The Gerontologist Vol. 48, No. 2, 235–244 Copyright 2008 by The Gerontological Society of America Falling Through the Social Safety Net: Food Stamp Use and Nonuse Among Older Impoverished Americans Esme Fuller-Thomson, PhD, MSW,1 and Melissa Redmond, MSW1 Purpose: Older adults are less likely than any other age group to use the federal Food Stamp Program. The personal and social costs of elderly diet insufficiency include disease exacerbation, depression, and increased hospitalization. In order to improve targeting and outreach efforts, this study identifies the characteristics of eligible older Americans who are not receiving food stamps and assesses the validity of the Andersen behavioral model for predicting impoverished older adults food stamp use. Design and Methods: We conducted a secondary analysis of the 2003 American Community Survey, which is a nationally representative survey with a response rate of 96.7%. We restricted our study subsample to the 14,724 impoverished American citizens who were aged 65 years and older. We used bivariate and logistic regression analyses to compare the 2,796 food stamp recipients with the 11,928 nonrecipients. Results: One in five impoverished older American citizens had received foods stamps in the preceding year. Female respondents, renters, younger respondents, disabled individuals, and those who received Supplemental Security Income or welfare were more likely to receive food stamps. The pseudo-R-square value indicated that the Andersen Behavioral Model explained 28% of the model’s variability. Implications: Improved targeting is needed to enhance older adults’ participation rates. Mobile and satellite food stamp offices in lower income neighborhoods and other innovative outreach programs that collaborate with community We acknowledge the generous support of the Royal Bank of Canada Graduate Research Fellowship in Applied Social Work and the assistance of Leanne McCormack, MSW, with manuscript preparation. Address correspondence to Esme Fuller-Thomson, Factor-Inwentash Faculty of Social Work, University of Toronto, 246 Bloor Street W, Toronto, Ontario M5S 1A1, Canada. E-mail: esme.fuller.thomson@ utoronto.ca 1 Faculty of Social Work, University of Toronto, Canada. Vol. 48, No. 2, 2008 235 partners could also improve access. With the vast majority of impoverished older adults not receiving food stamps, strategies such as these are extremely important to rectify this situation among the most vulnerable group of older Americans. Key Words: Access, Andersen Behavioral Model, Food Stamp Programs, Poverty, Social Welfare Programs Providing more support to poor Americans than any other public assistance plan, the federal Food Stamp Program (FSP) is designed to increase the food-purchasing power of impoverished households (Wolkwitz, 2007). Despite the program’s $23.9 billion budget allocation, recent estimates suggest that only 54% of eligible Americans actually receive food stamp benefits (Castner & Schirm, 2005). Concerned with this underutilization, the FSP is striving to reach a participation rate of 68% by 2010 (Castner & Schirm). Hunger in the United States is still a pressing social issue. One in eight households below the poverty line has at least one member who has experienced involuntary hunger because of insufficient funds to buy food (Nord, Andrews, & Carlson, 2004). Although access to the FSP is ostensibly offered to all eligible individuals living in poverty, research literature points to differential participation rates for diverse sectors of the American public. As a result, specific age cohorts and ethnic or racial groups require tailored outreach strategies to maximize their program participation. These strategies must be informed by solid research focused on each group’s needs and patterns of utilization. In this article we study the characteristics of a particularly vulnerable subgroup: impoverished older Americans who are not receiving food stamps (The term impoverished refers to individuals who have incomes that are less than 100% of the poverty level.) Numerous studies have reported that less than a third of eligible older Americans actually receive FSP benefits (Cunnyham, 2004; Haider, Jacknowitz, & Schoeni, 2003; Wilde & Dagata, 2002). These participation percentages are much lower than those for any other age group (Cunnyham, 2004; Rosso, 2001). Surprisingly, little research has focused on the characteristics of elderly persons who are eligible for the FSP but are not using the service. Even fewer studies have used multivariate models and theoretically informed analyses while attempting to describe this population. Reasons for low program utilization rates among American seniors include the following: (a) barriers to information concerning eligibility, program benefits, and policies, which include confusion concerning changing eligibility standards; (b) enrollment impediments, which include difficulty accessing local welfare offices, misconceptions regarding application process complexity, and language issues; (c) concern about stigma, which may be more pronounced among older adults; and (d) perceptions concerning lack of need or inadequate benefits (Gabor, Williams, Bellamy, & Hardison, 2002; McConnell & Ponza, 1999; Parker, Jog, & Kennedy, 2005; U.S. General Accounting Office, 2000; Wilde & Dagata, 2002; Zedlewski & Rader, 2005). As increasing numbers of Americans live to older ages, concern regarding their food security and the adequacy of their nutritional intake takes on added importance (Guthrie & Lin, 2002; Wellman & Kamp, 2004). The personal and social costs of elderly diet insufficiency include disease and disability exacerbation, depression, individual and caregiver strain, increased hospitalization, and earlier mortality (Food Security Institute, 2003; Lee & Frongillo, 2001). There is a need for timely research that focuses on the demographic characteristics of eligible elderly persons who do not use food