Skip to main content
PMC home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2022 Sep 1.
Published in final edited form as: Prev Med. 2021 Jun 5;150:106678. doi: 10.1016/j.ypmed.2021.106678

Supplemental Nutrition Assistance Program 2009 Expansion and Cardiometabolic Markers Among Low-Income Adults

Laura J Samuel 1, Sarah L Szanton 1,2, Jennifer L Wolff 2, Darrell J Gaskin 2
PMCID: PMC8316430  NIHMSID: NIHMS1716894  PMID: 34097951

Abstract

A 2009 Supplemental Nutrition Assistance Program (SNAP) policy change that expanded eligibility and increased benefit amounts has been associated with reduced food insecurity. This study tests the hypothesis that the SNAP policy change corresponds with improved stress- and nutrition-sensitive cardiometabolic markers. This study included non-pregnant participants aged 18–59 with annual family incomes ≤185% of the federal poverty guideline from the repeated cross-sectional NHANES study. Those living in SNAP eligible households (income ≤130% of the poverty guideline) were compared to those who were likely non-eligible (income 131%-≤185%). Difference-in-differences analyses compared hemoglobin A1c (%), CRP (mg/dL), total cholesterol (mg/dL), LDL (mg/dL) and waist circumference (cm) across groups before (2007–2008) and after (2009–2010) the SNAP policy change. Sampling weights were applied. Adjusting for demographic, socioeconomic, household and health factors, there were statistically significant difference-in-differences estimates for hemoglobin A1c (p= 0.003, n=3,723) and total cholesterol (p=0.028, n=3,710). SNAP eligible adults had no difference in hemoglobin A1c after the policy change and, among those less than 40 years of age, 5 mg/dL lower total cholesterol levels whereas likely non-SNAP eligible adults had 0.14% higher hemoglobin A1c and no difference in total cholesterol after the policy change. The 2009 SNAP expansion was associated with improved nutrition-sensitive cardiometabolic markers in SNAP-eligible adults. This study found less of an upward trend in hemoglobin A1c levels for young and middle aged adults and decreased total cholesterol for young adults. These results highlight the potential role of SNAP to prevent costly chronic conditions among low-income U.S. adults.

Keywords: socioeconomic factors, food assistance, Glycated Hemoglobin A, cholesterol, C-reactive protein

Introduction

Despite good evidence that low-income adults have poorer health than their wealthier peers, we lack interventional evidence about whether increasing income for food could improve health (Glymour et al., 2014). The American Recovery and Reinvestment Act of 2009 temporarily increased Supplemental Nutrition Assistance Program (SNAP) maximum monthly benefits by 13.6% ($80 for a family of four) (McGuire, 2013) and waived the typical three-month enrollment limit that was in place for childless adults aged 18–49 (Nord and Prell, 2011). The policy change was associated with a 3% increase in SNAP participation between 2008 and 2009 (Nord and Prell, 2011), which provides an opportunity to use a natural experiment to evaluate the association of this increase in income for food with changes in health risk markers.

The 2009 SNAP policy change has been associated with better health outcomes among likely-eligible low income adults. Although the policy change was not associated with improved dietary quality (McClain et al., 2019; Waehrer et al., 2015), it was associated with reduced food insecurity (McClain et al., 2019; Nord and Prell, 2011), and increased food spending (Kim, 2016). Notably, the increased benefit amounts may have also allowed families to better afford other basic necessities related to health, like housing, transportation, health care and medications (Kim, 2016; Morrissey and Miller, 2020). One Massachusetts study also found slower Medicaid inpatient hospital spending growth after the policy change (Sonik, 2016). However, whether and how the SNAP policy change temporally relates to shifts in cardiometabolic measures has not been examined but is of clinical significance due to disproportionate cardiometabolic disease burden among lower income adults (Gruenewald et al., 2012; Park et al., 2003).

This study tests the hypothesis that the 2009 SNAP policy change corresponded with improved stress- and nutrition-sensitive cardiometabolic markers, including hemoglobin A1c, C-reactive protein (CRP), total cholesterol, low-density lipoprotein (LDL) and waist circumference, in low-income young and middle-aged adults eligble for these expanded benefits. Although the SNAP policy change could only improve health through household program participation, this study focused on SNAP income eligibility rather than participation due to under-reporting of SNAP participation in surveys (Meyer et al., 2018). Secondly, because the policy change included both expanded eligibility and larger benefit amounts, this study examines changes in self-reported SNAP participation and benefit amounts to elucidate mechanisms that may explain hypothesized associations. These results have timely policy implications because 2020 and 2021 federal legislation temporarily re-instituted similar changes but the health impact of these policy changes is not well tested. Such knowledge could be used to inform future policy development.

Methods

Study Sample

The National Health and Nutrition Examination Survey (NHANES) provides a unique opportunity to evaluate the 2009 policy change. NHANES is a repeated semi-annual cross-sectional survey which conducts in-home interviews and in-clinic examinations among a nationally representative sample of approximately 10,000 noninstitutionalized individuals in the U.S. As described elsewhere (Services, 2013), NHANES uses complex, multistage probability sampling with oversampling for persons who are Hispanic, black, ≤130% of the federal poverty level and over age 79. This study uses data from the 2007–2008 and 2009–2010 NHANES examinations. Interview questions, laboratory assays and facilities were consistent across both examinations. This study included 4,240 NHANES participants aged 18–59 with annual family incomes ≤185% of the federal poverty guideline and excluded 77 of those because of pregnancy. No study participants were taking oral corticosteroids, antibiotics or anti-neoplastic medications; thus none were excluded due to medication use. One individual with a CRP level >10 μg/ml who had a high likelihood of active infection (Vanderschueren et al., 2006) was excluded from CRP analyses. These analyses were deemed exempt by the Johns Hopkins Medical Institutions IRB.

