2.87–20.54)*SNAP throughout pandemic (2020–2022) Yes23 (13%)53 (30%)34 (19%)25 (14%)40 (24%) No76 (32%)64 (27%)39 (16%)27 (11%)43 (14%) Crude OR (95% CI)REF2.73 (1.51–4.95)*2.88 (1.50–5.55)*3.06 (1.49–6.26)*4.01 (2.08–7.72)* Adjusted OR 95% CI)REF2.25 (1.20–4.19)*2.50 (1.26–4.98)*3.51 (1.63–7.58)*3.41 (1.71–6.79)*Pandemic job stability (2020,2021) Yes70 (27%)70 (27%)39 (15%)35 (14%)42 (16%) No29 (18%)47 (30%)34 (22%)17 (11%)31 (20%) Crude OR (95% CI)REF0.62 (0.35–1.09)0.48 (0.25–0.89)*0.85 (0.41–1.76)0.56 (0.30–1.06) Adjusted OR 95% CI)REF0.82 (0.45–1.52)0.52 (0.27–1.03)0.70 (0.33–1.51)0.67 (0.34–1.33)Pandemic housing stability (2020, 2021) Yes62 (49%)34 (27%)4 (3%)8 (6%)18 (14%) No37 (13%)83 (29%)69 (24%)44 (15%)55 (19%) Crude OR (95% CI)REF0.24 (0.14–0.43)*0.04 (0.01–0.10)*0.11 (0.05–0.26)*0.20 (0.10–0.38)* Adjusted OR 95% CI)REF0.22 (0.12–0.41)*0.03 (0.01–0.09)*0.12 (0.05–0.28)*0.20 (0.10–0.41)*aOdds ratios (OR) adjusted for age, education (high school or less vs. greater than high school), city, gender (female vs. not), race/ethnicity (Non-Hispanic Black vs. not), household composition (living in a household with children vs. not)*p < 0.05 Household economic factors by food security (FS) trajectories from 2018 to 2022 among low wage workers aOdds ratios (OR) adjusted for age, education (high school or less vs. greater than high school), city, gender (female vs. not), race/ethnicity (Non-Hispanic Black vs. not), household composition (living in a household with children vs. not) *p < 0.05 After adjustment for age, education, race/ethnicity, gender, city, and living in a household with children, participants were more likely to be in one of the less food secure classes (ii-iv) than in the consistent high food security class (i) if they had ever received SNAP, received SNAP throughout the pandemic, and experienced housing instability (Table 3). For example, those who received SNAP throughout the pandemic were more than twice as likely to be in the consistent VLFS class than the consistent high food security class (2.50 (1.26 - 4.98), aOR = 2.61, 95% CI: 1.32–5.14). Additionally, those with pandemic housing stability were less likely to be in the class with consistent VLFS compared with the class with consistent high food security (aOR = 0.03, 95% CI: 0.01–0.09). Participants who received SNAP every year or experienced pandemic job stability were no more likely to be in one of the less food secure classes (ii-iv) than in the consistently high food security class (i) in adjusted models. Results in did not change in the sensitivity analysis in which participants who were unsure whether they had received SNAP were classified as having received it (results not shown). Discussion We observed five trajectories of household food insecurity from 2018 to 2022 for adult low-wage workers: consistent high food security, consistent moderate food security, consistent very low food security, improved food security, and temporarily improved food security. City of residence, age, and household composition were associated with food insecurity trajectories. Additionally, we found statistically significant differences in food insecurity trajectory by pandemic household economic factors including SNAP participation and housing stability. The study demonstrated that, despite all participants being low-wage workers, experiences of food insecurity varied substantially during this period of economic and social upheaval. Latent class analysis has been used in several other studies to identify patterns of food insecurity over time, but rarely during a crisis like the pandemic period [25–27]. One study used a nationally representative sample of U.S. residents from April 2020 to March 2021 to identify food insecurity trajectories and found ten latent classes, including classes that showed improved food security, consistent low food security, and consistent high food security [26]. However, that study did not distinguish between low and very low food security status. It also was not limited to a sample at high risk for food insecurity and the study duration was only 11 months, versus five years in the current study. The current study also examined whether demographic and household factors, including the receipt of pandemic-era relief measures, were associated with food insecurity trajectories. Together, these studies support the notion that the effects of the COVID-19 pandemic on food insecurity outcomes demonstrated different trajectories and were socially patterned [5, 28, 29]. Many of the associations between demographic groups and food insecurity outcomes observed in the current study are not unexpected. For instance, food insecurity trajectories were associated with sociodemographic characteristics like age and household composition. Previous studies have identified differences by age in food insecurity, where older participants experienced more food insecurity [30–32]. It is also not surprising that participants in the two cities had a different distribution of trajectories. Federal COVID-19 economic policies were largely similar across the two locations; both Minnesota and North Carolina opted in to the USDA SNAP Emergency Allotments and