stamps. This research will generate a detailed profile of impoverished older Americans who, though eligible, are not accessing the FSP. A better understanding of the factors that distinguish participating and nonparticipating impoverished elderly Americans will address many of the gaps in our theoretical and practical understanding of elderly food stamp utilization and inform much needed program outreach efforts. We use Andersen’s (1995) behavioral model to predict the use of food stamps among older adults who are living in poverty. Andersen’s framework has been one of the most influential models in the field of health care utilization over the past three decades (Calsyn & Winter, 2000; de Klerk, Huijsman, & McDonnell, 1997; Jang, Kim, & Chiriboga, 2005; Jheriot, Segal, & Cowser, 2003; Li, 2006). However, it has not commonly been used to assess utilization of social assistance programs. The model is a helpful theoretical framework for understanding the demographic, social, and situational factors influencing program-utilization patterns. 236 Andersen hypothesized that patterns of utilization are affected by predisposing, enabling, and need characteristics. In equitable services, need characteristics will be the strongest predictors of utilization. In inequitable services, enabling factors will provide the greatest explanatory power. We have modified these three groups of factors to address food stamp use. Predisposing characteristics include demographic characteristics such as age, gender, and race or ethnicity. Greater food stamp use has been associated with being female (Rosso & Faux, 2003), Hispanic (Faux & Rosso, 2003), African American (Faux & Rosso), and younger (Martin, Cook, Rogers, & Joseph, 2003). Characteristics that enable utilization of the FSP include skill in speaking English (Algert, Reibel, & Renvall, 2006), higher educational attainment (Kirsch, Jungeblut, Jenkins, & Kolstad, 1993; Rank & Hirschl, 2005; Zedlewski & Rader, 2005), family composition, and use of other social assistance programs such as Supplemental Security Income (SSI) and public assistance (U.S. General Accounting Office, 2004; Zedlewski & Rader). Haider and colleagues (2003) reported that utilization rates were much higher among those individuals with children than among those without them. One possibility is that knowledge about and access to social assistance programs is increased through outreach programs in the schools, such as lunch programs or school social workers. Need characteristics refer to health-related characteristics and financial status. In particular, we include the association between food stamp use and functional limitations, limitations in activities of daily living, and blindness or deafness. Rosso and Faux’s (2003) analysis found that food stamp use was higher for individuals with disabilities than it was for those without them. Further, Rank and Hirschl (2005) found that duration of use was higher for those individuals with disabilities. Although all individuals studied were living below the poverty line, we included a variable in our analyses indicating household income as a percentage of the poverty line in order to identify those persons living in the most extreme poverty. The proposed analyses were to validate the Andersen behavioral model for predicting impoverished older adults’ use of food stamps. Methods Participants The 2003 American Community Survey (ACS) is a nationally representative survey of 838,293 households with a response rate of 96.7% (U.S. Census Bureau, n.d.). A random sample of the ACS 2003 is available as a Public Use Microdata Set (PUMS). Because there is variability in the eligibility of legal immigrants, we subsequently excluded noncitizens from our sample. Thus, the final unweighted ACS PUMS sample used for this study was 14,724 The Gerontologist respondents who were American citizens living in poverty and were aged 65 years and older, of whom 2,796 had received food stamps at some point in the 12 months preceding the survey. Design and Procedure We used SPSS 15.0 for Windows to conduct our analyses. In order to validate the Andersen Behavioral Model, we conducted both bivariate and multivariate tests. To assess whether groups (e.g., men and women) differed significantly in the percentage using food stamps, we conducted chi-square analyses for each of the predisposing, enabling, and need characteristics. A logistic regression analysis with food stamp use as the dependent variable was also necessary for two reasons. First, we were interested in whether each of the variables discussed here were independently associated with food stamp use, once the other variables were also included in the analysis. Second, Andersen’s model is based on an investigation of groups of variables divided into predisposing, enabling, and need characteristics. An assessment of the change in the pseudo-R-squared value associated with each group of characteristics is an important way to ascertain whether service provision is based primarily on need or other factors. For the logistic regression of food stamp use, we entered independent variables in three blocks: predisposing, enabling, and need characteristics. We assessed the independent contribution of each block of predictors, and we generated odds ratios (ORs) and 95% confidence intervals. Theory-driven hierarchical entry of each of the blocks of variables produced results that were more clearly interpretable (Phillips, Morrison, Andersen, & Aday, 1998). Our analyses utilized a weighting variable designed by the U.S. Census Bureau. This weight variable represents the probability of selection, including noninterview adjustments and controlling for the 2000 Census housing units and population level. To avoid falsely increasing the likelihood that we would obtain a significant p value, we rescaled the weight variable to have a mean value of 1 for the subsample. We determined the receipt of food stamps by the participants’ response to the following question: At any time in the past 12 months, did anyone in this household receive food stamps? We based all variables on self-report, including ethnicity and race. We based household income on a summation of income for all household members from the following sources: (a) SSI; (b) public assistance or welfare payments from the state or local welfare office; (c) wages, salary, commissions, bonuses, or tips from all jobs; (d) self-employment income; (e) interest, dividends, net rental income, royalty income, or income from estates and trusts; (f) social security or railroad retirement; (g) retirement, survivor, or disability pensions; and (h) other sources Vol. 48, No. 2, 2008 237 of income received regularly such as veteran’s payments, unemployment compensation, child support, or alimony. We dichotomized responses to the first two questions (yes or no) to determine if the household received SSI or welfare. The census calculated the household’s family poverty level by using household composition and household income data. The sample was restricted to those below 100% of the poverty line. In 2003, the poverty line was $8,825 for an older adult living alone and $11,122 for an older couple (DeNavas-Walt, Proctor, & Mills, 2004). We then created a variable identifying four levels of poverty under the poverty line: (a) 0% to 24%, (b) 25% to 49%, (c) 50% to 74%, and (d) 75% to 99% of the poverty line. Further, respondents were asked if they had either or both of the following long-lasting conditions: (a) blindness, deafness, or severe vision or hearing impairment; or (b) a condition that substantially limits physical activities such as walking, climbing stairs, reaching, lifting, or carrying. The former we labeled severe vision or hearing difficulty and the latter we defined as functional limitations. Respondents were also asked if they had a physical, mental, or emotional condition lasting 6 months or more that made it difficult to do any of the following: (a) learning, remembering, or concentrating; (b) dressing, bathing, or getting around inside the home; (c) going outside the home alone to shop or visit a doctor’s office; and (d) working at a job or business. For the purposes of this study, we defined these as difficulty remembering, limitations in activities of daily living, difficulty going out, and difficulty working, respectively. Results Of older American citizens living in poverty, approximately one in five was receiving food stamps. The majority of our chi-square analyses indicated that, as we hypothesized, a wide range of predisposing, enabling, and need characteristics were associated with food stamp usage (see Table 1). All bivariate analyses were below the .001 level of significance. Food stamp utilization rates were higher among female, Hispanic, and African American respondents. Prevalence rates decreased with age. Individuals with underage children in their homes were twice as likely to use food stamps as those who were not coresiding with children. Older adults who received welfare or SSI had much higher rates of food stamp use than those not receiving these social services. With respect to need characteristics, older adults who rented their accommodation had three times the rates of food stamp use in comparison with those who owned their accommodation and did not have a mortgage. Having a disability was associated with higher rates of food stamp utilization for each of the six types of disabilities that we investigated. Table 1. Analysis of Predisposing, Enabling, and Need Characteristics Associated With FSP Utilization Among Poor Older Americans Total (n) FSP Use (%) p 4,198 10,526 16.4 20.4 ,.001 Ethnicity Not Hispanic Hispanic 13,665 1,059 17.9 31.0 ,.001 Race Not African American African American 12,062 2,662 16.5 30.1 ,.001 6,566 5,881 2,277 24.0 16.8 11.1 ,.001 12,699 1,180 845 17.5 19.2 36.1 ,.001 4,309 3,423 6,131 861 26.9 21.5 13.4 13.3 ,.001 Citizenship U.S. citizen by birth Citizenship through naturalization 13,487 1,237 17.9 29.5 ,.001 Children present in home No children 1 or more children 13,747 977 17.8 37.2 ,.001 Welfare No welfare received Welfare received 14,373 351 18.3 57.5 ,.001 Supplemental Security Income No Yes 12,881 1,843 14.5 51.3 ,.001 1,751 1,478 3,735 7,760 6.5 13.2 19.9 23.1 ,.001 4,627 2,034 7,419 644 32.4 15.7 10.4 17.3 ,.001 Limitations in Activities of daily living Yes No 2,156 12,568 27.2 17.8 ,.001 Severe vision or hearing difficulty Yes No 3,351 11,373 24.3 17.8 ,.001 Variable Predisposing characteristics Gender Male Female Age (years) 65–74 75–84 85þ Enabling characteristics English-language skills Speaks English at home Speaks English as second language, very well Speaks English not well or not at all Education Primary Some high school (no diploma) High school diploma University degree Need characteristics Poverty quartiles 0% to 24% 25% to 49% 50% to 74% 75% to 99% Housing tenure Rented for cash rent Owned with mortgage Owned free and clear Occupied without payment of cash rent (Table continues on next page) 238 The Gerontologist Table 1. (Continued) Total (n) FSP Use (%) p 3,926 10,798 25.8 16.8 ,.001 Functional limitations Yes No 6,479 8,245 25.7 14.1 ,.001 Difficulty remembering Yes No 2,596 12,128 27.7 17.4 ,.001 6,585 8,139 26.2 13.8 ,.001 Variable Difficulty going out Yes No Difficulty working Yes No Note: The analysis is bivariate; poor older Americans are individuals aged 65 years or older who are living below the poverty line (unweighted n = 14,724). The Total column is unweighted; the FSP Use column is weighted. The p-value column shows the value of the chi-square tests. Information is from the American Community Survey 2003 Public Use Microdata Set. FSP = Food Stamp Program. However, three of the enabling characteristics and one of the need characteristics reached statistical significance but were in the opposite direction of what we had anticipated: Those who did not speak English well had