Measures

Main outcomes, including hemoglobin A1c (%), CRP (mg/dL), total cholesterol (mg/dL), LDL (mg/dL) and waist circumference (cm), were selected because they predict mortality risk (Anderson et al., 1987; Elkind et al., 2009; Staiano et al., 2012; Stamler et al., 1993) and should decrease within six months in response to improvements in food access and stress, except for waist circumference which likely will only stop increasing over time (Zheng et al., 2013). Phlebotomy was performed and waist circumference was measured during the in-clinic examination. CRP was quantified using latex-enhanced nephelometry and analyses were conducted at the University of Washington. Hemoglobin A1c and lipids were analyzed at the University of Minnesota. LDL was calculated (Friedewald et al., 1972) for adults with morning examinations, who had provided fasting blood specimens. There is <10% missing data for each outcome.

Potentially explanatory factors included SNAP participation and SNAP benefit amount. SNAP participation was classified based on self-report receipt of SNAP by the participant or household member in the past 12 months. The SNAP benefit amount for the last month received was recorded among participants who reported ever receiving household SNAP benefits in their lifetime. Although SNAP participation and benefit amounts may be under-reported, the measurement error likely does not differ across groups or examination years (Meyer et al., 2018).

Based on prior work (Kim, 2016) and federal SNAP income eligibility guidelines, individuals were classified as SNAP eligible if family income was ≤130% poverty guidelines and likely not eligible if income was 131%-≤185% poverty. Although some states allow households with incomes exceeding 130% to participate in SNAP, the policy change should have a small effect in this group because only about 6% of SNAP participants have incomes above this threshold (Service, 2019).. Family income was used to calculate income to poverty ratio using the Department of Health and Human Services Poverty Guidelines. For 11.7% and 12.7% of 2007–2008 and 2009–2010 participants, respectively, who did not provide exact income amounts, follow up questions asked whether the income was <185% or ≤130% of the poverty guideline, resulting in limited missing data (<5%). Examination year was used to create a binary variable to differentiate observations from after (2009–2010) and before the policy change (2007–2008).

This study measured potentially confounding demographic, socioeconomic, household and health characteristics. Demographic characteristics included age, sex, and self-reported race/ethnicity (non-Hispanic white (ref.), non-Hispanic black, Hispanic, other race/multi-racial). Household characteristics included total number of people in the household and whether the participant owned his/her home. Socioeconomic characteristics included whether the participant was unemployed, level of education (<high school (ref.), high school/GED, some college, ≥college) and household food insecurity (none (ref.), low, moderate, high) (Bickel et al., 2000). Health characteristics included waist circumference (in cm), diagnosis of diabetes or high cholesterol and whether the participant reported taking prescription(s) to lower blood sugar or cholesterol.

Statistical analysis

The difference-in-differences approach (Wooldridge and Imbens, 2007) was used to test hypotheses by comparing differences in outcomes over time among adults eligible for new or expanded SNAP benefits (i.e. ≤130% poverty) to differences over time among other low-income adults who were likely not eligible for SNAP (i.e. 131–185% poverty). Since this approach assumes that the composition of the groups remains stable over time (Wing et al., 2018), analyses compared samples across examination years to see if the groups differed over time. The difference-in-difference estimate is obtained from a regression model, which includes binary variables classifying SNAP eligibility and examination year and an interaction term between them (Wooldridge and Imbens, 2007). The difference-in-difference estimate is the SNAP eligibility X examination year interaction term which represents the difference in the outcome that is attributed to the SNAP policy change and, therefore, estimates the policy change treatment effect. To aid interpretation, the ‘lincom’ command in Stata 14.1 (2015) was used to compute the linear combination of the examination year coefficient and interaction term coefficient as an estimate of the differences over time among those who were SNAP eligible. Linear regression was used for hemoglobin A1c, total cholesterol, LDL and waist circumference. Logistic regression was used for SNAP participation. Negative binomial regression was used for CRP, and for SNAP benefit amount among those who ever received benefits. Sample weights accounting for both the complex sampling design and nonresponse were applied so results can be generalized to the population of civilian noninstitutionalized U.S. adults.

All regression models adjusted for age, sex, race/ethnicity, household size, unemployment, and education. Models for hemoglobin A1c, CRP, total cholesterol and LDL additionally adjusted for waist circumference. The hemoglobin A1c model additionally adjusted for diabetes diagnosis and diabetic medication(s). The total cholesterol and LDL models additionally adjusted for high cholesterol diagnosis and cholesterol medication(s). Because associations between SNAP participation and weight-related outcomes have previously differed by gender (Fan, 2010; Jilcott et al., 2011), and because of potential age-related differences in diet and metabolism, analyses tested whether hypothesized relationships differed by age and sex for all main outcomes using three-way interactions with the difference-in-difference estimate (i.e. SNAP X examination year). If statistically significant interactions were found, interaction terms were retained and models were stratified. Since food insecurity likely predicts SNAP utilization (Nord and Golla, 2009) and may differ across examination years, models for SNAP participation and SNAP benefit amount adjusted for food insecurity and its interaction with the difference-in-differences estimate.

Four planned sensitivity analyses were conducted. First, falsification tests (Shadish et al., 2002) evaluated the hypotheses that there were no associations between the policy change and either serum cotinine levels or blood lead (i.e. no statistically significant difference-in-differences). Cotinine and lead were selected as control outcomes because they should be higher among lower income groups but unaffected by the SNAP policy change and therefore should not differ over time. Second, because the SNAP policy change was enacted on April 1, 2009 participants examined between January and March 2009 are misclassified as exposed in our study. Therefore, a sensitivity analysis excluded 2009–2010 participants who were examined before April 30th in either year to address misclassification bias. Third, hypothesized associations were tested using data that was multiply imputed by chained equations with five replications to address missing data whenever imputation was supported by the data for all measured variables. Finally, since unemployment benefits from the American Recovery and Reinvestment Act of 2009 may have affected health outcomes, analyses excluded unemployed individuals.