continued them through early 2023. Both states also had nearly identical periods during which a stay-at-home order was in effect [33]. However, there were contextual differences in the conditions for low-income communities across states [34]. For instance, SNAP eligibility was more restrictive in North Carolina than Minnesota prior to the pandemic, so that pandemic-era SNAP expansions and flexibilities may have disproportionately benefited Raleigh participants [35, 36]. These factors may partially explain why the vast majority of participants with an improved food security trajectory were from Raleigh rather than Minneapolis. We found that the duration and timing of SNAP benefit receipt was associated with low-wage households’ experiences of food insecurity from 2018 to 2022. Low-wage workers who had ever received SNAP or received SNAP throughout the pandemic when benefit amounts were highest were more likely to belong to the improved food security trajectory relative to the consistently food secure trajectory. For those in the improved trajectory, SNAP receipt may have played a stabilizing or buffering role for some, elevating them to a more food secure status. At the same time, SNAP participation was also associated with belonging to the consistently very low food security trajectory relative to the consistently food secure trajectory. For those participants in the very low food security group, who faced the most severe or persistent forms of food insecurity, their food needs remained high despite the receipt of SNAP. Our findings are likely due to issues of endogeneity, or the fact that individuals who sought assistance and were eligible for it may have been systematically worse off than those who did not. In this case, both those who improved and those with persistent very low food security were at a high need, and systematically more likely to receive SNAP, than those in the consistent high food security group. Because our study did not compare food insecurity trajectories between SNAP recipients and SNAP-eligible low-wage workers whose families did not receive this nutrition assistance, we cannot make causal claims about SNAP’s effect on food security trajectories. Instead, the results likely reflect both the buffering role of SNAP, the underlying characteristics of households that initiate and complete the procedures and paperwork needed to receive these benefits, and also the selection and retention of low-wage workers into the Wage$ cohort study. Other literature on the effects of SNAP expansions and flexibilities during the pandemic supports the interpretation that increased benefits were helpful for some during the pandemic, but did not curb food insecurity among others [36–39]. Narrative accounts from participants in qualitative studies suggest that expansions and flexibility in SNAP, along with other pandemic-era supports, supported economic stability among low-income groups, and also that the de-implementation of these benefit expansions posed a challenge to program recipients [35]. Another study noted that while SNAP Emergency Allotments increased the benefit amounts for participating households, enrollment expansions were limited, excluding many households with food needs [39]. Notably, recent federal legislation will further reduce eligibility, benefits, and program administration support for SNAP in future years [40], which is likely to have a detrimental effect on food security in the U.S. However, measuring the impact of SNAP program changes on food insecurity rates in future years will be challenging, as the USDA recently announced the discontinuation of national food insecurity surveillance using the HFSSM [41]. In our sample of low-income households, we identified a food insecurity trajectory that generally had low food security but showed temporary improvements that peaked in 2021. In 2021, not only were Emergency Allotments in place, but additional programs, such as Economic Impact Payments (“stimulus checks”) and expansion of the Child Tax Credit, were concurrently available to many families during this period [42]. One study of a sample of low-income U.S. households found that SNAP expansion did not improve outcomes for those most vulnerable, but direct cash assistance programs like stimulus payments was associated with decreasing food insecurity [29]. While it is not possible to disentangle the impact of these different supports, results support the notion that, especially for households with the most severe and persistent food needs, food insecurity was lowest when a combination of economic relief measures were in place. Based on data from both the study sample and national data, food insecurity began to rise in 2022 once the expansion of these programs ended, suggesting that an extension of these policies would be beneficial to the populations who relied upon them [43]. Results of the current study also add to the literature suggesting that housing stability and food security are closely linked [20, 44, 45]. Participants who reported housing stability during the pandemic were less likely to be in lower food security classes, suggesting a protective effect of housing stability on food security. The