twice the rates of food stamp use as those who spoke English at home. Individuals with less education were more likely to utilize food stamps. Naturalized citizens were more likely to use food stamps than were those born in the United States. Older adults living below 25% of the poverty line had less than one third the utilization rates as those living between 75% and 99% of the poverty line. The logistic regression (see Table 2) presents the independent contribution of each of the characteristics, controlling for the other variables in the analysis. All four of the predisposing characteristics were significant. Women, Hispanics, and African Americans had higher odds of food stamp use. In comparison with those individuals aged 85 years and older, those aged 65 to 74 had almost three times the odds of using food stamps. The Nagelkerke R-square statistic is a pseudo-R square used in logistic regression to estimate the percentage of variance in the outcome variable explained by predictors in the model (Nagelkerke, 1991). The pseudo-R square indicated that the model including only predisposing characteristics explained 7.4% of the variance in food stamp utilization for the impoverished older adults in the sample. With the addition of enabling characteristics, the change in the pseudo-R square suggested that an additional 14% of the variance in food stamp utilization among this sample was explained, of which 9.5% was associated with just two variables (welfare use and SSI). The majority of the enabling characteristics were significant. Those individuals who did not speak English well had 39% higher odds of using food stamps than those who spoke English at home. In comparison with college-educated respondents, Vol. 48, No. 2, 2008 239 those with only a primary school education had 64% higher odds of using food stamps and those who had some high school but had not graduated had 37% higher odds. Naturalized citizens had 27% higher odds of food stamp use than those born in the United States. Older adults with children in their homes had more than twice the odds of food stamp use than their peers without coresident children. Receiving welfare was associated with 3.7 times the odds of using food stamps. Similarly, SSI receipt was associated with three times the odds of food stamp utilization. The change in the pseudo-R-square value indicated that need characteristics explained a total of 6.6% of the model variance. In comparison with those individuals living below 25% of the poverty line, persons in each successive quintile had higher odds of using food stamps (OR = 1.50, 2.21, and 2.79, respectively). In comparison with those persons who owned their own home and had paid off their mortgage, the odds of food stamp use was significantly higher among owners with a mortgage (OR = 1.25), those who lived with others without paying rent (OR = 1.56), and highest among those who rented their accommodation (OR = 2.64). Those with vision or hearing difficulties, functional limitations, or disabilities that impacted their ability to work had elevated odds of using food stamps. Several of the need characteristics that had reached the level of significance in the bivariate analyses were no longer significant in the multivariate analyses (i.e., limitations in activities of daily living, difficulties in going out, and difficulties remembering). Discussion This study has several limitations. Two factors may result in the exclusion of some eligible older Table 2. Logistic Analysis of FSP Utilization Among Poor Older Americans Variable OR 95% CI p Predisposing characteristics Female Male Hispanic Not Hispanic African American 1.32 1.00 1.23 1.00 1.73 (1.19–1.47) Referent (1.03–1.47) Referent (1.55–1.94) ,.001 1.00 Referent 2.81 1.90 1.00 (2.38–3.31) (1.61–2.23) Referent ,.001 ,.001 0.93 1.39 1.00 (0.76–1.14) (1.12–1.74) Referent ,.48 ,.003 Education Primary Some high school (no diploma) High school diploma University Graduate 1.64 1.37 0.97 1.00 (1.31–2.07) (1.08–1.73) (0.77–1.22) Referent ,.001 .009 .81 Citizenship Through naturalization U.S. citizen by birth 1.27 1.00 (1.06–1.51) Referent .009 Children present in home 1 or more No children 2.51 1.00 (2.15–2.93) Referent ,.001 Welfare recipient Not a welfare recipient SSI recipient Not a SSI Recipient 3.74 1.00 2.97 1.00 (2.93–4.79) Referent (2.64–3.37) Referent ,.001 1.00 1.50 2.21 2.79 Referent (1.16–1.94) (1.78–2.75) (2.27–3.43) 1.00 1.25 2.64 1.56 1.05 1.00 1.18 1.00 0.99 1.00 1.38 1.00 1.08 1.00 1.35 1.00 Referent (1.07–1.46) (2.37–2.94) (1.23–2.00) (0.91–1.22) Referent (1.06–1.33) Referent (0.87–1.13) Referent (1.23–1.55) Referent (0.96–1.23) Referent (1.20–1.52) Referent Not African American Age: years 65–74 75–84 85 and over Enabling characteristics English-language skills Speaks as second language, very well Speaks not well or not at all Speaks English at home Need characteristics Poverty quartiles Household income , 25% of poverty line 25% to 49% 50% to 74% 75% to 99% Housing tenure Accommodation owned mortgage free Owned with mortgage Rented for cash rent Occupied without payment of cash rent Limitations in activities of daily living (ADL) No limitations in ADL Severe vision or hearing difficulty No severe vision or hearing difficulty Difficulty going out No difficulty going out Functional limitations No functional limitations Difficulty remembering No difficulty remembering Difficulty working No difficulty working .02 ,.001 ,.001 .002 ,.001 ,.001 .006 ,.001 ,.001 .49 .003 .86 ,.001 .21 ,.001 Note: Poor older Americans are individuals aged 65 years or older who are living below the poverty line (unweighted n = 14,724). Information is from the American Community Survey 2003 Public Use Microdata Set. The Nagelkerke pseudo-R-square value that is due to the complete model is 0.282. The change in this value per block is as follows: predisposing characteristics, 0.074; enabling characteristics, 0.142 (receipt of SSI and welfare explain 0.095); need characteristics, 0.066. FSP = Food Stamp Program; OR = odds ratio; CI = confidence interval; SSI = Supplemental Security Income. 