Results

The sample was, on average 37 years of age (SE=0.36), predominately female (54%) and white (53%), with 17% of the sample black, 24% Hispanic and 6% other racial/ethnic groups. Approximately 68% (n=3,037) of the sample resided in households that were eligible for SNAP, which translates to an estimated 42,544,049 adults in the U.S. population. The remainder of the sample (32%, n=1,126) were just above the income threshold for SNAP eligibility, which translates to about 20,122,687 U.S. adults (Table 1). Compared with non-eligible adults, adults in SNAP-eligible households were younger, more likely to be either black or Hispanic, more likely to be unemployed, less likely to own a home or have higher than a high school level of education, more likely to have food insecurity, and had larger average household sizes. They did not differ with regard to waist circumference, diabetes or cholesterol diagnoses and medication use. Sample characteristics did not differ across examination years for either group, except that food insecurity worsened for SNAP eligible, but not non-eligible, adults.

Table 1.

Selected sample characteristics, NHANES participants aged 18–59 years, stratified based on SNAP eligibility and examination years (2007–2008 and 2009–2010)

Examination year Variable Likely not SNAP eligiblea (n=1,126, 32%) SNAP eligiblea (n=3,037, 68%) p value
2007–2008 Mean age (SE) 37.71 (0.64) 35.86 (0.80) 0.034
2009–2010 38.24 (0.42) 36.73 (0.36) 0.006
p value 0.494 0.330
Sex (%)
2007–2008  Female (ref.) 256 (51) 748 (57) 0.062
 Male 261 (49) 619 (43)
2009–2010  Female (ref.) 299 (52) 911 (54) 0.166
 Male 310 (48) 759 (46)
p value 0.984 0.185
Race/ethnicity (%)
2007–2008  Non-Hispanic white 186 (60) 493 (52) (ref.)
 Non-Hispanic black 111 (14) 288 (16) 0.033
 Hispanic 201 (22) 529 (25) 0.138
 Other/Multi-racial 19 (5) 57 (6) 0.376
2009–2010  Non-Hispanic white 248 (59) 628 (47) (ref.)
 Non-Hispanic black 115 (15) 347 (20) 0.077
p value 0.871 0.328
 Hispanic 209 (18) 597 (27) 0.009
p value 0.674 0.697
 Other/Multi-racial 37 (8) 98 (7) 0.717
p value 0.429 0.479
2007–2008 Mean household size (SE) 3.49 (0.16) 3.77 (0.13) 0.166
2009–2010 3.45 (0.10) 3.82 (0.10) 0.002
p value 0.800 0.733
Unemployed (%)
2007–2008  No (ref.) 377 (73) 732 (54) <0.001
 Yes 140 (27) 635 (46)
2009–2010  No (ref.) 410 (68) 849 (53) <0.001
 Yes 199 (32) 821 (47)
p value 0.097 0.666
Own home (%)
2007–2008  No (ref.) 238 (41) 871 (61) <0.001
 Yes 279 (59) 496 (39)
2009–2010  No (ref.) 321 (50) 1,091 (61) 0.025
 Yes 288 (50) 579 (39)
p value 0.194 0.931
Education (%)
2007–2008  <high school (ref.) 156 (25) 618 (38) 0.003
 High school/GED 146 (28) 335 (27)
 Some college 145 (33) 315 (26)
 ≥college 65 (14) 81 (9)
2009–2010  <high school (ref.) 195 (24) 623 (34) <0.001
 High school/GED 146 (24) 444 (27)
 Some college 179 (32) 439 (29)
 ≥college 83 (20) 137 (10)
p value 0.206 0.303
Household food insecurity (%)
2007–2008 None (ref.) 322 (67) 632 (53) 0.012
Low 85 (16) 231 (14)
Moderate 77 (12) 332 (22)
High 33 (6) 171 (11)
2009–2010 None (ref.) 350 (66) 697 (47) <0.001
Low 96 (13) 277 (16)
Moderate 105 (13) 415 (22)
High 58 (8) 280 (15)
p value 0.636 0.042
2007–2008 Mean waist circumference (SE) 97.15 (0.92) 96.05 (1.02) 0.297
2009–2010 94.85 (1.05) 97.06 (0.83) 0.056
p value 0.108 0.448
Diabetes diagnosis (%)
2007–2008  No (ref.) 466 (91) 1,257 (93) 0.296
 Yes 50 (9) 110 (7)
2009–2010  No (ref.) 557 (95) 1,543 (93) 0.222
 Yes 51 (5) 127 (7)
p value 0.097 0.709
Diabetes medication (%)
2007–2008  No (ref.) 478 (93) 1,279 (95) 0.495
 Yes 39 (7) 88 (5)
2009–2010  No (ref.) 570 (96) 1,578 (95) 0.382
 Yes 39 (4) 92 (5)
p value 0.131 0.647
High cholesterol diagnosis (%)
2007–2008  No (ref.) 423 (82) 1,146 (84) 0.387
 Yes 94 (18) 221 (16)
2009–2010  No (ref.) 498 (82) 1,381 (83) 0.691
 Yes 111 (18) 289 (17)
p value 0.938 0.521
Cholesterol medication (%)
2007–2008  No (ref.) 466 (91) 1,233 (90) 0.854
 Yes 51 (9) 134 (10)
2009–2010  No (ref.) 545 (90) 1,504 (90) 0.846
 Yes 64 (10) 166 (10)
p value 0.722 0.979
Open in a new tab

Note: Sample weights were used to estimate all values except for frequency counts. P values were obtained from bivariate logistic regression using either SNAP eligibility or study year as the dependent variable.

a

SNAP eligibility was classified based on income ≤130% poverty guideline vs. 131% to <185%.