eviction moratorium that was in place during the first year of the pandemic and emergency rental assistance programs that were subsequently available may have temporarily increased the money participants had available for food [35]. Housing stability may have also directly mitigated food insecurity by providing a consistent place for storing and preparing food. This study has several limitations. First, our sample was comprised of low-wage workers, the majority of whom were racial/ethnic minorities living in urban areas. Therefore, latent classes likely would have looked different in a nationally representative sample. Additionally, this study is limited in its ability to draw causal conclusions on the predictors of these trajectories. Potential endogeneity issues with SNAP participation complicate the causal interpretation. The presence of other economic relief measures that were implemented nationally at the same time also complicates the interpretation of SNAP effects. Moreover, all variables, including receipt of SNAP benefits, were self-reported. We relied upon participants to report these items accurately, but some misreporting is possible. In other surveys using self-reported data, SNAP receipt was > 90% accurate [46]. Finally, while this study examined some demographic and policy factors related to food insecurity trajectories, it did not examine other factors, including behavioral adaptations, that households may have made during the pandemic period. Households may, for instance, have relied on assistance from relatives or other households, or relied on food acquired from food pantries or home food production; such coping mechanisms have been reported in other studies [47, 48], but are not fully explored in the current study. Conclusion This study identifies food insecurity trajectories in a high-risk sample that can inform policy responses during the next health or economic crisis. Patterns of food security from 2018 to 2022 were socially stratified, even within this high-risk group. Periods of strong federal economic relief measures corresponded with improvements in food security among these low-income households. The study demonstrates that food insecurity status is not a static condition. It is urgent that regular surveillance of food insecurity is reinstated, especially during periods when food assistance programs are being scaled back. Combined with other economic strains on households, such as the expiration of marketplace healthcare subsidies [49] and the continued rise in food costs [50], households at risk require careful monitoring and sustained food assistance support to mitigate worsening food insecurity. Acknowledgements Not applicable. Author contributions CA: Conceptualization, Formal analysis, Visualization, Writing-original; VC: Data Curation, Writing-review & editing; MGV: Conceptualization, Data Curation, Writing-review & editing; KG: Conceptualization, Supervision, Writing-review & editing; MW: Conceptualization, Methodology, Supervision, Writing-review & editing; RX: Conceptualization, Methodology, Supervision, Writing-review & editing; CC: Conceptualization, Methodology, Supervision, Funding acquisition, Visualization, Writing-original. Funding This research project was supported by the National Institute of Diabetes and Digestive and Kidney Diseases of the National Institutes of Health (R01DK118664); the National Center for Advancing Translation Sciences (UL1TR002494) supported data management; the National Institute of Food and Agriculture (7002638) supported research personnel. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health or National Institute of Food and Agriculture. Data availability The study codebook is publicly available at [https://doi.org/10.6084/m9.figshare.28360268.v3] (https:/doi.org/10.6084/m9.figshare.28360268.v3) and study instruments are publicly available at [https://doi.org/10.6084/m9.figshare.28282910.v1] (https:/doi.org/10.6084/m9.figshare.28282910.v1) . The informed consent language precludes publicly sharing individual research records. Please contact the corresponding author with data inquiries. Declarations Ethics approval and consent to participate This study was conducted according to the guidelines outlined in the Declaration of Helsinki and all procedures involving research study participants were approved by the Institutional Review Boards at the University of Minnesota, University of North Carolina, and University of Connecticut. Consent for publication Written informed consent to publish findings without individually identifiable data was obtained from all subjects. Informed consent Written informed consent to participate was obtained from all subjects. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement The study codebook is publicly available at [https://doi.org/10.6084/m9.figshare.28360268.v3] (https:/doi.org/10.6084/m9.figshare.28360268.v3) and study instruments are publicly available at [https://doi.org/10.6084/m9.figshare.28282910.v1] (https:/doi.org/10.6084/m9.figshare.28282910.v1) . The informed consent language precludes publicly sharing individual research records. Please contact the corresponding author with data inquiries.