240 The Gerontologist adults from our sample. Some legal immigrants are eligible for food stamps whereas others are not. First, because the ACS did not contain sufficient information to determine food stamp eligibility for noncitizens, we excluded all noncitizens from the analysis. Second, we restricted the sample to those with gross household incomes under the poverty line. Eleven percent of food stamp recipients have gross incomes over 100% of the poverty line, although their net incomes meet the eligibility criteria of being below the poverty line (U.S. Department of Agriculture, Food and Nutrition Service, n.d.a). Thus, we may have excluded a minority of eligible older adults. We may also have erred in the opposite direction. Unfortunately, the ACS does not include information on financial or vehicle assets. Therefore, some of the respondents in our analysis may not be eligible for food stamps because their assets exceed prescribed levels. Furthermore, the ACS does not have information on whether American Indian and Alaskan Native respondents, who make up 1% of the sample, were participating in food distribution programs on Indian reservations and were therefore ineligible for food stamps. Despite these limitations, the analyses do provide some helpful insights into the characteristics of impoverished older Americans who do not use food stamps. Using 2003 data, we found an overall prevalence rate of 19.2%. This percentage is considerably lower than the 26.9% (Cunnyham, 2004) and 30% (Wilde & Dagata, 2002) previously identified in studies of those individuals aged 60 and older. In contrast, our sample included only those aged 65 and older. We found that participation rates were much higher among the younger cohorts, and thus the different age minimum may partially explain these disparities. Furthermore, as we already discussed, we may have included in our sample some older adults who were ineligible for food stamps as a result of wealth, which we could not assess. Predisposing Characteristics Our results support the findings of the previous literature indicating that food stamp rates are higher among women, African Americans, and Hispanics (Faux & Rosso, 2003; Rosso & Faux, 2003). However, even in these groups, utilization rates are still abysmal. At least two thirds of those persons eligible in each of these populations are not receiving food stamps. This study’s finding that the younger cohort (age 65–74) had almost three times the odds of food stamp use as compared with the oldest respondents (age 85 and older) is provocative. Rogers-Dillon (1995) suggests that welfare stigma is produced through an interaction between an individual’s life history, situation, and social audience. An earlier incarnation of the FSP was inaugurated in 1939 to Vol. 48, No. 2, 2008 241 provide support to the neediest among the poor. One explanation for the reticence of the oldest respondents is the possibility of greater concern over ‘‘welfare stigma’’ among those who lived through the Depression era during their adolescence and young adulthood. In addition, although food stamps should not be synonymous with welfare, it is important to note that many states require that applicants visit the same location in order to enroll in either assistance program. Florida, Kansas, New Jersey, and West Virginia are among a handful of states that permit online applications. Furthermore, the historic practice in which individuals physically hand over the food stamps to their local store when they pay for their groceries may be associated with public shame. Most states are now using electronic benefit transfer cards. Similar to debit cards, electronic benefit transfer cards were ostensibly introduced to counteract fraud but may also serve to minimize stigma (Food Stamp Program: Frequently Asked Questions, n.d.). This being said, the imposition of new technology may still serve to alienate the oldest eligible Americans. With nine out of ten of the oldest cohort not receiving food stamps, it is clear that focused outreach and public education to destigmatize the FSP are necessary. Enabling Characteristics Several of our findings on enabling characteristics were in opposition to our hypotheses and the previous research on the topic. For instance, although we were not able to assess differential access rates among various linguistic groups, we anticipated that, because they would have more difficulties learning about and accessing the service, those persons with low English proficiency would have lower rates of FSP participation. Although our expectations were in keeping with the research by Algert and colleagues (2006), this did not prove to be the case. Perhaps our finding of higher use among nonanglophones may be due to the fact that a lack of English skills is highly correlated with duration of poverty (Urban Institute, 2002, 2006). Greater duration of poverty, in turn, is associated with a higher likelihood of accessing social services. Another potential explanation is that the FSP has made a substantial effort to provide Spanish material and bilingual staff. The greater rates of utilization among non-English-speaking older adults may reflect the success of this outreach. This might also explain our surprising finding that, among impoverished Americans, those who were citizens through naturalization rather than through birth were more likely to use food stamps. We had anticipated that education would facilitate access to services, as is the case in medical care (U.S. Department of Health and Human Services, 1998). However, our findings indicate that there is a higher likelihood of food stamp utilization among the poorly educated. Typically, lower levels of educational attainment are associated with lower wages and less job security, and thus a greater need over the years to use food