NHANES = National Health and Nutrtion Examination Study; SNAP = Supplemental Nutrition Assistance Program

In unadjusted analyses, adults in SNAP-eligible households had higher CRP than likely non-eligible adults during the 2009–2010 examination. Among females, SNAP eligible adults had larger waist circumference and among adults less than 40 years old, SNAP eligible adults had lower total cholesterol than likely non-eligible individuals during the 2009–2010 examination (Table 2). SNAP eligible adults also had a higher prevalence of SNAP participation (39% and 45% vs. 11% and 14%) during both examinations and, among those who ever received SNAP benefits, received larger monthly amounts ($316 vs. $191) than non-eligible adults during the 2009–2010 examination. Outcomes did not change over time except that total cholesterol decreased among SNAP eligible adults less than 40 years old and SNAP benefits increased by about $97 among SNAP eligible adults who ever received benefits.

Table 2.

Unadjusted stress- and nutrition-sensitive cardiometabolic outcomes, NHANES non-pregnant participants aged 18–59 years with income <185% poverty (n=4,163)

Likely not SNAP eligible (131% to <185% poverty) (n= 1,126, 32%) SNAP eligible (≤130% poverty) (n=3,037, 68%) p value
Mean hemoglobin A1c, in %, (SE) (n=3,912) 2007–2008 5.48 (0.06) 5.53 (0.03) 0.405
2009–2010 5.57 (0.05) 5.55 (0.03) 0.738
p value 0.307 0.645
Mean LDL cholesterol, in mg/dL, (SE) (n=1,830) 2007–2008 116.06 (3.20) 114.35 (1.94) 0.619
2009–2010 120.10 (3.35) 117.47 (1.79) 0.480
p value 0.370 0.292
Mean total cholesterol, in mg/dL, (SE)
 Among those 18–39 years (n=2,213) 2007–2008 186.59 (2.84) 186.19 (1.55) 0.923
2009–2010 188.40 (3.85) 180.73 (1.69) 0.042*
p value 0.700 0.024*
 Among those 40–59 years (n=1,681) 2007–2008 208.48 (4.72) 205.20 (2.15) 0.387
2009–2010 207.12 (4.83) 206.93 (1.45) 0.968
p value 0.841 0.504
Mean waist circumference, in cm, (SE)
 Among males (n=1,878) 2007–2008 99.02 (1.47) 97.27 (0.85) 0.260
2009–2010 97.39 (1.63) 96.96 (0.89) 0.807
p value 0.454 0.800
 Among females (n=2,099) 2007–2008 95.31 (1.23) 95.07 (1.38) 0.901
2009–2010 92.43 (1.38) 97.16 (1.14) 0.012*
p value 0.137 0.255
Mean CRP, in mg/dL, (SE) (n=3,900) 2007–2008 0.35 (0.03) 0.41 (0.03) 0.276
2009–2010 0.31 (0.02) 0.41 (0.02) 0.043*
p value 0.274 0.951
Intermediary outcomes
SNAP participation, (%) (n=4,159) 2007–2008 No 448 (89) 788 (61) <0.001*
Yes 69 (11) 578 (39)
2009–2010 No 491 (86) 883 (55) <0.001*
Yes 118 (14) 784 (45)
p value 0.292 0.287
Mean recent SNAP benefit amount among those who ever received, in dollars, (SE) (n=1,852) 2007–2008 146.33 (23.15) 218.80 (9.75) 0.050
2009–2010 190.65 (30.07) 315.97 (15.59) 0.014*
p value 0.293 <0.001*
Open in a new tab

Note: P values were obtained from logistic regression using either SNAP eligibility or study year as the dependent variable. Models were stratified by age or sex if statistically significant (p<0.05) interaction terms were found. Sample weights accounting for study design and nonresponse were applied so results can be generalized to the population of civilian noninstitutionalized U.S. adults. Sample sizes differ across outcomes because data was not obtained for all participants based on NHANES protocol. There is <10% missing data for all outcomes.

NHANES = National Health and Nutrtion Examination Study; SNAP = Supplemental Nutrition Assistance Program

Adjusting for demographic, socioeconomic, household and health factors, there were statistically significant difference-in-differences estimates for hemoglobin A1c (p= 0.003), total cholesterol (p=0.028), and waist circumference (p=0.004) (Table 3). Adults in SNAP eligible households did not have higher hemoglobin A1c values after the policy change, but likely non-eligible adults had 0.14% higher hemoglobin A1c levels after the policy change. Policy change associations differed based on age for total cholesterol (p value for interaction term =0.013) but not for other outcomes (p >0.05 for all interaction terms). Adults in SNAP eligible households had 14 mg/dL lower total cholesterol levels after the policy change for 18 year old adults and the association attenuated by 0.33 mg/dL for each additional year of age, but likely non-eligible adults had no change in total cholesterol. In age-stratified analyses, total cholesterol levels decreased an average of 5 mg/dL after the policy change among SNAP eligible adults less than 40, but did not decrease among those older than 40 years. Policy change associations differed by sex for waist circumference (p value for interaction term = 0.020) but not for other outcomes (p >0.05 for all interaction terms). Despite a statistically significant difference-in-difference estimate, there were statistically significant changes in waist circumference after the policy change in neither the SNAP eligible nor the likely non-eligible group. The policy change was not associated with differences in either CRP or LDL cholesterol. Although food insecurity worsened over time in the SNAP eligible group, there were no statististically significant interactions between food insecurity and the difference-in-differences estimates for main outcomes.

Table 3.

Estimated associations between the 2009 SNAP policy change and stress- and nutrition-sensitive cardiometabolic markers among NHANES non-pregnant participants aged 18–59 years with income <185% poverty

Baseline difference between SNAP eligible and likely not eligible Difference after policy change Difference-in-differences
Among those likely not SNAP eligible (131%–≤185% poverty) Among those SNAP eligible (≤130% poverty) Estimate p value Interaction with D-in-D estimate
Hemoglobin A1c, in %, b (95% CI)a (n=3,723) 0.07 (0.02, 0.13)* 0.14 (0.05, 0.23)* 0.01 (−0.06, 0.07) −0.14 (−0.22, −0.05)* <0.01* None
Total cholesterol, in mg/dL, b (95% CI)bc (n=3,710) 1.17 (−4.10, 6.45) 0.87 (−6.21, 7.94) −13.99 (−24.08, −3.90)* −14.85 (−28.00, −1.71)* 0.03* Interaction with age: b= 0.33, p=0.01*
 Among those 18–39 years (n= 2,088) 2.72 (−4.19, 9.63) 2.80 (−5.45, 11.04) −5.19 (−10.20, −0.17) −7.99 (−17.98, 2.01) 0.11 N/A
 Among those 40–59 years (n=1,622) −0.16 (−8.06, 7.73) −0.49 (−12.99, 12.00) 2.38 (−3.59, 8.34) 2.87 (−9.65, 15.39) 0.64 N/A
LDL cholesterol, in mg/dL, b (95% CI)bd (n=1,746) −0.89 (−7.06, 5.28) 3.86 (−5.90, 13.62) 2.63 (−2.69, 7.96) −1.23 (−11.04, 8.59) 0.77 None
Waist circumference, in cm, b (95% CI)f (n=3,926) −0.78 (−2.77, 1.22) −2.10 (−4.58, 0.37) 2.49 (−0.24, 5.22) 4.59 (1.59, 7.60)* <0.01* Interaction with sex: b=−3.53, p=0.02*
 Among males (n=1,851) −1.24 (−4.53, 2.04) −1.31 (−5.11, 2.49) −0.28 (−2.66, 2.10) 1.03 (−3.63, 5.70) 0.66 N/A
 Among females (n=2,075) −0.76 (−4.24, 2.71) −2.95 (−6.37, 0.47) 1.80 (−1.31, 4.91) 4.75 (0.23, 9.26)* 0.04* N/A
CRP, in mg/dL, b (95% CI)e (n=3,716) 0.04 (−0.18, 0.26) −0.14 (−0.34, 0.06) −0.01 (−0.14, 0.12) 0.13 (−0.15, 0.41) 0.35 None
Open in a new tab

Note: Linear regression was used for all outcomes except for CRP. Regression models included SNAP eligibility, examination year, SNAP eligibility X examination interaction (difference-in-differences estimate), age, sex, race/ethnicity, education, household size, unemployment, and home ownership. Models were stratified by age or sex if statistically significant (p<0.05) interaction terms were found. Sample weights accounting for study design and nonresponse were applied so results can be generalized to the population of civilian noninstitutionalized U.S. adults. NHANES = National Health and Nutrtion Examination Study; SNAP = Supplemental Nutrition Assistance Program

a

Additionally adjusted for waist circumference, diabetes diagnosis and diabetic medication(s).

b

Additionally adjusted for waist circumference, diagnosis of high cholesterol, cholesterol-lowering medication(s).

c

Additionally adjusted for three-way interaction of age X SNAP X exam year. The difference-in-difference estimate from this model represents the estimate for individuals 18 years of age.

d

Restricted to participants who provided fasting specimens.

e

Additionally adjusted for waist circumference.

f

Additionally adjusted for three-way interaction of gender X SNAP X exam year. The difference-in-difference estimate from this model represents the estimate for females.

In models that additionally adjusted for food insecurity and its change over time among the SNAP eligible group (food security X SNAP X exam), adults in SNAP eligible households had four-fold higher odds of participating in SNAP at baseline and, among those who participated, received higher average benefit amounts (Table 4). There were no statistically significant difference-in-differences with regard to SNAP participation, although participation increased among SNAP eligible adults and not among non-eligible. There was no difference-in-difference for recent average benefit amounts, but SNAP eligible adults who participated had increased benefit amounts after the policy change and non-eligible adults did not. Also, there was an interaction between food insecurity with the difference-in-difference estimate; the increased benefit amount after the policy change among SNAP eligible adults was greater among those who were food secure compared to the food insecure.

Table 4.

Estimated associations between the 2009 SNAP policy change and intermediary outcomes among NHANES non-pregnant participants aged 18–59 years with income <185% poverty

Baseline difference between SNAP eligible and likely not eligible Difference after policy change Difference-in-differences
Among those likely not SNAP eligible (131%–≤185% poverty) Among those SNAP eligible (≤130% poverty) Estimate p value Food insecurity interaction with D-in-D estimate
SNAP participation, OR (95% CI) (n=4,103) 4.07 (2.49, 6.66)* 1.31 (0.73, 2.36) 1.99 (1.06, 3.73)* 1.52 (0.66, 3.50) 0.32 0.80 (0.63, 1.01)
 Among food secure (n= 1,978) 3.48 (1.91, 6.31)* 1.45 (0.80, 2.65) 1.68 (0.91, 3.10) 1.16 (0.54, 2.49) 0.70 N/A
 Among food insecure (n=2,125) 5.07 (2.46, 10.45)* 1.31 (0.55, 3.09) 1.01 (0.66, 1.56) 0.77 (0.33, 1.82) 0.55 N/A
Most recent SNAP benefit amount, in dollars, b (95% CI) (n=1,830) 0.32 (0.07, 0.56)* 0.22 (−0.15, 0.59) 0.57 (0.33, 0.82)* 0.35 (−0.10, 0.81) 0.13 −0.10 (−0.19, −0.01)*
 Among food secure (n=644) 0.16 (−0.26, 0.59) 0.28 (−0.26, 0.82) 0.52 (0.31, 0.72)* 0.24 (−0.37, 0.84) 0.43 N/A
 Among food insecure (n=1,186) 0.48 (0.13, 0.83)* 0.18 (−0.30, 0.67) 0.24 (0.11, 0.37)* 0.05 (−0.43, 0.53) 0.82 N/A
Open in a new tab