stamps. Furthermore, well-educated older adults are more likely to have accrued wealth and thus be ineligible for services. As we already mentioned, our inability to determine wealth may have resulted in the improper inclusion of some individuals in our subsample. Three of our enabling characteristics did support our original hypotheses: Children present in the home, receipt of welfare, and receipt of SSI were each associated with higher food stamp utilization. The pseudo-R-squared statistics indicate that almost 10% of the variability in the logistic regression was explained by the latter two variables. Although it appears that many welfare workers are connecting clients to the FSP, there are still more than 40% of welfare recipients who are not getting food stamps. Many states may benefit from adopting a strategy such as the one used in Maine. Maine has provided leadership in the linking of social welfare benefits through an integrated computer system. When an individual applies for one benefit program (e.g., Medicaid or welfare), eligibility for the FSP is concurrently assessed. Eligible individuals are then immediately invited to apply for food stamps and other appropriate programs (Wallack, 2006). This outreach project contributed to a 49% increase in food stamp participation between 2002 and 2005 (Acheson, 2006). Initiatives such as this hold great promise for improving access to food stamps for impoverished older Americans. Need Characteristics Andersen’s framework proposes that those individuals with greater need are more likely to use services. In the bivariate analyses, all six types of disabilities were associated with greater utilization of food stamps. This was in keeping with earlier research (Rosso & Faux, 2003). Those persons with disabilities may have both a greater need for food stamps and a longer duration of need (Rank & Hirschl, 2005). However, in the logistic regression analyses, only three of the six variables remained statistically significant. It is not surprising that we found that those individuals who owned their accommodation mortgage free had much lower food stamp utilization rates. Low-income older adults who are renters and, to a lesser extent, those who continue to have mortgage payments devote a higher proportion of their income to housing and related costs. This results in less disposable income for purchasing food and other necessities, thus contributing to a greater need for social service support such as the FSP. Although 242 subsidized housing is one way of increasing the disposable incomes of older renters living in poverty, applicants often face overly long waitlists. In one tenth of the Section 202 sites, waitlists have been closed because demand has vastly outstripped supply (Kochera, 2006). In contrast to our expectations, we found that those who were in the poorest quartile of the poverty index (0% to 24% of the poverty line) had the lowest levels of food stamp utilization. In 2003, this level of poverty indicated an annual income of $2,206 or less for an older adult living alone and $2,781 or less for an older couple. Only 1 in 15 older adults living in this level of extreme poverty was receiving food stamps. We hypothesize two reasons for this. First, if people in this economic situation applied for food stamps, it is probable that they would also be referred to welfare or SSI for assistance. Once welfare or SSI was received, their income would be considerably above the 25% of the poverty-line cutoff we have specified here. Second, the strongest association we found in our analyses was between welfare, SSI, and food stamp use. We anticipate that much of this association is related to the situation in which welfare recipients are referred to or screened for the FSP program by their welfare or SSI worker. It is probable that those in the lowest poverty quartile are not receiving any government subsidies such as welfare, a program designed to move families well above 25% of the poverty line. Thus, those in this lowest income category are probably not in contact with welfare workers who could inform them about the FSP. Although only 12% of older adults living under the poverty line are in this lowest income category, it is imperative that future outreach targets this particularly needy and underserved group. Although the pseudo-R-square value indicates that Andersen’s framework explained a substantial proportion (i.e., 28%) of the variability in food stamp utilization in older adults living below the poverty line, the model definitely requires further refining. The fact that the direction of the association between several key determinants and utilization was the reverse of what we had anticipated indicates the pressing need for further research. Qualitative research might help researchers understand how impoverished older adults learn about their eligibility for the FSP and what factors motivate or dissuade them from program participation. A comparison of pseudo-R-square values suggest that enabling characteristics explained twice the variability associated with either predisposing or need characteristics. This is indicative of an inequitable pattern of service utilization. In an equitable system, need should be the predominant factor associated with utilization. Our findings may have been influenced by the narrow subsample we selected: All older adults in our sample were living below the poverty line and therefore, by definition, The Gerontologist had need. However, additional need-related factors such as degree of poverty within this subsample, housing tenure, and disability levels were associated with a change in the pseudo-R-square value of only 7%. The gap in service delivery to older adults living in poverty is stark. Even among those persons receiving welfare or SSI, fewer than 60% were concurrently receiving food stamps. In no other examined subgroup did we find more than 40% accessing the program. Nine out of ten impoverished