Note: SNAP participation in past year were analyzed with logistic regression. SNAP benefit amounts were analyzed with negative binomial models among those who ever received SNAP benefits. Regression models included SNAP eligibility, examination year, SNAP eligibility X examination interaction (difference-in-differences estimate), age, sex, race/ethnicity, education, household size, unemployment, home ownership, household food insecurity and three-way interaction of food security X SNAP X exam. Sample weights accounting for study design and nonresponse were applied so results can be generalized to the population of civilian noninstitutionalized U.S. adults. NHANES = National Health and Nutrtion Examination Study; SNAP = Supplemental Nutrition Assistance Program

Sensitivity analyses were conducted, as described above. For the falsification test, there was no difference-in-difference for cotinine or lead levels and no difference over time in either the SNAP eligible nor non-eligible groups (Supplemental Tables 1 and 2). In other sensitivity analyses described in the methods inferences remained unchanged except the difference-in-difference estimates were not statistically significant for waist circumference in models restricted to either the 2,203 participants examined between May 1 and October 31 or the 2,368 employed participants and not statistically significant for hemoglobin A1c in the model restricted to employed adults (Supplemental Tables 3, 4 and 5).

Discussion

The 2009 SNAP policy change that expanded benefits for income-eligible adults was associated with less of an upward trend in hemoglobin A1c levels and, among young adults, a decrease in total cholesterol for those whose households were eligible for the expanded benefits. Importantly, population characteristics did not differ over time, suggesting that results are attributable to the policy change and not to population changes. The study also found increased SNAP participation among adults in eligible households and, among those who received SNAP, increased benefit amounts, highlighting that greater money for food may explain the blood sugar and cholesterol results. These results address a gap in health disparities literature by suggesting metabolic response to increased income for food for low-income adults.

Results from this study for hemoglobin A1c and total cholesterol may be explained due to the increased SNAP participation and larger SNAP benefits that were found among eligible adults. Although this study found increased food insecurity, that result may be due to the dramatically increased food insecurity during the Great Recession (Gundersen and Ziliak, 2018). Other studies that examined the 2009 policy change found it to be associated with reduced food insecurity (McClain et al., 2019; Nord and Prell, 2011) and less food consumed away from home (Burney, 2018), suggesting that it not only increased the amount of food purchased but also the quality of food consumed (Wolfson and Bleich, 2015). Importantly, low-income families often simultaneously face food insecurity and cost-related medication non-adherence and make trade-off decisions between food, health care and medications (Berkowitz et al., 2014; Mayer et al., 2016). The SNAP policy change was associated with increased spending on health care and other basic needs (Kim, 2016; Morrissey and Miller, 2020), suggesting spill-over effects that may improve health. A similar study estimated that fruit and vegetable subsidies could reduce incident diabetes and obesity (Choi et al., 2017). Together, these studies suggest reduced financial stress and/or more resources may help to prevent and manage chronic disease.

There are several potential reasons for the lack of change in CRP, LDL and waist circumference in this study. With regard to LDL cholesterol and waist circumference results, there is good evidence linking SNAP participation with improved food security (Gundersen et al., 2017), which is associated with lower risk of high cholesterol (Berkowitz et al., 2013). Lack of associations in this study may be due to longer lag time between exposure and clinical changes or because the population was already predominately healthy before the policy change. CRP may respond differently to chronic stress than other inflammatory cytokines because it typically responds less than IL-6 or IL-1β to acute stressors (Marsland et al., 2017; Steptoe et al., 2007). Additional studies are warranted to evaluate the effect of food policy changes and other stress- and nutrition-related outcomes.

These results contribute to the literature by showing that expansion of SNAP corresponds to improvements in blood sugar and cholesterol levels in young and middle-aged adults. These results are important because diabetes and high cholesterol levels have long been known to be important predictors of cardiovascular disease and mortality (Anderson et al., 1987; Kannel and McGee, 1979). Notably, the hemoglobin A1c and total cholesterol levels in this sample, were, on average, less than the thresholds for diagnosing diabetes or hyperlipidemia and associations with cholesterol were strongest among the youngest adults, suggesting that SNAP access may play a role in preventing chronic conditions in young and middle aged adults. Prior studies have found that SNAP participation is associated with less health care utilization (Berkowitz et al., 2017; Samuel et al., 2018; Szanton et al., 2017), less cost-related medication non-adherence (Pooler and Srinivasan, 2018) and, among those with diabetes, better glucose control (Mayer et al., 2016). Together, these studies show that improved SNAP access and benefit amounts may prevent and control burdensome and costly chronic conditions.

This study has timely policy implications. The increased SNAP benefit amounts arising from the American Recovery and Reinvestment Act in 2009 expired in November 2013 and the expanded eligibility criteria for most states has expired since then as their unemployment rate fell. However, several federal policy changes during the coronavirus pandemic temporarily re-implemented the same changes. Results from this study add to a growing body of evidence linking the 2009 policy change with improved health and food-related outcomes by linking the policy change with blood sugar and cholesterol levels. Given the increasing cost and public health burden of chronic conditions, these results suggest that expanding SNAP access and benefits may be a strategy for preventing and managing chronic disease among young and middle-aged low-income adults. For example, continuing the current SNAP expansion may correspond to population health improvements for income-eligible adults. In this study, benefit amounts increased more among food secure than insecure adults, which may identify a need to help food insecure adults receive the maximum benefit amount to which they are entitled.

Limitations

The repeated cross sectional study does not allow examination of individual trajectories over time, so this study focuses on population-level change. However, the study is strengthened by including a nationally representative sample of adults. Although the SNAP eligible and likely non-eligible groups were not balanced, the study applied a quasi-experimental approach and difference-in-differences results should be valid (Wing et al., 2018) because measured group characteristics were stable over time. Also, the SNAP eligible group had relatively more risk factors relevant to stress- and nutrition-sensitive outcomes. Therefore, the results in this study are unlikely to be attributable to observed group differences. Importantly, according to USDA data, about 70% of young and middle-aged adults who were eligible for SNAP between 2007 and 2010 SNAP actually received SNAP benefits; therefore, these analyses only estimate the population effect of the policy change, not the effect of the SNAP program itself. The study is limited to CRP for measurement of inflammatory biomarkers and cannot capture regional cost of living differences.

Conclusions

This study leveraged a unique opportunity to evaluate the effect of a SNAP policy change. The 2009 policy change that temporarily expanded SNAP access for childless adults and increased benefit amounts for all recipients was associated with population-level improvements in hemoglobin A1c and total cholesterol levels over time among adults eligible for the program. Greater SNAP access and larger benefits may help to prevent and manage diabetes and lipid disorders among young and middle-aged low-income adults.

Supplementary Material

1
2
3
4
5

Highlights:

  • A 2009 policy change expanded the Supplemental Nutrition Assistance Program (SNAP)

  • Difference-in-differences analyses were conducted among U.S. adults 18–59 years

  • SNAP-eligible adults had less increased hemoglobin A1c after the policy change

  • SNAP-eligible adults had decreased total cholesterol after the policy change

  • SNAP use and benefits increased among eligible adults after the policy change

Acknowledgements

LJS was supported by the NIA (K01AG054751). NHANES data are publicly available at https://wwwn.cdc.gov/nchs/nhanes/. LJS, SLS, and DJG designed the research, LJS obtained the data and conducted analyses; all authors contributed to the writing of the manuscript. The authors have no conflicts of interest to disclose.

Funding:

This study was funded by the National Institute on Aging (K01AG054751).

Abbreviations:

SNAP

Supplemental Nutrition Assistance Program

NHANES

National Health and Nutrition Examination Survey

CRP

C reactive protein

LDL

low density lipoprotein

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  1. 2015. Stata, in: StataCorp (Ed.). https://www.stata.com/, College Station, Texas. [Google Scholar]
  2. Anderson KM, Castelli WP, Levy D, 1987. Cholesterol and mortality. 30 years of follow-up from the Framingham study. JAMA 257:2176–80. [DOI] [PubMed] [Google Scholar]
  3. Berkowitz SA, Baggett TP, Wexler DJ, Huskey KW, Wee CC, 2013. Food insecurity and metabolic control among U.S. adults with diabetes. Diabetes Care 36:3093–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Berkowitz SA, Seligman HK, Choudhry NK, 2014. Treat or eat: food insecurity, cost-related medication underuse, and unmet needs. Am J Med 127:303–10 e3. [DOI] [PubMed] [Google Scholar]
  5. Berkowitz SA, Seligman HK, Rigdon J, Meigs JB, Basu S, 2017. Supplemental Nutrition Assistance Program (SNAP) Participation and Health Care Expenditures Among Low-Income Adults. JAMA Intern Med 177:1642–49. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Bickel G, Nord M, Price C, Hamilton W, Cook J, 2000. Guide to Measuring Household Food Security. Food and Nutrition Service, USDA, Alexandria, VA. [Google Scholar]
  7. Burney S, 2018. In-kind benefits and household behavior: The impact of SNAP on food-away-from-home consumption. Food Policy 75 134–46. [Google Scholar]
  8. Choi SE, Seligman H, Basu S, 2017. Cost Effectiveness of Subsidizing Fruit and Vegetable Purchases Through the Supplemental Nutrition Assistance Program. American Journal of Preventive Medicine 52:E147–E55. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Elkind MS, Luna JM, Moon YP, Liu KM, Spitalnik SL, Paik MC, Sacco RL, 2009. High-sensitivity C-reactive protein predicts mortality but not stroke: the Northern Manhattan Study. Neurology 73:1300–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Fan M, 2010. Do food stamps contribute to obesity in low-income women? Evidence from the National Longitudinal Survey of Youth 1979. Am J Agr Econ:1–16. [Google Scholar]
  11. Friedewald WT, Levy RI, Fredrickson DS, 1972. Estimation of the concentration of low-density lipoprotein cholesterol in plasma, without use of the preparative ultracentrifuge. Clin Chem 18:499–502. [PubMed] [Google Scholar]
  12. Glymour MM, Avendano M, Kawachi I, 2014. Socioeconomic status and health, in: Berkman LF, Kawachi I, Glymour MM (Eds.), Social Epidemiology. Oxford University Press, Oxford, pp. 17–63. [Google Scholar]
  13. Gruenewald TL, Karlamangla AS, Hu P, Stein-Merkin S, Crandall C, Koretz B, Seeman TE, 2012. History of socioeconomic disadvantage and allostatic load in later life. Social science & medicine 74:75–83. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Gundersen C, Kreider B, Pepper JV, 2017. Partial Identification Methods for Evaluating Food Assistance Programs: A Case Study of the Causal Impact of Snap on Food Insecurity. Am J Agr Econ 99:875–93. [Google Scholar]
  15. Gundersen C, Ziliak JP, 2018. Food Insecurity Research in the United States: Where We Have Been and Where We Need to Go. Applied Economic Perspectives and Policy 40:119–35. [Google Scholar]
  16. Jilcott SB, Wall-Bassett ED, Burke SC, Moore JB, 2011. Associations between Food Insecurity, Supplemental Nutrition Assistance Program (SNAP) Benefits, and Body Mass Index among Adult Females. Journal of the American Dietetic Association 111:1741–45. [DOI] [PubMed] [Google Scholar]
  17. Kannel WB, McGee DL, 1979. Diabetes and glucose tolerance as risk factors for cardiovascular disease: the Framingham study. Diabetes Care 2:120–6. [DOI] [PubMed] [Google Scholar]
  18. Kim J, 2016. Do SNAP participants expand non-food spending when they receive more SNAP Benefits?—Evidence from the 2009 SNAP benefits increase. Food Policy 65 9–20. [Google Scholar]