Americans aged 85 and older were not receiving food stamps. Clearly targeted outreach to this population is essential. It would be beneficial for the FSP to set older adult participation rate targets for 2010 and 2020 in order to keep this population’s needs in the forefront of program planning. Concomitantly, the FSP will have to annually assess the participation of older adults. Several promising strategies have been proposed to improve food stamp utilization rates. Community education could address erroneous perceptions regarding low program benefits. Creating mobile and satellite food stamp offices in lower income neighborhoods would improve access, as would offering online application forms in all states (Martin et al., 2003). Furthermore, having these sites be distinct from welfare offices may decrease the perceived stigma attached to food stamp use. Although there is a general online prescreen tool available (U.S. Department of Agriculture, Food and Nutrition Service, n.d.b), the application process itself could be improved. For example, in the state of Maine, employees determine eligibility before individuals begin the social service application process (Wallack, 2006). Reducing the length of the application form would also facilitate greater utilization (Martin et al., 2003). Innovative outreach programs that collaborate with community partners, such as faith-based organizations, health organizations, and school programming, could positively influence participation rates (Food Research and Action Center, 2006). These strategies could easily be adapted to focus specifically on older adults. More extensive multilingual service provision, and more assistance for the illiterate and those with visual or hearing impairments could also be of value. Home visits by FSP workers to the functionally impaired could also improve access. Further education of SSI, welfare, public housing, and senior community center workers on the importance of referral to the FSP could also improve access for those older adults who are using other services. Mailed brochures on the FSP, eligibility criteria, and local contact information could be sent to all older adults receiving SSI, welfare, public housing, and Medicaid, as well as to those whose Social Security payments are below the poverty line. With the vast majority of impoverished older adults not receiving food stamps, strategies such as these are extremely Vol. 48, No. 2, 2008 243 important to rectify this situation among the most vulnerable group of older Americans. References Acheson, A. W. (2006). Poverty in Maine 2006. Orono, ME: Margaret Chase Smith Policy Center. Algert, S. J., Reibel, M., & Renvall, M. J. (2006). Barriers to participation in the Food Stamp Program among food pantry clients in Los Angeles. American Journal of Public Health, 96, 807–809. Andersen, R. M. (1995). Revisiting the behavioral model and access to medical care: Does it matter? Journal of Health and Social Behavior, 36, 1–10. Calsyn, R. J., & Winter, J. P. (2000). Predicting different types of service use by the elderly: The strength of the behavioral model and value of interaction terms. Journal of Applied Gerontology, 19, 284–303. Castner, L. A., & Schirm, A. L. (2005). Reaching those in need: State food stamp participation rates in 2002. Washington, DC: Mathematica Policy Research, Inc. Retrieved March 11, 2007, from http://www. fns.usda.gov/oane/menu/Published/FSP/FILES/Participation/Reaching 2002.pdf Cunnyham, K. (2004). Trends in food stamp participation rates: 1999–2002 (MPR Ref. No. 6044-109). Washington, DC: Mathematica Policy Research, Inc. Retrieved March 11, 2007, from http://www.mathematicampr.com/publications/redirect_pubsdb.asp?strSite=pdfs/fns99-02rates.pdf de Klerk, M. M. Y., Huijsman, R., & McDonnell, J. (1997). The use of technical aids by elderly persons in the Netherlands: An application of the Andersen and Newman model. The Gerontologist, 37, 365–373. DeNavas-Walt, C., Proctor, B. D., & Mills, R. J. (2004). Income, poverty, and health insurance coverage in the United States: 2003 (U.S. Census Bureau, Current Population Reports P60-226). U.S. Washington, DC: U.S. Government Printing Office. Retrieved May 30, 2007, from http:// www.census.gov/prod/2004pubs/p60-226.pdf Faux, M., & Rosso, R. (2003). Characteristics of food stamp households: Fiscal year 2002—Advance report (No. PR03-35). Washington, DC: Mathematica Policy Research, Inc. Retrieved March 11, 2007, from http://www.mathematica-mpr.com/PDFs/charfood2002.pdf Food Research Action Center. (2006, September). FRAC’s guide to food stamp outreach collaborations. Washington, DC: Author. Food Security Institute. (2003). Hunger and food insecurity among the elderly (Hunger Issue Brief). Waltham, MA: Brandeis University. Retrieved March 11, 2007, from http://www.centeronhunger.org/pdf/ Elderly.pdf Food Stamp Program: Frequently asked questions (n.d.). Retrieved March 11, 2007, from http://www.fns.usda.gov/fsp/faqs.htm Food Stamp Program: Overview (n.d.). Retrieved March 11, 2007, from http://www.aspe.hhs.gov/hsp/abbrev/fsp.htm Gabor, V., Williams, S. S., Bellamy, H., & Hardison, B. L. (2002). Seniors’ views of the Food Stamp Program and ways to improve participation—Focus group findings in Washington state (Rep. E-FAN-02-012). Washington, DC: U.S. Department of Agriculture, Economic Research Service. Retrieved March 11, 2007, from http://www.ers.usda.gov/ publications/efan02012/ Guthrie, J. F., & Lin, B. (2002). Overview of the diets of lower- and higherincome elderly and their food assistance options. Journal of Nutrition Education and Behavior, 34(Suppl.), 31–41. Haider, S. J., Jacknowitz, A., & Schoeni, R. F. (2003). Food stamps and the elderly: Why is participation so low? Journal of Human Resources, 38(Suppl.), 1080–1111. Jang, Y., Kim, G., & Chiriboga, D. A. (2005). Health, healthcare utilization, and satisfaction with service: Barriers and facilitator for older Korean Americans. Journal of the American Geriatrics Society, 53, 1613–1617. Jheriot, M. T., Segal, S. P., & Cowsert, M. J. J. (2003). African-Americans and comprehensive service use. Community Mental Health Journal, 39, 225–237. Kirsch, I. S., Jungeblut, A., Jenkins, L., & Kolstad. A. (1993). Adult literacy in America: A first look at the findings of the National Adult Literacy Survey (Pub. NCES 93275). Washington, DC: U.S. Department of Education. Retrieved March 12, 2007, from http://nces.ed.gov/pubs93/ 93275.pdf Kochera, A. (2006). Developing appropriate rental housing for low-income older persons: A survey of Section 202 and LIHTC property managers. Washington, DC: American Association of Retired Persons (AARP). Retrieved February 22, 2007, from http://www.aarp.org/research/housingmobility/affordability/dd149_lihtc.html Lee, J. S. & Frongillo, E. A. (2001). Nutritional and health consequences are associated with food insecurity among U.S. elderly persons. Journal of Nutrition, 131, 1503–1509. Li, H. (2006). Rural older adults’ access barriers to in-home and communitybased services. Social Work Research, 30, 109–118. Martin, K. S., Cook, J. T., Rogers, B. L., & Joseph, H. M. (2003). Public versus private food assistance: Barriers to participation differ by age and ethnicity. Journal of Nutrition Education and Behavior, 35, 249–254. McConnell, S., & Ponza, M. (1999). The reaching the working poor and poor elderly study: What we learned and recommendations for future research (MPR Ref. No. 8305-013). Washington, DC: Mathematica Policy Research, Inc. Retrieved March 11, 2007, from http://www. mathematica-mpr.com/PDFs/reach.pdf Nagelkerke, N. J. D. (1991). A note on a general definition of the coefficient of determination. Biometrika, 78, 691–692. Nord, M., Andrews, M., & Carlson, S. (2004). Household food security in the United States, 2003 (Food Assistance and Nutrition Research Rep. No. 42). Washington, DC: U.S. Department of Agriculture. Retrieved March 11, 2007, from http://www.ers.usda.gov/publications/fanrr42/ Parker, S., Jog, M., & Kennedy, T. S. (2005). Factors affecting food stamp participation among the elderly. Journal of Nutrition Education and Behavior, 37(Suppl. 1), 56. Phillips, K., Morrison, K., Andersen, R., & Aday, L. A. (1998). Understanding the context of healthcare utilization: Assessing environmental and provider-related variables in the behavioral model of utilization. Health Services Research, 33, 571–596. Rank, M. R., & Hirschl, T. A. (2005). Likelihood of using food stamps during the adulthood years. Journal of Nutrition Education and Behavior, 37, 137–146. Rogers-Dillon, R. (1995). The dynamics of welfare stigma, Qualitative Sociology, 18, 439–457. Rosso, R. (2001). Trends in Food Stamp Program participation rates: 1994–1999. Washington, DC: Mathematica Policy Research, Inc. Retrieved March 11, 2007, from http://www.fns.usda.gov/oane/menu/ Published/FSP/FILES/Participation/1999Trendsreport.pdf Rosso, R., & Faux, M. (2003). Characteristics of food stamp households: Fiscal Year 2002 (No. PR03-59). Washington, DC: Mathematica Policy Research, Inc. Retrieved March 1, 2007, from http://www.mathematicampr.com/publications/PDFs/charfood2002full.pdf U.S. Census Bureau. (n.d.). American Community Survey 2003 quality measures. Retrieved December 5, 2006, from http://www.census.gov/ acs/www/UseData/sse/res/03.htm U.S. Department of Agriculture, Food and Nutrition Service. (n.d.a). Characteristics of food stamp households: Fiscal year 2000 (Executive Summary). Retrieved December 3, 2006, from http://www.fns.usda.gov/ oane/menu/Published/FSP/FILES/Participation/2000Characteristics.htm 244 U.S. Department of Agriculture, Food and Nutrition Service. (n.d.b). Food stamps pre-screening eligibility tool. Retrieved February 6, 2007, from http://65.216.150.143/fns/ U.S. Department of Health and Human Services, Centers For Disease Control and Prevention. (1998). Health, United States, 1998 with socioeconomic status and health chart book. Hyattsville, MD: National Center for Health Statistics. Retrieved February 22, 2007, from http:// www.cdc.gov/nchs/data/hus/hus98.pd U.S. General Accounting Office. (2000). Food assistance: Options for improving nutrition for older Americans (Publication No. GAO/RCED00-238). Washington, DC: Author. Retrieved March 11, 2007, from http://www.gao.gov/cgi-bin/fetchrpt?rptno=RCED-00-238 U.S. General Accounting Office. (2004). Food Stamp Program: Steps have been taken to increase participation of working families, but better tracking of efforts is needed (Publication No. GAO-04-346). Washington, DC: Author. Retrieved March 11, 2007, from http://www.gao. gov/new.items/d04346.pdf Urban Institute. (2002, August). Immigrant families and workers: Immigrant well-being in New York and Los Angeles (Brief No. 1). Washington, DC: Author. Retrieved March 10, 2007, from http:// www.urban.org/publications/310566.html Urban Institute. (2006, May). Children of immigrants: Facts and figures (Fact Sheet). Retrieved March 10, 2007 from http://www.urban.org/ UploadedPDF/900955_childrenFAct Sheet)_of_immigrants.pdf Wallack, V. (2006, October 20). Food stamp use surge linked to better outreach. Times Record. Retrieved February 26, 2007, from http:// www.timesrecord.com Wellman, N. S., & Kamp, B. (2004). Federal food and nutrition assistance programs for older people. Generations, 28, 78–85. Wilde, P., & Dagata, E. (2002). Food stamp participation by eligible older Americans remains low. Food Review, 25, 25–30. Wolkwitz, K. (2007). Trends in Food Stamp Program participation rates: 1999 to 2005. Washington, DC: Mathematica Policy Research, Inc. Retrieved December 16, 2007 from http://www.fns.usda.gov/OANE/ MENU/Published/FSP/FILES/Participation/Trends1999-2005.pdf Zedlewski, S. R., & Rader, K. (2005). Have Food Stamp Program changes increased participation? Social Service Review, 79, 537–561. Received March 14, 2007 Accepted July 5, 2007 Decision Editor: William J. McAuley, PhD The Gerontologist