  19. Marsland AL, Walsh C, Lockwood K, John-Henderson NA, 2017. The effects of acute psychological stress on circulating and stimulated inflammatory markers: A systematic review and meta-analysis. Brain Behav Immun 64:208–19. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Mayer VL, McDonough K, Seligman H, Mitra N, Long JA, 2016. Food insecurity, coping strategies and glucose control in low-income patients with diabetes. Public Health Nutr 19:1103–11. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. McClain AC, Tucker KL, Falcon LM, Mattei J, 2019. Food insecurity and dietary intake by Supplemental Nutrition Assistance Program participation status among mainland US Puerto Rican adults after the 2009 American Recovery and Reinvestment Act. Public Health Nutrition 22:2989–98. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. McGuire S, 2013. Supplemental Nutrition Assistance Program: Examining the Evidence to Define Benefit Adequacy. Adv Nutr 4:477–78. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Meyer BD, Mittag N, Goerge RM, 2018. Errors in survey reporting and imputation and their effects on estimates of food stamp program participation. National Bureau of Economic Research; Cambridge, MA. [Google Scholar]
  24. Morrissey TW, Miller DP, 2020. Supplemental Nutrition Assistance Program Participation Improves Children’s Health Care Use: An Analysis of the American Recovery and Reinvestment Act’s Natural Experiment. Academic Pediatrics 20:863–70. [DOI] [PubMed] [Google Scholar]
  25. Nord M, Golla AM, 2009. Does SNAP Decrease Food Insecurity? Untangling the Self-Selection Effect. Economic Research Service, United States Department of Agriculture, Washington, DC. [Google Scholar]
  26. Nord M, Prell M, 2011. Food Security Improved Following the 2009 ARRA Increase in SNAP Benefits. U.S. Department of Agriculture, Economic Research Service, Washington, DC. [Google Scholar]
  27. Park YW, Zhu SK, Palaniappan L, Heshka S, Carnethon MR, Heymsfield SB, 2003. The metabolic syndrome - Prevalence and associated risk factor findings in the US population from the Third National Health and Nutrition Examination Survey, 1988–1994. Archives of Internal Medicine 163:427–36. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Pooler JA, Srinivasan M, 2018. Association Between Supplemental Nutrition Assistance Program Participation and Cost-Related Medication Nonadherence Among Older Adults With Diabetes. JAMA Intern Med. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Samuel LJ, Szanton SL, Cahill R, Wolff JL, Ong P, Zielinskie G, Betley C, 2018. Does the Supplemental Nutrition Assistance Program Affect Hospital Utilization Among Older Adults? The Case of Maryland. Popul Health Manag 21:88–95. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Service, C.R., 2019. The Supplemental Nutrition Assistance Program (SNAP): Categorical Eligibility. Congressional Research Service, Washington, DC. [Google Scholar]
  31. Services, U.S.D.o.H.a.H., 2013. National Health and Nutrition Examination Survey: Analytic Guidelines, 1999–2010. National Center for Health Statistics. [PubMed] [Google Scholar]
  32. Shadish WR, Cook TD, Campbell DT, 2002. Quasi-experiments: interrupted time-series designs, Experimental and Quasi-experimental Designs for Generalized Causal Inference. Wadsworth Cengage Learning, Belmont, CA, pp. 171–206. [Google Scholar]
  33. Sonik RA, 2016. Massachusetts Inpatient Medicaid Cost Response to Increased Supplemental Nutrition Assistance Program Benefits. Am J Public Health 106:443–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Staiano AE, Reeder BA, Elliott S, Joffres MR, Pahwa P, Kirkland SA, Paradis G, Katzmarzyk PT, 2012. Body mass index versus waist circumference as predictors of mortality in Canadian adults. Int J Obesity 36:1450–54. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Stamler J, Vaccaro O, Neaton JD, Wentworth D, 1993. Diabetes, other risk factors, and 12-yr cardiovascular mortality for men screened in the Multiple Risk Factor Intervention Trial. Diabetes Care 16:434–44. [DOI] [PubMed] [Google Scholar]
  36. Steptoe A, Hamer M, Chida Y, 2007. The effects of acute psychological stress on circulating inflammatory factors in humans: a review and meta-analysis. Brain Behav Immun 21:901–12. [DOI] [PubMed] [Google Scholar]
  37. Szanton SL, Samuel LJ, Cahill R, Zielinskie G, Wolff JL, Thorpe RJ Jr., Betley C, 2017. Food assistance is associated with decreased nursing home admissions for Maryland’s dually eligible older adults. BMC Geriatr 17:162. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Vanderschueren S, Deeren D, Knockaert DC, Bobbaers H, Bossuyt X, Peetermans W, 2006. Extremely elevated C-reactive protein. Eur J Intern Med 17:430–3. [DOI] [PubMed] [Google Scholar]
  39. Waehrer G, Deb P, Decker SL, 2015. Did the 2009 American Recovery and Reinvestment Act affect dietary intake of low-income individuals? Economics & Human Biology 19:170–83. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Wing C, Simon K, Bello-Gomez RA, 2018. Designing Difference in Difference Studies: Best Practices for Public Health Policy Research. Annu Rev Publ Health 39:453–69. [DOI] [PubMed] [Google Scholar]
  41. Wolfson JA, Bleich SN, 2015. Is cooking at home associated with better diet quality or weight-loss intention? Public Health Nutrition 18:1397–406. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Wooldridge J, Imbens G, 2007. Difference-in-differences estimation, What’s New in Econometrics? National Bureau of Economic Research. [Google Scholar]
  43. Zheng H, Tumin D, Qian Z, 2013. Obesity and mortality risk: new findings from body mass index trajectories. Am J Epidemiol 178:1591–9. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1
NIHMS1716894-supplement-1.docx (14.6KB, docx)
2
NIHMS1716894-supplement-2.docx (15.1KB, docx)
3
NIHMS1716894-supplement-3.docx (15.4KB, docx)
4
NIHMS1716894-supplement-4.docx (14.6KB, docx)
5
NIHMS1716894-supplement-5.docx (14.3KB, docx)

RESOURCES